A Level Computer Science

for AQA Unit 2

Kevin R Bond Educational

Computing
Services
Single licence - Abingdon School
Structure of the book
The structure of this book follows closely the structure of AQA’s A Level Computer Science
specification for first teaching from September 2015. The content of the book has been constructed
with the aim of promoting good teaching and learning, so where relevant practical activities have
been suggested and questions posed for the student to answer. The book includes stimulus material
to promote discussion and deeper thinking about the subject. Additional material to support
teaching and learning will be available from the publisher’s website.

About the author

Dr Kevin R Bond is an experienced author. Kevin has 24 years of examining experience.
He also has many more years of experience teaching AS and A Level Computing and
Computer Science. Before becoming a computer science teacher, he worked in industry as
a senior development engineer and systems analyst designing both hardware and software
systems.

Single licence - Abingdon School

A Level Computer Science
for AQA Unit 2

Kevin R Bond

Educational Computing Services Ltd

Single licence - Abingdon School

Published in 2016 by
Educational Computing Services Ltd
42 Mellstock Road
Aylesbury
Bucks
HP21 7NU
United Kingdom
Tel: 01296 433004
e-mail: [email protected]

Every effort has been made to trace copyright holders and to obtain their permission for the use
of copyrighted material. We apologise if any have been overlooked. The author and publisher will
gladly receive information enabling them to rectify any reference or credit in future editions.

First published in 2016

ISBN 978-0-9927536-6-5

Text © Kevin R Bond 2016

Original illustrations © Kevin R Bond 2016
Cover photograph © Kevin R Bond 2016

The right of Kevin R Bond to be identified as author of this work has been asserted
by him in accordance with the Copyright, Designs and Patents Act 1988.

All rights reserved. No part of this publication may be reproduced or transmitted

in any form or by any means, electronic or mechanical, including photocopy,
recording or any information storage retrieval system, without permission in
writing from the publisher or under licence from the Copyright Licensing Agency
Limited, of Saffron House, 6 -10 Kirby Street, London, EC1N 8TS.

Approval message from AQA - Digital

The core content of this digital textbook has been approved by AQA for use with
our qualification. This means that we have checked that it broadly covers the
specification and that we are satisfied with the overall quality. We have also approved
the printed version of this book. We do not however check or approve any links or
any functionality. Full details of our approval process can be found on our website.
We approve print and digital textbooks because we know how important it is for
teachers and students to have the right resources to support their teaching and
learning. However, the publisher is ultimately responsible for the editorial control
and quality of this digital book. Please note that when teaching the A-level (7517)
course, you must refer to AQA’s specification as your definitive source of information.
While this digital book has been written to match the specification, it cannot
provide complete coverage of every aspect of the course. A wide range of other useful
resources can be found on the relevant subject pages of our website: aqa.org.uk

Single licence - Abingdon School

Acknowledgements
The author and publisher are grateful to the following for permission to reproduce images,
clipart and other copyright material in this book under licence or otherwise:

Chapter 5.1.1
Figure 1.1.1 “Late Babylonian clay tablet: table of numerals representing
lunar longitudes”, image ID 00851897001, British Museum
Pages 1, 2: Clip art www.123rf.com: green apple: 123rf / 14199537; red apple: 123rf / 1419906;
banana: 123rf / 39056131; orange: 123rf / 38547844; purse: 123rf / 27347756
Chapter 5.1.2
Page 4: Thermometer - “Thermometre froid a plat” Fotolia / 11368653 © Albachiara
Page 5: Cake - 123rf / 33382329 (www.123rf.com)
Chapter 5.1.4
Page 12: Greek character - 123rf / 32698394 (www.123rf.com)
Chapter 5.1.5
Page 14: Road going off into the desert - www.canstockphoto / csp9388362
Chapter 5.1.7
Page 20: Dreaming sheep - Shutterstock / 110338271; Page 20: Ruler - Shutterstock / 198850166
Chapter 5.2.1
Figure 2.1.1 Microsoft® Windows® 7 Calculator screenshot used with permission from Microsoft
Microsoft and Windows are either registered trademarks or trademarks of
Microsoft Corporation in the United States and/or other countries
Figure 2.1.4 Microsoft® Windows® 7 Device Manager version
6.1.7600.16385 screenshot used with permission from Microsoft
Chapter 5.3.1
Page 34: Penguins - Shutterstock / 114208987; Page 34: Peacock - 123rf / 36970936 (www.123rf.com)
Page 34/35: Highway code signs used in question 2, and page 35 are based on Highway Code signs,
© Crown copyright 2007, and are reproduced under Open Government Licence v3.0.
Page 35: Tree rings - Shutterstock / 97674011; Figure 3.1.4 red apples - 123rf / 1419906 (www.123rf.com)
Page 36: pound coin showing head - 123rf / 20150613_ml (www.123rf.com)
Page 36: pound coin showing tail - 123rf / 35831780 (www.123rf.com)
Chapter 5.3.2
Figure 3.2.2 Microsoft® Windows 7 command line window screenshot used with permission from Microsoft
Figure 3.2.3 Apple® MacBookPro® and OS X® are trademarks of Apple Inc., registered in the U.S. and other countries
Chapter 5.4.3
Page 52: CPU - Shutterstock / 222009121
Chapter 5.5
Figure 5.5.1 Shutterstock / 157001045; Figure 5.5.2 Shutterstock / 1226401
Figure 5.5.3 Shutterstock / 185237537; Figure 5.5.5 123rf / 31206024 (www.123rf.com)
Figure 5.5.6 123rf / 32168839 (www.123rf.com)
Chapter 5.6.2
Figure 5.6.2.2 Adapted from www.engineeringtoolbox.com/air-altitude-pressure-d462.html
with kind permission of the editor

Single licence - Abingdon School

Chapter 5.6.10
Figure 6.10.14 “A one-time pad” - reproduced with kind permission of Paul Reuvers, Crypto Museum
(www.cryptomuseum.com)
Figure 6.10.16 (a) “Image generated from random numbers generated by the PHP rand() function on Microsoft Windows.”
idea for this courtesy of Bo Allen, http://boallen.com/ who kindly provided permission
to use his PHP script to generate this image.
Figure 6.10.17 “Gilbert Vernam” - image in public domain
Figure 6.10.19(a) “Plaintext image to be encrypted using a one-time pad.” - Mathematician & computer scientist
Claude Shannon, Getty Image library, image 5337874
Chapter 7.4.1
Figure 7.4.1.3 CCD SONY ICX493AQA 10.14 Mpixels APS-C 1.8” (23.98 x 16.41mm) sensor side
CC A-Share Alike 4.0 Int. license, Andrzej w k 2.
Subtractive colour model image - Public domain SharkD
Figure 7.4.1.5 Schematic of the operation of a laser printer
Reproduced with permission from Computer Desktop Encyclopedia, www.computerlanguage.com
Chapter 7.4.1
Figure 7.4.2.5 SSD drive image reproduced with kind permission of StorageReview.com from
http://www.storagereview.com/samsung_ssd_840_pro_review
Chapter 8.1
Page 258 “Information Technology alone has this capacity to both automate and reflect information (informate)” - Professor
Shoshana Zuboff, Charles Edward Wilson Professor of Business Administration at the Harvard Business School.
Page 261 Memories for life: Page 99 “The Spy in the Coffee Machine” © Kieron O’Hara and Nigel
Shadbolt 2008, reproduced with permission of the publishers Oneworld Publications.
Page 260 Case study: “From Forbes.com, 16/02/2012 © 2012 Forbes LLC. All rights reserved. Used by
permission and protected by the Copyright Laws of the United States. The printing, copying, redistribution,
or retransmission of this Content without express written permission is prohibited.” http://www.forbes.com/
sites/kashmirhill/2012/02/16/how-target-figured-out-a-teen-girl-was-pregnant-before-her-father-did/
Chapter 11.1
Figure 11.1.2 Schematic of several servers connected via a network switch
Shutterstock / 53738941 (server), Shutterstock / 133413857 (switch)
Figure 11.1.3 Racks of commodity servers at one of Google’s data centres
(image Google/Connie Zhou) reproduced with permission of Google.
Chapter 12.1
Figure 12.1.1 screenshot of WinGhci, a simple GUI to run GHCI (the Glasgow Haskell Interpreter) on Windows - written by Pepi
Gallardo (http://www.lcc.uma.es/~pepeg/). Haskell - https://www.haskell.org/
- is a purely functional programming language developed by an open source community.
Figure 12.1.2 another screenshot of WinGhci.
The author would like to thank
• Richard Dobson for helpful comments on drafts of Chapters 5.6.1, 5.6.2, 5.6.3, 5.6.7, 5.6.8, and Chapter 6.3
• Marjory Joan and Edwin Sydney Bond for their unstinting and devoted support over
the years and for laying the foundations that made this book possible
• Sue Poh-Cheng Bond for her patience and constant support during the making of this book

Single licence - Abingdon School

Contents
Contents
How to use this book xi
Introduction xii
Numbering of chapters follows AQA’s specification numbering
5.1 Number systems 1
1.1 Natural numbers 1
1.2 Integer numbers 4
Whole numbers 5
1.3 Rational numbers 6
1.4 Irrational numbers 11
1.5 Real numbers 14
1.6 Ordinal numbers 18
1.7 Counting and measurement 20

5.2 Number bases 24

2.1 Number base 24

5.3 Units of information 34

3.1 Bits and bytes 34
3.2 Units 40

5.4 Binary number system 45

4.1 Unsigned binary 45
4.2 Unsigned binary arithmetic 48
4.3 Signed binary using two’s complement 52
4.4 Numbers with a fractional part 58
4.5 Rounding errors 69
4.6 Absolute and relative errors 75
4.7 Range and precision 77
4.8 Normalisation of floating point form 85
4.9 Underflow and overflow 91

5.5 Information coding systems 96

ASCII 96
Unicode 98
Character form of a decimal digit 99

Error checking and correction 100

5.6 Representing images, sound and other data 106

6.1(1) Bit patterns, images, sound and other data 106
6.1(2) Bit patterns, images, sound and other data 110

Single licence - Abingdon School

 6.1(3) Bit patterns, images, sound and other data 116
6.2 Analogue and digital 123
6.3 Analogue/digital conversion 129
6.4 Bitmapped graphics 139
6.5 Vector graphics 153
6.6 Vector graphics versus bitmapped graphics 160
6.7 Digital representation of sound 164
6.8 Musical Instrument Digital Interface(MIDI) 171
6.9 Data compression 177
6.10 Encryption 184

6.1 Hardware and software 216

6.1.1 Relationship between hardware and software 216
6.1.2 Classification of software 216
6.1.3 System software 218
6.1.4 Role of an operating system 219

6.2 Classification of programming languages 220

6.2.1 Classification of programming languages 220

6.3 Types of program translator 225

6.3.1 Types of program translator 225

6.4 Logic gates 231

6.4.1 Logic gates 231

6.5 Boolean algebra 247

6.5.1 Using Boolean algebra 247

7.1 Internal hardware components of a computer 257

7.1.1 Internal hardware components of a computer 257

7.2 The stored program concept 267

7.2.1 The meaning of the stored program concept 267

7.3 Structure and role of the processor and its components 270
7.3.1 The processor and its components 270
7.3.2 The Fetch-Execute cycle and the role of the registers within it 278
7.3.3 The processor instruction set 280
7.3.4 Addressing modes 286
7.3.5 Machine-code and assembly language operations 288
7.3.6 Interrupts 304
7.3.7 Factors affecting processor performance 308

7.4 External hardware devices 316

7.4.1 Input and output devices 316

Single licence - Abingdon School

 7.4.2 Secondary storage devices 323

8.1 Individual (moral), social (ethical), legal and cultural issues and
opportunities 330
8.1 Introduction 330

9.1 Communication 344

9.1.1 Communication methods 344
9.1.2 Communication basics 351

9.2 Networking 356

9.2.1 Network topology 356
9.2.2 Types of networking between hosts 362
9.2.3 Wireless networking 368

9.3 The Internet 383

9.3.1 The Internet and how it works 383
9.3.2 Internet security 400

9.4 The Transmission Control Protocol/Internet Protocol

(TCP/IP) protocol 420
9.4.1 TCP/IP 420
9.4.2 Standard application layer protocols 430
9.4.3 IP address structure 437
9.4.4 Subnet masking 444
9.4.5 IP standards 447
9.4.6 Public and private IP addresses 449
9.4.7 Dynamic Host Configuration Protocol (DHCP) 451
9.4.8 Network Address Translation (NAT) 454
9.4.9 Port forwarding 457
9.4.10 Client-server model 459
9.4.11 Thin- versus thick-client computing 470

10.1 Conceptual data models and entity relationship modelling 474

Data modelling 474
Entity relationship modelling 479

10.2 Relational databases 486

Relational database model 486

10.3 Database design and normalisation techniques 498

Normalisation techniques 498

10.4 Structured Query Language (SQL) 519

Using SQL to retrieve, update, insert and delete data 519

Single licence - Abingdon School

 Using SQL to create a database 527

10.5 Client server databases 533

10.5 Client server databases 533

11.1 Big Data 543

Section 1 - What is Big Data? 543
Section 2 - Functional programming is a solution 555
Section 3 - Fact-based model 556

12.1 Functional programming paradigm 561

12.1.1 Function type 561
12.1.2 First-class object 565
12.1.3 Function application 569
12.1.4 Partial function application 572
12.1.5 Composition of functions 578
12.2 Writing functional programs 581
12.2.1 Functional language programs 581

12.3 Lists in functional programming languages 592

12.3.1 List processing 592

Index 597
13.1 Aspects of software development - See Unit 1
Glossary - www.educational-computing.co.uk/CS/Unit2/Glossary.pdf
Exam practice questions -
www.educational-computing.co.uk/CS/Unit2/ExamPracticeQuestions.pdf

Exam practice solutions -

www.educational-computing.co.uk/CS/Unit2/ExamPracticeSolutions.pdf

Single licence - Abingdon School

■■ How to use this book 

The structure and content of this book maps to sections 4.5 to 4.12 of AQA’s A-level Computer Science specification
(7517). For example, the chapter number of the first chapter is 5.1.1 and its title is Number systems: Natural
numbers. This chapter maps to section 4.5.1.1 of AQA’s A-level Computer Science specification (7517). The chapters
in the book do not use the leading 4 as this designates Subject content – A-level in the specification.
Flipped classroom
This textbook has been written with the flipped classroom approach very much in mind. This approach reverses the
conventional classroom lesson and homework model of teaching. Instead, chapters in this textbook should be used
to prepare for a lesson so that classroom-time can be devoted to exercises, projects, and discussions.

The features in this book include:

Learning objectives
Learning objectives linked to the requirements of the specification are specified at the beginning of each chapter.

Key concept Concepts that you will need to understand and to be able to define or
explain are highlighted in blue and emboldened, e.g. Integers. The same
concepts appear in the glossary for ease of reference.

Key principle Principles that you will need to understand and to be able to define or
explain are highlighted in blue and emboldened, e.g. Abstraction. The same
principles appear in the glossary for ease of reference.

Key fact Key point Key term

Facts, points and terms that are useful to know because they aid in understanding concepts and principles are
highlighted in blue and emboldened, e.g. Whole number: Whole number is another name for an integer number.

Information Background
References information that has the potential to assist and contribute to a student’s learning, e.g. Read Unit 1 section
4.2.2 for more background on sets and set comprehension. Background knowledge that could also contribute to a
student’s learning.
Did you know? Extension Material
“Did you know?” - interesting facts to enliven learning. “Extension Material” - content that lies beyond the
specification.
Task Activity to deepen understanding and reinforce learning.

Programming tasks Practical activity involving the use of a programming language to deepen
understanding and reinforce learning of concepts and principles.

Questions Short questions that probe and develop your understanding of concepts
and principles as well as creating opportunities to apply and reinforce your
knowledge and skills.

■■ Web links for this book

The URLs of all websites referenced in this book are recorded at
www.educational-computing.co.uk/aqacs/alevelcs.html
Educational Computing Services is not responsible for third party content online, there may be some changes to this
content that are outside our control. If you find that a Web link doesn’t work please email webadmin@educational-
computing.co.uk with the details and we will endeavour to fix the problem or to provide an alternative.

Single licence - Abingdon School xi

Introduction 

If you are reading this book then you will already have So far it has not been necessary to mention digital
chosen to be a part of an exciting future, for Computer computers. Digital computers are just the current
Science is at the heart of an information processing means by which algorithms can be implemented to
revolution. This revolution applies not just to seeking execute on data. Both algorithms and the models on
patterns of meaning in data accumulated on an which they act need to be implemented: algorithms in
unprecedented scale by the huge growth in connected the form of code or instructions that a digital computer
computing devices but also the realisation that all can understand, i.e. a computer program; models in
forms of life are controlled by genetic codes. Genetic data structures in a programming language. Unit 1
codes are instructions in a procedural information sense was largely about the fundamentals of programming,
that together with the environment that they inhabit data structures, algorithms and their efficiency,
control and guide the development of organisms. i.e. algorithms to run quickly while taking up the
minimal amount of resources (e.g. memory, hard disk,
Computer scientists concern themselves with electricity), and the limits of computation.
• representations of information in patterns of
symbols, known as data or data representations, Unit 2 covers the fundamentals of computing devices,
• the most appropriate representation for this data how data is represented and communicated between
• the procedures in the form of instructions devices, the logic gate circuits that enable computing
that can transform this data into new forms of devices to perform operations and to store information.
information. It covers the fundamentals of computer organisation
The procedures themselves are also a form of and architecture, the structure and role of the processor,
information of an instructional kind. the language of the machine, binary (machine code)
and how it is used to program the hardware directly.
The key process in Computer Science is abstraction The fundamentals of networking are covered which
which means building models which represent aspects leads onto data models for storing structured and
of behaviour in the real-world which are of interest. unstructured data that can be accessed from networked
For example, if we wanted to build an automated machines. The simpler case of structured data is
recommendation system for an online book store, we covered first using the relational database model. The
might choose to record the types of book and number limitations of this model are exposed for data that
of each type purchased as well as details that identify lacks structure and which is too big to fit into a single
the respective customer. server. Such data is known as “Big Data”. Machine
learning techniques are needed to discern patterns
Computer Science is not alone in building abstractions, in this data and to extract useful information. Big
mathematics and the natural sciences also build Data also requires a different programming paradigm,
abstractions but their models only serve to describe and functional programming, one that facilitates distributed
explain whereas Computer Science must, in addition, programming.
perform actions on and with the data that has been
modelled if it is to solve problems. These actions are It is right that having journeyed through Unit 1 and
described by algorithms or step-by-step instructions Unit 2 that a student should have an opportunity to
which form what is called the automation stage of discuss using hypotheticals and case studies what kind
problem solving. Whilst it is true that automation of of philosophy of information is appropriate for any
tasks existed before Computer Science, their nature advanced information society. This is explored in Unit
involved concrete, real-world objects, e.g. the Jacquard 2 in the section Consequences of uses of computing
loom, not informational abstractions such as an online where guiding principles of behaviour are explored.
book recommendation system.
Single licence - Abingdon School xii
 5 Fundamentals of data representation
5.1 Number systems
Learning objectives:
■■Concept of number
■■The natural numbers
■■ 5.1.1 Natural numbers
What does it mean to count?
■■Numerals We learn to count from an early age. We notice that in the real world objects
can be grouped together in collections, for example
Task three apples. In doing so, we use abstraction in ignoring
Try your hand at counting: the differences between the individual apples in the
1
https://www.youtube.com collection – for example, one of them is green, the other two are red.
/watch?v=vJG698U2Mvo
video.
The concept of number
By considering collections of items we can get an understanding of the concept
of number. For example, a collection of three
Key principle
oranges, a collection of three bananas. If we
Abstraction: choose to ignore the differences between these
An abstraction is a collections and concentrate on their similarity, then we can form a relatively
representation that is arrived
abstract concept of the number three. The same process could lead to the
at by ignoring or removing
unnecessary detail. concept of the number 4, 5 and so on.

Numerals – representation of number

Key concept Representations of the concept of
Number: number have been carved in stone,
Quantity of things. and scratched on clay tablets since
Numeral:
early times. The representation
The representation of a number
is called a numeral.
of a number is called a numeral.
Numerals are written symbols The early Roman numerals were
for numbers. originally pictorial. For example,
three strokes carved in stone, III,
represented the number three.
Task
2 Investigate the Babylonian
numeral system.
What symbols did the Figure 1.1.1 Late Babylonian
Babylonian numeral system
clay tablet: table of numerals
use?
Evaluate the following: representing lunar longitudes

(Use http://en.wikipedia.org/
wiki/Babylonian_numerals)

Single licence - Abingdon School 1

5 Fundamentals of data representation

The Arabic (decimal) representations are less

0 0 3
pictorial, but again there is some choice in the
numerals to represent a number. For example,
both 3 and 03 (and indeed 003 and so on) are all
recognised as valid numerals, representing the same number.
Numeral systems
V A numeral system (or system of numeration) is
VI a writing system for expressing numbers, using
VII symbols in a consistent manner.

Questions
1 For the following numbers represented by Roman numerals,
change the symbols from Roman numeral representation to
the equivalent Arabic numeral representation:
(a) VII (b) LXXVII (c) MCMXCVI

Digits of numerals
In a basic digital system, a numeral is a sequence of digits, which may be of
arbitrary length. The most commonly used system of numerals is the Hindu–
Arabic numeral system, based on Hindu numerals. It uses ten symbols called
digits (0, 1, 2, 3, 4, 5, 6, 7, 8, and 9) to represent any number, no matter how
large or how small. This system is referred to as the decimal or denary system.
Key concept
What is a counting number?
Natural number:
We use the counting numbers {1, 2, 3, 4, ... } to keep
Natural numbers are the count-
ing numbers, either track of things such as how much money we have in our
{1, 2, 3, ... }, or {0, 1, 2, 3, ... }. pocket. The braces {} indicate a set (a collection of objects).
The objects in the set are written inside the braces. “…”
The symbol ℕ1 is used to indicates that there are infinitely more objects. Informally,
denote the set {1, 2, 3, ... },
the counting numbers are all the numbers you can get to by counting, starting
and the symbol ℕ0 is used to
denote the set {0, 1, 2, 3, ... }. at 1.

Where it is clear which set What is a natural number?

applies, the symbol ℕ is used.
Counting numbers are known as natural numbers.
Thus natural numbers can mean either “Counting Numbers”
ℕ
{1, 2, 3, ... }, or the “Counting Numbers” and zero, {0, 1, 2, 3, ... }. Sometimes
the special symbol ℕ or ℕ1 is used to denote {1, 2, 3, … } and the special
symbol ℕ0 is used to denote {0, 1, 2, 3, … }.

2 Single licence - Abingdon School

 5.1.1 Natural numbers

Did you know?

What is the sum of the first 100 natural numbers?
The mathematician Karl Friedrich Gauss when in elementary school in the eighteenth century amazed his teacher
by finding, in a few minutes, the sum of the natural numbers from 1 to 100. Gauss wrote the sum down twice as
shown below, once in ascending order, the second time in descending order, directly beneath the first.

1 + 2 + 3 + 4 + 5 + ....... + 96 + 97 + 98 + 99 + 100
100 + 99 + 98 + 97 + 96 + ....... + 5 + 4 + 3 + 2 + 1

Each vertical pair adds up to 101.

In total, there are 100 vertical pairs.
This makes the sum of all the natural numbers across the two rows = 100 × 101 = 10100.
The sum for one of the rows is one-half of this, i.e. 5050.

The alternative to Gauss’ method involves laboriously performing 99 addition steps, adding 1 to 2 then 3 to the
resulting sum and so on. This long-winded and laborious calculation is an example of a brute-force approach.

Questions
2 Find the sum of the following range of natural numbers using Gauss’s method

(i) 1 to 50 (ii) 1 to 200

3 Write a formula, in terms of n, to calculate the sum of all the natural numbers from 1 to n.

Programming tasks
1 Write a program to find the sum of the natural numbers from 1 to n. Your program should use
the brute-force approach. Test your program with the following values of n
(i) 100 (ii) 1000 000
2 Write a program to find the sum of the natural numbers from 1 to n. Your program should use
the formula approach. Test your program with the following values of n.
(i) 100 (ii) 1000 000

Investigation
1 Compare the execution times of programming tasks 1 and 2 for the two test values. What do
you observe?

In this chapter you have covered:

■■ What it means to count
■■ The concept of number
■■ Numerals – representation of number
■■ Numeral systems
■■ Digits of numerals
■■ What is a counting number?
■■ What is a natural number?

Single licence - Abingdon School 3

 5 Fundamentals of data representation
5.1 Number systems
Learning objectives:
■■Integer numbers
■■ 5.1.2 Integer numbers
Is the set of natural numbers, ℕ, enough?
Key concept
Are the natural numbers sufficient for all simple
Integer numbers: arithmetic? What about 3 – 5? The answer is clearly
Integer numbers are the natural
not a counting number so negative numbers have
numbers, ℕ, plus the negative
numbers formed by subtracting
to be added to the set of natural numbers to create
one natural number from the integers.
another. Integers are like the set of natural numbers, ℕ, but
The special symbol ℤ is used to they also include negative numbers. For example,
denote the set of integers
when the temperature is 10 degrees below zero, it is
ℤ = { ..., -5, -4, -3, -2, -1, 0,
-10 degrees.
1, 2, 3, 4, 5, ... }

So, integers can be negative {-1, -2, -3, -4, -5, … }, positive {1, 2, 3, 4, 5, … },
Information or zero {0}.

ℤ
Read Unit 1 section 4.2.2 for The special symbol ℤ is used to denote the set of integers
more background on sets and
set comprehension: ℤ = { ..., -5, -4, -3, -2, -1, 0, 1, 2, 3, 4, 5, ... }
{m – n: m ∊ ℕ and n ∊ ℕ}
means to generate the members
There are infinitely many elements in this set.
of the set ℤ by subtracting n
from m where m ∊ ℕ means m Questions
is a member of the set ℕ and n
∊ ℕ means n is a member of the 1 Are the following numbers (Hindu-Arabic numerals
set ℕ. representing numbers) integers?
(a) -10 (b) 5⅓ (c) 3.5?
Programming
2 Is the result of evaluating the following expression an integer?
Task
1 What are the maximum and (367 × 42) ⧸ 7
minimum integers for the
programming language that
you use?
If your programming
language has several integer
data types, find these values
for all the supported integer
data types.

Single licence - Abingdon School 4

5 Fundamentals of data representation

■■ Whole numbers
Key concept What are whole numbers?
We will consider whole numbers to be numbers without a fractional part
Whole number:
although other definitions of whole numbers exist which take a different
Whole number is another name
for an integer number. interpretation. A fractional part is a fraction. A
fraction is any number greater than 0 and less
than 1. For example, a slice of cake is a part of the
whole, say ∕₁₀ and clearly not the whole. On the
other hand, we may have 3 whole cakes or 3 whole degrees
of temperature below zero, i.e. -3.

Whole numbers can be positive, negative or zero according to this

interpretation. Whole number is another name for integer.

In this chapter you have covered:

■■ Integer numbers which are numbers which belong to the set
{ ..., -5, -4, -3, -2, -1, 0, 1, 2, 3, 4, 5, ... }
■■ Whole number is another name for integer

5 Single licence - Abingdon School

 5 Fundamentals of data representation
5.1 Number systems
Learning objectives:
■■Rational numbers
■■The set ℚ of
■■ 5.1.3 Rational numbers
rational numbers Is the set of integers, ℤ, enough for all arithmetic operations?
10 If we carry out an arithmetic operation of division such
1
as 3 ÷ 5 the result is not a natural number or an integer.
Key concept 7 Therefore, we need to extend our number system to include 4
Rational number: the rational numbers, e.g. ½, ⅓, ¼, ⅜, ⅛, 10⁄7 and so on.
A rational number, x, is one
that can be expressed in the
The definition of a rational number is as follows
form x = m⁄n where m and
n are integers, excluding zero
for n. “For x to be a rational number it must be expressible in the form
x = m⁄n

where m and n are integers, excluding zero for n.”

Key point
By this definition, 2⁄₁, 11⁄₁, 3⁄₂, 4⁄₁, 5⁄₃, − 3⁄₂, 24⁄₁₁, 9⁄₇, − 1⁄₂,
11
Simplest form:
A rational number in its are rational numbers as are 8⁄₄, 6⁄₃, 24⁄₃. 1
simplest or lowest form is
one that cannot be reduced For x to be a unique rational number then we need to insist that m and
any further because the only n have no common factor except 1. 8⁄₄, is not a unique rational number
common factor of m and n, the because 8 and 4 have the common factors 2 and 1. 8⁄₄, can be reduced to 2⁄₁
numerator and denominator which is a unique rational number because the only common factor is 1. A
respectively, is 1.
rational number in its simplest form is one that cannot be reduced any further
because the only common factor of m and n, the numerator and denominator
respectively, is 1. Such rational numbers are called simple fractions.

Information
The word “rational” comes
Formally, ℚ is the set of rational numbers.
ℚ
from the word “ratio” because a There are infinitely many elements in this set.
rational number can always be
written as the ratio, or quotient, Questions
of two integers.
1 Determine whether each of the following statements is true
or false:
(a) -13 ∊ ℕ (b) ³⁶∕₇₇ ∊ ℚ (c) -11 ∊ ℚ?

Single licence - Abingdon School 6

5 Fundamentals of data representation

Key term Programming Task

Set of rational numbers, ℚ: 1 A program that computes the quotient q and the remainder r when
Formally, the symbol ℚ is used dividing an integer x by an integer y must satisfy the following two
to mean the set of rational conditions as follows:
numbers.
1) x ≥ 0 and y > 0
2) x = y * q + r and 0 ≤ r < y

Information Pseudo-code for the algorithm to compute q and r when dividing an

integer x by an integer y:
Read Unit 1 section 4.2.2 for
more background on sets and r ← x
set comprehension: q ← 0
While r >= y
ℚ = {m⁄n : m ∊ ℤ and n ∊ ℕ1} r ← r − y
q ← q + 1

Write a program for this algorithm. Does your program meet the
two conditions? If your programming language supports assertions,
then use assertions to check that the two conditions are met. If not
use appropriate tests.

Is the set of rational numbers ℚ countable?

Just for the moment, consider the set of positive rational numbers
{¹⁄₁, ²⁄₁, ¹⁄₂, ¹⁄₃, ³⁄₁, ⁴⁄₁, ³⁄₂, ²⁄₃, ¹⁄₄, ¹⁄₅, ⁵⁄₁, …}

arrived at by following the arrows in Table 1.3.1.

This set is represented by the symbol ℚ⁺.
Key point
1 2 3 4 5 6 7 8 …
Simplest form: 1 1/1 1/2 1/3 1/4 1/5 1/6 1/7 1/8
The entries in Table 1.3.1 that
2 2/1 2/2 2/3 2/4 2/5 2/6 2/7 2/8
have common factors greater
3 3/1 3/2 3/3 3/4 3/5 3/6 3/7 3/8
than 1 are greyed out because
they can be reduced to their 4 4/1 4/2 4/3 4/4 4/5 4/6 4/7 4/8
simplest form which is already 5 5/1 5/2 5/3 5/4 5/5 5/6 5/7 5/8
in the table e.g. ²⁄₈ becomes 6 6/1 6/2 6/3 6/4 6/5 6/6 6/7 6/8
¹⁄₄. 7 7/1 7/2 7/3 7/4 7/5 7/6 7/7 7/8
8 8/1 8/2 8/3 8/4 8/5 8/6 8/7 8/8
…

Table 1.3.1

7 Single licence - Abingdon School

 5.1.3 Rational numbers

If you think about it all possible positive rational numbers will be generated,
e.g.¹⁴⁷⁄₉₁₄₅₇ will be in the table at the intersection of row 147 and column Key point
91457 and will get added in turn. Therefore, it is possible to order the rational
ℚ is countable:
numbers of the set ℚ⁺ and to show a one-to-one correspondence with the It is possible to order the
natural numbers as indicated in Table 1.3.2. rational numbers, ℚ,
and to show a one-to-one
correspondence with the natural
1 2 3 4 5 6 7 8 9 10 11 … numbers. Therefore, ℚ is
1⁄1 2⁄1 1⁄2 1⁄3 3⁄1 4⁄1 3⁄2 2⁄3 1⁄4 1⁄5 5⁄1 … countable.

Table 1.3.2
To generate ℚ, we can place zero /1 before 1/1 and insert the negative of each
positive rational number other than zero immediately after the positive rational
number as shown in Table 1.3.3.

1 2 3 4 5 6 7 8 9 …
0⁄1 1⁄1 -1⁄1 2⁄1 -2⁄1 1⁄2 -1⁄2 1⁄3 -1⁄3 …
Table 1.3.3
Table 1.3.3 shows that it is possible to order the rational numbers ℚ and
to place them in a one-to-one correspondence with the natural numbers.
Therefore, the set of rational numbers ℚ is countable. It is also infinite because
the set of natural numbers is infinite.
In addition, all integers are in ℚ because every integer n can be expressed as
n/1.
Representing rational numbers as terminating decimals
Long division is used to convert a rational number into decimal form.
Examples: (a) 189⁄9 = 189 ÷ 9 = 21 (b) 13⁄20 = 13 ÷ 20 = 0.65

9 goes into 189 20 goes into 13
Key point

21 times remainder 0 0 times remainder 13
∙ 20 goes into 130 The rational numbers in lowest
form whose only prime factors
6 times remainder 10
in their denominator are 2 or 5
20 goes into 100 or both, convert to terminating
5 times remainder 0 decimals.

If the final remainder is 0, e.g. 189⁄9, the quotient is a whole number, e.g. 21,
or a finite or terminating decimal, i.e. a decimal with a finite number of digits
after the decimal point, e.g. 0.65.
The rational numbers in lowest form whose only prime factors in their
denominator are 2 or 5 or both, convert to terminating decimals.

Single licence - Abingdon School 8

5 Fundamentals of data representation

Examples: (a) ¹⁄₂ = ¹⁄₂ × ⁵⁄₅ = ⁵⁄₁₀ = 0.5

Key point (b) ¹⁄₄ = ¹⁄₂ × ¹⁄₂ × ⁵⁄₅ × ⁵⁄₅ = ²⁵⁄₁₀₀ = 0.25
All recurring decimals are Questions
infinite decimals. This occurs
when the denominator involves 2 Without performing a decimal conversion, determine
the prime factors from the set whether the following rational numbers will convert to a
{3, 7, 11, 13, 17, 19, …}.
terminating decimal
(a) ⁵⁄ ₁₂₅ (b) ¹⁶∕₁₀₂₄ (c) ²∕₃

Representing rational numbers as recurring decimals

Sometimes when converting a rational number by long division, the
division never stops as there is always a remainder. Such rational numbers
convert to a recurring decimal.
All recurring decimals are infinite decimals.
This occurs when the denominator involves the prime factors from the set
{3, 7, 11, 13, 17, 19, …}

Examples: (a) ¹⁹⁄₁₂ = 19 ÷ 12 = 1.58333… (b) ¹⁄₃ = 1 ÷ 3 = 0.333…

Remainder 70 10
100 10
40 10
40 1
40 …
4
…

The repeating pattern may consist of just one digit or of any finite number
of digits. The number of digits in the repeating pattern is called the period.
The repeating pattern is indicated by placing a period mark or a bar over
each digit in the repeating pattern, e.g.
(a) ¹⁄₃ = 0.333… = 0.3
(b) ⁸⁄₁₁ = 0.727272… = 0.72

Questions
3 Convert the following rational numbers to decimal:
(a) ¹⁶⁄ ₃ (b) ¹⁰∕₇ (c) ¹³∕₁₁

4 What is the repeating pattern for each decimal in Q3?

9 Single licence - Abingdon School

 5.1.3 Rational numbers

Rational numbers are terminating or recurring decimals Key point

A rational number is either a terminating or recurring decimal. Every
terminating or recurring decimal can be converted to a⁄b A rational number is either
a terminating decimal or a
where a ∊ ℤ and b ∊ ℕ1. recurring decimal.

Questions
5 Convert the following decimals to their rational number
equivalent:
(a) 5.25 (b) 0.90

In this chapter you have covered:

■■ The set of integers ℤ is not enough for all arithmetic operations
■■ A rational number, x, is one that can be expressed in the form x = m/n
where m and n are integers, excluding zero for n.
■■ The set of rational numbers ℚ is countable
■■ The rational numbers in lowest form whose only prime factors in their
denominator are 2 or 5 or both, convert to terminating decimals.
■■ Rational numbers are terminating or recurring decimals
■■ Recurring decimals are infinite decimals

Single licence - Abingdon School 10

 5 Fundamentals of data representation
5.1 Number systems
Learning objectives:
■■Irrational numbers
■■Decimal expansion of an
■■ 5.1.4 Irrational numbers
Are rational numbers sufficient to model all numbers?
irrational number
The answer is no. The following boxed yellow section explains why but this is
for information only.
Figure 1.4.1 shows a rectangle with sides of length a, and b. We use
multiplication to work out the area of the rectangle as follows
Area = ab
where ab means a times b. To measure the length of the sides we use a ruler
marked with the rational numbers as shown in Figure 1.4.2. The integers
were marked first. Next the multiples of ¹⁄₂ were
added followed by the multiples of ¹⁄₃ and so on. b
It would seem that this process would leave little
room for any further points on the line. a
Figure 1.4.1 Rectangle of sides a and b

0 1 2
1
1 3
1
1
2 3 5
2 7
1 4 3 5 4 7 9 4 11 13 4 15
8 8 8 8 8 8 8 8

1 2 4 5
3 3 3 3

Figure 1.4.2 Ruler of the rational numbers

However, this intuition is not consistent with Pythagoras’ theorem which

Information requires that the length h of the hypotenuse in the right-angled triangle in
Pythagoras’ theorem states that Figure 1.4.3 should satisfy
the square on the hypotenuse
of a right-angled triangle is the
sum of the squares on the other h h2 = 12 + 12 = 2
two sides. 1

1 Figure 1.4.3 Right-angled triangle sides 1, 1, h

Single licence - Abingdon School 11

5 Fundamentals of data representation

If it is true that every length in Euclidean geometry

can be measured by a rational number, then it must
Information be true that there is a positive rational number such
that x = h, x2 = 2, and x = √2. That makes √2 a
Euclidean geometry is the
geometry described by Euclid in rational number but we will discover that it can’t be.
his textbook the Elements and If x is rational then it can be expressed as the ratio
which we have used for over
of two integers, m and n with
2000 years.
common factor 1 only and n <> 0.

x = m⁄n
Information
2
Proof by contradiction:
It follows that x2 = m ⁄n2 because x = √2
We assume that what we
want to prove is not true, and
Therefore, m2 = 2n2
then show the consequences
contradict either what we have And so m2 is even. This implies that m is even.
just assumed, or something
we already know to be true (or
both). We may therefore write m = 2k (Multiplying k, any natural number, by 2
ensures evenness).
Substituting 2k for m in m2 = 2n2

we get (2k)2 = 2n2

Or 4k2 = 2n2

2k2 = n2
Rearranging n2 = 2k2

Thus n is even.

For both m and n to be even they must be divisible by 2 but

by definition m⁄n is a rational number divisible by 1 only.

Therefore, √2 cannot be defined as a rational number and therefore it is not a

member of the set of rational numbers.

12 Single licence - Abingdon School

 5.1.4 Irrational numbers

What is the set of irrational numbers?

Conclusion, we need more than the set of rational numbers. We require in Key point
addition, a new set which contains those numbers that like √2 are not rational Irrational number:
numbers. We call this new set the set of irrational numbers. √2 is therefore an Irrational numbers are numbers
irrational number. that can be written as decimals
but not as simple fractions.
Irrational numbers are numbers that can be written as decimals but not as
simple fractions. Irrational numbers have decimal expansions that neither Irrational numbers have
terminate nor are periodic with some repeating sequence. For example, the decimal expansions that neither
terminate nor are periodic with
decimal expansion of √2 to 50 decimal places is
some repeating sequence.
1.41421356237309504880168872420969807856967187537694
To see more decimal places go to
http://apod.nasa.gov/htmltest/gifcity/sqrt2.1mil

Square roots and irrational numbers?

If a number could be the area of a square with a side that is a whole number,
then the number is called a “perfect square”, e.g. 4. However, if the area of
a square is not a perfect square, then the side of the square is an irrational
number, i.e. the square root of the area. For example, if the area is 3 cm2 then
1.73205….. is an irrational number.

Questions
1 Which of the following numbers are irrational?

(a) √⁸∕₂ (b) √8 (c) √300

(d) √361 (e) 3.777… = 3.7
(f ) 0.12112111211112… (g) 325∕7

In this chapter you have covered:

■■ The set of rational numbers ℚ is not enough to model all numbers
■■ An irrational number is a number that can be written as a decimal but not
as a simple fraction.
■■ Irrational numbers have decimal expansions that neither terminate nor are
periodic with some repeating sequence

Single licence - Abingdon School 13

 5 Fundamentals of data representation
5.1 Number systems
Learning objectives:
■■Real numbers
■■Real numbers form a
■■ 5.1.5 Real numbers
Real number system forms a continuum
continuum
We need more than the set of rational
■■Real number line numbers to model numbers because marking
■■Set of real numbers a straight line with the rational numbers
will still leave points of the line unmarked.
■■Decimal expansion of a These ‘holes’ are filled by irrational numbers.
real number When both rational and irrational numbers
are marked on the line, they fill it completely
Key concept
and stretch unbroken in both directions to form the real number system.
Real number: The real number system of rational numbers and irrational numbers forms a
A real number is either a continuum.
rational number or an irrational
number.
Real numbers are represented What is the real number line?
by decimals using an infinite The real number line is a useful way of modelling the set of real numbers. It is
decimal expansion.
an infinite line on which points are taken to represent the real numbers by their
distance from a fixed point labelled O and called the origin. Every point of
Key point this line represents a real number. Some real numbers are shown on this line in
The real number system of Figure 1.5.1.
rational numbers and irrational
numbers forms a continuum. -1 0⅓ 1 2 3 4

Key point
½ √3
�
Figure 1.5.1 The real number line
Real number line:
The real number line is a useful
Ideally, one would like to show and label every point on the number line, but
way of modelling the set of real no matter how dense one makes the points there are always points in between.
numbers. What is the set of real numbers?
The set interpretation of real numbers
ℚ Irrational is an alternative to thinking of real
Key concept
numbers numbers as points on an infinitely long
Set of real numbers:
ℤ line. The special symbol ℝ is used to
The set of real numbers, ℝ, is
ℕ denote the set of real numbers. It is
formed from the union of the a set formed from the union of the
set of rational numbers and the set of rational numbers and the set of
set of irrational numbers.
irrational numbers – Figure 1.5.2.
Figure 1.5.2 The composition of the set of real numbers, ℝ

Single licence - Abingdon School 14

5 Fundamentals of data representation

Questions
Key point
1 Determine whether each of the following statements is true
The set ℝ is an uncountable set. or false:
Real numbers have the property
(a) ¹⁶∕₃ ∈ ℝ (b) 3.142 ∈ ℝ
that between any two of them,
no matter how close, there lies (c) ¹∕₄ ∈ ℚ and ¹∕₄ ∈ ℝ (d) π ∈ ℝ (e) √2 ∈ ℚ
another real number. (f ) √361 ∈ ℕ and √361 ∈ ℝ (g) −3 ∈ ℝ

What is a real number?

Real numbers describe real-world quantities such as distances, amounts of
things, temperature, and so on. They are represented by decimals using an
Key point infinite decimal expansion and which define a real number.
Using this definition a real number is an expression of the form
Real numbers describe real-
world quantities such as ± a1a2a3…ak• b1b2b3…

For example, -4328.5000000… where … indicates infinitely many zeroes.

distances, amounts of things,
temperature, and so on.
± represents a choice between plus and minus.
Each of the digits a1, a2, a3, …, ak is an integer between 0 and 9 inclusive
except a1 when there is more than one digit in which case a1 is restricted to
being between 1 and 9, since it is the leading digit, e.g. 13•56….
The infinitely many b digits are integers between 0 and 9 inclusive.
Key point
Every real number has a unique
decimal expansion unless it is a Questions
rational number of the form
m 2 A line of exact length 1 is repeatedly shortened an infinite
⁄10n
number of times by cutting exactly ⁹⁄₁₀ths from the line each
For the latter number, there
time. Each bit removed is added to the end of its immediate
are two forms of the decimal
expansion, e.g. 0.9999… and
predecessor to make a new and separate line. How much in
1.0000..., 4.9999… and total is removed after
5.0000... with no real number (a) one cut (b) two cuts (c) four cuts
existing between each form. (d) a very large but finite number of cuts
(e) infinitely many cuts

Investigation
Find out why 0.9999… and 1.0000… represent the same real
number.

15 Single licence - Abingdon School

 5.1.5 Real numbers

Programming Task Information

1 The ratio of the circumference of a circle to its diameter, π can n=6
be calculated using the formula below. The more terms in the ∑n is shorthand for
1
sequence the more accurate the calculated value for π will be. 1 + 2 + 3 + 4 + 5 + 6.

π = 3 + ⁴ ⁴ ⁴ ⁴ … ∑ means summate or sum.

₂.₃.₄ − ₄.₅.₆ + ₆.₇.₈ − ₈.₉.₁₀ +
! means factorial,
… means that there are more terms modelled on what is e.g. 3! = 3×2×1.
given already, e.g. the next two terms in the sequence are
This special number e was
known in the early 17th

⁴
₁₀.₁₁.₁₂ − ⁴
₁₂.₁₃.₁₄ century, and probably was
discovered around this date in
connection with the growth
Write a program that uses this formula to calculate π for a
of money with time when
given number of terms, n. Test your program with the invested.
following values of n
(a) 2 (b) 3 (c) 5 (d) 7

Did you know?

If we evaluate
There are some irrational
numbers that can be
1+ ¹₁ + ¹₂ + ₃¹ + ₄¹ + … + ¹ + …
! ! ! ! r! contemplated in their entirety.
For example,
the result tends towards the value 2.718… which is given
the symbol e. 0.2200022200000
222222000000022…

We can arrive at a value for e by evaluating the formula

defines an irrational number
n in which each string of 0’s and
e = ∑ (2n + 2) 2’s increases in length each
0 (2n + 1)! time by two of each. Thus we
have a rule for constructing
Using a value for n of 6 we can evaluate e to 9 decimal
the decimal expansion of this
places of accuracy. irrational number. The rule can
Write a program that uses this formula to calculate e for be expressed as an algorithm to
the following values of n. generate successive digits.
(a) 1 (b) 3 (c) 6 (d) 8 Real numbers whose expansions
can be generated by algorithms
are called computable
In this chapter you have covered: numbers. They are the real
■■ Real number system forms a continuum numbers that can be computed
to within any desired precision
■■ A real number is either a rational number or an irrational number
by a finite, terminating
■■ Real numbers are represented by decimals using an infinite decimal algorithm.
expansion However, there are also many
real numbers that are not
■■ Every real number has a unique decimal expansion unless it is a rational
m computable in this sense.
number of the form ⁄10n

Single licence - Abingdon School 16

5 Fundamentals of data representation

■■ The set of real numbers, ℝ, is formed from the union of the set of rational
numbers and the set of irrational numbers
■■ The set ℝ is an uncountable set
■■ Real numbers describe real-world quantities such as distances, amounts of
things, temperature, and so on.

17 Single licence - Abingdon School

 5 Fundamentals of data representation
5.1 Number systems
Learning objectives:
■■Ordinal or ordinal number
■■ 5.1.6 Ordinal numbers
What is an ordinal number?
The natural numbers are used for counting or quantifying something, i.e. how
Key concept
much of something we have, but we have another type of number which we use
Ordinal or ordinal number: when we need to talk about where something comes in relation to something
Ordinals or ordinal numbers
else, e.g. first or second and so on. This number type is called ordinal number
are used to label the positions of
objects in a list, ordered set or
or just ordinal.
sequence of objects.
Ordinal numbers are therefore used when we need to position something.
Ordinal numbers are used to label the positions of objects in a list, ordered
set or sequence of objects. The objects must be ordered so that there is a first
element, a second element and so on. We use the natural numbers to describe
the position of an element in a sequence as follows
1st, 2nd, 3rd, 4th, 5th, etc.

In English, this is
first, second, third, fourth, fifth, etc.
We could also have started at 0 and labelled the first element as the zeroth
element in which case we use the natural numbers including zero as follows

0th, 1st, 2nd, 3rd, 4th, 5th, etc.

Questions
1 An index is used in programming to specify the position of an
Information
element in an ordered collection, e.g. an array. An index may
An array index is a way of start from 0 or 1 depending on programming language used or
labelling a cell of an array. The programmer preference.
term subscript is also used to
(a) If you were told that i was the 100th element of an array:
mean the same thing.
(i) How many elements would you consider come
before this element?
(ii) What would be the index for this 100th element
if the indexing starts at 0?
(iii) What would be the index for this 100th element if the
indexing starts at 1?
(b) If you were told that there are 100 elements in an array:
(i) What would be the index of the last element if
the indexing starts at 0?

Single licence - Abingdon School 18

5 Fundamentals of data representation

(ii) What would be the index of the last element if the indexing
starts at 1?

2 An element’s ordinal (index) equals the number of elements

preceding it in the sequence, e.g. an array or a
string.
Should the index start with 0 or 1 to make this
statement true?

3 The length of a range is the difference of its endpoints,

e.g. 0 <= i < 5 has 5 elements, as does 5 <= i < 10.
Give the length of these two ranges if in each case
‘<’ is replaced by ‘<=’. Comment on your answer.

Discussion

4 Some programming languages only allow array indexing to

begin from zero whilst other languages are more flexible
and allow the programmer to choose whether to start
indexing from 0, 1 or any integer, e.g. -2.
Discuss the advantages and disadvantages of each.

Programming Task
1 Write a program that accepts a letter of the alphabet typed
at the keyboard, uppercase or lowercase, and outputs its
numeric position in the alphabet followed by either “st”,
“nd”, “rd” or “th” as appropriate.

In this chapter you have covered:

■■ Ordinals or ordinal numbers which are used to label the positions of
objects in a list, ordered set or sequence of objects.

19 Single licence - Abingdon School

 5 Fundamentals of data representation
5.1 Number systems
Learning objectives:
■■ Be familiar with the use of
• natural numbers for counting
■■ 5.1.7 Counting and measurement
• real numbers for measurement Enumeration
When we count things we start at 1 for the first item, 2 for the second and so
on. Anything that can be counted out is said to be enumerable. The process of
Key point counting is known as enumeration and
The process of counting
we use the natural numbers for counting.
is known as enumeration. Can we count the integers? If we can they
To enumerate is to count.
can be said to be enumerable. We can in
Enumeration is counting with
actual fact. Similarly, we can enumerate,
natural numbers.
i.e. count, the positive rational numbers
greater than 0 using a for loop as follows
for i ← 1 to infinity
Information for j ← 1 to i
url for Cantor’s diagonal display i/j
argument:
By a similar for-loop argument we could enumerate all rational numbers and
Su, Francis E., et al. “Cantor
Diagonalization.” thereby count them. We conclude that the set of rational numbers is also
Math Fun Facts. countable.
https://www.math.hmc.edu/
We are not so lucky when it comes to the irrational numbers. The irrational
funfacts/ffiles/30001.4.shtml
numbers cannot be enumerated (for a formal proof see Cantor’s diagonal
argument).

Using real numbers for measurement

When we make a physical measurement, we use a measuring instrument. For
example, for measuring the lengths of pieces of wood we could use a ruler such
as the one shown in Figure 1.7.1. This ruler is marked off in whole centimetres
and tenths of centimetres. A tenth of a centimetre in the decimal numeral
system is the fraction ¹⁄₁₀.

Figure 1.7.1 Ruler marked in centimetres and tenths of a centimetre

Single licence - Abingdon School 20

5 Fundamentals of data representation

Clearly, using this ruler, we would not be able to measure the length of a piece
of wood to hundredths of a centimetre, but instead are limited to measuring
Key point to tenths of a centimetre. We conclude that when dealing with fractions, we
When making measurements shall have to approximate some by using a suitably close fraction that does
we are limited to using a have a representation on the ruler, e.g. ¹⁄₉ would be approximated by ¹⁄₁₀
rational number approximation with an error of ¹⁄₉ − ¹⁄₁₀ = ¹⁄₉₀. The more divisions that we have on the
to a real number.
ruler, the better our approximation can be, but the need for approximation
cannot be removed. Between one tenth and two tenths, for example, there are
infinitely many fractions. We are limited therefore to using a rational number
approximation to a real number.

Questions
1 Calculate the decimal expansion of ¹⁄₉.
Key point
2 If the decimal expansion of ¹⁄₉ is restricted to each of the
All fractional representations
following number of decimal places, what fraction results in
run the risk of being imprecise.
each case
(a) 2 (b) 3 (c) 6?

3 What is the difference between ¹⁄₉ and each of the fractions

that result from answering (a), (b), (c) in Q2?

Programming task
1 Write a program that outputs the result of performing the
following calculations:
(a) ¹⁄₃
(b) ¹⁄₇
(c) ¹⁄₅
(d) ¹⁄₁₀
Comment on the results.

21 Single licence - Abingdon School

 5.1.7 Counting and measurement

Rational number approximation to a real number

Key point
Any real number can be approximated to any desired degree of accuracy by
rational numbers with finite decimal representations, i.e. terminating decimals. Any real number can be
If the real number is x and assuming approximated to any desired
degree of accuracy by rational
x ≥ 0 numbers having finite decimal
representations.
Then for every integer n ≥ 1 there is a finite decimal
rn = + a1a2a3…ak • b1b2b3…bn

such that rn ≤ x < rn + ¹⁄₁₀n

For example, assuming x = 368.78456789123………

then for n = 2, r2 = 368.78 and r2 + ¹⁄₁₀₀ = 368.79
That is, x, lies somewhere between 368.78 and 368.79.

Questions
4 For each real number x where x ≥ 0 write down

rn and rn + ¹⁄₁₀n

where rn = + a1a2a3…ak• b1b2b3…bn and n ≥ 1

(a) x = 0.3245995632……… and n = 4

(b) x = 85.994467285……… and n = 3

Rounding off
We usually apply the process of rounding off to real numbers when using a
rational number approximation. The rules for rounding off to n decimal places
are:
■■ If the value of the (n + 1)th digit is less than five (0, 1, 2, 3, or 4), we
leave the nth digit alone.
■■ If the value of the (n + 1)th digit is greater than or equal to five
(5, 6, 7, 8, or 9), we increase the value of the nth digit by one.

For example, if the real number is 368.78456789123………

then rounding off to 2 decimal places, it becomes 368.78.
Whilst if the real number is 368.78546789123………
then rounding off to 2 decimal places, it becomes 368.79.

Single licence - Abingdon School 22

5 Fundamentals of data representation

Questions
5 Round off the following real numbers to the specified
number of decimal places
(a) x = 0.3245995632……… to 4 decimal places
(b) x = 85.994467285……… to 3 decimal places
(c) x = 5.884467285……… to 3 decimal places

In this chapter you have covered:

■■ To count is to enumerate.
■■ Enumeration is counting with natural numbers.
■■ When making measurements we are limited to using a rational number
approximation to a real number.
■■ All fractional representations run the risk of being imprecise.
■■ Any real number can be approximated to any desired degree of accuracy
by rational numbers having finite decimal representations.
■■ We usually apply the process of rounding off to real numbers when using
a rational number approximation.

23 Single licence - Abingdon School

 5 Fundamentals of data representation
5.2 Number bases
Learning objectives:
■■ Number base

■■ Decimal (base 10)

■■ 5.2.1 Number base
Meaning of number base
■■ Binary (base 2) The number base system specifies how many digits are used in constructing a
■■ Hexadecimal (base 16) numeral (representation of a number) and by how much to multiply each digit.
For example, in the decimal system the numeral 734 is interpreted as meaning
■■ Converting between decimal,
binary and hexadecimal 7 × 100 + 3 × 10 + 4 × 1

■■ Hexadecimal as a shorthand Decimal (base 10)

for binary and why The number base of the decimal system is ten because it has ten digits 0, 1, 2, 3,
4, 5, 6, 7, 8, 9 and the digit multiplier is a power of ten, 10n where n is
… −3, −2, −1, 0, 1, 2 , 3, …
Background
The number represented by the numeral 734 in base 10 is
The abacus is a calculating tool
based on moving beads in a constructed using the place values indicated in Table 2.1.1 as follows
counting frame to positions that
7 × 100 + 3 × 10 + 4 × 1
represent the size of a number.
It was invented long before the … 102 101 100 …
adoption of the written modern … 100 10 1 …
numeral system and is still in
7 3 4
use today.

Table 2.1.1 Place values for the decimal system

Table 2.1.2 shows how the place values can be extended to include fractions,
Information thousands and ten thousands.

Base 10 system is an example

… 104 103 102 101 100 10−1 10−2 …

of a positional number system. … 10000 1000 100 10 1 ¹⁄₁₀ ¹⁄₁₀₀ …

This type of system was first 2 7 7 3 4 3 5
used by the Babylonians over
4000 years ago in Mesopotamia, Table 2.1.2 Some more place values for the decimal system
modern day Iraq. Positional
number systems are good for
The number represented by the numeral 27734•35 in base 10 is
doing arithmetic with. constructed using the place values shown in Table 2.1.2 as follows
2 × 10000 + 7 × 1000 + 7 × 100 + 3 × 10 + 4 × 1 + 3 × ¹⁄₁₀ + 5 × ¹⁄₁₀₀
To indicate the base we use a suffix attached to the numeral, e.g. 3410.

Binary (base 2)
The number base of the binary system is two because it has two digits 0, 1 and
the digit multiplier is a power of two, 2n where n is … −3, −2, −1, 0, 1, 2 , 3, …

Single licence - Abingdon School 24

5 Fundamentals of data representation

Key concept
… 24 23 22 21 20 2−1 2−2 …
Decimal: … 16 8 4 2 1 …
¹⁄₂ ¹⁄₄
The number base of the decimal
1 0 1 1 1 1 1
system is ten because it has ten
digits 0, 1, 2, 3, 4, 5, 6, 7, 8, Table 2.1.3 Place values for the binary system
9 and the digit multiplier is a
power of ten, 10n where n is The number in decimal represented by the binary numeral 10111•11
…, −3, −2, −1, 0, 1, 2 , 3, …
is constructed using the place values in Table 2.1.3 as follows
1 × 16 + 0 × 8 + 1 × 4 + 1 × 2 + 1 × 1 + 1 × ¹⁄₂ + 1 × ¹⁄₄

To indicate the base we use a subscript attached to the numeral, e.g. 10111•112.
Now the “There are 10 types of people in the world those that understand
Key point binary and those that don’t” quote might make more sense because
To indicate the base we use a 10Binary = 2Decimal
suffix attached to the numeral,
e.g. 3410.
Questions
1 Write out each of the following in the form

digit × multiplier + digit × multiplier + …

(a) 10102 (b) 1111•112 (c) 10•01012

2 Convert the following numbers expressed in binary to their

decimal equivalent:

Key concept (a) 10102 (b) 11112 (c) 100101112 (d) 111111112

Binary:
3 Convert the following numbers expressed in binary to their
The number base of the binary
system is two because it has decimal equivalent:
two digits 0, 1 and the digit
multiplier is a power of two, 2n (a) 10•102 (b) 0•11112 (c) 100•101112 (d) 1•11111112
where n is
…, −3, −2, −1, 0, 1, 2 , 3, …

Hexadecimal (base 16)

The number base of the hexadecimal system is sixteen because it has sixteen
digits 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, A, B, C, D, E, F and the digit multiplier is a
power of sixteen, 16n where n is …, −3, −2, −1, 0, 1, 2 , 3, …
The number in decimal represented by the hexadecimal numeral D4
is constructed using the place values in Table 2.1.4 as follows
13 × 16 + 4 × 1

where D has been replaced by 13.

25 Single licence - Abingdon School

 5.2.1 Number base

Key concept
… 161 160 …
… 16 1 … Hexadecimal:
D 4 The number base of the
hexadecimal system is sixteen
Table 2.1.4 Place values for the hexadecimal system because it has sixteen digits 0,
1, 2, 3, 4, 5, 6, 7, 8, 9, A, B, C,
The hexadecimal digits A, B, C, D, E and F are in decimal 10, 11, 12, 13, 14
D, E, F and the digit multiplier
and 15 respectively. is a power of sixteen, 16n where
The number in decimal represented by the hexadecimal numeral 38AD4•95 is n is
…, −3, −2, −1, 0, 1, 2 , 3, …
constructed using the place values in Table 2.1.5 as follows
3 × 65536 + 8 × 4096 + 10 × 256 + 13 × 16 + 4 × 1 + 9 × 1/16 + 5 × 1/256

To indicate the base we use a subscript attached to the numeral, e.g. Information
38AD4•9516.
In some programming
languages, e.g. Java, a number
… 164 163 162 161 160 16−1 16−2 …
represented in hexadecimal is
… 65536 4096 256 16 1 ¹⁄₁₆ ¹⁄₂₅₆ …
indicated by placing 0x before
3 8 A D 4 9 5 the numeral, e.g. 0x3c4 or
0x3C4.
Table 2.1.5 Some more place values for the hexadecimal system

Questions
4 Write out each of the following in the form

digit × multiplier + digit × multiplier + …

(a) 102316 (b) 1F16 (c) F•1316

5 Convert the following numbers expressed in hexadecimal to their
decimal equivalent:

(a) 102316 (b) 1F16 (c) FFFF16 (d) DEAD16

Converting from decimal to binary

Method 1
Take the decimal number to be converted and find between which two column
place values it lies, e.g. 3510 lies between columns with place values 32 and 64,
respectively. Place 1 in the column with the lower of the two place values and 0
in the higher of the two as shown in Table 2.1.6. With the given example, take
the place value 32 away from the decimal number, leaving 310. Place 0 in all the
columns with place values greater than 310. It is then trivial to see that we need
one 2 and one 1 to match 310.

Single licence - Abingdon School 26

5 Fundamentals of data representation

… 26 25 24 23 22 21 20 …
… 64 32 16 8 4 2 1 …
0 1 0 0 0 1 1

Table 2.1.6 Some place values for the binary system

Questions
6 Convert the following numbers expressed in decimal to their binary
equivalent using Method 1.

(a) 3310 (b) 2410 (c) 5810 (d) 12710

Key method Method 2 - the method of successive division

Take the decimal number and repeatedly divide by 2 writing down the
Example:
Decimal to decimal by remainder each time as shown in Table 2.1.7, stopping when zero is reached.
successive division, picks out The binary equivalent of 3510 is read from the remainder column beginning at
the individual digits
the last row and working up the table.
e.g. n = 46210

Quotient New number Remainder

n n⁄₁₀ r
35/2 17 1
462 46 2
17/2 8 1
46 4 6
8/2 4 0
4 0 4
4/2 2 0
2/2 1 0
Where r is the remainder.
1/2 0 1
The remainder supplies the
individual digits, one at a time,
e.g. 2.
Table 2.1.7 Successive division by 2 method

Questions
7 Convert the following numbers expressed in decimal to their binary
equivalent using Method 2. Show the intermediate results in a table
with structure similar to Table 2.1.7.

(a) 3310 (b) 2410 (c) 5810 (d) 12710

Why does the method of successive division work?

We note that if a decimal number, n, is even then there is some integer, k for
which
n = 2k i.e. n = 2k + 0

27 Single licence - Abingdon School

 5.2.1 Number base

e.g. n = 62, k = 31,

∴ 62 = 2 × 31 + 0
We call 0 the remainder. In this example, 2 goes into 62, 31 times exactly.
On the other hand, if a decimal number, n, is odd then there is some integer, k
for which
n = 2k + 1

e.g. n = 63, k = 31

∴ 63 = 2 × 31 + 1
We call 1 the remainder. In this example, 2 does not divide 63 exactly.
The first 1 or 0 remainder is the least significant bit of the decimal number’s
binary equivalent and the final remainder 1 or 0 remainder, the most significant
bit of the binary equivalent.
Successive division algorithm decimal to binary
For decimal number, n.
Make k the value of n
If k is equal to 0 write down the answer 0
While k is not equal to 0
Make the new value of k the old value
divided by 2 using integer division
If this is the first pass write down remainder
Else write down remainder to the left
of the previous remainder

Programming task
1 Code this successive division algorithm in a programming language
with which you are familiar. Test your program by converting the
following decimal numbers

(a) 010 (b) 2410 (c) 5910 (d) 12710 (e) 3310

Converting from decimal to hexadecimal

We can use the method of successive division similar to the one used for
decimal to binary conversions, this time dividing by 16. Table 2.1.8 shows
a worked example for n = 31910 . The last column is read from the last row
upwards giving 13F16.

Single licence - Abingdon School 28

5 Fundamentals of data representation

Quotient New number Remainder

319/16 19 15 (F)
19/16 1 3
1/16 0 1

Table 2.1.8 Successive division by 16 method

Questions

8 Convert the following numbers expressed in decimal to their

hexadecimal equivalent using the algorithm above. Show the
intermediate results in a table with structure similar to
Table 2.1.8.

(a) 4710 (b) 30210 (c) 6551710 (d) 28556210

Successive division algorithm decimal to hexadecimal

For decimal number, n.
Make k the value of n
If k is equal to 0 write down the answer 0
While k is not equal to 0
Make the new value of k the old value
divided by 16 using integer division
If this is the first pass write down remainder
using hexadecimal digit
Else write down remainder to the left
of the previous remainder
using hexadecimal digit

Programming task
2 Code this successive division algorithm in a programming language
with which you are familiar. Test your program by converting the
following decimal numbers

(a) 010 (b) 4710 (c) 30210 (d) 6551710 (e) 28556210

Use of Microsoft® Windows® programmer calculator

It is possible to use Microsoft Windows’ calculator to perform number
conversions by selecting the programmer view mode, entering a value in
the chosen base and then by changing to one of the other available bases.

29 Single licence - Abingdon School

 5.2.1 Number base

You could use this calculator to check your answers to questions about number
bases.

Figure 2.1.1 Screenshot of Microsoft® Windows® calculator in decimal mode

Converting from hexadecimal to binary

This can be done in a straightforward way as follows:

Write down the number in hexadecimal

Replace each hexadecimal digit
by its binary equivalent
using 4 binary digits B47A

B47A16 = 10110100011110102
1011 0100 0111 1010

The method relies on the fact that the hexadecimal digits 0 to F map to 0 to
15 in decimal and this decimal range can be coded by just four binary digits.
When a number represented in four binary digits is multiplied by 1610, it
becomes a number represented by eight binary digits with zeroes in the least
significant four bit positions, twelve binary digits when multiplied by 1610 again
and so on.

1 0 1 1 × 1610 = 1 0 1 1 0 0 0 0

Questions
9 Convert the following numbers expressed in hexadecimal to their
binary equivalent using the method described above.

(a) 4716 (b) 3A216 (c) 6FE716 (d) BEEF16

Single licence - Abingdon School 30

5 Fundamentals of data representation

Converting from binary to hexadecimal

This can be done in a straightforward way as follows:

Write down the number in binary

Add leading 0s to the left-hand side of
the bit pattern so that the number of bits
is a multiple of 4 (if necessary)
Replace each block of four binary digits

1011 0100 0111 1010 by their hexadecimal equivalent

10110100011110102 = B47A16

B47A

Questions
10 Convert the following numbers expressed in binary to their
hexadecimal equivalent using the method described above.

(a) 11112 (b) 101011012 (c) 1011002

(d) 1100111000112

Hexadecimal as shorthand for binary

Long strings of 1s and 0s are difficult for a human to read so programmers
often switch to the hexadecimal equivalent because it is much easier to read. If
the strings of 1s and 0s represent executable code then debugging this code is
much easier if the code is displayed in hexadecimal form. Its meaning is easier
to determine than its binary
form.
Similarly, writing numbers in
hexadecimal form is less error
prone than writing the same
numbers in binary especially if
the binary form consists of long
strings of 1s and 0s. For example,
it would be cumbersome and
Figure 2.1.2 Machine code displayed in
error prone to specify the colour
binary
for text on a page of HTML in
24 binary digits, better to use the
shorthand form of hexadecimal, e.g. #1F040A. Here the # symbol is used to
indicate that the numeral is in hexadecimal.

31 Single licence - Abingdon School

 5.2.1 Number base

The contents of memory or registers of a computer system

can be displayed for debugging purposes. It is usual for Key point
the software that does this to display these contents in
Long strings of 1s and 0s are
hexadecimal because it is much easier for a human to read difficult for a human to read
the numbers in this form as well as taking up less space so programmers often switch
on the display screen. Software is needed because the to the hexadecimal equivalent
because it is
numbers are actually stored in the memory locations and
• much easier to read
the registers in base 2 form.
• more compact,
Memory addresses are more conveniently expressed in 4x fewer digits
hexadecimal than binary. For example, the memory • less error prone
Figure 2.1.3 The • easier to debug code
limit of Windows 7 is 4 GiB. This requires the use of same machine expressed in hexadecimal
32 binary digits to express the address of a particular
code expressed in • suitable for working with
memory word or location but in hexadecimal it requires hexadecimal digital hardware because
only 8 hexadecimal digits. Incidently, it would require an integral factor
relationship with binary
10 decimal digits. However, hexadecimal is more suitable
unlike decimal.
when working with digital hardware than decimal because hexadecimal uses
4x fewer digits than binary (³²⁄₄) but decimal uses 3.2x fewer (³²⁄₁₀), an
awkward factor to work with.

Figure 2.1.4 Microsoft® Windows® Device manager showing the allocation of memory

Single licence - Abingdon School 32

5 Fundamentals of data representation

Task

1 Explore memory with the memory viewer of a debugger in

the programming language environment that you use, e.g.
memory window in Visual Studio 2013, Xcode on Apple Mac, or
use a command line tool such as
cat /proc/<processid>/maps in Linux, e.g. using Raspberry Pi.

In this chapter you have covered:

■■ The meaning of number base
■■ The decimal number base
■■ The binary number base
■■ Converting from decimal to binary
■■ The method of successive division
■■ Converting from decimal to hexadecimal
■■ Converting from hexadecimal to binary
■■ Converting from binary to hexadecimal
■■ Hexadecimal as a shorthand for binary.

33 Single licence - Abingdon School

 5 Fundamentals of data representation
5.3 Units of information
Learning objectives:
■ The bit is the fundamental
unit of information
■ 5.3.1 Bits and bytes
Information
■ A byte is a group of 8 bits
The number of penguins can be represented by many
n
■ Know that 2 different values symbols,
can be represented with n bits e.g. 6 VI six 0110 |||||| 3 + 3 六
We use symbols all the time when
Background
we communicate. Animals also use
Sounds can have symbolic symbols to communicate. Special sounds or movements
meaning i.e. are symbols for
are used by animals to attract a partner and other sounds
something.
and movements are used to warn of danger. Figure 3.1.1
Gestures or movement can have shows a peacock with its tail fully extended. This display of
symbolic meaning i.e. are symbols tail feathers is a form of communication.
for something.
Figure 3.1.1 Peacock display
Special sounds are made by
humans too when we speak but our use of such symbols is
considerably more advanced than that of animals.
Key point Humans also use symbols when writing words and sentences on paper and on
electronic devices; when drawing pictures on paper and painting paintings.
Symbols communicate
information. We also use gestures and write music using musical notation and so on.
The use of the symbols is not decorative instead their use is
A symbol is an information carrier.
to communicate something. That something is information.
In other words, the symbol is an information carrier.

Questions
Key principle
A representation is a pattern of
1 State the information conveyed by the following symbols:

♩ ♪
symbols that conveys information,
e.g. a pattern of 1s and 0s.  He Ar

2 The word sign is sometimes used in place of the word

symbol. Figure 3.1.2 shows road signs. State the
Background information conveyed by these signs.

The words sign and symbol are (a) (b) (c)

equivalent.
Figure 3.1.2 Road signs

Single licence - Abingdon School 34

5 Fundamentals of data representation

Key concept Information and data

Information is made of data put together according to the rules (syntax) that
Datum: govern the way the chosen symbols are used. Syntax determines the form,
A datum (plural data) is any
construction, composition, or structuring of something. The data must also
physical phenomenon or object
that carries information, e.g. road have meaning or be meaningful. This means that data must comply with
sign object, speech. the meanings (semantics) of the chosen symbol system, code, or language
There can be no information in question. It is not restricted to language but could, for example, be
without data. Data is how
pictorial. The data-based definition of information is thus summarised as
information is represented.
Information = data + meaning
Key concept Types of information
The road signs shown in Q2(a), (b) and (c) are informational of a factual kind,
Physical phenomenon:
e.g. Q2(a) has the meaning, the road ahead narrows, that is
A natural phenomenon involving
the physical properties of matter a fact. On the other hand, information of the instructional
and energy, e.g. production or kind is supplied by, for example, the GIVE WAY sign. The
transmission of sound. sign has a meaning but that meaning is an instruction.
Both factual and instructional data belong to a category of
Background information called semantic content. Semantic content is associated with an
Genetic information in the form intelligent producer/consumer pair. Figure 3.1.3 shows the datum True being
of genes are instructions which
sent from a producer to a consumer. This datum has the status of information
together with other essential
ingredients serve the purpose of
at the producer end and at the consumer end where the information = it is
controlling and guiding the devel- raining. If the datum False was sent instead, then this would be interpreted at
opment of organisms very much the producer and the consumer ends as the information = it is not raining.
in the way that procedures in an
The datum True in transit is an uninterpreted symbol, i.e. its
imperative programming language
control the physical state in which meaning is not yet processed. It is the responsibility of the
an imperative program executes. consumer to interpret this datum and extract its meaning.
True

Key point ItIsRaining ⟵ True ItIsRaining ⟵ True

Datum
Information as semantic content Producer Consumer
can also be described as queries +
data, e.g. IsItRaining? + Yes, where Information Information
the query is IsItRaining? and the Figure 3.1.3 Semantic content information = data + meaning
datum is Yes.
Yes can be coded as 1. No as 0.
Environmental information on the other hand is information that is defined
relative to an observer who relies on it instead of having direct access to the
original data, e.g. the concentric rings visible in the wood of a cut tree trunk
provide information on the age of the tree and the growing conditions at the
time a ring was laid down. Note, environmental doesn’t mean it has to be
natural, e.g. the low battery indicator in Q1 is environmental information
because it reflects the state of the battery but it is not the battery.

35 Single licence - Abingdon School

 5.3.1 Bits and bytes

Bits
We have seen how to represent the
natural numbers in the decimal Figure 3.1.4 A collection of apples Did you know?
and the binary numeral systems.
Bit is derived from the b of binary
In the binary system, we use just two symbols, the binary digits 0 and 1 and the it of digit.
to represent a natural number, for example, the number of apples shown
in Figure 3.1.4 in binary is 1012. Rather than use the term binary digit
we can abbreviate it to bit. So we need three bits to count five apples.
This digital data 1012 encodes the information that we have five apples. If
we remove one apple, our digital data must change to 1002 to convey the
new information that we have four apples. Note that digital data always
changes in discrete steps. The minimum step in our apples’ example is one.
Removing more apples, say three, leaves just one apple and to convey this
information we must change our digital data to 12. Removing the last apple
leaves none and our digital data becomes 02 to convey this information.
In a similar manner, the datum True and the datum False in
Figure 3.1.3 could be encoded as 1 and 0, respectively.

Coin tossing
Now rather than counting objects and recording their number, let’s suppose
that we wish to convey the outcome of a coin tossing experiment. For the
experiment, assume our coin will land either head up (h) or tail up (t) with h
equal probability when tossed and we will call a single toss of the coin, a trial.

Now before we toss the coin, we cannot say that the outcome is (h) or the
outcome is (t). We are in a state of data deficit and therefore, in possession t
of no information about the outcome. However,
2-bit

ht
Outcome as soon as the coin is tossed we have an outcome,
encoding
(h) or a (t) which we can represent as 1 or 0. We
(tt) 00
now have data, a single bit of data which conveys
(th) 01
information, the result of the trial. We say that
(ht) 10
to represent the outcome we require one bit
(hh) 11
of information or one bit of information per
Table 3.1.1 symbol where a symbol is either (h) or (t).

Now let’s conduct a coin tossing experiment

using two unbiased coins. We have four possible outcomes, encoded as the
symbols (hh), (tt), (ht) and (th). Before conducting the experiment, the
data deficit is four units, the four symbols, because we have no information
about which outcome will actually occur. However, when we have an
outcome, say the symbol (ht), we remove a greater data deficit than we
do for the single coin experiment because each symbol in the two coin
experiment provides more information by excluding more alternatives.

Single licence - Abingdon School 36

5 Fundamentals of data representation

We need two bits to encode the possible two-coin experiment outcomes

as shown in Table 3.1.1, or two bits of information per symbol.

Questions
3 Complete a copy of Table 3.1.2 by replacing the blanks in the
Alphabet column and entering the missing information in the Bits of
information per symbol column.

No of Bits of information
Alphabet
coins per symbol
2 equiprobable symbols
1 1
(h), (t)
4 equiprobable symbols
2 2
(hh), (ht), (th), (tt)
8 equiprobable symbols
3 (hhh), (___), (___), (___)
(thh), (___), (___), (ttt)
16 equiprobable symbols
(hhhh), (____), (____), (____)
4 (hthh), (____), (____), (____)
(thhh), (____), (____), (____)
(tthh), (____), (____), (tttt)

Table 3.1.2

Key principle
The bit is the fundamental unit of information
Unit of information:
The bit is the fundamental unit of Imagine a machine that can answer only “42” to any question. There
information. is no data deficit because the answer to any question can be predicted
with absolute certainty, it is always the symbol “42”. Therefore, the
machine produces an amount of information which is zero.

The smallest amount of information occurs when we have two

equally likely choices which we know requires one bit. We therefore
use the bit as the fundamental unit of information.
Key concept
Byte: A byte is a group of 8 bits
The name used for a group of 8 It is convenient to group together bits and refer to the group
bits is byte.
by name. The name used for a group of 8 bits is byte.

Task
1 Investigate whether or not the programming language that you use
has a byte data type. If it doesn’t how could one be created for use in
programs that you might write?

37 Single licence - Abingdon School

 5.3.1 Bits and bytes

How different arrangements of n bits are there?

Figure 3.1.5 shows all possible 0 0 0 0
arrangements for 1, 2, 3, 4 bits. 0 0 0 1
Notice that the number of 0 0 1 0
arrangements doubles each time 0 0 1 1
0 1 0 0
we add another bit. Starting at one
0 1 0 1
bit, the number of arrangements
0 1 1 0 Key point
is 2 or 21. To double the number 0 1 1 1
of arrangements to 4 or 22 we just 0 0 0 1 0 0 0 Bit arrangements or bit patterns
add another bit making 2 bits. 0 0 1 1 0 0 1 can represent natural numbers.

To double again from 4 to 8 or 0 1 0 1 0 1 0 Decimal

3 bits
3
2 we add another bit making 3 0 1 1 1 0 1 1 value
bits. Doubling again we obtain 0 0 1 0 0 1 1 0 0 000 0
0 1 1 0 1 1 1 0 1 001 1
16 or 24 different arrangements
0 1 0 1 1 0 1 1 1 0 010 2
and we now have used 4 bits.
1 1 1 1 1 1 1 1 1 1 011 3
This suggests the relationship 100 4

16 ways
2 ways

4 ways

8 ways

between number of bits 101 5

and number of different 110 6
arrangements is as follows Figure 3.1.5 No of different arrangements 111 7
of n bits where n = 1, 2, 3, 4

Number of different arrangements of n bits = 2n

If each arrangement represents a value, e.g. a natural
number, then we can also say that Key concept
Number of different values that can be represented in n bits = 2n
Bit pattern:
An arrangement of bits is called a
No of bits Decimal integers No of integers Binary integers bit pattern, e.g. 100101.
1 0, 1 2 0, 1
2 0, 1, 2, 3 4 00, 01, 10, 11
000, 001, 010,
0, 1, 2, 3,
3 8 011, 100, 101, Key point
4, 5, 6, 7
110, 111
Number of different values that
Table 3.1.3 Number of different values for a given number of bits can be represented in
n bits = 2n
Questions
4 How many different arrangements are possible if the number of bits is
(a) 5 (b) 8 (c) 16 (d) 24 (e) 32?
Express your answers as both powers of 2 and fully evaluated.

5 How many different values are possible for the following number of
bytes
(a) 1 (b) 2 (c) 8?
Express your answers as powers of 2.

Single licence - Abingdon School 38

 5.3.1 Bits and bytes

Questions
6 Write all possible bit patterns for 4 bits and their corresponding
decimal natural number values in table format.

In this chapter you have covered:

■■ Symbols are used to communicate information
■■ The data-based definition of information:
Information = data + meaning

■■ Data is how information is represented

■■ Information can be factual or instructional
■■ A bit is a single binary digit, 0 or 1
■■ The bit is the fundamental unit of information
■■ Byte: The name used for a group of 8 bits is byte.
■■ Number of different values that can be represented in n bits = 2n

Single licence - Abingdon School 39

 5 Fundamentals of data representation
5.3 Units of information
Learning objectives:
■ Naming quantities of bytes

■ kibi, mebi, gibi, tebi

■ 5.3.2 Units
Quantities of bytes
■ kilo, mega, giga, tera Storage device manufacturers measure
capacity using the decimal system
(base 10), so 1 gigabyte (GB) is
calculated as exactly 1,000,000,000 bytes
or 1 billion bytes.
Figure 3.2.1 shows the reporting of the
capacity of a Western Digital hard disk.
On the other hand, the memory
capacity of RAM installed in machines
and quoted in GB is usually reported
by the OS using the binary system (base
2) of measurement. In binary, 1 GB
means 1,073,741,824 bytes, 2 GB is
Figure 3.2.1 Image of a part of the
therefore 2147483648 bytes as shown in
exterior of a hard disk showing
Figure 3.2.2. The RAM capacity GB is
storage capacity of 160.0 GB quoted
therefore a different unit from the disk
to 4 significant figures
storage GB.

Figure 3.2.2 Command line Microsoft® Windows® 7 showing capacity in

bytes of RAM chips – 2 in total each 2GB
This rather confusing situation has been resolved by the gradual adoption of the
International Electrotechnical Commission (IEC) standard for binary prefixes,
Background
which specify the use of gigabyte (GB) to strictly denote 1000000000 bytes
The bi in prefix gibi refers to and gibibyte (GiB) to denote 1073741824 bytes. This standard is now part of
binary. the International System of Quantities.

Single licence - Abingdon School 40

5 Fundamentals of data representation

Figure 3.2.3 shows the use of Gi and Ki for reporting disk storage capacity
using the command df –h in terminal mode on an Apple® MacBook Pro®
running operating system OS X® 10.8.5. The About This Mac window is also
shown.

Figure 3.2.3 shows the use of the units Gi and Ki

Task

1 Using the command line of your computer, investigate the

Information capacity of the
The following command to (a) hard disk drive attached to your computer
obtain disk capacity is available (wmic diskdrive get size on Microsoft Windows,
at the command line in wmic diskdrive get /? for more options)
Windows:
(b) RAM installed in your computer
wmic logicaldisk get size,
(hostinfo | grep memory on Apple Mac computers,
freespace, caption
wmic memorychip get capacity on Microsoft Windows)

Powers of 2
Table 3.2.1 shows some numbers in decimal numerals expressed as powers of
2 and their equivalent binary numeral. In 2 raised to the power of 10, 10 is
known as the exponent. The exponents 10, 20, 30, 40 specify the number of
zeroes in the binary numeral.

Decimal Power
Binary numeral
numeral of 2
1024 210 10000000000
1048576 220 100000000000000000000
30
1073741824 2 1000000000000000000000000000000
40
109951162776 Table
2 3.2.1100000000000000000000
Decimal numerals expressed as powers of 2
00000000000000000000

Questions
1 Express the following binary numerals in the form 2n.
(a) 10002 (b) 10000002 (c) 10000000000000002

2 Express the following decimal numerals in the form 2n.

(a) 1024 (b) 512 (c) 2048 (d) 4096 (e) 2097152

41 Single licence - Abingdon School

 5.3.2 Units

To avoid writing out long strings Name Symbol Power of 2

of zeroes, the names, symbols and kibi Ki 210
corresponding powers of 2 are used as mebi Mi 220
shown in Table 3.2.2. gibi Gi 230

If the binary numeral refers to a tebi Ti 240

quantity of bytes then we can express

Table 3.2.2 Unit name, symbol and
the quantity using the units of Ki, Mi,
corresponding power of 2
Gi and Ti as shown in Table 3.2.3.
B refers to byte.

Using symbol Using named

Key fact
Decimal Power Using form of unit unit for kibi, Ki - 210
numeral of 2 units for quantities quantities of mebi, Mi – 220
of bytes bytes gibi, Gi – 230
tebi, Ti – 240
1024 210 1Ki 1KiB 1 kibibyte
1048576 220 1Mi 1MiB 1 mebibyte
kilo, k - 103
1073741824 230 1Gi 1GiB 1 gibibyte
mega, M – 106
109951162776 240 1Ti 1TiB 1 tebibyte
giga, G – 109
tera, T – 1012
Table 3.2.3 Quantities of bytes expressed in units

Questions
3 Convert the following to bytes
(a) 1MiB (b) 1.5KiB (c) 1.75GiB

4 Convert the following quantities in bytes to KiB

(a) 1024 (b) 512 (c) 2048 (d) 4096

5 Convert the following quantities in bytes to MiB

(a) 1048576 (b) 6291456 (c) 4718592 (d) 9437184

Single licence - Abingdon School 42

5 Fundamentals of data representation

Power of Denary numeral Exponent Powers of 10

Ten
Figure 3.2.4 shows the decimal numeral corresponding to a given power of 10.
1012 1000000000000 12
The power is known as the exponent. The exponent specifies the number of
11
10 100000000000 11
zeroes in the decimal numeral.
1010 10000000000 10 Power of
109 1000000000 9
Name Symbol
To avoid writing out long strings 10
103
8
10 100000000 8
of zeroes, the names, symbols and kilo k
10 7
10000000 7
corresponding powers of 10 are used as mega M 106
10 6
1000000 6 giga G 109
shown in Table 3.2.4.
105 100000 5 tera T 1012
104 10000 4 Table 3.2.4 Unit name, symbol and
3
10 1000 3 corresponding power of 10
102 100 2

101 10 1
Table 3.2.5 shows how to express a decimal numeral which is a power of 10 in
Figure 3.2.4 Powers of 10 units of k, M, G and T.
If the decimal numeral refers to a quantity of bytes then we can express the
quantity using the units of k, M, G and T.

Using symbol Using named

Decimal Power Using form of unit unit for
numeral of 10 units for quantities quantities of
of bytes bytes
1000 103 1k 1kB 1 kilobyte
10000 104 10k 10kB 10 kilobytes
100000 105 100k 100kB 100 kilobytes
1000000 106 1M 1MB 1 megabyte
10000000 107 10M 10MB 10 megabytes
100000000 108 100M 100MB 100 megabytes
1000000000 109 1G 1GB 1 gigabyte
10000000000 1010 10G 10GB 10 gigabytes
100000000000 1011 100G 100GB 100 gigabytes
1000000000000 1012 1T 1TB 1 terabyte

Table 3.2.5 Quantities of bytes expressed in units k, M, G and T

Questions
6 Express the following decimal numerals in the form 10n
(a) 1000 (b) 1000000 (c) 10000000
7 Convert the following quantities in bytes to kB
(a) 1000 (b) 10000
8 Convert the following quantities in bytes to MB
(a) 500000 (b) 2000000 (c) 30000000

43 Single licence - Abingdon School

 5.3.2 Units

Data transfer units

Data transfer rates are normally expressed in bits per second using the units k,
M, G, T, e.g. 1Mb/s, where the meaning of 1Mb/s is 1 megabit per second or
1000000 bits per second. A lowercase b is used to indicate bits.

Questions
9 Convert the following data transfer rates to bits per second
(a) 1Mb/s (b) 100kb/s (c) 1Gb/s

In this chapter you have covered:

■■ How quantities of bytes are named
■■ The use of the prefixes kibi, mebi, gibi, tebi
■■ The use of the prefixes kilo, mega, giga, tera

Single licence - Abingdon School 44

5 Fundamentals of data representation
5.4 Binary number system
Learning objectives
■■Unsigned binary ■■ 5.4.1 Unsigned binary
■■Range of unsigned binary Non-negative values
Min value = 0
In this coding scheme, the numbers that can be coded are limited to
Max value = 2n – 1 for n bits
nonnegative values. For example, the numbers expressible in four bits for
unsigned binary are as shown in Figure 4.1.1. This figure also shows the
decimal equivalent values.

Key fact Decimal Unsigned Decimal Unsigned

Unsigned binary: value binary value binary
In unsigned binary numbers are value value
limited to non-negative values. 0 0000 8 1000
1 0001 9 1001
2 0010 10 1010
3 0011 11 1011
4 0100 12 1100
5 0101 13 1101
6 0110 14 1110
7 0111 15 1111

Figure 4.1.1 Table of unsigned binary codes in four bits and their
decimal equivalent values.

When coding numbers in unsigned binary, the weights of each binary position
in decimal are as shown in Figure 4.1.2. Notice that the next significant digit
weighting in decimal is obtained from the previous one by multiplying by 2.
Figure 4.1.2 shows that the number with decimal representation 12 has an
unsigned binary representation of 00001100 in 8 bits.

Decimal
Most significant weighting Least significant
binary digit binary digit

128 64 32 16 8 4 2 1
0 0 0 0 1 1 0 0

Figure 4.1.2 Decimal weighting of binary digits in unsigned binary coding.

Single licence - Abingdon School 45

5 Fundamentals of data representation

Questions
1 Convert the following decimal values to unsigned binary using 8
bits:
(a) 5 (b) 129 (c) 253

2 Convert the following unsigned binary numbers to decimal:

(a) 10100001 (b) 01111010 (c) 11111111

Minimum and maximum values

The range of numbers that can be coded in unsigned binary depends upon the
Key fact number of bits that are allocated to represent the number. Obviously, with just
Range of numbers in
one bit only two numbers can be coded, zero and one, giving a range of
unsigned binary in n bits: minimum = 02
Min value = 0
Max value = 2n – 1 maximum = 12
With two bits, four numbers can be coded in unsigned binary as follows:
00, 01, 10, 11, giving a range with

minimum = 002
maximum = 112
The minimum number is always zero but the maximum varies with the number
of bits used to represent the number. Figure 4.1.3 shows the maximum binary
numeral for 1, 2, 3, 4, 5, 6, 7 and 8 bits and the weighting in decimal for each
bit position.
128 64 32 16 8 4 2 1
1 1 1 1 1 1 1 1
1 1 1 1 1 1 1
1 1 1 1 1 1
1 1 1 1 1
1 1 1 1
1 1 1
1 1
1
Figure 4.1.3 Maximum binary numeral for a given number of bits

Figure 4.1.4 shows the maximum number expressed as a decimal numeral for
1, 2, 3, 4, 5, 6, 7, 8 and n bits.

46 Single licence - Abingdon School

 5.4.1 Unsigned binary

No of Maximum In unsigned In compact

bits number binary decimal form
in decimal
1 1 1 21 - 1
2 3 11 22 - 1
3 7 111 23 - 1
4 15 1111 24 - 1
5 31 11111 25 - 1
6 63 111111 26 - 1
7 127 1111111 27 - 1
8 255 11111111 28 - 1
n 1111…1111 2n - 1

Figure 4.1.4 Maximum number for a given number of bits

Generalising, the minimum and maximum values expressible in unsigned

binary for a given number of bits n is in decimal as follows

Minimum value = 0
Maximum value = 2n – 1

Questions

3 What is the largest number that can be represented in

unsigned binary for the following number of bits?
Express your answer in binary and decimal.

(a) 6 bits (b) 10 bits (c) 16 bits

In this chapter you have covered:

■■ Unsigned binary
■■ Range of unsigned binary
Min value = 0
Max value = 2n – 1 for n bits

Single licence - Abingdon School 47

5 Fundamentals of data representation
5.4 Binary number system
Learning objectives
■■Adding two unsigned binary
integers
■■ 5.4.2 Unsigned binary arithmetic
Adding two unsigned binary integers
■■Multiplying two unsigned The rules for adding numbers expressed in the binary numeral system are
binary integers
basically the same as for any other system. We add the contents of each column
in turn, starting from the right with the least significant digit column and
moving progressively leftward. Any carry from a column must be added to the
sum of the digits in the next column as shown in Figure 4.2.1 which shows the
Key principle
sum 011011002 + 001010102 of two 8-bit unsigned binary integers.
Addition of two unsigned binary
integers:
Apply the following rules to each This column uses the rule This column uses the rule
digit column 12 + 02 + carry 12 = 02 carry 12 12 + 12 = 02 carry 12

02 + 02 = 02
02 + 12 = 12 27 26 25 24 23 22 21 20
12 + 02 = 12 0 1 1 0 1 1 0 0
12 + 12 = 02 carry 12 0 0 1 0 1 0 1 0
02 + 02 + carry 12 = 12 1 Carry 1 0 Carry 1 0 1 Carry 1 0 1 1 0
02 + 12 + carry 12 = 02 carry 12
12 + 02 + carry 12 = 02 carry 12 Least
12 + 12 + carry 12 = 12 carry 12
significant digit
This column uses the rule
column
02 + 02 + carry 12 = 12
This column uses the rule
12 + 12 = 02 carry 12

Figure 4.2.1 Addition of two 8-bit unsigned binary integers

The basic rules are as follows

02 + 02 = 02

02 + 12 = 12

12 + 02 = 12

12 + 12 = 02, carry 12 to the next column

since there is no symbol for 2.

The last rule states that 12 + 12 = 102.

Single licence - Abingdon School 48

5 Fundamentals of data representation

If we have a carry from the previous column then the carry must be added to
the sum of the two digits of the current column. So we have the additional
rules
02 + 02 + carry 12 = 12

02 + 12 + carry 12 = 02 carry 12

12 + 02 + carry 12 = 02 carry 12

12 + 12 + carry 12 = 12 carry 12

Normally addition of two binary numerals representing unsigned binary

integers is set out in the manner of the example below
01101011
+00011011
10000110

Questions

1 Complete the following additions of two 4-bit unsigned binary

integers:
(a) 0 1 1 0 (b) 0 1 0 1
+ 0 0 0 1 +0101

2 Complete the following additions of two 8-bit unsigned binary

integers:
(a) 0 1 1 0 1 0 1 1 (b) 1 1 0 1 0 1 0 1
+ 0 0 0 1 1 0 1 1 +01011101

Multiplication of two unsigned binary integers

Multiplication in binary is performed in a similar manner to a decimal long
multiplication problem. For example, the decimal long multiplication problem
45610 x 4310 would be done as follows

weighting 104 103 102 101 100

multiplicand 4 5 6 456 x 3 = 1368

multiplier x 4 3

1st partial product 1 3 6 8

2nd partial product 1 8 2 4 0

1 9 6 0 8 456 x 40 = 18240
product

49 Single licence - Abingdon School

 5.4.2 Unsigned binary arithmetic

The multiplicand 45610 is multiplied by each digit of the multiplier 4310

separately and then the partial products are added giving appropriate weighting
to the implied power of 10 of each digit of the multiplier.
If we wanted the result of 1012 x 112 where each numeral represents an
unsigned binary integer then the binary long multiplication would be done as
follows
weighting 24 23 22 21 20

multiplicand 1 0 1 1012 x 12 = 1012

multiplier x 1 1

1st partial product 1 0 1

2nd partial product 1 0 1 0

1 1 1 1 1012 x 102 = 10102

product

Notice that because each digit of the multiplier 112 in the above example is a 1,
the multiplicand 1012 is just copied and then shifted either zero or one places
to the left to produce the corresponding partial product, 1012 or 10102. The
number of shifts to perform is the same as the exponent of the weighting, i.e. 0
or 1.
Extending this to a multiplication of larger numbers we see that binary
multiplication consists of copying and shifting the multiplicand,
e.g. 110111012 x 10112 = 1001011111112 as shown in Figure 4.2.2.
Key principle
Multiplication of two unsigned
211 210 29 28 27 26 25 24 23 22 21 20
binary integers:
For each 1 in the multiplier, copy
1 1 0 1 1 1 0 1 22110
the multiplicand and place below
the last partial product but shifted
x 1 0 1 1 1110
left by a number of columns equal
to the exponent of the weight for
1 1 0 1 1 1 0 1
this 1.
1 1 0 1 1 1 0 1 0 For each 0 in the multiplier,
change every digit in a copy of the
0 0 0 0 0 0 0 0 0 0 multiplicand to 0 and place the
copy as before.
1 1 0 1 1 1 0 1 0 0 0 Sum the partial products to obtain
the product.
1 0 0 1 0 1 1 1 1 1 1 1 243110 12 + 12 + carry 12 = 12 carry 12

Figure 4.2.2 Long multiplication of 110111012 by 10112

Single licence - Abingdon School 50

5 Fundamentals of data representation

Questions

3 Use the long multiplication method for the multiplication of

two unsigned binary integers to evaluate the following. Check
your answer is correct by converting the multiplicand and
multiplier to decimal, multiplying out and then comparing
with your unsigned binary answer.
(a) 1012 x 102 (b) 1012 x 112 (c) 10012 x 102

(d) 10012 x 112 (e) 10012 x 1012

(f ) 10010112 x 11012 (g) 10111002 x 10102

(h) 101111012 x 111012

In this chapter you have covered:

■■ Adding two unsigned binary integers
■■ Multiplying two unsigned binary integers

51 Single licence - Abingdon School

5 Fundamentals of data representation
5.4 Binary number system
Learning objectives
■■Two’s complement
representation of negative and
■■ 5.4.3 Signed binary using two’s complement
positive integers Representing negative integers
Numbers smaller than zero are called negative numbers. Most humans but not
■■Converting between signed all, accountants are the exception, place the symbol, -2 -1 0 1 2 3 4
binary and decimal and vice
'-', before a natural number greater than zero, e.g. 2,
versa
to indicate a negative integer, e.g. -2.
■■Subtraction using two’s The term, negative or minus sign, is used for '-'. The '+' symbol used to indicate
complement a positive integer, is called the positive or plus sign. If there is no sign before
■■Range of two’s complement the number, it is assumed to be positive. Integers can be positive or negative or
representation for a given zero.
number of bits Computations are carried out in digital computers using
binary to represent numbers because the binary system is
ideally suited to the electronic circuits in digital computers.
Did you know?
These circuits operate using two different levels of voltage
Accountants are different, they
place brackets around a natural which map easily to the two symbols, 0 and 1, of the binary system. However,
number to indicate a negative there is no third level of voltage to map to the symbol '-'. Therefore, we have to
integer, e.g. (567) means -567. rely on the two symbols, 0 and 1 to indicate
Bin Dec
both the magnitude (size) and the sign of a
000 0
number.
001 1
Key fact There are several choices of representation 010 2
For integers represented in for positive and negative integers in binary 011 3
two’s complement binary: one of which is two’s complement. 100 -4
• 1 in the most significant bit 101 -3
In Table 4.3.1, the column headed Bin
position indicates a negative 110 -2
integer and a 0, a positive contains 3-bit binary integer numerals 111 -1
integer and the column headed Dec contains the
• The most negative integer
Table 4.3.1 Two’s complement
corresponding decimal numerals for these
occurs with 1 in the most representation of negative and
integers. If you study Table 4.3.1 carefully,
significant bit position and all 0s positive integers
in the other positions
you will observe that Dec is a negative
• For -1, every bit is 1 integer only when the most significant digit
• For the most positive integer of Bin is 1 and Dec is a positive integer or zero only when the most significant
every bit is 1 except the most digit of Bin is 0. This representation in binary of integers is known as two’s
significant bit which is 0.
complement.

Single licence - Abingdon School 52

5 Fundamentals of data representation

Table 4.3.2 shows Bin using 4 bits to represent integers in binary. Again,
Bin Dec
negative numbers are represented in binary with the most significant digit 1
0000 0
and positive numbers with the most significant digit 0.
0001 1
0010 2 Whatever the number of bits:
0011 3
• 1 in the most significant bit position indicates a negative integer and a 0, a
0100 4
positive integer
0101 5
0110 6 • The most negative integer occurs with 1 in the most significant bit position
0111 7 and all 0s in the other positions
1000 -8
• For -1, every bit is 1 -8 4 2 1
1001 -7
1010 -6 • For the most positive value every bit is 1 0 0 0 0
1011 -5 except the most significant bit which is 0. 0 0 0 1
1100 -4 0 0 1 0
The sign bit in two’s complement is always the
1101 -3 0 0 1 1
most significant digit. 0 1 0 0
1110 -2
1111 -1 To achieve the range -810 to +710 the weighting for 0 1 0 1
each bit must be as shown in Table 4.3.3. Notice 0 1 1 0
Table 4.3.2 Two’s complement
that the most significant digit, the sign bit, also 0 1 1 1
representation of negative and
1 0 0 0
positive integers has magnitude or size, i.e. a weighting of -810.
1 0 0 1
Table 4.3.4 shows how the weighting of each bit 1 0 1 0
position varies for integers in two’s complement 1 0 1 1
Key fact binary for a given number of bits, e.g. for 5 bits 1 1 0 0
Two’s complement sign bit: the most significant bit has a weighting of -16 in 1 1 0 1
The most significant bit is the decimal. The most significant digit is always the 1 1 1 0
sign bit. 1 1 1 1
sign bit and its weighting is always negative.
Its weighting is always negative. Table 4.3.3 4-bit
The bit positions are labelled starting with the
two’s complement
least significant digit which is given bit position 0.
representation of integers
n – 1
The most significant bit has weighting, -2
where n is the number of bits,
e.g. n = 8, -28 – 1 = -27 = -12810.

No of
Weighting
bits
8 -128 64 32 16 8 4 2 1
7 -64 32 16 8 4 2 1
6 -32 16 8 4 2 1
5 -16 8 4 2 1
4 -8 4 2 1
3 -4 2 1
2 -2 1
7 6 5 4 3 2 1 0
Bit position
Table 4.3.4 Two’s complement representation of integers showing weighting
of bit positions for different numbers of bits

53 Single licence - Abingdon School

 5.4.3 Signed binary using two’s complement

Questions
1 What is the weighting in decimal of the most significant bit if the
following number of bits are used to represent integers in two’s
complement binary
(a) 3 (b) 5 (c) 8 (d) 10 (e) 16?

2 Express your answers to Q1 in 2x format.

3 What is the binary numeral for the most negative integer in two’s
complement binary when the number of bits for the numeral is as
follows
(a) 3 b) 5 (c) 8 (d) 10 (e) 16?

4 What is the binary numeral for the most positive integer in two’s
complement binary when the number of bits for the numeral is as
follows
(a) 3 (b) 5 (c) 8 (d) 10 (e) 16?

Converting an integer from decimal to two’s complement Key principle

binary
We have two cases to consider, negative integers and non-negative integers. To convert from decimal to 2’s
complement binary:
Non-negative integers Non-negative
Treat the integer as unsigned and convert using one of the available methods • Treat the integer as unsigned
for converting unsigned decimal numerals such as Repeated Division By Two. • Convert to unsigned binary
• Pad out with leading zeroes
Write down the result in binary and place one or more zeroes in front of the up to the specified number
binary numeral up to the specified number of bits. For example, convert +1310 of bits
to two’s complement binary using 5 bits as follows Negative
+1310 → 11012 → 011012 Method 1
• As for non-negative
The most significant bit will always be 0 for non-negative integers. • Flip the bits
• Add 1
Negative integers
Treat the integer as unsigned and convert using one of the available methods Method 2
• As for non-negative
for converting unsigned decimal numerals such as Repeated Division By Two.
• Starting from the right, leave
Write down the result in binary and place one or more zeroes in front of the all the digits alone up to and
binary numeral up to the specified number of bits. including the first 1
• Flip all the other digits
Now there are two possible methods for the next stage.
Method 1
• Flip the bits
• Then add 1.

Single licence - Abingdon School 54

5 Fundamentals of data representation

For example, if the specified number of bits is 5 proceed as follows

Change to
unsigned

decimal Convert Insert 0 Flip bits Add 1
-1310 → 1310 → 11012 → 011012 → 100102 → 100112

Check: 100112 = -1610 + 210 + 110 = -1310

Method 2
• Starting from the right, leave all the digits alone up to and including the
first 1
• Then flip all the other digits.
For example,
Change to Flip all
unsigned Leave 1st other
decimal Convert Insert 0 bit alone bits
-1310 → 1310 → 11012 → 011012 → 0110 │12 → 100112

Questions
5 Convert the following integers expressed in decimal to 5-bit two’s
complement binary.
(a) +12 (b) -12 (c) +7 (d) -7 (e) -1

6 Convert the following integers expressed in decimal to 8-bit two’s

complement binary.
(a) +12 (b) -12 (c) -7 (d) +32 (e) -32 (f ) -128
(g) -1 (h) -63 (i) -76

Converting an integer from two’s complement binary to

Key principle decimal
• Set out the decimal weighting for each binary digit remembering that the
Two’s complement to decimal: most significant digit’s weighting is negative.
Sum the products of each bit
value and the corresponding • Sum the products of each bit value and the corresponding decimal
decimal weighting. weighting.
For example, for 111100012 proceed as follows
-12810 6410 3210 1610 810 410 210 110
1 1 1 1 0 0 0 1

1 x -128 + 1 x 64 + 1 x 32 + 1 x 16 + 0 x 8 + 0 x 4 + 0 x 2 + 1 x 1 = -1510

55 Single licence - Abingdon School

 5.4.3 Signed binary using two’s complement

Questions
7 Convert the following integers expressed in two’s complement
binary to decimal.

(a) 01011100 (b) 10100100 (c) 1000000 (d) 11111111

(e) 10000000 (f ) 01111111

Subtraction in two’s complement

Given two integers in two’s complement binary, it is possible to subtract one, B, Key principle
from the other, A, by two’s complementing B and then adding the result to A. Subtraction:
A–B → A + (-B) Perform addition with 2’s
complement of B
For example, 01012 – 00112 would be evaluated as follows
A–B → A + (-B)
01012 – 00112 → 01012 + (-00112) → 01012 + 11012 → (1)00102

The addition carried out is just binary addition but this can generate a carry (1)
which is ignored because we restrict the answer to the same number of bits as
we started with, i.e. 4.
Check: 01012 – 00112 = 510 – 310 = +210 = 00102
Another example, 01012 – 11112 would be evaluated as follows
01012 – 11112 → 01012 + (-11112) → 01012 + 00012 → 01102

Check: 01012 – 11112 = 510 – (-110 )= +610 = 01102

Questions
8 Evaluate the following 4-bit two’s complement binary integer
expressions using steps that involve only binary.

(a) 0111 - 0100 (b) 0100 - 1110 (c) 1101 - 1110

(d) 1111 - 1100 (e) 1100 - 0011

Computer hardware engineers like to use two’s complement binary for

arithmetic because they only need to design addition circuits and circuits that
flip bits (complement), no subtraction circuitry is required. The addition and
complementing circuits are easy to design.
Computer engineers also like to use two’s complement binary because there
is only one binary numeral for zero. Other representations have two binary
numerals for zero. Comparisons of two numerals is often done by subtracting
one from the other and checking to see if the answer is zero or not. Zero means
that the two numerals represent the same number.

Single licence - Abingdon School 56

5 Fundamentals of data representation

Range for a given number of bits

Key fact The range of integers that can be coded in two’s complement binary depends
Range: upon the number of bits that are allocated to represent the integer.
For 2’s complement the range of For example, in 8 bits, the most negative integer that can be represented in
a given number of bits, n is
two’s complement binary is 100000002 whose bits are associated with the
-2n –1 to 2n – 1 – 1 following decimal weightings:

-27 26 25 24 23 22 21 20
-12810 6410 3210 1610 810 410 210 110
1 0 0 0 0 0 0 0

and the most positive integer is 011111112 whose bits are associated with the
following decimal weightings:

-27 26 25 24 23 22 21 20
-12810 6410 3210 1610 810 410 210 110
0 1 1 1 1 1 1 1

Thus, the range in decimal is

-27 to (26 + 25 + 24 + 23 + 22 + 21 + 20)

but (26 + 25 + 24 + 23 + 22 + 21 + 20) = 12710 = 12810 – 1 = 27 - 1

Therefore,
the range for 8 bits is -27 to 27 – 1
In general,

for n bits the range is -2n –1 to 2n – 1 – 1

Questions
9 What is the range in decimal of integers represented in two’s
complement binary using

(a) 4 bits (b) 6 bits (c) 10 bits

(d) 16 bits?

Express your answers first in 2x format before evaluating.

In this chapter you have covered:

■■ Two’s complement representation of negative and positive integers
■■ Converting between signed binary and decimal and vice versa
■■ Subtraction using two’s complement
■■ Range of two’s complement representation for a given number of bits

57 Single licence - Abingdon School

5 Fundamentals of data representation
5.4 Binary number system
Learning objectives
■■Representing numbers with a
fractional part in
■■ 5.4.4 Numbers with a fractional part
Fixed point form
• Fixed point form
Calculations often produce results that are not whole numbers, e.g. 5¼, so
• Floating point form there is a need to represent values that have a fractional part in the language of
the digital computer, i.e. binary. The decimal system gives a clue to how to do
■■Decimal to binary fixed point
this in binary
■■Binary to decimal fixed point 100 10 1 1/10 1/100
■■Decimal to binary floating ------------------------------
point 1 3 6 • 7 5

■■Binary to decimal floating The number 136 • 75 represents 1 hundred, 3 tens, 6 units, 7 tenths and 5
point hundredths.
Figure 5.4.4.1 shows how unsigned numbers with a fractional part can be
represented in binary using 8 bits. The weighting of each bit has been selected
to allow three bits for the fractional part but we could have chosen a different
number of bits for the fractional part, if we had wanted to. Notice that the
weighting decreases by a factor of 2 between adjacent columns as shown.

÷2 ÷2 ÷2 ÷2 ÷2 ÷2 ÷2

Weighting 16 8 4 2 1 ½ ¼ ⅛
16 8 4 2 1 0.5 0.25 0.125
1 0 1 1 0 1 1 1

Figure 5.4.4.1 Interpreting a bit pattern when it represents an unsigned

number with a fractional part

10110 • 1112 = 16 + 4 + 2 + ½+ ¼ + ⅛ = 22⅞ = 22.87510

This coding is known as fixed point coding because the binary point is fixed in
position, in this example between the third and fourth bits from the right.

Questions
1 Given 8 bits with the binary point fixed in position between the
third and fourth bits from the right as in Figure 5.4.4.1, what is the
decimal representation for each of the following unsigned binary
numbers?
(a) 00011•100 (b) 00101•110 (c) 10000•101
(d) 11111•111

Single licence - Abingdon School 58

5 Fundamentals of data representation

Questions
2 Given 8 bits with the binary point fixed in position between
the fourth and fifth bits from the right what is the decimal
representation for each of the following unsigned binary numbers?

(a) 0001•0001 (b) 0010•0011 (c) 1111•0100 (d) 1010•0111

3 Given 12 bits with the binary point fixed in position between

the sixth and seventh bits from the right what is the decimal
representation for each of the following unsigned binary numbers?

(a) 100000•000001 (b) 111000•000010

(c) 001111•000011 (d) 110001•000111

Fixed point form of signed numbers

We use two’s complement representation to represent signed numbers with a
fractional part in binary as shown in the example in Figure 5.4.4.2.

-16 8 4 2 1 ½ ¼ ⅛
1 0 1 1 0 1 1 1

Figure 5.4.4.2 Interpreting a bit pattern when it represents a signed number

with a fractional part

10110 • 1112 = -16 + 4 + 2 + ½ + ¼+ ⅛ = -9⅛ = -9.12510

Questions

4 Given 8 bits with the binary point fixed in position between the
third and fourth bits from the right as in Figure 5.4.4.2, what is
the decimal representation for each of the following signed binary
numbers?

(a) 11100•100 (b) 11010•010 (c) 10111•011

(d) 11100•001

59 Single licence - Abingdon School

 5.4.4 Numbers with a fractional part

Floating point form Information

In decimal, large and small numbers are often represented using scientific
Scientific notation:
notation which is of the form
In scientific notation real no
A x 10B 13.5610 is written as 1.356 x 101.

where A is any real number greater than -10 and less than +10 and B is any
Using the letter E instead of 10,
integer, e.g. A = 1.356, B = 1. and omitting the multiplication
For example, sign, the real no above can be
rewritten as follows
6•58723 x 104 = 65872•3 where A = 6•58723 and B = 4, the
13.56 = 13.56E0
number of decimal places to shift the point right
= 1.356E1
= 0.1356E2

6 • 5 8 7 2 3 = 65872•3 = 135.6E-1
= 1356E−2, etc
-8•0000 x 10 = -8000•0 where A = -8•0000 and B = 3
3
These numerals are sometimes
6•0 x 10 = 0•0006 where A = 6•0 and B = -4, the number of
-4 called floating point numerals as
opposed to the decimal numerals
decimal places to shift the point right, i.e. shift right -4 places
such as 13.56 that are called fixed
which translates to a shift left of 4 places. point numerals.
A similar notation is used when two’s complement binary is used to represent
Floating point numeral:
signed numbers that range from small to large
Consists of an integer numeral,
M x BaseE e.g. 1, followed by a fraction
numeral, followed by an exponent
where M is called the mantissa or significand, E the exponent and the number
part.
base Base equals 2 in decimal.
The mantissa is any real number greater than or equal to -110 and less than +110. Fraction numeral:
A fraction part is a decimal point
The exponent expresses the number of binary places to shift the point right
followed by a positive integer
or left. numeral, eg .356
For example,
Exponent part:
M = 0•10000002 and E = 00102 = 210 An exponent part is the letter E
followed by an integer numeral,
e.g. E1.
0 • 1 0 0 0 0 0 02 = 10•0000002

and another example,

M = 0•10000002 and E = 11102 = -210 Key principle
Exponent:
0 0 0 • 1 0 0 0 0 0 02 = 0•0010000002 The exponent expresses the
number of binary places to shift
Shifting the point by one binary place to the right is equivalent to multiplying the point right or left.
by 2 and shifting the point by one binary place to the left is equivalent to
dividing by 2. This floating or shifting of the point gives floating point
representation its name.

Single licence - Abingdon School 60

5 Fundamentals of data representation

Information Representation of two’s complement floating point binary

Figure 5.4.4.3 shows how, given 8 bits, the representation can be divided into
IEEE floating point standard: a mantissa field, 4 bits, and an exponent field, 4 bits, and their corresponding
IEEE floating point standard is
weighting.
another representation of real
numbers commonly used in
computer design.
0 1 1 0 0 1 0 1
31 30 23 22 0
Mantissa Exponent
S E F

S is one bit representing the sign

of the number -1 0.5 0.25 0.125 -8 4 2 1
E is an 8-bit biased integer
representing the exponent 0 1 1 0 0 1 0 0
F is an unsigned integer

The true value represented is Figure 5.4.4.3 shows mantissa and exponent fields of a floating point
(-1)S x f x 2e number, 0•110 x Base0101 where Base equals 2 in decimal
S = sign bit
(-1)S → (-1) 0 = +1
The most significant bit of the mantissa is the sign bit. Its weighting is always
and (-1)1 = -1 -1. Therefore, the binary point is situated between the most significant bit and
where the next most significant bit of the mantissa.
e= E – bias
f = F/2n + 1
Likewise, the most significant bit of the exponent is a sign bit. The exponent
For single precision numbers is always integral, i.e. a whole number, either negative or positive or zero.
n=23, bias=127.
Example 1: To evaluate 0•110 x Base0101 in decimal, where Base equals 2 in
decimal, first calculate exponent in decimal
Exponent = 01012 = 4 + 1 = +510
Then move the binary point of the mantissa +510 places to the right and then
convert the mantissa to decimal (plus sign means move binary point right)
0•1102 → 011000•02 = +2410

Example 2: To evaluate 0•110 x Base1000 where Base equals 2 in decimal.

0 1 1 0 1 0 0 0

Exponent = 10002 = -810 (minus sign means move binary point left)
Move binary point of the mantissa 810 places to the left and convert mantissa
to decimal
0•1102 → 0•00000000112 = ¹⁄₅₁₂ + ¹⁄₁₀₂₄ = + 0.002929687510

Questions
5 Given 8 bits to represent a signed number in two’s complement floating point form, with 4 bits for the
mantissa and 4 bits for the exponent as shown in Figure 5.4.4.3, write down separately, for each of (a) to
(f ), the binary forms of the mantissa and the exponent then the decimal expansion

(a) 01000100 (b) 10100100 (c) 01001111 (d) 01110011 (e) 10000000 (f ) 11111100

61 Single licence - Abingdon School

 5.4.4 Numbers with a fractional part

Converting from decimal to fixed point binary

Unsigned decimal to unsigned fixed point binary
Key principle
To convert an unsigned decimal number, W . F, e.g. 5.7510 where W = 5 and
Decimal to binary fixed point:
F = 0.75, to fixed point binary proceed as follows
1. Consider the whole number
1. Consider the whole number part, W, and the fractional part, F part, W, and the fractional
separately. part, F separately.

2. Convert the whole number part, W, from decimal to binary using, for 2. Convert the whole number
example, the repeated division by two algorithm. part, W, from decimal to
binary using the repeated
3. Convert the fractional part, F, from decimal to binary in a given number
division by two algorithm.
of bits using the following algorithm.
Repeated multiplication by two algorithm: 3. Convert the fractional part,
F, from decimal to binary in
n ← 0 a given number of bits using
OrigF ← Fractional part F the repeated multiplication by
two algorithm.
Repeat

R ← F x 2

Write down the digit to the left of the

decimal point of R, call it D

n ← n + 1

F ← R - D

Until F = 0 Or F = OrigF

Or n = AllocatedFractionalNoOfBits

Table 5.4.4.1 shows how the fractional part 0.7510 is converted to 0.112 by this
algorithm. The algorithm terminates for 0.7510 on the condition F = 0. It
will always terminate on F = 0 when the denominator of the fractional part
involves the prime factor 2 only, if a sufficient number of bits are allocated to
the fractional part.

Fractional
R Digit D
part, F
0.75 0.75 x 2 = 1.5 1
0.5 0.5 x 2 = 1.0 1 Most
0 significant

Table 5.4.4.1 Conversion of fractional part, 0.7510 to 0.112 bit

Table 5.4.4.2 shows an example where the algorithm terminates on the

condition F = OrigF. Under these circumstances, the fractional decimal part
converts to a repeating binary part, e.g. 0.810 is converted to 0 • 11002 .

Single licence - Abingdon School 62

5 Fundamentals of data representation

Fractional
R Digit D
part, F
0.8 0.8 x 2 = 1.6 1
0.6 0.6 x 2 = 1.2 1 Most
0.2 0.2 x 2 = 0.4 0 significant
0.4 0.4 x 2 = 0.8 0 bit
(Previous 4 steps now
0.8
repeat)

Table 5.4.4.2 shows an example of a conversion which

results in a repeating binary pattern.
Table 5.4.4.2 shows an example of a conversion which results in a repeating
binary pattern.
Table 5.4.4.3 shows an example of a conversion which results in a repeating
binary pattern for which F ≠ 0 and F ≠ OrigF so condition
n = AllocatedFractionalNoOfBits is necessary to terminate loop.

Fractional
R Digit D
part, F
0.1 0.1 x 2 = 0.2 0
0.6 0.2 x 2 = 0.4 0 Most
0.2 0.4 x 2 = 0.8 0 significant
0.4 0.8 x 2 = 1.6 1 bit
0.6 0.6 x 2 = 1.2 1
0.2 0.2 x 2 = 0.4 0
0.4 0.8 x 2 = 1.6 1
0.6 0.6 x 2 = 1.2 1
(Previous 3 steps now
0.2
repeat)

Table 5.4.4.3 Repeating binary pattern for which F ≠ 0

and F ≠ OrigF.

Repeating bit patterns occur whenever the denominator of the fractional part
involves prime factors other than 2, e.g. 0.8 = ⁸⁄₁₀ = ⁴⁄₅ so the denominator
has a prime factor (5) other than 2.

Questions
6 Convert the following decimal numbers to fixed point binary
using the repeated multiplication by two algorithm.

(a) 0.375 (b) 0.4 (c) 0.7 (d) 0.703125

(e) 0.1

63 Single licence - Abingdon School

 5.4.4 Numbers with a fractional part

Questions
7 Convert the following decimal numbers to fixed point binary
using the repeated division by two algorithm for the whole number
part and the repeated multiplication by two algorithm for the
fractional part.

(a) 101.875 (b) 333.55

Programming tasks
1 Write a program that implements the repeated multiplication by
two algorithm. Test your program for cases (a) to (e) in Question 6.

Signed decimal to signed two’s complement fixed point binary

To convert a signed non-zero decimal number to two’s complement fixed point Key principle
binary: Signed decimal to signed 2's
• Ignoring the sign, convert decimal number to two’s complement fixed complement binary:
point binary in given number of bits
Method 1: Flip all bits then add
• If the sign was negative 1 to the least significant bit
Use one of the following two methods
Method 2: Starting from the
Either right, leave all the digits alone
up to and including the first 1
Method 1: Flip all bits then add 1 to the least significant bit
then flip all the other digits.
Or
Method 2: Starting from the right, leave all the digits alone up to
and including the first 1 then flip all the other digits.
For example, -3.7510 becomes in 8-bit fixed point binary 11100.0102 if three
bits are allocated to the fractional part.
Using method 1:
3.7510 → 00011.1102 → 11100.0012

→ 11100.0012 + 12 → 11100.0102

Questions
8 Convert the following decimal numbers to 8-bit fixed point
binary in which three bits are allocated for the fractional part.

(a) -1.25 (b) -7.5 (c) -1 (d) -15.875

Single licence - Abingdon School 64

5 Fundamentals of data representation

Key principle Converting from decimal to binary floating point form

The following algorithm is used to convert decimal numbers to binary floating
Decimal to binary floating point form when both mantissa and exponent use two’s complement and
point form:
the binary point is positioned between the most significant and next most
Convert decimal to two’s
complement fixed point significant bit of the mantissa.
binary using just as many
Convert decimal number to two’s complement fixed point
bits as required
binary using just as many bits as required
While point is not between
most significant and next While point is not between most significant and next most
most significant bit significant bit
Shift point left
(divide by 2) Shift point left (divide by 2)
Increment exponent
Increment exponent

For example, 3.7510 is in two’s complement 8-bit floating point binary with 3
bits for the exponent and 5 bits for the mantissa
0.11112 0102

as shown in Table 5.4.4.4, which starts from the fixed point form 011.11002.

Divide by 2 Mantissa Exponent Decimal value

Weighting -4 2 1 • ¹⁄₂ ¹⁄₄ ¹⁄₈ ¹⁄₁₆ -4 2 1

Number 0 1 1 • 1 1 0 0 0 0 0 3.7510
Most
-2 1 • ¹⁄₂ ¹⁄₄ ¹⁄₈ ¹⁄₁₆ Increment exponent
significant
0 1 • 1 1 1 0 0 0 1 (1 + ¹⁄₂ + ¹⁄₄ + ¹⁄₈ ) x 21 = 3.7510
bit
Move point -1 • ¹⁄₂ ¹⁄₄ ¹⁄₈ ¹⁄₁₆ Increment exponent

left 0 • 1 1 1 1 0 1 0 (¹⁄₂ + ¹⁄₄ + ¹⁄₈ + ¹⁄₁₆) x 22 = 3.7510

Table 5.4.4.4 Stages of conversion of +3.7510 to 8-bit floating point form
Table 5.4.4.5 shows the algorithm applied to -3.7510 to produce 1.00012 0102
(mantissa exponent) in 8-bit floating point form with 3 bits for the exponent.
The tables starts with the fixed point form 100.01002.

Mantissa Exponent Decimal value

Weighting -4 2 1 • ¹⁄₂ ¹⁄₄ ¹⁄₈ ¹⁄₁₆ -4 2 1

Number 1 0 0 • 0 1 0 0 0 0 0 (-4 + ¹⁄₄ ) = -3.7510

-2 1 • ¹⁄₂ ¹⁄₄ ¹⁄₈ ¹⁄₁₆ Increment exponent

1 0 • 0 0 1 0 0 0 1 (-2 + ¹⁄₈ ) x 21 = -3.7510

Move point -1 • ¹⁄₂ ¹⁄₄ ¹⁄₈ ¹⁄₁₆ Increment exponent

left 1 • 0 0 0 1 0 1 0 (-1 + ¹⁄₁₆) x 22 = -3.7510

Table 5.4.4.5 Stages of conversion of -3.7510 to 8-bit floating point form

65 Single licence - Abingdon School

 5.4.4 Numbers with a fractional part

Questions
9 Convert the following decimal numbers to floating point binary
storing the mantissa in 5 bits in two’s complement form and the
exponent in 3 bits, also in two’s complement form. The binary
point should be between the most significant bit and the next most
significant bit of the mantissa.

(a) 1.25 (b) -5.5 (c) 7.5 (d) +1 (e) -1

(f ) 7.25 (remember mantissa is stored in 5 bits not 6)

Converting from binary floating point to decimal

Key principle
Floating point representation in any number base takes the form
Binary floating point to
M x BaseE
decimal:
where M is the mantissa, E is the exponent and Base is the number base, for M x BaseE
example, it is 2 in decimal for binary floating point numbers.
1. Convert M to decimal → Md
Method 1: 2. Convert E to decimal → Ed
3. Calculate Md x 2Ed
If we know the values of the mantissa and exponent we can use the formula
above to convert from binary floating point to decimal as follows Alternatively, convert the
1. Convert M to decimal → Md exponent to decimal and
move binary point of mantissa
2. Convert E to decimal → Ed right if exponent positive, left
3. Calculate Md x 2Ed otherwise.

For example,
M = 0•1102 (two’s complement)
E = 01012 (two’s complement)

-1 ½ ¼ ⅛
Md = + ( ¹⁄₂ + ¹⁄₄) = ³⁄₄ = 0.7510
0 1 1 0

-8 4 2 1
Ed = 01012 = +510 0 1 0 1

Md x 2Ed = ³⁄₄ x 25 = ³⁄₄ x 32 = 2410

Therefore,
Mantissa Exponent
0•110 0101 → 2410

Single licence - Abingdon School 66

5 Fundamentals of data representation

Questions
10 Using method 1 convert the following floating point binary
numbers which store the mantissa in 4 bits in two’s complement
form and the exponent in 4 bits, two’s complement form, into
decimal. The binary point is between the most significant bit and
the next most significant bit of the mantissa.
(a) 0•101 0111 (b) 1•000 0110 (c) 0•100 1000
(d) 0•111 1011 (e) 1•001 1111

Method 2:
Alternatively, convert the exponent to decimal and move the binary point of
the mantissa right if exponent positive, left otherwise.
For example,
M = 0•1102 (two’s complement)
E = 01012 (two’s complement)

Ed = 01012 = +510

Shift binary point 510 places to the right, but first add trailing zeroes
0•1102 → 0•1100002 -32 16 8 4 2 1 ¹⁄₂
→ 011000•02 0 1 1 0 0 0 0

The result of the conversion to decimal is 2410

Another example but this time with a negative exponent,

M = 0•1102 E = 11012
Ed = 11012 = -310
Shift binary point 3 places to the left, but first add leading zeroes
0•1102 → 0000•1102 -1 ¹⁄₂ ¹⁄₄ ¹⁄₈ ¹⁄₁₆ ¹⁄₃₂ ¹⁄₆₄
→ 0•0001102 0 0 0 0 1 1 0

The result of the conversion to decimal is ³⁄₃₂ = 0.0937510

67 Single licence - Abingdon School

 5.4.4 Numbers with a fractional part

Example with a negative mantissa and a negative exponent,

M = 1•0102 E = 11002
Ed = 11002 = -410
Shift binary point 4 places to the left, but first add leading 1s,
(note: 1•0102 is equivalent to 11111•0102, to check just convert each to
decimal remembering that the most significant bit is negative in each
case).
1•0102 → 11111•0102 -1 ¹⁄₂ ¹⁄₄ ¹⁄₈ ¹⁄₁₆ ¹⁄₃₂ ¹⁄₆₄ ¹⁄₁₂₈
→ 1•11110102 1 1 1 1 1 0 1 0

The result of the conversion to decimal is - ³⁄₆₄ = -0.04687510

Example with negative mantissa and positive exponent,
M = 1•0102 E = 01002
Ed = 01002 = 410
Shift binary point 4 places to the right, but first add trailing zeroes
1•0102 → 1•010002 -16 8 4 2 1 ¹⁄₂
→ 10100•02 1 0 1 0 0 0

The result of the conversion to decimal is - 1210

Questions
11 Using method 2, convert the following floating point binary
numbers which store the mantissa in 4 bits in two’s complement
form and the exponent in 4 bits, two’s complement form, into
decimal. The binary point is between the most significant bit and
the next most significant bit of the mantissa.
(a) 0•101 0111 (b) 1•000 0110 (c) 0•100 1000
(d) 0•111 1011 (e) 1•001 1111

In this chapter you have covered:

■■ Representing numbers with a fractional part in
• Fixed point form
• Floating point form
■■ Decimal to binary fixed point
■■ Binary to decimal fixed point
■■ Decimal to binary floating point
■■ Binary to decimal floating point

Single licence - Abingdon School 68

 5 Fundamentals of data representation
5.4 Binary number system
Learning objectives
■■For both fixed point and
floating point representation of
■■ 5.4.5 Rounding errors
real numbers, know and The units of the decimal and binary numeral systems
explain why when the decimal Remember that a numeral system is a writing system for expressing numbers.
form is converted to binary the In the decimal numeral system, we work with the decimal digits and use these
result may be inaccurate. to express a number as multiples of units such as 1000, 100, 10, 1, ¹⁄₁₀, ¹⁄₁₀₀,
etc. For example,
(6 x 10) + (0 x 1) + (7 x ¹⁄₁₀) + (1 x ¹⁄₁₀₀) = 60.71

Questions
1 Write down the following decimal numerals as sums of the
decimal units 1000, 100, 10, ….., shown above

(a) 302.034 (b) 5120.2007 (c) 0.4567

In binary, we are restricted to the numerals 0 and 1.

Also, when we convert from decimal to binary, we need to break the decimal
numeral into powers of 2 such as 32, 16, 8, 4, 2, 1, ¹⁄₂, ¹⁄₄ , ¹⁄₈ , ¹⁄₁₆ .
For example, 5.2510 is broken into
(1 x 4) + (0 x 2) + (1 x 1) + (0 x ¹⁄₂) + (1 x ¹⁄₄) = 101.012

Questions
2 Write down the following binary numerals as sums of the
decimal units 32, 16, 8, 4, …., shown above
(a) 1100.11 (b) 101.0101 (c) 11.1011

3 Write down the binary numerals from Q2 in the decimal form

x 21 21
⁄ n , e.g. 101.012 = ⁄ 2 (because ⁄ = 5.25)
2 2 4

(a) 1100.11 (b) 101.0101 (c) 11.1011

Single licence - Abingdon School 69

5 Fundamentals of data representation

Information Representation problem

We don’t have a representational problem in binary with the whole
Factoring calculator:
number part of a decimal providing we have enough bits but we do have a
http://www.solvemymath.com/
representational problem with the fractional part. We break a decimal into
online_math_calculator/general_
x
math/factoring_calculator/ ⁄ n when we convert it exactly to its binary equivalent, e.g.
2
factoring.php 5.2510 = (1 x 4) + (0 x 2) + (1 x 1) + (0 x ¹⁄₂) + (1 x ¹⁄₄)

= (1 x ¹⁶⁄₄) + (0 x ⁸⁄₄) + (1 x ⁴⁄₄) + (0 x ²⁄₄) + (1 x ¹⁄₄) = ²¹⁄₄ = ²¹⁄₂₂

Notice in binary the only possible prime factor for the denominator is 2.
(2 x 2)
However, the denominator of a decimal such as 0.8 = ⁄ doesn’t consist
5
of multiples of 2. There are in fact many decimals that cannot be broken down
x
into the form ⁄ n .
2

Questions
4 Write down the factors of the following fixed point decimal
Key fact numbers in Factors⁄Factors format, for some you may want to use a
factoring calculator
In binary, the only possible prime
factor for the denominator is 2. (a) 0.1 (b) 0.7 (c) 5.7 (d) 8.75 (e) 67.03125
x
We break a decimal into ⁄ n
2
when we convert it exactly to its
5 For each of the fixed point decimal numbers in Q4, state whether
binary equivalent. x
There are in fact many decimals
it can be represented in the form ⁄ n.
2
that cannot be broken down into
x 6 Using a decimal to binary converter, write down the fixed point
the form ⁄ n .
2
binary equivalent of the decimal numerals in Q4. Comment on
your answers.

Information
Fixed point Fixed point Table 5.4.5.1 shows the fixed point binary
Decimal to binary converter
and binary to decimal decimal binary equivalent of some unsigned fixed point
converter: 0.1 0.00011 decimals.
http://www.exploringbinary.com/ 0.2 0. 0011 Notice that some binary equivalents
binary-converter/
0.25 0.01 contain a repeating sequence of 1s and 0s
0.3 0.01001 indicated with a line.
0.4 0.0110 Table 5.4.5.2 shows the same decimals
0.5 0.1 expressed in rational form x/y with the
0.6 0.1001 numerator and denominator factored. It
0.7 0.10110 also shows the closest binary fixed point
0.75 0.11 numeral in 8 bits with binary point
0.8 0.1100 between the most significant digit and the
0.9 0.11100 next.
1.0 1.0
Table 5.4.5.1 Fixed point decimals and their fixed point binary equivalent

70 Single licence - Abingdon School

 5.4.5 Rounding errors

Unsigned x Closest
Closest ⁄ n Closest
Fixed 2 8-bit binary
x/y form for 8-bit fixed point
point numeral in
0≤n≤7 binary form
decimal decimal form
0.1 1⁄ 13⁄ 0.0001101 0.1015625
(2 x 5 ) 27

0.2 1⁄ 13⁄ 0.0011010 0.203125

5 26

0.25 1⁄ 1⁄ 0.0100000 0.25

(2 x 2) 22

0.3 3⁄ 19⁄ 0.0100110 0.296875

(2 x 5 ) 26

0.4 2⁄ 51⁄ 0.0110011 0.3984375

5 27

0.5 1⁄ 1⁄ 0.1000000 0.5

2 21

0.6 3⁄ 77⁄ 0.1001101 0.6015625

5 27

0.7 7⁄ 45⁄ 0.1011010 0.703125

(2 x 5 ) 26

0.75 3⁄ 3⁄ 0.1100000 0.75

(2 x 2) 22

0.8 (2 x 2)⁄ 51⁄ 0.1100110 0.796875

5 26

0.9 9⁄ 115⁄ 0.1110011 0.8984375

(2 x 5 ) 27

1.0 2⁄ 1⁄ 1.0000000 1.0

2 20
Table 5.4.5.2 Decimals expressed in rational form x/y

Missing rational numbers Key fact

Figure 5.4.5.1 shows the weighting for the 8-bit unsigned fixed point binary
that was used for Table 5.4.5.2. Inaccuracies arise when
representing a given fixed point
¹⁄₁ ¹⁄₂ ¹⁄₄ ¹⁄₈ ¹⁄₁₆ ¹⁄₃₂ ¹⁄₆₄ ¹⁄₁₂₈ decimal in fixed point binary using
0 0 0 0 0 0 0 0 a specified number of bits because
it is not possible to represent all
Figure 5.4.5.1 Weighting for 8-bit fixed point binary
the possible rational numbers
within the range.
If we construct all the possible fixed point binary numerals from these
Rounding errors occur when this
weightings we end up with a finite subset of rational numbers, S, as follows happens.
S={ 255
⁄
128
,128
254
⁄ , 253
⁄
128
, .…, 239
⁄
128
, …, 47
⁄
128
46
, ⁄128, .…, 128
⁄
128
,
The same is true of the mantissa
31 30
.…, ⁄128, ⁄128, …., 8
⁄
128
, 7⁄128, …., 2
⁄
128
}
, 1⁄128, 0⁄128 of floating point binary.
Rational numbers are missing from this subset, S.
Take any two consecutive members of the set, e.g. 255⁄128 = 510⁄256 and
254
⁄ = 508⁄256. Notice that, for this example, 509⁄256 is missing from the
128
set S. Lots more rational numbers can be found that are missing from the set S.

Single licence - Abingdon School 71

5 Fundamentals of data representation

This leads to inaccuracies when we represent a given fixed point decimal in

fixed point binary using the specified number of bits, e.g. we cannot represent
509
⁄
256
for example using 8-bit fixed point binary.

Questions
7 For each of the following values of x, write down the nearest
x
8-bit fixed point binary numeral to ⁄128 using Figure 5.4.5.1
weighting

(a) 0 (b) 1 (c) 2 (d) 3 (e) 5 (f ) 254

8 For each of the following values of x, write down the nearest

x
8-bit fixed point binary numeral to ⁄256 using Figure 5.4.5.1
weighting.

(a) 510 (b) 508 (c) 509

9 For each part of Q8 write down the decimal expansion of x⁄256

(e.g. 1.99…) and the decimal equivalent of your answers to Q8.
Comment on your results.

Rounding error Closest 8-bit Difference

Fixed point
Table 5.4.5.3 shows for each binary in or
decimal
given unsigned fixed point decimal Error
decimal, the difference or 0.1 0.1015625 0.0015625
error between it and the 0.2 0.203125 0.003125
closest decimal that can 0.25 0.25 0
be represented in 8-bit 0.3 0.296875 0.003125
unsigned binary fixed point 0.4 0.3984375 0.0015625
form. We call this error the 0.5 0.5 0
rounding error.
0.6 0.6015625 0.0015625
0.7 0.703125 0.003125
0.75 0.75 0
0.8 0.796875 0.003125
0.9 0.8984375 0.0015625
1.0 1.0 0
Table 5.4.5.3 Fixed point decimals and
the error difference

72 Single licence - Abingdon School

 5.4.5 Rounding errors

Truncation or rounding down

Unsigned fixed point decimal 0.110 converts to fixed point binary 0.00011,
0.310 to 0.01001 and so on as shown in Table 5.4.5.4.

When these fixed point binary numerals with recurring bit patterns are limited
to a given numbers of bits, e.g. 9, we can just drop the bits in bit positions
greater than this given number. This is known as truncation or rounding down.
For example, 0.110 truncated to 9 bits becomes 0.000110012 just by dropping
10011 from the 10th bit position onwards.
Rounding off
Alternatively, we can choose to round off. This means
• Add 1 to the last retained digit if the following digit is 1 otherwise leave
unaltered.
For example, if we round off 0.00011 to 8 bits then we need to look at the
binary expansion as far as the 9th bit which is 0.00011001.
The 9th bit is 1 so we drop the 9th bit but add 1 to the 8th bit so arriving at
0.00011012, the rounded off result in 8 bits.

On the other hand, if we round off 0.01001 to 8 bits then we don’t add 1
because the binary expansion to 9 bits is 0.010011002 and the 9th bit is 0.
Therefore, 0.01001 rounded off to 8 bits is 0.01001102.
Rounding off is usually used where the representation is inexact, because less
error can result compared with rounding down.

Fixed Fixed point

Fixed point Closest binary
point binary truncated
binary form in 8 bits
decimal after 9-bits
0.1 0.00011 0.00011001 0.0001101
0.2 0. 0011 0.00110011 0.0011010
0.25 0.01 0.01000000 0.0100000
0.3 0.01001 0.01001100 0.0100110
0.4 0.0110 0.01100110 0.0110011
0.5 0.1 0.10000000 0.1000000
0.6 0.1001 0.100110011 0.1001101
0.7 0.10110 0.10110011 0.1011010
0.75 0.11 0.11000000 0.1100000
0.8 0.1100 0.11001100 0.1100110
0.9 0.11100 0.11100110 0.1110011
1.0 1.0 1.00000000 1.0000000

Table 5.4.5.4 Fixed point decimals and their equivalents

Single licence - Abingdon School 73

5 Fundamentals of data representation

Questions
10 For each decimal in Table 5.4.5.4, write down the nearest
unsigned fixed point binary numeral in 10 bits, rounding off. Also
write down the rounding error if any.

Rounding errors in signed fixed point

Everything that has been stated for unsigned fixed point binary is also true of
signed fixed point binary.

Rounding errors in floating point

Everything that has been stated for fixed point binary is also true of the
mantissa of floating point representation because the mantissa takes on fixed
point form.

In this chapter you have covered:

■■ Why both fixed point and floating point representation of decimal
numbers may be inaccurate. For both fixed point and floating point
representation of real numbers, there are in fact many decimals that
x
cannot be broken down into the form ⁄ n . This leads to inaccuracies
2
when we represent a given fixed point decimal in fixed point binary using
a specified number of bits or in the mantissa of a floating point binary.

74 Single licence - Abingdon School

5 Fundamentals of data representation
5.4 Binary number system
Learning objectives
■■Be able to calculate for stored
and processed numerical data
■■ 5.4.6 Absolute and relative errors
Approximating a number
• absolute error
In the previous chapter we learned that when storing a number in a
• relative error computer, if the number contains more digits than can be accommodated,
an approximation to the number is stored (obtained by either rounding off
■■Compare absolute and
or truncating). When using truncated results, the machine representation is
relative errors for large and
constructed by simply discarding significant digits that cannot be stored; when
small magnitude numbers,
rounding off it approximates a quantity with the closest machine representation
and numbers close to one.
possible.
11

01

10

00

01
00

11
10

11

11

00

00
11

11
00

01
00

00

00

00

01
0.0

0.0

0.0
0.0

0.0
0.0

0.0

0.0001100110011....
Figure 5.4.6.1 Binary number line showing 8-bit unsigned
fixed point binary
For example, if only 8 bits are available as shown in Figure 5.4.6.1 then
0.110 which in unsigned fixed point binary is 0.00011 will be represented by
0.00011002 if truncated and 0.00011012 if rounded off because it lies between
these two values as shown.
Absolute error
Key principle
The difference between the actual number and the nearest representable value is
Absolute error: known as the absolute error. For example, 0.110 is stored as 0.00011012 in 8-bit
The difference between the unsigned fixed point binary form which is 0.101562510. Therefore,
actual number and the nearest
representable value. Absolute error = 0.101562510 - 0.110 = 0.001562510
Absolute means that the sign is ignored e.g. differences of 0.0015625 and
-0.0015625 are the same absolute error.

Questions

1 Calculate the absolute error when the following fixed point

decimal numbers are stored in 8-bit unsigned fixed point binary
as shown in Figure 5.4.6.1. Round off if a number cannot be
represented exactly.

(a) 0.210 (b) 0.610

Single licence - Abingdon School 75

5 Fundamentals of data representation

Key principle Relative error

The relative error is defined as the absolute error divided by the actual number.
Relative error: For example, the absolute error when 0.110 is stored as 0.00011012 in 8-bit
The absolute error divided by
unsigned fixed point binary form is 0.001562510. Therefore,
the actual number.
Relative error = 0.001562510 ⁄
0.110
= 0.015625

This is 1.5625% when represented as a percentage.

Questions
2 Calculate the percentage relative error when the following fixed point decimal numbers are stored in 8-bit
unsigned fixed point binary as shown in Figure 5.4.6.1. Round off if a number cannot be represented
exactly.

(a) 0.210 (b) 0.610

Comparing absolute and relative errors

Absolute error calculations are not as useful as relative error calculations.
For example, Table 5.4.6.1 shows how the relative error can vary for a given
absolute error if the magnitude of the actual value varies from small to large.

Absolute error Actual value in Relative error

in decimal decimal %
1.0 1000000.0 0.0001
1.0 10.0 10
1.0 1.0 100
16 128 12.5
1⁄ 1/ 12.5
2048 256
Table 5.4.6.1 Variation of relative error for a given absolute error

Questions
3 The percentage relative error is 1% for the following decimal numbers. Calculate the absolute error.

(a) 1.00 (b) 1.00 x 1038 (c) 1.00 x 10-39

In this chapter you have covered:

■■ How to calculate for stored and processed numerical data
• absolute error: the difference between the actual number and the nearest
representable value.
• relative error: the absolute error divided by the actual number.
■■ Comparing absolute and relative errors for large and small magnitude
numbers, and numbers close to one - see Table 5.4.6.1.

76 Single licence - Abingdon School

5 Fundamentals of data representation
5.4 Binary number system
Learning objectives
■■Be able to compare the
advantages and disadvantages
■■ 5.4.7 Range and precision
of fixed point and floating Using fixed point representation
point forms in terms of In the previous chapter, we learned that it becomes necessary to approximate a
number sometimes if a representation of it is to be stored in a digital computer.
• range
A digital computer is not designed to allocate an infinite number of bits instead
• precision it must allocate a fixed number of bits because computer memory is finite.

• speed of calculation In a simplified example of fixed point, we allocate 8 bits and use these as shown
in Figure 5.4.7.1.

128 128
values values

00

00

00

11
00
00

00

00

00

00

00

11
00
000
00

00

00

00

10

10

11
00
00

10

00

10

0.0

0.0

0.1

0.1
0.1
1.0

1.0

1.1

1.1

-128 -96 -64 -32 0 32 64 96 127

128 128 128 128 128 128 128 128 128

Figure 5.4.7.1 Number line for 8-bit signed two’s complement

fixed point binary.
For 8 bits there are 28 or 256 different arrangements of the bits or bit
Tasks patterns. The example uses these to code numbers in fixed point signed two’s
1 Write down all the bit
complement binary arranged in decimal from ⁻¹²⁸⁄₁₂₈ to ⁺¹²⁷⁄₁₂₈ or
patterns consisting of the binary 1.0000000 to 0.1111111 with the most significant bit as the sign bit, and
following number of bits weighted ⁻¹²⁸⁄₁₂₈ as shown in Figure 5.4.7.2 .
(a) 1 (b) 2
(c) 3 (d) 4 ⁻¹²⁸⁄₁₂₈ ⁶⁴⁄₁₂₈ ³²⁄₁₂₈ ¹⁶⁄₁₂₈ ⁸⁄₁₂₈ ⁴⁄₁₂₈ ²⁄₁₂₈ ¹⁄₁₂₈
0 0 0 0 0 0 0 0
2 For each answer in Task 1 Figure 5.4.7.2 Weightings in the example for 8-bit signed two’s complement
count the number of bit
fixed point binary
patterns.
Do these counts follow the The (128 + 128) or 256 different representations referenced in Figure 5.4.7.1
formula 2n where n is the are distributed evenly across the range. Any number that needs to be stored is
number of bits?
coded by assigning to it the nearest representation, e.g. 0.210 will be stored as
the bit pattern 0.00110102 or in decimal, ²⁶⁄₁₂₈, an approximation to 0.210.
The smallest difference of any two 8-bit patterns in the example coding is
¹⁄₁₂₈ or ¹⁄₂₈₋₁ in decimal. In general, for n bits, the smallest difference for
fixed point binary with the given weighting is 1⁄2n - 1 .

Single licence - Abingdon School 77

5 Fundamentals of data representation

If we think of the number line shown in Figure 5.4.7.1 as a ruler then the
precision with which we can record measurements with this ruler is to the
nearest ¹⁄₁₂₈ .
Thus, for any positive or negative number inside the range that can be
represented, the maximum absolute error in a measurement in this coding
scheme will be ¹⁄₂₅₆ (one half of ¹⁄₁₂₈ because we round off). The
maximum percentage relative error will be ¹⁄₂₅₆ ⁄ x 100 = 0.39%
¹²⁷⁄₁₂₈
and the largest ¹⁄₂₅₆⁄¹⁄ x 100 = 50%.
₁₂₈

Questions
1 With the binary point placed between the sign bit and
the next bit what is the smallest positive number that can be
represented in two’s complement fixed point binary for the
following number of bits ?

(a) 4 (b) 16 (c) 24 (d) 32 (e) 64

Using floating point representation

The range of numbers that can be stored in 8-bit signed two’s complement fixed
point form as shown above, ⁻¹²⁸⁄₁₂₈ to ⁺¹²⁷⁄₁₂₈ , is very limited. If we
divide these 8 bits into a 4-bit signed fraction, f, and a 4-bit signed multiplier ,
2e, then we can represent numbers differently in 8-bits as follows

f x 2e

We are still restricted with 8 bits to choosing from 28 or 256 different bit
patterns but we achieve a much greater range of representable numbers than
with fixed point coding. If we use the weightings shown in Figure 5.4.7.3 then
the range of representable numbers is
⁻⁸⁄₈ x 27 to ⁺⁷⁄₈ x 27 (i.e. -128 to 127)
or
⁻¹²⁸⁄₁₂₈ x 27 to ⁺¹¹²⁄₁₂₈ x 27

This range is larger than the 8-bit fixed point coding by a factor 27.

Mantissa Exponent
⁻⁸⁄₈ ⁴⁄₈ ²⁄₈ ¹⁄₈ -8 4 2 1
0 0 0 0 0 0 0 0
Figure 5.4.7.3 Weightings for 8-bit floating point, two’s complement 4-bit
mantissa and 4-bit exponent
Figure 5.4.7.4 shows the range of expressible negative and positive numbers
for a normalised 4-bit mantissa and a 4-bit exponent weighted as described
above in Figure 5.4.7.3. The negative value closest to zero is 1011 10002 in
4-bit mantissa 4-bit exponent form or -0.625 x 2-8 in decimal. The mantissa has

78 Single licence - Abingdon School

 5.4.7 Range and precision

been normalised for maximum precision as will be explained in Chapter 5.4.8.

The positive value closest to zero is 0.100 1000 in 4-bit mantisa 4-bit exponent
form or +0.5 x 2-8 in decimal. The mantissa has been normalised for maximum
precision.
Zero

-1.0 x 2 7 -0.625 x 2 -8 +0.5 x 2 -8 0.875 x 2 7

5 0 1
-128 - + +112
2048 512
Figure 5.4.7.4 The range of expressible negative and positive numbers for a normalised
4-bit mantissa and a 4-bit exponent weighted as shown in Figure 5.4.7.3

Essentially, the mantissa f is restricted to the following ranges to ensure

maximum precision
½ ≤ f < 1 and -1 ≤ f < -½
or in two’s complement binary
0.1002 ≤ f ≤ 0.1112 and 1.0002 ≤ f < 1.1002
or
0.1002 ≤ f ≤ 0.1112 and 1.0002 ≤ f ≤ 1.0112

0.1112 is the nearest positive mantissa to 12 or 110.

The nearest two’s complement binary representation less than 1.1002 is

1.1002 - 0.0012 = 1.0112 = -0.62510.

We consider zero as a special case. For zero, f is 0 and exponent e is also 0. This
means that for our 8-bit example we will use only 128 bit patterns excluding
zero, but 129 bit patterns if zero is included. For zero to be excluded then the
bit before the binary point must always be different from the bit after this
point.
The 129 different representations are distributed unevenly across the range as
illustrated in a section of the number line in Figure 5.4.7.5.

Difference is 1 Difference is 1 Difference is 1

8 4 2
0.100 x 20

0.100 x 21

0.100 x 22

0.100 x 23
0.101 x 20
0.110 x 20
0.111 x 20

0.101 x 21

0.110 x 21

0.111 x 21

0.101 x 22

0.110 x 22

0.111 x 22

Figure 5.4.7.5 Distribution of floating point representations

Thus for largest positive numbers, the difference between the representations
is as much as ⅛ x 27 or 16, so a somewhat large absolute error of half of this
because we round off when approximating. The approximate percentage

Single licence - Abingdon School 79

5 Fundamentals of data representation

relative error is ⁸⁄₂₇ x 100 = ⁸⁄₁₂₈ x 100 = 6.25% for the largest positive
number.
For the smallest positive numbers, the difference between the representations
is as little as ¹⁄₈ x 2-8 or ¹⁄₂₀₄₈, a small absolute error of half this. The
approximate percentage relative error is ¹⁄₄₀₉₆ ⁄ -8 x 100 = 6.25% for the
2
smallest positive number. The relative error is similar across the whole range.

Questions
2 Evaluate the following where the mantissa is in two’s complement form
(a) 0.1112 x 22 (b) 0.1012 x 26 (c) 1.0112 x 22 (d) 0.1012 x 2-6

3 Explain why the gaps between adjacent numbers expressed in floating point form are not constant.

Comparing fixed point and floating point ranges

Floating point representation can store numbers chosen from a much greater
range, larger and smaller, positive and negative, than fixed point, for a given
number of bits n for each.

RepresentationFor example, Table 5.4.7.1 shows how fixed point Rangerepresentation compares
The fixed point range in decimal forwith floating
n bits wherepoint
n for 32 bits.
= 32, and using two’s complement representation
+8.4 x 106 to +3.9 x 10-3, 0, -3.9 x 10-3 to -8.4 x 106
with 8 of the 32 bits allocated to the fractional part is
approximately
The floating point range in decimal, for the same
number of bits 32, for an 8-bit two’s complement
+ 1.7 x 1038 to + 1.5 x 10-39, 0, -1.5 x 10-39 to -1.7 x 1038
exponent and 24-bit normalised mantissa, is
approximately
Table 5.4.7.1 Comparison of ranges of fixed point and floating point representations
where the number of bits is 32 in each case

Questions
4 What are the most positive and most negative numbers that can be represented using the following fixed
point two’s complement representations? Give your answers in decimal.

(a) 4 bits of which 2 bits are allocated to the fractional part

(b) 8 bits of which 4 bits are allocated to the fractional part
(c) 10 bits of which 5 bits are allocated to the fractional part

5 What are the most positive and most negative numbers that can be represented using the following
floating point two’s complement representations? Give your answers in decimal.

(a) 16 bits of which 8 bits are allocated to mantissa and 8 bits to the exponent
(b) 32 bits of which 16 bits are allocated to mantissa and 16 bits to the exponent
(c) 64 bits of which 32 bits are allocated to mantissa and 32 bits to the exponent

80 Single licence - Abingdon School

 5.4.7 Range and precision

Precision and significant figures or digits

To compare the precision of fixed point with floating point we need first of all to understand what is meant by
precision. To understand precision we need first to understand what is meant by significant figures or digits.
Significant figures or digits
Rulers, tape measures and other measuring devices enable length measurements to be made. Here are some
measurements and their units
11.1 cm 120 cm 12.23 km 12.2 km

The first 11.1 cm is also 111 mm. This indicates that this measurement has been made to the nearest millimetre. The
second, 120 cm is ambiguous. It is not clear whether the measurement is exactly 120 cm or the measurer was just
measuring to the nearest 10 cm.
The third, 12.23 km in metres is 12230 m. The measurement was performed to the nearest 10 m, i.e. the real value
lies somewhere between 12225 and 12235. The last measurement 12.2 km in metres is 12200 m, a measurement
performed to the nearest 100 m, i.e. the real value lies somewhere between 12150 and 12250.
The latter two measurements, 12.23 km and 12.2 km, clearly show a different degree of precision, one is a
measurement to the nearest 10 m and the other to the nearest 100 m. We call the digits which provide information
about the precision of a measurement, significant digits or significant figures and the more significant digits
used the greater the precision of the measurement.
Table 5.4.7.2 shows the number of significant digits for some measurements.

Measurement Precision Explanation

111 11.1 cm is 111 mm
11.1 cm
3 significant figures measurement has been made to the nearest millimetre
1223 12230 m
12.23 km
4 significant figures measurement has been made to the nearest 10 metres
122 12200 m
12.2 km
3 significant figures measurement has been made to the nearest 100 metres
120 Decimal point indicates measurement made to nearest centimetre,
120. cm
3 significant figures the 0 is not just a placeholder
By changing the units to centimetres it becomes 3000 cm which
3000
0.03000 km suggests measurement to nearest centimetre and therefore the
4 significant figures
three zeroes after 3 are significant, the zeroes before are not
90 Measurement made to nearest millimetre (90 millimetres) therefore
9.0 cm
2 significant figures the zero is significant
2001 2.001 m is 2001 mm
2.001 m
4 significant figures Measurement made to nearest millimetre

Table 5.4.7.2 Number of significant digits for various measurements

Single licence - Abingdon School 81

5 Fundamentals of data representation

Key concept Rules of significant figures or digits

1. Any non-zero digits and zeroes between non-zero digits are significant, e.g.
Significant digits/figures: in 2.001m the two zeros are significant
Digits which provide
information about the precision 2. Leading zeroes, i.e. zeroes that come before the non-zero digits are not
of a measurement, are called significant, e.g. in 0.03000 km the first two zeros are not significant
significant digits or significant
figures.
3. Trailing zeroes after the last non-zero digit are significant if a decimal
point occurs anywhere in the number, e.g. 0.03000 km, the last three zeros
are significant
4. If there is no decimal point anywhere in the number then its precision is
ambiguous.

Questions
6 State the number of significant figures for each of the following
measurements made in the decimal system
(a) 12.23 (b) 130.04 (c) 0.03 (d) 0.00450 (e) 34 (f ) 0254.

Significant digits in floating point representation

0111
The number of significant digits for the measurement 10102 is ambiguous, it
1000 is not clear whether the measurement was exactly 10102, i.e. 1010.02 or the
measurement was made to the nearest unit, i.e. the actual value was between
1001
1001.12 and 1010.12.
1001.12
1010
If the measurement was between 1001.12 and 1010.12, the ambiguity in 10102
1010.12
could be removed if we expressed the measurement in the form showing that
1011
we have just 3 significant digits of precision
1100 1.012 x 23
Multiplying by 23 is equivalent to shifting the binary point three places to the
1101
right as follows
x 2 x 2 x2
1110 1.01 → 10.1 → 101. → 1010

1111
The number of significant digits is unambiguous, it is 3, i.e. 101. This
representation is known as scientific form. Floating point representation
Figure 5.4.7.6 Section of a resembles scientific form in structure but differs slightly on detail. The
binary ruler significant digits are contained in the mantissa. The exponent records how
many places to shift the binary point, left or right to obtain the fixed point
form of the number.
Questions
7 State the number of significant figures, if you can, for each of the following measurements made in fixed
point binary. If you cannot, explain why.

(a) 10.12 (b) 0.1112 (c) 0.0112 (d) 0.1012 (e) 0.11001002 (f ) 10102

82 Single licence - Abingdon School

 5.4.7 Range and precision

Precision Key concept

The more significant digits recorded for a measurement the greater its precision.
Precision is indicated by the number of significant digits. Precision:
The number of significant digits
used to represent the number.
Comparison of the advantages and disadvantages of fixed point The more significant digits used
and floating point the greater the precision.

Precision versus range

We focus on the mantissa of the floating point representation when comparing
the precision of floating point with fixed point as the size of the mantissa in bits
determines the number of significant digits that can be stored.
If fixed point representation has the same number of bits as the mantissa in a Key fact
floating point representation then both forms will (assuming the floating point Precision of floating point vs
representation is normalised – see Chapter 5.4.8) have the same precision. fixed point:
For a given number of bits, fixed
However, if the total number of bits is the same for each then the number
point representation can store
of bits in the floating point’s mantissa will be less and so will possess fewer more significant digits than
significant digits than fixed point. Therefore, a number stored in floating floating point.
point form will be stored with less precision than it will in fixed point form Therefore, a number stored in

if the same total number of bits are used for each form. floating point form will be stored
with less precision than fixed
However, floating point, for a given number of bits, can store numbers point representation which uses
chosen from a greater range than fixed point, but only at the expense of the same total number of bits.
precision.
For example, Figure 5.4.7.7 shows a floating point number representation in
32 bits using a 24-bit mantissa and an 8-bit exponent. Key fact
The mantissa has storage space for 23 significant digits + 1 sign bit. Range of floating point vs
fixed point:
However, the same 32 bits could store 31 significant digits + 1 sign bit in fixed
For a given number of bits,
point form. floating point represents a much

Mantissa Exponent greater range of numbers than

fixed point.
Sign Sign
bit bit
-1 -128
1 0 ... 1 0 0 1 ... 1 0
Figure 5.4.7.7 32-bit floating point number, two’s complement
24-bit mantissa and 8-bit exponent

Questions
8 How many significant digits can two’s complement floating
point store if the representation is as follows

(a) 16 bits in total of which 6 bits are allocated to the exponent?

(b) 32 bits in total of which 6 bits are allocated to the exponent?
(c) 64 bits in total of which 8 bits are allocated to the exponent?

Single licence - Abingdon School 83

5 Fundamentals of data representation

Questions
9 How many significant digits can two’s complement fixed point store
in the following number of bits

(a) 16 bits?
(b) 24 bits?
(c) 64 bits?

Speed of calculation
In general, calculations take longer with numbers stored in a digital computer
Key fact in floating point form than they do with numbers stored in fixed point form
Speed of calculation of
because floating point inevitably involves floating or shifting the decimal or
floating point vs fixed point: binary point of operands whereas fixed point doesn’t.
In general, calculations take longer
If the central processing unit (CPU) does not have circuitry to perform floating
with numbers stored in floating
point form than they do with
point operations directly then the steps of the operations have to be written in
numbers stored in fixed point software and accessed by the CPU. Fetching and executing code to perform
form. calculations is considerably slower than doing these calculations directly in
hardware designed just for this task.
Although modern general purpose computers contain CPUs that have access to
hardware floating point units, even with a hardware floating point unit floating
point calculations still take longer than fixed point calculations.

In this chapter you have covered:

■■ The advantages and disadvantages of fixed point and floating point forms
in terms of
• Range: For a given number of bits, floating point represents a much
greater range of numbers than fixed point.
• Precision: For a given number of bits, fixed point representation can
store more significant digits than floating point. Therefore, a number
stored in floating point form will be stored with less precision than
fixed point representation which uses the same total number of bits.
• Speed of calculation: In general, calculations take longer with numbers
stored in floating point form than they do with numbers stored in fixed
point form.

84 Single licence - Abingdon School

5 Fundamentals of data representation
5.4 Binary number system
Learning objectives
■■Know why floating point
numbers are normalised
■■ 5.4.8 Normalisation of floating point form
Uniqueness of representation
■■Be able to normalise The decimal number, 323.142 can be represented in floating point form in
un-normalised floating point
many different ways, some of which are as follows
numbers with positive and
32314.2 x 10-2
negative mantissas
323.142 x 100

0.323142 x 103

0. 00323142 x 105

Questions
1 Write down the floating point form of 323.142 for which the
power of ten multiplier is

(a) 107 (b) 109 (c) 10-5

Having more than one representation is not a good idea. For example, adding
323.142 to itself using two different floating point representations is not
straight forward. Try this for yourself
32314.2 x 10-2

+ 0.323142 x 103

-----------------
-----------------
Neither is comparing two numbers straightforward if they use different floating
pointing representations, e.g. 323.142 x 100 and 0.00323142 x 105.
It makes good sense therefore to allow just one floating point representation
of a number to ensure the uniqueness of the representation and to reduce the
effort required to perform arithmetic operations.
Normalisation in decimal
Computer memory is finite and therefore it is necessary to allocate a fixed
number of bits to each representation of a number. Just for the moment,
imagine that the computer is able to store the decimal digits, sign and decimal
point as shown in Figure 5.4.8.1 with the position of the mantissa’s decimal
point fixed to ensure uniqueness of representation. This memory is shown to
store up to six significant decimal digits in the mantissa.

Single licence - Abingdon School 85

5 Fundamentals of data representation

Mantissa Exponent
+ 3 2 3 1 4 2 + 2

Figure 5.4.8.1 Mantissa-exponent store for decimal digits, sign and decimal
point in fixed position as shown to ensure uniqueness of representation

Now imagine that after a floating point number calculation the result obtained
is
+ 0.00516838 x 106

We now face a problem. Although the mantissa of the answer has six significant
digits, and our fixed point memory arrangement can accommodate six
significant digits, if we do not adjust the exponent of the result we will lose
precision as shown in Figure 5.4.8.2 where rounding off has been applied

Mantissa Exponent
+ 0 0 0 5 1 7 + 6

Figure 5.4.8.2 Loss of precision

On the other hand, we can preserve precision by adjusting the exponent as

follows
Mantissa Exponent
+ 5 1 6 8 3 8 + 3

Figure 5.4.8.3 Regaining maximum precision

This form of the result is said to be normalised

The goal for normalising floating point representations of numbers is to
maximise precision by maximising the number of significant digits in the
representation.

Questions
2 Using the form of representation shown below, in which six
significant digits are allowed in the mantissa and one in the
exponent, normalise the following floating point decimal numbers
Mantissa Exponent
sign d d d d d d sign d

(a) 0.0000456789 x 107 (b) 0.0000456789 x 104

(c) 0.00456789 x 10-5 (d) 0.00004567895 x 106

86 Single licence - Abingdon School

 5.4.8 Normalisation of floating point form

Normalising an un-normalised floating point binary Key point

representation
Numbers expressed in floating point form are normalised to maximise the Normalisation:
1. Maximises precision by
precision with which the number is expressed, i.e. to maximise the number of
maximising the number of
significant digits present in the representation.
significant digits that are
Normalisation also provides a unique representation for a number. represented in mantissa.

0.187510 is 0.00112 in binary fixed point form. In un-normalised floating point This is achieved by ensuring
that bit to the immediate right
form, this is
of point is a significant digit,
0.00112 00002 1 for positive mantissas, 0 for
negative mantissas
where the mantissa is 0.00112 and the exponent is 00002.
2. Ensures that the representation
What if only 4 bits instead of 5 bits were available for the mantissa? Do we just is unique.
round off to 0.0102? No, we normalise the 5 bits first and then round off, if we
have to.
Figure 5.4.8.4 shows in stages how this is done, for a 4-bit mantissa and a Key fact
4-bit exponent for the result, by shifting the bits of the mantissa left whilst
Increment:
decrementing the exponent. Shifting/incrementing stops when the bit before
Means increase by 1, or more if
the binary point is different from the bit after the point. The mantissa bit
size of increment specified.
shown in red is unavailable to use for the result. The normalised representation
Decrement:
is
Means decrease by 1, or more if
0.112 11102 size of decrement specified.

Mantissa Exponent Decimal value

Weighting -1 ¹⁄₂ ¹⁄₄ ¹⁄₈ ¹⁄₁₆ -8 -4 2 1

0 • 0 0 1 1 0 0 0 0 ³⁄₁₆ x 20 = 0.187510
0 • 0 1 1 1 1 1 1 ³⁄₈ x 2-1 = 0.187510
0 • 1 1 1 1 1 0 ³⁄₄ x 2-2 = 0.187510
Figure 5.4.8.4 Normalising a positive number expressed in un-normalised floating point binary form

Key principle
The normalisation process essentially identifies where the significant digits of
the representation begin and then adjusts the mantissa so that digit positions Normalisation process:
Move significant digits so
are not wasted by being taken up by non-significant digits, For example, in the
that bit before binary point in
two’s complement floating point representation 0.00112 00002 , the mantissa mantissa is different from bit
is 0.00112 and has significant digits that begin at the first 12 after the point, after point, adjust exponent
i.e. 112. The 02 before the binary point is the sign bit so needs to be retained. accordingly.
The 002 between the point and 112 is not significant. So moving the point two
places to the right results in a mantissa of 0.112. To compensate we need to take
2 away from the exponent, 00002 → 11102.

Single licence - Abingdon School 87

5 Fundamentals of data representation

Key fact Normalisation algorithm

We start by expressing the decimal number in floating point form with the
Significant bit:
leftmost bit position of the mantissa weighted -110. For example, 10.7510 in
Significant digit and significant bit
two’s complement fixed point binary is shown in Table 5.4.8.1.
are used interchangeably because
they reference the same thing for
Fixed point 2's Normalised
binary. Floating point 2's
Decimal No complement floating
complement binary
binary point form?
(a) 10.7510 1010.112 0.101112 00112 Yes
(b) 0.23437510 0.00000112 0.00000112 00002 No
(c) 0.23437510 0.00000112 0.112 10112 Yes
Table 5.4.8.1 Normalised floating point form?

Normalisation algorithm for positive mantissa

Algorithm:
While bit before point = bit after point
Do
Remove bit after point
then shift all remaining mantissa bits left one place and
insert a zero in least significant bit position
Decrement exponent
EndWhile

For example, the algorithm applied to (b) in Table 5.4.8.1 produces

0.112 10112 when truncated to a 3-bit mantissa, which is (c) in the table.

Questions
Did you know?
3 Using the form of representation shown below for your answer in
Arithmetic Shift Left: which six bits are assigned to the mantissa and four to the exponent,
This operation is supported by normalise the following un-normalised floating point binary
most CPUs as a built-in machine
numbers whose mantissa and exponent are shown below in
operation belonging to the
instruction set of the processor. (a) to (c)
In assembly language it usually is
Mantissa Exponent
given the mnemonic ASL. It shifts
the bits left one position whilst -1 ¹⁄₂ ¹⁄₄ ¹⁄₈ ¹⁄₁₆ ¹⁄₃₂ -8 4 2 1
preserving the value of the sign d d d d d d d d d d
bit.
(a) 0.0000111012 00112 (b) 0.00101012 11002
(c) 0.000000111012 01112

88 Single licence - Abingdon School

 5.4.8 Normalisation of floating point form

Normalisation algorithm for negative mantissa

Express decimal number in floating point form with the leftmost bit position of
the mantissa weighted -110.

Algorithm:
While bit before point = bit after point
Do
Remove bit after point
then shift all remaining mantissa bits left one place and
insert a zero in least significant bit position
Decrement exponent
EndWhile

Figure 5.4.8.5 shows this algorithm applied to 1.11012 00002. Four bits
are allocated to the mantissa and four to the exponent for the result. The
fifth mantissa bit shown in red is not available to use in the result. For each
iteration of the algorithm a 1 in the ¹⁄₂ column is removed and then the lesser
significant bits are shifted left one bit position and a zero inserted in the least
significant bit position. The normalised form shows that we have avoided
rounding off and thereby a loss of precision.

Mantissa Exponent Decimal value

Weighting -1 ¹⁄₂ ¹⁄₄ ¹⁄₈ ¹⁄₁₆ -8 -4 2 1

1 • 1 1 0 1 0 0 0 0 -³⁄₁₆ x 20 = -0.187510
1 • 1 0 1 0 1 1 1 1 -³⁄₈ x 2-1 = -0.187510
1 • 0 1 0 0 1 1 1 0 -³⁄₄ x 2-2 = -0.187510
Figure 5.4.8.5 Normalising a negative number expressed in un-normalised floating point binary form

Also note that the algorithm for normalizing a negative mantissa is exactly the
same as for normalizing a positive mantissa.
The normalisation process for a negative mantissa essentially identifies where
the significant digits of the representation begin and then adjusts the mantissa
so that digit positions are not wasted by being taken up by non-significant
digits. For example, in the two’s complement floating point representation
1.11012 00002, the mantissa is 1.11012 and has significant digits that begin
at the first 02 after the point, i.e. 012 (this is the opposite of the case when the
mantissa is positive). The 12 before the binary point is the sign bit so needs to
be retained. The 112 between the point and 012 is not significant. So moving
the point two places to the right results in a mantissa of 1.012. To compensate
we need to take 2 away from the exponent, 00002 → 11102. Notice that when
the mantissa is negative we look for the occurrence of the first 02 after the
binary point as this is where the significant digits begin.

Single licence - Abingdon School 89

5 Fundamentals of data representation

By a similar argument, if the mantissa had been 111.11012 , the first two 1’s are
insignificant, they are equivalent to leading zeroes in a positive mantissa, e.g.
000.10102, and therefore can be dropped producing 1.11012 which then needs
to undergo normalisation.

Questions
4 Write down the result of eliminating unnecessary digits, if
possible, from the following two’s complement representations

(a) 1111.012 (b) 1101.0112 (c) 00011.012

(d) 101.012 (e) 101.01002

5 Using the form of representation shown below for your answer in

Key fact which six bits are assigned to the mantissa and four to the exponent,
Significant digits for a
normalise the following un-normalised floating point binary
negative mantissa: numbers whose mantissa and exponent are shown below in
The significant digits in a negative (a) to (c)
mantissa start after the point at
Mantissa Exponent
the position of the first zero.
-1 ¹⁄₂ ¹⁄₄ ¹⁄₈ ¹⁄₁₆ ¹⁄₃₂ -8 4 2 1
d d d d d d d d d d

(a) 1.1111011102 00112 (b) 1.11010112 11002

(c) 1.111111000102 01112

In this chapter you have covered:

■■ Why floating point numbers are normalised:
• To maximise precision by maximising the number of significant digits
that are represented in mantissa. This is achieved by ensuring that bit
to the immediate right of point is a significant digit, 1 for positive
mantissas, 0 for negative mantissas

• Ensures that the representation is unique.

■■ Normalising un-normalised floating point numbers with positive and
negative mantissas: see algorithms

90 Single licence - Abingdon School

5 Fundamentals of data representation
5.4 Binary number system
Learning objectives
■■Be able to explain ■■ 5.4.9 Underflow and overflow
• underflow
Underflow
• overflow Underflow occurs when a number is too small in magnitude to be
represented.
■■Be able to explain the
circumstances in which they Fixed point underflow
occur If the smallest fraction that can be represented is in decimal, let’s say, ¼, for the
given number of bits and fixed point representation (see Table 5.4.9.1), then
multiplying together ½ and ¼ results in something smaller than ¼, i.e. ⅛. For
Key principle
the given fixed point representation shown in Table 5.4.9.1, e.g. 8 bits with 2
Underflow: bits after the binary point, ⅛ is too small to be represented. We say underflow
Underflow occurs when has occurred. The number is too small in magnitude to be represented.
a number is too small in
magnitude to be represented. 32 16 8 4 2 1 ½ ¼
0 0 0 0 0 0 0 1
Table 5.4.9.1 8-bit fixed point representation with 2 bits allocated to the
fractional part
Floating point underflow
When we normalise a two’s complement floating point binary representation
of a number we ensure that the mantissa lies in the decimal ranges as shown in
Table 5.4.9.2.

Positive mantissa +½ ≤ mantissa < +1

Negative mantissa -1 ≤ mantissa < -½

Table 5.4.9.2 Ranges of positive and negative mantissa

Therefore, the least positive and the least negative decimal numbers that can be
represented in two’s complement normalised binary floating point form are as
shown in Table 5.4.9.3 where n is the number of bits assigned to the exponent.
n = no of
Magnitude Example n = 3
exponent bits
(n – 1)
½ x 2-2
(n – 1) (3 – 1)
Least positive number ½ x 2-2 = 0.5 x 2-2
2
= 0.5 x 2-2 = 0.5 x 2-4

< -½ x 2-2(n – 1)
less negative than (3 – 1)
Least negative number (n – 1) = -0.5 x 2-2
-½ x 2-2
2
< -0.5 x 2-2 = -0.5 x 2-4
Table 5.4.9.3 Least positive and negative numbers

Single licence - Abingdon School 91

5 Fundamentals of data representation

This means that there is a range of numbers either side of zero for which
a floating representation does not exist as shown in Figure 5.4.9.1. Zero
can be treated as a special case by assigning it with its own un-normalised
representation, e.g. an all zeroes mantissa and exponent or we have to use the
least positive or least negative number representations for zero. For the specifics
of how it is done in current computer systems look up the IEEE standard.
Negative Positive
Underflow Underflow
Expressible Expressible Positive
Negative
Negative Positive Overflow
Overflow
Numbers Zero Numbers
(n -1) -1 (n -1) (n -1) (n -1) -1
- 1 x 2+2 approx. -0.5 x 2 -2 +0.5 x 2 -2 approx. +1 x 2+2

0
Figure 5.4.9.1 Number range for floating point two’s complement binary where n is the
number of exponent bits
If a calculation produces a final result that lies in this range then it cannot be
represented and underflow has said to have occurred. Intermediate results
of the calculation may encroach into this range but after normalisation the
final result is restricted to the range of representable numbers for the specified
number of storage bits.
Simplified example
Figure 5.4.9.2 shows the least positive and the least negative normalised
Key fact
floating point binary representations in 7 bits for a 4-bit mantissa and a 3-bit
Underflow can occur when exponent, 01002 1002 and 10112 1002, respectively, or in decimal +½ x 2-4 and
dividing a small number by a very -⅝ x 2-4.
large number or when subtracting
two numbers of the same sign
Mantissa Exponent
and close in magnitude. -1 ½ ¼ ⅛ -4 2 1
Least positive normalised 0 1 0 0 1 0 0
Least negative normalised 1 0 1 1 1 0 0
Figure 5.4.9.2 7-bit two’s complement floating point binary least positive
and least negative number normalised representations
If we represent in two’s complement floating point form, the fraction ¹⁄₆₄
which is in fixed point two’s complement representation 0.000001002, we
obtain 0.1002 10112 or in decimal ½ x 2-5. This requires a 4-bit exponent.
Therefore, ¹⁄₆₄ cannot be represented in the 4-bit mantissa, 3-bit exponent
form shown in Figure 5.4.9.2. This is an example of underflow.
Circumstances when underflow can occur
Underflow can occur when dividing a small number by a very large number
or when subtracting two numbers of the same sign which are close in
magnitude.

92 Single licence - Abingdon School

 5.4.9 Underflow and overflow

Questions
1 Using the form of representation shown below in which six
bits are assigned to the mantissa and four to the exponent, state
whether underflow occurs when attempting to represent the
following numbers expressed in fixed point two’s complement
binary.
Mantissa Exponent
-1 ¹⁄₂ ¹⁄₄ ¹⁄₈ ¹⁄₁₆ ¹⁄₃₂ -8 4 2 1
d d d d d d d d d d

(a) 0.0000000001110112 (b) 0.0000000012

(c) 1.11111111011112 (d) 1.111111111011112

2 The least positive number that can be represented in a particular

computer system is 2-129. The following expression is to be
evaluated by this computer system
-2
210 x 2 ∕2129
In which order should the evaluation be arranged and why?

Overflow
Overflow occurs when a number is too large in magnitude to be represented.
Key principle

Integer overflow Overflow:

Overflow occurs when a number
Overflow can occur when representing integers. For example, when 8 bits
is too large in magnitude to be
are allocated to store an integer in two’s complement form, as shown in Figure represented.
5.4.9.3, the most positive number that can be represented is 011111112 in
binary. If 1 is added to this then the representation becomes 100000002, a
change of sign. Overflow has occurred.
Similarly, if 1 is subtracted from the most negative number that can be
represented in binary, 100000002, the representation becomes 011111112, a Key fact
change of sign. Again overflow has occurred. Subtracting 12 is equivalent to
Overflow can never occur if
adding -12. adding two numbers of opposite
-128 64 32 16 8 4 2 1 sign only when adding numbers of
the same sign.
0 1 1 1 1 1 1 1
Figure 5.4.9.3 Most positive number that can be represented in 8-bit two’s
complement binary
It should be clear that overflow can occur when adding two positive integers Key point
or two negative integers because in each case there is the potential to produce
When adding numbers of the
an even larger positive or negative integer that cannot be represented.
same sign, overflow can be
On the other hand overflow can never occur if adding two numbers of detected by observing a change
opposite sign. of sign.

Single licence - Abingdon School 93

5 Fundamentals of data representation

Questions
3 State whether performing the following arithmetic with integers represented in 8-bit two’s complement
form will result in overflow.

(a) 111100112 + 000011012 (b) 100000112 – 000001002

(c) 100011112 + 111100002 (d) 10000002 + 011111112

Floating point overflow

The most positive and negative decimal numbers that can be represented using
two’s complement binary floating point form are as shown in Table 5.4.9.4
where n is the number of bits assigned to the exponent.

n = no of
Magnitude Example n = 3
exponent bits
(3 – 1)
<+1 x 2(2 - 1)
less than 2
Most positive number = +1 x 2(2 - 1)
+1 x 2 (2(n – 1) - 1)
= +24-1 = +23 = 8
(3 – 1)
-1 x 2(2 - 1)
more negative than 2
Most negative number = -1 x 2(2 - 1)
-1 x 2 (2(n - 1) - 1)
= -24-1 = -23 = -8
Table 5.4.9.4 Two’s complement binary floating point representation of
most positive and negative decimal numbers
If a calculation produces a final answer which is more positive than the most
Key fact
positive number then positive overflow has occurred.
Circumstances when
Conversely, if a calculation produces a final answer which is more negative
overflow can occur:
Overflow can occur when adding than the most negative number that can be represented then negative
two very large numbers of the overflow has occurred.
same sign or in multiplication
Circumstances when overflow can occur
involving large numbers or when
dividing a number by a very small Overflow can occur when adding two very large numbers of the same sign or
number. in multiplication involving large numbers or when dividing a number by a
very small number.

Questions
4 Using the form of representation shown below in which six bits are assigned to the mantissa and four
to the exponent, state whether overflow occurs when attempting to represent the following numbers
expressed in signed fixed point binary.

Mantissa Exponent
-1 ¹⁄₂ ¹⁄₄ ¹⁄₈ ¹⁄₁₆ ¹⁄₃₂ -8 4 2 1
d d d d d d d d d d

(a) +10000000.02 (b) +1111100.02 (c) -10000000.02 (d) -100000000.02

94 Single licence - Abingdon School

 5.4.9 Underflow and overflow

Questions
5 The most negative number that can be represented in a particular computer system is -2127.
The following expression is to be evaluated by this computer system

-2127 x 220∕230

In which order should the evaluation be arranged and why?

6 Discuss whether it is preferable to use the right hand side of the identity shown below, or the left, when
writing a program to evaluate x2 - y2. State the circumstances for your choice(s).
Both x and y represent real numbers.

x2 - y2 = (x + y) ( x – y)

In this chapter you have covered:

■■ Underflow - underflow occurs when a number is too small in magnitude to be represented.
■■ Overflow - overflow occurs when a number is too large in magnitude to be represented.
■■ The circumstances in which they occur
• Underflow can occur when dividing a small number by a very large number or when subtracting
two numbers of the same sign and close in magnitude.
• Overflow can occur when adding two very large numbers of the same sign or in multiplication involving
large numbers or when dividing a number by a very small number.
• Overflow can never occur if adding two numbers of opposite sign only when adding numbers of the same
sign.
• When adding numbers of the same sign, overflow can be detected by observing a change of sign.

Single licence - Abingdon School 95

5 Fundamentals of data representation
5.5 Information coding systems
Learning objectives:
■ Describe the following coding
systems for coding character
■ 5.5 Information coding systems
data ASCII
Figure 5.5.1 shows the states of traffic lights,
• ASCII
amber, green, red & amber, red. Each state encodes
• Unicode a message using light codes, e.g. green means GO,
red means STOP.
■ Explain why Unicode was
Traffic lights are able to use light codes successfully
introduced
to convey information of an instructional form to
■ Differentiate between the Figure 5.5.1 Traffic lights
road users because the communication is between
character code representation and their states
machine, the traffic lights, and humans in control of
of a decimal digit and its pure another kind of machine, e.g. a motor car. Humans
binary representation can interpret the light codes from the traffic lights and decide whether to
proceed through the junction controlled by the lights or not.
■ Describe and explain the use
In computing, we often want to send
of :
data between one computer or part of
• parity bits one computer and another, e.g. between
• majority voting the central processing unit (CPU) and a
flat screen display, that map to symbols
• checksums
which humans use for communication,
• check digits i.e. letters of an alphabet. Light codes
Key concept are not used because the language of
Figure 5.5.2 Two computing digital computers is binary not coloured
Machine to machine
machines communicating in binary lights, instead binary codes in electrical
communication of human
form are used. Human readable text and
readable text:
Digital computers and their their binary-coded equivalent are mapped to each other in terminal devices
components send and receive such as keyboards and their controllers, and visual display units and their
binary codes that are mapped controllers. Two coding schemes that map between human readable symbols
in terminal devices such as
and binary codes that represent them are ASCII and Unicode.
keyboards and visual display
units to symbols which humans In ASCII, the symbols corresponding to the letters of the alphabet (upper case
use to communicate, i.e. letters and lower case), punctuation marks, special symbols and the decimal digits 0 to
of an alphabet. Two such
9 are assigned different 7-bit binary codes according to a look up table, Table
coding schemes that do this are
5.5.1 which shows 96 of the possible 128 codes (27). For example, the ASCII
ASCII and Unicode.
code for the letter A is 1000001 in binary and 65 in decimal.
All 128 codes are called character codes because they encode what is
collectively known as characters. However, only 95 codes are actually used for
symbols, the other 33 are control codes, codes 0 to 31 and the code 127 which
is reserved for an instruction delete a character code.

Single licence - Abingdon School 96

5 Fundamentals of data representation

ASCII was invented in the 1960s so that information could be exchanged over telephone wires between data
processing equipment. ASCII stands for American Standard Code for Information Interchange. Messages were
prepared on paper tape, similar Code
Character
Code
Character
Code
Character
Code
Character
in decimal in decimal in decimal in decimal
to that shown in Figure 5.5.3, by
32 Space 56 8 80 P 104 h
punching holes in the tape (a hole
33 ! 57 9 81 Q 105 i
= 1, an absence of a hole = 0) and 34 “ 58 : 82 R 106 j
then the tape was read by a sending 35 # 59 ; 83 S 107 k
machine connected to a telephone 36 $ 60 < 84 T 108 l
line. At the other end of the line 37 % 61 = 85 U 109 m
was another machine that would 38 & 62 > 86 V 110 n
interpret the received ASCII codes 39 ‘ 63 ? 87 W 111 o
40 ( 64 @ 88 X 112 p
and then print the corresponding
41 ) 65 A 89 Y 113 q
message in symbol form on paper
42 * 66 B 90 Z 114 r
for a human to read.
43 + 67 C 91 [ 115 s
44 , 68 D 92 \ 116 t
45 - 69 E 93 ] 117 u
46 . 70 F 94 ^ 118 v
47 / 71 G 95 _ 119 w
Figure 5.5.3 5-bit Punched paper
48 0 72 H 96 ` 120 x
tape large black dot = a hole = 1,
49 1 73 I 97 a 121 y
absence of a hole = 0 50 2 74 J 98 b 122 z
51 3 75 K 99 c 123 {
Table 5.5.2 shows a lookup table 52 4 76 L 100 d 124 |
for ASCII control codes, 0 to 31. 53 5 77 M 101 e 125 }
The codes with a blank character 54 6 78 N 102 f 126 ~
field are codes used for controlling 55 7 79 O 103 g 127 DEL

communication over a telephone Table 5.5.1 ASCII code lookup table

line. Line feed and carriage return Code Code
Character Character
are used to break a long string of characters into in decimal in decimal

separate lines. When characters are organised on a 0 Null 16

1 17
line-by-line basis we call this text, e.g. the text that
2 18
you are reading on this page.
3 19
Text files therefore consist of one long string 4 20
of ASCII character codes with the line breaks 5 21
marked by a combination of ASCII code 10 (line 6 22
feed) and ASCII code 13 (carriage return). These 7 Bell 23
8 Backspace 24
control codes reposition a VDU’s cursor at the
9 Horizontal tabulation 25
beginning of the next line when displaying a text
10 Line feed 26
file on a VDU.
11 Veetical tabulation 27 Escape

12 Form feed 28
13 Carriage return 29
14 30
15 31
Table 5.5.2 ASCII code lookup table for some control codes
97 Single licence - Abingdon School
 5.5 Information coding systems

Questions Key concept

1 What is the ASCII character code for ASCII or American Standard
Code for Information
(a) the letter H (b) the decimal digit 3 (c) the symbol ?
Interchange:
2 What is the symbol or character corresponding to the following ASCII In ASCII, the symbols
corresponding to the letters of
character codes
the alphabet (upper case and
(a) 97 (b) 37 (c) 48? lower case), punctuation marks,

special symbols and the decimal
3 Encode the message “Hello” in ASCII.
digits 0 to 9 are assigned
different 7-bit binary codes
4 Why is ASCII code 127 the control code for the instruction delete a
according to a look up table.
character code (HINT: a clue is in the holes punched in 5-bit paper
tape - see Figure 5.5.3)?
Information
5 Encode the text
“Hello Extended ASCII:
This is the 8-bit version of ASCII
World!”
consisting of 28 or 256 character
in ASCII. codes or code points. The code
points beyond 127 use the eighth
6 Convert the following string of ASCII character codes to its equivalent
bit. These map to symbols that are
text form
not covered in 7-bit ASCII, e.g. £
72 101 108 108 111 10 13 87 111 114 108 100 33 sign.
A code point or code position is
any of the numerical values that
Unicode make up the code space.
ASCII provides only 128 numeric values, and 33 of those are reserved for
special functions - the control codes and delete. Many of the control codes are
Information
no longer needed because they have their origin in the days of the teletype,
punched cards and paper tape. ASCII does not cater for many Western Scan codes:

European languages which have accented letters, and special symbols such as £, When a key on a keyboard is
pressed a scan code is generated.
as it was designed for the North American market and it certainly doesn’t cater
Scan codes are binary codes
for Asian languages which are logogram-based (symbols represent concepts), as well. The scan code that is
not alphabetic. The 95 ASCII codes for characters found in text are wholely generated is converted into an
inadequate for a universal standard for information interchange. ASCII code that corresponds
to the current setting for the
Unicode was designed to provide a single character set that covers the languages
keyboard’s keys. The mapping
of the world. Unicode UTF-16 uses either one or two 16-bit code units for
between scan codes and ASCII
its character codes. A single 16-bit unit supports 216 or 65536 different codes. codes can be changed. For
Unicode UTF-32 uses 32-bit code units each representing a single character example, the mapping for a key
code. Unicode includes all the ASCII codes in addition to codes for characters marked in one currency symbol
can be changed so that when
in foreign languages (e.g. complete sets of Chinese characters), and many
pressed it maps to the ASCII code
mathematical and other symbols. for another currency symbol,
UTF-8 encodes each of the 1,112,064 valid code points in the Unicode code e.g. $ to £ (extended ASCII code
16310). This was the only way
space using one to four bytes. The first 128 characters of Unicode, which
to overcome ASCII’s limited
correspond one-to-one with ASCII, are encoded using a single byte with the character set until the adoption
same binary value as ASCII. of Unicode.

Single licence - Abingdon School 98

5 Fundamentals of data representation

Character form of a decimal digit

Table 5.5.3 has been constructed by copying the code points for the decimal
digits 0 to 9 from Table 5.5.1.
Humans work with numerals consisting of decimal Code
Symbol
in decimal
digits, e.g. 261, when they do a calculation or
48 0
record a number. If a decimal numeral sent from
49 1
one computer or computer component to another 50 2
is used by a human at the receiving end for a 51 3
calculation, the decimal digits of the numeral must 52 4
first be mapped to their ASCII code equivalents 53 5
before sending, and mapped back on receipt from 54 6
ASCII code to decimal digit form. 55 7
56 8
For example, if 261 is typed at the keyboard, the 57 9
sequence of ASCII codes 50, 54, 49 is generated
Table 5.5.3 ASCII codes
and sent. A visual display unit (VDU) receiving
for the decimal digit
these ASCII codes knows that it should display 261
symbols 0 to 9
on its screen - see Figure 5.5.4.
Decimal number To Decimal number
typed at keyboard To 49 54 50 decimal displayed on VDU
261 ASCII digits 261

Figure 5.5.4 From decimal numeral to ASCII codes and back to decimal numeral
The ASCII codes 50, 54, 49 are called the character code form of the decimal
digits 261 e.g. 50 is the character code form of the decimal digit 2. To convert
this character code form 50 into the number 2 we need to subtract 48. The
character code form of the decimal number 2 in 7-bits is 0110010 whereas its
pure binary representation is 000 0010 in 7-bits.
Symbolically, the character code form 50, 54, 49 can be written as ‘2’ ‘6’ ‘1’. The
single apostrophes around each digit are used to differentiate the character form
from the decimal digit form.

Questions
7 What needs to be done to convert the following ASCII codes to their equivalent decimal digit
(a) 53 (b) 48 (c) 57?

8 What is the ASCII character code form of the following decimal digits and combination of decimal digits
(a) 6 (b) 34 (c) 908 (d) 444?
9 Why is it difficult to do arithmetic with the character form of a decimal numeral?

10 What would need to be done with the character form of a decimal numeral in order to do arithmetic in the
conventional way?

11 What is the ASCII character code form of the following characters and character strings
(a) '6' (b) '54'?

99 Single licence - Abingdon School

 5.5 Information coding systems

Error checking and correction

Key fact
Every time information is transmitted it may get corrupted by electrical
interference or faulty hardware, and result in errors in the information Errors:
Every time information is
received. Faulty hardware may also cause errors to suddenly appear in
transmitted it may get corrupted
information stored in a storage device. by electrical interference or faulty
The solution to this problem is to use hardware, and result in errors in
the information received.
redundancy to add reliability to information
Faulty hardware may also cause
in transit or in storage. The data (data is errors to suddenly appear in
how information is represented) is extended information stored in a storage
by including additional data used for error device.
checking and correction.
Majority voting
Majority voting is an error correction Key concept
Figure 5.5.5 Error detected method that duplicates or copies each bit in
Majority voting:
in data bits the message an odd number of times before
Majority voting is an error
sending these copies. For example, if the correction method that
message consists of three bits, 101, then the thrice duplicated message would duplicates or copies each bit
consist of nine bits as follows 111 000 111. The size of the message is thus in the message an odd number
of times before sending these
increased but without increasing the amount of information. The message
copies, e.g. 101 becomes 111
therefore contains additional redundancy (it may already be redundant, e.g. 000 111.
message = “ It is hot. It is hot.”). However, this additional redundancy can be If for each triplet all three bits
used for error correction. are identical then the receiver
assumes that they are correct (it
Let’s first see how error detection can be achieved by just duplicating the is possible but very unlikely that
message bits twice. If the data 1011 have to be transmitted then the bits 11 00 111 gets corrupted to 000). If
11 11 are sent instead. If the receiver receives a pair of bits with non-identical only two bits in a triplet are the
bits then it knows an error has occurred but it won’t know if, for example, 01 same and the third is different,
the receiver assumes that the
was originally 00 or 11. Duplication twice has allowed error detection but not
two bits the same are correct
correction. and the third bit is in error.
To allow for error correction, we need to copy the message bits an odd number Majority voting does not
guarantee absolute reliability.
of times. For example, 1011 becomes 111 000 111 111.
On receipt of this redundant bit pattern, the receiver compares the three bits
of each triplet. If for each triplet all three bits are identical then the receiver
assumes that they are correct (it is possible but very unlikely that 111 gets
corrupted to 000). If only two bits in a triplet are the same and the third is
different, the receiver assumes that the two bits the same are correct and the
third bit is in error. This is what is meant by majority voting. The message bits
need to be duplicated an odd number of times, n, for majority voting to make
a decision.
For the above example, if transmission errors change 111 000 111 111 to 110
010 101 111, majority voting applies error correction producing 111 000 111
111 and the recovered message 1011.

Single licence - Abingdon School 100

5 Fundamentals of data representation

Majority voting does not guarantee absolute reliability. Careful consideration

of this example will tell you that majority voting can get it wrong, but the
probability of this happening can be minimised if it isn’t already low enough by
choosing a bigger value for n.
Parity bits

Key concept If error detection rather than error correction is sufficient then the parity bit
method can be used. The parity bit is computed from a group of n data bits
Parity bit: and then added to the group, making it n + 1 bits long. For example, a 7-bit
The parity bit is computed from
ASCII code becomes 8 bits long after a parity bit is added. The parity bit is
a group of n data bits and then
added to the group, making it computed by counting the number of
n + 1 bits long. ones in the n bit data group, and then
The parity bit is computed by setting the parity bit to make the count
counting the number of ones in
for the n + 1 group (parity + data) either
the n bit data group, and then
setting the parity bit to make
even or odd. The former is called even
the count for the n + 1 group parity and the latter odd parity.
(parity + data) either even or
For example, the count of 1s for the Figure 5.5.6 7-bit ASCII
odd. The former is called even
7-bit ASCII code 0101101 is 4. character code
parity and the latter odd parity.
A transmission or disk read is With even parity this becomes the 8-bit
judged reliable if the parity bit code 00101101 with the parity bit set to 0 to make the count of 1s across the
regenerated from the n data bits 8 bits an even number. With odd parity this becomes the 8-bit code 10101101
agrees with the received parity
with the parity bit set to 1 to make the count of 1s across the 8 bits an odd
bit.
Parity bit checking only works number. Now suppose that even parity is used and 00101101 is sent. If the
if an odd number of bits have pattern 01101101 is received then an error has occurred because the count of
been flipped. 1s is now odd.

The parity bit can be computed by applying the exclusive-OR (XOR) to the n
data bits because an XOR operation performs modulo-2 addition. Thus a series
of XOR operations can perform the counting.
Suppose the data is the 7-bit ASCII code, 0101101, and the XOR operation is
denoted by ⊕ then
0 ⊕ 1 ⊕0 ⊕1 ⊕1 ⊕0 ⊕1 = 1 ⊕1 ⊕1 ⊕1 = 0 ⊕0 = 0
Information
The XOR-computed parity bit for the 1011010 is 0 for even parity. Inverting
Exclusive-OR (XOR): the computed XOR-computed result gives 1 for odd parity. The result for
This performs modulo-2 addition, parity + data is thus as follows
i.e. integer addition modulo 2.
EVEN parity: 00101101 ODD parity: 10101101
+ 0 1 Now suppose that the byte 00101101 (most significant bit(MSB) a parity bit)
0 0 1 is read from disk and even parity is used. To check that this byte has been read
1 1 0 reliably, the parity bit for its 7 data bits is computed (by hardware or software)
using XOR (0) and compared with the MSB using XOR again (0 ⊕ 0 = 0). The
transmission or disk read is judged reliable if the regenerated parity bit agrees
with the received parity bit. This judgement is not always correct as two bits or
an even number of bits may be corrupted during transmission. However, use
101 Single licence - Abingdon School
 5.5 Information coding systems

of a single parity bit is usually sufficient except when circumstances dictate that
full error-detection capability is required.

Questions
12 Calculate

the parity bit using even parity for the following 7-bit codes (a) 0111000 (b) 1110010

13 Calculate the parity bit using odd parity for the following 7-bit codes (a) 0111000 (b) 1110010

14 Explain how a receiver of a data transmission consisting of one parity bit and 7 data bits can detect that an
error has occurred affecting an odd number of the 8 received bits.

Checksums
Parity checking is good for checking asynchronous serial transmission of data Key concept
over short distances but not very good for synchronous serial transmission over
Checksum:
long distances. For the latter, checking must be applied to a block of data. The
The checksum is a number
data is treated as a sequence of fixed size numbers, e.g. each one byte in size. appended to the end of a block
These numbers are added together to form a total which is then truncated to of data and used for error
the same size as the number size, e.g. one byte, often by hashing. This truncated detection and correction. The
total is known as the checksum. The checksum is appended to the end of the block of data is treated as a
sequence of fixed size numbers,
block and for this reason is also referred to as a block check character. When
e.g. each one byte in size which
the data arrives at the receiver, the checksum is regenerated and compared are added together to form a
with the transmitted checksum. In this way the received data can be checked total which is then truncated
for errors that have arisen in to the same size as the number
Parity bit

transmission. size, e.g. one byte, often by

hashing. This truncated total is
Two common checksum methods the checksum for the block.
are LRC and CRC. LRC is
1 1 0 0 1 0 1 0
0 0 1 0 1 1 0 1 an acronym for Longitudinal
1 1 1 0 1 0 0 0 Redundancy Check or Longitudinal
1 0 1 1 1 1 1 0 Redundancy Character. CRC is an
0 0 1 1 1 1 1 1
1 1 0 0 0 1 0 1 acronym for Cyclic Redundancy
1 1 0 1 1 1 1 0 Check or Cyclic Redundancy
Checksum 1 0 0 1 0 1 0 1 Character.
Figure 5.5.7 Longitudinal LRC uses a block check character
Redundancy Check using a made up of a parity bit for each
checksum formed by computing the column as shown in Figure 5.5.7.
parity bit for each column (vertical By using horizontal parity bits as
parity) well, it is possible to correct some
errors. If the indicated bit  is
flipped to become a 0 then both the vertical parity and the horizontal parity
checking will indicate an error has occurred. The horizontal parity bit will
indicate the row and the vertical parity the column. Thus the data bit in error
can be located and corrected.

Single licence - Abingdon School 102

5 Fundamentals of data representation

Questions
15 What is a checksum?
1 1 0 0 1 0 1 0
16 The checksum for the block of data opposite uses even vertical parity. 0 0 1 0 1 1 0 1
1 1 1 0 1 0 0 0
Horizontally the most significant bit is an even parity bit. There is a 1 0 1 0 1 1 1 0
single bit error in this block. Can you identify which bit is in error? 0 0 1 1 1 1 1 1
1 1 0 0 0 1 0 1
1 1 0 1 1 1 1 0
1 0 0 1 0 1 0 1

Check digits
Check digits and parity bits are special cases of checksums. The maths used for
Key concept parity bits works for binary numbers but not decimal numbers. Thus different
methods must be used for making decimal number data such as credit card
Check digit:
numbers and book ISBNs reliable.
A check digit is a decimal digit
added to a number (either at A check digit is a decimal digit added to a number (either at the end or the
the end or the beginning) to beginning) to validate the number, e.g. a valid book ISBN.
validate the number, e.g. a valid
book ISBN. For example, the check digit in ISBN 978-0-
The main task of a check digit 9927536-2-7 shown in Figure 5.5.8, is the
is to detect a single corrupted rightmost 7 digit. This 7 is computed by an
digit and a transposition of two
algorithm applied to the information digits
adjacent digits.
of the number, i.e. 978-0-9927536-2. On Figure 5.5.8 ISBN-13
entering this ISBN into a computer, the check- book code
digit generating algorithm is applied to the 978-0-9927536-2-7
information digits of the ISBN as before, and showing check digit 7
the re-computed check digit compared with the
check digit that was entered (see later for a more
efficient way of doing this). In this way it is possible to check that the book
ISBN has been read correctly.
The three most common errors made by humans when keying numbers
into a computer, or reading and saying them, are omitting or adding a digit,
transposing adjacent digits and changing a single digit.
For example, transposing the digits 2 and 7 in 978-0-9927536-2-7 or changing
the triplet 992 to 922. The omission or addition of a digit is easily detected
without a check digit. Therefore, the main task of a check digit is to detect a
single corrupted digit and a transposition of two adjacent digits. Other types of
error are rare.
Check digits normally use modular arithmetic. The mathematical function
a mod b returns the remainder of the integer division a /b, an integer in the
range 0 to b - 1. Given a number N that consists of decimal digits d1 d2 d3 ...,
the simplest way to compute a check digit C for N is to solve the equation
(C + d1 + d2 + d3 + ...) mod p = 0

103 Single licence - Abingdon School

 5.5 Information coding systems

choosing an appropriate value for p. Note that for the lefthand side of this
equation to be 0
(C + d1 + d2 + d3 + ... ) must be a multiple of p
Therefore, this equation can be solved by first computing the sum S as follows
S = (d1 + d2 + d3 + ... ) mod p
and then using the fact that if C is restricted to the range 0 to p - 1,
C+S=p
Rearranging, C=p-S

Example
Suppose N is a three-digit number and each digit is in the range 0 to 4, inclusive, then a good choice for
p is 5.
If N = 342, S = (3 + 4 + 2) mod 5 = 4
then C = 5 - 4 = 1
The check digit 1 is appended to the number N and the 4-digit number 3421 is given over the
telephone, stored in a computer or transmitted over a communication line. At the receiving end, the
4-digit number is checked. If no digits have been corrupted, the calculation (3 + 4 + 2 + 1) mod 5 will
yield 0 (remember (C + d1 + d2 + d3 + ...) mod p = 0). However, if the received 4-digit number has been
corrupted in a single digit, e.g. it became 3221, then the calculation (3 + 2 + 2 + 1) mod 5 yields 3 when
it should be 0. Detection of single-digit errors are possible with this simple check digit mechanism.

However, it is not possible to detect any transposition of digits.

Therefore, the check digit is calculated by applying weights to each digit as
follows
(C + w1∙ d1 + w2∙ d2 + w3∙ d3 + ...) mod p = 0
or S = (w1∙ d1 + w2∙ d2 + w3∙ d3 + ...) mod p
and C = p - S

Example
Suppose N is a three-digit number and each digit is in the range 0 to 4, inclusive, then a good choice for p
is 5. The weights chosen are 2, 3, and 4 because they are relatively prime to 5, i.e. 5 does not divide any of
them evenly.
If N = 342, S = (2∙3 + 3∙4 + 4∙2) mod 5 = 1
then C = 5 - 1 = 4
The check digit 4 is appended to the number N and the 4-digit number 3424 is read over the telephone,
stored in a computer or transmitted over a communication line. At the receiving end, the 4-digit number is
checked. If no digits have been corrupted, the calculation (2∙3 + 3∙4 + 4∙2 + 4) mod 5 will yield 0. However,
if two adjacent digits of the 4-digit number have been swapped because of an error, e.g. it became 3244,
then the calculation (2∙3 + 3∙2 + 4∙4 + 4) mod 5 yields 2 when it should be 0, thereby detecting an error.

Single licence - Abingdon School 104

5 Fundamentals of data representation

ISBN
ISBN-13 has a total of 13 digits and includes a check digit. It conforms to EAN-13, the
European Article Numbering barcode system. The commonly used ISBN-10 book codes
have been turned into ISBN-13 by prepending 978. ISBN-13 book codes can use an EAN-13
barcode and therefore be barcode scanned. For example, ISBN-13 book code 978-0-9927536-
2-7 has 978 followed by language/country code 0, publisher code 9927536, book number 2,
and check digit 7.
To calculate the check digit:
Add up all the even numbered positions and multiply the sum by 3.
Sum the odd numbered positions.
Total the two sums.
Add a number that rounds up this total to the nearest multiple of ten.
This number is the check digit.
Algebraically
S = (1∙9 + 3∙7 + 1∙8 + 3∙0 + 1∙9 + 3∙9 + 1∙2 + 3∙7 + 1∙5 + 3∙3 + 1∙6 + 3∙2 ) mod 10 = 3
C=p-S
C = 10 - 3 = 7
The check digit C is therefore 7.

Questions
17 What is a check digit?

18 What are the three most common errors made by humans when keying numbers into a computer, or
reading and saying them?

Using an example, describe how a check digit is calculated so that it can be used to detect two of these
19 commonest errors?

In this chapter you have covered:

■■ The following coding systems for coding character data
• ASCII
• Unicode
■■ Why Unicode was introduced
■■ The difference between the character code representation of a decimal
digit and its pure binary representation
■■ The meaning of and uses of
• parity bits
• majority voting
• checksums
• check digits

105 Single licence - Abingdon School

 5 Fundamentals of data representation
5.6 Representing images, sound and other data
Learning objectives:
■■ Describe how bit patterns may
represent other forms of data,
■■ 5.6.1(1) Bit patterns, images, sound and other data
including graphics and sound Binary, the language of the machine
The language of digital computers is binary. Whether the communication
is instructions, e.g. calculate the square of 9, or data, e.g. speech, the
communication must be transformed into discrete signals of a binary nature for
the hardware of the computer to be able to process them.
Instructions or data at this level are seen logically as sequences of bit patterns
or bits, e.g. 01101010 10001111 11000010 …., although physically they are
patterns of electrical voltage (or electric charge) in the memory of a computer,
for example zero volts and five volts.
A bit pattern is just a unit of bits (binary digits) such as a byte. 01101010
is an 8-bit bit pattern. For convenience, bit patterns are usually shown in
hexadecimal or decimal form to make viewing easier for humans - Figure
5.6.1.1.

Figure 5.6.1.1 Binary bit patterns and their equivalent hexadecimal

We can view a sequence of bit patterns representing instructions or data as just

a sequence of numbers. Each bit pattern can be treated as a binary value with
an equivalent hexadecimal or decimal value, e.g. 011010102 is 6A16 or 10610 if
treated as an unsigned integer.
When data or instructions are organised as files and stored on a computer’s
backing store, e.g. magnetic disk, a stream of bits is sent to the backing store
device. Similarly, when a file is opened for reading and its contents transferred
to the CPU or main memory of a computer, the contents are transferred as
a stream of bits. To interpret a bit stream of bit patterns as a digitised image/
digitised sound/text instructions, for example, requires that the sequence of
bit patterns is organised into an appropriate structure for viewing/playing/
displaying/executing. Applying the wrong structuring can have unintended
consequences, e.g. interpreting data as code and vice versa.

Single licence - Abingdon School 106

5 Fundamentals of data representation

Questions
1 Explain why instructions, e.g. calculate square of 9, and data,
e.g. speech, must be transformed before the hardware of a digital
computer is able to process these instructions or data.

2 What is a bit pattern?

Graphics
One way of structuring bit patterns is the Joint Photographic Experts Group
(JPEG) method for images produced by digital photography.
A JPEG file stores a digitised image as a sequence of bit patterns obtained,
for example, from a digital camera that captures a scene photographically by
sampling the brightness (or intensity) of the colour components of the scene
before digitising the result 0 255 255 255 255 255 255
Information in numbers to produce a 255 255 255 255 255 255 255

MatLab: JPEG formatted digital image 255 255 255 255 255 255 255
255 255 255 255 255 20 0
http://uk.mathworks.com/ representation of the scene –
255 255 255 255 255 255 255
products/matlab see Figure 5.6.1.2. 255 255 255 255 255 255 255

GNU Octave: When this JPEG file’s contents 255 255 255 255 255 255 255
255 255 255 21 0 255 255
http://mxeoctave.osuv.de/ are accessed and processed
correctly the digitised Figure 5.6.1.2 Image data taken from a
Redang.jpg:
recording of the original section of the JPEG formatted file, Redang.jpg
www.educational-computing.
co.uk/CS/Images/Redang.jpg
scene can be displayed as and displayed in decimal for ease of viewing.
shown in Figure 5.6.1.3. The
sequence of bit patterns serves to convey both
the digitised image itself plus information
(metadata) about the image such as its
dimensions, in this case 600 x 800.

Questions
3 Outline a method by which an image of a scene can be captured in Figure 5.6.1.3 600 × 800
digital form so that it can be displayed on an image display device. digital image stored in file
Redang.jpg

A relatively easy way to explore digital images is to use either Matlab from
Information MathWorks or GNU Octave, an open source system. The same scripts and
Octave: commands execute in either. For example, the following script
You will need to add the Z = imread('Redang.jpg');
command disp(info); to output
the value of the info. info=imfinfo('Redang.jpg');
image(Z);

107 Single licence - Abingdon School

 5.6.1(1) Bit patterns, images, sound and other data

executes in either Matlab or GNU Octave and extracts and

displays image and format information data from Redang. Filename: 'C:\Images\Redang.jpg'
jpg. FileModDate: '11-Jul-2003 12:12:16'
Figure 5.6.1.4 shows the extracted format information and FileSize: 58014
Figure 5.6.1.5 the image displayed by the command Format: 'jpg'
image (Z). FormatVersion: ''
The digital image is actually made of three separate Width: 800
monochrome digital images, one red, one green and one Height: 600
blue that are combined by the command image(Z) to BitDepth: 24
produce the 600 × 800 image shown in Figure 5.6.1.5 with ColorType: 'truecolor'
labelled x and y axes. FormatSignature: ''
When the digital camera snapped the scene it sampled the NumberOfSamples: 3
scene through three filters: a red filter, recording each red CodingMethod: 'Huffman'
sample’s intensity value in 8 bits, a green filter recording CodingProcess: 'Progressive'
each green sample’s intensity value in 8 bits and a blue filter
Comment: {}
recording each blue sample’s intensity value in 8 bits.
The red, green and blue samples are combined to produce
Figure 5.6.1.4 Produced in MatLab’s command
an RGB image of 600 × 800 samples in all. For each sample,
window by >>info
a total of 8 + 8 + 8 = 24 bits is used as indicated by the
BitDepth field.

A quick calculation indicates by comparison with the format information

FileSize that the whole collection of digital samples has undergone
compression. The JPEG format uses compression throwing away image
information that the viewer would not notice.
To process the bit patterns from the file Redang.jpg appropriately, i.e.
according to the JPEG standard, the bit patterns must be structured as
follows using two-dimensional arrays of the following dimensions: Figure 5.6.1.5 The output of the script
■ The 600 × 800 red samples into a 600 × 800 array command image(Z). Z contains the
■ The 600 × 800 green samples into a 600 × 800 array image data.

■ The 600 × 800 blue samples into a 600 × 800 array

It is the metadata on image dimensions 600 × 800 extracted from this file
that is used to determine the dimensions 600 × 800 of the arrays.
Therefore when all three two-dimensional arrays are stacked together
we obtain a 600 × 800 × 3 three-dimensional array as shown in Figure
5.6.1.6.
The script command:
Z = imread('Redang.jpg');
Figure 5.6.1.6 Three-dimensional
reads the contents of Redang.jpg, decompresses it and performs the
array Z with dimensions
processing just described, storing the image samples’ intensity values in a
600 × 800 × 3.
three-dimensional array Z with dimensions 600 × 800 × 3.
Single licence - Abingdon School 108
5 Fundamentals of data representation

Using MatLab’s Pixel region Image Tool as shown in Figure 5.6.1.7, the Red
(R), Green (G) and Blue (B) sample values of any region of the displayed image
can be retrieved.

Figure 5.6.1.7 Pixel Region Image Tool showing Red (R), Green (G)
and Blue (B) values in a region of the image.

Questions
4 A digital image file stores bit patterns representing intensity values
of samples of the scene captured by the imaging device. What other
information about the image is also stored in the image file and why?

In this chapter you have covered:

■ How bit patterns may represent graphics

109 Single licence - Abingdon School

 5 Fundamentals of data representation
5.6 Representing images, sound and other data
Learning objectives:
■■ Describe how bit patterns may
represent other forms of data,
■■ 5.6.1(2) Bit patterns, images, sound and other data
including graphics and sound Manipulating digital images
Having digitised an image, it is now just a sequence of numbers (bit patterns)
Information
to which arithmetic operations may be applied to produce new numbers
MatLab: and new forms of the digital image. For example, the following MatLab/
http://uk.mathworks.com/ GNU Octave script will double every value in the Red array, C(:, :, 1)
products/matlab
obtained after reading the JPEG image file Redang.jpg with the command
GNU Octave:
http://mxeoctave.osuv.de/ W = imread('Redang.jpg');

Redang.jpg: and storing a copy of W in C with the command

www.educational-computing. C = W;
co.uk/CS/Images/Redang.jpg

% Introduces a comment in the script

close all; % Closes all figures
clear all; % Deletes all stored variables in workspace
clc; % Removes all lines in the command window
W = imread('Redang.jpg'); % Populate 3-D array W
figure(1); % Draw contents of W as figure 1 appropriately rendered as an image
image(W); % Renders the digital image for values in W
C = W; % makes a copy of W and assigns it to C
% Every value in the 600 x 800 Red array (1) of C is now doubled and written
back into the corresponding cell of this array. This will enhance the redness
of the image
% :, :, means the entire 600 x 800 array
C(:,:,1) = 2*C(:,:,1);
figure(2); % Draw the result as Figure 2
image(C); % Render C as a digital image
% Write C to a new JPEG file RedangChanged.jpg.

imwrite(C, 'RedangChanged.jpg');

Single licence - Abingdon School 110

5 Fundamentals of data representation

The outcome is shown in Figure 5.6.1.8(b) alongside the original image,

Figure 5.6.1.8(a).

Figure 5.6.1.8(a) Array W rendered. Figure 5.6.1.8(b) Array C rendered showing the
effect of doubling every red value in W

Questions
5 Explain how each of the red and the green components of an RGB image can be reduced by 50% in
MatLab or GNU Octave.

The greyscale digitised image shown in Figure 5.6.1.9(a) occupies a single

480 × 640 two-dimensional array, C, when loaded by the MatLab/GNU

Information Octave script

PlaneGrey.jpg:
www.educational-computing. clear all;
co.uk/CS/Images/PlaneGrey.jpg C = imread('PlaneGrey.jpg’);
figure(1);
image(C);
C(:,:) = 255 - C(:,:);
Information
figure(2);
Octave:
imshow(C);
You will need to add the
command disp(info); to output imwrite(C, 'PlaneGreyNegative.jpg');
the value of the info.

If the intensity values in array C are subtracted from 255 then an intensity value
of 255 becomes an intensity value of 0, and an intensity value of 0 becomes an
intensity value of 255, and so on.
Thus we get the negative of this image when we update C as follows
C(:,:) = 255 - C(:,:);

111 Single licence - Abingdon School

 5.6.1(2) Bit patterns, images, sound and other data

Figure 5.6.1.9(b) shows the result.

Figure 5.6.1.9(a) 480 × 640 greyscale image Figure 5.6.1.9(b) 480 x 640 negative greyscale image

Programming tasks
1 Whenever the red, green and blue components of an image sample have the same value, the colour
displayed is a shade of grey. This means that a digitised image of sampled red, green and blue colours has
the potential for 256 shades of grey if each colour is encoded with 8 bits (0..255). We can use the intensity
of the overall colour, i.e. red + green + blue, to assign a shade or level of grey. The intensity of a colour
called the luminance is calculated as follows

red + green + blue

3

Write a program or script in MatLab or GNU Octave that uses this formula to set the colour of each pixel
of an RGB image to a shade of grey to produce an equivalent greyscale image.

2 If you succeeded in turning an RGB image into a greyscale image you may have noticed that the result is
not as expected. This is because the formula method used in Programming task 1 did not take into account
the way that the human eye perceives luminance, e.g. the eye is less sensitive to blue light than red. We
need to adjust for this by weighting as follows

0.299 × red + 0.587 × green + 0.114 × blue

3
Change your program or script to take account of this new formula.

3 Write a program or MatLab/GNU Octave script to rotate an image through 180 degrees, i.e. turn the
image upside down.

A digital image can be created without using a camera. We can instead create a digital coloured image
by creating a three-dimensional array of numbers, D, as shown in Figure 5.6.1.10. D is populated with
values, 0 and 255 or in binary 00000000 and 11111111, representing the intensity of red, green and blue
with 255 being the strongest and 0 the weakest.

The MatLab/GNU Octave script to generate this array, to render it as an image and write the data to a
file Squares.jpg is as follows

Single licence - Abingdon School 112

5 Fundamentals of data representation

close all; % Closes all figures

clear all; % Deletes all stored variables in workspace
D(:,:,1) = [0 255 0 255; 255 0 255 0; 0 255 0 255];
D(:,:,2) = [0 255 0 255; 255 0 255 0; 0 255 0 255];
D(:,:,3) = [0 255 0 255; 255 0 255 0; 0 255 0 255];
figure(4);
image(D);
imwrite(D, 'Squares.jpg');

[0 255 0 255; 255 0 255 0; 0 255 0 255] is the way that MatLab/GNU Octave creates a two-
dimensional array, each sequence of numbers is a row vector with rows separated by ';' so putting the
row vectors together we get, in this instance,
0 255 0 255
255 0 255 0
0 255 0 255

As we have three primary colours, three of these 2-D arrays are required, one for each colour, Red, Green,
Blue. (3) BLUE

0 255 0 255
(2) GREEN
0
0 255 0 255
(1)RED 255
0
0 255 0 255
255
255 0 255 0

0 255 0 255

Figure 5.6.1.10 Three-dimensional array, D, containing cells values, 0 or 255.

The outcome when the command image(D) is executed is a 3 x 4 grid of black and white squares on the
screen. The black square is produced by the triplet 0, 0, 0 taken from the arrays for (1) RED, (2) GREEN,
(3) BLUE. The white square is produced by the triplet 255, 255, 255 taken from the arrays for (1) RED,
(2) GREEN, (3) BLUE.

The command:
imwrite(D, ‘Squares.jpg’)

113 Single licence - Abingdon School

 5.6.1(2) Bit patterns, images, sound and other data

scans array D,as it does so writing its values to a bit stream for file Squares.jpg
using the format required by JPEG.
(3) BLUE

0 255 0 255
(2) GREEN
0
0 255 0 255
(1)RED 255
0
0 255 0 255
255 Figure 5.6.1.11 Outcome
255 0 255 0 1 0 0 1 of executing image(D), an
255 255 0 1
0 0
1 1 image of 3 by 4 squares.
DISK

Figure 5.6.1.12 Writing array D to disk

Information
Reading file contents byte by byte
Spyder python:
Files of any type, e.g. JPEG, BMP, XLS, TXT, can be opened as a file of byte https://store.continuum.io/
and their contents read as bit patterns of unit size one byte. For example, cshop/anaconda/
given access to a bitmapped file Fruit1.bmp the following Python 3.4 script Spyder is part of the Anaconda
will open, read and display both a running count and each byte of this file in system that gives access to
scientific routines including
decimal.
support for arrays and digital
signal processing in Python.
Programming tasks
4 Write a script for execution in MatLab or GNU Octave that creates
a file Squares.bmp for a black and white chequer board image with
dimensions 4 × 4 with white as the colour of the top left square.

Figure 5.6.1.13 Python 3.4 script to read byte by byte contents of a bitmap file Fruit1.bmp

Single licence - Abingdon School 114

5 Fundamentals of data representation

The size of this file is calculated as follows.

The total number of bytes necessary to store one row of pixels is
RowSize = BitsPerPixel × ImageWidth
8

where ImageWidth is expressed in pixels. A pixel is a picture element and is the smallest area of the picture that is
sampled and digitised.
The total number of bytes to store an array of pixels, ArraySize, is
RowSize × ImageHeight

where ImageHeight is measured in pixels.

Image Fruit1.bmp when displayed has dimensions 126 × 161, i.e. 126 rows each of 161 pixels. This bitmap stores 8
bits per pixel.
Therefore, RowSize = (8 × 161) / 8 = 161
and
ArraySize = 161 × 126 = 20286 bytes

The metadata occupies 1078 bytes

Therefore, total size in bytes of Fruit1.bmp = 20286 + 1078 = 21364
This calculation is close to the result obtained from running the Python 3.4 script in Figure 5.6.1.13 above. The
discrepancy is caused by the fact that RAM stores bytes in groups of four so our calculation for the RowSize is
an underestimate. It should be 164 bytes. This gives 164 × 126 bytes for total file size, i.e. 20664 + 1078 = 21742
bytes. This agrees exactly with the output of the Python 3.4 script.

Programming tasks
5 Write a program that opens a BMP image file as a file of byte. The program should copy the first 1078
bytes of the file into a new file, then write the 8-bit ASCII codes for "HELLO WORLD" to the new file after
this. It should skip copying the next 11 bytes of the original file (which are effectively replaced by "HELLO
WORLD") and then copy the rest of the data in the original file into the new BMP file. Note where the
message starts. View the new BMP file in an image viewer. Can you detect where the original image has
been altered?

Now write a program to extract the message that has been stored in the image file. The program should
use the same message starting position as was used in the program that stored the message.

Tasks
1 Investigate steganography and digital watermarking.

In this chapter you have covered:

■■ How bit patterns may represent graphics

115 Single licence - Abingdon School

 5 Fundamentals of data representation
5.6 Representing images, sound and other data
Learning objectives:
■ Describe how bit patterns may
represent other forms of data,
■ 5.6.1(3) Bit patterns, images, sound and other data
including graphics and sound
Sound
Information A WAV file, Me2.wav, is just a sequence of bit patterns or numbers
Audacity:
http://audacity.sourceforge.net/ recording the sampled and digitised waveform of a sound.
JES: File Me2.wav was sampled, and recorded in digitised form, using a
http://coweb.cc.gatech.edu/ microphone connected to a computer running Audacity, the free, open source,
mediaComp-teach
cross-platform software for recording and editing sounds.
Information This WAV file was then read from disk as a bit stream of bit patterns using
The beginning of a WAVE file JES, free, cross-platform software for interacting with graphics and sound files.
comprises a “header” storing The sequence of bit patterns read from the disk was stored in sound, a one-
information about the sound
dimensional array. JES’ Sound Tool is able to render the bit patterns stored in
data :
• number of channels
array sound as an on screen waveform of amplitude against sample number
• number of sample frames as shown in Figure 5.6.1.14. Each sample value can be shown on screen using
• word size (16bit, 24bit, etc) this tool. The samples are stored in as 16-bit twos’ complement integers
• sample type (int, float) (−32768 to 32767). JES displays the sample values in decimal.
• sample rate

Figure 5.6.1.14 JES GUI showing the Command window and the Sound
Tool window and sample 326368 whose value is -2963.
The command makeSound(bitStream) reads the bit patterns from bit
stream bitStream which itself is connected to WAV file Me2.wav. It
extracts the sampling rate, the number of bits per sample and the type of
recording (mono or stereo) all of which are stored in this file. With this
information, makeSound(bitStream) constructs either a one-dimensional

Single licence - Abingdon School 116

5 Fundamentals of data representation

Information array (mono) or a two-dimensional array (stereo) and then stores the bit stream
bit patterns in the constructed array.
For sound file I/O by far the
Figure 5.6.1.15 shows Me2.wav opened by a Python 3.4 script running in
best add-in module for Python
is “pysoundfile”. Spyder. It extracts the sampling rate and assigns this to variable samplingRate
http://pysoundfile.readthedocs. and the sound data which it assigns to variable soundData. Before printing
org/en/0.8.1/ both, samples per second (44100) and array soundData [9 -2 12 …, -50,
-48 -47].

Information
Spyder:
https://store.continuum.io/
cshop/anaconda/
Spyder is part of the Anaconda
system that gives access to
scientific routines including
support for arrays and digital
signal processing.

Figure 5.6.1.15 WAV file Me2.wav opened in Spyder by a Python 3.4 script.

Programming tasks
6 Using JES, Spyder Python 3.4 or another programming/scripting system that supports exploration of
digitally recorded sound, write a program/script/commands to open WAV files, read the stored sampled
sound values and display these. Try also to extract the sampling rate and bits per sample.

Information Creating digital sound files

The MatLab/GNU Octave script shown in Figure 5.6.1.16 creates a sequence
MatLab of numbers or bit patterns, allocating 16 bits to each bit pattern, to represent
http://uk.mathworks.com/
the digital equivalent of a continuous tone of a frequency/pitch 1000 Hz
products/matlab 1th
sampled every 20000 of a second. The bit patterns or numbers are stored in
GNU Octave WAV format in file Tone.Wav together with the sampling rate and the bits per
http://mxeoctave.osuv.de/ sample.
Tone.wav can be played using Windows Media Player or any other suitable
media player.

117 Single licence - Abingdon School

 5.6.1(3) Bit patterns, images, sound and other data

SampleRate = 2e4; % 20000 samples per second

t = 0:1/2e4:1-(1/2e4); % time step 1/2e4 from 0 to 1 - 1/2e4
x = 1/2*cos(2*pi*1000*t); % cosine value at time t
% write the signal x to Tone.wav file using 16 bits per sample
wavwrite(x, SampleRate,16, 'Tone.wav');

Figure 5.6.1.16 Generating mathematically a sequence of numbers that represent a time sequence of samples of
a continuous tone of frequency 1000 Hz sampled at a rate of 20000 samples per second or one every 1⁄20000th
of a second. The sequence is written together with the sampling frequency and the bits per sample, to file
Tone.wav.

Programming tasks
7 Using MatLab or GNU Octave, mathematically generate separate WAV files of the following tones (use
trigonometric function cosine and then repeat using trigonometric function sine)

(a) 500 Hz (b) 2000 Hz (c) 4000 Hz (d) 8000 Hz

Use sampling rate 20000 samples per second, bits per sample 16 and collect 20000 samples
(0 to 1 − 1⁄20000 in time steps of 1⁄20000 second).

Play your generated tones in a media player.

Manipulating digital recordings of sounds

Just as it is possible to manipulate digital images because they are represented by bit patterns/numbers so it is
possible to manipulate digital recordings of sounds
because they too can be accessed as a sequence of bit
patterns/numbers. A simple way of demonstrating
this is to create a WAV file using a script similar to
that shown in Figure 5.6.1.16.
The sampled points of the wave are indicated
in Figure 5.6.1.17 with and . The height
(amplitude) of the wave is normalised (adjusted
to a desired value) in the figure for convenience, 1
corresponds to +32767 and -1 to -32768.
The chosen frequency for this explanation is
deliberately low in order that the numbers are Figure 5.6.1.17 Cosine wave marked with
manageable. sample points  and 
Using normalised values we have a sequence of samples

1.0, 0.9239, 0.707, 0.3826, 0.0, -0.3826, -0.707, -0.9239, -1.0, -0.9239, -0.707, -0.3826, 0.0, 0.3826, 0.707, 0.9239, etc
1th
The values in this sequence are separated in time by of a second because the sampling rate used was 20000
20000 1th
samples per second. The “sampling interval” or “sampling period” for this sample rate is .
20000

Single licence - Abingdon School 118

5 Fundamentals of data representation

If we read this sequence from the beginning and write the sequence to a new WAV file, ToneFreqDoubled.wav,
omitting every other value, then the sequence in the new file is
1.0, 0.707, 0.0, -0.707, -1.0, -0.707, 0.0, 0.707, 1, etc

These are the samples indicated by  in Figure 5.6.1.17.

If we record the sampling frequency as 20000 samples
per second in this new file, then when it is read back, a
sample will be separated in time from the next sample
1th
by of a second. If the sequence of numbers is
20000
plotted on the same time scale as Figure 5.6.1.17 then
we get the waveform shown in Figure 5.6.1.18. This has
5 complete waves to the 2.5 waves in Figure 5.6.1.17,
i.e. the frequency of the wave has been doubled. A script
to double frequencies of digitally recorded sounds in Figure 5.6.1.18 new waveform
WAV files is shown in Figure 5.6.1.19. We appear to
have brought about a doubling of frequency of the sound by halving the sampling rate. We have to be careful when
sampling a waveform to sample at a sufficiently high rate to avoid creating frequencies which don’t exist in the
waveform, i.e. spurious frequencies. If we get spurious frequencies we have produced a situation called aliasing.

#Spyder (Python 3.4) script

import numpy as np
import scipy.io.wavfile
samplingRate, soundSamples = scipy.io.wavfile.read('Tone.wav')
soundSamplesNew = []
for i in range(len(soundSamples)):
if (i % 2) == 0:
soundSamplesNew.append(soundSamples [i])
#Convert from soundSamplesNew list to array
soundSamplesNew = np.asarray(soundSamplesNew)

scipy.io.wavfile.write('ToneFreqDoubled.wav', samplingRate, soundSamplesNew)

Figure 5.6.1.19 Spyder Python 3.4 script to double frequencies of digitally recorded sounds in a WAV file.
Play Tone.wav and ToneFreqDoubled.wav in a media player such as Windows Media Player and note the
difference in frequency.

Questions
6 Writing every other sample is one way of doubling frequencies of digitally recorded sound. Can you think
of another way that this could be done without having to omit sampled values and which could alter
frequencies by factors other than 2?

119 Single licence - Abingdon School

 5.6.1(3) Bit patterns, images, sound and other data

Programming tasks
8 Using JES, Spyder Python 3.4 or another programming/scripting system that supports exploration of
digitally recorded sound, write a program/script/commands to double frequencies of digitally recorded
sounds in WAV files.

Test your results in a media player.

Sound and text files

The numbers representing samples of digitised
sound may be read from a WAV file, converted
to their string equivalent and then written to a
text file, one sample per line (text files are strings
of characters organised on a line-by-line basis).
The text file may now be opened in a spreadsheet
and the numbers displayed on a chart as shown in
Figure 5.6.1.20.
A Python script that creates the text file equivalent
of a sound file, Tone.wav, is shown in Figure
5.6.1.21.

Figure 5.6.1.20 Excel spreadsheet that displays and charts

Tone.txt

#Spyder (Python 3.4) script

import scipy.io.wavfile
samplingRate, soundSamples = scipy.io.wavfile.read('Tone.wav')
bitStream = open('Tone.txt', "wt") # open file in write text mode
for i in range(len(soundSamples)):
# str converts number to string representation, \n add end of line
bitStream.write(str(soundSamples[i]) + "\n")
bitStream.close()

Figure 5.6.1.21 Spyder Python 3.4 script to transfer sound samples to a text file

Likewise, it is possible to convert a text file into a sound file. Using Tone.txt for convenience, the Spyder Python
3.4 script shown in Figure 5.6.1.22 creates a WAV file, TextToSound.wav, of digitised sound samples. It sets
the sampling rate to 20000 samples per second but this can be changed easily to change the frequency of the tone
represented by this file and it sets the number of bits per sample to be 16.

Single licence - Abingdon School 120

5 Fundamentals of data representation

#Spyder (Python 3.4) script

import numpy as np
import scipy.io.wavfile
bitStream = open('Tone.txt', "rt")
contents = bitStream.readlines()
bitStream.close()
fileIndex = 0
soundData = []
samplingRate = 20000
while (fileIndex < len(contents)):
sample = int(contents[fileIndex].replace("\n", ""))
soundData.append(sample)
fileIndex = fileIndex + 1
soundData = np.asarray(soundData, dtype='int16')
scipy.io.wavfile.write('TextToSound.wav', samplingRate, soundData)

Figure 5.6.1.22 Spyder Python 3.4 script to create a sound file from a text file

The Matlab command audioinfo can be used as shown in Figure 5.6.1.23 to obtain the metadata stored in file
TextToSound.wav.

Information
>> info = audioinfo('TextToSound.wav')
audioinfo: info = Filename: 'TextToSound.wav'
The audioinfo command is not
CompressionMethod: 'Uncompressed'
yet implemented in Octave
NumChannels: 1
SampleRate: 20000
Principle TotalSamples: 20000
Duration: 1
Text, digitised sound and
images: Title: []
Text, digitised sound and Comment: []
images are all just bits or bit
Artist: []
patterns under the hood. As
such they can be mapped BitsPerSample: 16
between each other by
transforming the way that the
bit patterns are arranged and Figure 5.6.1.23 MatLab command line >>info =
interpreted. audioinfo('TextToSound.wav')

121 Single licence - Abingdon School

 5.6.1(3) Bit patterns, images, sound and other data

Programming tasks
9 Using JES, Spyder Python 3.4 or another programming/scripting system that supports exploration of
digitally recorded sound and text files, write a program/script/commands to convert WAV files to text files
and vice versa.

Test your results in a media player.

Questions
7 It has been demonstrated that it is possible to transform sound and image files to text files and back again.
Give three reasons why this is useful.

In this chapter you have covered:

■■ How bit patterns may represent sound

Single licence - Abingdon School 122

 5 Fundamentals of data representation
5.6 Representing images, sound and other data
Learning objectives:
■■ Understand the difference
between analogue and digital
■■ 5.6.2 Analogue and digital
What is data?
• data
Recording your body weight over time, say six months, would generate a set of
• signals values of a quantitative and discrete nature. Discrete because the values are not
recorded continuously but sampled at intervals of time. The recorded values
are known individually by the term datum and collectively as data. The data
is quantitative in nature because it is obtained by measurements performed by
some measuring instrument calibrated by reference to some continuous scale of
values.
Data may also be qualitative and discrete. For example, recording name and
eye colour of every individual in a class of students, e.g. “John Smith, blue”,
“Carol Jennings, green”, produces a set of values or value-pairs of a qualitative
nature. The recorded values or value-pairs are also known collectively as data
and a single value or value-pair as a datum. The data is qualitative because it
is descriptive in nature and constitutes a characteristic, e.g. eye colour or a
property, e.g. a person has a name rather than a measurement.
Key concept Vacuum Scale indicates
What is analogue data? air pressure in
Analogue data: mm of mercury
Data that varies in a continuous Air temperature and air pressure vary Air pressure
pushes down
manner or is recorded in a in a continuous manner. For example,
on mercury
continuous form and that is if you were to climb a mountain you forcing it to rise
similar to its original structure. up tube
would find that as you rose in height
760mm
the air pressure would lessen in a
continuous manner as the total amount
of air pressing down on you from above
became less – see Figure 5.6.2.1.
The relationship between air pressure Mercury bath
and height above sea level is shown
Figure 5.6.2.1 Toriccelli
in Figure 5.6.2.2. This variation in
barometer
pressure could have been observed with
a Torricellian barometer carried up the
mountain. The height of the column of mercury, the data, would have been
observed to vary in a continuous manner. Data that varies in a continuous
manner is known as analogue data. The barometer is a source of analogue data.

Single licence - Abingdon School 123

5 Fundamentals of data representation

Elevation and Atmospheric Pressure

140
Atmospheric pressure (kPa)

120
100
80
60
40
20
0

Elevation above sea level (m)

Figure 5.6.2.2 Relationship between air pressure and height above sea level (Adapted from
www.engineeringtoolbox.com/air-altitude-pressure-d462.html with kind permission of the editor)

Information that is recorded in a continuous form and that is similar to its
Key concept
source’s original structure is also analogue data. The phonograph invented in
Discrete data: 1877 by Thomas Edison, known today as a record player, recorded speech
Information represented by directly onto wax cylinders by making physical deviations of a groove,
separate values is discrete. We
impressed into the wax, a replica of the variation in air pressure caused by the
say that these values are discrete
data. speech.

The pattern of variation recorded on the wax cylinder is an example of analogue

data because it varies in a continuous fashion and is similar in form to that
which caused it, the variation in air pressure caused by the spoken word. The
modern equivalent of the wax cylinder is the vinyl LP.

What is discrete data?

Information represented by separate values (quantities), e.g. words in a list, is
“discrete”. Here are three sets of discrete quantities:
Hour Temperature Hour Temperature
■■ 1, 2, 3, 4 (set 1)
1 8 7 13
2 7 8 14 ■■ 0, 1, 0, 1, 1, 0 (set 2)
3 6 9 16 ■■ A, B, C, D (set 3)
4 8 10 16
When analogue data are sampled and their values
5 10 11 17
recorded, in the appropriate units, they become
6 10 12 16
discrete data. The decimal number 45 is discrete
Table 5.6.2.1 Discrete temperature data (sampled from
because it belongs to a set of discrete numbers,
analogue data)

124 Single licence - Abingdon School

 5.6.2 Analogue and digital

the set of all positive integers. Table 5.6.2.1 shows discrete data in the form of Key concept
temperature readings taken at hourly intervals.
Digital data:
What is digital data?
Digital data is discrete data
To store data digitally in a computer, it has first to be represented in discrete form, which has been encoded in
and then converted (encoded) to digital (binary) values. digital form, i.e. binary, using
some algorithm.
Figure 5.6.2.3 shows discrete data being encoded in binary by a process which
Since discrete information is
represents each discrete datum by a specific binary value, e.g. 4.7010 and 4.9310 conveyed by the sequence in
are both represented by 1002. This digitising process introduces errors called which the encoding symbols
quantisation errors, e.g. 4.7010 is represented by 1002 which is 410. are ordered, there must be
some way of determining the
beginning of a sequence. This
Quantisation is known as synchronisation.
Synchronisation is a property of
Discrete data digital data that distinguishes it
111 from analogue data. Machine
6.60 110 communications typically
use special synchronization
5.65 5.38 5.44 101 sequences to enable machines
4.70 4.93 100 to extract discrete information
Digital data
3.40 011 represented by digital data.

2.10 2.50 2.80 010

1.00 1.17 001
0.25 0.55 0.75 000

Figure 5.6.2.3 Digitising discrete data by encoding the data in 3 bits

Key concept
What is a signal? Signal:
A signal is that which conveys
Many countries around the world have used beacons, i.e. bonfires, strategically a message or information from
sited, to warn of or signal danger. Some animals use sound for a similar purpose. one place to another.
Internally, the human body uses both electrical and chemical means to convey
signals some of which are in response to danger, e.g. to cause an adrenaline response to a threatening situation.

Signals are used for all sorts of purposes. Essentially a signal is that which conveys a message or information from
one place to another. As such, signals are subject to the laws of Physics, in particular Einstein’s special theory of
relativity that states that signals or the information that they carry cannot travel faster as a group than the speed of
light which is 3 × 108 metres per second in a vacuum.

The information carried by a signal is in the form of energy that can activate a detector or sensor in a receiver
of the signal. For example, the light from a warning beacon is conveyed as photons or light particles, each of
which carries a certain amount of electromagnetic energy, enough to stimulate cells in the retina of the eyes of
the receiver. This stimulation of the retina results in an electrical signal to the receiver’s brain which responds
accordingly.

Single licence - Abingdon School 125

5 Fundamentals of data representation

What are analogue signals?

Key concept In order to process analogue data it must be sensed and then converted into an
Analogue signal: equivalent electrical form. The electrical equivalent for this purpose is called an
In telecommunications and analogue electrical signal or just analogue signal. In telecommunications and
computer engineering, an
computer engineering, an analogue signal is an electrical or electromagnetic
analogue signal is an electrical
signal that varies in a continuous manner. The conversion process takes place
or electromagnetic signal that
varies in a continuous manner. in a device known as a transducer. A transducer is designed to convert energy
from one form to another. A microphone is an example of a transducer.
It converts continuously varying sound pressure waves into an equivalent
continuously varying electrical signal. Another example of a transducer is a
loudspeaker. A loudspeaker converts electrical energy into sound energy.
Figure 5.6.2.4 shows an electrical circuit for converting sound energy into
electrical energy. Figure 5.6.2.5 shows the variation in pressure produced
by the speaker whistling a pure tone. Figure 5.6.2.6 shows the equivalent
analogue signal. Fluctuating current, I

Analogue
0 voltmeter
R showing a
fluctuating
Microphone voltage,
Sound waves Battery V = IR

Figure 5.6.2.4 Electrical circuit for converting sound energy into electrical energy

Information
Speech:
Pressure/Pa

When a person speaks, they

emit a continuous stream of 0
sound, essentially - the final 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 Time/milliseconds
syllable of one word prefixes
the starting syllable of the
next. However, the words
spoken are nevertheless
semantically discrete, Figure 5.6.2.5 Variation in pressure produced by speaker whistling a pure tone
and can be written down
accordingly. The “raw” data Analogue
is arguably the (continuous) voltmeter
sound. The information readings
Voltage/volts

it carries is discrete – the

words and their meaning. 0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 Time/milliseconds

Figure 5.6.2.6 Equivalent analogue electrical signal

126 Single licence - Abingdon School

 5.6.2 Analogue and digital

What are digital signals? Key concept

In contrast to analogue signals, a digital signal is a signal that represents a
sequence of discrete values. It may be considered to be a sequence of codes Digital signal:
A digital signal is a signal
represented by a physical quantity such as an alternating current or voltage, the
that represents a sequence
signal strength of a radio signal, the light intensity of an optical signal, etc. of discrete values. It may be
Figure 5.6.2.7 shows a digital signal with 7 distinguishable voltage levels. In this considered to be a sequence of
codes represented by a physical
example, the voltage levels available to the signal were -7.5, -5 , -2.5, 0, 2.5, 5, 7.5.
quantity such as an alternating
Each voltage level encodes a binary datum (single item of binary data) as shown current or voltage or the signal
in Table 5.6.2.2. The most significant binary digit is a sign bit with 0 representing strength of a radio signal or
the light intensity of an optical
+ and 1 representing −. Unfortunately this leads to two binary patterns
signal.
representing zero.

7.5 Voltage level Binary

5
+7.5 011
+5 010
-2.5
+2.5 001
Voltage/volts

0 0 000
1 2 3 4 5 6 7 8 9 10 11 12 Time/milliseconds 0 100
-2.5
−2.5 101
-5
−5 110
-7.5 −7.5 111
Figure 5.6.2.7 Digital signal
Table 5.6.2.2 Using digital
signals to encode binary data

It is possible to use just two distinguishable levels of voltage, 0 volts and 5 volts as shown in Figure 5.6.2.8. The
digital signal is then a binary digital signal. The shape in Figure 5.6.2.8 is called a voltage pulse.
Each voltage represents a binary datum, binary datum 1 by +5 volts and binary datum 0 by 0 volts as shown in
Table 5.6.2.3.
Voltage level Binary
5 1
0 0
Table 5.6.2.3 Using binary digital signals to encode binary data

The stream of voltage pulses shown in Figure 5.6.2.8 encodes the binary data 0110101010 (the least significant
digit is the first pulse to be produced).
Voltage/volts

0
0 1 2 3 4 5 6 7 8 9 10 Time/milliseconds

Figure 5.6.2.8 Using binary digital signals to encode binary data

Single licence - Abingdon School 127

 5.6.2 Analogue and digital

Questions
1 What is analogue data?

2 What is digital data? Give an example.

3 What is a signal?

4 Differentiate between analogue and digital signals.

In this chapter you have covered:

■■ The difference between
• analogue data: data that varies in a continuous manner or is recorded in
a continuous form and that is similar to its original structure
and
• digital data: discrete data which has been encoded in digital form, i.e.
binary, using some algorithm
■■ The difference between
• analogue signals: in telecommunications and computer engineering, an
analogue signal is an electrical or electromagnetic signal that varies in a
continuous manner
and
• digital signals: a digital signal is an electrical signal which conveys
information represented by digital data, i.e. it is a signal that represents
a sequence of discrete values. It may be considered to be a sequence of
codes represented by a physical quantity such as an alternating current
or voltage or the signal strength of a radio signal or the light intensity
of an optical signal. The digital signal can also change voltage level or
amplitude in an abrupt manner or in abrupt steps.

Single licence - Abingdon School 128

5 Fundamentals of data representation
5.6 Representing images, sound and other data
Learning objectives:
■■Describe the principles of
operation of:
■■ 5.6.3 Analogue/digital conversion
Analogue to digital converter (ADC)
• An analogue to digital Using a transducer to generate an analogue signal
converter (ADC)
Sound waves travel through air causing vibrations in your ear that you
• A digital to analogue perceive as sound. Sound waves are classified as analogue data because they
converter (DAC) vary continuously in shape and size. Sound waves may be converted into an
equivalent analogue electrical current or voltage using a microphone which
■■Know that ADCs are used
is an example of a transducer, a device for converting energy from one form
with analogue sensors
to another. The variation in frequency (pitch) and amplitude (loudness) of
■■Know that the most common the sound is converted to an equivalent electrical form in the microphone to
use for a DAC is to convert produce an analogue signal.
a digital audio signal to an
Converting to digital form
analogue signal
An analogue signal representing a sound may be recorded by converting it
with an analogue to digital converter (ADC) into a digital signal suitable for
transmitting and storing in a digital computer system. Figure 5.6.3.1 shows an
Key principle analogue signal plotted on a voltage-time graph.
Analogue to digital converter
(ADC):
Converts an analogue signal
into an equivalent digital signal.
Voltage/volts

Key principle 0
Time/milliseconds
Pulse Amplitude
Modulation(PAM):
Pulse Amplitude Modulation
is a process of measurement
of the amplitude (height) of
an analogue signal at fixed
and regular intervals of time Figure 5.6.3.1 Analogue signal plotted on a voltage-time graph
determined by the sampling
The analogue to digital conversion process consists of several stages:
frequency. The process outputs
a series of pulses whose 1. The analogue signal is sampled at fixed and regular intervals of time
amplitudes correspond to using sample and hold circuitry – see Figure 5.6.3.2 - to produce an
these measurements and whose
equivalent digital signal as shown in Figure 5.6.3.3.
duration in time is the time
elapsed between one sampling This form of digital signal is known as a Pulse Amplitude Modulation
and the next (the sampling (PAM) signal.
interval).

Single licence - Abingdon School 129

5 Fundamentals of data representation

2. The size or amplitude of each sample is measured and coded in binary

in a given number of bits, e.g. 4 bits, as shown in Figure 5.6.3.4.
3. The binary form of the measurements is represented by electric pulses
Key principle suitable for transmission over a bus system, serial or parallel, connected
Pulse Code Modulation to the ADC. This form of the digital signal is known as a Pulse Code
(PCM): Modulation (PCM) signal.
Pulse Code Modulation is a
process for coding sampled Red lines indicate when
analogue signals by recording amplitude is sampled
the amplitude of each sample in
a binary electrical equivalent.
Voltage/volts

0
Time/milliseconds

Figure 5.6.3.2 Analogue signal sampled at fixed and regular intervals of time
Figure 5.6.3.2 shows this analogue signal sampled at fixed and regular time
Questions intervals. A sample is a single measurement of amplitude. The number
of measurements of amplitude per second is known as the sampling rate.
1 What is a sample?
Sampling rate is expressed as number of samples per second, e.g. 1000
samples per second. Sampling rate is also called sampling frequency. Sampling
2 What is the sampling rate
frequency is expressed in Hz, e.g. 1000 Hz is the equivalent of 1000 samples
for the following sampling
per second, and 1 KHz, which is the equivalent of 1000 samples per second.
frequencies
Figure 5.6.3.3 shows the digital signal produced from the sampled analogue
(a) 20000 Hz (b) 40 kHz
signal by a circuit that holds the current sampled value steady until the next
(c) 44.1 kHz?
sampled value is obtained.
Equivalent digital signal
7

4.5
Voltage/volts

0
Time/milliseconds
-2.4

Figure 5.6.3.3 Digital signal produced from the sampled analogue signal.

130 Single licence - Abingdon School

 5.6.3 Analogue/digital conversion

A 4-bit ADC is helpful in explaining how the measurements of voltage are converted to binary but not very useful
in practice; commercially available ADCs use a higher number of bits, e.g. 8, 10, 12, 16.
If we are dealing with a bipolar signal, i.e. one where the voltage may be positive or negative then the ADC must
be set to work across a voltage range that includes both positive and negative values. For the conversion shown in
Figure 5.6.3.4 the range is set from −8.5 to +7.5 volts i.e. 16 volts.

4-bit ADC
Voltage range ± 8 volts
No of bit patterns or levels = 16
Resolution = Voltage range = 1 volt
No of levels Amplitudes here all
+7.0 0111 recorded as 0100
+6.0 0110
+5.0 0101
+4.0 0100
+3.0 0011
+2.0 0010
+1.0 0001
+0.0 0000 Questions
-1.0 1111
-2.0 1110 3 Describe the stages of
-3.0 1101 the analogue to digital
-4.0 1100
-5.0 1011 conversion process.
-6.0 1010 Amplitudes here all
-7.0 1001 recorded as 1110
-8.0 1000

Figure 5.6.3.4 Levels for a 4-bit ADC and voltage range -8.5 to + 7.5 volts
coded in 4-bit two’s complement binary

Encoding samples using 4 bits gives 16 different bit patterns from 0000 to 1111. To cover both positive and negative
values of voltage, these bit patterns are interpreted as representing two’s complement binary, so voltages in the range
−0.5 to +0.5 are coded as 0000, −0.5 to -1.5 volts are coded as 1111, +6.5 to +7.5 volts as 0111 and −7.5 volts to
−8.5 volts as 1000. Table 5.6.3.1 shows the correspondence between voltage and binary code.

Sample Binary Voltage Sample value Binary Voltage

value Two’s equivalent in volts Two’s complement equivalent
in volts Complement of code of code
-0.5 to +0.5 0000 0 -0.5 to -1.5 1111 -1.0
+0.5 to +1.5 0001 +1.0 -1.5 to -2.5 1110 -2.0
+1.5 to +2.5 0010 +2.0 -2.5 to -3.5 1101 -3.0
+2.5 to +3.5 0011 +3.0 -3.5 to -4.5 1100 -4.0
+3.5 to +4.5 0100 +4.0 -4.5 to -5.5 1011 -5.0
+4.5 to +5.5 0101 +5.0 -5.5 to -6.5 1010 -6.0
+5.5 to +6.5 0110 +6.0 -6.5 to -7.5 1001 -7.0
+6.5 to +7.5 0111 +7.0 -7.5 to -8.5 1000 -8.0

Table 5.6.3.1 4-bit ADC set to range -8.5 to +7.5 volts

Single licence - Abingdon School 131

5 Fundamentals of data representation

We can imagine that the ruler shown in Figure 5.6.3.5 has been used to measure the amplitude of the digital signal
shown in Figure 5.6.3.4, rounding up or down to a value from the set {−8.0, −7.0, …, +6.0, +7.0}. For example,
3.6 volts would be rounded up to 4.0 volts and coded as 0100, as would 3.5 volts. But 3.4 volts would be rounded
down to 3.0 volts and coded as 0011.

-8.0 -7.0 -6.0 -5.0 -4.0 -3.0 -2.0 -1.0 0.0 +1.0 +2.0 +3.0 +4.0 +5.0 +6.0 +7.0

1000 1001 1010 1011 1100 1101 1110 1111 0000 0001 0010 0011 0100 0101 0110 0111
Figure 5.6.3.5 Ruler for a 4-bit ADC
The ADC stores the binary code, e.g. 0110, for the current measurement of amplitude in an internal register before
transfer to the processor of the computer to which the ADC is connected. ADCs may be connected by serial (SIP
or I2C) or parallel interface depending on its design. Figure 5.6.3.6 shows the pulse code form of one 4 bit sample.
Binary code 1 is a 5 volts high pulse and binary code 0 is a 0 volts high pulse.

0 1 1 0
Figure 5.6.3.6 Pulse Code Modulation (PCM) form of one four bit sample.

Questions
4 Draw a ruler for a 3-bit ADC to measure an analogue signal that varies from −9.0 to +7.0 volts over a
range of 16 volts. The ruler uses two’s complement representation (HINT: see Figure 5.6.3.5).

5 What binary code would be used for a voltage of

(a) +2.0 volts (b) +2.9 volts (c) −2.9 volts (d) +5.0 volts?
6 Draw a ruler for a 3-bit ADC to measure an analogue signal that varies within the range 0.0 to 8.0 volts.
The ruler uses unsigned binary representation (HINT: see Figure 5.6.3.5).

Resolution
The purpose of an ADC is to output a PCM digital signal that represents measurements of the amplitude of
an analogue signal at fixed and regular intervals of time. The accuracy of the measurements are determined by
the number of bits that the ADC uses for its measurements, the more bits the greater the accuracy. Stating the
resolution of the ADC is one way of expressing this accuracy.
Resolution of an ADC is measured in terms of the number of bits per sample. The number of bits per sample is
referred to as the bit depth of the ADC or word length.

132 Single licence - Abingdon School

 5.6.3 Analogue/digital conversion

Resolution for a given analogue signal is defined in terms of the range of voltage measured and the number of levels
or bit patterns available as follows
Voltage range
Resolution =
No of levels
where No of levels = 2No of bits
Table 5.6.3.2 shows resolution for a voltage range 0 to +8 volts and various number of bits

No of bits No of Resolution
available to levels in volts
ADC 2No of bits
4 16 0.5
8 256 0.03125
12 4096 0.001953125
16 65536 0.0001220703125

Table 5.6.3.2 Resolution for a voltage range of 0 to +8 volts

Quantisation
The measurement process can be visualised using a ruler to measure the amplitude of an analogue signal to the
nearest binary code or corresponding voltage. Imagine that the range of voltage for the analogue signal is from 0 to
4 volts and the number of bits available to represent the measurement is 2 then the ruler would be marked as shown
in Figure 5.6.3.7 with 0.5 volts corresponding to 00, 1.5 volts to 01 and so on.

0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4

00 01 10 11
Figure 5.6.3.7 Ruler for measuring voltages in range 0 to 4 volts in 2 bits.
A voltage measurement that lies between 0 and 1 volts would be therefore be coded as 00, a voltage measurement
between 1 and 2 volts as 01, and so on. The resolution is 4 volts = 1 volt .
4
However, given, for example, 01 as the coded measurement, we can only say that Key concept
the analogue signal’s amplitude at the time of measurement was in the range 1 to 2 Quantisation error:
volts or 1.5 ± 0.5 volts. The error in measurement
introduced by an ADC because
If the actual amplitude was 1.7 volts then the measurement would be rounded
of rounding down or up when
down to 1.5 volts and coded as 01. measuring the amplitude of an
If the actual amplitude was 1.2 volts then the measurement would be rounded up analogue signal.

to 1.5 volts and coded as 01.

Therefore, the ADC can introduce errors when it converts an analogue signal to a PCM signal.
The error in measurement introduced by an ADC because of rounding down or up is known as quantisation error.
The process of rounding up or down is called quantisation. It results in a distorted recording of the true shape of
the original analogue signal. This is known as quantisation distortion.
The maximum possible error because of rounding down or up is, in this example, ± 0.5 volts. This is known as the
maximum quantisation error.

Single licence - Abingdon School 133

5 Fundamentals of data representation

The effects of quantisation errors produced by an ADC are most apparent when the number of bits is small. The
greater the number of bits the smaller the effects of quantisation error. Unfortunately, more bits means the quantity
of digital data is greater and therefore file sizes that store this data are greater too. Quantity of data is calculated as
follows
Quantity of data = No of bits per sample × Sample rate × Length in time of analogue signal
Music CDs are PCM recordings of analogue signals sampled at 44,100 samples per second using ADCs with a
resolution of 16 bits. So three minutes of mono sound would occupy 15.876 MB of storage according to Table
5.6.3.3. If two channels are used then three minutes of stereo sound would occupy 31.752 MB of storage.

No of bits Sample rate Length in time of Quantity of data

per sample (samples per analogue signal (megabytes)
second) (seconds)
8 40000 60 2.4
16 40000 60 4.8
16 44100 180 15.876

Table 5.6.3.3 Quantity of data for various no of bits per sample and sample rates

Questions
7 (a) Draw a ruler for a 4-bit ADC to measure an analogue signal that varies in the voltage range 0 to 4
volts.
(b) What is the resolution of this ADC?
(c) What is the resolution in volts for the measurements of this analogue signal?
(d) What is the maximum quantisation error?

8 An ADC with a resolution of 10 bits is used to digitize an analogue signal of duration 180 seconds using
a sampling rate of 40000 samples per second. How many bytes will the ADC’s PCM produce?
9 A CD-ROM has a capacity of 737 MB (1MB = 1000000 bytes). How many 3 minute two-channel stereo
music recordings can be stored on this CD-ROM if the recordings were made in PCM from an ADC
with a resolution of 16 bits using a sampling rate of 44,100 samples per second per channel?

Digital to analogue converter (DAC)

To turn a PCM signal back into an analogue signal requires the use of a digital to Key principle
analogue converter (DAC). The DAC produces an analogue signal which is an
Digital to analogue converter:
approximation of the original analogue signal as illustrated in Figure 5.6.3.9. The Converts a digital signal into an
PCM signal is first turned into a PAM signal - Figure 5.6.3.8. The staircase effect analogue signal approximately
is a result of the approximation at the PCM quantisation stage of the analogue to equivalent to the original
digital conversion of the original analogue signal. The deviation from the original analogue signal from which the
digital signal is derived.
is known as quantisation noise. The DAC applies smoothing to the PAM signal
before it is output as shown in Figure 5.6.3.9.

134 Single licence - Abingdon School

 5.6.3 Analogue/digital conversion

DAC PAM signal before smoothing applied Key concept

Quantisation noise:
The deviation in the DAC –
produced analogue signal from
the original analogue signal.
Voltage/volts

0
Time/milliseconds

Original analogue signal

Figure 5.6.3.8 DAC reconstructed analogue signal and the original analogue signal

Questions
10 What is the purpose
of a digital to analogue
converter?
Voltage/volts

0
Time/milliseconds

Figure 5.6.3.9 Output of DAC after smoothing applied Key concept

ADCs and analogue sensors Sensor:

A sensor is a device that
What is a sensor?
measures something of interest
A sensor is a device that measures something of interest using a variety of in the physical world and using
mechanisms. A sensor is usually integrated with a transducer which converts the a transducer converts what
output of the sensing into a signal as shown in Figure 5.6.3.10. This conversion is sensed into an equivalent
electrical signal.
process is known as transduction. Sensors play a key role in connecting the
physical world (temperature, light level, pressure, moisture, concentration levels of
gases such as CO2) with the digital world.

Physical quantities
Chemical
Biological
Sensor

Temperature Measurement Output

Light Electrical
Sound signal
Transduction
Motion

Figure 5.6.3.10 The sensing process

Single licence - Abingdon School 135

5 Fundamentals of data representation

Analogue sensors
The output signal from the majority of sensors is analogue so the signal must first be converted into a digital
signal before it can be passed to a digital computer system for recording and further processing.

Key concept To convert the analogue signal from an analogue sensor, the signal is fed to an
analogue to digital converter (ADC). The output of the ADC is a PCM
Analogue sensor:
signal (digital) suitable for transmission to a digital computer system.
A sensor whose output is an
analogue signal. Transmitting the PCM signal to a digital computer system is usual done through
a serial interface such as a UART or I2C or SPI (see Chapter 9.1.1).
The analogue signal may also need to undergo some conditioning before being applied to the ADC.
This signal conditioning takes the form of filtering
• to remove unwanted frequency components
• signal conversion to ensure its voltage range is correct for the ADC
• signal isolation for safety reasons in healthcare applications where there may be direct contact between a
patient’s body and the sensor.
The need to perform sense-transduce-signal condition-signal convert-output PCM onto a serial bus with analogue
sensors has led to the development of integrated circuits called MEMS that do all this.

Information
MEMS
MEMS stands for microelectromechanical systems. They consist of mechanical microstructures, microsensors,
microactuators, and microelectronics, all integrated onto the same silicon chip. Figure 5.6.3.11 shows a schematic
for a MEMS integrated circuit digital gyroscope and an actual MEMS 3-axis gyroscope that can be connected to a
Raspberry Pi.

Directions of + Ωz Yaw
+ Ωy
detectable Yaw
angular rates Ω
Roll
D OS
G

IG CO
YR

IT P
AL E

+ Ωx
Pitch

Figure 5.6.3.11 MEMS 3-axis gyroscope

MEMS are known as smart sensors because they incorporate into a single integrated package or chip,
• sensing + transduction with an analogue signal conditioning interface circuit
• an integrated analogue-to-digital converter (ADC)
• a microcontroller and an I/O bus to provide serial output to other computer systems.
Figure 5.6.3.12 shows a simplified block diagram of a smart sensor on a chip.
136 Single licence - Abingdon School
 5.6.3 Analogue/digital conversion

Figure 6.5.3.12 Single integrated circuit smart sensor

MEMS applications
MEMS can be found in smartphones, tablets, game console controllers, digital cameras and camcorders as well as
healthcare devices such as pacemakers. Two of the most important and widely used forms are accelerometers and
gyroscopes.
Smartphones often have embedded within them a range of analogue smart sensors such as accelerometers,
gyroscopes, magnetometers, pressure sensors, optical sensors, silicon microphones, etc.
Sensor platforms
Sensor platforms are a subset of smart sensors. Like smart sensors they feature a microcontroller, a wired/wireless
interface, and memory. However, sensor platforms are designed for non-specific platforms, i.e. not just dedicated to
generating a PCM signal from an analogue sensor such as a gyroscope. Sensor platforms can provide their services
to a range of sensors that may be optionally connected to them by direct wiring, Wi-Fi or Bluetooth. Examples are
the Arduino, smartphones, and the electric imp.
Converting digital audio signals to analogue using a DAC
Much of today’s music is available in digital format (digital audio) as are radio broadcasts and sound tracks
accompanying video. The digital format is not suitable for direct replay through loudspeakers, it would sound like
a morse code transmission, so the digital signal from a digital recording or a digital broadcast must be converted by
a DAC into an analogue signal that approximates closely the original audio. The loudspeakers convert the electrical
energy in the DAC-produced analogue signal into sound energy. If the quantisation noise is low then the quality
of the sound produced in the loudspeakers will be high, reproducing faithfully the original analogue sound from
which the digital form was created. Figure 6.5.3.13 shows a schematic for a typical sound card.
Sound Card
10011010

Sound
waves Parallel
Amplifier DAC to Serial
converter

Loudspeaker

Figure 6.5.3.13 Use of a DAC in a sound card

Single licence - Abingdon School 137

5 Fundamentals of data representation

Questions
11 What is the most common use for a Digital to Analogue Converter (DAC)?

12 What is a sensor? Why is an Analogue to Digital Converter (ADC) often required before the signal from a
sensor can be processed by a digital computer?

13 Name three analogue sensors found in smartphones.

14 Name four components of a single integrated circuit smart sensor.

15 Why is a Digital to Analogue Converter (DAC) needed in order to play digitally recorded sound?

16 With the aid of diagrams, describe the process of converting a PCM signal into its equivalent analogue
signal.

17 An audio signal from a microphone was converted into a PCM signal using an 8-bit ADC and replayed
through a loudspeaker via a sound card employing an 8-bit DAC. A listener complained that the quality
of the reproduced sound was inferior to a PCM signal generated from the same audio signal using a 16-bit
ADC and replayed through the same loudspeaker via a sound card employing a 16-bit DAC. Explain why
the quality of the reproduced sound could have been perceived as different for the two systems.

In this chapter you have covered:

■■ The principles of operation of:
• An analogue to digital converter (ADC)
ŠŠ Sample analogue signal
ŠŠ Measure amplitude of sample
ŠŠ Encode amplitude in binary to produce PCM signal
• A digital to analogue converter (DAC)
ŠŠ Convert PCM signal into a PAM signal
ŠŠ Smooth PAM signal to produce analogue signal
■■ ADCs are used with analogue sensors
• Analogue sensors produce analogue signals which must be converted into digital form to be stored and
processed by a digital computer. The conversion is performed by an ADC.
■■ The most common use for a DAC is to convert a digital audio signal into an analogue signal
• Digital audio signals are not suitable for direct replay through loudspeakers. This form of signal would sound
through a loudspeaker like morse code. Therefore, a DAC is required to convert the digital audio signal
into an analogue signal that approximates closely the original audio. The output of the DAC when played
through a loudspeaker should then resemble the sound of the original audio signal.

138 Single licence - Abingdon School

 5 Fundamentals of data representation
5.6 Representing images, sound and other data
Learning objectives:
■ Explain how bitmaps are
represented
■ 5.6.4 Bitmapped graphics
■ Explain the following for Image sensing and acquisition
bitmaps If an object is illuminated by a source of Light
Energy
light it will reflect that light to varying
• resolution Power
degrees, reflecting some colours more than in

• colour depth others. If the reflected light is captured in,

• size in pixels say, a digital camera then the energy in the Voltage
out
light is converted by light-sensitive sensors
■ Calculate storage requirements (photosensors) into an analogue electrical Array of
for bitmapped images voltage as shown in Figure 5.6.4.1. photosensors

■ Be aware that bitmap image This analogue electrical voltage must then be
files may also contain digitised to produce digital output.
metadata
In a digital camera, many such photosensors Figure 5.6.4.1 Sensor and its
■ Be familiar with typical array of photosensors
are arranged as shown in Figure 5.6.4.1. The
metadata
whole array is just called a sensor.
Did you know? Sampling and quantisation
Digital single-lens reflex
When, for example, a digital camera takes a picture of an object such as shown
cameras: in Figure 5.6.4.2, light from the object is projected through the imaging
These use an aspect ratio of system onto an array of light sensitive sensors (photosensors).
3:2. Aspect ratio is the ratio
of the width of the image to
its height. The Canon EOS
600D (released February 2011) Real world object
Imaging system
uses an APS-C CMOS sensor
consisting of a sensor array of
dimensions 5184 x 3456.

Output (digitised image)

Tasks Figure 5.6.4.2 Digital imaging

1 Digital cameras currently system
use either a CCD or a (Internal) image plane
CMOS light sensor array. Light source
How does each work and
why are two types used? The intensity of the image is sampled in the photosensor array at specific X-
and Y-coordinate positions. Each photosensor produces a voltage proportional
2 Find out the dimensions of
the array of photosensors to the intensity of the light falling on it. Making intensity measurements at
for a digital camera that you specific X- and Y- coordinates positions is called sampling. Digitising the
have access to (it will be analogue voltages representing intensity of light is called quantisation.
specified in pixels).

Single licence - Abingdon School 139

5 Fundamentals of data representation

Pixel
Key concept
Figure 5.6.4.3 shows the result of sampling the image in Figure 5.6.4.2 at
Pixel: discrete coordinate positions ranging in the X-direction from 0 to 27 and the
A pixel is the smallest Y-direction from 0 to 19, and digitising the analogue voltage representing the
addressable region or element of
intensity of the light of the primary colours Red(R), Green(G) and Blue(B) in
a digital image. Each pixel is a
sample of the original image.
this light, e.g. at X = 18, Y = 12.

X
0 27
0

Key concept
Pixel:
A pixel is also the smallest Y
controllable element of a digital
image represented on the screen Pixel
i.e. the smallest element or at
X = 18, Y = 12
region of a digital image that Colour = A8231616
can be changed or edited when
19
editing bitmapped images
using software such as iPhoto, Pixel
or Picture element
Photoshop, Windows Paint, and
at position X = 3, Y = 14
other “paint” style packages. Colour = FFFFFF16

Figure 5.6.4.3 Result of sampling the image plane and digitising

the intensity of light of the primary colours for each sample.
Key fact Each coordinate position in this discrete coordinate space is known as a pixel
Pixel-based graphics: or picture element. It is the smallest addressable region of the image plane that
Pixel-based graphics are made can be sampled and the intensity of light falling on it quantised.
up of small individual pieces
At the position (18, 12),
of the whole, and each can be
changed via editing. ■ the red component has an 8-bit value representing its intensity of
Its strength is in creating A816 or 16810 (measured on a scale that ranges from 0 to 255)
complex patterns and displaying
photographs with many colour ■ the green component has an intensity of 2316 or 3510
changes. ■ the blue component has an intensity of 1616 or 2210

Its weakness is in changing size. At position X = 3 and Y = 14, the corresponding red, green and blue intensities
Pixel images can be reduced in are each represented by FF16 or 25510, the maximum value.
size, but lose quality when they
Figure 5.6.4.4 shows a smiley face drawn in pixel mode using Photoshop. The
are increased in size.
pixels for the eyes, nose and mouth were drawn individually by selecting the
relevant pixel and changing its colour.

140 Single licence - Abingdon School

 5.6.4 Bitmapped graphics

Figure 5.6.4.4 Smiley face drawn in Photoshop in pixel mode at 50 x 50 pixels

Questions
1 What is a pixel?

2 Explain sampling and quantisation in the context of a taking a picture

with a digital camera.

3 How many pixels make up a digitised image if the image plane is

sampled over the following X and Y coordinates:

(a) X from 0 to 719, Y from 0 to 479?

(b) X from 0 to 1919, Y from 0 to 1279?
(c) X from 0 to 5183, Y from 0 to 3455?

4 Express the result for 3(c) in megapixels by dividing your answer by

1,000,000.

Bitmapped image or bitmap Key concept

If we wish to store a digitised image, such as the one shown in Figure 5.6.4.2
Bitmapped image or bitmap:
then each quantised sample must be stored, i.e. pixel by pixel, by recording the A bitmapped image is a pixel-
bit pattern representing the digitised intensity of each pixel. based digital image.
The digitised image is mapped
Figure 5.6.4.5 shows a section of memory from locations 308 to 335 and the
to bits in memory representing
corresponding row of pixels that it maps to. Note that the white pixels are the intensity and colour of light
stored as FFFFFF and the red pixels as A82316, CA1719, F29476. of each pixel.

We say that the digitised image is mapped to bits in memory.

The stored bits in memory are a digital representation of the image or just a
bitmap. We say that the image has been bitmapped.
A bitmapped image is a pixel-based digital image.

Single licence - Abingdon School 141

5 Fundamentals of data representation

MEMORY

MAPPED TO

DIGITISED 308 FFFFFF

309 FFFFFF
SAMPLES 310 FFFFFF
311 FFFFFF
312 FFFFFF
313 FFFFFF
314 FFFFFF
315 FFFFFF
316 FFFFFF
317 FFFFFF
318 FFFFFF
319 A82316
320 A82316
321 CA1719
322 F29476
323 A82316
324 A82316
325 A82316
326 FFFFFF
327 FFFFFF
Figure 5.6.4.5 shows a section of memory from 328 FFFFFF
329 FFFFFF
locations 308 to 335 and the corresponding row of 330 FFFFFF
331 FFFFFF
pixels that they map to. 332 FFFFFF
333 FFFFFF
334 FFFFFF
335 FFFFFF

MEMORY
307
308
309
Questions
310
311
312 5 Figure 5.6.4.6 shows an image of a section of a chequered board
313
314 and a section of memory for storing the bitmap for this image.
315
316 The pixel size is defined as the size of a single square. Each
317
318 memory cell, e.g. cell 311, can store one byte. A bitmap is to be
319
320
321
created in the memory that will record the colour of each pixel as
322
any one of 256 different colours. White will be coded as 255,
Figure 5.6.4.6 black as 0 and the red used in the image as 125.

(a) Show how this image could be stored as a bitmap. Use the
given memory cells in your explanation.

(b) With each memory cell still representing one pixel the
memory is changed so that each cell can store 3 bytes. How
many different colours can be coded in one memory cell?
Express your answer as a power of 2.

142 Single licence - Abingdon School

 5.6.4 Bitmapped graphics

Bitmap size in pixels Did you know?

The output of the sampling and quantisation processes is a sequence of digital
values, one per sample, corresponding to each discrete coordinate position or Digital cameras, and scanners
(both film and paper type)
pixel. Image size is usually expressed as number of pixels in the X-direction
capture information in pixel
by number of pixels in the Y-direction, e.g. 28 × 20 in the example in Figure format.
5.6.4.3.

Tasks Key concept

Bitmap size in pixels:
3 What is the image size in pixels produced by the digital camera in a
Bitmap size = w x h
typical smartphone? Use number of columns (X) × number of rows where
(Y) notation. w = width of image in pixels
h = width of image in pixels

Colour depth or bit depth

Colour depth, also known as bit depth, is expressed as the number of bits used Key concept
to indicate the colour of a single pixel, e.g. 8 bits, in a digitised image.
Colour depth or bit depth:
When the voltage representing intensity of light is quantised, it is represented Colour depth, also known as
by an integer number chosen from some range beginning at zero, e.g. 0 to 255 bit depth, is expressed as the
in decimal. number of bits used to indicate
the colour of a single pixel, e.g.
A range of 0 to 255 can be represented in binary by 8 bits.
255 8 bits, in a digitised image.
If the range was instead 0 to 65535 then 16 bits would be
required to represent the colour of a single pixel.
192
The example shown in Figure 5.6.4.5 allocates 8 bits to Information
represent red intensity, 8 bits to green and 8 bits to blue, a
Did you know?
128 total of 24 bits. Figure 5.6.4.7 shows the intensity of red, The Hubble telescope used
coded in 8 bits, for some selected values. a CCD detector array size of
4096 x 4096 and a field of view
Each possible combination of quantised red, green and
64 of 160 x 160 arcsecs. This gave
blue intensities represents a different resultant colour. a pixel size of 160/4096 or
The number of different bit patterns that 24 bits can 0.04 arcsec, i.e. a pixel for every
0 represent is
1/90000 degrees of view.
224 = 16777216
Figure 5.6.4.7 1⁄90000 degrees

Therefore, the number of different intensities of colour that can be recorded

using 24 bits for each is 16777216.

Questions
This is approximately the angle
6 How many different intensities of colours can be represented for
subtended by a penny viewed
an individual pixel if the number of bits used for each quantised at 52 km, meaning it could
sample is be distinguished from another
(a) 12 (b) 15 (c) 18? penny immediately next to it.
Express your answer as a power of 2.

Single licence - Abingdon School 143

5 Fundamentals of data representation

Resolution
A popular convention is to describe the resolution of a bitmapped image as
Key principle
the number of pixel columns (width) by the number of pixel rows (height), for
Resolution of a bitmapped example, 3264 × 2448.
image:
Image resolution = width of image in pixels × height of image in pixels
Image resolution = w x h
where For a given dimension of image, say 1.5 inches by 1 inch, the more samples
w = width of image in pixels that are taken across the image the smaller the pixels and the greater the
h = width of image in pixels
recorded detail.
Two objects
that cannot be resolved by sensor Voltage
To resolve the detail in a barcode, for as two objects out
example, it must be possible to pick
Key principle out both white and black bands in the
image. If the photosensors are too big Sensor
Displayed or printed
then this detail will be missed. Imaging
resolution of a bitmapped system
Power
image: However, the pixel dimensions of a in
This is expressed as pixels per bitmapped image, such as 5184 × Figure 5.6.4.8 Sensor too
inch or ppi.
3456 pixels, doesn’t give the image a large to resolve two objects
The choice of ppi for a clear, physical size. The bitmapped image’s
sharp image depends on the dimensions in pixels state only how many pixels there are, not how big each is.
viewing distance. Display and print resolution
The size of each pixel is set by specifying how many pixels should be fitted into
an inch when the bitmapped image is displayed or printed.
This is expressed as pixels per inch or ppi.
Key fact The choice of ppi for a clear, sharp image depends on the viewing distance.
Scanned and digital camera At the viewing distance, it should not be possible to see individual pixels.
images:
For images produced by For example, a 4"× 6" standard photographic print, printed at 300 ppi and
scanning or by a digital camera, viewed at about 11 inches appears fine, whereas a billboard-sized photograph
the clarity and resolution of a also appears fine because its viewing distance is so much greater, even though it
captured image is determined
is only printed at about 15 ppi.
by the size of the photosensors
(one photosensor = one pixel). Figure 5.6.4.9 shows a bitmapped image originally prepared for printing at (a)
This can be expressed as ppi. 10 ppi, (b) 50 ppi and (c) 300 ppi.

The image in (a) of 40 × 30 pixels still reveals its pixels at normal viewing
distance. We say that the
image is pixelated because
the individual pixels are
visible. Image (b) is 200
× 150 pixels and shows
much less pixelation but
(a) (b) (c) the sharpest image is (c)
Figure 5.6.4.9 bitmapped image printed at (a) 10 ppi, (b) 50 ppi and (c) 300 ppi which has dimensions
1200 × 900 pixels.

144 Single licence - Abingdon School

 5.6.4 Bitmapped graphics

Physical dimensions of printed images Information

It should be clear that for practical purposes, the clarity of the image displayed
Spatial resolution:
or printed is decided by its spatial resolution, not the number of pixels in an
Spatial resolution is the capability
image.
of the sensor to observe or
In effect, when a digital image is displayed or printed, resolution refers to the measure the smallest object clearly
number of independent pixel values per unit length, e.g. pixels per inch or and distinctly, distinguishing it
from other objects that surround
ppi. This is also known as pixel density.
it.
This is an alternative meaning of digital image resolution or bitmap resolution. Figure 5.6.4.10
is composed
Bitmap resolution is measured in number of pixels per inch (ppi).
of two objects
This definition determines the size of the pixel of the display unit when which are
displaying digital images or the number of image pixels that will fit inside each straight lines
Figure 5.6.4.10
separated from
inch of paper when printed.
each other by a small amount of
Specifying a resolution gives a size to the pixels of the printed image. white space. At the distance at
which you are viewing this page,
Scanned and digital camera images
most people will be able to see
For images produced by scanning or by a digital camera, the clarity and these as two separate objects, i.e.
resolution of a captured image is determined by the size of the photosensors lines. The sensor in this case is
(one photosensor = one pixel). your eye.
However, if you gradually increase
Resolution of computer displays
the distance between your eye and
A bitmapped digital image produced by a digital camera is composed of the page, you will reach a distance
digitised or quantised samples that a computer screen displays as pixels because at which the two lines are seen
the screen of a computer display is divided as just one thicker line. You are
into pixels. 1920 pixels no longer able to resolve the two
lines spatially (i.e. in space). Thus
The pixels are the addressable units of the 20” spatial resolution depends upon
1080 pixels 11.25”
screen that are individually illuminated to object separation and viewing
create an image or text on the screen. Pixels distance, ignoring any deficiencies
of the sensor itself.
per inch (ppi) or pixels per centimetre
(ppcm) is a measure of pixel density and Figure 5.6.4.11 LCD monitor
Did you know?
therefore screen or display resolution. It with screen dimensions 20
is defined as the number of pixels in the inches by 11.25 inches, filled Retina displays:

horizontal direction per unit measurement, with 1920 x 1080 pixels. Retina displays use 326 ppi
When introducing the iPhone 4,
e.g. inch, or the number in the vertical
Steve Jobs said that the number of
direction per unit measurement which is pixels needed for a Retina Display
the same thing for square pixels. The ppi of a computer display is therefore is about 326 ppi for a device held
related to the size of the display in inches and the total number of pixels in the 10 to 12 inches from the eye. At a

horizontal or vertical directions. distance of 12 inches, the average

eye will not be able to resolve the
For the display in Figure 5.6.4.11, individual pixels of the screen
and therefore the display will be
Resolution = 1920 pixels/20 inches
acceptable for viewing.
= 1080 pixels/11.25 inches
= 96 ppi
PPI is a display resolution not an image resolution.

Single licence - Abingdon School 145

5 Fundamentals of data representation

Image resolution is measured in samples per inch or loosely in horizontal pixels Did you know?
× vertical pixels. Display screens used by desktop computers typically have a
Screen resolution:
resolution of 96 ppi or lower.
The iPhone 5s (released
For the LCD monitor in Figure 5.6.4.11, 96 ppi is the maximum pixel density. September 2012) has a screen
Display devices usually allow the display settings to be changed, e.g. for the given resolution of 326 ppi and a
display, choosing 1280 pixels by 720 pixels changes the pixel density or resolution screen size of 1136 by 640-pixel.
Sony’s Xperia Z’s screen has
to
1280 pixels/20 inches or 720/11.25 = 64 ppi a screen size of 1080 × 1920
pixels and a resolution or pixel
for the same screen real estate of 20 × 11.25 inches. density of approximately 441
ppi.
Figure 5.6.4.12 shows the same section of a digitised image at different screen
resolutions or ppi. The dimensions of the image in pixels is unchanged but the
image’s size in the display goes from small to large as the resolution of the screen is reduced. The displayed pixels
become larger as the screen resolution is lowered.

Figure 5.6.4.12 The same section of a digitised image at different screen resolutions, with the pixel density of
the screen or ppi decreasing from left to right.

Questions
7 Calculate the screen resolution in number of pixels per inch for the following
(a) Apple MacBook Pro Retina 13.3"
screen size in pixels = 2560 × 1600 screen dimensions = 11.3 × 7.04 inches
(b) Microsoft Surface Pro 3 12"screen size in pixels = 2160 x 1440 screen dimensions = 10.0 × 6.67 inches
(c) Dell Ultrasharp U2414M 24 inch monitor
screen size in pixels = 1920 × 1200 screen dimensions = 20.43 × 12.77 inches
(d) Google Nexus 6
screen size in pixels = 2560 × 1440 screen dimensions = 5.19 × 2.92 inches
8 Large screens can get away with lower pixel densities because viewing distance is important with regards to
resolution. Use your answers for question 7 to justify this statement.

146 Single licence - Abingdon School

 5.6.4 Bitmapped graphics

Camera resolution
The iPhone 5s’ iSight camera uses a chip with light sensors of width and height 1.5 × 10-6 metre in size, giving
dimensions for each pixel of 1.5 × 10-6 metres (1.5µm where µm is a micrometre) . This camera uses an aspect ratio
of 4:3 and therefore is 3264 pixels across by 2448 pixels down (3264 × 2448) or 7990272 pixels in total. Expressed
to 2 decimal places this is 7.99 megapixels or to none, 8 megapixels.
Apple actually increased pixel size in the iPhone 5s to 1.5µm (from 1.4 µm in iPhone 5) and kept the pixel count
the same by using a 15% larger sensor. The slightly larger sensor size and therefore pixel size improved low light
sensitivity and reduced the ratio of image signal to noise emanating from within the sensor in low light conditions.
This actually improved the quality of the image.
Sony Xperia ’s primary camera uses a chip of width and height 4128 × 3096 pixels. The total quoted number of
pixels is 13.1 megapixels. This is actually greater than the number which contribute to the final image because some
pixels are unused or are shielded from the light because they are around the edges of the sensor.

Questions
9 Why is it useful for smartphone cameras to have 4 megapixels or greater?

Printing a bitmapped image on paper

Paper is an analogue material so it differs from a typical computer screen which is digital, i.e.
divided up into pixels. The coordinate system for drawing by hand on paper is a continuous
one whereas that for a digital screen is a discrete one. However, when a digital image is printed
on paper, it is pixels which are printed, i.e. discrete units. There is a difference, however, for
the printer controls the size of the printed image pixel, all that the digitised image
supplies by way of control is the number of pixels horizontally and the number of Key fact
pixels vertically. Printer resolution and
Dots per inch (DPI):
For example, a 100 × 100 pixel image that is printed in a 1 inch square has
Dots per inch is a measure of
a resolution of 100 pixels per inch (ppi). To produce good quality printed the resolution of a printer. It
photographs, the printer must be capable of printing 300 pixels per inch, at 100% properly refers to the dots of ink
size, and the paper printed on must be coated paper stock. or toner used by an imagesetter,
laser printer, or other printing
A printer creates an image on paper by laying down a series of dots of ink or
devices to print text and
toner. Its resolution is therefore measured in graphics. In general, the more
number of dots per inch (DPI). dots, the better and sharper
the image. DPI is printer
An ink-jet printer prints by moving a
resolution.
printhead across and down the paper. It has
a basic movement of 1200 steps across and A pixel made up of 16 blue dots
1200 steps down, typically. Each pixel of the
image is created by a series of tiny dots and every pixel output is made up of different coloured
inks (usually 4 colours, CMYK - Cyan, Magenta, Yellow, and Key which is black - though
professional printing uses more) deposited by the print head on the paper.
If the printer can print 1200 dots of ink per inch (1200 dpi) and a bitmapped image is sent
to the printer for printing at 300 pixels per inch, then each printed pixel will be consist of 16
smaller ink dots.
Single licence - Abingdon School 147
5 Fundamentals of data representation

Questions
10 An image of size 640 × 480 pixels is to be printed at 300 pixels per inch. What will be the size of the
printed image in inches to one decimal place?

11 The size of a photographic print printed at 300 ppi is 4.2 × 3.2 inches. What was the size in pixels of the
digital image that was printed?

Stretch & challenge question

12 A bimapped image is produced by scanning a 35mm film slide (0.94 inches by 1.42 inches) with a
scanner designed for this purpose. A print of size 9.4 × 14.2 inches is to be made of the bitmapped image
on a printer that prints at 300 pixels per inch, i.e. photographic quality.

(a) How many pixels are printed in a 9.4 inch wide row?

(b) If the scanning resolution is n samples per inch, how many samples, in terms of n, would be taken
in a scan of one row across the film slide (0.94 inches)?

(c) What must the minimum value of the scanning resolution be in samples per inch to produce a print
of acceptable quality?

(d) Using your answer to part (c), what is the size of the bitmap in pixels produced by the scanner?
Express your answer in the form of number of pixel columns (width) by the number of pixel
rows (height). Both numbers are integers.

Key concept Metadata

Metadata: Microsoft’s Paint program that comes with the Windows operating system
The header part of the bitmap enables bitmaps to be created and saved. Bitmap files saved in Windows
file contains information about Paint with file extension ".bmp" have a file structure which conforms to the
the bitmap data part of the file,
Windows bitmap format shown in Figure 5.6.4.13. The header contains
such as number of bits per pixel.
This is metadata because it is
information about the bitmap data part of the file, such as
data about data, i.e. the data in ■■ number of bits per pixel
the bitmap part of the file.
■■ horizontal width of bitmap in pixels
■■ whether it is compressed or not, etc.
This is called metadata because it is data about data.
HEADER
The actual detailed structure and content of the header is shown in Figure
5.6.4.14 for an uncompressed, RGB, 24 bits per pixel, 4 × 2 pixel bitmap
produced with Microsoft Paint and shown in Figure 5.6.4.15.
Figure 5.6.4.16 shows the data part of the bitmap file for
BITMAP DATA
Figure 5.6.4.13. Bytes 54, 55 and 56 correspond to the first pixel in the
bottom row of the image. The colour of each pixel is controlled by three bytes,
the first in the triplet controls red, the second green, and the third blue. Thus
Figure 5.6.4.13 Structure of a this first pixel of the bottom row is green because its colour is controlled by
the triplet 0, 255, 0. The second pixel in the bottom row is black because it is
Windows bitmap file
controlled by the triplet, 0, 0, 0.

148 Single licence - Abingdon School

 5.6.4 Bitmapped graphics

Figure 5.6.4.16, byte 10 states that the bitmap data begins at byte 54. Byte 77 0 Top row fourth pixel
34 states that the length of the bitmap data is 24 bytes, which indeed it is 76 0 BLACK
as Figure 5.6.4.15 shows. Microsoft chose to store the bytes in little-endian 75 0 0, 0, 0
fashion. In little-endian, the least significant byte is stored in the smallest 74 0 Top row third pixel
address. For example, "Where the data starts" is four bytes long and has value 73 0 BLUE
0, 0, 0, 54 which in decimal is just 54. 72 255 0, 0, 255

71 255 Top row second pixel

Questions 70 0 RED
69 0 255,0 ,0
13 Using Figure 5.6.4.16 as a reference, 68 0 Top row leftmost pixel
list nine items of metadata found in 67 0 BLACK
the header of a bitmap file. 66 0 0, 0, 0

Figure 5.6.4.15 Image 65 0 Bottom row fourth pixel

produced when 24-bit 64 0 BLUE
63 255 0, 0, 255
bitmap rendered on screen
62 255 Bottom row third pixel
61 0 RED
60 0 255, 0, 0

59 0 Bottom row second

58 0 pixel BLACK
57 0 0, 0, 0

56 0 Bottom row leftmost

55 255 pixel GREEN
54 0 0, 255, 0

Figure 5.6.4.14 Bytes 54 to 77

of bitmap file represent bitmap
data for 4 x 2 pixels image,
24-bit Windows bitmap.

Information
Reading and writing bytes:
The source code of programs
to read and write a file of bytes
can be downloaded from www.
educational-computing.co.uk.

Single licence - Abingdon School 149

5 Fundamentals of data representation
0 identifier The file type 66 'B'
Bitmap, BM

}
1 identifier must be ‘BM’. 77 ‘M’
2 file size The size, in bytes, of the bitmap file. 78
4 bytes,
3 file size 0
4 file size 0 0, 0, 0, 78
5 file size 0
6 reserved Reserved; must be zero. 0
7 reserved 0
8 reserved 0

}
9 reserved 0
10 bitmap data offset The offset, in bytes, from the 54 Where the data
11 bitmap data offset beginning of the 0 starts
12 bitmap data offset BITMAPFILEHEADER 0
0, 0, 0, 54

}
13 bitmap data offset structure to the bitmap bits. 0
14 bitmap header size 40
15 bitmap header size 0 4 bytes,
16 bitmap header size 0
0, 0, 0, 40

}
17 bitmap header size 0
18 Horizontal width of bitmap in pixels 4
19 Horizontal width of bitmap in pixels 0 4 bytes,
20 Horizontal width of bitmap in pixels 0 0, 0, 0, 4

}
21 Horizontal width of bitmap in pixels 0
22 Vertical width of bitmap in pixels If Height is positive, the 2
23 Vertical width of bitmap in pixels bitmap is a bottom-up DIB. 0 4 bytes,
24 Vertical width of bitmap in pixels DIB = Device Independent 0 0, 0, 0, 2
25 Vertical width of bitmap in pixels Bitmap. 0
26 Number of planes in the bitmap The number of planes for the target 1
27 Number of planes in the bitmap device.This value must be set to 1. 0
28
29
30
Bits per pixel
Bits per pixel
Compression The type of compression for a
24
0
0
} 2 bytes,
0, 24

31 Compression compressed bottom-up bitmap 0

32 Compression RGB uncompressed= 0x0000 0

}
33 Compression JPEG = 0x0004, PNG = 0x0005 0
34 Bitmap data size Size in bytes 24
35 Bitmap data size 0 4 bytes,
36 Bitmap data size 0 0, 0, 0, 24
37 Bitmap data size 0

Horizontal resolution in pixel/metre An application can use this value

38 0
of the target device to select a bitmap from
39 Horizontal resolution in pixel/metre a resource group that best 0

Horizontal resolution in pixel/metre matches the characteristics

40 0
Figure 5.6.4.16
41 Horizontal resolution in pixel/metre of the current device. 0
Header part of
Vertical resolution in pixel/metre
42 0
of the target device bitmap file. It
43 Vertical resolution in pixel/metre 0
contains metadata.
44 Vertical resolution in pixel/metre 0
45 Vertical resolution in pixel/metre 0
46 Number of colours used If zero, the bitmap uses the maximum 0
47 Number of colours used number of colours corresponding to the 0
48 Number of colours used value of the bits per pixel. 0
49 Number of colours used 0
50 Number of important colours used If zero, all colours are important. 0
51 Number of important colours used 0
52 Number of important colours used 0

150 53 Number of important colours used

Single licence - Abingdon School 0
 5.6.4 Bitmapped graphics

Programming Tasks

1 Using Microsoft Paint (or equivalent), create and save a 4 × 2 pixels 24-bit uncompressed Windows
bitmap similar to Figure 5.6.4.15. Use the pencil tool to change the colour of individual pixels. You will
need to zoom in to make the pixels large enough to manipulate.

2 Write a program that opens and reads the contents of this file, byte by byte, displaying each byte as a
decimal integer on the console. Number these bytes starting from 0 so that the console output displays
number followed by byte value read from file. Check that your output shows similar values to those shown
in Figures 5.6.4.14 and 5.6.4.16.

3 Using Paint, change the colours of the pixels in the bitmap, noting the RGB values and re-run your
program. Check that the output from your program now reflects the new RGB values of the colours.

4 Using Paint, create and save an 8 x 2 pixels 24-bit uncompressed Windows bitmap, with differently
coloured pixels.

5 Re-run you program and note the relevant changes in the header (metadata) and the data part of the
bitmap displayed on the console. Do the displayed changes agree with what you expect?

6 Edit your program so that it also writes each byte that it reads from the opened bitmap file to a new file.
Save your edited program under a new name. The new bitmap file produced by running the new program
should be given a suitable name and the extension “.bmp”. At the moment it should be just a copy of the
original.

7 Edit your new program so that it alters the three bytes of a chosen pixel before writing these to the new
bitmap file. Now open the changed bitmap file in Paint and check that your program has changed the
colour of a pixel.

8 How could a short sequence of 8-bit ASCII character codes be placed in a bitmap file? Choose a much
larger image bitmap file than you have been working with, e.g. 640 x 480 pixels, and use your program
suitably modified to replace pixel bytes in this bitmap with a sequence of 8-bit ASCII codes. Check the
result by displaying the new image bitmap file in an image viewer, e.g. Paint. Can you detect the ASCII
codes?

9 Write a program that reads an altered image bitmap file and recovers the sequence of 8-bit ASCII codes.
Display these as characters on the console.

Single licence - Abingdon School 151

5 Fundamentals of data representation

Calculating storage requirements for bitmapped images

Ignoring the storage space taken up by metadata, the storage requirements of
Key fact
the data part of a bitmapped image is calculated as follows
Storage requirements for
Storage requirements = width in pixels × height in pixels × colour depth
bitmapped images):
Storage requirements = width in This is sometimes referred to as being the minimum file size for a bitmapped
pixels × height in pixels × colour image.
depth For example, a bitmapped image has dimensions 5184 × 3456 pixels and uses
24-bit colour. What is the size of the data part of the bitmap in bits? In bytes?
What is it in megabytes (1000000 bytes), to 1 decimal place?
Size in bits � 5184 × 3456 × 24 � 429981696
Size in bytes � (5184 × 3456 × 24)/8 � 53747712
Size in megabytes � (5184 × 3456 × 24)/(8 x 1000000) � 53.7

Questions
14 What is the minimum file size in bytes, for a bitmapped image that has a colour depth of 12 bits and
dimensions 640 × 480 pixels?

In this chapter you have covered:

■■ How bitmaps are represented
■■ For bitmaps the meaning of
• resolution

• colour depth

• size in pixels

■■ How to calculate storage requirements for bitmapped images

■■ Bitmap image files may also contain metadata
■■ Typical metadata

152 Single licence - Abingdon School

 5 Fundamentals of data representation
5.6 Representing images, sound and other data
Learning objectives:
■■ Explain how vector graphics
represents images using lists of
■■ 5.6.5 Vector graphics
objects How does a vector graphic represent images?
A vector graphic image is created in a similar
■■ Give examples of typical
way to the way we draw with coloured pencils by
properties of objects
hand, i.e. by drawing lines from point to point
■■ Use vector graphic primitives in different colours, drawing and shading closed
to create a simple vector shapes such as rectangles in different colours, etc.
graphic
A vector graphic such as the one shown in
Figure 5.6.5.1 is just a collection of objects of
Figure 5.6.5.1 Vector
various types, (e.g. circle, line, arc) together with
graphic image
their properties (e.g. radius of circle).
A vector graphic can be represented as a list of objects or a list of drawing
Key principle commands that reference geometric objects such as

Vector graphic:
■■ points in space,
A vector graphic can be ■■ straight lines or curves connecting these points
represented as a list of objects or
a list of drawing commands that ■■ shapes formed from closed paths (such as rectangles, circles, ellipses,
reference geometric objects triangles, other polygons, and non-regular shapes created from
straight and curved lines).
■■ Other commands apply fill operations to shapes or set the thickness
of lines.
All of these points, lines, curves and shapes make up what is seen as the image.
newpath
moveto 100 300
lineto 200 250
lineto 200 200
setlinewidth 2
stroke

showpage
Figure 5.6.5.2 Vector graphic image and the commands that produced it
Figure 5.6.5.2 shows a vector graphic image of two black lines and the
SCRIPT commands that produced it.
A command in this vector graphic scripting language consists of operands
(parameters) and an operator written operator operand1 operand2.

Single licence - Abingdon School 153

5 Fundamentals of data representation

For example,
■■ moveto 100 300, moves a “phantom” pen to the point whose
coordinates are 100,300.
■■ The command lineto 200 250 creates the path for a straight
line from the position of the current point to the new current point
at 200,250.
■■ The command setlinewidth 2 sets the pen to draw lines with a
thickness of 2 units which means 2/72 inch if the unit is 1/72 inch
(72 ppi).
■■ The command stroke causes the pen to move along the defined path
inking it in.
■■ The final command in the script, showpage displays the page on the
screen.
The parameters supplied to the commands, e.g. 200 and 250 in command
lineto 200,250 are just numbers. Therefore, to apply a change, such as
resize or reshape, to a part of a vector graphic image we simply do arithmetic
Key fact on these numbers and then draw the vector graphic again.
Image quality: This means we can resize a vector shape as many times as we like, making it any
A vector graphic can be scaled size we need, without any loss of image quality.
without any loss of image
quality. This is why vector graphic images retain crisp, sharp edges to graphic elements/
As a consequence, the quality shapes no matter by how much we enlarge them, and, unlike pixel-based
of a vector graphic image is graphics, vector graphic images are resolution-independent.
independent of the resolution
of the display device, i.e. This means that when printing, the resolution of the printed image is
vector graphic is resolution determined by the highest resolution that the printer can print at.
independent.
Figure 5.6.5.3 shows the original vector graphic and a separate scaled by 2
version.
In Figure 5.6.5.3, the scale 2 2 command is applied to all the operand
numbers in the other commands, e.g. moveto 100 300 becomes moveto
200 600, setlinewidth 2 becomes setlinewidth 4.
newpath
scale 2 2
moveto 100 300
lineto 200 250
lineto 200 200
setlinewidth 2
stroke

showpage
Figure 5.6.5.3 Vector graphic image from Figure 5.6.5.2, its scaled
counterpart, and the commands that produced the scaled version.

154 Single licence - Abingdon School

 5.6.5 Vector graphics

Questions
1 How does a vector graphic represent images?

2 Why are vector graphic images resolution-independent?

Display devices are mainly digital so to display a vector graphic image, it must
be rasterised, i.e. turned into pixels. There are some display devices that can
display the vector graphic image without needing to rasterise at all. These
essentially are Cathode Ray Tube-based (CRT) displays with a single colour
continuous phosphor coating applied to the screen. These are ideal for being
controlled by commands in a vector graphic script. Some radar displays,
monitors used in medicine and laser shows are operated directly in this way as
are X-Y plotters which are a kind of printer that uses vector data.

Questions
3 Explain why a vector graphic image is still displayed without any loss of image quality on a digital display
when it is scaled.

Tasks Information
These activities are designed to familiarise you with a vector graphic Postscript interpreter and
format. viewer Ghostscript:
Download from
Download and install Ghostscript 9.15 or later for Windows and
http://www.ghostscript.com/
GSview 5.0 or later. download/gsdnld.html and
1 Create the following file in a text editor, e.g. Microsoft Wordpad, and instal then download and install
GSview 5.0 or later: Download
save with extension .ps
from
newpath http://pages.cs.wisc.edu/~ghost/
0 1 0 setrgbcolor gsview/get50.htm
256 500 100 0 360 arc fill
newpath Download postscript language
1 0 0 setrgbcolor tutorial and cookbook,
100 500 50 0 360 arc fill PSBlueBook.pdf from
showpage http://partners.adobe.com/
public/developer/ps/sdk/
Open this file in GSview.
sample/index_psbooks.html
(a) What do you see?
(b) Change the fill colour of the first object to blue.
(c) Change the dimension of the second object from 100 to 50
(d) Change the arc from 0 360 to 0 180.

Single licence - Abingdon School 155

5 Fundamentals of data representation

Tasks

2 The code folder of the cookbook, PSBlueBook.pdf contains several

postscript vector graphic files. Explore the following in GSview and
Wordpad. You are not required to understand the commands only to
observe that the vector graphic files contain lists of commands that
reference objects and that the contents can be viewed in a text editor
(a) Prog_01.ps
(b) Prog_02.ps
(c) Prog_03.ps

3 GSView can convert postscript files to bitmapped files and pdfs.

Convert Prog_02.ps to a bmp16m 72ppi bitmapped image file,
Prog_02.bmp. Compare the file size of Prog_02.ps with its bitmapped
equivalent. What do you observe?
4 Convert Prog_02.ps to a pdfwrite 72ppi file, Prog_02.pdf. Launch
Acrobat reader and open Prog_02.pdf. What do you observe? Is the
pdf format vector graphic or bitmapped?

Using vector graphic primitives to create a simple vector

graphic
Scalable Vector Graphics (SVG) is an XML-based vector image format for two-
dimensional graphics.
The specification of SVG is an open standard developed by the World Wide
Web Consortium (W3C) since 1999.
SVG images can be created and edited with any text editor because the images
are defined in XML text files.
Figure 5.6.5.4 shows an SVG image rendered in a browser window. All
modern web browsers have some degree of support for rendering SVG images.
The XML-text of this SVG image shown below was produced with a text editor
and saved as BasicRecSVG.svg.

<svg xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" >

<rect x="10px" y="10px" width="150px"
height="150px" fill="rgb(0,255,0)"
stroke-width="1" stroke="rgb(0,0,0)" />
</svg>

Figure 5.6.5.4 Scalable Vector Graphics image produced from XML and rendered in a browser

156 Single licence - Abingdon School

 5.6.5 Vector graphics

The root element of all SVG images is the <svg> element. Here is how it looks:
<svg xmlns=”http://www.w3.org/2000/svg” xmlns:xlink=”http://www.w3.org/1999/xlink” >

Namespaces are defined in the <svg… /> part to avoid the names in the rest
of this file clashing with similar names elsewhere:
xmlns=”http://www.w3.org/2000/svg” xmlns:xlink=http://www.w3.org/1999/xlink

The vector graphic part <rect…/> element specifies a rectangle of

width 150 pixels, height 150 pixels, 10 pixels in from the left side of the
browser window and 10 pixels from the top.
In the SVG coordinate system the point x = 0, y = 0 is the upper left corner.
The rectangle has a black outline
stroke="rgb(0, 0, 0)" of width 1px (stroke-width="1")

and is filled in the green colour defined in the rgb triplet,

rgb(0, 255, 0)

Nesting SVG elements, as shown in Figure 5.6.5.5, enables shapes to be drawn

relative to the position (x, y) of its enclosing svg element.

<svg xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" >

<svg x="10">
<rect x="10px" y="10px" width="150px" height="150px" fill="rgb(0,255,0)"
stroke-width="1" stroke="rgb(0,0,0)" />
</svg>
<svg x="300">
<rect x="10px" y="10px" width="150px" height="150px" fill="rgb(255,0,0)"
stroke-width="1" stroke="rgb(0,0,0)" />
</svg>
</svg>

Figure 5.6.5.5 Scalable Vector Graphic script

Figure 5.6.5.6 Scalable Vector Graphic image produced

using nested svg elements and rendered in a browser.

Single licence - Abingdon School 157

5 Fundamentals of data representation

So the position of the first rectangle will be measured from x = 10, y = 10 and
therefore be drawn starting from x = 10+10, y = 10.
The position of the second rectangle will be measured from x = 300, y = 10
and therefore be drawn starting from x = 300 + 10, y = 10.
The result is shown in Figure 5.6.5.6.
Figure 5.6.5.7 shows another SVG image produced from XML and rendered
in a browser. The XML defines two shapes, a green rectangle and a red circle.
The XML-text of this SVG image was also produced with a text editor and is
shown in Figure 5.6.5.8.
Figure 5.6.5.7 SVG image

<svg xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" >

<rect x="50" y="150" width="150" height="150" fill="rgb(0,255,0)" stroke-width="1"
stroke="rgb(0,0,0)" />
<circle cx="50" cy="150" r="30" fill="red" stroke-width="4" stroke="black" />
</svg>

Figure 5.6.5.8 SVG image file contents

Tasks

5 Using a text editor such as Wordpad, create a copy of the vector graphic script shown in Figure 5.6.5.5
and save as TwoRects.svg. Experiment with changing the properties of each rectangle and observe the
outcome by opening TwoRects.svg in a web browser.

6 Repeat activity Task 5 for the vector graphic script shown in Figure 5.6.5.8. Save as RectAndCircle.svg.

7 Using a text editor such as Wordpad, create a copy of the vector graphic script shown below, save it as
Text.svg.

<svg
xmlns=”http://www.w3.org/2000/svg”
xmlns:xlink=”http://www.w3.org/1999/xlink”>
<text x=”10” y=”30” fill=”red”
font-family=” ‘Lucida Grande’,sans serif”
font-size=”32” transform=”rotate(90 20,30)”>
I love SVG!
</text>
</svg>
(a) Display Text.svg in a web browser.
(b) Experiment with changing the properties of the text and observe the outcome by
opening Text.svg in a web browser.

158 Single licence - Abingdon School

 5.6.5 Vector graphics

Tasks Information
8 Inkscape is an open-source vector graphics editor similar to Adobe Inkscape:
Download and install https://
Illustrator, Corel Draw, Freehand, or Xara X but it differs from these
inkscape.org/en/download/
because it uses Scalable Vector Graphics (SVG), an open XML-based
W3C standard, as the native format. Download and install.

Examples of typical properties of objects

The properties of a vector graphic object specify its dimensions, its position in
coordinate space and its appearance, e.g.
Rectangle
■■ Dimensions: width, height
■■ Position: bottom left or top left corner x, y coordinates
■■ Appearance: filled with colour green, rgb(0,255,0),
outline stroked in black, stroke width 1 px
Circle
■■ Dimensions: radius
■■ Position: centre x, y coordinates
■■ Appearance: filled with colour green, rgb(0,255,0),
outline stroked in black, stroke width 1 px

Questions
4 List three types of object that could be found in a vector graphic image file.

5 List four properties associated with one of the objects in your

answer to Q4 that could be found in a vector graphic image file.

In this chapter you have covered:

■■ How vector graphics represents images using lists of objects
■■ Examples of typical properties of objects:
• dimensions - width, height, radius, line thickness; position in coordinate
space - x, y coordinates;
• appearance - filled and fill colour, outline stroke colour
■■ Using vector graphic primitives to create a simple vector graphic

Single licence - Abingdon School 159

5 Fundamentals of computer systems
5.6 Representing images, sound and
Learning objectives: other data
■■Compare the vector graphics
approach with bitmapped
■■ 5.6.6 Vector graphics versus bitmapped graphics
graphics approach Figure 5.6.6.1 shows a vector graphic image of a black circle on a white
background and a bitmap image of the same circle, at their original size and
■■Understand the advantages
each magnified seven times. The vector graphic image was produced from the
and disadvantages of each
following script and rendered in GSview, a graphical interface for displaying
■■Be aware of appropriate uses interpreted PostScript:
of each
newpath
300 500 10 0 360 arc fill

Information Vector graphic magnified 7 times

Original size of vector
GSView:
graphic, radius 10 x 1/72"
GSview is a graphical interface
for Ghostscript, an interpreter
for the PostScript language and
Portable Document Format
(PDF). GSview allows selected
pages to be viewed, printed, or Bitmap magnified 7 times
converted to bitmap, PostScript
or PDF formats. Converted to bitmap,
GSview can be downloaded from resolution 72 ppi
http://pages.cs.wisc.edu/~ghost/
gsview/index.htm

Figure 5.6.6.1 Vector graphic image and its equivalent bitmap image,
original size and magnified 7 times

GSview was also used to produce the 24 bit, 72 ppi bitmap image from its
vector graphic equivalent.
Bitmap, or raster images as they are sometimes called, use a rather primitive
form of representation consisting of no more than a lattice of small rectangular
areas called pixels. In a bitmap file, the only information that is stored about
a pixel is its colour. Therefore, for the bitmap image of a black circle on a
white background displayed in Figure 5.6.6.1, there is no stored black circle
as such in the bitmap file. All that the computer knows about the image from
the bitmap file is that some of its pixels are black and some are white. When
the bitmap image is rendered on screen it appears as a black circle to the viewer
because that is how the viewer perceives the black and white pixels, but the
larger the pixels, in the displayed image, the harder this task for the viewer.

Single licence - Abingdon School 160

5 Fundamentals of computer systems

As the computer does not see the bitmap image representation as a circle object with properties, such as position,
radius, stroke (outline) colour, it cannot move the circle or transform it by, for example, changing its radius. All it
can do is to change the colour of pixels, e.g. all white pixels to red or all black pixels to green, say.
Even using a pixel editing package, moving the black circle is difficult to do precisely, especially if the edge of the
circle is antialiased so that some pixels on the edge have intermediate values between
black and white as in Figure 5.6.6.2.

Questions
1 For a bitmapped graphic, explain why it is difficult to move objects
from one position in the image to another.

However, storing the black circle on a white background image as a vector graphic
avoids using pixels altogether. Instead, the actual definition of the circle is stored as Figure 5.6.6.2 Enlarged
a circle object with its properties specified. Its properties are its (x, y) centre position antialiased bitmap
in the coordinate system of the image, its radius, its stroke (or outline) colour, stroke
thickness and its fill colour.
The immediate advantage is that the circle object can be treated separately from the rest of the image and its
properties adjusted to move or transform it.
Indeed, if the vector graphic image contains many circles, each can be selected separately and individually
manipulated. Also, with such an image, a computer can automatically, say, delete all circles or scale all black circles
to three times their size or paint all green objects red. There is no pixel selection taking place only object selection.

Questions
2 For a vector graphic, explain why it is relatively easy to move objects from one position in the image to
another and to manipulate objects in other ways

As an example of the way that a vector graphic allows easy selection of image
Information
objects, Figure 5.6.6.3 shows the contents of the part of an SVG file produced
Inkscape:
in Inkscape by drawing a red circle. If there was another circle in the image then Download and install
another <circle ……./> block of text would be stored in the file. https://inkscape.org/en/download/

Some of the properties of the circle object

<circle Figure 5.6.6.3 Part of SVG
referenced in Figure 5.6.6.3 are
r=”117.14286” file produced in Inkscape by
radius of circle, r : 117.14286 cy=”495.21939” drawing a red circle
centre of circle x coordinate, cx : 382.85715 cx=”382.85715”
centre of circle y coordinate, cy : 495.21939 id=”path3338”
fill colour : #ff0000 (red) style=”fill:#ff0000;
colour of stroke : #000000 (black) fill-rule:evenodd;stroke:#000000;
stroke-width : 0px stroke-width:0px;
stroke-linecap:butt;
stroke-linejoin:miter;stroke-opacity:1” />

161 Single licence - Abingdon School

 5.6.6 Vector graphics versus bitmapped graphics

Advantages of the vector graphics approach

Vector images are scalable
Since vector shapes are essentially represented by their geometric properties,
Key fact
each time a change is made to the shape, either by resizing or reshaping it in
some way, it is these properties that are changed. The vector graphic image Advantages of the vector
graphics approach:
is simply redrawn on the screen using the new definition. This means that a
1. Vector images are scalable
vector graphic shape can be scaled in size any number of times without any without loss of image quality.
loss of image quality. Vector graphic drawn shapes retain their crisp, sharp 2. Vector images tend to
edges no matter how large they are made. Vector graphic images are resolution- produce smaller file sizes
than bitmapped image files.
independent. This is in contrast to bitmapped (raster) images which are not.
3. Vector images are easy to
A big disadvantage of bitmapped images or shapes is that they don’t scale very create, read and edit.
well, at least not when making them larger than their original size. Enlarge a 4. Vector objects are reusable.

bitmapped image or shape too much and it will lose its sharpness. Enlarge it
even more and the pixels that make up the image or shape can become visible,
resulting in a blocky or pixelated appearance.
Vector images tend to produce smaller file sizes than bitmapped image files
The original vector graphic image file for the black circle on a white background consisted of 44 bytes. Its
bitmapped equivalent was 1400 bytes. Thus, a vector image usually takes up less storage space than a bitmapped
image.
Bitmapped (raster) images always have a fixed size in pixels, e.g. 300 x 240 pixels. In an uncompressed bitmap,
doubling the dimensions, e.g. from 300 x 240 to 600 x 480 pixels, quadruples the size of its file because extra pixels
need to be stored. This is not the case for vector graphic image files because the number of vector shape definitions
remains the same no matter by what factor the image is scaled. All that changes are the property values for each
shape; the file size, or computer memory requirements, remain the same.
Vector images are easy to create, read and edit
Vector graphic image files are (generally) plain text files and as such they can easily be generated by hand or as the
output of user written programs. As text files, they are human-readable as well as generally free format, that is,
the text can be split across lines and indented to highlight its logical structure. A vector graphic image file can be
directly edited in a text editor. This is not the case with bitmap images.
Vector image has no resolution
Unlike bitmaps, vector graphic images are resolution-independent. Vector graphic images always print at the highest
possible resolution of the printer that they are sent to. They also display without loss of image quality at different
screen resolutions, keeping their dimensions the same, more or less.
On the other hand, when bitmapped images are printed, how they look always depends very heavily on their
resolution. Even though the bitmapped image may look good on the computer screen, printing requires much
higher resolution than computer screens display at. If the bitmapped image doesn’t have enough pixels to print at
the size required, its printed equivalent will be of poor quality.
Vector objects are reusable
It is very easy to pick an object from one vector graphic image,
Questions
transform it or restyle it without any loss of quality, and then to
insert it into another vector graphic. 3 State four advantages that vector graphics
has over bitmapped graphics.

Single licence - Abingdon School 162

5 Fundamentals of computer systems

Disadvantages of the vector graphic approach compared with

the bitmapped approach
Vector graphic images are constructed from a limited number of object types,
e.g. circles, rectangles, lines. There are many images that are difficult or even
Key fact impossible to reproduce exactly in vector form. For example, images that
require complex textures, such as human skin or hair. These images are typically
Disadvantages of the vector
graphics approach compared present in digital photographs. For these, tasks such as colour corrections and
with the bitmapped approach: retouching are best done in a bitmap or raster image editor such as GIMP or
There are many images that are Photoshop. Similarly, for any image where the texture of the coloured surface
difficult or even impossible to
is important, vector graphics’ flat colours and colour gradients are insufficient
reproduce exactly in vector form:
whereas the bitmap approach can emulate painting in oils, pastels and
• Images that require complex
textures such as human skin watercolours when done in a specialised raster tool such as ArtRage or Corel
• Where the texture of the Painter.
coloured surface is important
such as when emulating Questions
painting in oils, pastels and
4 State two disadvantages that vector graphics has compared with
watercolours.
bitmapped graphics.

Appropriate uses of vector graphics

Vector graphics are appropriate to use when the task is a drawing one. For
example, when producing charts, diagrams, cartoons, maps, illustrations,
typography of all kinds, leaflets, posters, web graphics, etcetera, because these
Key fact can all be constructed from shape and text objects.
Appropriate uses of vector Appropriate uses of bitmapped graphics
graphics: Bitmapped graphics are best suited to capturing a record of the real world such
Vector graphics are appropriate as in photographic images or where the task is a painting one because they can
to use when the task is a drawing
represent texture in a realistic way.
one, e.g. maps.
Appropriate uses of bit-
mapped graphics: Questions
Bitmapped graphics are best 5 For what type of use is vector graphics appropriate and why?
suited to capturing a record of the
real world such as in photographic 6 For what type of use is bitmapped graphics appropriate and why?
images because they can represent
texture in a realistic way.
In this chapter you have covered:
■■ Comparing the vector graphics approach with bitmapped graphics
approach
■■ The advantages and disadvantages of each approach
■■ Appropriate uses of each approach

163 Single licence - Abingdon School

5 Fundamentals of data representation
6 Representing images, sound and other data
Learning objectives:
■ Describe the digital
representation of sound in
■ 5.6.7 Digital representation of sound
terms of: Classification of waveforms
Periodic and aperiodic
• sample resolution
It is useful when working with sounds to graph their waveforms (amplitude
• sampling rate and of air pressure, or voltage from a microphone, as a function of time). Figure
Nyquist theorem. 5.6.7.1(a) shows the waveform of the sound “Laa” spoken into a microphone.
Figure 5.6.7.1(b) shows the waveform of white noise, the sort of sound heard
■ Calculate sound sample sizes
from an analogue radio not tuned to any radio station.
in bytes

Figure 5.6.7.1(b) White noise sound recording

Figure 5.6.7.1(a) “Laa” sound recording
The waveform in Figure 5.6.7.1(a) consists of a repeating pattern called a
Key concept
periodic oscillation, with maybe a few minor deviations. This is characteristic
Frequency of a sound: of sounds that have pitch. Pitch is what humans perceive as frequency and
The pitch of a sound is what recognise by its position in a range of audible frequencies that range from low
humans perceive as frequency.
to high. The pitch of a sound is varied when you sing or whistle a song melody.
Frerquency is measured as the
number of cycles per second Musicians use a notation for indicating the pitch of a sound to be played.
of a repeating pattern in a The waveform in Figure 5.6.7.1(b) does not repeat in time (aperiodic) because
periodic waveform or vibration
it is essentially random in nature.
or oscillation. Its unit is the Hz
which is one cycle per second. Waveforms are broadly divided into two classes:
1. periodic (repeat in time)
Sounds are caused by vibrations
or oscillations of a column of 2. aperiodic (don’t repeat in time)
air in a woodwind instrument
or in a stretched string such as The first class is subdivided into simple (sinusoidal) and complex (non-
a violin or guitar string when sinusoidal) waveforms. It turns out that these complex (non-sinusoidal)
bowed or plucked. waveforms are composed of sinusoidal waveforms of different frequencies
and amplitudes added together. This means that complex waveforms can be
synthesised by selecting the right sinusoids to add together.
The second class is subdivided into impulsive (occur once) and noise
(continuous but random) waveforms.

Single licence - Abingdon School 164

5 Fundamentals of data representation

Sinusoids
If we imagine the red dot inside the circle in Figure 5.6.7.2 is a cyclist cycling
round the circle at constant speed of, say, one complete loop of the circle
per minute then we are imagining
One cycle Sine
a periodic system because it has a
Time repeating pattern. If we were to lie
down in the plane of the circle, i.e.
view the circle sidewise-on, then
looking along the time axis across the
When the red dot returns to its starting page we would see the vertical distance
position after completing a revolution of of the cyclist from this axis vary in time
Cosine the circle it is said to have gone through
one cycle as shown in red. Mathematically, this
waveform is called a sine waveform or
Time sine wave.
Figure 5.6.7.2 Generating sinusoids - sine and cosine waveforms
If we look along the time axis going
down the page we see the vertical
Task distance of the cyclist from this new axis also vary in time as shown in
To see a demonstration of red. Mathematically, this waveform is called a cosine waveform or cosine
generating a sinusoid from wave. Both sine and cosine waveforms are called sinusoids. A sinusoid is
circular motion visit characterised by three quantities:
http://treeblurb.com/dev_math/
sin_canv00.html
1. Peak amplitude or just amplitude which is the maximum vertical
distance of the waveform from the time axis. In circle terms this is the
radius of the circular path that the cyclist follows.
Task 2. Frequency which in circle terms is the number of cycles of the circle
per second, e.g. 60 loops of the circle per second. In pitch or frequency
Download and install
SFS/ESynth from terms it is expressed in Hz, e.g. 60 Hz.
https://www.phon.ucl.ac.uk/ 3. Phase which in circle terms is where the red dot starts. Conventionally,
resource/sfs/esynth.php this is measured in angle from the time axis, e.g. for the sine waveform
Experiment with generating
time axis, angle = 0 degrees; for the cosine waveform time axis this is
sinusoids using ESythn.
90 degrees.

Figure 5.6.7.3(a) Frequency 200Hz, Figure 5.6.7.3(b) Frequency 200Hz,

amplitude 0.5, phase 0 degrees amplitude 0.5, phase 90 degrees
165 Single licence - Abingdon School
 5.6.7 Digital representation of sound

Figure 5.6.7.3 shows waveforms generated by ESynth with frequency 200 Hz,
amplitude 0.5 and phases 0 and 90 degrees.
Sampling a waveform Key concept
Sampling rate and bit depth (sample resolution) Sampling rate:
We have learned that the term periodic refers to any waveform that can be Sampling rate or sampling
described in terms of going round in a circle. frequency is the number of
samples taken per second.
Figure 5.6.7.4 shows one cycle of a sine wave generated along the time axis
by recording the vertical distance of the clock hand from this axis in time. The Bit depth:
clock hand rotates anticlockwise and the recording starts when this hand is Bit depth or sampling
resolution is the number of bits
pointing along the time axis (phase = 0 degrees).
allocated to each sample.
To create a sine-wave generator (oscillator) in a digital computer all that is
needed is to store, at
successive intervals of time, Positions marked | on clockface
the vertical distance of when waveform sampled Sample point
One cycle
the clock hand from the per second B

time axis. This is called B

sampling. C Time/s
C A
0 0.25 0.5 0.75 1.0
The positions marked A
D
in red on the clock face
indicate the moments in Sampling clock D

time when the vertical One cycle

Figure 5.6.7.4 Generating a sine wave by
distance is sampled.
the circle method and sampling it 4 times
Let’s suppose that the clock per cycle
hand rotates at 1 revolution
per second, i.e. 1Hz, then in one revolution 4
samples are taken at A, B, C and D, measured
and recorded, i.e. a sampling rate of 4 samples
per second (4 Hz).
Each measurement made by the digital
computer is stored in binary. This is the process
of quantisation. The number of binary digits
used for each measurement is called the bit
depth. Bit depth is one way of specifying
sample resolution.
Figure 5.6.7.5 shows the first stage in the
process of generating a sine wave digitally with
Adobe® Audition CC 2015. At this stage, the Figure 5.6.7.5 Setting sampling rate
sampling rate is set at 44100 Hz and the bit and bit depth
depth at 16 bits.

Single licence - Abingdon School 166

5 Fundamentals of data representation

Figure 5.6.7.6 shows sampling occurring

Positions marked | on clockface at a rate of 8 samples per cycle. If the
when waveform sampled Sample point clockhand rotates at 1 revolution per
C
second then the sampling rate is 8 samples
B D per second, i.e. one sample every ⅛
C
D B second.
E Time
E A To convey an important point about
A
F
G
H sampling frequency Figure 5.6.7.7 and
F H Figure 5.6.7.8 have been simplified
G by omitting any filtering which
Sampling clock
Figure 5.6.7.6 Generating a sine wave by the circle would normally be applied during the
method and sampling it 8 times per cycle reconstruction process.

Sample point
B

Figure 5.6.7.7 One cycle of Figure 5.6.7.7 shows one cycle of the
Time
reconstructed wave, sampling rate A
C
reconstructed wave for a sampling rate of 4
4 samples per second samples per second - see Figure 5.6.7.4.
D

One cycle Figure 5.6.7.8 shows several cycles of the

reconstructed wave for a sampling rate of 4 samples
Time per second. It is still periodic with a frequency of
1 Hz, the same frequency as the original sine
wave. However, its shape is no longer a sine wave.
The waveform is now triangular in shape. We
Figure 5.6.7.8 Three cycles of reconstructed wave need to sample at a higher rate to get a better
approximation to a sine wave - see Figure 5.6.7.6.
However, if this triangular waveform was played through a sound card and
Key concept loudspeakers it would still have a pitch of 1 Hz (we would need to work with
higher frequencies to make a sound that the ear would perceive in a tone-like
Jean Baptiste Joseph Fourier
(1768-1830), introduced the way, e.g. 200 Hz) but it would sound different from a sine wave of the same
concept by which a signal can frequency. We say that its timbre is different. Higher frequencies have been
be synthesised by adding up its added which are whole number multiples of the frequency with which the
constituent frequencies.
waveform repeats. These are called harmonics. The repetition frequency of the
He introduced the concept of waveform is called the fundamental frequency.
frequency for elementary signals
that belong to a set of sinusoidal The triangular waveform is thus made up of a fundamental frequency plus
signals (sines and cosines) with harmonics of the fundamental frequency. Figure 5.6.7.9 shows that adding two
various periods of repetition. harmonics, 3f and 5f to the fundamental frequency f in just the right amounts
and phase produces a triangle-like waveform.

167 Single licence - Abingdon School

 5.6.7 Digital representation of sound

Figure 5.6.7.10 shows a Frequency f Amplitude 0.8106

screenshot of a triangle-
like waveform being
synthesised in ESynth Combining the fundamental frequency f
with two harmonics 2f and 3f
using a fundamental of Frequency 3f Amplitude 0.0901
Phase 90o
200 Hz and harmonics
of 600 Hz and 1000 Hz. +
Joseph Fourier was the
Frequency 5f Amplitude 0.0324
first person to realise
Result is a triangular-like waveform
that complex periodic
waveforms could be Figure 5.6.7.9 Fundamental + two harmonics = triangle-like waveform
synthesised in this way.
It led to the concept of bandwidth. To preserve
the shape of this signal any communication system
through which it passes must pass not only the
fundamental frequency but also its harmonics, 600
Hz and 1000 Hz, i.e. frequencies located in the
band 0 to 1000 Hz.
Lower limit on sampling rate - Nyquist’s theorem
We have seen that to achieve a better
approximation to the original signal we need to
sample at a higher rate but is there a lower limit?
Figure 5.6.7.10 ESynth screenshot
The answer is yes. Figure 5.6.7.11 shows a 2.5 Hz
sine waveform sampled at 4 samples per second (4 Hz). The sampling points in the sampling cycle are A, B, C, D.

Sampling rate = 4 samples per second

Rotating blue One cycle
clockhand 1 Hz
produces
2.5 Hz signal
A B C D A 6 7
2.5 cycles per second

Sampled B
Z
Spurious frequency Time/s
C A
frequency
D

Sampling clock

1 Hz
0 1 2 3 4 Hz 2.5 cycles
Sampling frequency 2.5 Hz 2.5 Hz
The fundamental frequency of the 2.5 Hz signal Reconstructed
when reconstructed from the samples does not
match the original’s
Figure 5.6.7.11 Sampling a waveform at a sampling frequency which is less than twice the waveform’s
frequency results in an alias (spurious) frequency replacing the sampled waveform’s frequency

Single licence - Abingdon School 168

5 Fundamentals of data representation

However, the waveform constructed from these samples has a repeating pattern
Key principle
frequency which is not 2.5 Hz. Its frequency is approximately 1.25 Hz. This is
Nyquist’s theorem: an artifact called a spurious or alias frequency, i.e. one that does not really exist.
When sampling a (complex)
This known as aliasing. However, the waveform that could be constructed
periodic waveform, we must
sample at twice the highest from samples of the 1 Hz sine waveform (red dotted curve) does have a
frequency present in the repeating pattern frequency of 1 Hz.
waveform, at least, if all the
frequencies present in the It turns out that when sampling a (complex) periodic waveform, we must
(complex) periodic waveform sample at twice the highest frequency present in the waveform, at least, if all
are to be preserved. the frequencies present in the (complex) periodic waveform are to be preserved.
This is known as Nyquist’s theorem. Figure 5.6.7.12 illustrates this with a
sinusoid (cosine waveform) of frequency 1 Hz. The sampling rate is twice this
Task at 2 samples per second but it is still possible to construct a waveform with
Sampling a rotating image at fundamental frequency, 1 Hz, the same frequency as the original.
too low a frequency can result in
the rotating image appearing to Sample points A and B Sampling rate = 2 samples in 1 cycle
rotate at a lower frequency than it Signal = 2 samples per cycle
actually is. amplitude Reconstructed signal
A
Try observing rotating ceiling fan
blades whilst blinking your eyes.
The fan blades may appear to Time
rotate at a lower frequency than
they really are. We call the false
frequency of rotation a spurious
B
frequency. It is a consequence of
Nyquist’s theorem.
Original signal

1 cycle First cycle consists of samples A and B

Figure 5.6.7.12 Applying Nyquist’s theorem, sampling rate is at least twice
highest frequency in waveform

Questions
1 Explain the terms (a) sampling rate (b) bit depth or sample resolution.

2 An analogue waveform made up of the following sinusoids with frequencies 1 kHz, 5 kHz, and 10 kHz is
sampled and the samples digitised. When the digitised result is processed, it is discovered that it is made up
of sinusoids with frequencies, 1 kHz, 5 kHz and 7.5 kHz but not 10 kHz. Suggest the most likely reason
why this has happened and suggest one possible solution.

3 Why are music CDs recorded at a sampling rate of 44100 samples per second?

4 State Nyquist’s theorem.

169 Single licence - Abingdon School

 5.6.7 Digital representation of sound

Nyquist’s theorem and recording sound Did you know?

Music CDs are recorded at a sampling rate of 44100 samples per second for a The seemingly arbitrary choice
good reason. The human ear is capable of hearing sound over a frequency range of of 44100 Hz arose in order to
20 Hz to 20 kHz with its greatest sensitivity to frequencies between 2000 and accommodate early Video Tape
5000 Hz. Thus the sampling rate at which music is recorded for music CD recorders.
production is greater than the minimum sampling frequency according to See e.g. https://en.wikipedia.
Nyquist’s theorem. A note from a violin pitched at 2000 Hz still must be sampled org/wiki/44,100_Hz

at 44100
samples per
Did you know?
The most general term for a
second because
frequency component of a
what makes complex tone is a “partial” or
the note sound “overtone”. All harmonics are
like a violin partials, but partials can be
note are the either harmonic or inharmonic.
harmonic Instruments such as bells, and
frequencies other metallophones, generate
Amplitude (dB)

a multitude of inharmonic
that are also
partials. Waveforms containing
present. This Figure 5.6.7.13 Spectral content of the inharmonic partials will not
is called the sound “Laa” recorded at 44100 samples per
be periodic at the fundamental
second
quality of frequency, but may or may not
the note or be over a longer time.
timbre that
distinguishes Did you know?
Frequency (kHz)
it from other Modified Shannon-Nyquist
Theorem:
sounds of the same pitch and volume, e.g. a violin note from a trumpet note. The
States that the highest frequency
fundamental plus harmonics of a sound are called the spectral content. Figure
component in the source must
5.6.7.13 shows the spectral content of a recording of the sound “Laa”. A Discrete be less than half the sampling
Fourier Transform(DFT) has been applied to the recorded waveform to reveal frequency.
a fundamental at 155 Hz and 22 harmonics, some of which are not displayed
because their amplitude is too small. The highest harmonic frequency is 3895 Information
Hz. Fourier analysis is the process of finding which sine waves need to be added Discrete Fourier Transform:
together to make a particular waveform shape. The DFT works with digital Converts from the time domain
samples. If the sound “Laa” had been sampled at, say 4000 samples per second, to the frequency domain and lists
which is not at least twice the frequency of the highest frequency and ten other the frequency components of the
harmonics, the recording would have been distorted by including frequencies not signal.

in the original - see Figure 5.6.7.11.

Calculate sound sample sizes in bytes - See Chapter 5.6.3.
In this chapter you have covered:
■ The digital representation of sound in terms of:
• sample resolution or bit depth which is the number of bits allocated to each sample.
• sampling rate or sampling frequency is the number of samples taken per second.
• Nyquist theorem -when sampling a (complex) periodic waveform, we must sample at twice the highest
frequency present in the waveform, at least, if all the frequencies present in the (complex) periodic waveform
are to be preserved. If we don’t then spurious(false) frequencies appear called alias frequencies and their
corresponding original frequencies do not.

Single licence - Abingdon School 170

5 Fundamentals of data representation
6 Representing images, sound and other data
Learning objectives:
■ Describe the purpose of MIDI
and the use of event messages
■ 5.6.8 Musical Instrument Digital Interface(MIDI)
in MIDI What is MIDI?
MIDI stands for Musical Instrument Digital Interface. It is a hardware/
■ Describe the advantages of
software protocol adopted in the 1980s to enable electronic instruments to
using MIDI files for
communicate with each other using the same set of agreed-upon codes and
representing music
numbers. For example, a Korg keyboard (MIDI controller) can instruct suitable
Key principle software running on a computer (MIDI instrument) to play a note by sending
MIDI: MIDI stands for
the software a “Note On” message. Figure 5.6.8.1 shows a Korg MIDI 61-key
Musical Instrument Digital
Interface. It is a hardware keyboard connected via USB (using a USB-MIDI driver) to a computer
and software specification for running an emulator for a Korg synthesiser called Wavestation.
the exchange of information
(musical notes, expression
control, etc) between different
musical instruments or other
devices such as sequencers,
computers, lighting controllers,
etc.

Korg Wavestation
synthesiser
emulator running
on Windows 7
USB cable

Figure 5.6.8.1 Korg 61-key keyboard connected to Korg’s Wavestation synthesiser emulator

Information Pressing a key on the Korg keyboard sends a message to the computer program
Introduction to computer to play the note corresponding to this key. A note number is assigned to each
music: key on a MIDI keyboard. For the keyboard in Figure 5.6.8.1, note numbering
http://www.indiana.edu/~emusic/ starts at 36 and runs consecutively up to 95 as shown in Figure 5.6.8.2.
etext/toc.shtml
MIDI note number 60 has been assigned to middle C on this keyboard. Note,
number 60 corresponds to frequency 261.63 Hz but the MIDI specification
allows this mapping to be changed.
Single licence - Abingdon School 171
 5.6.8 Musical Instrument Digital Interface(MIDI)

Table 5.6.8.1 shows the usual correspondence between MIDI note number and frequency for this and some other
MIDI note numbers. The name given to each note is also shown. Note that musical pitch (note frequency) is
not embedded in any way in MIDI Note messages, thereby allowing mapping from note number to pitch to be
changed. Nor is note name tied to a specific frequency (tuning a musical instrument adjusts frequency).
37 39 42 44 46 49 51 54 56 58 61 63 66 68 70 73 75 78 80 82 85 87 90 92 94

Middle C
D4 E4 F4 G4 A4 B4
C2 C3 C4 C5 C6 C7

36 38 40 41 43 45 47 48 50 52 53 55 57 59 60 62 64 65 67 69 71 72 74 76 77 79 81 83 84 86 88 89 91 93 95 96

Figure 5.6.8.2 Some notes from an octave (white and grey keys only), their MIDI note
number and one possible assignment of frequencies
MIDI 60 62 64 65 67 69 71 72
note no
Frequency Hz 261.63 293.67 329.63 349.23 392.00 440 493.88 523.25
Note name C4 D4 E4 F4 G4 A4 B4 C5
Table 5.6.8.1 MIDI note no, its usual corresponding frequency in Hz and its usual note name
Information
MIDI itself does not make sound. It is just a series of messages to turn notes on
and off, etc. These messages are interpreted by a MIDI instrument to produce Octave:
The range from C4 to C5 is an
sound. A MIDI instrument can be a piece of hardware (a synthesizer) or a
octave. The grey keys indicate
software tool (Wavestation emulator, MuLab, Logic Pro). the musical interval of an octave
The most common tool used to generate MIDI messages is an electronic (12 semitones) between notes.
keyboard. These messages may be routed to a digital synthesiser inside the An octave has the property that
the ratio of the frequencies at the
keyboard or they may be patched (wired) to some other MIDI instrument such as
ends of the range is 2:1. In equal
a computer running synthesiser software. Almost all MIDI devices are equipped to temperament, an octave is defined
receive MIDI messages on one or more of 16 selectable MIDI Channel numbers, to be 12 equal semitones in the
labelled 1 to 16 (supports “multi-timbral” performance). modern scale. Each semitone
1/12
therefore has a ratio of 2
MIDI messages
(approximately 1.059). Note A4
The most common MIDI messages are Voice Channel messages. Voice Channel is assigned frequency 440 Hz.
messages convey information about whether to turn a note on or off on a Therefore, the frequency of the
particular channel, what instrument sound to change to, and so on. nth semitone above or below A4
Voice Channel MIDI messages consist of two or three bytes as shown in Figure is
2n/12 x 440 Hz.
5.6.8.3 (Status byte followed by one or two Data bytes). For the serial hardware
interface, each byte is surrounded by a start bit and a stop bit, making each packet
10 bits long. Within a MIDI software system data is 8-bit bytes. The first byte, called the Status byte, takes on
values ranging from 0x80 to 0xFF in hexadecimal or 128 to 255 in decimal - most significant bit (MSB) is ‘1’. The
Data bytes, take on values in the range 0x00 to 0x7F or 0 to 127 - most significant bit of each byte is a ‘0’.
The transmission bit rate of the hardware interface in the MIDI standard is 31,250 bits per second. Therefore, one
start bit, eight data bits, and one stop bit result in a maximum transmission rate of 3125 bytes per second.
MIDI uses the fact that the Status byte is in a different range from the Data bytes. If MSB = 1, the byte is a “Status”
byte. If MSB = 0, the byte is a “Data” byte. The first four bits of a Status byte are the code for the command, and
the last four bits the channel to which the command applies (e.g. 00002 is Channel 1, 11112 is Channel 16).

Single licence - Abingdon School 172

5 Fundamentals of data representation

For example, when a key is pressed, the keyboard creates a “Note On” (Status byte = 0x91, 14510, 100100012)
message for Channel 2 consisting of three bytes, e.g. 145 45 100. The first four bits of the Status byte (10012) tell
MIDI that the message is a Note On command, while the last four bits tell MIDI what MIDI channel the message
is for (00012= MIDI Channel 2).

Loudspeaker

Virtual Machine
Channel 2 Wavestation
Piano Keyboard
MIDI
(VMPK) 145 45 100
instrument

LoopMIDI
virtual MIDI
connection

144 43 58 Channel 1

Korg 61-key keyboard

Figure 5.6.8.3 Two channel MIDI

system

Figure 5.6.8.4 MIDI-OX showing other

status byte values
The second byte, the nore number 60, selects the frequency used by the receiving Information
instrument, in this example middle C (261.63 Hz). The third byte, 85, specifies
VMPK:
how fast the key was pressed (velocity).
Information on VMPK can be
Velocity is a number that is used mainly to describe the volume (gain) of a viewed at
MIDI note (higher velocity = greater volume or loudness) because it refers to how http://vmpk.sourceforge.net/.
hard a key was pressed. The harder a key is pressed It can be downloaded from http://
Information vmpk.sourceforge.net/#Download
the greater will be the volume or loudness but the
mapping is performed by the receiving instrument. MIDI-OX:
http://www.midiox.com/

173 Single licence - Abingdon School

 5.6.8 Musical Instrument Digital Interface(MIDI)

To generate messages that use different velocities requires a MIDI keyboard. Computer keyboards are not velocity
sensitive. Using a computer’s keys to play notes into a software synthesiser will generate note messages that all have
the same velocity.
When a key is released the keyboard creates another MIDI message, a
Information
“Note Off” message, e.g. 129 60 85. The first byte, 129, is the Status
byte - the first four bits of the Status byte (10002) correspond to “Note LoopMIDI:

Off”, the second four bits (00012) to the channel, i.e. Channel 2 (0000 is Information on LoopMIDI can be viewed
at http://www.tobias-erichsen.de/software/
Channel 1). The second byte is the key, 60 (middle C) in this example, loopmidi.html where it can also be
and the third byte is the velocity which indicates how quickly the key downloaded.
was released. The MIDI instrument can use the velocity value of 85 to
know how quickly it should dampen the note.
Figure 5.6.8.3 shows an on screen piano keyboard (VMPK) connected Information
via a virtual MIDI connection (LoopMIDI) to a running copy of the
Korg Wavestation:
Wavestation emulator. A Korg 61-key keyboard is connected via USB to Information on the Korg Wavestation is
the computer and Wavestation emulator. The output from Wavestation available at http://www.korg.com/us/products/
goes to a loudspeaker. The input channels 1 & 2 are monitored by a software/korg_legacy_collection/. It is not free
software. MULAB is an alternative that is free
piece of software called MIDI-OX.
- http://www.mutools.com/mulab-downloads.
In Figure 5.6.8.3 MIDI-OX shows that the “Note On” and “Note Off” html
Status bytes for Channel 1 have codes 144 and 128, respectively, whilst
for Channel 2, these are 145 and 129, respectively. Wavestation uses the
Channel values to route the messages on each Channel to different voices.
Messages can also have other purposes, e.g. to change the instrument sound. Figure 5.6.8.4 shows messages that do
this, e.g. 193 41 which is a two-byte code to change the MIDI instrument for Channel 2 (code 193 - 1100 00012,
11002 is the change code and 00012 is the Channel) to a viola (code 41). Such messages are control messages.

Pitch Bend (control to vary pitch) is another type of control message that a MIDI controller can send, e.g. 225 43
0 (Figure 5.6.8.4), causes pitch bend in a Channel 2 note. For Channel 1 the same control message would use a
leading byte with value 224 (1110 00002). There are many more ways to control the playing of a note and each has a
corresponding control code.
The playing of multiple notes “together” events in MIDI are sent as a string of serial commands so, for example, a
2-note chord will be transmitted as two separate messages, Status(Note
While True
On, Ch 1) key1-velocity Status(Note On, Ch 1) key2-velocity unless the
Do
synthesiser supports Running Status. In this case, a single Status byte’s
Wait for message
action is allowed to persist for an unlimited number of Data byte pairs
Process message
which follow.
MIDI messages Figure 5.6.8.5 Event handler
A MIDI message is the means by which an event in one system, e.g. key
pressed on a keyboard, is communicated or transported to another to produce an Information
event in the receiving system, e.g. a synthesiser plays a note. Chuck download:
http://chuck.cs.princeton.edu/
Keyboard-event → MIDI message → synthesiser event.
release/
For example, the “Note On” message sent by a MIDI controller to a MIDI
instrument causes an event to take place, i.e. the synthesiser plays the note

Single licence - Abingdon School 174

5 Fundamentals of data representation

specified in the message by note number. The Virtual MIDI Piano Keyboard shown in Figure 5.6.8.3 is a MIDI
events generator and receiver. Event-driven systems rely upon a piece of software called an event handler which
consists of a non-terminating loop that “sleeps” when there are no messages to process, i.e. is suspended - Figure
5.6.8.5 - and springs into action when there is (in Chuck the loop takes the form of a “polling loop”).

Extension material
Chuck is an open-source and freely available programming language for real-time sound synthesis and music creation.
Figure 5.6.8.6 shows an event handler written in Chuck.
MidiIn midiIn; //create an event object
0 => int port; // select MIDI port 0
if( !midiIn.open(port) ) // if MIDI port 0 not open exit
{
<<< “Error: MIDI port did not open on port: “, port >>>;
me.exit();
}
MidiMsg msg; // makes object to hold next MIDI message
Wurley piano => dac; // select Wurley piano to play with MIDI controller
while( true ) // loop forever
{
midiIn => now; // wait on MIDI event, shred suspended but time advances
while( midiIn.recv(msg) )
{
if (msg.data1 == 144) //check that status byte = 144 which is Note On Channel 1
{
Std.mtof(msg.data2) => piano.freq; //convert MIDI no to corresp. frequency
msg.data3/127.0 => piano.gain; //set piano gain (data3 in range 0 to 127)
1 => piano.noteOn; //trigger note on
}
else //status byte not equal to 144 so switch note off
{
1 => piano.noteOff; //trigger note off
}
} Figure 5.6.8.6 Event handler written in Chuck programming language
}

Chuck
Virtual
LoopMIDI Virtual Loudspeaker
keyboard
Machine

Figure 5.6.8.7 Playing music via an executing Chuck event handling program
Figure 5.6.8.7 shows the use of VMPK virtual piano connected via LoopMIDI to the running Chuck event handler shown in
Figure 5.6.8.6. The output of the Chuck program is sent to a loudspeaker connected to the computer.

175 Single licence - Abingdon School

 5.6.8 Musical Instrument Digital Interface(MIDI)

Advantages of using MIDI files for representing music

MIDI consists of a series of event messages that instruct a MIDI controlled instrument how to play music. These
messages can be stored in a file before being read from the file and transmitted serially byte by byte to a
MIDI-controlled instrument.
This has four main advantages over audio data produced from analogue sounds by sampling thousands of times per
second and recording the digitised samples (sounds) in, for example, a .wav file.
• compact compared to sampled audio data. With MIDI, an entire song can be stored within a few hundred
MIDI messages saving on memory whilst the equivalent sampled audio data would occupy many more
bytes, possibly millions
• easy to modify/manipulate notes, e.g. change pitch, duration, and other parameters without having to
record the sounds again which would be the case with sampled audio data recordings
• easy to change instruments - MIDI only describes which notes to play, these notes can be sent to any
instrument to change the overall sound of the composition whilst with sampled audio data the sampling
and recording process would have to be repeated
• it offers a simple means to compose and notate algorithmically which sampled audio does not. MIDI data
is mostly a glorified note list, and such lists can easily be generated by code, and translated as needed into
MIDI, whether for live output or via a MIDI file.

Questions
1 What is MIDI?

2 “MIDI itself does not make sound”. Explain this statement.

3 The following MIDI message consisting of three bytes is generated when a key is pressed on a MIDI
keyboard: 144 60 64
Explain the purpose of each of three bytes.

4 Note On is one example of a MIDI message. Give three other examples, each must be a different type.

5 Explain the statements “A MIDI keyboard is an events generator” and “MIDI messages are associated with
events”. What is the fundamental structure of event-handling software such as that found running in a
MIDI instrument?

6 State three main advantages of MIDI file representation of music over audio data file representation, e.g.
.wav file

In this chapter you have covered:

■■ The purpose of MIDI - to instruct via messages a MIDI controlled instrument how to make sound, e.g. Note
On, Note Off, pitch, duration of note, loudness
■■ The use of event messages in MIDI - MIDI controller sends messages to a MIDI controlled instrument to turn
notes on and off, etc. These are events that a MIDI controlled instrument responds to. It waits in a loop for
messages and then acts on these received messages accordingly
■■ The advantages of using MIDI files for representing music - compact, easy to modify/manipulate and change
instruments compared with sampled audio data stored in, e.g., .wav files

Single licence - Abingdon School 176

 5 Fundamentals of data representation
6 Representing images, sound and other data
Learning objectives:
■■Know why images and sound
files are often compressed and
■■ 5.6.9 Data compression
that other files, such as text Why are images, sound files and other files compressed?
files, can also be compressed There are two main reasons why files are compressed:
• To reduce the amount of storage space required to store the data
■■Understand the differences
between lossless and lossy • To reduce the time taken to transmit the data because fewer bytes need
compression and explain the to be transmitted.
advantages and disadvantages Essentially, the purpose of data compression is to squeeze the data into a
of each smaller number of bytes than the data would occupy if uncompressed.
■■Explain the principles behind For example, text may be compressed by replacing each common character/
the following techniques for letter combination with a single byte-coded integer number from Table 5.6.9.1.
lossless compression:
Character
• run length encoding Integer Code
Combination
(RLE) 1 ‘TH’
• dictionary-based methods 2 ‘BL’
3 ‘CK’
Key principle 4 ‘AT’
5 ‘ON’
Compression:
Data can be compressed because Table 5.6.9.1 Codes for common character combinations
its original representation is
Uncompressed text = ‘THE BLACK CAT SAT ON A MAT’
not the shortest possible. The
original data has redundancies Compressed text = ‘1E 2A3 C4 S4 5 A M4.’
and compressing the data If each character in the uncompressed text is coded in one byte (including
reduces or eliminates these
spaces and full stop) then this text requires 27 bytes of storage. For the
redundancies.
Non-random data is non- compressed text the storage requirement is just 20 bytes, a saving of seven
random because it has structure bytes. This represents a 26% saving, approximately.
in the form of regular patterns. Not every file can be compressed significantly, in fact most files cannot.
It is this structure that is the
cause of redundancy in the Beyond A level
data. Random data has no Suppose that we arbitrarily but quite reasonably decide that significantly means at least 50%
structure and therefore has no or greater, i.e. an n-bit file should be compressed to one of length n/2 or less. There are
n
redundancy. Therefore, random 2n of these n-bit files. The number of n/2 bit compressed files will be 2 /2 , the number of
n/ - 1 bit compressed files will be 2n/2 - 1, ....., the number of 2-bit files will be 22 = 4, the
data cannot be compressed. 2
number of 1-bit files will be 2.
n
Total number of compressed files, S = 2 + 4 + ... + 2 /2 = 21 + n/2 - 2 ≈ 21 + n/2
For n = 800 bits (100 bytes), the total number of different files is 2800 and the number of
files that can be compressed by at least 50% or more is 251.
51
The fraction of compressed files is thus 2 / 2800 = 2-749 ≈ 3 x 10-226. This is an
extremely small number. Further analysis shows that no compression method can compress
all files or even a significant percentage of them.

Single licence - Abingdon School 177

5 Fundamentals of data representation

The redundancies in data depend on the type of data (text, images, audio, etc)
which is why different compression methods have been developed. Each works
best with a particular data type.

Questions
1 What does it mean to compress data?

2 Why is it possible to compress data that has structure without losing

Key principle
information?
Lossless and lossy
compression: 3 Give two reasons why files are compressed.
Data is how information is
represented. 4 Can random data be compressed?
It is possible using compression
to alter the representation 5 Why is it necessary to have different compression methods?
without losing information
- this is called lossless
compression.
It is also possible using What are the differences between lossless and lossy
compression to alter the compression?
representation and lose Lossless compression
information - this is called lossy In lossless compression, the compression algorithm does not remove
compression.
information from the original uncompressed data only redundancies. This
allows the original uncompressed data to be restored by reversing the process.
Lossless compression is used for text because it must be compressed without
any loss of information. Imagine uncompressing an essay that you wrote for an
Key concept assignment and finding that it looked nothing like the original that you spent
hours constructing.
Lossless compression and
redundancy: Lossy compression
In general, information can be In lossy compression, the compression algorithm may remove information
compressed if it is redundant. which is irrelevant from the original uncompressed data. For example, in
Lossless compression is possible
audio data, harmonics to which the human ear is not sensitive may be removed
when information is redundant.
because they are not important to the listener. However, this means that the
original uncompressed data cannot be fully restored when the reverse process is
carried out. This does not matter for most images, video and audio data because
Key concept these can tolerate much loss of data when compressed and later decompressed.
Lossy compression and Some exceptions are text files, executable files, and medical X-ray images where
irrelevancy: artefacts introduced into lossy compressed images could matter.
Even when no redundancy
Advantages and disadvantages of lossless and lossy compression
exists it is still possible to
compress by removing irrelevant Better compression ratios can be achieved with lossy compression than with
information, e.g. removing lossless compression. Compression ratio is the size of the compressed file as a
image features to which the eye fraction of the uncompressed file, e.g. 50% expressed as a percentage.
is not sensitive.
This means that data compressed with a lossy compression method will occupy
less storage space than with a lossless compression method. The time taken to
transmit the data will also be less, e.g loading a file from disk.

178 Single licence - Abingdon School

 5.6.9 Data compression

However, a disadvantage with lossy compression is that the lost data are not retrievable. The compressed data will
have very limited potential for adjustments or changes and every time the compressed data is uncompressed, edited,
compressed again and saved, more data is lost.
With lossless compression the original uncompressed data is always recoverable.
Online high-quality image retailers often display their images in low quality form, i.e. they use a lossy, compressed
version of high compression ratio, so potential customers can view what is on offer before purchasing. This protects
against theft of data as it prevents customers from accessing and downloading a higher-quality version. It is the
ability of lossy compression methods to allow the compression ratio to be varied from low to high that supports this
way of marketing images. Alternative, an uncompressed or lossless version can be made available to customers on
receipt of payment.

Questions
6 What is meant by lossless compression?

7 What is meant by lossy compression?

8 State one advantage that lossy compression has over lossless compression.

9 State one advantage that lossless compression has over lossy compression.

10 For each of lossless and lossy compression, give one example where it is used and why.

Principles of lossless compression

Run length encoding (RLE)
In run length encoding a run of contiguous bytes all with the same value can be condensed into two bytes, one
byte that stores the count or run length and a second byte that stores the value in the run. These two bytes are called
an RLE packet. Figure 5.6.9.1 shows run length encoding applied to a run of six contiguous bytes each of value
128.
Run of 6 bytes 2 bytes
128 128 128 128 128 128 6 128

Figure 5.6.9.1 Run length encoding compression of 6 bytes into 2 bytes

Information
RLE can be used to compress greyscale images. Each run of pixels of the same
Contiguous:
intensity (gray level) is encoded as a pair (run length pixel value). It doesn’t make
Means next to each other or
sense to encode a run of one and so the raw value is used. The following example together in sequence.
shows how RLE could be applied to a greyscale bitmap that encodes the gray level
of each pixel in 8 bits and that starts with the sequence
15, 15, 15, 15, 15, 15, 15, 15, 46, 81, 123, 58, 98, 98, 98, 98, 7, 7, 7, 8, ...

The compressed sequence of bytes is

8, 15, 46, 81, 123, 58, 4, 98, 3, 7, 8, ...

where the red values indicate counts. The problem is to distinguish a byte containing a greyscale value (such as 15)
from one containing a count (such as 8). There are several possible solutions.

Single licence - Abingdon School 179

5 Fundamentals of data representation

In one solution, the 256 different greyscale values are reduced to 255 so that the 256th can be used as a flag to
precede every byte containing a count. Suppose this flag value is 255 then the sequence above becomes
255, 8, 15, 46, 81, 123, 58, 255, 4, 98, 255, 3, 7, 8

RLE works well with images that contain large areas of the same colour e.g. black and white images which are
mostly white, such as the page of a book. This is due to the large amount of contiguous bytes that are all the same
colour.
However, an image with many colours and relatively few runs of the same colour such as a photograph containing a
high degree of colour variation will not lend itself to compression using RLE so well.
The direction of scan can also affect the compression ratio. For example, an image that has lots of vertical lines will
not compress well if it is scanned horizontally for same-pixel runs but will if scanned vertically. A good RLE image
compressor should be able to scan a bitmap by rows, columns, or in a zig-zag pattern and be able to choose the scan
output that produces the best compression ratio.

Questions
11 Explain the principles of run length encoding lossless compression.

12 The following numbers, restricted to the range 0..254, represent the intensities of a contiguous block of
pixels in a greyscale bitmap
15, 112, 112, 112, 98, 76, 76, 15, 46, 46, 46, 46, 46, 19, 101, 6, ...

Using run length encoding, compress this block of pixels using 255 as the flag that prefixes an RLE packet.

13 Run length encoding works well, i.e. achieves a good compression ratio, with some images but not others.
Why?

14 Why is run length coding normally not a good choice for text compression?

Dictionary-based methods
We compress naturally in everyday life when, for example, referring to months of the year by number, e.g.
September by the number 9. Dictionary-based methods compress by using this technique. The dictionary is a kind
of look-up table, e.g. entry 9 is September. Dictionary-based compression methods vary in how the dictionary is
constructed and represented but they all use the principle of replacing substrings in a text, e.g. ‘th’ in ‘the’ with a
codeword, e.g. 1, that identifies that substring in a dictionary or codebook - see
Key concept
Table 5.6.9.1. The substring is called a phrase. Codewords for the dictionary
Token: are chosen so that they need less space than the phrase that they replace, thus
A unit of data written on
achieving compression. The process of compression is called encoding. The reverse
the compressed file. A token
consists of two or more fields.
process is called decoding. The compressor is an encoder and the decompressor is
In LZ78, the token consists of a decoder.
two fields, the first is a pointer
If we have to use a dictionary containing a large number of entries then the
to an entry in the dictionary
and the second is the code of a
overhead of storing or transmitting the dictionary is significant, and choosing
symbol, e.g. “A”. which substrings to place in the dictionary to maximise compression is also
A token is sometimes written difficult.The solution is to use an adaptive dictionary scheme based on methods
surrounded by chevrons < and > developed by Jacob Ziv and Abraham Lempel in the 1970s.
e.g. <2, A>

180 Single licence - Abingdon School

 5.6.9 Data compression

In the 1978, Ziv and Lempel described a dictionary-based algorithm (LZ78) that encodes a phrase (substring) of
n characters from the input in a codeword that points back to an earlier phrase in the input which it matches in
all but the last character, e.g. the B in BA matches B at
Input string A B AA BABAA
position 2 in the example shown in Figure 5.6.9.2 so
Dictionary
is encoded as the two-field token 2, A. The first field of
A B AA BA BAA
the token is the pointer 2 and the second the code of the Index 1 2 3 4 5
symbol, e.g. ASCII A. A B AA BABAA

The dictionary starts empty with the empty string at Encoder 0, A 0, B 1, A 2, A 4, A

output
position 0 (not shown in Figure 5.6.9.2). As substrings
Decoder
or phrases are read and encoded, phrases are added to the A B AA B A BA A
output
dictionary at positions 1, 2, and so on. For the given input 0 means no dictionary entry,
i.e. nothing for a substitution
string, ABAABABAA, the first phrase consisting of the
Figure 5.6.9.2 Lev-Zimpel 1978 (LZ78) compression
single character A is added at position 1, the next, B, at
simple example
position 2.
This happens because when the first substring, A, is read
from the input, no dictionary entry with the one-character string A is found, so A is added at the next available
position in the dictionary which is 1, and the token
0, A is output. This token indicates the string empty string followed by A.

The next symbol read from the input is the character B but there is no entry yet in the dictionary for this phrase,
and so B is added at position 2.
The third character read from the input, is A. This is matched with the A at position 1 in the dictionary. The goal
of dictionary encoding is to find the longest dictionary substring that matches the input so the next symbol is read.
This is another A. The dictionary is now searched for an entry containing the two-symbol string AA.
None is found, so the string AA is added to the next available position in the dictionary which is 3, and the token
1, A is output. This is the “compressed” version of the substring AA. We actually need to build up phrases in the
dictionary with at least three characters before we can replace a phrase with something shorter. Figure 5.6.9.3
shows that this happens at entry 5 in the dictionary.
An efficient way of representing the dictionary is a tree-like structure called a trie as shown in Figure 5.6.9.3 which
grows as more characters of the input are processed. All strings that start with the empty string (dictionary index 0)
are added as children of the root which is labelled 0.
Information
In the example, all strings that begin with A are located in the subtree with node
Trie:
labelled 1 (index 1 in the dictionary).
A tree data structure in which the
All strings that begin with B are located in the subtree with root labelled 2. string of characters represented
by node n is the sequence of
All strings that begin with AA are located in subtree with root labelled 3 and so
characters along the path from
on. the root to n. Given a string, the

The example in Figure 5.6.9.2 is too trivial to achieve a reduction in the number trie consists of nodes for exactly
those substrings that are prefixes
of bytes (assuming each character is represented in one byte). We need a much
of some other substring.
longer input string to achieve compression. The word ‘trie’ comes from the
The dictionary shown in Figure 5.6.9.2 and Figure 5.6.9.3 is constructed as middle of the word ‘retrieval’.

the input string is parsed (processed). This dictionary is empty when the first

Single licence - Abingdon School 181

5 Fundamentals of data representation

Dictionary Encoded Empty string character, A, of the input string is read. The characters about to be
output
0 encoded are used to traverse the tree until the path is blocked, either
0 Empty
string
1 A <0, A> A B
because there is no onward path for the current character or because a
2 B <0, B> 1 2 leaf is reached.
3 AA <1, A> A A
The node at which the block occurs gives the index/phrase number to be
4 BA <2, A> 3 4
used in the output, e.g. 1 in <1, A>.
5 BAA <4, A>
A A new node, e.g. 3, is added and joined by a new branch to the node at
5 which the block occurred.
Figure 5.6.9.3 Data structure for The new branch is labelled with the last character of current string, e.g.
LZ78 coding - numbers in nodes refer A.
to dictionary index
For example, suppose we append BAB to the input string to form the
new input ABAABABAABAB. We need to add a new node 6 to Figure 5.6.9.3 and connect it to node 4. The path
from node 4 to node 6 is labelled BAB.
The encoder output is now the string of tokens
0, A 0, B 1, A 2, A 4, A 4, B

If we extend the input with BABB to form input ABAABABAABABBABB the encoder output is now the string of
tokens
0, A 0, B 1, A 2, A 4, A 4, B 6, B
The compressed form (ignoring commas which have been included to aid readability) occupies less space than the
uncompressed form, 14 bytes compared with 16 bytes.
On decoding, the decoder reconstructs the tree data structure so it can decode the token string - Figure 5.6.9.4.
Empty string Empty string Empty string Empty string
0 0 0 0

A A B A B
1 1 2 1 2
A A
Figure 5.6.9.4 LZ78 decoder reconstructs the 3 4
dictionary tree from the token string
A B
0, A 0, B 1, A 2, A 4, A 4, B
5 6
token by token
Dictionary Encoded
Empty string output
Our examples have used an alphabet of two symbols 0 0 Empty
string
A and B but what if the alphabet of symbols was A, A B E C 1 A <0, A>
B, C, D, E? 1 2 4 5

B
2 B <0, B>
The tree data structure dictionary would then A D 7 3 AA <1, A>
3 6
consist of a root for the empty string and then all E 4 E <0, E>
strings that start with the empty string (strings 8
5 C <0, C>
for which the token pointer is zero) are added to 6 BD <2, D>
Figure 5.6.9.5 LZ78 coding
the tree as children of the root. Figure 5.6.9.5 7 CB <5, B>
dictionary tree constructed for
shows the dictionary tree for the input string
ABAAEACBDCBBDE 8 BDE <6, E>
ABAAEACBDCBBDE.

182 Single licence - Abingdon School

 5.6.9 Data compression

Using the token structure <pointer, symbol> the output of the encoder for this
input string is,
<0, A> <0, B> <1, A> <0, E> <0, A> <0, C> <2, D> <5, B> <6, E>

Questions
15 Explain the principles of dictionary-based lossless compression. Empty string
0
16 This tree data structure was created when encoding a string using a dictionary-based
B A
lossless compression technique.
1 2
(a) What is the dictionary? B B

(b) What was the input if the encoded output was 3 4

<0, B>, <0, A>, <1, B>, <2, B>, <0, B>, <1, B> <4, A> <4, B> A B
5 6
17 The output from a dictionary-based encoder is
<0, B> <1, B> <0, A> <1, B> <3, A> <1, D> <1, A> <2, B>

Draw the dictionary tree for this output and then decode the output.

In this chapter you have covered:

■■ Why images and sound files are often compressed and that other files,
such as text files, can also be compressed
• to reduce the amount of storage space required to store the data
• to reduce the time taken to transmit the data
■■ The differences between lossless and lossy compression and the advantages
and disadvantages of each
• In lossless compression,
ŠŠ the compression algorithm does not remove information from the
original uncompressed data only redundancies
ŠŠ the original uncompressed data is always recoverable.
• In lossy compression,
ŠŠ the compression algorithm may remove information which is
irrelevant from the original uncompressed data
ŠŠ the original uncompressed data cannot be fully restored when the
reverse process is carried out but compression ratios are higher than
for lossless compression.
ŠŠ limited potential for adjustments or changes and every time the
compressed data is uncompressed, edited, compressed again and
saved, more data is lost
■■ The principles behind the techniques for lossless compression:
• run length encoding (RLE)
• dictionary-based methods

Single licence - Abingdon School 183

 5 Fundamentals of data representation
5.6 Representing images, sound and other data
Learning objectives:
■■ Understand what is meant by
encryption and be able to
■■ 5.6.10 Encryption
define it What is cryptography?
Cryptography has typically concerned itself with methods of protecting
■■ Be familiar with Caesar cipher
information by transforming the contents of messages and documents into
and be able to apply it to
representations called “secret codes” that make the message and document
encrypt a plaintext message
contents incomprehensible except to those granted the means to reverse the
and decrypt a ciphertext
process. A cryptosystem, or cipher, is a system or method for achieving this.
■■ Be able to explain why it is
Figure 6.10.1 illustrates this with two pieces of similar-looking text, one an
easily cracked
encrypted message and the other gobbledegook. The text on the left hand
■■ Be familiar with Vernam side is a message rendered incomprehensible by a process called encryption,
cipher or one-time pad and be i.e. turned into text that resembles random gibberish, whilst the text on the
able to apply it to encrypt a right hand side is random gibberish. The reverse of the encryption process
plaintext message and decrypt is known as decryption. Decryption restores the message to a form that is
a ciphertext comprehensible.

■■ Explain why Vernam cipher is Encrypting a message is one way to keep the message’s contents secret from
considered as a cipher with others who are not authorised to view its contents. The encrypted messages
perfect security look just like random gibberish as illustrated in Figure 6.10.1.

■■ Compare Vernam cipher

VWXGHQWV VKRXOG EH LUBD YEAQJFF TEW PBDNJKFD
with ciphers that depend on IDPLOLDU ZLWK WKH WHUPV FDSFD ND JQHD JBDCJBC
computational security FLSKHU SODLQWHAW DQG EQBVAX VSS MCVN VAJGOTH
FLSKHUWHAW FDHVDU DQG HGVCCB HSXGWEFR GVCBC
YHUQDP FLSKHUV DUH DW BCB BC W APQV VXVBBSCX
RSSRVLWH HAWUHPHV RQH YREGEVD NHEPHJVBBSP NDJ
RIIHUV SHUIHFW VHFUHFB BGBWRSE SFDHJ GDHJBD
WKH RWKHU GRHVQW MJOVQZXGT VBSH HBSV

Figure 6.10.1 A secret message and random gibberish

A cryptographer is someone who uses and studies secret codes (encrypted
messages). On the other hand, someone who analyses other peoples’ secret
codes in order to discover the secret message is a cryptanalyst. Cryptanalysts
are also known as code breakers. The most famous of code breakers is Alan
Turing who during the Second World War was a key member of the team
which broke the German’s Enigma machine coded messages daily during World
War 2, revealing important information that aided the war effort.

Single licence - Abingdon School 184

5 Fundamentals of data representation

What is encryption?
Encryption is the process of obtaining ciphertext from plaintext. Before
Key concept
it is encrypted, the understandable (English) text is normally referred to as
Encryption: plaintext and after encryption it is referred to as ciphertext. The left hand side
Encryption is the process of
of Figure 6.10.1 is an example of ciphertext.
obtaining ciphertext from
plaintext. The encryption process requires two inputs: the plaintext and the key.
The decryption process also requires two inputs, the ciphertext and the key, in
Key concept order to produce as output the plaintext equivalent of the ciphertext.

Decryption: The processes of encryption and decryption are called the cryptosystem or
Decryption is the process cipher. Thus a cryptosystem is a set of rules for converting between plaintext
of obtaining plaintext from and ciphertext.
ciphertext.

Codes versus ciphers

Common parlance and the media apply the term code to the practice and
Key concept science of transforming messages in order to protect a secret when in fact the
Cipher: correct term is cipher.
The processes of encryption Morse code and ASCII are examples of codes.
and decryption are called the
cryptosystem or cipher. Morse code is a system which was designed
A • ▬ S • • •
to allow communication across a telegraph
B ▬ • • • T ▬
link by translating English into electrical
pulse codes in a process called encoding C ▬ • ▬ • I • •
Key point
and electrical pulse codes into English D ▬ • • J • ▬ ▬ ▬
Encoding and decoding are
in a process called decoding. Unlike a E • K ▬ • ▬
processes that are applied
using coding systems that are
cipher system, codes are intentionally F • • ▬ • L • ▬ • •
publicly available and open, understandable and publicly available so G ▬ ▬ • M ▬ ▬
e.g. ASCII, whereas encrypting cannot be used to protect a secret. H • • • • N ▬ •
and decrypting are processes
Figure 6.10.2 shows a sample of Morse Figure 6.10.2 Sample of Morse
in a cipher system which is by
definition closed to all but the code. A short pulse is called a dot (•) and a code
participants using it to exchange long pulse a dash (▬).
secret or private messages or
Caeser cipher
information.
The ciphertext (secret message) shown in Figure 6.10.1 was produced using a
cipher called the Caesar cipher, so named because it is believed that it was first
used by Julius Caesar two thousand years ago. A writer at the time, Suetonius,
Key concept
wrote that Julius Caesar’s cryptosystem replaced the plaintext letter A by the
Caesar cipher: letter D, B by E and so on. The last three letters of the alphabet were replaced,
The Caesar cipher is a shift respectively, by the first three letters of the alphabet. The Caesar cipher is a
cipher which shifts plaintext
type of cipher called a shift cipher because the plaintext letters are shifted to
letters by an amount called the
key to produce ciphertext. produce the ciphertext.
The easiest way of visualising this is to use something called a cipher wheel or
disk to convert plaintext to ciphertext. The wheel consists of an inner wheel of
letters + numbers and an outer wheel of letters as shown in Figure 6.10.3.

185 Single licence - Abingdon School

 5.6.10 Encryption

The outer wheel in Figure 6.10.3 is set to map the plaintext letter to its
equivalent ciphertext letter, e.g. letter A maps to letter D for the current setting.
The encryption key is the number on the inner wheel corresponding to the
letter A on the outer wheel. For the current wheel setting it is the number 3.
The dot under the letter A in the outer wheel is to remind the user that the
encryption key is the corresponding number on the inner wheel.

W X
V Y
U A B Z
Y Z
T X 24 25 0 1

C
23 2

A
20 2 V W

D
22
S

B
1

E
4
R
U

C
F G H
5 6 7 I
Q
T
R 18 19

D E
S
P

Q 17

O
16

9
F

10 15 J
K 11 12 13 14 P G
N

O
L M
H N M
I L
J K

Figure 6.10.3 Cipher wheel showing inner and outer wheels

You can make your own cipher wheel by downloading copies of the inner
and outer wheels from www.educational-computing.co.uk/cipherwheels and
cutting out and pinning these shapes with a brad fastener or you can try an
online version at https://www.khanacademy.org/computing/computer-science/
cryptography/ciphers/a/ciphers-vs-codes.

How to encrypt with the cipher wheel

First write down the plaintext form of the message to be encrypted across the
page, as shown below. Set the cipher wheel for the given key, let’s say 3. For
each letter of the message find the corresponding letter on the outer cipher
wheel then read off the corresponding letter from the inner wheel. Write this
below the plaintext letter as shown below.

COMPUTER SCI ENCE I S COOL

F R P S X W H U V F L HQ F H L V F R RO

Single licence - Abingdon School 186

5 Fundamentals of data representation

Decrypting with the cipher wheel

To decrypt ciphertext, first write it down as shown below. Set the cipher wheel
to the key used to encrypt the plaintext. For each letter of the ciphertext,
find the corresponding letter on the inner cipher wheel then read off the
corresponding letter from the outer wheel. Write this below the ciphertext
letter as shown below. When finished you have the decrypted message.

F R P S X W H U V F L HQ F H L V F R RO

COMPUTER S C I E NC E I S COOL

Questions
1 Convert the following plaintext message using the Caesar
cipher and a key of 7:

THE SUN HAS GOT ITS HAT ON

2 Decrypt the following ciphertext which was produced

using Caeser’s cipher and a key of 5:

MNU MNU MTTWFD

3 Decrypt the ciphertext shown in the left hand side of Figure

6.10.1 which was produced by Caesar cipher using a key of 3.
Online exercises for encryption and decryption using the Caesar
cipher are available at
https://www.khanacademy.org/computing/computer-science/
cryptography/ciphers/e/

Mathematical description
To describe the Caesar cipher mathematically, we represent each letter of the
alphabet by an integer between 0 and 25:
0 for A, 1 for B, …, 25 for Z as shown in Figure 6.10.4.

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

Figure 6.10.4 alphabet and integer equivalent representation

To encrypt the plaintext COMPUTER SCIENCE IS COOL with key 8, first
convert the plaintext letters to their integer equivalent using Figure 6.10.4.
Next add the key 8 to each integer. If the resulting integer is 26 or greater
then subtract 26 to convert 26 to 0, 27 to 1, 28 to 2, and so on, otherwise
leave alone. Finally, convert each resulting integer to its equivalent letter using
Figure 6.10.4 again. The steps of the process are shown in Figure 6.10.5.

187 Single licence - Abingdon School

 5.6.10 Encryption

C O M P U T E R S C I E N C E I S C O O L

2 14 12 15 20 19 4 17 18 2 8 4 13 2 4 8 18 2 14 14 11
+8
10 22 20 23 28 27 12 25 26 1016 12 21 10 12 16 26 10 22 22 19

10 22 20 23 2 1 12 25 0 1016 12 21 1012 16 0 10 22 22 19

KW U X C B M Z A KQM V KM Q A K W W T
Figure 6.10.5 Encrypting with key 8
To decrypt the ciphertext KWUXCBMZ AHQMVKM QA KWWT with key
8, first convert the plaintext letters to their integer equivalent using Figure
6.10.4. Next subtract the key 8 from each integer. If the resulting integer is less
than 0 then add 26 to convert −1 to 25, −2 to 24, −3 to 23, and so on. Finally
convert each integer to its equivalent letter using Figure 6.10.4 again. The steps
of the process are shown in Figure 6.10.6.

KW U X C B M Z A KQM V KM Q A K W W T

10 22 20 23 2 1 12 25 0 10 1612 21 1012 16 0 10 22 22 19
-8
2 14 12 15 -6 -7 4 17 -8 2 8 4 13 2 4 8 -8 2 14 14 11

2 14 12 15 20 19 4 17 18 2 8 4 13 2 4 8 18 2 14 14 11
C O M P U T E R S C I E N CE I S C O O L
Figure 6.10.6 Decrypting with key 8
The number circle
The number circle shown in Figure 6.10.7 is useful for visualising addition and
subtraction performed in the Caesar cipher.
25 0 1
24 2
23
3
22

4
9 20 21

5
7 6
18 1

8
17

10
16

5 11
12 13 14 1

Figure 6.10.7 Number circle

If we add 6 to 22 then we move to position 2 on the number wheel. If we
subtract 6 from 2 then we end up at position 22 on the number wheel as shown
in Figure 6.10.8.

Single licence - Abingdon School 188

5 Fundamentals of data representation

+6
-6 2

22 24
25 0 1
2 2
23

3
22

22

4
9 20 21

5
7 6
18 1

8
17

9
10

16
5 11
12 13 14 1

Figure 6.10.8 Number circle

Modular arithmetic
Key concept Addition using this number circle is called modulo 26 addition. Likewise
subtraction using this number circle is called modulo 26 subtraction. In
Modular arithmetic:
the world of modulo 26, there are exactly 26 numbers, {0, 1, 2, …, 23, 24,
Addition using the number
circle is called modulo n 25}. These are indicated on the number circle as 0 through 25. The number
addition where n is the 26 is called the modulus. This particular arithmetic is called modular
modulus. Likewise subtraction arithmetic.
using the number circle is called
modulo n subtraction. Questions
4 Evaluate the following using modulo 26 addition:

(a) 13 + 13 (b) 20 + 16 (c) 19 + 26 (d) 10 + 26

5 Evaluate the following using modulo 26 subtraction:

(a) 6 – 12 (b) 11 – 20 (c) 13 – 26 (d) 18 – 26

6 (a) How many numbers are there in the world of modulo

12 arithmetic?

(b) What are the numbers in modulo 12 arithmetic?

It is conventional to indicate modulo arithmetic as follows

13 + 13 = 0 (mod 26)
Where (mod 26) indicates that modulo 26 arithmetic has been used.

189 Single licence - Abingdon School

 5.6.10 Encryption

Modular arithmetic in daily life

People use modular arithmetic in their daily lives often without realising that
they are doing so. For example, consider the clock face shown in Figure 6.10.9,
ignore the fact that 12 has been replaced by 0. Suppose it is eight o’clock, and
you want to know what time it will be in 7 hours. You would use modulo 12
arithmetic: 8 + 7 is 3 (mod 12).

0
11 1

10 2

9 3

8 4

7 5
6

Figure 6.10.9 Unconventional 12 hour clock

In the case of days of the week represented by numbers as shown in Table

6.10.1 it is useful to use modulo 7 arithmetic.
Day Number
Sunday 0
Monday 1
Tuesday 2
Wednesday 3
Thursday 4
Friday 5
Saturday 6

Table 6.10.1 Numbered days of the week

Suppose today is Wednesday.
What will the day of the week be in 6 days?
To answer this we add 6 to 3, the latter being the number for Wednesday,
obtaining 9. But 9 modulo 7 is 2 (mod 7) which is Tuesday.
What day of the week will it be in 490 days?
To solve this, imagine travelling around the number circle for modulo 7. How
many times does one travel around in 490 days, i.e. how many times does 7 go
into 490? The answer is 70 times. This means we would arrive on a Wednesday,
i.e. 3 + 490 = 3 (mod 7).

Single licence - Abingdon School 190

5 Fundamentals of data representation

Congruence
This last example illustrates that adding 7 has the same effect as adding 14,
as adding 21 or adding 0 or subtracting 7 and so on. We call this congruence
(remember congruent triangles from maths lessons, it is a similar idea).

Questions
7 Suppose today is Tuesday.
(a) What day of the week will it be in 492 days?
(b) What day of the week was it 210 days ago?

8 In dealing with compass bearings, the modulus to use is 360.

Whole number bearings are chosen from the set
{0, 1, 2, …, 358, 359}.
Suppose that you are headed due East, your bearing is 90 degrees.
You turn right 130 degrees onto a bearing 220 degrees. You then
turn right 150 degrees, what is your new bearing?

9 Suppose that you are headed due East, your bearing is 90 degrees.
You turn left 130 degrees, what is your new bearing?

Key concept Two integers are said to be congruent with respect to a given modulus if they
Congruence: differ by a multiple of that modulus. For example, if the modulus is 12 then
Two integers are said to be 2, 14, 26, 36 are congruent. Figure 6.10.10 shows some more congruence for
congruent with respect to a modulus 12.
given modulus if they differ by
a multiple of that modulus.
A statement that two expressions are congruent is called a congruence.
For example, if the modulus
is 12 then 2, 14, 26, 38 are
48
congruent. 36
47 24 37
35 12 25
23 13
38
46 0
34 11 1 26
22 14
10 2

45 33 21 9 3 15 27 39

8 4
20 16
32 7 5 28
44 6 40
19 17
31 18 29
30 41
43
42

Figure 6.10.10 Unconventional 12 hour clock showing congruence

191 Single licence - Abingdon School

 5.6.10 Encryption

Congruence modulo 12
Two integers are congruent modulo 12 if they differ by a multiple of 12.
For example 5 is congruent to 17 (which is 5 + 12) and to
29 (which is 5 + 2 • 12).

5 is also congruent to −7 (which is 5 + (−1) • 12).

The mathematical notation for writing a congruence is similar to the

mathematical notation for writing an equation but where the equality symbol
has two horizontal bars (“=”), the congruence symbol has three (“≡”).
For example, we write the congruence
5 ≡ 17 (mod 12)

to state that 5 is congruent (modulo 12) to 17.

The notation requires that the modulus is specified in brackets together with
the word “mod” short for “modulo”.
Here are more modulo 12 examples. To verify that the numbers are indeed
congruent check that the difference between them is a multiple of 12:
6 ≡ 42 (mod 12)

−13 ≡ 11 (mod 12)

−13 ≡ −1 (mod 12)
−21 ≡ 3 (mod 12)
12 ≡ 0 (mod 12)

7 + 5 ≡ 12 (mod 12)
Representative theorem
Every integer is congruent modulo m to exactly one of the integers
0, 1, 2, 3, …, m −1.

For example, modulo 7: every integer is congruent to exactly one of the integers
0, 1, 2, 3, 4, 5, 6.

The 0, 1, 2, 3, …, m − 1 are called the representatives.

Questions
10 What are the modular arithmetic representatives of
(a) modulo 5 (b) modulo 9?

Single licence - Abingdon School 192

5 Fundamentals of data representation

Quotient and remainder theorem

For every integer b and every positive integer m, there is exactly one integer q
Key principle
and exactly one integer r among 0, 1, 2, 3, …, m – 1 such that
Quotient and remainder
b=q•m+r
theorem:
For every integer b and every Example:
positive integer m, there is
Let b = 23 and let m = 7.
exactly one integer q and
exactly one integer r among 0, Then the above equation is satisfied by q = 3 and r = 2
1, 2, 3, …, m – 1 such that (That is, 23 = 3 • 7 + 2)
b=q•m+r
As this example suggests, r is the remainder when b is divided by m.
r is the remainder when b is q is the quotient.
divided by m.
The remainder r is thus the value of (b mod m)’s representative.
q is the quotient.
The equation b = q • m + r shows that b and r differ by a multiple of m, which
The equation b = q • m + r shows that b is congruent to r (mod m).
shows that b and r differ by a
multiple of m, which shows that
b is congruent to r (mod m). Questions
11 Find the remainder r and quotient q if
(a) b = 37 m = 12 (b) b = 38 m = 24 (c) b = 76 m = 60
(d) b = 576 m = 365

Representatives and negative integers

What is the remainder for −15 mod 7?
To answer this we must remember that the result must be a mod 7
representative, i.e. one of 0, 1, 2, 3, 4, 5, 6.
−15 divided by 7 is −2 with −1 left over.

The representative that is congruent to −1 is 6

(difference between 6 and −1 is 7).
Therefore −15 mod 7 = 6.

Questions
12 Find the remainder r and quotient q if
(a) b = −37 m = 12 (b) b = −38 m = 24 (c) b = −76 m = 60
(d) b = −576 m = 365

193 Single licence - Abingdon School

 5.6.10 Encryption

Programming Tasks
1 Write a program to encrypt a line of uppercase text using the Caesar
cipher. Represent each character by a number between 0 and 26
(A ↦ 0, B ↦ 1, …, Z ↦ 25, space ↦ 26).
Use only these characters. Allow a user to choose a key within the
range 0…26.
Encrypt each character of the text using the equation

ciphertext_character = (plaintext_character + key ) mod 27

2 Write a program to decrypt a line of text encrypted using the

Caesar cipher. Assume that each character was represented by a
number between 0 and 26 (A ↦ 0, B ↦ 1, …, Z ↦ 25, space ↦
26) and only these characters were used when producing the line of
encrypted text. The user should enter a key in range 0…26.

Decrypt each character of the text using the equation

ciphertext_character = (plaintext_character − key ) mod 27

Questions
13 What is it unnecessary to use a key range wider than 0..26 for the
Caesar cipher in this case?

Breaking the Caesar cipher

Brute force approach
The Caesar cipher is easily broken by an attacker thus revealing the plaintext
and the key used to produce the ciphertext. A brute-force search is sufficient.
Assuming that the plaintext consisted of the 26 uppercase letters of the
alphabet. A brute force search on the ciphertext consists of just trying all the
possible keys except 0 on the ciphertext until the plaintext is discovered. It is
assumed that key 0 would not have been used when encrypting the plaintext
because it doesn’t alter the plaintext.
For example, take a five letter name, choose a key between 1 and 25 and then,
with this key, use the Caesar cipher to encrypt the name. If we choose ALICE
for the name and 3 for the key then the ciphertext is DOLFH. Table 6.10.2
shows the outcome of the brute force search from which it can be deduced that
the key was 3 and the plaintext was ALICE.

Single licence - Abingdon School 194

5 Fundamentals of data representation

Key Plaintext Key Plaintext Key Plaintext

1 CNKEG 10 TEBVX 19 KVSMO
2 BMJDF 11 SDAUW 20 JURLN
3 ALICE 12 RCZTV 21 ITQKM
4 ZKHBD 13 QBYSU 22 HSPJL
5 YJGAC 14 PAXRT 23 GROIK
6 XIFZB 15 OZWQS 24 FQNHJ
7 WHEYA 16 NYVPR 25 EPMGI
8 VGDXZ 17 MXUOQ
9 UFCWY 18 LWTNP
Table 6.10.2 Brute force attack on ciphertext DOLFH

Task
1 Choose a four letter name and a key. Use the key and the Caesar
cipher to encrypt the name. Now give the ciphertext to another
student and ask them to use a brute force attack to discover the
name and the key used.

Letter frequency attack

The Caesar cipher is also susceptible to letter frequency analysis. If an attacker
knows that the plaintext was written in English then because the Caesar cipher
applies the same shift to each plaintext letter, the frequencies of occurrence
of the letters in the ciphertext match those in the plaintext shifted by the key.
When the plaintext is sufficiently long or a series of ciphertexts are intercepted,
a good guess is that the ciphertext(s) matches the relative frequency of letters
common to a large number of English texts when shifted by the key.
Frequency analysis of a large number of English texts has revealed that each
letter of the alphabet occurs with unequal likelihood as shown in Figure
6.10.11. The letter E occurs most frequently, 12.7% of the time on average,
and so is roughly twice as likely on average to occur in a piece of text as the
letter S which has relative frequency of approximately 6.3%.

195 Single licence - Abingdon School

 5.6.10 Encryption

%
14

12
RELATIVE FREQUENCY

10

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

Figure 6.10.11 Relative frequency analysis for English

%
14

Ciphertext
12
RELATIVE FREQUENCY

10

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

Figure 6.10.12 Ciphertext relative frequency analysis

Single licence - Abingdon School 196

5 Fundamentals of data representation

Figure 6.10.12 shows what the relative frequency distribution would be if the
Caesar cipher was applied with key 3 to plaintext with relative letter frequency
distribution as shown in Figure 6.10.11. It is relatively easy to see that an E
has been shifted to become an H, therefore the key must be 3. The ciphertext is
said to leak information about the plaintext.

Task
2 Try the Caesar frequency analysis exercise at
https://www.khanacademy.org/computing/computer-science/
cryptography/ciphers/e/

Programming Task
1 Write a program that performs a relative frequency analysis of
English text obtained from a text file. You will need access to text
files of appropriate length. Whilst developing and testing your
program you could use any file on your local machine such as a
ReadMe.txt or you could write your own in a text editor. For more
substantial text files you could download an ebook from http://
www.gutenberg.org.
The NLTK toolkit from www.nltk.org written for Python is a very
powerful text processing resource that could be used for this and
other work.

Caesar cipher weaknesses

Key point Summarising, the Caesar cipher has three major weaknesses:
1. The number of possible keys is too small
Caesar cipher weaknesses:
1. The number of possible 2. The same shift is applied to each character making it easy to use
keys is too small relative letter frequency analysis.
2. The same shift is applied
to each character making 3. The same shift is likely to be used for each message.
it easy to use relative The solution:
letter frequency analysis.
3. The same shift is likely to 4. Make the number of possible keys so large that it becomes infeasible
be used for each message. to employ a brute force approach of trying all possible keys
5. Arrange for the occurrence of each letter/character in the ciphertext
to be equally likely by applying a random shift to each.

197 Single licence - Abingdon School

 5.6.10 Encryption

One-time pad Background

One way of making the number of possible keys large is to choose a new key
value for each letter/character of the plaintext message and a new set of key The use of a truly random
key, as long as the plaintext,
values for each new message. If key values are chosen randomly then the second
is an essential part of the
bullet point above can also be satisfied. one-time pad algorithm. The
For example, if the plaintext message is one-time pad algorithm itself
is mathematically secure. Thus
CLOCK TOWER USUAL TIME TONIGHT J the codebreaker cannot retrieve
the plaintext by examining the
then 32 key values are needed because there are 32 characters – 27 letters and
ciphertext. The best that the
5 spaces in this message. We randomly choose a different combination of 32
codebreaker can do is to try to
key values for each new 32 character long message from the set of all possible retrieve the key. If the random
permutations of 32 key values. Each key value can be one of 26 possible letters values for the one-time key are
or a space. This is done 32 times therefore there are 2732 different key patterns not truly random but generated
by a deterministic mechanism
to choose from at random.
or algorithm then there is a
Applying a random shift to each letter/character in the plaintext ensures that possibility of predicting the key.
each ciphertext character is equally likely. Thus, selecting a good random
number generator is the most
For our example, first convert each plaintext character to numeric form. The 26 important part of the system.
letters of the alphabet are coded as 0…25, respectively and the space character To see one way of manually
as 26. generating a truly random key
using ten-sided dice visit
If we use pi to refer to the numeric code for the ith character of the plaintext, http://users.telenet.be/d.
and ci for the numeric code of the corresponding letter in the ciphertext, then rijmenants/en/onetimepad.htm
to obtain ci from pi and key ki we use
ci = (pi + ki) mod 27
To obtain the ciphertext character we convert ci into its equivalent character.
Let’s suppose the 32 key values, ki where i is in {1…32}, chosen at random
from the range {0…26} are
22 23 8 8 3 13 14 15 24 22 5 9 8 18 25 16 10 7 21 1 2 4 23
1 12 11 4 14 4 23 15 6
Call this sequence of keys the cipher key K. (To obtain these 32 keys you
could use all the hearts and diamonds from a pack of cards and a joker because
this gives 27 cards. Shuffle the pack then take the top card, write down its
corresponding number: the joker is 0, hearts are 1…13, diamonds are 14…26
with Ace 1 or 14, Jack 11 or 24, etc, Put the card back and shuffle the pack
again, repeat the process until you have 32 randomly chosen numbers.)

Questions
14 Why is it unnecessary for this cipher to choose numbers greater
than 26?

Single licence - Abingdon School 198

5 Fundamentals of data representation

The plaintext codes pi where i is in {1…32} are

2 11 14 2 10 26 19 14 22 4 17 26 20 18 20 0 11 26 19 8 12 4 26 19 14 13 8
6 7 19 26 9
The ciphertext codes [ci = (pi+ ki) mod 27 where i is in {1…32}]
are
24 22 22 10 1 12 6 2 19 26 15 7 1 9 18 16 21 6 13 9 14 8 22 20 26 24 12
20 11 15 14 15
The ciphertext is
YGWKLMGCT FGBJSQVGNJOIWU YMULPOP
The tendency of applying random shifts is to flatten the distribution. Analysis
reveals that applying random shifts to a large number of plaintext messages
leads to the following two powerful properties possessed by ciphertexts
1. The shifts do not fall into a repetitive pattern, e.g. E → H every time
is avoided
2. The ciphertext distribution is flattened and has a uniform frequency
distribution
Achieving a uniform frequency distribution as in Figure 6.10.13 will mean
that there is no frequency differential and therefore no leak of information
about the plaintext message that an attacker or eavesdropper could exploit to
guess the plaintext.
RELATIVE FREQUENCY

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z
space

Figure 6.10.13 Uniform relative frequency distribution

199 Single licence - Abingdon School

 5.6.10 Encryption

Task
3 Watch the Khan Academy polyalphabetic video and try the tool to
see how a non-uniform plaintext distribution can be flattened as
described in this section.
https://www.khanacademy.org/computing/computer-science/
cryptography/crypt/v/polyalphabetic-cipher

Questions
15 Suppose Alice and Bob communicate messages to each other which
have been encrypted using the Caesar cipher and a previously
agreed secret key. Now suppose that Eve intercepts the ciphertext
and that she happens to know or suspect that Alice starts all
her messages to Bob with the characters “DEAR BOB”. The
corresponding ciphertext is GHDUCERE. Alice and Bob make
no secret of the fact that they use the Caesar cipher believing that
keeping the key secret is sufficient to maintain the security of their
messages. The encryption equation that Alice and Bob use is
ciphertext_character = (plaintext_character + key ) mod 27

Explain how Eve using the plaintext and ciphertext could recover
the secret key used by Alice and Bob.

16 Alice and Bob decide to use a new key for each character, generated
randomly, and apply the stream of random keys using the Caesar
cipher to future messages. Explain why this could offer greater
security over the scheme described in Q15 even though Eve
continues to intercept the ciphertext and Alice continues to start
messages with “DEAR BOB”.

To subject this cipher to closer scrutiny we ask how many ways can a particular
plaintext message consisting of only uppercase letters of the alphabet be
encrypted by a shift cipher which chooses key values randomly? Well consider
that for the first character, there are 26 different possible key values, for the
second character, 26 different possible values again and so on. If plaintext
messages of length 32 characters are encrypted then a key will consist of 32 key
values. The total number of keys of length 32 is therefore
26 x 26 x 26 x …….. x 26 x 26 = 2632 ≈ 2 x 1045

Single licence - Abingdon School 200

5 Fundamentals of data representation

Key principle The total number of possible ciphertexts corresponding to any particular
32 character plaintext message is thus 2 x 1045 one for each possible key
One-time pad: consisting of 32 key values.
Plain text of a message is ‘mixed’
with random text taken from a If each of these possible ciphertexts is written on a separate piece of paper
one-time pad resulting in cipher then the entire stack would be
text which is truly random.
2 x 1045 x 5 x 10-5 metres high = 1 x 1040 metres
The same one-time pad is used
to ‘unmix’ the random text from taking the thickness of a piece of paper to be 5 x 10-5 metres.
the cipher text, which results in
the original plain text. By comparison the Milky Way galaxy is estimated to be 9.5 x 1020 metres
across. If the key values were generated randomly then each ciphertext will
One only has to guarantee
that the one-time pad is be equally likely.
safe, that it comprises truly The decryption cipher is
random numbers, that there
are only two copies of it, and pi = (ci - ki) mod 27
that both copies are destroyed
immediately after use to prevent Likewise a particular ciphertext could have come from any one of 2 x 1045
it being used again (the one- possible plaintext messages. The chance of an attacker guessing which
time property), for it to be used
one correctly is therefore vanishingly small. It is impossible therefore for
to send a message safely without
the risk of being deciphered by an attacker or eavesdropper to break this encryption scheme because the
an attacker or eavesdropper. ciphertext yields no possible information about the plaintext (except its
length).
Background This is the strongest possible method of encryption. It is known as the
The “red phone” used one-time pad because when first used the key was written on a sheet of
in the 1980s for secure paper or pad and used only once.
communication between the
Summarising, the one-time pad method is based on the principle that
USA and the USSR was based
on a one-time pad. The random
the plain text of a message is ‘mixed’ with truly random text taken from
key sequences or pads were a one-time pad. Because the resulting cipher text is still truly random it
delivered by courier. can safely be sent without the risk of being deciphered by an attacker or
eavesdropper.
At the receiving end, the same one-time
pad is used to ‘unmix’ the random text
from the cipher text, which results in
the original plain text. One only has
to guarantee that the one-time pad is
safe, that it comprises truly random
numbers, that there are only two copies
of it, and that both copies are destroyed
immediately after use to prevent it being
used again (the one-time property) on
another plaintext message.
Figure 6.10.14 A one-time pad reproduced with kind
permission of Paul Reuvers, Crypto Museum
(www.cryptomuseum.com)

201 Single licence - Abingdon School

 5.6.10 Encryption

Task Background
4 Watch the Khan Academy one-time pad video Hardware random number
https://www.khanacademy.org/computing/computer-science/ generators have been built into
some processor systems or made
cryptography/crypt/v/one-time-pad
possible in some operating
system. E.g. Raspberry Pi
5 The Venona project was a counter-intelligence program initiated includes a hardware-based
by the United States Army Signal Intelligence Service (a forerunner random number generator that
of the National Security Agency) that lasted from 1943 to 1980. can generate cryptographic
quality random numbers.
The program attempted to decrypt messages sent by Soviet Union
In Unix-like operating systems
intelligence agencies, including its foreign intelligence service and /dev/random is a special
military intelligence services. The project produced some of the file that serves as a blocking
most important breakthroughs for western counter-intelligence in pseudorandom number
this period, including the discovery of the Cambridge spy ring and generator:
.dev>more –f random
the exposure of Soviet espionage targeting the Manhattan Project.
The NSA declassified the program in 1995. It can be read about in a
NSA document at
https://www.nsa.gov/about/_files/cryptologic_heritage/publications/
coldwar/venona_story.pdf

Key principle
Randomness
Randomness means lack of pattern or predictability of events. Randomness Randomness:
abounds in the physical world and in man-made devices such as electrical Randomness means lack of
pattern or predictability of
circuits as fluctuations of an unpredictable nature which we call noise.
events.
Electrical storms, the microwave background left over from the Big Bang and
other events induce random currents of electricity in aerials connected to
radio receivers and televisions that cause the hissing noise that we hear in their
loudspeakers. This atmospheric noise can be captured, sampled and digitised to
provide a source of truly random bits. Such a service is provided by
https://www.random.org. Background
Another source is https://www.fourmilab.ch/hotbits/secure_generate.html RandomX package for Java has
which uses the unpredictable nature of radioactive decay to generate truly an option to get random bits
random bits. from hotbits.
http://www.fourmilab.ch/
A random number is one that is drawn from a set of possible values, each hotbits/source/randomX/
of which is equally probable, i.e., a uniform distribution, e.g. the throw of a randomX.html
six-sided die. When discussing a sequence of random numbers, each number
drawn must be statistically independent of the others, i.e. knowledge of an
arbitrarily long sequence of numbers is of no use whatsoever in predicting the
next number to be generated. Each possible arbitrarily long sequence is thus
equally likely.

Single licence - Abingdon School 202

5 Fundamentals of data representation

Pseudorandom
Pseudorandomness is an important concept in cryptography.
Informally pseudorandom means:
cannot be distinguished from uniform i.e. random.
The cryptographic definition of pseudorandom however is
a distribution is pseudorandom if it passes all efficient statistical tests.
This definition has been arrived at by considering the need to resist an attack
from an adversary who is trying to obtain information from ciphertexts about
the corresponding plaintext messages.
Generating large numbers of truly random numbers is extremely difficult
Key point
so people have turned to the computer and algorithms programmed into
Pseudorandom numbers: the computer to generate pseudorandom numbers from an initial seed.
Pseudorandom numbers are These pseudorandom numbers are generated deterministically and can only
generated deterministically
approximate a truly random distribution because numbers calculated by a
and can only approximate
a truly random distribution computer through a deterministic process, cannot, by definition, be random.
because numbers calculated Given knowledge of the algorithm used to create the numbers and the seed,
by a computer through a it is possible to predict all the numbers returned by subsequent calls to the
deterministic process, cannot,
algorithm, whereas with genuinely random numbers, knowledge of one
by definition, be random.
number or an arbitrarily long sequence of numbers is of no use whatsoever in
predicting the next number to be generated. Therefore, computer-generated
“random” numbers are more properly referred to as pseudorandom numbers,
and pseudorandom sequences of such numbers. Pseudorandom generated
sequences eventually repeat with the periodicity determined by the seed and the
algorithm used. Pseudorandom generated sequences are also reproducible, i.e.
for a given algorithm, starting from the same seed generates the same sequence.
Pseudorandom number generators (PRNGs)
Key point A pseudorandom number generator is an efficient, deterministic algorithm
Pseudorandom number that expands a short, uniform seed into a longer, pseudorandom output in
generators: polynomial time. It is useful whenever
A pseudorandom number
1. It would be difficult to communicate a long sequence of numbers
generator is an efficient,
deterministic algorithm that needed in a symmetric key cipher instead the seed is communicated
expands a short, uniform seed 2. A large number of random numbers are required and access to truly
into a longer, pseudorandom
random numbers is restricted to a much smaller number
output in polynomial time.
The seed may be chosen from The seed may be chosen from the small number of truly random numbers that
the small number of truly are available but doesn’t have to be.
random numbers that are
available but doesn’t have to be.

203 Single licence - Abingdon School

 5.6.10 Encryption

SEED

OUTPUT
Figure 6.10.15 A pseudorandom number generator, G, producing a longer
stream of “random” bits from a shorter length seed of true random bits Background
Care must be taken when relying on pseudorandom number generators Security experts have long
for cryptographic purposes because they are deterministic. However, a class suspected the National Security
of improved random number generators, termed cryptographically secure Agency (NSA) has been
introducing weaknesses into
pseudorandom number generators (CSPRNG) exist that rely on truly random
CSPRNG standard 800-90 that
seeds external to the software.
they can exploit in ciphers that
use this standard; this being
confirmed for the first time by
one of the top secret documents
leaked to the Guardian by
Edward Snowden.

Figure 6.10.16 (a) Image Figure 6.10.16 (b) reproduced with

generated from random numbers permission of RANDOM.ORG
generated by the PHP rand() – image generated from random
function on Microsoft Windows. numbers obtained from atmospheric
noise.
The image in Figure 6.10.16(a) ( reproduced using a PHP script with kind
permission of Bo Allen, http://boallen.com) exhibits patterns because the
pseudorandom number generator, the programming language PHP’s rand()
function, is deterministic with a relatively short periodicity whereas the bitmap
in Figure 6.10.16(b) does not because it relies on truly random numbers.

Single licence - Abingdon School 204

5 Fundamentals of data representation

Task
6 Download the test program at http://www.fourmilab.ch/random/
Use it to test sequences of bytes/bits for their randomness. Run
this program on data generated by a high-quality pseudorandom
sequence generator. You should find it generates data that are
indistinguishable from a sequence of bytes chosen at random.
Indistinguishable, but not genuinely random.

Background
7 Watch the video on pseudorandom number generators at https://
The Vernam cipher was www.khanacademy.org/computing/computer-science/cryptography/
exploited in the design of a
crypt/v/random-vs-pseudorandom-number-generators
high security teleprinter cipher
machine that the Lorenz
company made for the German
Army High Command to The Vernam cipher
enable them to communicate The Vernam Cipher is named after Gilbert Sandford Vernam (1890-1960)
by radio in complete secrecy
who, in 1917, invented the stream cipher and later co-invented the one-time
during WW2.
These transmissions were pad (OTP). His patent US1310719 was filed in 1918 and is, according to the
broken by Bill Tutte. The National Security Agency (NSA), perhaps one of the most important in the
process of decrypting Lorentz history of cryptography.
machine ciphertexts was later
automated using a refinement
At the time of the invention, Vernam was working at AT
suggested by Max Newman & T Bell Labs in the USA. Messages were then sent by
and some clever engineering by telegraph, a system that used pulses of electrical current
Tommy Flowers who designed to encode characters according to the Baudot code. The
and built Colussus, the world’s
characters were entered into and read from the system
first stored program computer
to decrypt Lorentz encrypted using a teleprinter.
messages. Colussus reduced Vernam proposed a teleprinter cipher in which a
the time taken from weeks to
previously prepared key, kept on paper tape, was combined Figure 6.10.17
hours. Colussus came online
just in time to decrypt messages character by character with the plaintext message to Gilbert Vernam
which gave vital information to produce the ciphertext. To decrypt the ciphertext, the
Eisenhower and Montgomery same key would be again combined character by character, producing the
prior to D-Day - plaintext.
http://www.codesandciphers.
org.uk/lorenz/colossus.htm
Working with Joseph Mauborgne, at that time a captain in the US Army
Signal Corps, they proposed that the paper tape key should contain random
information (the key stream). The incorporation of this proposal into Vernam’s
machine implemented an automatic form of the one-time pad.

Task
8 Watch AT & T Labs’ video of the Vernam cipher http://techchannel.att.com/play-video.
cfm/2009/10/12/From-the-Labs:-Encryption1
9 Read about the details of the Vernam cipher machine by visiting the website
http://www.cryptomuseum.com/crypto/vernam.htm

205 Single licence - Abingdon School

 5.6.10 Encryption

The Vernam cipher relies on the bit-wise eXclusive-OR (XOR) Boolean Key principle
function. This is symbolised by ⊕ and is represented by the following truth
table, Table 6.10.3, where 1 represents true and 0 represents false Vernam cipher:
Encrypts and decrypts a
INPUT OUTPUT message using a one-time pad
A B A⊕B approach in which the plaintext
0 0 0 message M = {0, 1}n is bitwise
0 1 1 XORed with a uniformly
1 1 1 random key k {0,1}n where n is
the number of bits to encrypt
1 0 0
the message.
Table 6.10.3 Exclusive-Or truth table
A very useful property of the exclusive-or operation is that it is possible to Background
recover an input given the output and the other input. For example, if the
The eXclusive-OR function is
inputs A and B are 0 and 1 respectively, then A ⊕ B = 1. If we exclusive-or this an example of an involution.
output 1 with, say, input B which was 1, 1 ⊕ 1 = 0, we recover input A which An involution is a function
was 0. It works for all inputs.
f:X →X
Therefore the same key stream can be used both to encrypt plaintext to
That, when applied twice,
ciphertext and to decrypt ciphertext to yield the original plaintext:
brings one back to the starting
Plaintext ⊕ Key = Ciphertext point

and: f(f(x)) = x
Ciphertext ⊕ Key = Plaintext
If the key stream is truly random, and used only once, this is effectively a
one-time pad. Information
Visit the cryptomuseum at
0

http://www.cryptomuseum.
0

key stream k key stream k

1

com/crypto/vernam.htm
1
1

for a good demonstration of the

1
0

Identical key streams Vernam cipher.

1

1
0

0
0

0
1

ciphertext message plaintext message M

1 0 1 1 0 0 1 0 1 1 0 1 1 0 1 1

Encrypt Decrypt

Figure 6.10.18 Encrypting and

1

decrypting a message using

1

the Vernam cipher machine in

0

plaintext message M
1

one-time pad mode. Plaintext

1

message M = {0, 1}n is bitwise

0

XORed with a uniformly

1

random key k {0,1}n where n is

1

the number of bits to encrypt

the message.

Single licence - Abingdon School 206

5 Fundamentals of data representation

Substituting pseudorandom data generated by a cryptographically secure

Background
pseudorandom number generator is a common and effective construction for a
RC4 is an example of a Vernam stream cipher.
cipher. It has been and still is
RC4 has been a very widely used software stream cipher. RC4 is an example
used in popular protocols such
as Transport Layer Security of a Vernam cipher. It has been and still is used in popular protocols such as
(TLS) (to protect Internet Transport Layer Security (TLS) (to protect Internet traffic) and WEP (to secure
traffic) and WEP (to secure wireless networks) although it is now considered insecure.
wireless networks) although it is
now considered insecure. WEP relies on a short secret key that is shared between a mobile station (e.g. a
laptop with a wireless Ethernet card) and an access point (i.e. a base station).
The short secret key is expanded into an infinite pseudorandom key stream
which is XORed with the message packets before they are transmitted.

Programming task
3 RC4 is a stream cipher used in WEP. The infinite pseudorandom key stream for RC4 is generated from
a secret key using the following two algorithms. Code these in your preferred language and run the test
below on secret key = AQACS and plaintext = Computer Science
Key-scheduling algorithm
for i from 0 to 255
S[i] := i
endfor
j := 0
for i from 0 to 255
j := (j + S[i] + key[i mod keylength]) mod 256
swap values of S[i] and S[j]
endfor

PseudoRandom Number Generator

i := 0
j := 0
while PseudoRandom Numbers required:
i := (i + 1) mod 256
j := (j + S[i]) mod 256
swap values of S[i] and S[j]
PSRNumber := S[(S[i] + S[j]) mod 256]
output PSRNumber
endwhile
Test
The keys and plaintext are ASCII, the keystream and ciphertext are expressed below in hexadecimal but
stored as bytes.
Key: AQACS
Keystream: F163D4497F1C801DCB4E3C...
Plaintext: Computer Science
Ciphertext: B20CB9390A68E56FEB1D5FC4720A7BD7

207 Single licence - Abingdon School

 5.6.10 Encryption

Task Key principle

Perfect secrecy:
10 Attacks on RC4 have shown that it is possible to distinguish
Perfect secrecy means an
its output from a random sequence. Why does this make RC4 eavesdropper would not, by
insecure? Use a search engine to research an RC4 attack. gaining knowledge of the
ciphertext but not of the key,
be able to improve their guess
RC4 is still installed on some operating systems. For example, running the
of the plaintext even if given
openssl ciphers command on an Apple Mac running Mac OS X 10.8 reveals unlimited computing power.
that the SSL(Secure Sockets Layer) cipher RC4 (128 bit) is being used for Such cryptosystems are
encryption on this machine because it has not been deselected: considered cryptoanalytically
unbreakable and information-
$ openssl ciphers -tls1 -v RC4-SHA
theoretically secure meaning
RC4-SHA SSLv3 Kx=RSA Au=RSA Enc=RC4(128) Mac=SHA1 they will not be vulnerable
to future developments in
Perfect secrecy
computer power such as
Claude Shannon, at Bell Labs, proved that the one-time pad is unbreakable, quantum computing.
and that it is the only cryptosystem that achieves perfect secrecy. He published The term perfect security is
his proof in a research paper in 1949. In it he defined a mathematical model given to such systems.
of what it means for a cryptosystem to be secure. Essentially, any unbreakable
system must have the same characteristics as the one-time pad:
The key
1. must be truly random
2. must be as long as the plaintext message
3. must never be reused in whole or part
4. and must be kept secret.
Claude Shannon
Points 1 and 2 mean that the number of possible keys must be at least as large (Getty Image library)
as the number of possible messages of a given length.

Task
11 Watch the video on Perfect Secrecy from Khan Academy -
www.khanacademy.org/computing/computer-science/cryptography/crypt/v/perfect-secrecy

Perfect secrecy means an eavesdropper would not, by gaining knowledge of

the ciphertext but not of the key, be able to improve their guess of the plaintext
even if given unlimited computing power.

Key point
Unconditional or Perfect Security (Perfect secrecy)
Regardless of any prior information the attacker has about the plaintext, the ciphertext leaks no additional
information about the plaintext in a ciphertext-only attack.

Single licence - Abingdon School 208

5 Fundamentals of data representation

Such cryptosystems are considered cryptoanalytically unbreakable and

information-theoretically secure meaning they will not be vulnerable to
future developments in computer power such as quantum computing.
The one-time pad is an example of an information-theoretically secure
cryptosystem. These systems have been used for the most sensitive
governmental communications, such as diplomatic cables and high-level
military communications.
What if the plaintext was a sequence of bits that represented an image? If
we apply a bitwise exclusive-or to this image using a sequence of randomly
generated bits for the key, we get an image that contains no information
about the original image because each ciphertext bit is just as likely to be a
Information 0 as a 1 - Figure 6.10.19(a) and Figure 6.10.19 (b).

Figure 6.10.19(a) Claude The original image can only be recovered by using the same key, i.e. the
Elwood Shannon was an exact sequence of randomly generated bits that produced the ciphertext.
American mathematician,
electronic engineer, and Applying the “wrong” key will result in recovering a different image.
cryptographer famous for
having founded information
theory with a landmark paper
that he published in 1948.

Shannon was also the first

person to show how the
logical algebra of 19th-century
mathematician George Boole
could be implemented using
electronic circuits of relays
and switches in which open or
closed switches could represent.
Figure 6.10.19(a) Plaintext Figure 6.10.19(b)
“true” and “false” and “0” and
“1”. Furthermore he showed image to be encrypted using a One-time pad ciphertext
how the use of electronic logic one-time pad. image of (a).
gates could be used to make (Getty Image Library)
decisions and to carry out
arithmetic.

Task
12 Try the reversible XOR demonstration at
https://www.khanacademy.org/computer-programming/reversible-
xor-demo/5580322717564928

209 Single licence - Abingdon School

5 Fundamentals of data representation

How XOR and a random key achieves perfect secrecy

Let’s suppose that Alice needs to send a private message to Bob and
therefore encrypts the message with a secret key known only to her
and Bob. Furthermore before encrypting the message, the letters of the
plaintext message are replaced by their equivalent ASCII values expressed
in binary. We would then have the problem of encrypting 0s and 1s. For
the sake of argument, let’s just focus on one bit of the message, call it p,
and encrypt this obtaining a one-bit ciphertext, c.
Next, Alice chooses the key, k, at random and uniformly (with no bias)
from the set of symbols {, , }. These symbols have equal likelihood,
⅓, of being chosen by Alice, i.e. ⅓ of the time Alice chooses , ⅓ of
the time  and ⅓ of the time . The chosen secret key is known only to
Alice and Bob.
When Alice needs to send the message she uses the chosen key to encrypt
her plaintext message bit p obtaining ciphertext, c according to the
following table, Table 6.10.4.
p k c
0  0
0  1
0  1
1  1
1  0
1  0

Table 6.10.4 Look up table for encrypting p into c using key k

Let’s suppose that Alice has chosen  for k. The values in Table 6.10.4
have been carefully chosen so that no two rows have the same k-value and
c-value, so on receipt of c, Bob will be able to decrypt c.
Unfortunately, this encryption scheme leaks information to Eve, an
eavesdropper who wishes to learn something about the plaintext so that she
can read what Bob is able to read.
Here is the method that Eve uses:
Suppose plaintext message bit p = 1,
Then if the key was , the ciphertext c = 1
But if it was  or  then c = 0
However, one of  or  are twice as likely as  to have been chosen
Therefore, for p = 1, c = 0 is twice as likely as c = 1.
Suppose, plaintext message bit p = 0,
Then if the key was , the ciphertext c = 0

210 Single licence - Abingdon School

 5.6.10 Encryption

But if it was  or  then c = 1

However, one of  or  are twice as likely as  to have been chosen
Therefore, for p = 0, c = 1 is twice as likely as c = 0.
Although having knowledge of c doesn’t allow Eve to determine the value of p
with certainty, it does allow her to revise her estimate of the chance that p = 0
or p = 1, as follows: Value of c Probability Probability
If before seeing c, Eve believed that p = 0 and p = 1 were equally seen by Eve p=1 p=0
likely, then if she sees that c = 1 she can infer that p = 0 is twice as not seen ½ ½
likely as p = 1. On the other hand, if she sees c = 0 then she can infer 0 ⅔ ⅓
that p = 1 is twice as likely as p = 0 - Table 6.10.5. 1 ⅓ ⅔
Table 6.10.5 How probability changes
The solution is to remove  as a possible value for k. Encryption then
when Eve gets sight of value of c
takes place using the values in Table 6.10.6.
Alice randomly chooses the key, k, from the set of symbols {, }. These
symbols have equal likelihood, ½, of being chosen, i.e. ½ of the time p k c
Alice will choose , ½ of the time . The chosen secret key is known 0  0
only to Alice and Bob. 0  1
1  1
Why does this new cryptosystem thwart Eve’s attempt to learn something
1  0
about the plaintext by examining the ciphertext?
Table 6.10.6 Look up table for
Suppose p = 0,
encrypting p into c using key k
Then if the key was , c = 0
But if it was  then c = 1
Since  is equally likely as  to have been chosen
Key principle
c = 1 is as equally likely to occur as c = 0.
Perfect security:
Suppose, p = 1, When the probability
Then if the key was , c = 1 distribution of the output, the
ciphertext, for the encryption
But if it was  then c = 0 system does not depend
Since  is equally likely as  to have been chosen upon whether 0 or 1 is being
encrypted we say that the
c = 0 is as equally likely to occur as c = 1. scheme achieves perfect secrecy
or perfect security because the
If before seeing c Eve believed that p = 0 and p = 1 were equally likely,
output will leak no information
then seeing c = 1 or seeing c = 0 cannot alter that belief because c = 1 about the input, the plaintext.
and c = 0 are equally likely whichever value of key k is chosen.
For this encryption scheme, the probability distribution of the output does not
depend upon whether 0 or 1 is being encrypted, so knowing the output gives
Eve no information about which is being encrypted. We say that the scheme
achieves perfect secrecy or perfect security.

Single licence - Abingdon School 211

5 Fundamentals of data representation

Encrypting long messages

If we replace  with 0 and  with 1, the encryption Table 6.10.6 becomes the
Background modulo 2 addition table for GF(2), Table 6.10.7.
GF(2) is covered in A Level p k c
Computer Science for AQA 0 0 0
Unit 1 in section 2.8.1. 0 1 1
GF is short for Galois Field 1 0 1
and is applied to arithmetic
1 1 0
in which there are a limited
number of elements, i.e. a finite Table 6.10.7 Look up table for encrypting p into c using key k
field; all operations performed
The exclusive-or operator ⊕ can be used to implement this table and encrypt
in the finite field result in an
element within that field. GF(2) plaintext p using key k to produce ciphertext c as follows
means that there are just two c=k⊕p
elements.
Similarly, the exclusive-or operator can be used to decrypt ciphertext c using
key k to produce plaintext p as follows
p=k⊕c

Key point
VERY IMPORTANTLY by using XOR and choosing the key randomly, the XOR
operator has a 50% chance of outputting a 0 or a 1.

To encrypt a long message, we first represent it as a string of n bits. Next, Alice

and Bob should agree an equally long sequence of key bits, k1, …, kn chosen
randomly. Now once Alice has produced the plaintext p1, …, pn, she obtains
the ciphertext c1, …, cn, one bit at a time as follows:
c 1 = k1 ⊕ p 1
c2 = k2 ⊕ p2


cn = kn ⊕ pn
The previous section argued that each bit ci of ciphertext tells Eve nothing
about the corresponding bit pi of plaintext and nothing about any of the other
bits of plaintext. From this we can draw the conclusion that the cryptosystem
has perfect secrecy or security.

212 Single licence - Abingdon School

 5.6.10 Encryption

Questions
17 A 3-symbol message, AQA is encrypted as follows. Each symbol is represented by a number between
0 and 26 (A ↦ 0, B ↦ 1, …, Z ↦ 25, space ↦ 26). Each number is represented by a five-bit binary
sequence (0 ↦ 00000, 1 ↦ 00001, …, 26 ↦ 11010). Finally, the resulting sequence of 15 bits is
encrypted using the key consisting of 15 randomly chosen bits 110000000101110 (obtained from
random.org) and modulo 2 addition.

Compute the ciphertext.

Computational security
Limitations of the one-time pad
The success of the one-time pad is that it achieves perfect secrecy. However the
following limitations have prevented more widespread use:
■■ The key is as long as the message
■■ Only secure if each key is used to encrypt a single message (i.e. key
Information
not used more than once) The “red phone” used
in the 1980s for secure
This means that the parties wishing to communicate in secret, e.g. Washington,
communication between the
DC and Moscow, Russia via the “red phone” must share keys of total length USA and the USSR was based
equal to the total length of all messages that they might ever send. on a one-time pad. The random
key sequences or pads were
If the same key k is used twice, e.g
delivered by courier.
c1 = k ⊕ m 1
c2 = k ⊕ m2
the attacker can compute
c1 ⊕ c2 = (k ⊕ m1) ⊕ (k ⊕ m2) = m1 ⊕ m2
This leaks information about m1 and m2 because it reveals where these differ:
the characteristics of the ASCII coding scheme can be exploited to identify
some letters and frequency analysis can be brought to bear as well.

Questions
18 Study the ASCII code table and note that letters all begin with 01 and the space character begins with
00. Also note that XOR of two letters gives 00… and XOR of a letter and a space gives 01… It is easy to
identify XOR of letter and space. If the identified XORed letters and spaces are XORed with space’s ASCII
code, the plaintext letter is recovered. The following two ciphertexts were intercepted. It is suspected that
the same random key has been used to produce these. Assuming 8-bit ASCII, can you recover any letters
of the plaintext messages?

c1 = 010100001010110011001000111110111101000011101110
c2 = 010000011010010011011101100100111101110011110011

(The key that was used so that you can check your answer:
000100011111110110001001110110111001001110111101)

Single licence - Abingdon School 213

5 Fundamentals of data representation

Computational secrecy
In practice it is more convenient to allow the leak of information with a tiny
probability to eavesdroppers with bounded computational resources, i.e. not
unlimited. This means relaxing perfect secrecy by
Key principle ■■ Allowing security to fail with tiny probability

Computational secrecy: ■■ Only considering “efficient” attackers

Computational secrecy relies on
To set this in perspective we need to consider what is meant by a tiny
allowing
1. security to fail but with
probability and “efficient” attackers.
a probability negligible Let’s say we allow security to fail with probability 2−60 or 1 in 1018 times.
in n where n is a measure
of the challenge of This is of the order of probability that a person will be struck by lightning in
breaking the system, the next year.
e.g. factoring a given
Now consider a brute-force search of the key space.
integer n
2. restricting attention to Assuming for argument’s sake that one key can be tested per clock cycle of the
attackers running in time CPU (2014 commodity PC CPU):
polynomial in n
■■ Desktop computer ≈ 257 keys per year
■■ Supercomputer ≈ 280 keys per year
Key principle ■■ Supercomputer since Big Bang ≈ 2112 keys

Computational security: The meaning of “efficient” attackers is attackers who can try 2112 keys.
An encryption method is
Well, if we choose a key space of 2128, i.e. keys of length 128 bits then we
computational secure if it
safe to assume that no known
should meet the requirement for secrecy.
attack can break it in a practical This kind of secrecy is called computational secrecy. It relies on allowing
amount of time.
1. security to fail but with a probability negligible in n where n is a
measure of the challenge of breaking the system, e.g. factoring a
given integer n
2. restricting attention to attackers running in time polynomial in n
The notion of computational secrecy leads to the classification of an encryption
method as being computational secure if it safe to assume that no known
attack can break it in a practical amount of time.
However, this is very different from a proof of security. Thus in theory, every
Information cryptographic algorithm except for the Vernam cipher (one-time pad) can be
IBM Quantum computer: broken, given enough ciphertext and time.
IBM makes quantum computer
available to members of the Task
public, 4th May 2016 -
www-03.ibm.com/press/us/en/ 13 Research why quantum computing might be a threat to ciphers
pressrelease/49661.wss that rely on computational security and not information-theoretical
See security.
www.research.ibm.com/
quantum/

214 Single licence - Abingdon School

 5.6.10 Encryption

Task
14 Look at the Kryptos transcript available from
www.elonka.com/kryptos/. Can you decrypt the four ciphertexts?
Don’t worry if you can’t –
see www.wired.com/2013/07/nsa-cracked-kryptos-before-cia

In this chapter you have covered:

■■ What is meant by encryption and its definition
■■ Caesar cipher and applied it to encrypt a plaintext message and to decrypt
a ciphertext
■■ The limitations of the Caesar cipher
■■ Vernam cipher or one-time pad and applied it to encrypt a plaintext
message and to decrypt a ciphertext
■■ Why Vernam cipher is considered as a cipher with perfect security
■■ Comparison of Vernam cipher with ciphers that depend on
computational security

Single licence - Abingdon School 215

6 Fundamentals of computer systems
6.1 Hardware and software
Learning objectives:
■■Understand the relationship
between hardware and
■ ■ 6.1.1 Relationship between hardware and
software
software and be able to define
What is hardware?
the terms:
The hardware of a computer is the physical components, electronic and
• hardware electrical, that it is assembled from. It is the platform on which software
• software executes.

■■Explain what is meant by: What is software?

Software consists of sequences of instructions called programs which can be
• system software understood and executed by the hardware in its digital electronic circuits or a
• application software virtual machine equivalent.

■■Understand the need for, and Questions

attributes of, different types of 1 What is meant by hardware?
software
2 What is meant by software?
■■Understand the need for, and
functions of the following

■■ 6.1.2 Classification of software

software:

• operating systems (OSs)

• utility programs Computer software may be classified as follows:
1. The system programs (or system software), which control the
• libraries
operation of the computer itself, e.g. the operating system
• translators (compiler, 2. The application programs (or application software), which solve
assembler, interpreter)
problems for their users, e.g. constructing a letter using word
■■Understand the role of the processing software for printing and sending to someone.
operating system
What is system software?
Key concept A computer system uses a layer or layers of software to enable users to operate
the computer without having to be familiar with its internal workings. This
Hardware:
The hardware of a computer layer or layers is called systems software and includes the operating system and
is the physical components, other forms of systems software.
electronic and electrical, that
it is assembled from. It is the What is application software?
platform on which software Applications software is an application program or programs designed to
executes. support user-oriented tasks which would need to be carried out even if
computers did not exist. For example, communicating in written form, placing
orders for goods, looking up information.

Single licence - Abingdon School 216

6 Fundamentals of computer systems

Key concept Questions

Software: 3 Describe the classification of computer software.
Consists of sequences of
instructions called programs
which can be understood and
executed by the hardware in its The need for and attributes of different types of software
digital electronic circuits or a Application software cannot execute unless it has been first translated into
virtual machine equivalent.
the language of the computer, machine code. or a form that is executable by a
computer.

Key concept It needs to be loaded into main memory and it needs to obtain input from
input devices such as keyboards and to write output to output devices such as
System software:
printers and it may need to communicate with other computers.
A layer or layers of software
which enables users to operate Application software may need to store information permanently and to
the computer without having subsequently access stored information. The stored information should be
to be familiar with its internal
backed up so, if necessary, it may be restored from a back-up copy. These
workings.
services are provided by the operating system and utility software without
which it would not be possible to run application software.
Key concept Application software may be classified as

Application software: • General purpose application software: software that is appropriate

Application software is an for many application areas is described as general-purpose application
application program or software. For example, word processing can be applied in writing-up
programs designed to support
project work, in personal correspondence, writing memos, writing a
user-oriented tasks which would
book, creating standard business letters. The software is relatively cheap
need to be carried out even if
computers did not exist. because its development costs are spread among all the purchasers of
the software, which in the case of popular application software will
be a large number. It is likely to be very reliable because it has been
Key concept produced by an experienced team of programmers and tested on a large
customer base.
Different types of software:
1. General purpose • Special purpose applications software: special purpose application
2. Special purpose software is used for a particular application. For example, a dentist
3. Bespoke
might use application software written specifically to record and process
dental treatments, a task that every dentist needs to do. A business
might use an accounting package for its accounts of sales. It is likely to
be very reliable because it has been produced by an experienced team of
programmers and tested on a large but specialised customer base.
• Bespoke software: when no general purpose or special purpose
software exists that could do the job, software must be written from
scratch to solve the specific problem or to support the required task.
This software is called bespoke (tailor-made) software. For example, a
teacher interested in finding out how frequently his students logged on
to the college’s computer network and for how long, wrote a program

217 Single licence - Abingdon School

 6.1.3 System software

using the programming language C to handle this task because no

application program existed which could do this job.

Questions
4 Describe the classification of application software?

5 Why is system software needed in addition to application software?

■■ 6.1.3 System software

Systems software can be classified as follows:
• Operating system software: an operating system is a program or suite
Key concept
of programs which controls the entire operation of a computer
• Utility programs: a utility program is a systems program designed System software classification:
1. Operating systems
to perform a common place task, for example, formatting and
2. Utility programs
partitioning a disk or checking a disk for viruses. Some utility programs 3. Library programs
are supplied with the operating system, others can be installed at a later 4. Translators
time. 1. Compilers
• Library programs: a program library is a collection of compiled 2. Assemblers
3. Interpreters.
routines that other programs can link to and use. Linking may be done
at compile-time when building an executable or at run-time. Run-time
library programs are loaded on demand and shared by different
software applications. Loaded run-time libraries remain resident in Information
memory until the last executing application is closed. In the Microsoft
Intermediate code:
Windows operating system, the run-time libraries are called dynamic This is a language which lies
linked libraries or dlls. between a high-level language
• Compilers, assemblers, interpreters: these are computer language (HLL) and machine code. It is
closer to machine code than an
translators.
HLL. It supports operations for
ŠŠ Compiler: a compiler translates a high-level language program a fictitious machine. Compilers
into a computer’s machine code or some other low-level language. consist of several stages, one

Machine code is a language that the hardware of a computer can of which is intermediate-code
generation. It is a much simpler
understand and execute. It consists of executable binary codes.
task to write an interpreter for
ŠŠ Assembler: an assembler translates a program written in assembly a new machine designed with
language into machine code. Assembly language is a symbolic form a different instruction set than
it is to write a compiler. Any
of machine code. The symbolic form consists of mnemonics such as
program in intermediate-code
ADD and SUB that denote the machine operation to be performed. form, including a compiler, can
An assembler simply substitutes the corresponding executable binary be “executed” by interpreting its
code for the mnemonics. intermediate-code form with the
interpreter written for the new
ŠŠ Interpreter: translates and executes a high-level language or machine.
intermediate-code program one statement at a time. It provides Examples of intermediate-code
a way of executing programs not in the machine code of the are p-code and bytecode.

computer.

Single licence - Abingdon School 218

6 Fundamentals of computer systems

Questions
6 What are the functions of each of the following software:

(a) operating systems (b) utility programs (c) libraries (d) translators?

7 Name three different types of utility program.

■■ 6.1.4 Role of an operating system

The most fundamental of all the system programs is the operating system.
An operating system has two major roles:
• Hide the complexities of the hardware from the user so that the user is presented with a machine which is
much easier to use.
• Manage the hardware resources to give an orderly and controlled allocation of the processors, memories
and input/output (I/O) devices among the various programs competing for them, and manage data storage.

Key concept Questions

Role of an operating system: 8 What is the role of the operating system?
1. To hide the complexities of
the hardware from the user.
2. Manage the hardware
resources.

In this chapter you have covered:

■■ The relationship between hardware and software and be able to define the terms:
• hardware
• software
■■ What is meant by:
• system software
• application software
■■ The need for, and attributes of, different types of software
■■ The need for, and functions of the following software:
• operating systems (OSs)
• utility programs
• libraries
• translators (compiler, assembler, interpreter)
■■ The role of the operating system

219 Single licence - Abingdon School

6 Fundamentals of computer systems
6.2 Classification of programming languages
Learning objectives:
■■Show awareness of the
development of types of
■■ 6.2.1 Classification of programming languages
programming languages and Low-level programming languages
their classification into low- Low-level programming languages are classified as
and high-level languages • machine code
■■Know that the low-level • assembly language.
languages are considered to be:
EDSAC and machine code
• machine code On May 6th, 1949, EDSAC ran its first program which printed a table of
• assembly language squares for integers in the range 0 to 99. The programme (sic) took two
minutes to run. The program of order codes had been punched on paper
■■Know that high-level tape as 5-bit binary codes (see Figure 5.3.3 in Chapter 5.5). The order
languages include imperative codes represented arithmetic and logical orders, shifts, jumps, data transfer
high-level language orders, input and output orders and stop orders. The word “order” was
■■Describe machine-code literally an order for EDSAC to do something. These order codes were the
language and assembly first programming language, a low-level language known as machine code
language that was interpreted directly by the hardware of EDSAC. Two examples of
these order codes are shown in Table 6.2.1.1 where each 5-bit order code
■■Understand the advantages is expressed as a single letter. The single letter order codes were typed on a
and disadvantages of machine-
machine that punched the corresponding 5-bit code directly onto paper tape
code and assembly language
(see Information panel opposite for the 1951 film on how EDSAC was used
programming compared with
in practice). Addresses were also expressed in decimal and then translated into
high-level language
binary.
programming
Order code Address Description
■■Explain the term ‘imperative
Add the content of location n to the
high-level language’ and its A n
accumulator.
relationship to low-level
Subtract the content of location n from the
languages. S n
accumulator.
Table 6.2.1.1 Examples of EDSAC order codes
Information
Figure 6.2.1.1 shows a snippet of an EDSAC order code program. Each
EDSAC film:
character represents a 5-bit code
http://www.tnmoc.org/special-
projects/edsac/edsac-history
Maurice Wilkes’ 1976 T123SE84SPSPSP10000SP1000SP100SP10SP1S
commentary on the 1951 film
QS#SA40S!S&S@SO43SO33SPSA46S
about how EDSAC was used in
practice. T65ST129SA35ST34SE61ST48SA47ST65SA33SA40S
Figure 6.2.1.1 EDSAC order code

Single licence - Abingdon School 220

6 Fundamentals of computer systems

What is machine code?

Key concept
Machine code is a language consisting of bit patterns/binary codes that a
Machine code: machine can interpret, i.e. execute. For this reason, machine code is referred
Machine code is a language
to as executable binary codes. For example, the EDSAC order code program
consisting of bit patterns/
binary codes that a machine can instruction
interpret, i.e. executable binary 0010100000010101
codes.
means “transfer the content of the accumulator to storage location 21.”
A machine code instruction is an operation which a machine is capable of
Key concept
carrying out. This direct relationship with the hardware gives machine code
Machine code instruction: instructions their low-level classification. Therefore, higher-level operations for
A machine code instruction is which there is no direct machine counterpart have to be broken down into a
an operation which a machine is
sequence of machine code instructions.
capable of carrying out.
What is a machine code program?
A machine code language program is a program consisting of executable binary
Key concept
codes.
Low-level programming
language: Questions
The direct relationship with the
1 What is machine code?
hardware gives machine code
instructions their low-level What is a machine code instruction?
2
classification.
3 Why is machine code classified as a low-level programming language?

Assembly language
Writing programs directly in machine code is challenging. The EDSAC programmers wrote their programs using
letters for the operation to be performed and addresses in decimal using the digit characters '0'..'9'.
The hardware on which they typed these letters and digit characters was wired to punch paper tape with the 5-bit
equivalent of each.
We would call the form of the program shown in Figure 6.2.1.1 which uses letters, an assembly language
program. In assembly language, a (symbolic) name is assigned to each operation/instruction code. The operation/
instruction code name is called a mnemonic or memory jogger. The operation code mnemonic should describe in
some way what the instruction does, e.g. LDR means LoaD a Register, ADD means add - see Table 6.2.1.2. The
address field &1234 is expressed in hexadecimal (& is used to indicate this).
Assembly language Description
LDR means LoaD a Register with content of a
memory location or word, Rd is the symbolic name
LDR Rd, &1234
for the register, &1234 is the memory location’s
address expressed in hexadecimal.
ADD means add content of registers Rn and Rm,
ADD Rd, Rn, Rm
store result in register Rd.
STR means STore the content of the specified
STR Rd, &4321
Register in a memory location or word.
Table 6.2.1.2 Some assembly language instructions
221 Single licence - Abingdon School
 6.2.1 Classification of programming languages

There is a ONE-to-ONE mapping between an assembly language instruction Key concept

and its equivalent machine code language instruction.
Assembly language:
For example, Assembly language is the
LDR Rd, &1234 might be assembled to 000000 0001 01001000110100 symbolic form of machine code.
Each operation/instruction code
The one-to-one mapping makes translating instruction mnemonics into the of machine code is assigned a
binary of machine code a simple task that can be assigned to a computer. The symbolic name or mnemonic
translator is called an assembler. describing what the instruction
does, e.g. ADD.
Questions There is a ONE-to-ONE
mapping between an assembly
4 What is assembly language code? language instruction and
its equivalent machine code
5 What is the mapping between assembly language instructions and
machine code? language instruction.

6 What language translator is required to translate assembly language Information

into machine code?
GNU Fortran:
GNU Fortran is the primary open
source version of the Fortran
High-level languages
compiler widely used both in and
As the 1951 EDSAC film showed, a problem had to be recast by hand into out of academia. It is one of the
a form that could use the machine code language of EDSAC. Wouldn’t it be Fortran compilers available for
much better if the problem could be expressed in a programming language the Raspberry Pi.

much closer to the problem space, leaving the task of translating to machine
code to the computer? This thought led to the development in the 1950s of Key concept
high-level languages, some of which are still used. For example, Fortran (1957) High-level programming
was designed for numerical applications and is still used by mathematicians, language:
scientists and engineers, today. High-level programming
languages are problem-oriented
High-level languages are closer to English than they are to the machine. This and therefore closer to English
means that the mapping from a high-level language statement to machine code than they are to the machine.
will be a one-to-many mapping because each high-level language statement This means that the mapping
from a high-level language
will need to be broken down into several machine code operations. For
statement to machine code will
example, the assignment statement be a one-to-many mapping
x ← y + z because each high-level
when translated could become in the assembly language form of machine code language statement will need
to be broken down into several
LDR R0, &1234 machine code operations.
LDR R1, &1235
ADD R2, R0, R1
Information
High-level language
STR R2, &1236 classification:
Imperative:
Questions • Procedural
• Object-oriented
7 What is meant by the term high-level programming language?
Declarative:
8 What is the mapping between high-level language statements and • Logic
machine code? • Functional

Single licence - Abingdon School 222

6 Fundamentals of computer systems

Information Imperative high-level languages (HLL)

The word “imperative” is derived from the Latin word imperare meaning “to
Imperative languages with
command”. High-level languages that are classified as imperative do just that.
support for functional
They consist of a sequence of commands for actions such as assign, add, write,
programming:
Several imperative languages such read which a programmer has written to solve some problem or accomplish
as Delphi, Visual Basic, C#, Python some task. Table 6.2.1.3 shows a snippet of program code for an imperative
and Javascript now have versions high-level language. Procedural and Object-Oriented Programming languages
of the language with support for
are classified as imperative high-level languages, e.g. Pascal, Delphi, Basic, C,
functional programming, a non-
imperative style of programming.
C++, Java, C#, Python and Javascript.

Imperative
Description
program
y := 6; assign 6 to y
z := 7; assign 7 to z
x := y + z; add z to y and store result in x

Table 6.2.1.3 Imperative high-level program

Another important feature of imperative languages is that their commands
change a program’s state, e.g. Table 6.2.1.3 shows that the variable y has its
state changed by the action of the assign command from whatever value it was
before to 6.

Questions
9 Explain the term imperative high-level language.

Advantages of programming in machine code and assembly

language compared with HLL programming
Information High-level language programs are converted into machine code by a translator
called a compiler. Most compilers attempt to optimise the machine code which
Code optimisers of
compilers:
is produced. The compiler scans the machine code to see if it contains any
The code optimisers built into unnecessary code which it then attempts to remove or adapt. Fewer machine
compilers are less effective for code instructions means the code will take up less memory (smaller footprint)
machine code operations such as well as running more quickly when executed. However, the process is
as complex bit manipulation and
not perfect, for example, where floating-point operations are concerned.
floating-point arithmetic because
these operations are not easily In embedded computer systems, where speed of execution is paramount or
expressed in a high-level language. memory is at a premium, the compiled code can be examined by hand and
sections that are not already optimised replaced by hand-coded assembly
language code, which is then assembled into machine code.
For short sections of code which need to run quickly or take up little space,
it may be better to code directly in assembly language. Some high-level
programming languages allow assembly language code to be embedded (inline)
in the HLL program to take advantage of the time and space efficiency of
assembly language coding.
223 Single licence - Abingdon School
 6.2.1 Classification of programming languages

Assembly language and machine code programming allow direct access Key fact
to registers and low-level operating system routines which is not generally
Adv. of programming in
possible with most high-level language programming languages.
machine code and assembly
language:
Questions
Hand-coded assembly language
when assembled can
10 State three advantages of programming in assembly language compared
• achieve a smaller memory
with programming in a high-level language.
footprint in machine code
than compiled high-level
language code
Disadvantages of programming in machine code and assembly • achieve better code
language compared with HLL programming optimisation than compiled
Code written in assembly language or machine code is less readable than code high-level language code and
written in a high-level language and therefore more difficult to understand therefore code that will run
and maintain, debug and write without making errors. Code written in faster
• directly access registers
assembly language or machine code uses the instruction set of a particular
and low-level operating
processor (processor family). It is therefore machine dependent and will system routines which is not
only execute on processors that use this instruction set. High-level languages possible with most high-level
are machine independent. An HLL program is expressed in an English- programming languages.

like language which is turned into machine code by a compiler. As long

as a compiler exists for a particular instruction set, the HLL program may
be ported to and its compiled version run on a computer with a different
instruction set processor from the one it was written on. HLL programs are Key fact
easier to understand and therefore maintain than assembly language programs
Disadv. of programming in
because they are written using statements that are close to English. They are machine code and assembly
less error-prone when writing for the same reason. language:
Code written in assembly
Questions language or machine code is less
readable than code written in a
11 State three disadvantages of programming in assembly language
high-level language and so more
compared with programming in a high-level language. difficult to
• understand and maintain
• debug
In this chapter you have covered: • write without making errors

■■ Classification of programming languages into low- and high-level Code written in assembly
languages language or machine code is
machine dependent making it
■■ Low-level languages classified as: difficult to port to a different
• machine code instruction set processor
compared with code written
• assembly language
using high-level languages which
■■ Imperative high-level language is a type of high-level language do port readily because they are
■■ Machine-code language and assembly language not machine-oriented.

■■ The advantages and disadvantages of machine-code and assembly language

programming compared with high-level language programming
■■ The meaning of the term ‘imperative high-level language’ and its
relationship to low-level languages.
Single licence - Abingdon School 224
 6 Fundamentals of computer systems
6.3 Types of program translator
Learning objectives:
■■Understand the role of each of
the following:
■■ 6.3.1 Types of program translator
Types of program translator
• assembler
There are three types of program translator:
• compiler • Assembler
• interpreter • Compiler

■■Explain the differences • Interpreter

between compilation and Role of an assembler
interpretation. Programs written in assembly language have to be translated into machine code
before they can be executed. This is done with an assembler.
■■Describe situations in which
each would be appropriate Machine code is a language that the machine can execute, i.e. it is executable
binary code (binary patterns for which machine operations are defined).
■■Explain why an intermediate
language such as bytecode is Assembly language is the mnemonic form of these executable binary codes.
produced as the final output Thus there is a one-to-one correspondence between an assembly language
by some compilers and how it statement and its machine code equivalent: one assembly language statement
is subsequently used. maps to one machine code statement. This is in contrast to a high level
language statement which typically maps to several machine code statements.
■■Understand the difference Role of a compiler
between source code and object
A compiler is a program that reads a program (the source code) written in
(executable) code
a high level programming language (the source language) and translates it
Key principle into an equivalent program (the object code) in another language - the target
Assembler:
language. As an important part of this translation process, the compiler reports
An assembler translates the presence of errors in the source code program.
assembly language into machine A compiler translates (compiles) a high level programming language source
code.
code program into a separate and independently executable object code
One assembly language
statement maps to one machine target language program. The target language program or object code
code statement. produced by the process could be
• Machine code of an actual machine ( in which case the compiler is
called a native language compiler)
• Intermediate code which can, if necessary, be interpreted by an
interpreter, e.g. Java bytecode is an intermediate language produced by
a Java compiler
• Executable code for execution by a virtual machine.
A compiler translates one high level language statement into several machine
code or target language statements.

Single licence - Abingdon School 225

6 Fundamentals of computer systems

Key principle A compiler only translates a high level language program (the whole of the
program), it does not execute it.
Compiler:
A compiler translates a high The process that the compiler engages in is called compiling.
level programming language
source code program into a
A compiler consists of several stages:
separate and independently • Lexical analysis – splits the source into user-defined “words”,
executable object code target
e.g. variable identifiers and language-defined “words”, e.g. While
language program. Object code
is typically machine code. • Syntax analysis – checks that statements are grammatically correct
• Semantic analysis – e.g. type checking, "A" + 3.142 is incorrect as you
A compiler translates one high
level language statement into can’t add a real to a string
several machine code or target • Intermediate code generation
language statements.
• Code optimising
• Code generation
Key principle Role of an interpreter
Interpreter: An interpreter is a program that executes a high level programming language
An interpreter is a program
program, statement by statement, by recognising the statement type of a
that executes a high level
programming language statement, e.g. X = X + 1, and then calling a pre-written procedure/function
program, statement by for the statement type, to execute the statement. Therefore, an interpreter does
statement, by recognising the not, unlike a compiler, produce an independently executable target language
statement type of a statement equivalent of the source language program. The application of interpreter to a
and then calling a pre-written
source code program is called interpreting.
procedure/function for the
statement type, to execute the The differences between compilation and interpretation
statement. The major differences between the compilation and interpretation are:
• An interpreter both “translates” and executes whereas a compiler only
translates.
• A compiler produces a separate independently executable form of the
source code program whereas an interpreter does not.
Key principle
• A compiler is not needed when target form of source program is
Interpreter vs compiler:
executed whereas in the case of the interpreter, execution requires the
An interpreter both “translates”
and executes whereas a source code form of the program together with the interpreter, i.e.
compiler only translates. the interpreter needs to be available on the machine where the program
is being run.
• If an interpreter is used then only the source code form of program is
Key principle needed to execute the program whereas, if a compiler is used then the
Interpreter vs compiler: object code form of program is needed in order to execute the program.
A compiler produces a separate • Interpreters are usually easier to write than compilers.
independently executable form
of the source code program • With the compiler approach, if an error is discovered while the
whereas an interpreter does not. program is executing the source form of program must be located. An
editor and the source form of the program must be loaded. The error
must be pin-pointed which is not always easy and then corrected.

226 Single licence - Abingdon School

 6.3.1 Types of program translator

The compiler must be loaded and a compilation carried out. The new
target form of program must then be loaded and executed. With an
interpreter, the execution is halted at the point where the error occurs.
The interpreter gives precise details of location of error. The error is
corrected with an editor which may be co-located with interpreter. If
it isn’t, an editor will have to be loaded. However, no time-consuming
compilation is involved and execution can resume immediately.
Situations in which assemblers, compilers and interpreters
would be appropriate
Key principle
Assemblers
Interpreter vs compiler:
For time-critical sections of code where execution speed is important, e.g.
Where speed of execution and/
interrupt service routines, assembly language still has a role to play because in or direct access to hardware is
the hands of a skilled programmer, assembly language code can be written that required, use assembly language
is highly optimised for speed. As an assembler simply translates one assembly and an assembler.
language statement into one machine code statement, that optimisation is
preserved. Compilers can optimise code but the binaries produced cannot be
guaranteed to be fully optimised for the given hardware. In the pecking order
of speed, interpreters come after compilers.
Key principle
Interpreter vs compiler:
Assembly language is still used where direct access to hardware is required e.g. Compiled code which has been
processor registers or I/O controller registers. This is the case when writing compiled into machine code of
device drivers, e.g. a screen driver. In this instance an assembler would be the computer will execute a lot
required to translate the assembly language program into machine code. faster than its interpreted source
code equivalent (i.e. interpreter
Compilers and interpreters + the source code equivalent of
It is considerably more productive to write programs in high-level languages the compiled code).
than in assembly language. There are relatively few programmers who are
skilled in writing assembly language programs compared with the number
of programmers skilled in writing in one or more high-level programming
languages. Key principle
Interpreter vs compiler:
Compiled code which has been compiled into machine code of the computer
Where rapid debugging and
will execute a lot faster than its interpreted source code equivalent (i.e. immediate feedback on errors is
interpreter + the source code equivalent of the compiled code). required including pinpointing
the location of both syntax
The immediate feedback and ease of locating errors in source code give
and runtime errors, use an
interpreters an advantage over compilers when developing programs. This interpreter.
advantage is particularly beneficial for novice programmers or when programs
are being prototyped and the write, compile, debug, edit cycle can be too time
consuming. Key principle
Compiling has an advantage over interpreting because it produces a separate Interpreter vs compiler:
Where a separate executable
executable which means that the source code program does not have to be
that can execute independently
distributed. There are plenty of situations where this is desirable such as when of its source code equivalent is
producing commercial software or where there is a requirement is to protect the required, use a compiler.
algorithm or coding technique used.

Single licence - Abingdon School 227

6 Fundamentals of computer systems

Key concept Bytecode

Bytecode is an intermediate language between machine code and high-
Bytecode:
Bytecode is an intermediate level language source code. Bytecode is produced by a compiler which has
language between machine code been designed to translate source code into object code for execution on
and high-level language source a virtual machine based on a type of machine architecture called a stack
code.
machine. You will learn about an alternative type of machine architecture
It is produced by a compiler
called a register machine in Chapter 7.3.1.
which has been designed to
translate source code into Compilers for stack machines are simpler and quicker to build than
object code for execution on a compilers for other machine architectures. For example, for a simple stack
virtual machine based on a stack
machine architecture, the compiled code for the statement x ← x ∗ y + z
machine
would take, minus the comments, the form:

push x /transfer a copy of local variable x to top of stack

push y /transfer a copy of local variable y to top of stack

multiply /multiply the top two items on the stack,replace with result

push z /transfer a copy of local variable z to top of stack

add /add the top two items on the stack,replace with result

pop x /remove top item from stack and store in local variable

The stack operations push and pop are covered in Chapter 2.3.1 of the
Unit 1 textbook and evaluating expressions using a stack covered in
Chapter 3.3.1 of the same textbook. A compiler for this simple stack
machine would output byte-long numeric codes, called opcodes, for the
operations push, pop, multiply and add. These opcodes are known as
bytecodes because they are one byte long and they form the instruction
set of the stack machine.
The bytecode stream issued by the compiler for this example might be as
follows 1a 1b 68 1c 60 3b. This bytestream example uses bytecodes
for the Java Virtual Machine which is a stack-based machine that is
able to interpret Java bytecodes. Table 6.3.1.1 shows the corresponding
interpretation of these bytecodes.
Bytecode Operation
1a push first local variable onto stack
1b push second local variable onto stack
68 pop top two items on stack, multiply them together, push result on stack
1c push third local variable onto stack
60 pop top two items on stack, add them together, push result on stack
3b pop top item on stack, store in first local variable
Table 6.3.1.1 Java bytecodes and their interpretation

228 Single licence - Abingdon School

 6.3.1 Types of program translator

To execute bytecode on a virtual machine requires that it be interpreted by the underlying real machine, i.e. the
bytecode is executed in software running on the underlying real machine. This software is called an interpreter.
Writing a software interpreter to interpret bytecode is an easier task than writing an interpreter to interpret high-
level language source code. All that a bytecode interpeter has to do is parse (identify) and directly execute the
bytecodes, one at a time. This also makes the bytecode interpreter very portable, i.e. very easy to move onto a new
machine with a different instruction set, and very compact.
Interpreting bytecode programs is also much faster than interpreting their high-level language source code program
equivalents because the interpreter written to interpret bytecode has to perform much less work and is therefore
simpler.
Bytecode targets a virtual machine not a real machine and so can run on any machine or operating system for
which a bytecode interpreter has been written.
This means that the same object code can run on different platforms by simply creating an interpreter for the
platform. A compiler that outputs bytecode thus produces object code that is portable.
However, bytecode may be further compiled into machine code for better performance. Some systems, called
dynamic translators, or “just-in-time” (JIT) compilers, translate bytecode into machine language as necessary at
runtime.

Questions
1 Explain the role of each of the following:
(a) assembler
(b) compiler
(c) interpreter
2 State three differences between compilation and interpretation.

3 (a) Give two reasons why programs are still written in assembly language
(b) What is the relationship between
(i) assembly language statement and machine code
(ii) high level programming language statement and machine
code?

4 Given a choice, under what circumstances would it be preferable to use:

(a) a compiler;
(b) an interpreter?

Single licence - Abingdon School 229

6 Fundamentals of computer systems

Stretch & Challenge question

5 A particular computer has two compilers for a high level language HLL. The compilers are called
HLL1 and HLL2. HLL1 compiles a program written in HLL into the machine code of this
computer, whereas HLL2 compiles an HLL program into intermediate code, which can then be
executed by an interpreter running on this computer, if one exists.

On purchase, compiler HLL2 was supplied in intermediate code form without an interpreter, the
same intermediate code that is produced by HLL2, and HLL1 in source code program form.
(a) With only a means to write assembly language programs and to run an assembler on the
computer at this stage, explain carefully what could be done to enable HLL2 to compile
HLL programs for this computer.
(b) Explain carefully how HLL1 can now be executed.
(c) Explain carefully how the machine code form of HLL1 can now be produced on this
computer.

In this chapter you have covered:

■■ The role of each of the following:
• assembler
• compiler
• interpreter
■■ The differences between compilation and interpretation
■■ Situations in which each would be appropriate.
■■ Why an intermediate language such as bytecode is produced as the final
output by some compilers and how it is subsequently used.
■■ The difference between source code and object (executable) code

230 Single licence - Abingdon School

6 Fundamentals of computer systems
6.4 Logic gates
Learning objectives
■■Construct truth tables for the
following logic gates:
■■ 6.4.1 Logic gates
Boolean variables
• NOT In 1847 George Boole, an English mathematician, introduced a shorthand
• AND notation for a system of logic originally set forth by Aristotle. Aristotle’s system
dealt with statements considered either true or false. Here are two examples:
• OR
It is sunny today.
• XOR
Today is Tuesday.
• NAND Quite clearly these two statements are either True or False. If today is
• NOR Wednesday then the statement “Today is Tuesday” is False. Table 6.4.1.1 shows
the possible outcomes of examining the truth of each statement.
■■Be familiar with drawing and
interpreting logic gate circuit Statement Outcome
diagrams involving one or It is sunny today False True
more of the above gates Today is Tuesday False True
■■Complete a truth table for a Table 6.4.1.1 Possible outcomes for truth of statements
given logic gate circuit Just as we might use an integer variable G to record the number of goats in a
farmer’s field so we can use variable X as shorthand for “It is sunny today”, and
■■Write a Boolean expression for Y for “Today is Tuesday”. The values that G can be assigned are the natural
a given logic gate circuit
or counting numbers. For X and Y, we have only two possible values, True or
■■Draw an equivalent logic gate False, to assign. We call X and Y Boolean variables, after George Boole who
circuit for a given Boolean introduced this form of algebra called Boolean algebra. Table 6.4.1.2 shows the
expression Boolean variable equivalent of Table 6.4.1.1 for “It is sunny today” expressed
as Boolean variable X. Boolean algebra deals with Boolean values that are
■■Recognise and trace the logic of
typically labelled True/False (or 1/0, Yes/No, On/Off).
the circuits of a half-adder and
a full-adder X
Meaning
(It is sunny today)
■■Be familiar with the use of the
False It is not sunny today
edge-triggered D-type flip-flop
True It is sunny today
as a memory unit
Table 6.4.1.2 Boolean variable representation of truth statements
Boolean algebra had very little practical use until digital electronics and digital
computers were developed. As digital computers rely for their operation on
using the binary number system, Boolean algebra can be applied usefully in
the design of the electronic circuits of a digital computer. Using Boolean values
1 and 0 instead of True and False, True in Table 6.4.1.2 becomes 1 and False
becomes 0 as shown in Table 6.4.1.3. X = 1 now means that “It is true that it
is sunny today” and X = 0 means “It is not true that it is sunny today”.

Single licence - Abingdon School 231

6 Fundamentals of computer systems

X Meaning
0 It is not sunny today
1 It is sunny today
Table 6.4.1.3 Boolean variable representation of truth statements using
0 in place of False and 1 in place of True
It is then a small step to use Boolean variables to represent the state of
components such as switches and indicator lamps as follows:
• a switch can be either closed (1) or open (0) and
• an indicator lamp can be either on (1) or off (0).

Y Meaning
0 Switch is not closed
1 Switch is closed
Table 6.4.1.4 Boolean variable representation for state of a switch Y

Z Meaning
0 Lamp is not on
1 Lamp is on
Table 6.4.1.5 Boolean variable representation for state of an indicator
lamp Z

Logical OR operation
Things become interesting when switches and lamps are combined together in
circuits. Figure 6.4.1.1 shows a simple circuit consisting of two switches wired
Switch in parallel, one indicator lamp and one battery.
X Y Q X Y Q
X Q Open Open Off 0 0 0
Open Closed On 0 1 1
Lamp 1 0 1
Closed Open On
Y Closed Closed On 1 1 1
Battery
Figure 6.4.1.1 OR logical operation: switch arrangement,
switch state combinations and corresponding lamp state
The lamp is on if switch X is closed OR if switch Y is closed OR if both are
closed, otherwise the lamp is off. The state of the switches can be expressed in
the two Boolean variables, X and Y, as open or closed or using 0 for open and
1 for closed. The state of the lamp can also be expressed in a Boolean variable,
Q, because the state has two possible values, off or on, which can be coded as 0
and 1, respectively.
Just as we can write the number equation for the total number of goats G, a
farmer possesses,
G=X+Y

232 Single licence - Abingdon School

 6.4.1 Logic gates

where X is the number in the first goat pen and Y is the number in the second,
X
so we can write for the lamp circuit the Boolean equation
Y OR Q
Q=X+Y
Figure 6.4.1.2 Logical OR
The operator "+" denotes the logical OR operation that behaves according to operation: block diagram
the tables in Figure 6.4.1.1, e.g. if X = 1 and Y = 1 then Q = 1, i.e. the lamp is showing inputs X and Y and
on. output Q
Logical OR Truth Table
If the logical operator "+" is represented by a rectangle labelled OR (Figure X Y Q
6.4.1.2) then Boolean variables X and Y become its inputs and Q becomes its 0 0 0

output. The inputs X and Y are transformed by the logical OR operation into 0 1 1
1 0 1
Q.
1 1 1
In fact, the logical OR operation defines a Boolean function OR because it
Figure 6.4.1.3 Logical OR
operates on binary inputs and returns a single binary output (Figure 6.4.1.4).
truth table
Figure 6.4.1.2 is called a block diagram. A single block in a block diagram is
sometimes called a black box even though it is not coloured black. Logical OR
The black box approach is a convenient way of representing the logical OR Inputs: X, Y
operation with the details of how it is implemented abstracted away. We now Output: Q
define the logical OR operation by its truth table (Figure 6.4.1.3) not by the Function: OR = X + Y
particular details of its implementation which could be, for example, electronic, Figure 6.4.1.4 Logical OR
magnetic, optical, biological, hydraulic, or pneumatic. function

Logical AND operation Key point

Figure 6.4.1.5 shows a simple circuit consisting of two switches wired in series,
Boolean function:
one indicator lamp and one battery.
A Boolean function is a function
that operates on binary inputs
X Y Q X Y Q
and returns a single binary output.
Open Open Off 0 0 0
Open Closed Off 0 1 0
Closed Open Off 1 0 0 Switches
1 1 1
Q
Closed Closed On
Figure 6.4.1.5 AND logical operation: switch arrangement, X Y
switch state combinations and corresponding lamp state Lamp
In this case, the lamp is only on if both switch X is closed
Battery
AND switch Y is closed, otherwise the lamp is off. Again,
the state of the switches can be expressed in the two Boolean
variables, X and Y, as open or closed or using 0 for open and
1 for closed. The state of the lamp can also be expressed in a Boolean variable,
Q, because the state has two possible values, off or on, which can be coded as 0
and 1 respectively.
We can write for the lamp circuit the Boolean equation
Q=X.Y

Single licence - Abingdon School 233

6 Fundamentals of computer systems

X Y Q The operator "." denotes the logical AND operation that acts according to the
0 0 0 truth table in Figure 6.4.1.6, e.g. if X = 1 and Y = 1 then Q = 1, i.e. the lamp is
0 1 0 on.
1 0 0 Figure 6.4.1.7 shows the block diagram representation of the logical AND
1 1 1
operation with inputs X and Y transformed into output Q.
Figure 6.4.1.6 AND truth
table
Questions
1 Draw the arrangement of switches that produce output Q
X
Y AND Q where Q = X.Y + X.Z

Figure 6.4.1.7 Logical AND Logical NOT operation

operation: block diagram It is the convention to use Boolean value 1 for the active state, e.g. lamp on,
Q = 1, and the Boolean value 0 for the inactive state, e.g. lamp off, Q = 0.
Another way of expressing lamp off is NOT lamp on. This insight leads to
NOT lamp on = lamp off X Q
X NOT Q Or, NOT 1 = 0 0 1
1 0
Figure 6.4.1.8 Logical NOT And, NOT lamp off = lamp on
Figure 6.4.1.9 NOT truth
operation: block diagram Or, NOT 0 = 1 table
The NOT black box in Figure 6.4.1.8 transforms input X into output Q using
the logical NOT operation which inverts its input, 0 → 1, 1 → 0.
Q = NOT X
Q=X
The line or bar placed over X is shorthand for NOT or the invert operation.

Logical NAND operation

If the output of the AND operation is inverted (Figure 6.4.1.10) then we have
the NAND logical operation. Its Boolean equation is X Y Q
0 0 1
X AND NOT Q
Q=X.Y
0 1 1
Y
1 0 1
Figure 6.4.1.10 Logical NAND 1 1 0
Its truth table is Figure 6.4.1.11.
operation: constructed from Figure 6.4.1.11 NAND
an AND and a NOT, block Logical NOR operation truth table
diagram If the output of the OR operation is inverted (Figure 6.4.1.12) then we have
the NOR logical operation. Its Boolean equation is
X Y Q
0 0 1
X OR NOT Q Q=X+Y
0 1 0
Y
1 0 0
Figure 6.4.1.12 Logical NOR 1 1 0
operation: constructed from an Its truth table is Figure 6.4.1.13. Figure 6.4.1.13 NOR truth
OR and a NOT, block diagram
table
234 Single licence - Abingdon School
 6.4.1 Logic gates

Logical XOR operation X Y Q

The truth table for the eXclusive-OR (XOR) operation (Figure 6.4.1.14) 0 0 0
shows Q to be 1 if X is 1 and Y is 0 (Y not 1) or if X is 0 (X not 1) and Y is 1. 0 1 1
Its Boolean equation is thus 1 0 1
Q=X.Y + X.Y 1 1 0

It has its own symbol ⊕ so the Boolean equation is written as follows Figure 6.4.1.14 XOR truth
table
Q = X ⊕Y
X X Y
AND Key point
Y NOT Y
OR Q Logic gate:
A logic gate is a physical device
NOT X
that implements a Boolean
AND function.
X Y
Figure 6.4.1.15 XOR logical operation: block diagram constructed from
NOTs, ANDs and an OR
Questions X Y Z Q
0 0 0
2 Q is only 1 if both X and Y are 1 and Z is 0 or if both Y and Z are 1 and X is 0
0 0 1
or if X, Y and Z are 1. Complete the truth table.
0 1 0
3 Draw the block diagram using AND and OR operations that produces output Q 0 1 1
where Q=X.Y+X.Z 1 0 0
1 0 1
4 Draw the block diagram for logical operations that produce output Q where 1 1 0
Q=X.Y+X.Z 1 1 1

Logic gates Logic gate symbol Logical operation

The logical operations above are implemented in electronic circuits
as logic gates. The circuit symbols for these logic gates are shown in OR
Table 6.4.1.6.

Drawing and interpreting logic gate circuit diagrams NOR

Logic gates may be connected together to perform a variety of
logical operations. The output of one gate is used as the input to
AND
other gates.
For example, in Figure 6.4.1.16, Boolean variable, E, is the output
of an AND gate and the input to an OR gate. NAND
The full circuit uses Boolean variables, A, B, C, D, E, F, Q
as follows E=A.B XOR
F = C . D
Q = E + F
NOT
therefore Q=A.B+C.D
Table 6.4.1.6 Logic gate symbols (ANSI/IEEE standard 91-1984)

Single licence - Abingdon School 235

6 Fundamentals of computer systems

Questions A
B E
5 What is the output of this logic gate circuit
when its input is (a) 0 (b) 1? Q
C F
6 What is the output of this logic gate circuit when its D
input is (a) 0 (b) 1? Figure 6.4.1.16 Three gates
connected to perform a logical
operation

Questions

C
7 What is the output, Q, of this logic gate circuit when its A
inputs A and B are (a) both 0 (b) both 1 Q
(c) different from each other?
B D
8 What is the output of this logic circuit when A1 A0
(a) A0 = B0 and A1 = B1?
(b) A0 ≠ B0 and A1 ≠ B1?
(c) A0 ≠ B0 and A1 = B1? Q
(d) A0 = B0 and A1 ≠ B1?

B1 B0

9 What is the purpose of the logic circuit in Q8?

Truth table equivalent of a logic gate circuit

A A truth table can be used to analyse the behaviour of a logic gate circuit when
C
inputs are applied to it. For the logic gate circuit shown in Figure 6.4.1.17,
Q
there are two inputs, A and B, and one output Q. The first two columns in
B D
Table 6.4.1.7 contain all possible combinations of values for inputs A and B.
Figure 6.4.1.17 Three gates Column C contains the values NOT A and column D the values NOT B. Q’s
connected to perform a logical column contains the values for C NORed with D, e.g. 1 NOR 1 → 0.
operation
A B C D Q
1
0 Figure 6.4.1.18 traces Boolean values through 0 0 1 1 0
0 the three gates, for A = 1 and B = 0. 0 1 1 0 0
1 1 0 0 1 0
0 Information 1 1 0 0 1
Figure 6.4.1.18 Tracing Logic simulator: Table 6.4.1.7 Truth table
Boolean values through the A freeware logic simulator is
for logic circuit shown in
three gates available from
http://www.cburch.com/logisim/ Figure 6.4.1.17

236 Single licence - Abingdon School

 6.4.1 Logic gates

Questions A B C D Q C
0 0 A
10 Complete the truth table for this logic gate Q
0 1
circuit using Boolean variables A, B, C, D
1 0 B
and Q. 1 1
D

11 Draw the truth table for this logic gate circuit using A
Boolean variables A, B and Q. B
Q
12 Draw the truth table for this logic gate circuit using
Boolean variables
A
A, B, C, D and Q. B
(Note: the table will
have 16 rows for the Q
values of
the Boolean variables). C
D

Questions A 0 Q
A 0 0
13 Complete the truth table for this logic circuit. Q
0 1 0

A A 1 Q
14 Complete the truth table for this logic circuit. 1
Q
0 1
1 1
15 Complete the truth table for this logic gate circuit.
A B D

A B D C3 C2 C1 C0
C3 0 0 1 See Figure 6.4.1.32 for
0 1 1 an example of how this
1 0 1 circuit could be used.
C2
1 1 1
0 0 0
C1 0 1 0
1 0 0
1 1 0
C0

Single licence - Abingdon School 237

6 Fundamentals of computer systems

Boolean expression equivalent of a logic gate circuit

The logic gate circuit in Figure 6.4.1.19 may be expressed using Boolean
variables, A, B, C, D and Q and the logical operators, NOT and NOR
as follows C = NOT A = A
A
C
Q D = NOT B = B

B D
Q=C+D
Figure 6.4.1.19 Three gates
connected to perform a logical therefore Q=A+B
operation
A + B is the Boolean expression equivalent of the logic gate circuit shown in
Figure 6.4.1.19. If we examine Figure 6.4.1.6 carefully we see that the output
from this logic gate circuit for inputs A and B is that of the truth table for an
AND logic gate. Therefore, another equivalent Boolean expression is A . B.
Thus for the two Boolean expressions
A . B is equivalent to A + B
This means that we could replace the logic circuit in Figure 6.4.1.19 by an
AND gate with inputs A and B.

Logic gate circuit equivalent of a given Boolean expression

Consider the following Boolean expression

A.B
To convert this into an equivalent logic gate circuit we must take each term in
the expression, starting with the innermost, and apply each operation in turn.
The innermost terms are
A and B
applying the NOT operation to each
A and B
turning these into equivalent logic gates

A A B B

applying the next operation AND

A.B
turning this Boolean expression into its equivalent logic gate circuit

A A
A B
B B

238 Single licence - Abingdon School

 6.4.1 Logic gates

Finally applying the NOT operation to

A A
A B
A B
B B

The number of gates can be reduced to three by replacing the AND-NOT combination by its logic gate equivalent
NAND. The logic gate circuit becomes

A A
A B
B B

For more complicated Boolean expressions the same approach is used to arrive at the equivalent logic gate circuit,
e.g.
A.B + A.B

A . B is converted to its equivalent logic gate circuit W

A . B is converted to its equivalent logic gate circuit Z

the output of these two circuits, W and Z is then ORed together, W + Z

If the order of evaluation needs to be controlled, brackets are used as the following example demonstrates

(A + B) . (A + B)
(A + B) is converted to its equivalent logic gate circuit W

(A + B) is converted to its equivalent logic gate circuit Z

the output of these two circuits, W and Z is then ANDed together, W . Z

Questions
A
16 Write an equivalent Boolean expression in terms of B
A, B, C and D for this logic gate circuit. Q
C
17 Write an equivalent Boolean expression in D
terms of A, B for this logic gate circuit.
A
B
Q

Single licence - Abingdon School 239

6 Fundamentals of computer systems

Half-adder
To make a system to add together two binary digits A and B to produce a Sum
and a Carry digit requires the following understanding of binary arithmetic

A A + B = Sum and Carry

Sum 0 + 0 = 0 and 0
B
0 + 1 = 1 and 0
1 + 0 = 1 and 0
Carry 1 + 1 = 0 and 1
Table 6.4.1.8 Binary addition with Sum and Carry
Figure 6.4.1.20 Half-adder
You will notice that the A, B and Sum columns mirror the truth table for
constructed from an
the exclusive-OR logical operation which means that A + B = Sum can be
exclusive-OR and an AND
implemented with an XOR logic gate (Figure 6.4.1.20). The A, B and Carry
logic gate
columns mirror the truth table for the AND logical operation and so Carry can
be generated with an AND logic gate.
1
0 Figure 6.4.1.21 shows the block diagram for a half-adder and Table 6.4.1.9
0 1 shows its truth table. This is the first step towards adding binary numbers
which is to be able to add two bits. The result of the operation A + B is a
Most Least two-bit number (Figure 6.4.1.22). If we label the least significant bit of the
significant significant addition Sum, and the most significant bit Carry, then we can treat the half-
bit bit adder as two functions as shown in Table 6.4.1.10.
(Carry) (Sum)
A HALF Sum
Figure 6.4.1.22 A + B where
A = 1 and B = 0 B ADDER Carry
Figure 6.4.1.21 Half-adder block diagram
Inputs Outputs
A B Carry Sum
Half-adder
0 0 0 0
Inputs: A, B
0 1 0 1
Outputs: Sum, Carry
1 0 0 1
1 1 1 0 Function: Sum = Least significant bit of A + B
Carry = Most significant bit of A + B
Table 6.4.1.9 Truth table for
Table 6.4.1.10 Half-adder as two Boolean functions, Sum and Carry
half-adder

Questions
18 Trace the operation of the half-adder logic gate circuit shown in
Figure 6.4.1.20 for inputs A = 1 and B = 1.

240 Single licence - Abingdon School

 6.4.1 Logic gates

Full-adder
0 0 0 1 Carry 1 1 1 1
A pair of binary numbers, A and B, can be added
0 1 0 1 A 0 1 1 1
digit by digit from right to left, by the simple rules
+ 1 0 0 1 B + 1 0 1 1
of arithmetic. First, the two right-most digits, also
0 1 1 1 0 A+B 1 0 0 1 0
called the least significant bits (LSB) of the two binary
numbers, are added. no overflow overflow
Figure 6.4.1.23 Adding two binary numbers with carry,
Next, the resulting carry bit (which is either 0 or 1)
one case with overflow and one without
is added to the sum of the next pair of bits up the
significance ladder. The process is continued until Inputs Outputs
the two most significant bits (MSB) are added. If the A B C Carry Sum
last bit-wise addition generates a carry of 1, overflow 0 0 0 0 0
0 1 0 0 1
has occurred, otherwise, the addition completes
1 0 0 0 1
successfully. Figure 6.4.1.23 shows two different cases,
1 1 0 1 0
one with overflow and one without.
0 0 1 0 1
Computer hardware for binary addition of two n-bit 0 1 1 1 0
numbers can be built from logic gates designed to 1 0 1 1 0
calculate the sum of three bits (pair of bits plus carry 1 1 1 1 1

bit). The transfer of the resulting carry bit forward

to the addition of the next significant pair of bits can A
be easily accomplished by appropriate wiring of logic FULL Sum
gates. The logic gate circuit that does this is called a
B ADDER
full-adder. C Carry
Figure 6.4.1.24 shows the block diagram for a full-
adder, designed to add three bits, a pair of bits A, B, Full-adder
and a carry bit C. In a similar way to the half-adder Inputs: A, B, C
case, the full-adder produces two outputs: the least Outputs: Sum, Carry
significant bit of the addition, and the carry bit. Function: Sum = Least significant bit of A + B + C
Carry = Most significant bit of A + B + C
You should note that the Sum column of the first half Figure 6.4.1.24 Full-adder block diagram and its
of the full-adder truth table (Figure 6.4.1.24) is the truth table
same as that for the half-adder Sum function (Table
6.4.1.9). A 1
Sum
You should also note that the Sum column of the B
second half of the truth table (Figure 6.4.1.24) is
A Sum
the same as that for the half-adder Sum function
B
plus another half-adder Sum function with one
0
input set to 1 (Figure 6.4.1.25) – see Q14.
In fact, the first half of the truth table (Figure
6.4.1.24) is the same as that for the half-adder Sum A Sum
function plus another half-adder Sum function B
with one input set to 0 (Figure 6.4.1.25).
Figure 6.4.1.25 XORs of two half-adders implement the
Setting this input to 0 simply causes the other Boolean function Sum = A + B + C
input to be unchanged (see Q13)

Single licence - Abingdon School 241

6 Fundamentals of computer systems

Figure 6.4.1.26 shows the conditions for generating Carry = 1.

C = 0/1
A=1
0 0
B=1

0
1 Carry
Carry

2nd HALF-ADDER
1st HALF-ADDER
C=1
A =1 B = 0
Figure 6.4.1.26 Conditions that generate
A =0 B = 1 1
carry = 1: two in the upper diagram, Sum = 1
two in the lower Carry = 0 1
1st HALF-ADDER Carry

2nd HALF-ADDER
Note that the first half-adder carry output is the inverse of the second half-
adder output: 1/0 and 0/1 respectively. Therefore if the two carry outs, one
from the first half-adder and one from the second are made the first and second
inputs of an OR-gate then its output will be the final carry. Figure 6.4.1.27
shows the final circuit. C becomes the CARRY IN from the previous digit
column. A and B are the current digit column’s bits to be added and CARRY
OUT is the final carry from the current column.

CARRY IN

A SUM
B
PARTIAL SUM 2nd HALF-ADDER

1st HALF-ADDER PARTIAL CARRY 2

CARRY
OUT
PARTIAL CARRY 1
Figure 6.4.1.27 Full-adder logic gate circuit

Questions
19 Trace the operation of the full-adder logic gate circuit shown in Figure 6.4.1.27 for inputs A = 1 and
B = 1 and CARRY IN = 1.

242 Single licence - Abingdon School

 6.4.1 Logic gates

The logic gates and logic gate circuits that we have considered so far are known
as combinational logic circuits. Combinational logic circuits compute functions
that depend solely on combinations of input values. Combinational logic
circuits are used in the construction of the Arithmetic and Logic Unit (ALU)
which lies at the heart of the central processing unit (CPU) – see Chapter
7.3.1.
Edge-triggered D-type flip-flop Key fact
Although combinational logic circuits provide many important processing The D-type flip-flop is the building
functions, they cannot maintain state. In addition to computing values such as block for all the hardware devices
5 + 6, computers must also be able to store and recall values, i.e. they must be that computers use to maintain
state, from single-bit cells to reg-
equipped with memory elements that preserve data over time. These memory
isters to arbitrarily large random
elements are built from sequential logic gate circuits called flip-flops.
access memory (RAM) units.
Flip-flops are the elementary building blocks of all memory devices used in
typical modern computers.
A flip-flop sequential logic circuit is one in which data is captured and
“committed to memory” at a specific moment in time. Time in a computer is
provided by a master clock that delivers a continuous train of alternating binary
signals. The master clock is an oscillator that alternates between two phases
labelled 0 and 1, or low and high with the transition between the two phases
called an edge (Figure 6.4.1.28).
ADVANCING TIME
HIGH 1
Rising edge Falling edge
LOW 0
Figure 6.4.1.28 Clock signal of alternating binary signals

A clock signal applied to a flip-flop triggers the flip-flop, on an edge of the

clock signal, into updating its “memory”. It is this characteristic that gives the D Q
flip-flop its name. The flip-flop is designed to be triggered by either a rising
edge (low to high) or a falling edge (high to low).
CLK
A data or D-type flip-flop is one that consists of a single-bit data input, a
single-bit data output and a single clock input as shown in Figure 6.4.1.29.
Figure 6.4.1.29 Edge-triggered
The description “edge-triggered” in edge-triggered D-type flip-flops is
D-type flip-flop triggered by a
redundant because all flip-flops are edge-triggered. If it wasn’t edge-triggered
rising edge
but was level-triggered it would be a latch.
D Clock Q
Table 6.4.1.11 describes the behaviour of an edge-
0 ↑ 0
triggered D-type flip-flop for the case when triggering
1 ↑ 1
takes place on a rising edge.
Q is
0 or 1 not rising
The symbol ↑ means a rising edge clock signal. "Not unchanged
rising" means the clock signal is not producing a Table 6.4.1.11 States of an edge-triggered D-type flip-flop
rising edge currently, either because it is at another triggered by a rising edge

Single licence - Abingdon School 243

6 Fundamentals of computer systems

part of its low-high cycle or it is disabled. The state of the device changes on the
rising edge of the clock signal only.

D Clock Q
0 ↓ 0
1 ↓ 1
Q is
0 or 1 not falling
D Q unchanged
Table 6.4.1.12 States of an edge-triggered D-type flip-flop triggered by a
falling edge
CLK
Table 6.4.1.12 describes the behaviour of an edge-triggered D-type flip-flop for
Figure 6.4.1.30 Edge-triggered the case when triggering takes place on a falling edge. Figure 6.4.1.30 shows
D-type flip-flop triggered by a the circuit symbol for an edge-triggered D-type flip-flop triggered by a falling
falling edge edge (○).
Figure 6.4.1.31 shows a typical circuit for using a D-type flip-flop to
“memorise” a datum. While the datum bit is being set up on the D input, the
clock signal is prevented from reaching the clock input, CLK, of the D-type
flip-flop by setting the Enable Clock Signal to the AND gate to 0.

The output of the AND gate is then guaranteed to be zero. Next, the Enable
Clock Signal is set to 1, enabling the AND gate to pass the clock signal through
to the CLK input of D-type flip-flop. On the rising edge of the clock signal,
the output Q assumes the same value as applied to input D. The Enable Clock
Signal is then set to 0, disabling the AND gate from
D Q passing the clock signal to input CLK.
Enable
Clock Signal The input to D may now be changed without
1/0 affecting output Q. The D-type flip-flop remembers
CLK
the value applied to input D at the time the clock
signal last reached CLK.
Clock Signal

Master
Clock

Figure 6.4.1.31 Edge-triggered D-type flip-flop

connected via an AND gate to master clock

Questions
20 The output Q of an edge-triggered D-type flip-flop is currently 0. Explain how this output could be
updated to Q = 1 using the circuit shown in Figure 6.4.1.31. Pay particular attention to the order of
events.

244 Single licence - Abingdon School

 6.4.1 Logic gates

Stretch and challenge extension

Questions
21 Figure 6.4.1.32 shows a 4 x 2 memory.
(a) Which row is selected when A0 = 0 and A1 = 0?
(b) What is the value of each Q for the selected row when D0 = 1 and D1 = 0 and the Write signal has
been applied?

D1
Data in
D0

Address Write
A1 A0 D Q D Q
Row 1
CLK CLK

D Q D Q
Row 2
CLK CLK

D Q D Q
Row 3
CLK CLK

D Q D Q
Row 4
CLK CLK

Figure 6.4.1.32 Logic gate diagram for a 4 x 2 memory. Each row is one of the four 2-bit words. A row is
selected by setting the value of A0 A1. Data is sent along the Data in wires D0 D1 and written to the selected
D-type flip-flops by the Write signal.

In this chapter you have covered:

■■ Constructing truth tables for the following logic gates:
• NOT
• AND
• OR
• XOR
• NAND
Single licence - Abingdon School 245
6 Fundamentals of computer systems

• NOR
■■ Drawing and interpreting logic gate circuit diagrams involving one or
more of the above graphs
■■ Completing a truth table for a given logic gate circuit
■■ Writing a Boolean expression for a given logic gate circuit
■■ Drawing an equivalent logic gate circuit for a given Boolean expression
■■ Recognising and tracing the logic of the circuits of a half-adder and a full-
adder
■■ The use of the edge-triggered D-type flip-flop as a memory unit

246 Single licence - Abingdon School

6 Fundamentals of computer systems
6.5 Boolean algebra

Learning objectives:
■■Be familiar with the use of
■■ 6.5.1 Using Boolean algebra
Boolean algebra
Boolean identities and
In Boolean algebra as in the algebra you have studied in Maths, variables are
De Morgan’s laws to manipulate
combined into expressions with Boolean operators that obey certain laws
and simplify Boolean expressions
(rules).
Boolean variables
A B A•B The variables that we have used so far are known as Boolean variables because
0 0 0 they are two-state variables whose states have the values 0 and 1. These are not
0 1 0 the 0 and 1 of arithmetic but represent True and False.
0 1 0 Boolean operators
1 1 1
We need only consider three operators because all other operators can be
Table 6.5.1.1 Truth table for expressed in terms of these. They are the
AND function + operator denoting Boolean addition

A B A+B • operator denoting Boolean multiplication

0 0 0 operator denoting Boolean inversion
0 1 1
Boolean functions
0 1 1
We have encountered the Boolean AND function, the Boolean OR
1 1 1
function and the Boolean NOT function in Chapter 6.4.1 where they were
Table 6.5.1.2 Truth table for
implemented by logic gates with inputs, A and B, for AND and OR, and A for
OR function
NOT. A and B are Boolean variables.

A B A • B A + B A • B + (A + B) These Boolean functions and Boolean expressions containing Boolean

0 0 0 0 0 operators are equivalent as shown below
0 1 0 1 1
AND(A, B) = A • B OR(A, B) = A + B and NOT(A) = A
1 0 0 1 1
This means that Boolean algebra can be used to design logic gate circuits.
1 1 1 1 1
Combining Boolean functions
Table 6.5.1.3 Truth table for
We also learned in Chapter 6.4.1 that we can combine Boolean
the two pairs of switches in
functions. The functions AND and OR may be combined in exactly the
parallel shown in Figure 6.5.1.1
same way that we can, for example, combine pairs of switches that can
perform these functions. A AND B

OR

A
OR

B
Figure 6.5.1.1 Pairs of switches connected in parallel
Single licence - Abingdon School 247
6 Fundamentals of computer systems

Table 6.5.1.3 shows the truth table for the combination of switches in Figure
6.5.1.1. The two switches labelled A are ganged together as shown by a dotted
line, as are the two switches labelled B. This means that when the first switch A
is open so is the second switch A and when the first switch A is closed so is the
second A. Likewise for the two switches B.
Note that the A + B column of Table 6.5.1.3 is exactly the same as the
A • B + (A + B) column. Therefore, we can say that A • B + (A + B) is
equivalent to A + B.
We draw the conclusion that the Boolean expression A • B + (A + B) can be
simplified to A + B.

In general, many Boolean expressions may be simplified to Boolean expressions

containing fewer terms. This is important if we are using Boolean algebra to
design logic gate circuits because fewer terms means fewer gates.

Simplifying Boolean expressions

Boolean identities
In mathematics, an identity is a statement true for all possible values of its
variable or variables. For example, the algebraic identity of A x 1 = A tells us
that anything (A) multiplied by 1 equals the original “anything,” no matter
what value that “anything” (A) may be. Like ordinary algebra, Boolean algebra
has its own unique identities based on the states 0 and 1 of Boolean variables as
shown in Figure 6.5.1.2.
Did you know? A 0 A+0 A 1 A+1
Simplifying Boolean A+0=A 0 0 0 A+1=1 0 1 1
expressions: 1 0 1 1 1 1
Inspecting truth tables is a
useful way of identifying A A A+A A A A+A
simplifications. A+A=A 0 0 0 0 1 0
A+A=1
1 1 1 1 0 1

0 A 0•A 1 A 1•A
0•A=0 0 0 0 1•A=A 1 0 0
0 0 0 1 1 1

A A A•A A A A•A
A•A=A 0 0 0 A•A=0 0 1 0
1 1 1 1 0 0

A A A
Figure 6.5.1.2 Boolean identities and
their truth tables
A=A 0 1 0
1 0 1

248 Single licence - Abingdon School

 6.5.1 Using Boolean algebra

Laws of Boolean algebra

Commutative law
A+B=B+A
Does it matter in which order inputs A and B are presented A B
A + B ≡ B + A
to an OR-gate or an AND-gate? The answer is no. B A

Similarly, placing A to the left of the + operator or to the ≡ means identical to

right doesn’t matter, we still get the same answer. The same
A•B=B•A
applies to the • operator. That is what is meant by saying A B
A • B ≡ B • A
that the + operator and the • operator are commutative. B A

Associative law
Does it matter whether the operator
is applied to B and C first or to A + (B + C) = (A + B) + C
A and B first? The answer is no as
A A
A + (B + C) A +B
long as it is the same operator. B
That is what is meant by saying that ≡

the + operator and the • operator B

B +C (A + B) + C
C C
obey the associative law.
Take care because the associative A • (B • C) = (A • B) • C
law does not say that A + (B • C)
A A
has to be the same as (A + B) • A • (B • C) A • B
B
C where the term in brackets is ≡

evaluated first. It is not. When B

B • C (A • B) • C
C C
written as A + B • C
B • C is evaluated first because • has
a higher precedence than +.

Distributive law operator1 operator2

This law applies where a term or terms have been bracketed as follows • +
+ •
A operator1 (B operator2 C) = A operator1 B operator2 A operator1 C
Table 6.5.1.4 Operators for
where operator1 may be • and distributive law
operator2 + or operator1 may be + A • (B + C) = A • B + A • C
and operator2 • as shown in Table A
A • (B + C)
A A • B
B
6.5.1.4. A • B + A • C
≡

Note the use of brackets to define B

B +C
A
C C A • C
the order of evaluation. A • B + C is
not the same as A • (B + C) because • A + (B • C) = (A + B) • (A + C)
has a higher order of precedence than
+ and is therefore evaluated before + A A A + B
A + (B • C)
B
unless brackets are used. ≡ (A + B) • (A + C)

B A
B • C
C C A + C

Single licence - Abingdon School 249

 6.5.1 Using Boolean algebra

However, A + (B • C) = A + B • C because the brackets are redundant since • has higher precedence than +.
Similarly, (A + B) • C is not the same as A + B • C because B • C is evaluated first in A + B • C whereas A + B is
evaluated first in (A + B) • C and (A + B) • C = A • C + B • C.
Examples

Simplify A•A+B•1 The Boolean identities used in

A • A + B • 1 = A + B • 1 using A • A = A the examples and their truth
tables are shown below
A + B • 1 = A + B using B • 1 = B
A A A•A
A•A=A 0 0 0
Simplify A • (B + 1)
1 1 1
A • (B + 1) = A • 1 using B + 1 = 1
1 A 1•A
A • 1 = A using A • 1 = A 1•A=A 1 0 0
1 1 1
Simplify B•A+B
A 1 A+1
B • A + B = B • (A + 1) using the distributive law A+1=1 0 1 1
1 1 1
B • (A + 1) = B • 1 using A + 1 = 1
B • 1 = B using B • 1 = B A A A•A
A•A=0 0 1 0
1 0 0
Simplify A • (A + B)
A 0 A+0

A • (A + B) = A • A + A • B using the distributive law A+0=A 0 0 0
A•A+A•B=0+A•B using A • A = 0 1 0 1

0+A•B=A•B using 0 + X = X

Show that (A + B) • (A + B) = A • B + B • A

(A + B) • (A + B) = A • A + A • B + B • A + B • B using distribution law

A•A+A•B+B•A+B•B=0+A•B+B•A+0 using A • A = 0 and B • B = 0

0+A•B+B•A+0=A•B+B•A using 0 + X = X and X + 0 = X

Questions
Information
Using Boolean algebra show that
Writing A • B as AB:
1 (A + B) • (A + B) = A • B + B • A We can omit the • operator and
write the Boolean variables one
2 A(A + B)(B + C) = ABC after another, e.g. A • B as AB.
Examples
3 Use a truth table for each question above to verify that the identities are true.

Single licence - Abingdon School 250

6 Fundamentals of computer systems

Key principle Show that A+A•B=A

Redundancy theorem: A + A • B = A • (1 + B) A is common factor

In a sum of products Boolean
expression, e.g. A + A • B, A • (1 + B) = A • 1 using (1 + B) = 1
a product such as A • B that A • 1 = A using A • 1 = A
contains all the factors of
another product, A • 1, is
redundant.
Show that B+A•B=A+B

B + A • B = (B + A) • (B + B) using distributive law
Distributive law for Boolean
(B + A) • (B + B) = (B + A) • 1 using B + B = 1
variables X, Y, Z:
X + (Y • Z) = (X + Y) • (X + Z) (B + A) • 1 = B + A using X • 1 = X
B + A = A + B using commutative law

Information Simplify A+A•B

Product: We could use the distributive law immediately but it

A • B is known as a product is useful to be aware of other techniques:
Boolean expression.
A + A • B = A • 1 + A • B using A • 1 = A
Sum:
A • 1 + A • B = A • (1 + B) using distributive law
A + B is known as a sum Boolean
expression. A • (1 + B) = A using 1 + B = 1
Sum of products:
A • B + B • C is known as a sum
of products Boolean expression.
Questions
Product of sums:
(A + B) • (B + C) is known Using Boolean algebra show that
as a product of sums Boolean
expression. 4 A+A•B= A

5 A+A•B= A+B

6 A+A•B+B•C= A+B+C

7 A + A • C + B + D • (B • C + A • C) = A + B + C + D
(HINT: Use the result proved in question 5)
Using Boolean algebra simplify the following

8 ABC+ABC

9 A • (A + B) + A • B

10 (A • B + A • B) • A • B + A • B

11 Show that (A + B) • (A + B) = A • B + B • A

251 Single licence - Abingdon School

 6.5.1 Using Boolean algebra

De Morgan’s laws Information

De Morgan’s laws expressed in a form that is useful for designing logic circuits
Propositional logic form of
are as follows
De Morgan’s laws:
A + B = (A • B)
1. ¬(A ⋁ B) = (¬A ⋀ ¬B)
A • B = (A + B)
Table 6.5.1.4 demonstrates the equivalence of A + B and (A • B). 2. ¬(A ⋀ B) = (¬A ⋁ ¬B)
Table 6.5.1.5 demonstrates the equivalence of A • B and (A + B) . where ⋁ means OR, ⋀ means
AND, ¬ means NOT.

A + B = (A • B) A B A+B A B (A • B) (A • B)
Using + and • in place of ⋁ and
0 0 0 1 1 1 0 _
⋀ respectively, and in place of
0 1 1 1 0 0 1 ¬, De Morgan’s laws become
1 0 1 0 1 0 1
1 1 1 0 0 0 1 1. (A + B) = (A • B)
Table 6.5.1.4 Truth table for A + B and (A • B)
2. (A • B) = (A + B)

A • B = (A + B) A B A•B A B (A + B) (A + B)
0 0 0 1 1 1 0
Key point
0 1 0 1 0 1 0
1 0 0 0 1 1 0 Cancelling NOTs:
1 1 1 0 0 0 1 Care should be exercised when
cancelling NOTs. The following
Table 6.5.1.5 Truth table for A • B and (A + B) examples illustrate when you may
Examples cancel and when you may not:

(A + B) = (A + B) cancellation
possible
Show using Boolean algebra and De Morgan’s laws that A • B = A + B
(A + B) = (A + B) cancellation
A•B = A+B Using De Morgan’s law X • Y = (X + Y) possible
(A • B) = (A • B) cancellation
A+B=A+B Using the Boolean identity X=X not possible

(A • B) = (A • B) cancellation
possible
Show using Boolean algebra and De Morgan’s laws that A + B = A • B

A+B = A•B Using De Morgan’s law X + Y = (X • Y)

A•B = A•B Using the Boolean identity X=X

Single licence - Abingdon School 252

6 Fundamentals of computer systems

Show using Boolean algebra and De Morgan’s laws that A • B + A • B = (A + B) • (A + B)

A • B + A • B = (A + B) + (A + B) Using De Morgan’s law X • Y = (X + Y)

= (A + B) • (A + B) Using De Morgan’s law X + Y = (X • Y) and X = X

=A•A+A•B+B•A+B•B Multiplying out the bracketed terms

=0+A•B+B•A+0 Using the Boolean identities A • A = 0 and B • B = 0

=A•B+B•A Using the Boolean identity 0+X=X

= (A • B) • (B • A) Using De Morgan’s law X + Y = (X • Y) and X = X

= (A + B) • (B + A) Using De Morgan’s law X • Y = (X + Y) and X = X

Show using Boolean algebra and De Morgan’s laws that (A + B) + (A + B) = (A • B) • (A • B)

Let (A + B) + (A + B) = X + Y Where X = (A + B) and Y = (A + B)

X + Y = X•Y Using De Morgan’s law C + D = (C • D)

(X • Y) = (A + B) • (A + B) Substituting for X and Y

(A + B) • (A + B) = A • A + A • B + B • A + B • B Multiplying out the bracketed terms

=0+A•B+B•A+0 Using A • A = 0 and B • B = 0

=A•B+B•A Using 0 + X = X and X + 0 = X

= (A • B ) • (B • A) Using De Morgan’s law C • D = (C + D)

= (B • A ) • (A • B) By commutative law

= (A • B ) • (A • B) By commutative law

A A•B
B (A • B) • (A • B)
The original Boolean expression has
A been transformed into one that can be
implemented just with NAND gates as
A•B shown in Figure 6.5.1.3.

B
(A • B) • (A • B)
Figure 6.5.1.3 NAND gate implementation

253 Single licence - Abingdon School

 6.5.1 Using Boolean algebra

Questions
Using Boolean algebra and De Morgan’s laws show for questions 12 to 15 that

12 A • B + A • B = (A + B) + (A + B)

13 (A + B) + (A + B) = A • B + B • A

14 A•B+A=1
15 A•B•A•B=B
16 Simplify the following:

A + B + (A • B)
(a) (b) (c) A • (A + B) A•B•C+A•B

(d) (e)
A + (A • B) (A • B) + (A • B)

17 An electronic control circuit is used to switch off an industrial process when certain parameters, indicated
by two-state electronic signals W, X, Y and Z, reach critical values. The process must be stopped if either
W and X or W, Y and Z become critical at the same time. Write a Boolean expression for these parameters
that when evaluated will output 1 to switch off the process and 0 otherwise.

18 The Boolean expression for EXCLUSIVE-OR is A • B + A • B.

(a) Convert this expression into a form that could be implemented with NAND logic gates each with two
inputs.
Draw the NAND logic gate circuit for this expression.
(b) Convert this expression into a form that could be implemented with NOR gates each with two inputs.
Draw the NOR logic gate circuit for this expression.

19 For a process to proceed the following Boolean expression must be true W • (X + Y • Z).
(a) Convert this expression into a form that could be implemented with NAND logic gates each with two
inputs.
Draw the NAND logic gate circuit for this expression.
(b) Convert this expression into a form that could be implemented with NOR gates each with two
inputs.
Draw the NOR logic gate circuit for this expression.

20 Given that
(1) Alice never gossips
(2) Bob gossips if anyone else is present
(3) Dick gossips under all conditions even when alone
(4) Soria gossips if and only if Alice is present
Determine the conditions when there is no gossip in the room.

Single licence - Abingdon School 254

6 Fundamentals of computer systems

Questions
21 A security light outside a house is controlled by two switches, which
can be turned on or off from inside the house, and a light level sensor.
The switches are named A and B. The light level sensor is named C.
The security light is labelled L.
If the light level is low (i.e. it is night time) the output of the sensor is
on otherwise it is off.

Information • If both switches A and B are off then the light L is always off.
• If switch A is on the light L is always on.
Dual-in-line package
• If switch B is on and switch A is off then:
(DIP) Digital Integrated
Circuits(ICs or chips): ºº the light L turns on if the light level is low
Logic gates are available as ºº the light L turns off if the light level is not low.
integrated circuits. An integrated Write a Boolean expression to represent the logic of the security light
circuit (IC) containing four
system.
NAND logic gates each with two
inputs is shown below in both 22 A second sensor is added to the system in Q21. This sensor is a
schematic form and as an actual movement detector. This second sensor is named M. The output of M
IC.
is on if it senses movement otherwise it is off.
14 13 12 11 10 9 8
Vcc

74LS00 • If switch B is on and switch A is off then:

1 2 3 4 5 6
GND
7 ºº the light L turns on if the light level is low and movement
is detected
7408 DIL IC containing four
NAND gates. ºº the light L turns off after one minute if movement is not
detected.
Write a Boolean expression to represent the logic of the security
light system.

Universality of NAND gates

Information Any logic circuit can be implemented using only NAND gates.
NAND logic gate:
NAND (A, B) = (A • B)
NOT(A) = NAND (A, A) = A • A = A
AND (A, B) = NOT(NAND (A, B)) =(A • B)
OR (A, B) = NAND (NOT(A), NOT(B)) = (A • B)
NOR (A, B) = NOT (OR (A, B)) = NOT (NAND (NOT(A), NOT(B)))

Information = (A • B)

NAND logic gate wired as a

NOT gate:

255 Single licence - Abingdon School

 6.5.1 Using Boolean algebra

Universality of NOR gates Did you know

Any logic circuit can be implemented using only NOR gates.
Gate universality of NAND
and NOR:
NOR (A, B) = (A + B)
NAND and NOR gates possess
NOT(A) = NOR(A, A) = A + A = A the property of universality. This
means, that a circuit consisting
OR (A, B) = NOT(NOR (A, B)) =(A + B)
only of NAND gates or a circuit
AND (A, B) = NOR (NOT(A), NOT(B)) = (A + B) consisting only of NOR gates is
able to perform the operation of
NAND (A, B) = NOT (NOR (NOT(A), NOT(B))) = (A + B)
any other gate type. The ability
of a single gate type to be able
to replicate the operation of any
In this chapter you have covered: other gate type is one enjoyed
only by NAND and NOR.
■■ Boolean expressions, e.g. (A • B)
NAND gates are preferred to
■■ De Morgan’s laws in a form for designing logic gate circuits NOR because
A + B = (A • B) • NAND cheaper to
fabricate than NOR
A • B = (A + B) • NAND has a lower
■■ Boolean identities propagation delay than
NOR.
A+0=A A+1=1 A+A=1 A+A=A
A+0=A A+1=1 A+A=A
0•A=0 1•A=A A•A=0 A•A=A
0•A=0 1•A=A A•A=A

A=A

■■ Distribution laws
A • (B + C) = A • B + A • C A + (B • C) = (A + B) • (A + C)

■■ Using De Morgan’s laws and Boolean identities to manipulate and simplify

Boolean expressions

Single licence - Abingdon School 256

7 Fundamentals of computer organisation and architecture

7.1 Internal hardware components of a computer

Learning objectives:
■ 7.1.1 Internal hardware components of a computer
■ Have an understanding and Structure of a simple computer
knowledge of the basic internal The architecture of a simple (traditional) computer system consists of a set
components of a computer of independent components or subsystems which may be classified as either
system internal or external. The internal subsystems are:

■ Understand the role of the • processor or Central Processing Unit (CPU)

following components and • main memory (RAM)
how they relate to each other • I/O controllers - input only, output only, both input and output
• processor
• buses
• main memory
The external subsystems are on the periphery of the computer system and
• address bus are known, therefore, as peripherals or peripheral devices - for example, the
• data bus keyboard, visual display unit, printer, magnetic disk drive. The main processor
• control bus or CPU exchanges data with a peripheral device through a part of an I/O
• I/O controllers controller called an I/O port. Peripheral devices are not connected directly
■ Understand the need for, to the CPU because the former often operate with signal levels, protocols and
and means of, communication power requirements which are different from those used by a CPU. Therefore,
between components. In peripherals are not under the direct control of the CPU. Figure 7.1.1 illustrates
particular understand the the structure of a simple (traditional) computer.
concept of a bus and how
Control
address, data and control buses bus
Processor
are used (CPU) Data &

■
address
Be able to explain the buses
Main memory
difference between von or
immediate access store
Neumann and Harvard (RAM)
architectures and describe
where each is typically used
■ Understand the concept of Keyboard
Keyboard
input
VDU
output
Visual
Display
Unit
addressable memory controller controller
(VDU)

Information (Disk) I/O controller

Processor (CPU):
The name processor was commonly
Secondary store or backing
used for the name of the central or store
general-purpose processor. Nowadays, (e.g. magnetic disk)

CPU or Central Processing Unit refers

to the general processor to distinguish Figure 7.1.1 Block diagram of the simplified structure of a
it from other processors, e.g. Graphics traditional (von Neumann) computer
Processing Unit (GPU). Originally, CPU
meant processor + main memory.

Single licence - Abingdon School 257

7 Fundamentals of computer organisation and architecture

Key concept Questions

Peripheral:
A peripheral is a device that 1 The components or subsystems of a traditional von Neumann
is connected to the computer computer system are classified as either internal or external.
system but which is not under (a) Name the four internal components
the direct control of the
(b) Give three examples of an external component
processor. Instead the processor
interacts with the peripheral
indirectly via the peripheral’s
The bus subsystem
I/O controller which sits
electrically between the Buses can be parallel buses, which carry data words in parallel on multiple
peripheral and the system bus. wires, or serial buses, which carry data in bit-serial form (one bit after another)
in one or more communication pathways or channels (channel = pair of wires
Key concept or equivalent). A parallel bus is a set of parallel wires connecting two or more
System bus: independent components of a computer; for example in Figure 7.1.2 control,
A bus that connects together
data and address buses are shown connecting the processor (CPU), memory
processor, main memory and
I/O controllers is called a and I/O controllers. A key characteristic of the parallel bus in this example is
system bus. that it is a shared transmission medium, so that only one component can
It consists of three dedicated successfully transmit at any one time.
buses:
1. data bus A bus that connects together processor (CPU), main memory and I/O
2. address bus controllers has traditionally been called a system bus. Typically such a bus
3. control bus. consists of from 50 to 100 separate wires (conducting pathways). Each wire
(line) conveys a single bit at a time. The number of wires is referred to as the
Information width of the bus. Although there are different bus designs, on a traditional
Bus line:
system bus the lines can be classified into three functional groups: data,
The wires of a bus are often
referred to as lines or bus lines address and control lines. The subsets of lines are known as the data, address
because they resemble tram lines and control buses, respectively.
used by trams.

Key fact Control bus

Bus width:
The number of wires in a bus is
referred to as the width of the
bus.
Visual Magnetic
Keyboard Processor Main
display disk
controller (CPU) memory
Information controller controller

Bus masters:
In the traditional shared bus
system, one device takes charge of
the system bus at a time, e.g. the Address bus

processor or the main memory.

A device that is granted access
to the system bus so that it can Data bus

communicate with another device

is called the master during the
Figure 7.1.2 Internal components of a
communication and the device traditional shared bus computer system (von Neumann)
receiving the communication the
slave.

258 Single licence - Abingdon School

 7.1.1 Internal hardware components of a computer

Questions Information
2 Distinguish between a parallel bus and a serial bus. Serial bus:
Parallel buses suffer timing skew
3 What is a parallel bus used for in the traditional von Neumann which limits their operating speed.
and they only operate one-way at
computer system and what name is given to this bus?
a time (half-duplex).

4 The system bus is subdivided into three functional groups of wires in A serial bus does not suffer from
timing skew and therefore can
the traditional von Neumann computer system. Name these groups.
operate at a much higher rate
than is possible with parallel
5 What is meant by saying that this von Neumann system bus is a shared
buses. They can also operate in
transmission medium? both directions simultaneously
just by adding another serial
channel. PCI Express, Serial ATA
(SATA) and USB are all examples
Modern bus systems that support serial bus operation.
The traditional computer system of von Neumann’s design required that both Address, data and control lines
processor (CPU) and main memory operate together at the same speed. are replaced by address, data
and control phases of the serial
Modern general purpose processors operate at speeds much faster than main
communication. which takes place
memory can operate. Thus the main processor or CPU in such a system will
in packets of bits.
be held up waiting for a requested data word to be fetched from memory
The serial bus is now considered
unless the CPU is decoupled from main memory in a way that allows it to do
to be a point-to-point connection
other tasks. Figure 7.1.3 shows how the bus architecture is evolving to take with one or more channels (pair
account of this and differences in speed of operation of other devices that are of wires).
connected to the computer system. Figure 7.1.4 shows a printed circuit board This has become an important
(motherboard) for a relatively modern general purpose computer. In the latest mechanism for communicating
processors and mother boards, the memory controller has been moved into the between multiple cores and
slices of main memory. One such
CPU and two separate buses are used, a memory bus and a packet-based (up to
system resembles the switching
32 bits) serial bus for peripherals. circuits of a telephone exchange
which connects multiple pathways
simultaneously to enable many
Traditional (von Neumann) computer system data bus
independent calls to be made.
The data bus, typically consisting of 8,16, 32, 64 separate lines, provides
a bidirectional path for moving data and instructions between system
components. The width of the data bus is a key factor in determining overall Key concept
system performance. For example, if the data bus is 16 bits wide, and each
Data bus:
instruction is 32 bits long, then the processor must access the main memory
The data bus provides a
twice during each instruction cycle. bidirectional path for moving
data and instructions between
system components.
Questions
6 What is the purpose of the data bus and why is its width a factor in
determining overall system performance?

Single licence - Abingdon School 259

7 Fundamentals of computer organisation and architecture

Main
memory
(RAM)
Cache slots
Memory
Did you know? bus
Memory (64 bit)
The Spinnaker Human Brain controller
hub 100 Gbit/s
project:
CPU
In this project 100000 to
1000000 simple microprocessors
containing ARM processor System bus 16 Gbit/s
50 - 400 Gbit/s Serial ATA
cores and main memory are (64 bit) 0.48 -10 Gbit/s
Bus USB
being connected together in a bridge 0.1 -100 Gbit/s
Ethernet
hexagonal grid to allow networks
of many millions of neurons to be
20-256
simulated in real time. PCI BIOS
HDMI Graphics Gbit/s
The communication between the express (Flash
controller Card ROM) PCI 64-bit/100 MHz
controller
microprocessors is based on an
6.4 Gbit/s
efficient multicast infrastructure
inspired by neurobiology. It uses
PCI bus
a packet-switched network to (32 bit)
emulate the very high connectivity
of biological systems.

PCI slots

Figure 7.1.3 Block diagram of an I/O controller

CPU beneath
fan and heat sink
PCI slots Memory
controller
hub
Main memory
(RAM)
Bus Quartz crystal
bridge oscillator for
clock timing
BIOS signals
(Flash
ROM) Figure 7.1.4 Motherboard with memory hub controller and bus bridge

260 Single licence - Abingdon School

 7.1.1 Internal hardware components of a computer

Traditional (von Neumann) computer system control bus Key concept

The control bus is a bidirectional bus meaning that signals can be carried in
Control bus:
both directions. The data and address buses are shared by all components of The control bus is a
the system. Control lines must therefore be provided to ensure that access bidirectional bus i.e. it carries
to and use of the data and address buses by the different components of the signals between the processor
system does not lead to conflict. The purpose of the control bus is to transmit and other system components
and vice versa. The purpose of
command, timing and specific status information between system components.
the control bus is to transmit
Timing signals synchronise operations by indicating when information on the command, timing and specific
data, control and address buses is ready for consumption. Command signals status information between
specify operations to be performed. Specific status signals indicate the state of a system components.
data transfer request, or the status of a request by a system component to gain
control of the system bus.
Typical control lines include:
• Memory Write: causes data on the data bus to be written into the
addressed location.
• Memory Read: causes data from the addressed location to be placed on
the data bus.
• I/O Write: causes data on the data bus to be output to the addressed
I/O port.
• I/O Read: causes data from the addressed I/O port to be placed on the
data bus.
• Transfer ACK: indicates that data have been accepted from or placed
on the data bus.
• Bus Request: indicates that a component needs to gain control of the
system bus.
• Bus Grant: indicates that a requesting component has been granted
control of the system bus
• Interrupt request: indicates that an interrupt is pending.
• Interrupt ACK: acknowledges that the pending interrupt has been
recognised.
• Clock: used to synchronise operations.
• Reset: initialises all components.

Questions
7 Using examples, explain the purpose of the control bus in the
traditional von Neumann computer system.

Single licence - Abingdon School 261

7 Fundamentals of computer organisation and architecture

Key concept Traditional (von Neumann) computer system address bus

When the processor wishes to read a word (say 8, 16, 32 or 64 bits) of data or
Address bus: an instruction from memory, it first puts the address of the desired word on the
The address bus is used to select
address bus. The width of the address bus determines the maximum possible
a specific memory location
containing a word of data or memory capacity of the system - its address space. For example, if the address
an instruction. It does this bus consisted of only 8 lines, then the maximum address it could transmit
by carrying the address of the would be (in binary) 11111111 or 255 - giving a maximum memory capacity
desired location on its bus, e.g.
of 256 (including address 0). A more realistic minimum bus width would be
23410 or 111010102.
20 lines, giving a memory capacity of 220, i.e. an address space of 1048576
addressable memory locations (words). The address bus is also used to address
I/O ports during input/output operations.
Key concept
Address space:
The range of memory addresses Maximum no of
that the machine can address. Maximum no of addressable locations
No of address lines, m
addressable locations expressed as a
power of two, 2m
1 2 21
2 4 22
3 8 23
4 16 24
8 256 28
16 65536 216
20 1048576 220
24 16777216 224

Table 7.1.1: Relationship between number of address lines m and maximum

number of addressable memory locations

Questions
8 The address space of a particular computer system is 220. What does
this mean?

9 The address bus of a traditional computer system consists of 16 lines.

What is the total number of memory locations that theoretically can be
addressed by this address bus?

262 Single licence - Abingdon School

 7.1.1 Internal hardware components of a computer

Main memory
Main memory consists of a contiguous block of read/write,
randomly accessible storage locations constructed from
semiconductor technology - Figure 7.1.5. It is a store for
Figure 7.1.5 Main memory RAM chips
addressable words, one word per location, with each word
composed of the same number of binary digits - Main Memory
Figure 7.1.6 (Random Access Memory)
(RAM)
Each location is
• capable of “remembering” what was written
to it
• able to change its contents to another bit
pattern when a write request is received if the Currently
addressed 9 Contents of
memory is read/write (selected) 7
6 1 0 11 0 0 1 1 1 0 1 1 1 1 0 1 addressed
memory 5 memory location
• assigned a unique integer address by which location 4
3
it may be located 2
1
0
• capable of providing a copy of its contents
Figure 7.1.6 Main memory
when a read request is received.
The semiconductor technology used in read/write main memory means Information
• that the main memory is volatile, i.e. the contents of each storage von Neumann architecture:
location is lost when the power is removed. Memory contains addressable
words each composed of the
• the contents of main memory are not restored when powered up again same number of binary digits;
but instead each location consists of a random pattern of bits. addresses consist of integers
running consecutively through the
Storage locations may be visited (selected) one after another in any order
memory, 0, 1, 2, .....
noncontiguously, starting from anywhere in the memory. The time taken to
access any particular storage location is the same.
These two facts have led to main memory being labelled Random Access
Memory or RAM.

Questions
10 How is main memory organised?

11 What is meant by volatile memory?

12 What is meant by random access in the context of main memory?

I/O Controllers Key concept

Peripheral devices cannot be connected directly to the processor. Each I/O controller:
peripheral operates in a different way and it would not be sensible to design An I/O controller is a board
processors to directly control every possible peripheral. Otherwise, the of electronics that enables
invention of a new type of peripheral would require the processor to be the processor to control and
communicate with a peripheral
redesigned. Instead, the processor controls and communicates with a peripheral
device through an I/O port.
device through an I/O or device controller.

Single licence - Abingdon School 263

7 Fundamentals of computer organisation and architecture

Information The controller is a board of electronics consisting of three parts:

• An interface that allows connection of the controller to the system or
Hardware interface:
An interface is a standardised
I/O bus.
form of connection defining • A set of data, command, address and status registers (for block
such things as signals, number
transfer devices the data register will be replaced by a block of storage
of connecting pins/sockets and
voltage levels that appear on
locations).
these pins/sockets. • An interface that enables connection of the controller to the cable
connecting the device to the computer.
Key concept An I/O controller presents a standard interface to the system bus so that the
peripheral device appears to the processor as just a set of registers mapped onto
I/O port:
An I/O port is a set of registers
the address space of the machine and which can be referenced by machine
(and block of storage cells in a instructions - see Figure 7.1.7. This set of registers (and for block transfer
block transfer device) located devices, block of storage locations) is known as an I/O port.
in an I/O controller connected
I/O controllers are available which can operate both input and output transfers
to the system bus. An I/O port
provides a standard interface of bits, e.g. magnetic disk controller. Other controllers operate in one direction
through which a processor only, either as an input controller, e.g., keyboard controller or as an output
connected to the system bus can controller, e.g., VDU controller.
control and communicate with
the peripheral device attached External device
to the I/O controller. e.g. magnetic disk drive

Information
Memory-mapped peripherals: I/O controller Block of Command
storage locations register
The registers and block of storage (512)
locations of an I/O controller Address
register &DF03 &DF04
can use the same address space
as main memory. In this case,
&DD00
memory addresses that would
address
have been used for main memory in Status Interrupt
are allocated to I/O controllers. address &DD03 register generator
space of
The peripherals attached to these machine &DF05
I/O controllers are then said to
be memory-mapped.

Information
Address Data Control
bus bus bus
Direct Memory Access
System
(DMA): bus
Block transfer devices such as Figure 7.1.7 Block diagram of an I/O controller
magnetic disks can take charge of
the system bus to transfer a disk Questions
block directly to main memory 13 What is an I/O controller?
bypassing the processor entirely.
This is called Direct Memory 14 Why is a processor not connected directly to external devices?
Access.
15 What is an I/O port?

264 Single licence - Abingdon School

 7.1.1 Internal hardware components of a computer

Processor (CPU)
The processor (CPU) executes machine instructions that have been fetched
along the data bus from main memory locations. The processor selects a
memory location by placing the address of the location on the address bus. The
data processed by machine instructions is also fetched along the data bus from
main memory and the results of processing returned the same way. The control
bus is used by the processor to assert actions, e.g. read from memory, write
to memory, and to allow devices such as the keyboard controller to grab the
attention of the processor via the interrupt mechanism when a key is pressed.

Von Neumann and Harvard architectures

General-purpose processors are designed to work well in a variety
of contexts. In the von Neumann architecture, programs and data share the
same memory which the processor
Instruction communicates with over a shared bus
Processor & called the system bus - Figure 7.1.8. Did you know?
(CPU) Data
As John von Neumann was working
memory Sales of processors for embedded
at Princeton university at the time
applications, e.g., mobile phones,
Figure 7.1.8 Von Neumann this architecture is also known as the far exceed the sale of processors
architecture Princeton architecture. for general purpose computing.

The Harvard architecture is often used

in the design of processors where the context in which the processor is required
to work is restricted or dedicated to a particular task, e.g. sampling and
recording data from sensors. Such processors are used in embedded systems
e.g., traction control
systems in automobiles. Data Processor Instruction
In the Harvard memory (CPU) memory
architecture, program
and data are allocated
Figure 7.1.9 Harvard architecture
separate memories
as shown in Figure 7.1.9. The processor is connected to both memories by
separate buses so that each memory can be accessed simultaneously. The benefit
of having a separate data memory is that data access is possible at a consistent
bandwidth (same bit rate) which is particularly important for sampled-data
systems.
The Raspberry Pi computer is based on the Harvard architecture.
The instruction sets of Harvard architecture processors can be different
from that of general purpose von Neumann processors because Harvard
processors need to support the context in which they will be used, e.g. graphics
processing.
In graphics processing, algorithms perform identical operations on each section
of the screen. For this type of processing, a processor that can multiply and

Single licence - Abingdon School 265

7 Fundamentals of computer organisation and architecture

accumulate a block of data in a single instruction is very useful. Processors that

perform this kind of specialised processing are called Digital Signal Processors
(DSP). They are usually based on the Harvard architecture because they need
to perform single instruction multiple data processing and accumulate results
in an accumulator register.

Questions
16 What is the major difference between the von Neumann computer
architecture and the Harvard architecture?

17 Where is each typically used?

In this chapter you have covered:

■■ The basic internal components of a traditional computer system
• processor
• main memory
• address bus
• data bus
• control bus
• I/O controllers
■■ The concept of a bus, parallel and serial and how address, data and control
buses are used
■■ The difference between von Neumann and Harvard architectures and
where each is typically used
■■ The concept of addressable memory

266 Single licence - Abingdon School

7 Fundamentals of computer organisation and architecture

7.2 The stored program concept

Learning objectives:
■■Be able to describe the stored
program concept
■■ 7.2.1 The meaning of the stored program concept
The stored program concept was
proposed by John von Neumann and
Alan Turing in separate publications
Information
in 1945. They proposed that both
EDSAC and SSEM: the program and the data on
The University of Manchester’s
which it performed processing and
Small-Scale Experimental Machine
(SSEM) is generally recognized calculations should be stored in
as the world’s first electronic memory together.
computer that ran a stored Specifically,
program—an event that occurred Figure 7.2.1.1 Plaque located at
• The program to be executed
on 21st June 1948. University of Manchester
However, the EDSAC (designed is resident in an electronic
commemorating the creators
and built at Cambridge university) memory directly accessible to
of the first stored program computer
is considered the first complete the processor
and fully operational electronic
digital stored program computer.
• Instructions are fetched one at a time (serially) from this memory and
It ran its first program on 6th May executed in a processor
1949. • Data is resident in an electronic memory directly accessible to the
processor which can change it if instructed to by the executing program.
The stored program model in which
program and data reside together in
main memory when the program
is being executed became known
as a von Neumann computer.
The world’s first stored program
electronic computer was designed
and built at the university of
Manchester.
The stored program concept enables
computers to perform any type of
computation, without requiring the
user to physically alter or reconfigure
the hardware.
Figure 7.2.1.2 ENIAC computer being reprogrammed by changing
In contrast, to program the
the wiring (U.S. Army photo, http://ftp.arl.army.mil/~mike/comphist/)
forerunners of the von Neumann
computer and to change the data, the programmer had to manually plug in
cables and set switches. This was quite tedious and time consuming. Figure
7.2.1.2 shows two programmers changing the program and data by literally
Single licence - Abingdon School 267
7 Fundamentals of computer organisation and architecture

rewiring the ENIAC computer, a non-stored program computer. This simple

Key concept
but fundamental idea of the stored program computer has been incorporated
Stored program concept: into all modern digital computers.
Both the program and the data
on which it performs processing Program code and data are the same
and calculations are stored in The stored program concept, as embodied in the von Neumann computer, of
memory together. having the program and data share the same memory means that the computer
Specifically,
can modify its data or the program itself while it is executing.
1. The program to be executed is
resident in a memory directly Figure 7.2.1.3 shows the basic architecture of a von Neumann computer.
accessible to the processor
Processor Main
2. Instructions are fetched one memory
at a time (serially) from this
Next instruction Program
memory and executed by the pointer
processor
Control
3. Data is resident in a memory Unit
directly accessible to the
Data
processor which can change Instruction
register
it, if instructed to by the
executing program.
Arithmetic
and
Logic Unit
Information Accumulator

Harvard architecture:
Another stored program
Figure 7.2.1.3 Stored program von Neumann computer basic architecture
computer is the Harvard
architecture computer, a rival
architecture to the von Neumann Program code and data can be treated as if they are the same when they occupy
architecture that came from the same memory. The memory is interpreted as an instruction when the next
Princeton university. instruction pointer references it, and as data when an instruction references it.
In the Harvard architecture, the
program resides in one memory Treating program code as data is useful when, for example, a program needs
and data in a separate memory to be downloaded from a remote location because it can be treated as data and
which is also directly accessible downloaded in the same way that an email can be.
by the processor. Nevertheless a
Harvard architecture computer is Programs such as compilers also treat other programs as data when they read
also a stored program computer. them. However, treating programs as data also has a downside. Computer
viruses are programs too which get treated as data when being downloaded but
Key principle as programs when the host computer is tricked into executing them.

Program code and data are the

Questions
same:
Program code and data can be 1 What is meant by the stored program concept?
treated as if they are the same
when they occupy the same 2 Explain what is meant by program code and data can be treated as the
memory. same thing in a von Neumann stored program computer.
The memory is interpreted as
3 State one advantage and one disadvantage of being able to do this.
an instruction when the next
instruction pointer references it,
and as data when an instruction
references it.

268 Single licence - Abingdon School

 7.2.1 The meaning of the stored program concept

Task
1 Explore the university of Manchester’s site on the world’s first stored program
computer at
http://curation.cs.manchester.ac.uk/digital60/www.digital60.org/birth/index.html

In this chapter you have covered:

■■ the stored program concept
■■ why program code and data can be treated as the same thing in a von
Neumann stored program computer.

Single licence - Abingdon School 269

7 Fundamentals of computer organisation and architecture

7.3 Structure and role of the processor and its components

Learning objectives:
■ 7.3.1 The processor and its components
■ Explain the role and operation Processor
of a processor and its major A simplified block diagram of a traditional processor is shown in Figure
components: 7.3.1.1.
A typical processor or Central Processing Unit (CPU) consists of the following
• arithmetic and logic unit
components:
• control unit
• Control Unit, which fetches instructions from memory, decodes and
• clock
executes them one at a time
• general purpose registers
• Arithmetic and Logic Unit (ALU) which performs arithmetic and
• dedicated registers, logical operations on data supplied in registers, storing the result in a
including: register. It can perform, for example, addition and subtraction, fixed
• program counter and floating point arithmetic, Boolean logic operations such as AND,
• current instruction OR, XOR and a range of shift operations.
register • Registers: general purpose, e.g. RegisterA, and special purpose or
• memory address register dedicated registers, e.g. Current Instruction Register (CIR), Program
• memory buffer register Counter (PC), Memory Buffer Register (MBR), Memory Address
Register (MAR), Status Register.
• status register
• System clock, which generates a continuous sequence of clock pulses to
step the control unit through its operation.

Register0

Register1

Memory Buffer
RegisterA RegisterB Registern Register
Operand 1 Data
Operand 2
32 32 bus
32 32
Control Unit

Current Instruction
Register
32 Instruction
Operation Instruction

ALU Operand
Program Counter address Address
RegisterC 32 Address of 32 32 bus
Next Instruction
Result
6
32 Memory Address
System Register Control
Status Register Clock bus
Central Processing
Unit
Tri-state gate: 1 or 0 or non-conducting
Figure 7.3.1.1 Simplified internal structure of a
Control
processor/central processing unit
Single licence - Abingdon School 270
7 Fundamentals of computer organisation and architecture

Key fact The processor or central processing unit is connected to main memory by
the system bus.
Processor (CPU):
The processor or central processing Questions
unit (CPU) consists of the following
1 General purpose registers are one major component of a
components:
traditional processor. Name and describe the four other major
• Control unit
• Arithmetic and Logic Unit components.
(ALU)
• General purpose registers
Processor operation with main memory
• Dedicated or special purpose A memory is a set of words, each with an address and a content:
registers
• The addresses are values of a fixed size, the address length
• system clock
• The contents are values of another fixed size, the word length
Key principle
• A load operation is used to obtain the content of a memory
Memory:
A memory is a set of words, each with word
an address and a content: • A store operation changes the content of a memory word.
• The addresses are values of
a fixed size, called the In a von Neuman computer both program and data reside in the same
address length memory. This memory is called main memory.
• The contents are values of
A processor interacts with this memory in 3 ways:
another fixed size, called
the word length • by fetching instructions
• A load operation is used
to obtain the content of a
• by loading a memory word into a processor register
memory word • by changing the content of a memory word by a store operation.
• A store operation changes
the content of a memory The size of the registers in the processor defines the size of the processor,
word. e.g. a 32-bit processor has registers that are 32 bits long. The length of a
register is known as the word length of the processor. This word length is
Key principle also usually the size of the memory word transferred in a load operation.
Processor and memory:
A processor interacts with memory in In Figure 7.3.1.1 the registers have a word length of 32 bits. Each is
3 ways: connected to a bus inside the processor which is also 32 bits wide shown
• by fetching instructions as /32 in the figure.
• by loading a memory word
into a processor register
Questions
• by changing the content of
a memory word by a store 2 State three ways that a processor interacts with memory.
operation.

Key fact
Control Unit
Processor word length:
The size of the registers in the processor
The control unit of the processor shown in Figure 7.3.1.1 controls
defines the size of the processor, e.g. a 32- fetching, loading and storing operations.
bit processor has registers that are 32 bits
It fetches an instruction into the Current Instruction Register via the
long. The length of a register is known as
the word length of the processor.
Memory Buffer Register and the data bus by
This word length is also usually the size • reading the contents of the Program Counter to obtain the
of the memory word transferred in a load
memory address of the memory word containing the instruction
operation.

271 Single licence - Abingdon School

 7.3.1 The processor and its components

• placing this memory address in the Memory Address Register Key principle
connected to the address bus so that the addressed memory word can
be selected and transferred across the data bus into the Memory Buffer Control unit:
Controls fetching, loading and
Register
storing operations.
• transferring the instruction fetched from memory from the Memory
Buffer Register into the Current Instruction Register
The control unit also Key principle
• decodes the instruction to determine if it is a load, store, arithmetic Control unit:
operation, or logic operation • Fetches an instruction
from memory and places
• executes the instruction by it in the Current
ŠŠ using the instruction’s operand fields as addresses to use in load or Instruction Register.
• Decodes the instruction
store operations, if required, or
to determine if it is one
ŠŠ loading a memory word into a register, or of load, store, arithmetic
operation, or logic
ŠŠ changing a word of memory in a store operation, or operation
ŠŠ controlling an arithmetic operation, e.g., ADD, or a logical • Executes the instruction.
operation, e.g., AND, in the Arithmetic and Logic Unit (ALU) using
as operands the instruction’s operand fields.

Questions
3 State the purpose of a processor’s control unit.

4 Describe in detail the operation of a processor’s control unit when executing a stored program, instruction
by instruction. You should state the name and describe the role of each register in this process.

System clock
The system clock or clock is a unit inside the processor that provides regular
Key fact
clock pulses that the control unit uses to sequence its operations. System clock:
The system clock is a unit inside
The clock signal is a 1-bit signal that oscillates between a “1” and a “0” with a
the processor that provides
certain frequency as shown in Figure 7.3.1.2. The change from “0” to “1” is regular clock pulses that the
called the positive edge, and the change from “1” to “0” the negative edge. control unit uses to sequence its
operations.
The time taken to go from one positive edge to the next is known as the clock
period, and represents one clock cycle. The number of clock cycles that fit one
second is called the clock frequency or clock speed. one clock cycle
1
1
Clock period =
Clock frequency 0
Figure 7.3.1.2 Clock signal

Table 7.3.1.1 shows some examples of clock speed/frequency for both current
processors/CPUs and a very popular processor from the 1980s.

Single licence - Abingdon School 272

7 Fundamentals of computer organisation and architecture

Information Clock frequency Clock period/cycle CPU

4GHz 0.25 nanoseconds AMD 6300
Instructions per second:
ARM Cortex-A7 (Raspberry
The number of instructions fetched 900MHz 1.1 nanoseconds
and executed per second is given by Pi 2)

Instructions per second Motorola 6502 (BBC Model

1 MHz 1 microsecond
Clock frequency B computer)
=
no of cycles per instruction Table 7.3.1.1 Clock speeds for some processors/CPUs
The number of clock cycles an instruction takes to be fetched and executed
varies from
Fetch
processor to
Information processor. A very
simple design Execute
Quartz crystal oscillator:
might use one
The processor is usually
connected to an external quartz
clock cycle to fetch
Clock
crystal oscillator which provides an instruction
a very stable and regular timing from memory Figure 7.3.1.3 Timing of Fetch/Execute, one clock
signal. and another clock cycle for Fetch, one for Execute
The clock unit in the processor
cycle to execute the
uses this timing signal to derive all
the timing signals that are needed instruction as shown
by the control unit to synchronise in Figure 7.3.1.3.
operations within the processor.
A fetch phase occurs when the Fetch signal is 1 and the Execute signal is 0. An
Execute phase occurs when the Execute signal is 1 and the Fetch signal is 0.
Both Fetch and Execute signals are derived from the (master) clock signal and
are therefore synchronised by this signal.
The number of instructions fetched and executed per second is given by
Clock frequency
Instructions per second =
no of cycles per instruction
If the clock frequency for the 2-cycle processor design is 1 GHz then the
number of instructions fetched and executed per second is
1GHz
Instructions per second =
2
i.e. 500 million instructions per second.

Questions
5 What is the purpose of a processor’s system clock?
6 What is meant clock speed?

7 With the aid of a diagram, explain how the control unit could use the
system clock when an instruction in memory is executed in two clock
cycles.
8 The clock frequency at which a particular processor is operated is
2GHz. The number of clock cycles per instruction is 2.
How many instructions can be executed per second in this processor?

273 Single licence - Abingdon School

 7.3.1 The processor and its components

Registers Key concept

Registers are memory locations internal to the processor that support fast access
as well as rapid manipulation of their contents because they are made from the Registers:
Registers are memory locations
fastest memory technology.
internal to the processor that
Moving data between the ALU and registers and between registers is facilitated support fast access to and
by dedicated pathways within the processor that the control unit can open or manipulation of their contents
because they are made from the
close relatively quickly.
fastest memory technology.
However, the memory technology used and the dedicated pathways make
implementing registers expensive and so the processor will have only a limited
number, typically 32 but the number can range from 4 to 256. This is in
contrast to memory locations in main memory which are made from much
slower but cheaper technology and which are accessed over a shared pathway, Key concept
the system bus. The cheap technology and shared bus make it possible to have General purpose registers:
a very large number of main memory locations, e.g. 1000 0000 0000 locations, General purpose registers are
but access is much slower than the speed the processor operates at. registers that can be used by the
programmer to store data, as
Questions needed.

9 What are processor registers?

10 Processor registers and main memory are located in separate areas of Key concept
a traditional computer system. State four other differences between Special purpose or dedicated
processor registers and main memory. registers:
These are registers that are used
General purpose registers by the control unit in a specific
or dedicated way, e.g. Program
General purpose registers are registers that can be used by the programmer to Counter.
store data, as needed. Each register will be capable of storing a memory word
of a fixed size and will have a unique address known to the control unit. For
example, if there are 16 general purpose registers then their addresses will be 0, Key concept
1, 2, 3, ..., 14, 15 (0, 1, 2, ..., E, F in hexadecimal).
Program Counter (PC):
Dedicated or special-purpose registers Points to the next instruction to
Some registers are designed to be used by the control unit in a specific way, be fetched and executed.
e.g. the Program Counter (PC) stores a memory address which is the
address of the next instruction to be fetched and executed. The control unit
sets this address to ensure it points to the next instruction. The control unit Key concept
increments this address during a Fetch. It also changes this address if the
Memory Buffer Register
current instruction is a branch instruction or a subroutine call instruction or an (MBR):
interrupt service routing call. Connected to the data bus and
contains a word to be stored in
The following special-purpose registers are dedicated as follows:
memory, or a word copied from
• Memory Buffer Register (MBR): Connected to the data bus and memory.
contains a word to be stored in memory, or a word copied from This is also called the Memory
Data Register (MDR).
memory. This is also called the Memory Data Register (MDR)

Single licence - Abingdon School 274

7 Fundamentals of computer organisation and architecture

Key concept • Memory Address Register (MAR): Connected to the address bus so
that the memory address it contains can appear on this bus and be used
Memory Address Register at the memory end of this bus to select a particular memory word
(MAR):
Connected to the address bus • Instruction Register (IR) or Current Instruction Register (CIR):
so that the memory address it When an instruction is fetched from memory it is stored in this register
contains can appear on this bus while the control unit decodes and executes it.
and be used at the memory end
of this bus to select a particular • Status register: This register stores single bit condition codes each
memory word. of which indicates the outcome of arithmetic and logical operations
carried out in the ALU - for example, an arithmetic operation may
produce any of the following a positive, negative, zero result, a carry,
Key concept overflow and the corresponding condition codes are set (made 1).
Current Instruction Register Sometimes the name flag is used for a single bit condition code, i.e. a
(CIR): flag is set. These condition codes may subsequently be tested by the
When an instruction is fetched control unit when it is executing a conditional branch operation. The
from memory it is stored in this
possible condition codes are
register while the control unit
decodes and executes it. ŠŠ Sign: Contains the sign bit of the result of the last arithmetic
operation
ŠŠ Zero: Set when the result is zero
Key concept
ŠŠ Carry: Set if an operation resulted in a carry (addition) into or
Status Register: borrow (subtraction) out of a high-order bit. Used for multi-word
This register stores single
arithmetic operations
bit condition codes each of
which indicates the outcome ŠŠ Equal: Set if a logical compare result is equality. (Alternatively the
of arithmetic and logical zero flag may be used)
operations carried out in the
ALU, e.g. Zero bit or flag is ŠŠ Overflow: Used to indicate arithmetic overflow.
set to 1 if the result of the last The status register will also have single bits to control the operation of
arithmetic operation is zero
the control unit:
otherwise it is set to 0.
The status register also has single ŠŠ Interrupt Enable/Disable: Used to enable or disable interrupts
bits to control the operation of
ŠŠ Supervisor: Indicates whether the processor is executing in supervisor
the control unit, e.g. Interrupt
Enable/Disable bit. or user mode. Certain privileged instructions can be executed only
in supervisor mode, e.g. disabling interrupts, and certain areas of
memory can be accessed only in supervisor mode.

Questions
11 What is meant by general purpose register?

12 What is meant by a dedicated or special purpose register?

13 State the role of each of the following:

(a) Program Counter (b) Memory Buffer Register (c) Memory Address Register
(d) Current Instruction Register (e) Status Register.

275 Single licence - Abingdon School

 7.3.1 The processor and its components

Questions
14 Name four condition code flags and two control flags present in a
typical status register.

Key concept
Arithmetic and Logic Unit (ALU)
Arithmetic and Logic Unit
Figure 7.3.1.4 shows the ALU performing the arithmetic operation 3 + (- 5 )
(ALU):
producing the result -2. The Arithmetic and Logic Unit
RegisterA RegisterB
The Negative flag (ALU) performs arithmetic and
3 -5 logical operations on the data.
condition code (N)
in the status register 32 32
is set to 1 because the
result is negative. The
Zero flag (Z), Carry ADD Control
Operation
flag (C), Overflow +
Unit Key concept
ALU
flag (O) are set to 0. Negative flag:
The Interrupt Enable RegisterC 32 The Negative flag condition
code is the Sign condition code.
flag is 1 therefore -2 6
enabling interrupts.
Z NC OI S
The Supervisor mode 32
0 1 0 010
flag (S) is 0, therefore
Status Register
the processor is in User
Figure 7.3.1.4 ALU performing an ADD operation
mode. The Supervisor
flag is set to 1 when the operating system needs to use the processor, otherwise
it is 0 when a user is executing a program in the processor.
The Arithmetic and Logic Unit (ALU) performs arithmetic and logical
operations on data supplied in registers, storing the result in a register as shown
in Figure 7.3.1.4. It can perform, for example, addition and subtraction, fixed
and floating point arithmetic, Boolean logic operations such as AND, OR,
Information
XOR and a range of shift operations.
ASM Tutor:

Questions Available from Educational

Computing Services Ltd -
15 What is the purpose of the Arithmetic and Logic Unit? www.educational-computing.co.uk
Visual X - Toy:
16 Describe, with the aid of a diagram, the role of registers in the Available for Princeton university.
execution of arithmetic operation 4 + (-4) in the Arithmetic and Logic http://introcs.cs.princeton.edu/
Unit. xtoy/

Tasks
1 Explore the operation of a processor using a simulator such as
ASMTutor or Visual X-Toy.

Single licence - Abingdon School 276

7 Fundamentals of computer organisation and architecture

In this chapter you have covered:

■■ The role and operation of a processor and its major components:
• arithmetic and logic unit
• control unit
• clock
• general purpose registers
• dedicated registers,
including:
ŠŠ program counter
ŠŠ current instruction register
ŠŠ memory address register
ŠŠ memory buffer register
ŠŠ status register

277 Single licence - Abingdon School

7 Fundamentals of computer organisation and architecture

7.3 Structure and role of the processor and its components

■■ 7.3.2 The Fetch-Execute cycle and the role of the registers within it
Learning objectives: Fetch-Execute cycle
■■Explain how the Fetch-Execute A machine code program is made up of machine code instructions which are
cycle is used to execute fetched from main memory, one at a time, and executed in the processor/CPU.
machine code programs In Chapter 7.3.1, we learned that a processor executes each machine code
including the stages in the instruction by breaking its execution into a three-step sequence with the
cycle (fetch, decode, execute) execution synchronised by the system clock and controlled by the control unit.
and details of the registers This sequence of three steps is called the Fetch-Execute cycle or instruction
used. cycle.
The first step is a fetch operation, the second a decode operation and the third
Key concept step is execution.
Machine code program: These steps may be further broken down as follows:
A program consisting of
machine code instructions.
(Fetch phase)
1. The address of the next instruction to be executed (held in the PC) is
copied to the MAR which is connected to the address bus.
2. The instruction held at that address is fetched from memory along the
Key principle
data bus and placed in the MBR.
Fetch-Execute cycle:
A processor executes each
3. Simultaneously with step 2, the contents of the PC are incremented by
machine code instruction by 1 to point to the next instruction to be fetched.
breaking its execution into a 4. The contents of the MBR are copied to the CIR. This frees up the
three-step sequence:
MBR for the execute phase.
1. Fetch
2. Decode (Decode phase)
3. Execute
5. The instruction held in the CIR is decoded.
(Execute phase)
6. The instruction is executed. The sequence of micro-operations in the
Key fact
execute phase depends on the particular instruction being executed.
Registers always involved in
In register transfer notation, the Fetch-Execute cycle is described as follows:
the Fetch-Execute cycle:
• Program Counter (PC)
MAR [PC]
• Memory Address Register(MAR)
• Memory Buffer Register (MBR) MBR [Memory]addessed ; PC [PC] + 1
• Current Instruction Register
CIR [MBR]
(CIR)
[CIR] opcode part decoded and executed
where [ ] means contents of and means assign.

Single licence - Abingdon School 278

7 Fundamentals of computer organisation and architecture

This cycle repeats until the execution of the machine code program terminates.
Machine interrupts to the processor, if enabled, are ignored until the current
Information
Fetch-Execute cycle is completed.
The PC is shown as being
Chapter 7.3.3 covers the meaning of the term opcode. Essentially it is the part
incremented by 1 in step 3 on the
previous page. This assumes that
of an instruction which specifies the type of operation to be carried out, e.g.,
every instruction is one memory ADD, SUBTRACT, AND, etc.
word in length and therefore
The operation specified by the opcode is applied to the operands part of the
occupies one memory address. If
this is not the case then the PC
instruction, i.e. the part which isn’t the opcode.
is incremented by the amount The word field is used to mean a part of an instruction, e.g. an operand field.
necessary for the PC to point to
In the case of multiple operands, there is more than one operand field.
the next instruction to be fetched
and executed. The execution step of the Fetch-Execute cycle will know from the opcode if an
operand field is a datum for immediate use or an address of a memory word
containing a datum. If the operand is an address, the execution step will fetch
Information this datum.
Branch instructions and effect
on PC:
Questions
Branch instructions can change
1 Name the four registers that are always used in the Fetch-Execute cycle.
the contents of the PC when
executed, e.g. Branch on zero
<memory address> checks to 2 Using both register transfer notation and prose, explain how the Fetch-
see if the Zero flag in the status Execute cycle is used to execute machine code programs.
register is set. If it is, the contents
of the PC will be replaced by
the value of <memory address>.
This value is the address of an In this chapter you have covered:
instruction which the PC will now ■■ How the Fetch-Execute cycle executes machine code programs, instruction by
point to.
instruction, in a repeating cycle consisting of three steps:
• Fetch
Information
• Decode
Role of ALU: • Execute
The execution of an arithmetic
or logical instruction will involve
■■ The registers that are always involved are
the ALU. The status register is • Program Counter (PC)
updated during the execution
• Memory Address Register (MAR)
to reflect the outcome of the
arithmetic or logical operation. • Memory Buffer Register (MBR)
• Current Instruction Register (CIR)
■■ These registers are
Information used together with
MAR [PC]
Role of other registers: main memory in the
The execution of an instruction Fetch-Execute MBR [Memory]addressed ; PC [PC] + 1
may involve one or more general cycle as shown here: CIR [MBR]
purpose registers as well as the
status register. [CIR] opcode part decoded and executed
where [ ] means contents of and means assign.
279 Single licence - Abingdon School
7 Fundamentals of computer organisation and architecture

7.3 Structure and role of the processor and its components

Learning objectives:
■■ 7.3.3 The processor instruction set
■■Understand the term ‘processor Processor instruction set
instruction set’ and know that Format of instructions
an instruction set is processor The language of instruction for a digital computer is machine code; instructions
specific consisting of sequences of binary digits which a machine can recognise and
interpret. Machine code instructions are interpreted (executed) in a digital
■■Know that instructions consist computer’s processor (CPU) which must be designed so that it can understand
of an opcode and one or more and execute valid instructions. To understand why, consider instead a processor
operands (value, memory designed to understand certain three letter instruction words.
address or register)
Table 7.3.3.1 shows examples of possible valid and invalid instructions formed
from letters of the alphabet.
Information
Valid instruction Invalid instruction
Risc simulator:
ADD DAD
A Risc simulator designed by
SUB BUS
Peter Higginson is available from
MUL ULM
www.peterhigginson.co.uk/RISC/
DIV VID
Table 7.3.3.1 3-letter valid and invalid English instruction words
Note that the instructions are of the same fixed length and only some particular
combinations of letters are valid.
If these examples of valid combinations of letters correspond to the arithmetic
operations, ADD, SUBTRACT, MULTIPLY, and DIVIDE then we need to
include operand fields, R, B and C in each instruction so that an instruction
such as ADD has something to add and somewhere to store the result. The
instruction ADD R 3 4 adds together the values 3 and 4 and stores the result
in R. We call ADD the operation and, R, B and C the operands. Table 7.3.3.2
shows the new structure of two valid instructions
Valid instruction Action
Add C to B store
ADD R B C
result in R
Subtract C from B
SUB R B C
store result in R
Table 7.3.3.2 Operation and two operands
Questions
1 What will be stored in R if the instruction is
(a) ADD R 3 4 (b) SUB R 4 3 (c) MUL R 4 3

Single licence - Abingdon School 280

 7.3.3 The processor instruction set

Similar design constraints apply when designing the set of instructions (instruction set) that a processor (CPU) is
capable of recognising as valid and then executing, i.e. instructions belonging to its instruction set. The processor
or CPU will have access to registers, to memory, to an Arithmetic and Logic Unit and will also be able to make
transfers of data to I/O devices such as magnetic disks.
The basic machine operations that a processor executes can be categorised as follows:
• Data processing: Arithmetic, logic and shift instructions
• Data transfers: Register and memory instructions
• I/O transfers: I/O instructions
• Control: Test, branch and halt instructions
Data processing instructions
The processor might support the instructions shown in column 1, Table 7.3.3.3, which also shows their abbreviated
form in brackets, e.g. SUB.
Data transfer instructions Data processing Data transfers Control
The processor might support the ADD
COMPARE (CMP)
instructions shown in column 2, SUBTRACT (SUB) LOAD (LDR)
Table 7.3.3.3. Bitwise logical AND
Unconditional branch (B)
Control instructions Bitwise logical OR (ORR)
STORE (STR)
The processor might support the Bitwise logical EOR (EOR)
Conditional branch (B)
instructions shown in column 3, Bitwise logical NOT(MVN)
Logical Shift Left (LSL) MOVE (MOV)
Table 7.3.3.3. HALT (HALT)
Logical Shift Right (LSR)
I/O transfer instructions
Table 7.3.3.3 Some examples of basic machine operations
Instructions for this are not shown
but I/O transfers could be done using the given data transfer instructions if the registers and data locations in the
I/O controllers for each peripheral are mapped into the addressable memory space by allocating main memory
addresses to these in the same way as locations in RAM are mapped into the addressable memory space.
Operands
The next choice is how many operands? ARM processors are very popular and successful processors. ARM is the
market leader (2015) for processors in smartphones and tablets and an ARM processor is the main processor in the
Raspberry Pi and in the Parallela platform. ARM is a three-register architecture, meaning that a single machine code
instruction can reference up to three registers. For example, the ADD instruction can specify two registers from
which to read the values to be added and a third register, the destination register to store the calculated sum.
Figure 7.3.3.1 shows that the
Machine code instruction
structure of a machine code
Opcode Operand field (3 operands in this example) instruction is divided into an opcode
Basic Address
field and an operand field. Register
machine mode Rd is the destination register, register
operation
Rn contains the first input to the
operation and Operand2, the second.
1 0 1 1 0 0 Rd Rn Operand2
Operand2 could be an actual value,
6, or it could be another register, Rm,
Figure 7.3.3.1 Structure of a machine code instruction with containing the value to be used. The
an example Address mode bit is used to select

Single licence - Abingdon School 281

7 Fundamentals of computer organisation and architecture
which is the case, e.g. 0 for value, 1 for value contained in specified register. The combined code for the basic
machine operation and this address mode bit is known as the opcode or operation code field of the machine
code instruction. The ARM processor has 16 programmer-accessible registers, R0 through R15. R15 is the Program
Counter and R14 the link register (used to store the PC contents when a subroutine is called). Usually, another
register is used for the Stack Pointer. Therefore, Rd, Rn and Rm can be any of the remaining thirteen registers,
R0 through R12, e.g. R1, R2 and R3,
respectively. No of bits, n No of possible opcodes No of possible
With six bits allocated to the opcode field in opcode field 2n opcodes
4 24 16
of the machine code instruction shown in
5
Figure 7.3.3.1 and every other instruction, 5 2 32

there are a possible 26 different opcodes, 6 26 64

one for each 6-bit pattern. Table 7.3.3.4 Table 7.3.3.4 No of possible opcodes for a given no of opcode field bits
shows how the number of possible opcodes
varies with the number of bits reserved for the opcode field of a machine code instruction (remember: opcode field
size is fixed at the design stage of a processor).
Machine code
Opcode Description
instruction format
000000 LDR Rd, <memory ref> Load the value stored in the memory location specified by <memory ref> into register Rd.
000010 STR Rd, <memory ref> Store the value that is in register Rd into the memory location specified by <memory ref>.
Add the value specified in <operand2> to the value in register Rn and store the result in register Rd.
000100
ADD Rd, Rn, <operand2> 000100 for when operand2 is a value, 000101 for when operand2 is another register, Rm. The same
000101
interpretation applies to the other two opcode instructions.
000110 Subtract the value specified by <operand2> from the value in register Rn and store the result in register
SUB Rd, Rn, <operand2>
000111 Rd. For interpretation of why two opcodes, see ADD for why two opcodes.
001000 Copy the value specified by <operand2> into register Rd. For interpretation see ADD.
MOV Rd, <operand2>
001001
001010 Compare the value stored in register Rn with the value specified by <operand2>. See ADD
CMP Rn, <operand2>
001011
001100 B <label> Always branch to the instruction at position <label> in the program.
011101 Conditionally branch to the instruction at position <label> in the program if the last comparison met
011111 the criteria specified by the <condition>. Possible values for <condition> and their meaning are:
B <condition> <label>
011101 EQ: Equal to NE: Not equal to GT: Greater than LT: Less than.
011111
010010 Perform a bitwise logical AND operation between the value in register Rn and the value specified by
AND Rd, Rn, <operand2>
010011 <operand2> and store the result in register Rd. See ADD for why two opcodes.
010100 Perform a bitwise logical OR operation between the value in register Rn and the value specified by
ORR Rd, Rn, <operand2>
010101 <operand2> and store the result in register Rd. See ADD for why two opcodes.
010110 Perform a bitwise logical eXclusive OR (XOR) operation between the value in register Rn and the
EOR Rd, Rn, <operand2>
010111 value specified by <operand2> and store the result in register Rd. See ADD for why two opcodes.
011000 Perform a bitwise logical NOT operation on the value specified by <operand2> and store the result in
MVN Rd, <operand2>
011001 register Rd. See ADD for why there are two opcodes.
011010 Logically shift left the value stored in register Rn by the number of bits specified by <operand2> and
LSL Rd, Rn, <operand2>
011011 store the result in register Rd. See ADD for why there are two opcodes.

111100 Logically shift right the value stored in register Rn by the number of bits specified by <operand2> and
LSR Rd, Rn, <operand2>
111101 store the result in register Rd. See ADD for why there are two opcodes.
111110 HALT Stops the execution of the program.
Table 7.3.3.5 6-bit opcodes mapped to machine operations. Reproduced with permission of AQA. Currently only in specimen
papers and has therefore not been through the complete rigorous question paper process and is liable to change. Please consult
AQA’s website for the most recent version of the specification.

282 Single licence - Abingdon School

 7.3.3 The processor instruction set

Table 7.3.3.5 shows a possible mapping of some 6-bit opcodes (5 bits for Key concept
basic machine operation, 1 bit for address mode) to machine operations for an
imaginary processor. Processor instruction set:
The set of bit patterns for which
Instruction set machine operations have been
The simple operations referenced in Table 7.3.3.5 may be combined together in defined.
sequences to perform quite complicated tasks.
Information
The set of 28 bit patterns (strings of bits) shown in the opcode column of Table
7.3.3.5 represent these operations for a given processor and are known as the ARM:
ARM processor technology is
processor instruction set. Note that if Table 7.3.3.5 shows all the operations that
controlled by ARM Holdings.
a processor has been designed to understand and interpret then four bit patterns ARM stands for Advanced
do not correspond to any defined machine operations because for a 6-bit opcode RISC Machine and RISC stands
field, there are 32 possible bit patterns but only 28 are used to define opcodes. We for Reduced Instruction Set
Computer. RISC is a design
therefore use the following definition of instruction set:
philosophy in which the
The set of bit patterns for which machine operations have been defined. instruction set is deliberately
designed as a small set of very
An instruction set is processor specific
simple instructions.
An instruction set is specific to a particular processor for the following reasons: It does mean that a machine
• The machine operations that a processor is designed to perform varies code program will consist of a
large number of these simple
in number and type from processor to processor, e.g. from the ARM®
instructions but execution of
Cortex®-A7 CPU used in the Raspberry Pi 2 to the Intel® Core™ i7 simple instructions is very fast.
CPU used in laptops and PCs.
The simplicity of the instructions
• The number of bits allocated to the opcode field can also vary from simplifies the design of the control
processor to processor as well as how they are mapped to the operations unit. In particular, the instruction
that the processor supports. Therefore, machine code programs written for decoder circuits. This turns out
the ARM Cortex-A7 CPU will not run on an Intel Core i7 CPU. to have the advantage of lowering
processor power consumption.
• The number of possible operands, their type and the number of bits
Low power consumption is
reserved for each may also vary from processor to processor. particularly desirable in portable
• The design of a processor's control unit instruction decoder circuits devices such as smartphones
reflects the structure of the machine code instructions and will not and other embedded systems.

therefore be able to decode instructions designed for a different processor. This accounts for why ARM
processors dominate the mobile,
Structure of machine code instructions
tablet, System On a Chip (SOC)
We have learned that a machine code instruction is divided into an opcode part and embedded systems markets.
and an operand part as shown again in Figure 7.3.3.2. Considerably more ARM CPUs
than Intel CPUs are sold.
Opcode Operand(s)

Figure 7.3.3.2 Format of a machine code instruction

The machine code instruction with format MOV Rd, <operand2> has two possible opcodes which in binary are
001000 and 001001.
MOV Rd <operand2>
• opcode 001000 moves the value that is <operand2>
into the register Rd, e.g., 001000 0000 0100 1111.This 001000 0000 01001111
instruction is broken down as shown in Figure 7.3.3.3.
The destination register is R0 (binary code 0000) and Opcode Register R0 Value 7910
<operand2> is replaced by the value 0100 11112 (7910). Figure 7.3.3.3 Machine code instruction

Single licence - Abingdon School 283

7 Fundamentals of computer organisation and architecture

MOV Rd <operand2> • opcode 001001 moves the value stored in the

register specified by <operand2> into the
001001 0001 0000 register Rd, e.g., 001001 0001 0000.
This instruction is broken down as shown in Figure 7.3.3.4. The
Opcode Register R1 Register R0 destination register is R1 (binary code 0001) and <operand2>
Figure 7.3.3.4 Machine code instruction is replaced by register R0 (binary code 0000). If when the
processor executes this instruction the value 4510 is stored in register R0
Information
ARM processor architecture:
then at the end of the execution register R1 will also contain 4510.
ARM is a load-store architecture. The structure that has been used to illustrate machine code is a simplified one.
This means that before data can
It is not a good idea to design an instruction set that doesn't map to the word
be processed, a LOAD instruction
must be executed to load data
length of a memory word which is typically a multiple of eight. The word
from memory into registers. length of registers is usually a multiple of eight too, e.g. 32 bits. The unit of
Similarly, after data has been transfer between processor and main memory for load and store operations is
processed, a STORE instruction usually the same as the word length of registers, e.g., 32 bits if registers are 32
must be executed to store the
bits. Bits of a memory word that are not needed for a particular machine code
data in memory. Arithmetic and
logic instructions cannot work
instruction are ignored by the instruction decoder.
directly with values in main The machine code instruction with format LOAD Rd, <memory ref> has a
memory.
single opcode which is 000000 in binary. Rd is the destination register and
<memory ref> is replaced by the memory address of a memory word stored in
Information
Possible binary codes for
main memory. This instruction is broken down as shown in Figure 7.3.3.5.
registers: For example, 000000 0011 0100 1111 copies the memory word whose memory
Binary address is 0100 1111 into register R3 (binary code 0011).
Register code for
register LOAD Rd <memory ref>
R0 0000
R1 0001 000000 0011 0100 1111
R2 0010
R3 0011 Opcode Register R3 Memory
R4 0100 address 7910
R5 0101
Figure 7.3.3.5 Machine code instruction
R6 0110
R7 0111
R8 1000 Questions
R9 1001
2 Eight bits are reserved for the opcode field of a particular
R10 1010
processor's instruction set.
R11 1011
(a) What is the total number of codes that could be used
R12 1100
as opcodes for this processor?
(b) Why might only some of these 8-bit codes be valid?
Key concept
Structure of a machine code instruction:
(c) The 8-bit opcode is subdivided into two parts. What
are the two parts?
A machine code instruction consists of an
opcode and one or more operands of type: value,
memory address or register.

284 Single licence - Abingdon School

 7.3.3 The processor instruction set

Questions
3 What is meant by processor instruction set?

4 Explain using LOAD, STORE, ADD that machine code instructions

consist of an opcode and one or more operands which may be value,
memory address or register.

5 What is meant by saying that an instruction set is processor specific?

Task
1 You are required to design an instruction set for a processor based on a
two register architecture in which the destination register is always the
register called the accumulator. The processor must be able to add and
subtract. The only instructions allowed to interact directly with main
memory are load and store. The first input to an arithmetic operation
is always read from the accumulator. The second input is read from
one of fifteen other registers. Values may be set up in all sixteen
registers either by a load instruction or by a move instruction. The
second operand of a move instruction is always a value. The processor's
instruction set must support four control instructions:
1. branch on zero
2. branch on negative
3. branch unconditionally
4. Halt
Branch instruction 1 tests the zero flag and branch instruction 2,
the negative flag of the status register. All three branches use a single
operand which is a value (positive or negative) to add to the current
value of the program counter. The status register and program counter
are separate from the sixteen registers also used by the instruction set.
Registers have a word length of 16 bits as does main memory.
Main memory consists of 256 memory words.
Some instructions may not use all 16 bits. The processor will use those
bits which define the instruction.

In this chapter you have covered:

■■ The meaning of ‘processor instruction set’ and that an instruction set is
processor specific
■■ The structure of machine code instructions which is an opcode and one or
more operands (value, memory address or register)

Single licence - Abingdon School 285

 7 Fundamentals of computer organisation and architecture

7.3 Structure and role of the processor and its components

■■Learning objectives:
• Understand and apply
immediate and direct
■■ 7.3.4 Addressing modes
addressing modes Immediate addressing
When the addressing mode is immediate addressing the operand is the datum.
Information
For example, the MOV operation copies the value specified by <operand2>
0x: into register Rd.
0x indicates a hexadecimal
number, e.g. 0x3F. MOV Rd,<operand2>
If Rd is register R0 and <operand2> is 0xFF in hexadecimal (255 in decimal)
then the assembly language instruction is as follows
R0 MOV R0,#0xFF
before ? The # in front of 0xFF indicates that the mode of addressing is immediate
addressing. The contents of register R0 will be the binary equivalent of
hexadecimal value FF after this instruction is executed - Figure 7.3.4.1.
after 0xFF

Figure 7.3.4.1 register R0 before Questions

and after execution of instruction
MOV R0,#0xFF 1 Register R0 contains the datum 4310, register R1 the datum 5610. What
will register R0 contain after each of the following assembly language
Key concept instructions have been executed?
Express your answer in decimal.
Immediate addressing:
The operand is the datum. (a) MOV R0,#0x5E
(b) ADD R0,R1,#0x5E

Key concept
Direct addressing
Direct addressing:
The operand is the address of
When the addressing mode is direct addressing the operand is the address in
the datum. memory where the datum can be found.
For example, the LDR operation loads the value stored in the memory location
specified by <memory ref> into register Rd.
LDR Rd,<memory ref>
If Rd is register R0 and <memory ref> is in hexadecimal 0xFCC0 (64704
in decimal) then the assembly language instruction is as follows
LDR R0,0xFCC0
Note that in this instruction there is no #. The absence of the # symbol
indicates that this is direct addressing.

Single licence - Abingdon School 286

7 Fundamentals of computer organisation and architecture

The operand 0xFCC0 is the main memory address of a memory location containing the datum to be used when
this instruction is executed. The contents of register R0 will therefore be 0x4D (we can omit the leading 00) after
this instruction is executed because memory location with address 0xFCC0 contains the datum 0x4D - Figure
7.3.4.2.
Main Memory
Address
R0
0xFFFF 0xFCC2
before ?
0x0000 0xFCC1

0x004D 0xFCC0
after 0x4D
0x014E5 0xFCBF
Figure 7.3.4.2 register R0 before
and after execution of instruction 0x3628 0xFCBE
LDR R0,0xFCC0

Questions
2 Register R0 contains the datum 4310, main memory contents are as shown in Figure 7.3.4.2. What will
register R0 contain after the following assembly language instruction has been executed?
Express your answer in decimal.
LDR R0,0xFCC0
3 In the assembly language instruction STR Rd,<memory ref>, the STR operation stores the value
that is in register Rd in a memory location specified by <memory ref>. Register R0 contains the
datum 4310, register R1 the datum 5610, main memory contents are as shown in Figure 7.3.4.2. What
will the memory locations 0xFCC1 and 0xFCC2 contain after the following two instructions have been
executed? Express your answers in decimal.
STR R0,0xFCC1
STR R1,0xFCC2

4 Main memory contents are as shown in Figure 7.3.4.2 What will be stored in register R1 after the
following instructions are executed? Express your answer in decimal.
LDR R0,0xFCC0
ADD R1,R0,#0xFCC0

In this chapter you have covered:

■■ Immediate addressing: the operand is the datum
■■ Direct addressing: the operand is the address of the datum

287 Single licence - Abingdon School

7 Fundamentals of computer organisation and architecture

7.3 Structure and role of the processor and its components

Learning objectives:
■ 7.3.5 Machine-code and assembly language operations
■ Understand and apply the basic Load-Store architecture
machine-code operations of In a load-store architecture the only instructions that work directly with
• load memory are load and store instructions or their equivalent. A value in memory
that needs to be processed
• add
must be loaded into the
• subtract processor (core) first,
• store processed and then stored
• branching (conditional back in memory.
and unconditional) Load
• compare A load register operation
is used to transfer a
• logical bitwise operators
copy of a datum from a
(AND, OR, NOT, XOR)
specified location, e.g. main Figure 7.3.5.1 ASMTutor after executing the
• logical memory location 102, to a machine code equivalent of MOVE 102, R0
 shift right symbolically named register,
 shift left e.g. R0.
• halt For example using direct addressing, LDR R0, 102 transfers the contents of
■ Use the basic machine-code memory location with address 102 into register R0. In some instruction sets,
operations above when machine- the mnemonic MOV or MOVE is used instead of the mnemonic LDR, and
code instructions are expressed the order of the operands can be reversed. The simulator ASMTutor shown in
in mnemonic form- assembly Figure 7.3.5.1 is one that uses MOVE instead of LDR for a load operation. It
language, using immediate and also reverses the order of the operands.
direct addressing Figure 7.3.5.1 shows an example of an assembly language program which
transfers a copy of the datum 32 from memory location 102 to register R0. The
Key concept assembly language program ends with RTS (ReTurn from Subroutine) because
Load-store architecture: ASMTutor expects the last
No direct manipulation of instruction to be RTS in
memory contents. A value
order to work correctly.
in memory that needs to be
Note that memory location
processed must be loaded
into the processor (core) first, with address 102 in this
processed and then stored back example contains 32.
in memory. Figure 7.3.5.2 shows the
register state just before the
Information machine code equivalent of
See Table 7.3.3.5 in Chapter 7.3.3 MOVE 102, R0 is executed. Figure 7.3.5.2 ASMTutor just before executing
for AQA instruction set.
the machine code equivalent of MOVE 102, R0

Single licence - Abingdon School 288

7 Fundamentals of computer organisation and architecture

Figure 7.3.5.3 shows

the result when memory
Memory Main memory
address contents
location 102 contains 64.
(in decimal) (in decimal)

102 21
103 42
104 84

Figure 7.3.5.4

Figure 7.3.5.3 Just after with 102 containing 64

Questions
1 The assembly language instruction LDR R0, 102 transfers a copy of the contents of memory location 102 to
register R0. Figure 7.3.5.4 shows the contents of memory locations 102, 103 and 104.
What does register R0 contain after the following instructions, expressed in assembly language, are executed
in machine code?
(a) LDR R0, 102 (b) LDR R0, 103 (c) LDR R0, 104

Figure 7.3.5.5 ARM® μVision® V 5.17 Debugger in single-step mode

Information Figure 7.3.5.5 shows a screenshot of the ARM μVision simulator in debugger
ARM μVision may be
mode single-stepping through an assembly language program written for an
downloaded from
ARM Cortex processor. ARM instruction sets use LDR but the reference to a
https://www.keil.com/download/
product/ memory location is not direct. Instead, the memory reference must be obtained
The evaluation mode is free. from a register. This example uses register R0.
To configure for Thumb-2 [R0] means contents of register R0. LDR R4, [R0] loads register R4 with the
instruction set, select Project/
contents of memory location 0x00000000 as this is the memory address stored
Options for Target.../Asm and tick
Thumb Mode.
in register R0. The notation 0x indicates a hexadecimal number.

289 Single licence - Abingdon School

 7.3.5Machine-codeandassemblylanguageoperations

Figure 7.3.5.6 shows the result of executing the instruction LDR R4, [R0].
The ARM μVision simulator
simulates the execution of
instructions for the Cortex™-M
family of ARM microcontrollers.
These microcontrollers implement
the ARMv7 instruction set. In order
for this simulator to function, every
assembly language program must
start with the preamble shown in
Figures 7.3.5.5, 7.3.5.6. The user
chooses the identifier in AREA
FirstASM1, CODE, i.e. FirstASM1,
but the rest of the preamble must
Figure 7.3.5.6 ARM® μVision® V 5.17 Debugger in single-step more
conform to that given.
STORE
A store operation, STR, transfers a copy of the contents of a register to a specified memory location,
e.g. STR R4, 0x20000000
If R4 contains 0x00000065 then
execution of the machine code
equivalent of this instruction will
change the contents of memory
location, address 0x20000000, to
the value 0x65.
Figure 7.3.5.7 shows register R4
preset with value 0x00000065,
register R0 preset with value
0x20000000, and memory Figure 7.3.5.7 ARM® μVision® V 5.17 Debugger in single-step more
location 0x2000000 initialised to
0x00000000. The next instruction
to be executed is STR R4, [R0]
which is equivalent to
STR R4, #0x20000000

Figure 7.3.5.8 shows memory

location 0x20000000’s contents
with value 0x65, the result of
executing STR R4, [R0].
The ARM μVision simulator is
configured so that 0x20000000 is
Figure 7.3.5.8 ARM® μVision® V 5.17 Debugger in single-step more
the first available location that a
program may write to.

Single licence - Abingdon School 290

7 Fundamentals of computer organisation and architecture

Questions
2 The assembly language instruction STR R0, 102 transfers a copy of the contents of register R0 to memory
location with decimal address 102. What are the contents of memory location 102 after this instruction is
executed in machine code when the value in R0 is decimal 67?

MOVE
A MOVE operation copies a value from source to destination. The source could be a register, an immediate value
(or a memory location but not in the case of ARM processors). The destination could be another register (or a
memory location but not in the case of ARM processors).
For example, the ARM processor instruction set has a MOV instruction (MOVS in ARMv7 to update status register as
well). MOV Rd, <operand2> copies the value specified by <operand2> into register Rd. For example, MOV R2, #36
copies the value 36 into the register R2. # indicates immediate addressing.
MOV R1, R2 copies the value in register R2 into register R1. Figure 7.3.5.9 shows an assembly language program,
MOV.s, prepared in Notepad++ and then loaded into ArmSim# version 1.9.1 for assembling and executing one
instruction at a time. Register contents are shown in the window on the left. Note how registers R2 and R3 are
changed by the two different MOV operations. The hash symbol # before the decimal value 36 indicates immediate
addressing, i.e. the operand is the value to be used. As with ASMTutor, the last instruction must be an instruction
that enables ArmSim# to function correctly. This instruction is SWI 0x11.

Figure 7.3.5.9 Single stepping through execution

of machine code equivalent of assembly language
program containing MOV Rd, <operand2> for an
immediate operand and a register operand

ArmSim simulates ARMv5 instruction set architecture.

It can be downloaded from
http://armsim.cs.uvic.ca/DownloadARMSimSharp.html

Questions
3 The assembly language instruction MOV R0, R1 transfers a copy of the contents of register R1 to register
R0. What does register R0 contain after the following instructions, expressed in assembly language, are
executed in machine code?
(a) MOV R0, #78 (b) MOV R1, #25 followed by MOV R0, R1

291 Single licence - Abingdon School

 7.3.5Machine-codeandassemblylanguageoperations

ADD
An add operation ADD Rd, Rn, <operand2> is
used to add the value specified in <operand2>
and the value in register Rn, storing the result in
register Rd. <operand2> may be an immediate
value or a register.
For example, ADD R2, R3, #1 when executed in
machine code adds 1 to the contents of register
R3 before storing the result in register R2.

Figure 7.3.5.10 shows ArmSim# stepping

through program ADD1.s one instruction at a
time. This program assigns 24 to R3 and 36 to
R2. It adds 1 to a copy of the 24 stored in R3
then stores the result 25 in R2.

Task
1 Using a text editor such as Notepad++,
create the file ADD32.s with the
following contents:

MOV R4, #25

MOV R5, #43
ADD R5, R4, #32
SWI 0x11
Load ADD32.s into ArmSim# and single
step through the instructions. Observe
how registers R4 and R5 change.

Task
2 Using a text editor such as Notepad++,
create the file ADD.s with the following
contents:

MOV R4, #25

MOV R5, #43
MOV R3, #1
Figure 7.3.5.10 Before and after state of registers R2, R3 for
ADD R5, R4, R3
execution of ADD R2, R3, #1
SWI 0x11
Questions
Load ADD.s into ArmSim# and single
4 Write an assembly language program that stores the
step through the instructions. Observe
result of adding the contents of registers R1, R2, R3 in
how registers R3, R4 and R5 change.
R0. The program will need to initialise R1 with decimal
5, R2 with decimal 3 and R3 with decimal 6.

Single licence - Abingdon School 292

7 Fundamentals of computer organisation and architecture

SUBTRACT
A subtract operation SUB Rd, Rn,
<operand2> is used to subtract the value
specified in <operand2> from the value
in register Rn before storing the result
in register Rd. <operand2> may be an
immediate value or a register.
For example, SUB R2, R3, #1 when
executed in machine code subtracts 1 from
the contents of register R3 before storing
the result in register R2.
Figure 7.3.5.11 shows the before and after
contents of the registers R2 and R3.

Task
3 Using a text editor such as
Notepad++, create the file SUB5.s
with the following contents:

MOV R4, #25

Figure 7.3.5.11 Before and after state of registers R2, R3 for
MOV R5, #43
execution of SUB R2, R3, #1
SUB R5, R4, #5
SWI 0x11 Questions
Load SUB5.s into ArmSim# 5 Write an assembly language program that stores the result of
and single step through the subtracting the contents of register R1 from R2 in R0. The
instructions. Observe how registers program will need to initialise R1 with decimal 5 and R2
R4 and R5 change. with decimal 13.

Status register
The status register contains flags called condition codes which are set or reset to reflect the outcome of the last
machine operation, e.g. if the result of an arithmetic operation was zero then the zero flag is set. A flag is a single bit
code that can be set (binary 1) or reset (binary 0). A status register consists of at least four condition codes:
• Zero flag - set if the result of the last machine operation stores zero in the results register
• Negative flag - set if the result of the last machine operation stores a negative value in the results register
• Carry flag - set if the result of an unsigned operation overflows the result register or as a result sometimes of
performing two’s complement signed arithmetic.
• Overflow flag - set if the result of a signed operation overflows the result register.
Moving zero into a register can set the Z(ero) flag as will an arithmetic operation if the result is zero.
Moving a negative value into a register can set the N(egative) flag as will an arithmetic operation if the result is
negative. A carry can be produced when a machine performs two’s complement arithmetic or when it performs
unsigned addition.

293 Single licence - Abingdon School

 7.3.5Machine-codeandassemblylanguageoperations

Figure 7.3.5.12 shows an assembly language program

created in ASMTutor. It has been assembled so it can be
executed. The first screenshot shows that the next instruction
to be executed is MOVE #0, R0. The second screenshot shows
the effect on the status register of executing this instruction.
The zero flag has been set in the status register (indicated by
a Z).
The third screenshot shows that the negative flag has been set
in the status register as a result of the machine executing the
instruction MOVE #-1, R1. Note the value stored in R1 is
6553510 or 11111111111111112 which is the two’s complement
representation for -110.
The fourth screenshot shows the effect on the status register
of subtracting decimal 5 (register R2) from 0 (register R0).
The negative flag is set because the result is negative.
Figure 7.3.5.13 shows that when the largest positive number
01111111111111112 (7FFF16) is added to itself, overflow
results and the overflow flag is set. The negative flag is also
set because the result, 6553410 , that is stored in register R0 is
interpreted as -210 by the machine.

Questions
6 The format of the MOVE operation in an
instruction set is MOV Rd, <operand2> which
is interpreted as copy the value specified by
<operand2> into register Rd. Assuming that MOV
can set the condition codes.
Which, if any, status register condition codes are
set when the machine code equivalents of the
following are executed
(a) MOV R0, #-1 (b) MOV R1, #0 Figure 7.3.5.12 Single-stepping an assembly
2 (c) MOV R2, #23? language program in ASMTutor to show the effect
7 The format of the SUBTRACT operation in an on the status register
instruction set is SUB, Rd, Rn, <operand2> which
is interpreted as subtract the value specified in
<operand2> from the value in register Rn and
store the result in register Rd. Register R0 stores
decimal 7. Which status register condition codes
are set, if any, when the machine code equivalents
of the following are executed
(a) SUB R1, R0, #9 (b) SUB R1, R0, #7
(c) SUB R1, R0, #5?

Figure 7.3.5.13 Shows the result of adding the

most positive value $7FFF or 7FFF16 to itself

Single licence - Abingdon School 294

7 Fundamentals of computer organisation and architecture

COMPARE
Compare instructions may be used to
compare the contents of two registers or
the contents of a register and an immediate
value. For example, CMP R0, R1 compares
the contents of registers R0 and R1. If
the contents of these registers are equal,
the zero flag in the status register is set. A
compare operation performs a subtraction
and uses the result to determine whether
the two operand’s values are equal or not. If
Figure 7.3.5.14 ARM µVision simulating the execution of CMP R0, #9 unequal then the negative flag will be set if
the subtraction result was negative. Figure
7.3.5.14 shows ARM μVision simulating
the execution of CMP R0, #9 with decimal
7 stored in R0. The negative flag is set
indicating that the operation [R0] - 9 has
been performed by CMP. The notation []
means ‘contents of ’.
Figure 7.3.5.15 shows the result of CMP
R0, #7 with decimal 7 stored in R0.
Note that the zero flag is set. The carry
flag is also set because two’s complement
arithmetic sometimes sets this flag.
Figure 7.3.5.15 ARM µVision simulating the execution of CMP R0, #7

Questions
8 What is the state of each condition code after the following comparison operations are executed in machine
code? R0 contains the value decimal 9. Assume that CMP behaves as shown above.
(a) CMP R0, #15 (b) CMP R0, #7 (c) CMP R0, #9

Branching (conditional and unconditional)

Normally, a processor executes one instruction after another in a linear fashion. This means the next instruction
to execute is found immediately following the current instruction. Branch instructions allow for a different order
of execution. For example, the B loop instruction in Table 7.3.5.1 causes the previous instruction to be repeated
indefinitely. The previous instruction is labelled loop so that the branch instruction can refer to it. The assembler
will convert this symbolic label into a memory address when it translates the assembly language program into its
machine code equivalent - object code column in Table 7.3.5.2.
Assembly language instructions or statements are divided into four fields separated by spaces or tabs as shown in
Table 7.3.5.1.
Label Opcode Operand Comment field
field field field(s)
MOV R0, #1 ; initialise counter to 1, R0 will hold a running count, R0 = 1
loop ADD R0, R0, #1 ; increment counter by 1, R0 = R0 + 1
B loop ; repeat previous instruction
END ; this is a pseudo-op that marks the end of the program to the assembler
Table 7.3.5.1 ARM assembly language program showing how instructions
AQA uses HALT are divided into four fields
295 Single licence - Abingdon School
 7.3.5Machine-codeandassemblylanguageoperations
Object code
The label field is optional and starts in
the first column. It is used to identify
the position in memory of the current
instruction. It must be unique within
the program. The opcode field expresses
the processor command to execute. The
operand field specifies where to find the
data the command uses when it executes.
ARM processor instructions have 0, 1, 2,
3 or 4 operands separated by commas. We Source code
will consider instructions that use only 0,
1, 2 or 3 operands.
The comment field is optional and
is ignored by the assembler. It allows
Figure 7.3.5.16 ARM µVision simulating the execution of a
a programmer to write a few words
program that uses an unconditional branch instruction, B
describing the purpose of the instruction,
e.g. ‘increment counter by 1’, to make it Address Object Label Opcode Operand
easier to understand. A semicolon (;) is used to separate the code
operand and comment fields. 0x0000008 2001 MOV R0, #1
The assembler translates assembly language source code 0x000000A 1C40 loop ADD R0, R0, #1
into object code. Object code consists of the machine 0x000000C E7FD B 0x000000A
instructions executed by the processor. Table 7.3.5.2 shows Table 7.3.5.2 ARM assembly language program
ARM processor (Thumb-2 instruction set) object code showing both source and object code
alongside its equivalent assembly language source code. The Object code
first column shows the address in RAM of each machine code Decimal Hexadecimal
byte
instruction, e.g. 0x00000008. The second column the opcode + 8 8 20
operands, e.g. 1C40. The third, fourth and fifth columns show 9 9 01
the corresponding assembly language source code. The comment 10 A 1C
field has been omitted. Figure 7.3.5.16 shows ARM μVision 11 B 40
simulating this program. The loop label has been replaced in 12 C E7
13 D FD
the instruction B loop by the memory address 0x000000A
corresponding to label loop. In object code this is translated Table 7.3.5.3 Memory map for ARM machine
into the value to ‘add’ to the current address because ARM uses code program
relative addressing1. The new address becomes the address of the next instruction to be fetched and executed. In this
case FD in hexadecimal or -3 in decimal because two’s complement coding is used for numbers.
Table 7.3.5.3 shows the memory map for the program’s machine code. Note that 1310 is the address of byte value
FD or -310. 1310 - 310 = 1010 = A16. A16 is the address of instruction opcode 1C which is ADD in assembly language
mnemonics.
Unconditional branch
The unconditional branch instruction B label always causes execution to branch (jump) to the instruction at the
address indicated by label. Using direct addressing this would be for the example program
B 0x000000A

1 relative addressing not covered in AQA specification

Single licence - Abingdon School 296
7 Fundamentals of computer organisation and architecture

Conditional branch
There is another kind of branch called a conditional branch. In this type of branch a condition must be true for
branching of program execution to occur.
The instruction immediately before a conditional branch must be a COMPARE instruction. Execution of this
instruction affects the condition code flags which conditional branch instructions examine before deciding whether
or not to branch (SUBTRACT can be used instead of COMPARE, e.g. is an alternative to CMP R0, R1).

CMP R0, R1 Condition Condition codes

R0 = R1 Equal Zero flag set, Z = 1
R0 <> R1 Not Equal Zero flag not set, Z = 0
R0 > R1 R0 Greater Than R1 Z = 0, N = 0
R0 < R1 R0 Less Than R1 Z= 0, N = 1

Table 7.3.5.4 Condition and condition codes for SUB and CMP

Table 7.3.5.5 shows the four conditional branch instructions, BEQ, BNE, BGT and BLT.

Condition
Instruction Description
codes
BEQ <label> Branch if operands being compared are equal Z=1
BNE <label> Branch if operands being compared are not equal Z=0
Branch if first signed operand is greater than
BGT<label> Z = 0, N = 0
second signed operand
Branch if first signed operand is less than second
BLT <label> Z = 0, N = 1
signed operand

Table 7.3.5.5 Conditional branch instructions

Figure 7.3.5.17 shows the simulation of

conditional branch BEQ loop.

Questions
9 Explain what the following snippet of
assembly language code does when its
machine code equivalent is executed

MOV R0, #12

MOV R1, #6
loop ADD R1, R1, #1
CMP R1, R0
BNE loop
HALT ; Stops the execution

Figure 7.3.5.17 ARM µVision simulating the execution of a

program that uses conditional branch instruction, BEQ

297 Single licence - Abingdon School

 7.3.5Machine-codeandassemblylanguageoperations

Questions
10 Explain what the following snippets of assembly language code do when their machine code equivalent is
executed

(a) MOV R0, #12 (b) MOV R0, #12

MOV R1, #6 MOV R1, #6
loop SUB R0, R0, #1 loop ADD R1, R1, #1
CMP R0, R1 CMP R0, R1
BGT loop BLT loop
HALT ; Stops the execution HALT

11 What other conditional branch instruction would result in the code behaving in a similar way if used in
place of BGT and BLT in (a) and (b)?

Logical bitwise operators

When designing digital logic gate circuits gates are used, such as AND, OR, NOT, which convert single bit input
signals into single bit output signals.
For example, with the AND gate, if the inputs are 1 and 0 then the output is 0 because 1 AND 0 = 0.
Using AND, OR, NOT and XOR as operators in assembly language programs is slightly different. The inputs are
typically 32-bit numbers and the output is a single 32-bit number. The inputs are transformed into the output by
applying 32 logic operations, e.g. AND, at the same time in a bitwise fashion.
The format for ARM processors for the logical operations AND, OR and XOR is
Logical operation Rd, Rn, <operand2>

This means perform a bitwise logical operation between the value in register Rn and the value specified by
<operand2> and store the result in register Rd. The symbolic opcode for the AND operation is AND; for the OR
operation it is ORR and for XOR it is EOR.
The format for ARM processors for the logical operation NOT is
MVN Rd, <operand2>

This means perform a bitwise logical NOT operation on the value specified by <operand2> and store the result in
register Rd.
AND
Figure 7.3.5.18 shows ARM μVision
simulating 11112 AND 00012. The result
is 00012 when AND R0, R0, R1 is executed
in machine code. This instruction ANDs the
contents of registers R0 and R1 and stores the
result in R0, the register specified as the first
operand.
A mask operation is one that isolates bits to
Figure 7.3.5.18 ARM µVision simulating the execution of a
be tested. The logical AND can be used in
program that applies a bitwise AND operation to operands
this role. Suppose that we need to test the
0xF and 0x1 i.e. 11112 AND 00012
three least significant bits of a 32-bit word
Single licence - Abingdon School 298
7 Fundamentals of computer organisation and architecture

then we would choose the mask 0x00000007

because the last three bits are 1112 (716) and
the other bits are 0. If the bit pattern to be
tested is stored in R0 then R1 will contain
the state of the three least significant bits and
zeroes everywhere else after executing
AND R1, R0, #0x7

To know if all three least significant bits are 1

then compare R1 with 0x7 as follows
CMP R1, #0x7

The zero flag will be set by CMP if they are. Figure 7.3.5.19 ARM µVision simulating the execution of a
program that applies a bitwise OR operation to operands 0x8 and
0x7 i.e. 10002 OR 01112
OR
Figure 7.3.5.19 shows ARM μVision
simulating 10002 OR 01112. The result is
11112 when ORR R0, R0, R1 is executed
in machine code. This instruction ORs the
contents of registers R0 and R1 and stores
the result in R0, the register specified as the
first operand.
XOR
XOR is the eXclusive-OR operation.
ARM names the operator for this operation
EOR (Exclusive-OR) Figure 7.3.5.20 ARM µVision simulating the execution of a
Figure 7.3.5.20 shows ARM μVision program that applies a bitwise XOR operation to operands 0x8
simulating 10002 XOR 01112. The result is and 0x7 i.e. 10002 XOR 01112
11112 when EOR R0, R0, R1 is executed in
machine code. This instruction Exclusive-
ORs the contents of registers R0 and R1 and
stores the result in R0, the register specified
as the first operand.
NOT
To perform a bitwise logical NOT operation
the instruction
MVN Rd, <operand2>

This instruction NOTs the value specified by

<operand2> and stores the result in register
Rd. Figure 7.3.5.21 ARM µVision simulating the execution of
a program that applies a bitwise NOT operation to operand
Figure 7.3.5.21 shows ARM μVision
0x00000000, i.e. NOT 000000000000000000000000000000002
simulating NOT 0x00000000.
299 Single licence - Abingdon School
 7.3.5Machine-codeandassemblylanguageoperations

The result is 111111111111111111111111111111112 when MVN R0, R1 is executed in machine code. The result
expressed in hexadecimal is FFFFFFFF. Note that R1 was assigned 0x0 in a MOV operation first.

Questions
12 What will the contents of register R0 be after the machine code equivalent of the following snippets of
assembly language code are executed
(a) MOV R0, #0xFF (b) MOV R0, #0xFF (c) MOV R0, #0x0
EOR R0, R0, R0 MOV R1, #0x7 MOV R1, #0x7
HALT AND R0, R0, R1 ORR R0, R0, R1
HALT HALT
13 A certain process may begin if bits 1, 3 and 5 of an 8-bit word are set. The state of the other bits may be
ignored. Write the assembly language instructions to determine if the process may begin. You should assume
that bit 1 is the least significant bit and that register R0 contains the 8-bit word.

14 Write an assembly language instruction using ORR to set bit 4 of register R0. Assume bits are numbered
from the right 1...8 with bit 1 the least significant bit.

15 Write an assembly language instruction to isolate bits 1 and 3 of register R0 so that the state of each may be
tested by other instructions. Assume bit numbering as in Q14.

16 “We use the logical OR to make bits become one, and we use the logical AND to make bits become zero.”
Explain using examples the meaning of this statement.

17 Register R0 contains a 32-bit word that represents the state of 32 pixels of a black and white image with
colour depth one bit per pixel. Write a single assembly language instruction to invert the state of each pixel
stored in R0. Write another instruction to restore the stored state.

Logical shift operations

A logical shift treats the bit pattern as being an unsigned pattern of bits. A shift operation takes two inputs, one the
number of shifts to apply, n, and the other the bit pattern to be shifted by n bits. For example, the bit pattern in
Figure 7.3.5.22 (a) when shifted by one bit to the left becomes the bit pattern shown in Figure 7.3.5.22(b).
1 0 1 1 0 1 1 0
Figure 7.3.5.22(a) 8-bit bit pattern before it is shifted left one bit
0 1 1 0 1 1 0 0
Figure 7.3.5.22(b) 8-bit bit pattern after it has been shifted left one bit
Logical shift left operation
With a logical shift left the bit pattern is moved to the left with the least significant bit position replaced by a zero.
The carry bit will contain the last bit shifted out. For the example in Figure 7.3.5.22 the carry bit will contain
1 after shifting left one bit. The bit pattern in Figure 7.3.5.22(a) is unsigned decimal 178 or B2 in hexadecimal.
In ARM assembly language, the 32-bit register R0 contains B216 after MOV R0, #0xB2 is executed. The ARM
instruction LSL R1, R0, #1 shifts the bit pattern in R0 left one bit and stores the result 16416 or 35610 in R1.
Notice that the value stored in R1 is double the value in R0. This is equivalent to multiplying by 21. If the shift
operation is LSL R0, #2 and R0 contains 17810 or B216 then 2C816 or 71210 will be stored in R1. This is equivalent to
multiplying by 22.

Single licence - Abingdon School 300

7 Fundamentals of computer organisation and architecture

Figure 7.3.5.23 shows the result of applying

logical shift left to 0x1 eight times. The loop
was stepped through eight times to change
the pattern in R0 from 0x00000001 to
0x00000100. (Note these are hex numbers)

Questions
18 Rewrite
loop LSL R0, R0, #1
B loop
END
to obtain the bit pattern Figure 7.3.5.23 Logical Shift Left by one bit applied 8 times
0x00000100 from the bit pattern by single-stepping through loop eight times
0x00000001 stored in R0 without
a loop.

Questions
19 The decimal number 4 is stored in register R0. Write an assembly language instruction that multiples this
number by 24.
20 Using Figure 7.3.5.22 as a template, record the state of R0 after the following assembly language program
is executed in machine code. Assume R0 is an 8-bit wide register. Include the carry bit in your answer.
MOV R0, #3
LSL R0, R0, #7
HALT

Logical shift right

With a logical shift right the bit pattern is moved to the right with the most significant bit position replaced by a
zero. The carry bit will contain the last bit shifted out.
For the example in Figure 7.3.5.24 the carry bit will contain 0 after right shifting one bit. The bit pattern in Figure
7.3.5.24(a) is unsigned decimal 178 or B2 in hexadecimal. In ARM assembly language, the 32-bit register R0
contains B216 after MOV R0, #0xB2 is executed. The ARM instruction LSR R1, R0, #1 shifts the bit pattern in R0
right one bit and stores the result 5916 or 8910 in R1. Notice that the value stored in R1 is half that in R0. This is
equivalent to dividing an unsigned number by 21. If the shift operation is LSR R1, R0, #2 and R0 contains 17810 or
B216 then 2C16 or 4410 will be stored in R1. This is equivalent to unsigned integer division by 22, with the remainder
1 stored in the carry bit.

1 0 1 1 0 1 1 0
Figure 7.3.5.24(a) 8-bit bit pattern before it is shifted right one bit
0 1 0 1 1 0 1 1 0 Carry bit
Figure 7.3.5.24(b) 8-bit bit pattern after it is shifted right one bit

301 Single licence - Abingdon School

 7.3.5Machine-codeandassemblylanguageoperations

Questions
21 R0 contains 0x00000100. What does it contain after LSR R0, R0, #8 in machine code is executed?

22
23 The decimal number 64 is stored in register R0. Write an assembly language instruction that divides this
number by 24. The result should be stored in register R1.

23 Using Figure 7.3.5.24 as a template, record the state of R0 after the following assembly language program
is executed in machine code. Assume R0 is an 8-bit wide register. Include the carry bit in your answer.
MOV R0, #195
LSR R0, R0, #7
HALT

HALT
When a HALT instruction is encountered in an executing machine code program the execution of the program is
stopped.

Questions
Use AQA’s instruction set from Table 7.3.3.5 in Chapter 7.3.3 to answer these questions.

24 The high level language program statement "Sum := Sum + 100;" assigns to variable Sum the result of
adding decimal number 100 to Sum. The symbol ":=" is the assignment operator. Write the equivalent
assembly language instructions for this statement. Assume that memory location with address 0x1000 is
used to store the current value of variable Sum.

25 Write the equivalent assembly language instructions for high level language statement
If Sum > 5 Then Sum := Sum + 1 Else Sum := Sum - 1;
Assume that memory location with address 0x1000 is used to store the current value of variable Sum.

26 Write the equivalent assembly language instructions for high level language statement
While Sum < 10 Do Sum := Sum + 1;
Assume that memory location with address 0x1000 is used to store the current value of variable Sum.

27 Write the equivalent assembly language instructions for high level language statement
Repeat Sum := Sum - 1 Until Sum = 0;
Assume that memory location with address 0x1000 is used to store the current value of variable Sum which
is decimal 10.

28 Write the equivalent assembly language instructions for high level language statement
Sum := Sum * 8;
Assume that memory location with address 0x1000 is used to store the current value of variable Sum, an
unsigned number.

29 Write the equivalent assembly language instructions for high level language statement
If (SwitchSettings BitWiseAND 4) = 1 Then Sum := 0 ;
Assume that memory location with address 0x1000 is used to store the current value of variable
SwitchSettings, an unsigned number and memory location 0x1004 the current value of variable Sum.
30 Use an assembly language simulator to check your answers.

Single licence - Abingdon School 302

7 Fundamentals of computer organisation and architecture

In this chapter you have covered:

■■ The basic machine-code operations of
• load - LDR Rd, <memory ref>
• add - ADD Rd, Rn, <operand2>
• subtract - SUB Rd, Rn, <operand2>
• store - STR Rd, <memory ref>
• branching (conditional and unconditional)
ŠŠ B <label>
ŠŠ BEQ <label>
ŠŠ BNE <label>
ŠŠ BGT <label>
ŠŠ BLT <label>
• compare - CMP Rn, <operand2>
• logical bitwise operators
ŠŠ AND - AND Rd, Rn, <operand2>
ŠŠ OR - ORR Rd, Rn, <operand2>
ŠŠ NOT - MVN Rd, <operand2>
ŠŠ XOR - EOR Rd, Rn, <operand2>
• logical
ŠŠ shift left - LSL Rd, Rn, <operand2>
ŠŠ shift right - LSR Rd, Rn, <operand2>
• halt - HALT
■■ The use of the basic machine-code operations above when machine-code
instructions are expressed in mnemonic form - assembly language, using
immediate and direct addressing
■■ The instructions set of assembly language mnemonics identified by AQA
to be used in questions - see Table 7.3.3.5 in Chapter 7.3.3. Question
papers will supply the list of mnemonics and their description so that they
do not need to be memorised.

303 Single licence - Abingdon School

 7 Fundamentals of computer organisation and architecture

7.3 Structure and role of processor and its

Learning objectives: components
■■Describe the role of interrupts
■■Describe the role of interrupt
■■ 7.3.6 Interrupts
The role of interrupts
service routines (ISRs) and
Virtually all computers provide a mechanism by which a program currently
• their effect on the fetch- executing on the processor may be interrupted by a module such as an I/O
execute cycle controller, seeking the attention of the processor. The module generates a signal
• the need to save the called an interrupt signal which is sent along a control line to the processor.
volatile environment while Thus an interrupt may be defined as follows:
the interrupt is being serviced
An interrupt is a signal from some device/source seeking the
attention of the processor.
Key concept
Interrupt: If interrupts are enabled then, on receipt of an interrupt, the currently
An interrupt is a signal from executing program is suspended in an orderly fashion and control is passed to
some device/source seeking the
an interrupt service routine. The currently executing program is suspended in
attention of the processor.
such a way that its execution can be resumed without error after the servicing
of the interrupting device has been carried out.

Single licence - Abingdon School 304

7 Fundamentals of computer organisation and architecture

Sources of interrupt
There are many sources of interrupt. Table 7.3.6.1 shows the main ones and
their priority with 1 being the highest and 4 the lowest.

Class of interrupt Source of interrupt Priority

Power failure
Hardware failure 1
Memory parity
Arithmetic overflow
Division by zero
Attempt to execute
an illegal machine
instruction 2
Program Reference outside a
user’s allowed memory
space
Supervisor call to cause
mode to switch from
user to privileged
Real time clock
This allows the
Timer operating system 3
to perform certain
functions at regular
intervals of time
Generated by an I/O
controller to signal
normal completion
of an operation or to
signal a variety of error
conditions, e.g.
I/O disk block of data 4
transfer into main
memory completed,
keyboard key pressed,
printer ready to
accept next block/
line of characters
Table 7.3.6.1 Classes of interrupt

305 Single licence - Abingdon School

 7.3.6 Interrupts

The role of interrupt service routines (ISRs) Key concept

What happens when, for example, a key on the keyboard is pressed, thus
Interrupt service routine:
generating an interrupt? A small program called an interrupt service routine
An interrupt service routine is
(ISR) or interrupt handler is executed to transfer the character code value a small piece of program code
of the key pressed into main memory. A different ISR is provided for each written to process an event such
different source of interrupt. as a key on a keyboard being
pressed.
The effect on the fetch-execute cycle
A typical sequence of actions when an interrupt occurs would be:

1. The processor must complete the current fetch-execute cycle for the
current program if begun;
2. The contents of the program counter, which points to the next
instruction of the current program to be executed, must be stored away
safely so it can be restored after servicing the interrupt;
3. The contents of other registers used by the current program are stored
away safely for later restoration;
4. The source of the interrupt is identified;
5. Interrupts of a lower priority are disabled;
6. The program counter is loaded with the start address of the relevant
interrupt service routine;
7. The interrupt service routine is executed.

After the interrupt service routine has completed its execution:

1. The saved values belonging to the current program for registers other
than the program counter are restored to the processor’s registers;
2. Interrupts are re-enabled;
3. The program counter is restored to point to the next instruction to be
fetched and executed in current program.

The need to save the volatile environment

The volatile environment of the processor refers to the contents of processor Key concept
registers, e.g. the program counter, the general purpose registers, the status
register. Running the interrupt service routine causes the contents of processor Volatile environment:
The volatile environment of the
registers to change. Unless the volatile environment is saved before these
processor refers to the contents
changes occur, restoring the contents of the affected registers will be impossible of processor registers.
as will returning the processor to the exact state it was in just before executing
the interrupt service routine.

Single licence - Abingdon School 306

7 Fundamentals of computer organisation and architecture

To service an interrupt, the program counter contents must be changed from

the memory address of the next instruction to be executed of the program
that is being interrupted to the memory address of the first instruction of the
interrupt service routine responsible for servicing the interrupt. The interrupt
enable/disable flag of the status register must be changed to disable interrupts
of a lower priority. The interrupt service routine may well make use of one
or more of the general processor registers. Hence the need to save the volatile
environment before switching execution to the interrupt service routine.

Questions

1 What is an interrupt?

2 Give three examples of sources of interrupt.

3 Describe the role of interrupts.

4 Describe the role of an interrupt service routine.

5 Describe the typical sequence of actions when an interrupt

occurs and its effect on the fetch-execute cycle.

6 What is meant by the volatile environment?

7 Why is it necessary to save the volatile environment while

an interrupt is being serviced?

In this chapter you have covered:

■■ The role of interrupts
■■ The role of interrupt service routines (ISRs) and
• their effect on the fetch-execute cycle
• the need to save the volatile environment while the interrupt is being
serviced

307 Single licence - Abingdon School

7 Fundamentals of computer organisation and architecture

7.3 Structure and role of the processor and

its components
Learning objectives:
■■ 7.3.7 Factors affecting processor performance
■■Explain the effect on processor How many instructions can be executed per second?
performance of: We have learned already about the basic
computational model of CPU and memory from CPU
• multiple cores earlier chapters in Section 7. In this model, the
Bus
• cache memory program is fetched instruction by instruction from
main memory and executed in the CPU. The Main
• clock speed executing program accesses data in main memory
memory

• word length while it is executing. Figure 7.3.7.1 shows this basic Figure 7.3.7.1 Basic
model. The fetching, decoding and execution of an computational model
• address bus width
instruction is synchronised with the CPU’s clock.
• data bus width
The number of clock cycles ("ticks") of the CPU’s clock it takes the CPU
Information to execute an instruction varies from instruction to instruction with load
instructions which load data from memory taking the most.
No of clock cycles per
instruction: Suppose, the average number of clock cycles per instruction = 2
The number of clock cycles that 1
then the average number of instructions executed per clock cycle = / = 0.5
an instruction takes to execute is 2
determined by its complexity and If the CPU operates at a clock frequency of 800 MHz then there are 800
the design of the control unit in
million clock cycles per second.
the CPU. For a given Instruction
Set Architecture (ISA), the same Using this, we calculate that
instruction may take longer to
the average number of instructions executed per second is 0.5 x 800
execute on one processor than
another operating at the same = 400 million per second
clock frequency. The difference is
Cycles per instruction (CPI)
in the design of the control unit.
Cycles per instruction (clock cycles per instruction) is one aspect of a
processor’s performance. When evaluating processor performance the average
number of clock cycles per instruction is often used.
Information
Questions
Average no of clock cycles The performance of two processors with the same instruction set
1
per instruction:
architecture but operating at different clock frequencies is assessed
The percentage of instructions
that are load, store, integer by measuring the average number of cycles per instruction (CPI) for
arithmetic, branches varies various programs compiled by the same compiler and executed on each
from program to program so processor. Which processor do you think was the faster at executing
the average no of clock cycles
these programs? Justify your answer.
per instruction will vary from
program to program. Processor 1: Clock frequency 5 GHz CPI = 3
Processor 2: Clock frequency 3 GHz CPI = 1.5

Single licence - Abingdon School 308

7 Fundamentals of computer organisation and architecture

Information CPU time

CPU Time is the amount of time it takes the CPU to execute a particular
Listed below in order of
program. CPU time is a function of the number of instructions in the program,
popularity, as of 2015, are some
the clock cycle time and average CPI:
of the most common ISAs (most
popular first) : CPU time = instruction count x CPI x clock cycle time
• ARM CPU time can be reduced by reducing any or all of the quantities on the right-
• IA-32 (Intel® x86) hand side of the above equation.
• Intel® 64 (Intel® x86-64 )
Instruction count can be reduced by
• IA-64 (Intel® Itanium® )
• inspecting the compiled code and replacing sections of it with code
that uses fewer instructions, written directly in assembly language by
Information hand
Clock speed or rate: • redesigning the compiler to produce fewer machine code instructions
Clock speed can be adjusted in for a given program, i.e. better optimisation.
the BIOS.
To reduce CPI and clock cycle time we must focus on the processor (CPU)
itself.
Questions
2 What affects the amount of time it takes a CPU to execute a particular program?

How can we improve processor performance?

CPI and clock cycle time are related to how the processor operates. To improve
its performance we need to reduce the
Information
• average CPI by redesigning the processor, using multiple cores,
Tools for exploring the
increasing memory bandwidth (number of bits transferred per second),
hardware of a computer
system:
or pre-fetching data and instructions and storing these in fast access
Speccy®: www.piriform.com/ memory (cache) located on processor chip.
speccy
• clock cycle time by clocking the processor at a higher rate.
CPUID: www.cpuid.com /
softwares/cpu-z.html Multiple cores
Arithmetic instructions are executed using the Arithmetic and Logic Unit
(ALU). If the number of ALUs is increased from one to four then a single
arithmetic instruction can use all four ALUs at the same time. For this to be
possible,
Key concept
1. The data must lend itself to being divided into four streams, one per
Core: ALU.
A processing unit consisting of
ALU + Control unit + Registers 2. Four cores (ALU + Control Unit + Registers [+ Cache]) are required
within a CPU. This means that all the arithmetic instructions to which this applies for a given
A CPU or processor with just
program can be executed in a quarter of the time. Single Instruction Multiple
one core is called a single-core
CPU. Data (SIMD) stream processing, as it is known, requires special control units
A CPU with more than one core to decode and execute instructions that are to be executed in parallel.
is called a multi-core CPU, e.g.
Data can also be pre-fetched at the same time the processors are busy decoding
a quad-core processor has four
cores. and executing arithmetic instructions. The pre-fetched data is stored in fast to

309 Single licence - Abingdon School

 7.3.7 Factors affecting processor performance

access memory on the processor chip so as not to hold up the

ARM® Cortex®-A7
processor.
Figure 7.3.7.2 shows a schematic for an ARM® Cortex®-A7 NEON Data
Engine
quad-core 32-bit processor based on ARM’s licensed v7-A ARMv7 32b
CPU Floating
instruction set architecture (ISA). It has four CPUs or cores Point Unit

labelled 1, 2, 3 and 4. Each core has its own data cache as 16-64k 16-64k Core
I-Cache D-Cache
well as an instruction cache enabling instructions to be pre- 1 2 3
fetched as well. SIMD operations for handling audio and 4
video processing as well as graphics and gaming processing
SCU L2 Cache
rely on a special control unit called the NEON Data Engine.
128-bit AMBA® ACE Coherent Bus Interface
Floating point operations take considerably longer than fixed
point and integer operations (fixed point data can be treated Figure 7.3.7.2 ARM® Cortex®-A7
and processed as integers). The A9 processor also includes a quad-core processor
dedicated Floating Point Unit specially designed to allow the
CPU to offload floating point operations to this unit.
Figure 7.3.7.3 shows the Parallella computer platform which is an energy efficient, high performance, credit card
sized computer based on the Epiphany multicore chips from Adapteva®. This desktop version cost £150 and is
used for developing and implementing high performance, parallel processing. It uses a Zynq dual-core ARM A9
processor to launch and run programs which use the 16-core Epiphany coprocessor for parts of the program that
can be executed in parallel. The Epiphany coprocessor is also available with 64 cores.The Parallela’s Ethernet
Zynq dual-core ARM A9 processor connection allows multiple
Gigabit
with Field Programmable Gate units to be interconnected to
Ethernet
Array (FPGA) make a cluster.

16-core Epiphany
coprocessor

Questions
3 Explain how a multi-core CPU
can improve the performance
of the CPU when executing a
program.

MESH NODE

1GB SDRAM RISC CPU

DMA
ENGINE

Memory Local Network

Memory Interface

Figure 7.3.7.3 Desktop Parallela platform

Router

16 cores

Single licence - Abingdon School 310

7 Fundamentals of computer organisation and architecture

Bus width effect on processor performance

Main memory or RAM is controlled by a circuit called a memory controller which is a part of the CPU. The
memory controller is connected to main memory by a memory bus as shown in Figure 7.3.7.4.
8192 MBytes
Processor
In modern computer systems every byte in memory has
Cache Slot 2 its own address. The data bus part of the memory bus
Control
is typically 64, 128, 192 or 256 bits wide on a modern
Memory
controller
Address Slot 1 general purpose Intel or AMD CPU. 64 bits means that
Data 8192 MBytes when 64 bits are transferred along the data bus, these 64
bits have come from 8 memory addresses.
Memory bus DIMM
Hypertransport Memory Modern computer systems use synchronous dynamic
bus modules
or similar random access memory (SDRAM) that is dynamic
random access memory (DRAM) synchronized with the
Figure 7.3.7.4 Memory bus
system/memory bus.
DDR3-1600 memory used in the computer on
which this book was written stands for double data
rate type three synchronous dynamic random-access
memory. It is operated at a clock rate or clock speed
of approximately 800MHz but because data is
transferred on both the rising edge of the clock signal
and the falling edge, twice as much data is transferred
per clock cycle - Figure 7.3.7.5.
Each transfer consists of 64 bits, the width of
the data bus (single channel). The memory data
bus must also operate at the same frequency as
memory because it has to be synchronised with the
DDR memory.
A processor should take less time executing a
program if more data can be transferred each time
memory is accessed. A wider data bus allows more
data to be transferred in one go. However, the
speed at which the transfer takes place is also a factor.
Figure 7.3.7.5 Details of RAM in author’s computer
This is why both data bus width and memory bus
clock rate must be taken into account.
Extension material: Memory bandwidth and effect on processor performance
Memory bandwidth is the rate at which data can be read from or stored in main memory by a processor.
Memory bandwidth is usually expressed in units of bytes per second.
Memory bandwidth = memory clock rate x bits transferred per clock cycle / 8
For example, bits transferred per clock cycle = 64 bits x 2 x no of channels
(Multiple by 2 when Double Rate Dynamic RAM (DDR3) is used)
No of channels = 2, clock rate = 800MHz (8 x 106),
Memory bandwidth = 800 x 106 x 64 x 2 x 2/8 = 256 x 108 bytes per second = 25.6 Gigabytes per second
= 25.6 GB/s
311 Single licence - Abingdon School
 7.3.7 Factors affecting processor performance

This is a theoretical maximum because memory doesn’t respond immediately Information

to a read or write request. This is called latency, in particular, Column Access Bus width cannot be increased
Strobe (CAS) latency - Figure 7.3.7.5. It is the delay time between the moment indefinitely because as more
a memory controller tells the memory module to access a particular memory lines are added, it becomes more
difficult to keep all the bits in step.
column on a RAM module, and the moment the data from the given location
A problem known as skew.
is available on the module’s output pins. To overcome this latency, which can The solution is to reduce the
be as much as 11 clock cycles or more (see Figure 7.3.7.5), cache memory is number of bus lines to 32 or
employed. less but compensate by clocking
these lines at a much higher
Cache memory is faster to access than main memory because
rate. The bus interfaces at either
(a) its technology is different from main memory technology end take care of aggregating bits
into the required size, e.g. 128,
(b) the bus speed of the bus that accesses cache memory is much higher before passing these on. The
than the memory bus speed connecting main memory to the CPU. Hypertransport bus is one way
that this is done. This is a packet
Questions based serial bus.
4 Describe the effect on memory bandwidth of increasing Just as USB has largely replaced
the Centronics parallel
(a) the width of the data bus from 64 bits to 128 bits
communication interface, so
(b) the clock rate that memory uses from 666.6MHz to 800MHz.
Hypertransport and similar
5 Explain why the transfer speed of bits along the data bus is not the systems are replacing the
traditional physical parallel
only factor that determines the time taken to transfer data between
computer bus.
processor and main memory.

The effect of cache on processor performance

CPU cache is memory on the CPU chip used by the central
Main memory (RAM)
processing unit (CPU) of a computer to reduce the average time to CPU
access data from main memory.
L2 cache
The cache is a small amount of fast but expensive memory which
stores copies of the data from frequently used main memory L1 Instruction
L1 Data cache
locations, data to be written to main memory and pre-fetched cache
Data
instructions. flow
Fetch unit
When the processor attempts to read a word of main memory, a Control Registers
flows
check is made first to determine if the word is in the cache. If it is, Decode unit Data
flow
a copy of the word is transferred to the processor. This is a much Data
flow
faster operation than accessing main memory. If not, a block of Execute unit ALU
main memory, consisting of a fixed number of words, is transferred
L1 Data cache Data
into the cache and then a copy of the referenced word is transferred flow
to the processor.
Main memory (RAM)
Similarly, when the processor needs to write to main memory it will
write to the cache instead which is a much faster operation than Figure 7.3.7.6 Cache hierarchy and role in
writing directly to main memory. Fetch-Execute cycle
Figure 7.3.7.6 shows the cache hierarchy and its role in the Fetch-
Execute cycle. L1 cache has near zero latency but only a limited amount is
provided because it is expensive. The L2 memory cache is cheaper to make than
Single licence - Abingdon School 312
7 Fundamentals of computer organisation and architecture

L1 cache but is slower to respond than the L1 cache,

and therefore it has some latency but still much less than
main memory. Some systems use an additional layer of
cache between main memory and L2 cache called L3
cache with latency greater than L2 cache but still less
than main memory. More L3 cache is provided than L1
and L2 cache because it is cheaper than these.

Figure 7.3.7.7 shows the L1, L2 and L3 caches

incorporated into AMD’s FX-6300 CPU and their
memory sizes. Note that there are two L1 caches, one for
data and one for instructions.

Figure 7.3.7.7 Cache present in AMD® FX-6300 CPU

Key concept Questions

Cache memory: 6 Explain how cache memory may be used to improve the performance
A small amount of faster of a processor.
memory than main memory,
that stores copies of the data
from frequently used main The effect of word length on processor performance
memory locations, data to be
The instruction set architecture of a processor is designed to work with registers
written to main memory and
pre-fetched instructions. L1 of a given word length (number of bits). In a 32-bit processor, the registers
and L2 cache are usually located are 32 bits in length, in a 64-bit processor, 64 bits. A machine code instruction
on the CPU chip. Slower L3 will manipulate 32 bits at a time in a 32-bit processor and the unit of transfer
cache is often located on the between processor and main memory will also be 32 bits. In each case, 32 bits
motherboard.
are presented for manipulation or transfer in 32-bit long registers.
The cache memory approach
relies for its effectiveness on the If we need to work with 64 bits but are restricted to using a 32-bit processor
fact that when a block of data is then we have to use more 32-bit machine code instructions to accomplish
fetched into the cache to satisfy
the same task than would be the case if we could use 64-bit machine code
a single memory reference, it is
likely that future references will instructions. A program compiled for a 32-bit machine is thus likely to have
be to other words in the block. more instructions to execute than the same program compiled for a 64-bit
processor other things being equal. More instructions to execute means more
CPU time.

Questions
7 Explain the effect on processor performance of processor word length.

313 Single licence - Abingdon School

 7.3.7 Factors affecting processor performance

The effect of address bus width on processor performance

The language of digital computers is binary. Addresses of
Processor No of bytes that can be addressed
memory words are no different and are also expressed at
8-bit 256
the machine level in binary. Addresses can and are treated
16-bit 65,536
as data to be manipulated. The address bus width tends
32-bit 4,294,967,296
therefore to mirror the word length of the processor and
64-bit 18,446,744,073,709,551,616
therefore its registers. Although 8-bit processors have tended
Table 7.3.7.1 Effect on no of bytes that can be
to be different by using an address bus width of 16 lines.
addressed when address bus width is increased
More address lines means more bytes can be addressed.
Table 7.3.7.1 shows how the number of bytes that a processor can address Questions
increases with word length of processor. At present 64-bit processors are
designed with a lower figure for the width of the address bus of 48 bits as 8 Explain the effect on
18,446,744,073,709,551,616 bytes is of the order of petabytes for 64-bit wide processor performance
address bus. Memory of this capacity would cost a lot of money! of address bus width.
Summary
To improve the performance of a processor timewise, CPU time needs to be
reduced so that programs take less time to execute. CPU time is defined as
follows: CPU time = instruction count x CPI x clock cycle time
To reduce CPU time the following need to be reduced

1. Instruction count
• influenced by the design of the compiler or whether sections of code
have been rewritten in assembly language to use fewer instructions.
2. CPI (cycles per instruction) and clock cycle time
• influenced by the design of the processor and its operation.
CPI can be reduced by
• increasing the word length of the processor
• redesigning its control unit to take less time decoding and executing
instructions
• using multiple cores
• increasing memory bandwidth (number of bits transferred per second)
ŠŠ by increasing the width of the data bus for a given clock rate
ŠŠ by clocking the memory bus at a higher rate for a given width of
data bus
• pre-fetching data and instructions and storing these in fast access
memory (cache) located on processor chip.
Clock cycle time can be reduced by increasing the clock speed and clocking
the processor at a higher rate.
To improve the performance of a processor regarding the number of memory
words it can address, the width of the address bus needs to be increased.

Single licence - Abingdon School 314

7 Fundamentals of computer organisation and architecture

In this chapter you have covered:

■■ Explanations of the effect on processor performance of:
• multiple cores
• cache memory
• clock speed
• word length
• address bus width
• data bus width

315 Single licence - Abingdon School

7 Fundamentals of computer organisation and architecture

7.4 External hardware devices

Learning objectives:
■■Know the main characteristics,
purposes and suitability of the
■■ 7.4.1 Input and output devices
devices and understand their Barcode reader
principles of operation A barcode reader, or barcode scanner, is an electronic device for reading
barcodes printed on items such as cans, packaging, and the covers of books or
• barcode reader magazines. A barcode is a sequence of white and black bars (Figure 7.4.1.1)
• digital camera that encodes information such as a product identifier. The product identifier is
usually printed in human-readable form beneath the barcode.
• laser printer
• RFID
Information
Barcode symbols:
A combination of several bars
that make up an individual
character or digit is often called
a symbol. The set of symbols
Figure 7.4.1.1 Barcode encoding the characters 978-0-9927536-2-7
available for a specific barcode A barcode reader consists of a light source (low-powered laser diode), a lens,
standard is referred to as its photoelectric detectors (photodiodes) and decoder circuitry to analyse the
symbology. All these different barcode’s image data and generate character codes. The scanner uses the light
symbologies can be read with a
source to illuminate the black and white bands. More light is reflected from a
laser beam.
white band than from a dark band. The pattern of reflection is converted from
optical form to electrical form by photoelectric detectors in the barcode reader.
Key concept The electrical form of the reflection data is
One-dimensional barcode: analysed and the barcode is decoded into
Barcodes are said to have one character form. The path of a red laser beam
dimension if there’s a single as it moves over the barcode is shown in
line (such as a line traced by a
Figure 7.4.1.2. The relative time the beam
scanner’s laser) that can cross all
lines of the symbol. spends scanning dark bars and light spaces
Figure 7.4.1.2 Laser beam scan
which encode a character is measured and
of barcode
a lookup table is then used to translate this
Did you know?
time into the corresponding character.
Universal Product Code (UPC):
The scanner outputs the character codes, e.g. ASCII codes, as a sequence of
The first-ever product carrying
a UPC code in its packaging binary digits for processing by a computer.
was scanned June 26th 1974. It The line of the laser beam shown in Figure 7.4.1.2 is the reason why barcodes
was a 10-pack of chewing gum,
that are scanned in this way are known as one-dimensional barcodes.
now on display at Smithsonian
Institution’s National Museum A major advantage of one-dimensional barcodes is that they can be decoded
of American History in
very reliably even when the items tagged with such barcodes are moving at high
Washington, D.C.
speed. They are also relatively cheap to use because the technology has been
Single licence - Abingdon School 316
7 Fundamentals of computer organisation and architecture

Information around for 40 years and the necessary components of laser diode and decoding
QR code: electronics have benefited from high volume of use which has led to economies
Refers to a single member of of scale in their manufacture.
the family of 2D barcodes 2D
barcodes are machine-readable
Information
codes that use markings forming The GS1 organization (www.gs1.org) maintains the standards related to the Global Trade
a two-dimensional grid. The Item Number (GTIN). There are several symbologies that belong to this family, all of them
example below is a QR code that representing a product code and all using the same kind of barcode symbols:
represents the text Unit 2 CS. UPC-A - The first product barcode (12 digits) now refered to as GTIN-12.
EAN-13 - European barcode (13 digits) now refered to GTIN-13.

Questions
1 What is a barcode?

2 Explain the operation of a barcode reader designed to read one-

dimensional barcodes.
A CCD camera is used to “see” 3 Why is the information encoded in barcode form also printed in
the squares that make up a 2D
human-readable form beneath the barcode?
barcode.
4 The automated luggage handling system at Heathrow airport reads
barcodes on labels attached to passengers’ luggage to route the luggage
to the correct conveyor belt among 30 miles of conveyors. The barcodes
encode flight and passenger information. Why are barcodes suitable for
this application?

Digital camera
Light reflected from objects is focussed by the lens of a digital camera onto
Figure 7.4.1.3 CCD SONY a two-dimensional array of light-sensitive cells (photosensors) to form an
ICX493AQA 10.14 Mpixels analogue image. Each cell or site accumulates an electric charge proportional
APS-C 1.8” (23.98 x 16.41mm) to the brightness of the illumination and as the latter varies in a continuous
sensor side manner so does charge accumulation. Both are analogue quantities.
To process the analogue image digitally, the magnitude of the charge in each
Information photosensor is sensed and converted into digital format by an analogue to
digital converter (ADC).
Digital single-lens reflex
cameras: The two-dimensional array (matrix) of photosensors and associated electronics
These use an aspect ratio of 3:2. to perform analogue to digital conversion is one of two types:
Aspect ratio is the ratio of the
width of the image to its height.
• Charge-Coupled Device (CCD)
The Canon EOS 600D (released • Complementary Metal-Oxide Semiconductor (CMOS)
February 2011) uses an APS-C
CMOS sensor consisting of a
Both are fabricated from metal oxide semiconductors with photodiodes used
sensor array of dimensions 5184 x as photosensors. A single chip may contain millions of photodiodes laid out in
3456 pixels. rows and columns forming a matrix.

317 Single licence - Abingdon School

 7.4.1 Input and output devices

Figure 7.4.1.4(a) illustrates with a 6 x 6 matrix how coloured images can be Information
sensed in a CCD-type camera sensor. Bayer filter:
Red, green and blue filters cover the A Bayer filter mosaic is a colour
One pixel filter array for arranging RGB
photodiodes. Each pixel is composed
colour filters on a square grid of
of one red, one blue and two green photosensors (U.S. Patent No.
filtered photodiodes reflecting the fact 3,971,065). developed by Kodak
that the human eye is more sensitive employee Bruce Bayer in 1976.

to green light. This filter arrangement Kodak produced the world’s

first digital camera but the
is known as a Bayer filter mosaic.
company went bankrupt because
Each row of photodiodes shares an it continued to prioritise its film
ADC which is located to the right of manufacturing business.

Figure 7.4.1.4(a) CCD-type Bayer each row. The accumulated charges in

filter sensor matrix for capturing a row are shifted rightwards one cell Information
coloured images at a time, converted into an analogue
DCIM (Digital Camera
voltage and then into a digitally
IMages):
equivalent voltage by the ADC. DCIM is the default directory
structure for digital cameras.
With some additional processing to take into account the filters, the whole
When you put a memory card
image is converted to an equivalent pixel-based digital one (see Chapter
into a camera, the camera
5.6.1.1). This is called a raw format digital image, e.g. Canon cameras save a immediately looks for a ‘DCIM’
raw format image in a file with extension CR2. folder. If it doesn’t find such a
folder, it creates one.
CR2 files use a format based on the TIFF specification. These files are extremely
Similarly, some desktop image-
high quality, and are the very best when it comes to editing. CR2 files can be editing programs are designed
converted into JPEG format once the need to adjust the RAW image file is no to look specifically for ‘DCIM’
longer needed. folders on any media inserted into
the PC.
In a CMOS sensor, each site in the matrix contains a photodiode plus some
transistors to do some of the pre-processing as well as to allow each photosite to
Key fact
be independently accessed. It does however mean a smaller photodiode.
CCD vs CMOS:
Figure 7.4.1.4(b) shows with a CCDs consume as much as
black square the area occupied 100 times more power than an
by a photodiode at a photosite equivalent CMOS sensor.
compared with the area occupied CMOS sensors are cheaper to
manufacture than CCD.
by a photodiode in the equivalent
The quality of manufacture of
CCD sensor. The rest of a CMOS CCD is higher than CMOS
photosite is occupied by transistors. because CCD has been around
Light falling on these does not get for a lot longer (since 1975).

captured which means the site is CCD sensors have greater light
sensitivity and produce images
less sensitive to light. It also leads
which are less noisy. Noise
Figure 7.4.1.4(b) CMOS-type sensor to images with more noise than produces grainy images.

photodiodes overlaid onto a CCD-type with the equivalent CCD-produced CCD sensors have a greater
image. resolution than CMOS sensors
sensor matrix for capturing coloured
because their packing density is
images higher.

Single licence - Abingdon School 318

7 Fundamentals of computer organisation and architecture

Key fact Questions

Colour laser printer:
5 Explain the operation of a digital camera capable of taking colour
A colour laser printer uses four
toners, Cyan, Magenta,Yellow, and
pictures.
Black, 6 State one situation where a digital camera with a CCD sensor would
Colour laser printers use the
be used and one situation where a CMOS sensor would be used.
CMYK colour model, Cyan,
Magenta,Yellow, and Key (Black), Explain your choice in each case.
This is a subtractive colour model,

Laser printer
A laser printer prints a whole page at a time. It prints high-quality text and
graphics on plain paper.
A page description language usually Key fact
describes the page to be printed as lines, Monochrome laser printer:
arcs and polygons. A processor in a A monochrome laser printer uses
CMY “primaries” are combined a single toner containing a black
monochrome laser printer generates a
at full strength, the resulting powder. It is only able to print text
bitmap of the page in raster memory from
“secondary” mixtures are red, in black.
green, and blue. Mixing all three the page description. A negative charge They are poor at printing greyscale
gives black. However, for a darker is applied to the photosensitive drum images which has to be done by
black the Key (Black) toner is at the heart of the laser printer. One or printing small dots and varying the
used. spacing and arrangement of those
more laser beams are directed onto the
dots. This is called dithering or
rotating drum’s surface (Figure 7.4.1.5).
halftoning.
Key fact The lasers are turned on or off at positions
Colour laser printer: determined by the bitmap data stored in
In a color laser printer, each of the raster memory. This causes the negative charge to be neutralised or reversed
the four CMYK toner layers is at positions corresponding to the black parts
stored as a separate bitmap.
of the page to be printed.
The resulting pattern of charges on the
drum’s surface is an image of the page to be
printed. The charged surface of the drum
is exposed to toner, fine particles of dry
plastic powder mixed with carbon black or
colouring agents. The charged toner particles
are given a negative charge so they attach to
the uncharged or positively charged regions
of the drum and not to the negatively
charged regions. Darker areas are achieved by
depositing thicker layers of toner. A higher
voltage applied to the gap between toner
cartridge and drum surface forces more toner
Figure 7.4.1.5 Schematic of the operation of a laser printer onto the drum. The raster memory stores the
Reproduced with permission from Computer Desktop greyscale data for each area of the page and
Encyclopedia, www.computerlanguage.com this is used to set the appropriate voltage level.

319 Single licence - Abingdon School

 7.4.1 Input and output devices

By rolling and pressing the rotating drum over a sheet of paper, the toner Key concept
is transferred onto the paper. Transfer may be assisted by using a positively
charged transfer roller on the back of the paper to pull the toner from the Radio frequency identification
(RFID):
surface of the drum to the paper. The paper is passed through heated rollers
Any method of identifying
that squeeze the paper and fuse the toner to the paper. and tracking items using
radio waves. Typically a reader
Questions (also called an interrogator)
7 Why is a laser printer described as a page printer? communicates with a
transponder, which holds digital
8 Explain the operation of a monochrome laser printer. information in a microchip.
Alternatively, a chipless RFID
9 What toners are used by a colour laser printer and why?
tag is used which just uses
material to reflect back a portion
RFID of the radio waves beamed at
them.
Radio frequency identification (RFID) uses radio frequencies (RF) to transmit
data, a timing signal and radio frequency energy if necessary between a reader
(Figure 7.4.1.6(b)) and an RFID device (transponder) as shown in Figure
Information
7.4.1.6(a). RFID tag:
RFID Device A microchip attached to an
Data antenna which is packaged in a
Clock Contactless way that allows it to be applied
RFID
data carrier
Reader Energy to an object such as an item of
(transponder)
clothing for sale. The tag picks up
signals from and sends signals to
a reader.
Coupling element Tags come in many forms such as
(microwave antenna) smart labels which have a barcode
printed on them and tags simply
Figure 7.4.1.6(a) RFID reader and transponder embedded in plastic ready for
attaching to an object.

RFID devices do not need a physical electrical contact to transfer data. Nor do Smart labels are manufactured to
be fed through an RFID printer
they need visible contact as they use radio waves to transfer data. This makes it
that both prints a bar code on
possible to read many codes simultaneously from afar without the need to open the label and writes to its RFID
boxes. microchip.

An RFID system has a transponder and a reader. The RFID transponder is

located on the object to be identified. The reader, or interrogator, may be able
to read data or to read and write data, but it is always called a reader.
The RFID transponder can be powered by RF energy from the reader if the
transponder is a passive RFID device or it may use an internal battery if it is an
active or semi-passive device.
The transponder has a small RF antenna and circuitry for transmitting and
receiving data.
Figure 7.4.1.6(b) RFID reader
Active RFID tags can continuously broadcast their own signal which is useful
when they are used to track the real-time location of an object but they are

Single licence - Abingdon School 320

7 Fundamentals of computer organisation and architecture

bulkier than passive RFID tags. Active tags provide a much

longer read range than passive tags, but they are also much
more expensive.
Passive RFID tags have applications including access control,
smart labels, race timing, and more.
Passive RFID are much cheaper to make than active RFID
tags as well as being much less bulky.
The data capacity of RFID transponders is normally a few
Figure 7.4.1.7(a) RFID price smart tag bytes to several thousand bytes, but a transponder with a
data capacity of just 1 bit can distinguish between
transponder present and transponder not present.
Most 1-bit transponders are used in electronic article
surveillance (EAS) systems to protect goods in shops
and businesses. They are removed or deactivated at
the till when the goods are paid for. A reader installed
at the shop’s exit raises an alarm if goods are removed
before the transponder has been deactivated.
RFID smart cards, such as Transport for London’s
Oyster card, are used as tickets for journeys on public
Figure 7.4.1.7(b) RFID price smart tag located on
transport.
underside of price tag
RFID devices are attached to products to respond
Key fact with a unique code when interrogated. This means each item can be recognized
RFID tag characteristics: individually instead of just recognising it as belonging to a particular product
RFID tags can include a unique type. A part of the unique code could be used to identify the product type, e.g.
code, which makes every RFID
12345-1, 12345-2 where 12345 is product code for blue t-shirt ,size small.
label individual, meaning that
each item can be recognized Marks and Spencer uses such tags in its stores and warehouses for stock control
individually instead of just purposes. For privacy reasons, this RFID tag is not read at the sales counter. If
recognizing it as belonging to a
it were then the customer could be tracked when they leave the store by linking
particular product type.
RFID devices do not need a
their identity, captured at the sales counter, to the unique serial number in the
physical electrical contact to RFID tag attached to the purchased item.
transfer data. Nor do they need
However, the time staff spend stock checking is greatly reduced because RFID
visible contact as they use radio
tags on store items can be read quickly with a handheld reader from a distance,
waves to transfer data. This makes
it possible to read many codes using a barcode scanner would take considerably longer.
simultaneously from afar without
Figure 7.4.1.7(a) shows an item of clothing’s price tag with an RFID
the need to open boxes.
transponder (Figure 7.4.1.7(b)) on its underside. The RFID transponder
responds with the item’s unique code or serial number when read by an RFID
reader from as far away as 70cm.
They can be used as contactless security badges to give access to protected
premises. Electronic immobilisers for cars use RFID; the ignition key is
combined with a transponder. An RFID device placed under an animal’s skin
can be used for tracking and identification (one use is a cat flap which can be
321 Single licence - Abingdon School
 7.4.1 Input and output devices

opened by cats with a recognised RFID tag, but not by other cats to stop these
entering the house). RFID devices are put in the stomach of cattle and remain
there for life.
Barcode scanning versus RFID scanning
RFID scanning allows many RFID tags to be read at a time even when the tags
are hidden inside boxes or behind panels. Barcode scanning scans one item at a
time and the barcode needs to be visible to the scanner.
When stocktaking it is important to avoid counting an item more than
once. RFID tagging and scanning enables each item to be uniquely tagged
and therefore uniquely identified when scanned. When combined with a
timestamp it is very easy to avoid counting an item more than once in a
stocktaking session. Barcodes do not usually identify an item uniquely but
instead encode the product type of an item. Timestamping would be of little
help in preventing an item being counted twice because items are not uniquely
identified.

Questions
10 What is Radio Frequency Identification (RFID)?

11 What is an RFID tag?

12 Give three uses of RFID tags.
13 Give two advantages of RFID scanning over barcode scanning.

In this chapter you have covered:

■■ The main characteristics, purposes and suitability of the following devices
and understand their principles of operation
• barcode reader
• digital camera
• laser printer
• RFID

Single licence - Abingdon School 322

7 Fundamentals of computer organisation and architecture

7.4 External hardware devices

Learning objectives:
■■Explain the need for secondary
storage within a computer system ■■ 7.4.2 Secondary storage devices
■■Know the main characteristics, Why do we need secondary storage?
purposes, suitability and The technology that primary storage (RAM) is built from and which supports
understand the principles of read and write random access to individual words requires a continuous
operation of the following supply of electrical energy in order to work. Unfortunately, when the supply
devices: of electrical energy is removed the information stored in the memory words
• hard disk is lost. We say that read/write main memory is volatile (analogous to liquids
• optical disk which disappear by the process of evaporation). To retain information and
• solid-state disk (SSD) programs after electrical power is removed requires a different form of storage,
■■Compare the capacity and one which is non-volatile. There are three technologies with which such
speed of access of various media storage is built currently:
and make judgement about
1. Magnetic
their suitability for different
applications 2. Optical

Did you know? 3. Solid-state

IBM obtained the technology If we want to retain a program we have created in RAM, or some information
for making magnetic disks from we have written to RAM, then we must transfer both to a non-volatile
Manchester University where a
secondary store. The commonest form of read/write secondary store is a
one kilobyte magnetic disk had
been made on a one metre-wide magnetic hard disk encased in a magnetic hard disk drive (HDD) - Figure
platter. 7.4.2.1. A newer form of read/write secondary store that is now shipping in
desktop PCs, laptops and tablets is a solid-state disk (SSD). Compact
Disc (CD) and Digital Versatile Disc (DVD) storage are optical media
that can be used for secondary storage. There are read only (CD-R,
DVD-R), write once read many times (CD-R, DVD-R) and read/write
Platter rotation
versions (CD-RW, DVD-RW) of these.
Questions
1 Why is secondary storage needed?
Read-write
Read-write head
head radial Magnetic hard disk
Spindle
movement IBM developed magnetic disk drives in the late 1950s. The disk drive
connected
allows rapid random (direct) access to large amounts of data. All disk
to motor
drives use a thin circular platter made of non-ferrous metal or plastic
Platter
which is rotated at up to 10,000 revolutions per minute beneath a
read-write head that moves radially across the surface of the platter.
Figure 7.4.2.1 shows a 20 GB hard disk drive with the cover removed.
Figure 7.4.2.1 Hard disk drive with The platter and read-write head can be clearly seen as well as the
cover removed photographer’s reflection in the platter.
Single licence - Abingdon School 323
7 Fundamentals of computer organisation and architecture

Key concept The platter is coated with an emulsion of iron or cobalt oxide (or a cobalt-based
alloy) particles that act as tiny magnets. Binary data is recorded by aligning
Track: these tiny magnets in one direction to represent a binary 0 and in the opposite
One of the concentric rings on a
direction to represent a binary 1. Binary data is recorded in concentric rings, or
platter of a hard disk.
Sector: tracks, subdivided into sectors that hold a fixed number of bytes, such as 512. A
A subdivision of a track. hard disk can store and retrieve a large volume of data.
Disk block:
To read data stored on the hard disk, the read-write head moves to the
The smallest unit of transfer
between a computer and a disk. desired track and waits for the relevant sector to pass beneath it. When data is
A disk block is one sector of a transferred from the hard disk to the computer and vice versa, a whole sector
track. of a track is read or written each time. A whole sector of a track is often called
a disk block or a block. For this reason, a magnetic hard disk drive is known as
Did you know? a block-oriented storage device. The smallest unit of transfer is a block which is
In 2015, the fastest rotation typically 512 bytes.
speeds of consumer disk drives The top and bottom surfaces of a platter may be used to store data. A block
was 10000 revolutions per
address for a single-platter system is composed of a surface address, a track
minute.
address and a sector address. Typically, the surfaces are numbered 0 and 1, the
Key point tracks 0 to 7,000 and the sectors 0 to 63. Figure 7.4.2.2 shows a schematic for
one surface of a magnetic hard disk.
Disk buffer:
Executing programs do not write Stepper
motor
directly to magnetic hard disks.
Instead, they write to an area Read-write
of main memory (RAM) called head
a disk buffer. Before a program
can write to a file, it has first
Sector Track
to open the file, if it exists, or
create the file if it doesn’t. This
open/create action creates a disk
buffer in main memory (RAM)
which is then associated with the
Disk block
file. The program writes to this (one sector
buffer. When the buffer becomes of a track)
full or the program closes the
corresponding file, the operating
system writes the buffer to disk.
The size of the buffer matches the N- Platter
S
size of a disk block or a multiple S-N
N-S
of this. S-N
To read a file it must first be Part of one
opened. This creates a disk buffer sector
which receives a block at a time Magnetic particle of a track
each encoding a 1 or 0
belonging to the file. The program
that opened the file then reads Figure 7.4.2.2 Hard disk platter showing concentric tracks and sectors
from this buffer. When the buffer
becomes empty, the operating Modern hard disks for a PC system are sealed units, called Winchester disks,
system transfers the next disk
containing several platters mounted on a common spindle. The platters are
block belonging to the file into
this buffer. sealed inside an assembly which allows the disk to operate with minimal risk of

324 Single licence - Abingdon School

 7.4.2 Secondary storage devices

damage from contaminants. The read-write heads are built into the assembly Key concept
with one head per surface. The greater the number of platters, the greater the Access time:
storage capacity. Access time for a magnetic hard
disk is the time interval between
Questions the moment the command is
given to transfer data from disk to
2 Why is a magnetic disk drive known as a block-oriented storage device?
main memory and the moment
3 In the context of a magnetic disk, what is (a) a track, (b) a sector, this transfer is completed.
It is made up of three
(c) a disk block and (d) disk block address?
components:
4 Explain the principle of operation of a magnetic disk drive. 1. seek time
2. rotational delay
5 What effect do you think that having smaller platters and faster 3. data transfer time
rotation speeds will have on the time taken to read a disk block?
Seek time:
The time it takes for the read/
Optical disc write heads to align on the desired
track.
An optical disc is a flat, usually circular disc which encodes binary data (bits)
Rotational delay:
in a special reflective layer. In one form of optical disc, binary data is encoded The time taken for the desired
in the form of pits (binary value of 0 due to lack of reflection when read) and sector to come under the read/
lands (binary value of 1 due to a reflection when read) on a reflective material, write head. On average, this is the
usually metallic, on one of its flat surfaces as shown in Figure 7.4.2.3. time taken for half a revolution of
the disk platters. This average is
CD-ROM called the latency of the disk.
The success of compact discs (CDs) for storing audio led to a new format, CD Data transfer time:
Read-Only Memory (CD-ROM). Introduced early in 1985, this format was The total time taken to read the
disk block and transfer it into
initially used to publish encyclopedias, reference works, professional directories
main memory
and other large databases. CD-ROMs were ideal for this because they had (for
the time) a high storage capacity of
600–700 million bytes, offered fast Magnified view
showing pits in
data access and were portable, rugged reflective metal
layer
and read-only. Today, CD-ROMs are
also used for software distribution.
The data is written on the discs
Label Protective
using disc-mastering machinery that
layer
impresses pits (physical depressions)
into a continuous spiral track. The
Reflective
silvery data surface contains pits in a 1.2 mm metal layer
thick
single track 3.5 miles (5.6 km) long.
The disc spins at 200-500 revolutions
per minute depending on which part
Polycarbonate
of the track is being read. disc

A data bit is read by focusing a laser Laser beam

beam onto a point in the reflective

metal layer where the pits are
Figure 7.4.2.3 CD-ROM cross-section through its layers
impressed (Figure 7.4.2.3).

Single licence - Abingdon School 325

7 Fundamentals of computer organisation and architecture

Key fact More laser light is reflected from the unpitted surface than from the
Optical disc: pitted surface. This is detected by a photodiode that outputs an equivalent
An optical disc is a flat, usually circular electrical signal. After some conditioning, the result is a digital signal
disc which encodes binary data (bits) in representing a single data bit. This data bit encodes the amount of
a special reflective layer. In one form of
reflection as 0 or 1.
optical disc, binary data is encoded in the
form of pits (binary value of 0 due to lack Did you know? Questions
of reflection when read) and lands (binary Blue-ray disc:
6 What is an optical disc?
value of 1 due to a reflection when read) A Blu-ray disc (BD) is a high-
on a reflective material, usually metallic, on density optical disc capable of 7 Explain how information is
one of its flat surfaces. storing 23.3 GiB (25 GB) in a recorded and then read from a
single-layer which is consider-
CD-ROM.
ably more than a DVD can
store.
Did you know?
CD-R:
Write Once, Read Many (WORM) times Solid-state disk (SSD)
optical disc. The solid-state disk (SSD) in a solid-state disk drive (Figure 7.4.2.4)
CD-R can record about 650 - 900 MiB operates by trapping electrons in a wafer of semiconducting material.
of data
These electrons and their electric charge remain trapped even when
CD-RW:
CD-ReWritable disc that can be read and electric power is removed, i.e. SSD is non-volatile storage. Binary 0 is
written to over and over again. represented by trapped electrons and binary 1 by absence of trapped
DVD-ROM: electrons.
Digital versatile disc or digital video disc
(DVD) is an optical standard offering The sites
much greater storage capacity than CDs. (floating gate
Storage capacity of a single-layer DVD- transistors)
ROM is 4.3 GiB (4.7 GB). where these
DVD-R:
electrons are
DVD-R is a WORM format similar to
trapped are
CD-R.
DVD-RW: organized in a
The DVD-RW format provides a grid. The entire
rewritable optical disc with a typical grid layout is Figure 7.4.2.4 Solid-state disk drive
capacity of 4.3 GiB (4.7 GB).
referred to as a ©D-Kuru/Wikimedia Commons
DVD+RW:
A competing rewritable format to
block, while the
DVR-RW. individual rows that make up the grid are called a page.
DVD-RAM:
Common page sizes are 2KiB, 4KiB, 8KiB, or 16KiB, with 128 to 256
DVD-RAM is a rewritable format
that has built-in error control and a
pages per block. Block sizes are typically between 256KiB and 4MiB. For
defect management system, so it is example, the Samsung™ SSD 840 EVO has blocks of size 2MiB, and
considered to be better than the other each block contains 256 pages of 8 KiB each. The Samsung SSD 840
DVD technologies for tasks such as data EVO comprises 8 NAND flash chips, each of capacity 64 GiB. Each
storage, backup and archiving.
Samsung flash chip contains 32 blocks.
Blu-ray disc:
A Blu-ray disc (BD) is a high-density Unlike magnetic disk drives, solid-state drives contain no moving parts
optical disc capable of storing 23.3 or spinning disks. The absence of moving parts means that solid state
GiB (25 GB) in a single-layer which is
disk drives can operate at speeds far above those of a typical hard disk
considerably more than a DVD can store.
drive.

326 Single licence - Abingdon School

 7.4.2 Secondary storage devices

Access time for a typical hard drive is on average 10-15 milliseconds whereas access time for an SSD drive is 25-100
microseconds (access time for RAM is typically 40 -100 nanoseconds).
The technology used is NAND flash memory which is a type of EEPROM (Electrically Erasable Programmable
Read Only Memory). A solid-state disk is a block-oriented storage device which has to erase a block first in order
to rewrite it because unlike magnetic hard disk drives, NAND flash memory can’t overwrite existing data. Erasing a
block in the SDD means “untrapping” electrons.
The solid-state disk drive requires an onboard controller which consists of an embedded microprocessor with RAM
buffer to perform reading and writing to the solid state disk (Figure 7.4.2.5). The controller is a very important
factor in determining the speed of the SSD drive.
SSD SSD
Controller PCB

SATA NAND
Interface Flash
Memory

Figure 7.4.2.5 SSD drive printed circuit board (PCB) showing the controller and the NAND flash memory
chips (image reproduced with kind permission of StorageReview.com). Information
To alter the contents of a particular memory location of SSD storage, an entire SSD vs other flash-based
devices:
block must be constructed containing the new information and written to SSD.
SSDs are much faster than any
The controller arranges for this new block to be written to a different area of SSD. of the other flash-based portable
The reason for this is that SSD blocks can be programmed for only a limited drives, e.g. USB memory stick.
amount of time before they become unreliable. This is known as write-endurance.
It is measured in number of program erase (P/E) cycles. To lessen this effect, a
controller uses a technique called wear-levelling, which effectively makes sure that all the drive’s memory chips are
used, cell by cell, before the first cell is written on again.
SSD secondary storage is increasingly being used in laptops, tablets and is an option now for desktop PCs. The
attraction is lower power consumption and faster booting of the operating system.

Questions
8 Explain the operation of a solid-state disk drive.

9 Give four reasons why a solid-state disk drive might be preferred to a magnetic disk drive.

Single licence - Abingdon School 327

7 Fundamentals of computer organisation and architecture

Capacity and speed of access

Hard disk drive manufacturers specify disk capacity using the SI prefixes mega Information
(106), giga (109), and tera (1012), abbreviated to M, G and T, respectively. Byte is IOPS (Input/Output
abbreviated to B. CD and DVD capacities are quoted in a similar fashion. Operations Per Second):
This is the unit for random access.
In the case of both magnetic hard disks and solid-state disks speed of access is Random access is the ability to
determined by whether access is random or sequential. Sequential operations access (read/write) data at an
access locations on the storage device in a contiguous manner and are generally arbitrary location, in a non-
contiguous manner. It is a unit
associated with large data transfer sizes, e.g., 128 KiB. Random operations access
used for storage devices such as
locations on the storage device in a non-contiguous manner and are generally hard disk and solid-state disks.
associated with small data transfer sizes, e.g. 4 KiB. Table 7.4.2.1 quotes read and
write speeds for random operations for these devices using a unit, IOPS (Input/
output Operations Per Second). For a magnetic hard disk drive (HDD), the random IOPS is primarily dependent
upon the HDD’s random seek time. For a solid-state disk drive, the random IOPS is primarily dependent upon its
internal controller and memory interface speed.

Storage medium Capacity Read speed Write speed

Magnetic hard disk 20 GB to 10 TB Random I/O Random I/O operations:
(SI units: powers of operations: 75 - 200 IOPS
10, G = 109, T = 1012) 75 -200 IOPS

CD-ROM, CD-R, CD-RW 650 - 900 MiB CD-ROM 150 CD-R 1.76 MiB/s (12x)
KiB/s (1x) to CD-RW 1.46 MiB/s
6,750–10,800 KiB/s (10x)
(72x)
DVD-ROM, DVD-R, DVD±RW, single layer: 4.3 GiB 1,353 KiB/s (1x) to 1,353 KiB/s (1x) to
DVD-RAM dual layer: 8.5 GiB 21,648 KiB/s (16x) 21,648 KiB/s (16x)

Blu-ray Disc (BD-ROM) single layer: 23.3 GiB 4.3 MiB/s (1x) to BD-R / BD-RE 50 MiB/s
Blu-ray Disc Recordable (BD-R) dual layer: 46.6 GiB 68.66 MiB/s (16x) (12x)

Blu-ray Disc Recordable (BD-RE) XL x3 layer: 93.2 GiB

XL x4 layer: 119.2 GiB
Solid-state disk 128 MB to 4 GB Random I/O Random I/O operations:
operations: 80,000 IOPS

150,000 IOPS
Table 7.4.2.1 Capacities and speed of access for various storage media

Information
CD drives:
1x means 150 KiB/s

DVD drives:
1x means 1,353 KiB/s

328 Single licence - Abingdon School

 7.4.2 Secondary storage devices

Storage medium Applications Did you know?

Magnetic hard disk Online storage of programs and data files
Backing up data means
CD-ROM Distributing software
taking a copy of data and
Distributing software, storing photographs, backing storing it somewhere safe,
CD-R
up data, archiving data e.g. in a fireproof safe or off-
CD-RW Backing up data, transferring files site. Archiving data means
DVD-ROM Distributing software or videos removing it from the online
storage medium, usually to
Distributing software or videos, storing photographs,
DVD-R free up space. Data qualifies
backing up data, archiving data for archiving if it has not been
DVD±RW Backing up data, transferring files accessed recently and will not be
DVD-RAM Backing up data accessed regularly in the future.
Blu-ray Distributing high definition videos and video games Programs and data may be
Solid-state disk Online storage of programs and data files backed up and archived.

Table 7.4.2.2 Storage media and typical applications to which they can be put

Questions
10 State what applications each of the following might be used for:
(a) magnetic hard disk drive (b) CD-ROM (c) DVD-R
(d) DVD-RAM (e) blu-ray disc.

11 What storage medium would be most suitable for distributing a 5 GiB

file in each of the following cases:
(a) a single individual requiring read-only access
(b) a single individual who will need write access
(c) a large number of people requiring read-only access?

In this chapter you have covered:

■■ the need for secondary storage within a computer system
■■ the main characteristics, purposes, suitability and the principles of
operation of the following devices:
• hard disk
• optical disk
• solid-state disk (SSD)
■■ the capacity and speed of access of various media and
their suitability for different applications

Single licence - Abingdon School 329

8 Consequences of uses of computing
8.1 Individual (moral), social (ethical), legal and cultural issues and opportunities
Learning objectives:
■■Show awareness of
current individual (moral),
■■ 8.1 Introduction
social(ethical), legal and When we talk about the morality of human actions we are referring to
cultural issues and whether the actions are right or wrong. A moral action is one that is right.
opportunities and risks of An immoral action is one that is wrong. Morality is primarily about making
computing correct choices between right and wrong or between good behaviour and bad
behaviour.
■■Understand that:
When an individual reaches a conclusion or decision as to the morally right
• developments in computer course of action they often draw on a framework or set of principles to help
science and the digital
their reasoning, e.g. our actions should do no harm. The framework or set
technologies have
of principles is called an ethical theory or just, ethics. When applied to
dramatically altered the
particular cases, the framework can provide clear choices. This can be helpful
shape of communications
because often there isn’t one right answer for a situation that requires a person
and information flows in
to act morally - there may be several right answers, or just some least worst
societies, enabling massive
answers - and the individual must take responsibility and choose between them.
transformations in the
When a person ‘thinks ethically’ they are giving some thought to human action
capacities to
that has moral consequences for someone beyond themselves and their own
▫▫monitor behaviour desires and self-interest. When new possibilities for human action arise as is
▫▫amass and analyse the case with computer technology, human beings face new ethical questions,
personal information i.e. questions requiring new ethics with which to reason.

▫▫distribute, publish, Case study

communicate and The case of the Crown versus Dudley and Stephens in 1884 established
disseminate personal a precedent, throughout the common law world, that necessity is not a
information defence to a charge of murder. The sailing ship Mignonette foundered in
• computer scientists and a storm on July 5th 1884 whilst on route to Southampton from Sydney.
software engineers therefore The crew of four, Tom Dudley, the captain, Edwin Stephens, Edmund
have power, as well as the Brooks and Richard Parker, aged 17, the cabin boy, took to a lifeboat.
responsibilities that go with On the nineteenth day of their ordeal, Dudley suggested drawing lots to
it, in the algorithms that determine who would die so that the others might live. Brooks refused
they devise and the code and so no lots were drawn. By this time, Parker, who had drunk sea water,
that they deploy appeared to be dying. On the next day and still with no ship in sight that
could rescue them, Dudley took out a penknife and whilst saying a prayer
• software and their over him, stabbed Parker in the jugular vein. For four days, the three men
algorithms embed moral fed on the body and blood of Parker.
and cultural values ...continued on next page

Single licence - Abingdon School 330

8 Consequences of uses of computing

Learning objectives:
Case study continued
■■Understand that: After their rescue on July 29th and arrest after returning to England
• the issue of scale, for Dudley and Stephens went to trial and Brooks turned state’s witness.
software the whole world Defending killing Parker before his natural death they argued that his
over, creates potential blood would be better preserved for them to drink. They freely confessed
for individual computer to killing and consuming Parker but claimed they had done so out of
scientists and software necessity. Parker was an orphan without dependents whilst Dudley and
engineers to produce great Stephens had families that depended upon them.
good, but with it comes the At the time maritime law was not clear cut on such cases and in such
ability to cause great harm circumstances the custom of the sea applied as practised by the officers
and crew of ships and boats in the open sea.
■■Be able to discuss the
challenges facing legislators in
the digital age

Questions for discussion

1 Suppose you were the judge at their trial. Putting aside the
question of maritime law and assuming that you were making a
moral judgement, how would you rule and what argument would
you use to justify your ruling?

(The actual ruling is available from https://en.wikipedia.org/wiki/

R_v_Dudley_and_Stephens)

The case of the Crown versus Dudley and Stephens in 1884 was important for
two reasons. Firstly, it established a precedent, throughout the common law
world, that necessity is not a defence to a charge of murder. Secondly and more
importantly, it highlighted that morality is more than a matter of cost-benefit
analysis and calculating consequences but has more to do with the proper way
for human beings to treat one another. Morality implies certain moral duties
and human rights so fundamental that they rise above a matter of simply
calculating consequences.

Questions for discussion

2 Suppose you were asked to draw up a list of fundamental moral
duties and human rights. What would be on your list? You could
use a search engine to research the general consensus on these
but try yourself or discuss the question with others first. Bear in
mind that your generation, the digital native one, may disagree
with what you find on the World Wide Web which could reflect
what has been framed by an older generation labelled digital
immigrants.

331 Single licence - Abingdon School

 8.1 Introduction

The Global Information Society and a new Digital Ethics Key principles
Society is shaped by the technology at its disposal:
Protection of personal data:
“The handmill gives you society with the feudal lord; the steam-mill the society EU Charter of Fundamental
with the industrial capitalist.” Karl Marx Rights Article 1 defines the
right of an individual to
The information society that we live in today has been shaped by the protection of their personal data
information and communication technologies of the global networking of the via data protection principles
digital computer. the following principles
-necessity, proportionality,
“The digital computer and networking give you the Global Information fairness, data minimisation,
Society” purpose limitation, consent and
transparency.
Whilst previous societies have evolved relatively slowly, the Global Information
These apply to data processing
Society has burst into existence in a relatively short period of time by in its entirety, to collection as
comparison. Berkeley’s School of Information Management and Systems well as to use.
estimated that humanity had accumulated approximately 12 exabytes of data
in the course of its entire history before the advent of the desktop digital Key concept
computer. Since then there has been an explosion in data. For example, in just Personal data:
Data about any living person.
one year, 2002, more than 5 exabytes of data were produced and recorded.
According to IDC, in 2013 the digital universe contained some 4.4 zettabytes, Task
a thousand-fold increase on 2002, roughly. Whilst bringing enormous benefits Data Protection Act
and opportunities, the information revolution’s enormous growth rate has 1. Look at the eight principles
posed a conceptual, ethical and cultural challenge of how to put in place a of the Data Protection Act
1998 at https://ico.org.uk/
viable philosophy and ethics of information.
for-organisations/guide-to-data-
The European Data Protection Supervisor (EDPS), an independent institution protection/ and see if you can
of the EU, published a report in 2015 entitled match these to the EU Charter
principles.
“Towards a New Digital Ethics”
2. Look at Regulation (EU)
(https://secure.edps.europa.eu/EDPSWEB/webdav/site/mySite/shared/ 2016/679 which applies to all 28
Documents/Consultation/Opinions/2015/15-09-11_Data_Ethics_EN.pdf ) EU countries from 25 May 2018
and Directive (EU) 2016/680
At its core is the protection of human dignity: which requires member states
“An ethical framework needs to underpin the building blocks of this digital to enact legislation by 6 May
2018.
ecosystem. The EDPS considers that better respect for, and the safeguarding
of, human dignity could be the counterweight to the pervasive surveillance and Key concept
asymmetry of power which now confronts the individual. It should be at the Asymmetry of power:
Computers do not only carry
heart of a new digital ethics.”
out actions using information
Asymmetry of power and pervasive surveillance in the form of programmed
Asymmetry of power instructions but they can also
produce information that
Coded algorithms automate operations or actions according to a logic that
coded algorithms can capture
differs little from that which humans have applied for centuries. Replacing the and report. However, whilst
human brain by a digital computer enables more consistency, greater speed you have some control over
of calculation and more control. However, it adds an additional dimension the former in that you have
that is not possible with a human agent: computers do not only carry out initiated the action, e.g. a
search, you seemingly have
actions using information in the form of programmed instructions but they
much less control over the
can also produce information that coded algorithms can capture and report. latter.

Single licence - Abingdon School 332

8 Consequences of uses of computing

Key concept “Information Technology alone has this capacity to both automate and reflect
information (informate)” - Professor Shoshana Zuboff. However, whilst you
Machine learning: have some control over the former in that you have initiated the action, e.g.
Machine learning is the
a search, you seemingly have much less control over the latter (e.g. collecting,
science of getting computers
to act without being explicitly storing and associating with you information about what you are searching for)
programmed. This approach and certainly many people do not perceive that the latter is happening behind
enables patterns in and links the scenes. This is what is meant by asymmetry of power.
between data to be discovered
autonomously once the Visiting a particular website results in your visit being logged, tracked with
machine has been “trained” on a cookies and then combined with information already held about you to send
large enough dataset. Examples you personalised advertisements. You can see what is happening behind the
are speech recognition and
scenes using packet capture software such as Wireshark. For example, visiting
language translation software,
the New York Times’ website results in the communication of information to
and self-driving cars.
the following other sites (as of October 2015):
https://www.doubleclickbygoogle.com/
Key concept http://www.conversantmedia.com/
Information: http://www.facebook.com
Data is how information is
Pervasive surveillance
represented.
The logic in automation is the logic of action, i.e. do this followed by do that,
whereas the logic of information reflection is the logic of accumulation.
Key concept The latter has given rise to huge datasets labelled Big Data (see Chapter 11.1)
and the impetus to analyse large datasets using machine learning techniques.
Knowledge:
Knowledge is usable Contributing to this Big Data phenomena in a major way is the capturing of
information. small data from individuals’ online actions and utterances as they go about
their daily lives. Nothing is too trivial or unimportant for this data harvesting,
from Facebook ‘likes’ and clicks on links, to smartphone location data.
Key concept
Such data (representation of information) is aggregated, analysed, packaged and
Pervasive surveillance: sold.
Capturing of small data from
These data flows have been labelled ‘data exhaust’ because they are the
individuals’ online actions and
utterances as they go about their byproduct of users’ actions as they go about their daily business interacting
daily lives. These data flows have with computer systems. Google and Facebook are among the largest and most
been labelled ‘data exhaust’. successful Big Data companies because they sweep up this data exhaust. In
2015, Google’s search engine was the most visited engine and Facebook the
most visited social media site.
Although Google started out with no intention of offering advertisers space on
their search results’ Web pages, they eventually gave in to the need to generate
revenue via an advertising model rather than a fees-for-service one because the
latter might have impacted on the expansion of their user base. This advertising
approach depended upon the acquisition of personal data as the raw material
that after analysis and application of machine learning would sell and target
advertising through a unique auction model reliant upon the accumulation of
huge quantities of personal data to make it work with increasing precision and
success.

333 Single licence - Abingdon School

 8.1 Introduction

Google’s business is thus the auction business and its customers are advertisers. Information
AdWords, Google’s algorithmic auction method for selling online advertising,
What They Know, is an in-depth
analyses massive amounts of data to determine which advertisers get which one
investigative series by the Wall
of eleven sponsored links on each search results page (http://archive.wired.com/
Street Journal. It found that one
culture/culturereviews/magazine/17-06/nep_googlenomics?currentPage=all). of the fastest growing Internet
business models is of data-
Case study gatherers engaged in intensive
surveillance of people visiting
In 2009 Google published a research paper in the prestigious scientific websites in order to sell data
journal Nature that described a method to track influenza-like illness in about, and predictions of, their
a population by analyzing large numbers of Google search queries. The interests and activities, in real
researchers reported that they could accurately estimate the current level time.
of weekly influenza activity in each region of the United States, with a http://juliaangwin.com/the-what-
reporting lag of about one day. they-know-series/
(http://static.googleusercontent.com/media/research.google.com/en//
archive/papers/detecting-influenza-epidemics.pdf )
“This seems like a really clever way of using data that is created Information
unintentionally by the users of Google to see patterns in the world that
would otherwise be invisible,” said Thomas W. Malone, a professor at the Using Big Data to track
Sloan School of Management at MIT. population movements in an
Shortly after its publication, H1N1, a new strain of a particularly virulent Ebola infected country

strain of flu hit the United States. In order to track and contain the http://www.pri.org/
stories/2014-10-24/how-big-data-
outbreak before it became pandemic, the Centers for Disease Control and
could-help-stop-spread-ebola
Prevention (CDC) requested that doctors inform them of new flu cases.
Unfortunately, this reporting at best involved a two-week lag, too long to
enable effective control of the outbreak.

Questions for discussion

3 “What companies know about me from my behavior online cannot hurt me. In fact, it is more likely
to benefit me.”
4 “What information I give out about myself is a fair tradeoff for benefits that I receive.”
5 “Surrendering personal data for perceived benefits is not a square deal because I cannot control all
the ways that my personal data will be used.”
6 “I am powerless to stop my personal data being used for purposes that I am unaware of because the
services that I use for free in exchange for surrendering my data are an integral part of my life.”
7 “I am just amazed at all the innovative services, e.g. Google translate, DropBox, GoogleDocs, Apps
or education, YouTube, etc, that I can get for free by interacting online. It doesn’t bother me that my
personal data and activities online can be used by the data scientists and data mining experts to
create the weird and wonderful algorithms behind all these services.”
8 Google’s use of user-generated data to support public health efforts in significant ways is a benefit to
society and therefore outweighs the concerns raised by two privacy organisations that Google could
be compelled by court order to release sensitive user-specific information - the Electronic Privacy
Information Center (https://www.epic.org) and the US Patient Privacy Rights organisation (https://
patientprivacyrights.org).

Single licence - Abingdon School 334

8 Consequences of uses of computing

Case study: “From Forbes.com,

Case study - adapted with permission from Forbes.com
16/02/2012 © 2012 Forbes LLC. All
rights reserved. Used by permission and The chief data scientist, Andrew Pole, of a US retail chain, Target,
protected by the Copyright Laws of the successfully figured out how to answer the question, “If we wanted to
United States. The printing, copying, know if a customer is pregnant, even if she didn’t want us to know, is it
redistribution, or retransmission of possible? ”, by analysing data that the company collected on customers’
this Content without express written spending habits.
permission is prohibited.” http://www. As Pole’s computers crawled through the collected personal data, he
forbes.com/sites/kashmirhill/2012/02/16/ was able to identify about 25 products that, when analyzed together,
how-target-figured-out-a-teen-girl-was- allowed him to assign each shopper a “pregnancy prediction” score.
pregnant-before-her-father-did/ More importantly, he could also estimate her due date to within a small
window, so Target could send coupons timed to very specific stages of
Key principle her pregnancy. The sending of such coupons did result in an unexpected
outcome when a Target store was visited by an angry man to complain
Linking of information: that Target was sending his teenage daughter coupons for baby clothes
Two pieces of information
and cribs even though she was certainly not pregnant and was still at high
about a person might
individually be harmless but less school.
so if linked. The manager was bemused and could only apologise profusely and then
again a few days later by phone.
For example:
On the phone, though, the father was somewhat apologetic. “I had a talk
1. Tony Blair is a former British with my daughter,” he said. “It turns out there’s been some activities in my
Prime Minister
house I haven’t been completely aware of. She’s due in August. I owe you
2. Tony Blair has a home in
London an apology.”
Google’s street view map of Target changed their mailing policy to mask the fact that they knew a lot
Tony Blair’s London home more about their customers than their customers realised. So they started
is now blurred. Google are
mixing coupons for other things with the baby item coupons so that the
required now to offer to blur
properties visible in Street View
pregnant woman did not think that she had been spied on.
after a European court ruling in
favour of people’s ‘right to be Questions for discussion
forgotten’.
Revisit questions 3 to 6 shown again below, to see if you would
1. Dick Cheney was US respond in the same way after reading the case study above.
Secretary of Defense during
3 “What companies know about me from my behaviour online
Operation Desert Storm, the
1991 invasion of Iraq cannot hurt me. In fact, it is more likely to benefit me.”
2. Dick Cheney has been
4 “What information I give out about myself is a fair tradeoff for
fitted with a heart pacemaker,
recently. benefits that I receive.”

The pacemaker was specially 5 “Surrendering personal data for perceived benefits is not a square
adapted for Dick Cheney so deal because I cannot control all the ways that my personal data
that it would be resistant to
will be used.”
hacking and disruption. This
must be very reassuring to the
6 “I am powerless to stop my personal data being used for purposes
rest of the population which has
that I am unaware of because the services that I use for free in
to make do with pacemakers
that are vulnerable to hacking. exchange for surrendering my data are an integral part of my life.”

335 Single licence - Abingdon School

 8.1 Introduction

The issue of scale Information

Google is the pioneer of hyperscale, the ability at relatively low cost to scale
processing quickly across thousands of commodity computers housed in data Hyperscale computing:
centres. Other hyperscale businesses such as Facebook, Twitter, Alibaba, Baidu, In computing, hyperscale is the
Amazon, and Yahoo also possess this ability. Smaller firms without hyperscale ability of an architecture to
scale appropriately and quickly
revenues can leverage some of these capabilities by using a cloud facility such as
in a cost-effective manner as
Amazon’s Elastic Cloud facility. increased demand is added to the
Having this hyperscale ability enables many results to be extracted from system. Hyperscale computing
is necessary in order to build
individuals’ personal data that would have remained unknown but for the
a robust and scalable cloud, big
scaling of the processing it makes possible, as well as the support for massive data, map reduce, or distributed
datasets of personal information it grants. The extracted information can be storage system and is often
of benefit to society but it also has the potential for misuse if the processing associated with the infrastructure
is used for social or economic discrimination, unsolicited advertising, or required to run large distributed
sites such as Facebook, Google,
reputational damage.
Microsoft Azure or Amazon
This hyperscale ability also provides the capacity for sharing information AWS.
amongst individuals on a global scale through social media sites, tweets, online
blogs, etc., from which individuals derive much benefit of a social nature. Key fact
All of this has been made possible by coded algorithms devised and deployed
by computer scientists and software engineers. However, the free access and Memories for life:
facilities provided by Internet companies to enable this sharing comes at a cost The ability to record memories,
and store them indefinitely in
which may be difficult for the individual to assess or for that matter for anyone
digital form in virtually unlimited
to know how the information surrendered freely will be used in the future.
quantities has been dubbed the
“The Web means the end of forgetting” Jeffrey Rosen phenomenon of memories for life.
Photographs and documents
YouTube demonstrates that people’s eccentric behaviour can be distributed featuring you may turn up in
around the world without their knowledge or control as they may not know other people’s memory banks.
their behaviour was captured in a video or that the video has been uploaded “The Spy in the Coffee Machine”
to YouTube. Whilst access to lots of information is made possible by the © Kieron O’Hara and Nigel
Shadbolt 2008, reproduced with
hyperscale reach provided through the Internet, it also provides access to
permission of the publishers
aspects of people’s lives that were formally private. Sometimes people do Oneworld Publications.
knowingly give away or surrender their privacy, but it is also the case that
sometimes they are innocent victims because of circumstances beyond their
control. It is very difficult to protect privacy in such cases.

Questions for discussion Information

Topic: Practical obscurity
Practical obscurity is an important factor in the preservation of privacy. If The Web means the end of
forgetting:
the representation of information does not permit it to be easily queried,
New York Times article by Jeffrey
e.g. the information is on paper in a filing cabinet, then the extraction of Rosen
important knowledge (usable information) is made more difficult. http://www.nytimes.
com/2010/07/25/
9 Why does the ability to collect and process data on a mammoth magazine/25privacy-t2.
scale in the way achieved by Google and other hyperscale companies html?pagewanted=all&_r=0
reduce practical obscurity?

Single licence - Abingdon School 336

8 Consequences of uses of computing

The challenges facing legislators in the digital age

Key concept There is a general feeling that a person’s privacy has shrunk in the global
Cookie: information society. There are something like five hundred companies that are
A cookie is the standard way able to track every move you make on the Internet, mining the raw material of
that a website uses to track
the Web and selling it to marketers.
its visitors to the site. They are
little pieces of data harmless “Personal data are purchased, aggregated, analyzed, packaged, and sold by
in themselves that are used data brokers who operate, in the US at least, in secrecy – outside of statutory
to inform a website that a
consumer protections and without consumers’ knowledge, consent, or rights
particular visitor to the website
has returned. This is generally
of privacy and due process” (U.S. Committee on Commerce, Science, and
seen as a positive thing. Transportation, 2013).
A benefit is that cookies allow The law often places constraints on what computer scientists and software
e-commerce sites to maintain
engineers are allowed to do, but equally the nature of software, data, and
a virtual shopping basket for
the visitor between visits. information, and the degree and scale of control over software available to this
However, some cookies collect group constrain what the lawyers and legislators can achieve when local laws
data across many websites, run up against the global Internet.
creating ‘behavioural profiles’
of people. These profiles can “In today’s digital environment, adherence to the law is not enough; we have to
then be used to decide what consider the ethical dimension of data processing.”
content or adverts to show
(Towards a New Digital Ethics)
you. This use of cookies for
targeting in particular is what This is the case whether the right under scrutiny is any one of copyright,
recent changes in UK law were trademark, privacy, or freedom of expression. Can a law made in one country
designed to address by requiring
be successfully applied to the global Internet whose content, algorithms and
websites to inform and obtain
consent from visitors for the access embed value judgments from different cultures, societies and legal
use of cookies. The law’s aim is systems?
to give web users more control
over their online privacy.
Case study
Guardian article: Right to be forgotten: Swiss cheese Internet, or
database of ruin?
Read this Guardian article by Julia Powles at
http://www.theguardian.com/technology/2015/aug/01/right-to-be-
forgotten-google-swiss-cheese-internet-database-of-ruin

Task
Questions for discussion
Cookie Law:
Find out about Cookie law which
10 Do you think that one country should have the authority to
was adopted by all EU countries
control what content someone in another country can access on
in May 2011 and required an
update in the UK to the Privacy the Internet? Justify your opinion.
and Electronic Communications
Regulations. 11 In respect of the Internet, should the reach of the law for each of
the following apply (i) globally or (ii) locally with each country
deciding what law to apply

(a) copyright (b) trademark (c) privacy (d) freedom of expression

337 Single licence - Abingdon School

 8.1 Introduction

Software and their algorithms embed moral and cultural values Background
Any artifact that interacts with human beings and that is able to change the
In Egypt on January 25th, 2011
dynamics of social processes is not value free. a Facebook page was posted
Social processes are the ways in which individuals and groups interact, adjust calling for a day of protest after
an incident in which a 28 year-
and re-adjust and establish relationships and patterns of behaviour.
old man was dragged from an
To be value free means to be without bias or to use criteria that do not reflect Internet cafe and beaten to death
prejudice or cultural attitudes. by Egyptian police officers. When
a large number of people signed
Cultural attitudes reflect the culture of a society and are expressed through the up for the protest on Facebook
ideas, customs, and social behaviour that are held to define the society. British the Egyptian government shut
culture, for example, means to act fairly and justly; respect for the right of free down Egypt’s Internet and cellular

expression; respect for the rule of law and the democratic process; respect for phone service. Three weeks
later, on February 11th, Egyptian
a free press (free from Government control); the right to protest in an orderly
President Hosni Mubarak was
manner; and much more. Other cultures place a different emphasis on free driven from office after 30 years
speech, etc. in power.

Email was the first killer application built on top of the Internet. It changed
a major social process dramatically, i.e. the way that we communicate, but
Background
if you were not digitally connected then you could not use it. Although this The built-environment:
is less true today there are sectors of the world’s population who are unable The environment that consists of
buildings, roads and railways.
to communicate via email because they do not have access to the necessary
resources.
Case study
The architecture of a place functions as a form of regulation; it constrains the behaviour of those who interact
with it, often without them even realizing it. Build a bridge so low that buses cannot pass under it then it is
possible to exclude people from travelling to what lies beyond the bridge, if economic circumstances mean that
buses are their only means of travel. If these circumstances are associated with lack of education and therefore
employment opportunities then it becomes a form of segregation, all the more worse if the lack of education
opportunities and employment are linked to ethnicity.
The architecture of a place is the way it is because of how the built-environment was designed. The design is
not value free if it has been based on criteria which are subjective, i.e. not objective. If the design was objective
then the object of connecting one side of the bridge to the other would have applied equally to both buses and
cars.

Questions for discussion

12 Think of two other ways in which the design of the architecture of the built-environment constrains access
to certain social groups?

13 Generally speaking, securing privacy by controlling access is an inalienable right. After all privacy is a
human right. What is morally wrong then with applying design to public places in a way that affects some
peoples’ access but protects other peoples’ right to privacy?

14 Can you imagine a situation where the designers set out with the intention of creating fair access but are
prevented from doing so by circumstances beyond their control?

Single licence - Abingdon School 338

8 Consequences of uses of computing

From your discussions you may have concluded that the architecture of the
Background
built-environment can be said to embed moral and cultural values, i.e. it is
The algorithms of the Internet not value free.
allow data about people to be
collected in the following ways: This is also true of the Internet, the World Wide Web and other applications
• Volunteered data – created and
built on top of the Internet such as social media because each of these has
explicitly shared by individuals, been designed. Each has a particular architecture shaped by its designers in its
e.g. social network profiles. hardware, software, and its algorithms, none of which can be value free because
• Observed data – captured design decisions are always taken which inevitably embed moral and cultural
by recording the actions of
values. Thus the design and operation of the Internet, the World Wide Web
individuals, e.g. location data
when using cell phones.
and other applications that rely on the Internet must also raise questions of
• Inferred data – data about justice, fairness and democracy.
individuals based on analysis
This matters even more because there are major differences between the built
of volunteered or observed
information, e.g., credit scores.
environment and the architecture of the Internet, World Wide Web and these
other applications.
• The first is one of scale, the Internet’s reach is global whereas that of
public space is local.
• Secondly, the Internet, World Wide Web and other Internet-based
applications embody not only the logic of action, i.e. automation,
but also the logic of information reflection or accumulation which
facilitates a form of control.
• Thirdly, the distinction between hardware and software is blurred
because cloud computing effectively delivers physical hardware
packaged as software i.e. a virtual machine running the operating
system of your choice and virtual storage such as DropBox. Users
can gain access in minutes to virtualised, scalable hardware resources
(e.g., Amazon’s Elastic Cloud) which obviates the need to purchase
the equivalent physical hardware. As there is no physical hardware
Information to dismantle, terminating or halting such a provision also takes only
Cloud service providers: minutes.
Amazon Web Services -
Cloud computing also decouples the physical possession of data from their
https://aws.amazon.com
ownership. It means dealing now with the issues of ownership of virtual assets
Google Cloud -
https://cloud.google.com and access to those virtual assets, i.e. the right to ownership and the right
Microsoft Azure: to usage. For example, if your personal data is aggregated and subjected to
https://azure.microsoft.com machine learning algorithms that derive new information about you, who owns
this new information and who has the right to access this information?

339 Single licence - Abingdon School

 8.1 Introduction

Case study
There are a number of large gateways into China through which Internet access for Chinese citizens
is controlled. The Chinese telecom companies that control these gateways are required by the Chinese
government to configure their routers to use DNS servers that screen and filter out content that the Chinese
government objects to because, according to Western opinion, of a fear of losing control to forces other than
the Chinese government. While the primary purpose of routers is to direct or route Internet traffic to its
correct destination, they can also be configured in this way to block content and thereby prevent information
from getting to its destination. Routers can also be configured to block access to websites by their URL, e.g.
Twitter.com and forbidden web page content by inspecting the packets of information, any content that
mentions Tiananmen Square. The router, DNS server, and the software they use become a censor in these
circumstances.

Questions for discussion

Information
15 (a) Do you think that the world at large has an inalienable right to
Example:
influence public opinion in China? URL to use to explore how China
(b) What would be your motives if you do believe this? controls its citizens access to the
Internet:
16 What should be the limits of freedom of expression on social http://www.howtogeek.
media sites? com/162092/htg-explains-how-
the-great-firewall-of-china-works/
17 Is it wrong for a search engine to return a list of web pages
according to a profile that they have built up about you? Why?

18 Should governments make policies to govern access to certain

Task
web sites?
Read
(1)
Software can produce great good but with it comes the ability to cause great
http://www.wired.com/2015/07/
harm hackers-remotely-kill-jeep-
It is not always enough to encrypt communications over the Internet because highway/
the packets that carry the encrypted communication also carry tracking (2)
data in plaintext form, i.e. the source and destination IP addresses, e.g. your http://www.theguardian.com/
technology/2014/jun/29/
computer’s IP address and the IP address of the website that you are visiting.
facebook-users-emotions-news-
These IP addresses must be machine readable for the Internet to function and feeds
route packets successfully. Internet Service Providers can sell this tracking
data to marketers and in some countries are required to keep and reveal this
information to the authorities on demand. The packets can also be examined in
transit by packet sniffing software snooping in on the communication. Background
Preserving privacy means not only hiding the content of messages, but also
How Tor works:
hiding who is talking to whom. The Tor project and the Tor software that it https://www.torproject.org/about/
developed has made this tracking much more difficult as well as providing overview
encryption of the content of messages. As a software tool for anonymous and
confidential communication, it has been used successfully by journalists to

Single licence - Abingdon School 340

8 Consequences of uses of computing

Background communicate more safely with whistleblowers and dissidents. It has proved
to be an effective tool to circumvent censorship of the Internet in countries
Deep Web DVD: such as China and Iran and as a building block for other software designed to
Now available to purchase from
protect privacy.
www.amazon.com.
URL: http://www.amazon.com/ However, the Tor software has also been used to hide criminal activity and
Deep-Web-Directors-Keanu- other undesirable activities. A question mark also hangs over whether Tor is
Reeves/dp/B017WUEJ52
secure against monitoring by governments as it was originally developed, built
and financed by the US military which released it for general use in 2004,
ostensibly to improve the cover of its spies overseas by masking their Internet
activities amongst the activities of a diverse group of people to whom the Tor
Information software was now available. If it is not secure against government monitoring,
Banning use of free or shared it might be difficult to know because the authorities would be careful about
WiFi and Tor: revealing their hand.
In December 2015 in the wake of
the Paris attacks, it was reported
that France’s law enforcement
Case study
authorities were proposing new
The Dark Web is the World Wide Web content that exists on
legislation to forbid the use of
free or shared WiFi during a
underground networks which use the public Internet but which require
state of emergency. They were specific software such as Tor for access. The Dark Web forms part of
also proposing that anonymous the Deep Web, the part of the Web not indexed by search engines. In
browsers like Tor should be the Deep Web there are several online shopping sites that specialise in
blocked in general.
connecting sellers of illicit goods with willing buyers.
The online black market Silk Road launched in February 2011 used
the anonymising tool Tor to protect the identities of buyers, sellers
and the site’s administrators. Payment was made in Bitcoin, allowing
buyers a relatively high amount of protection. Ross William Ulbricht, its
Background
creator, was arrested in October 2013. On May 29, 2015, Ulbricht was
Bitcoin: handed five sentences to be served concurrently, including two for life
Bitcoin is a decentralized digital
imprisonment, without the possibility of parole. He was also ordered to
currency or virtual currency. It
uses peer-to-peer technology to forfeit $183 million obtained from his criminal activities that he held in
operate with no central authority Bitcoins, beyond the reach of the authorities until he was forced to hand
or banks. over the encryption keys. In a letter to Judge Forrest before his sentencing,
Ulbricht stated that his actions through Silk Road were committed
through libertarian idealism and that “Silk Road was supposed to be about
giving people the freedom to make their own choices” and admitted that
he made a “terrible mistake” that “ruined his life”.

Watch the film Deep Web by Alex Winter, if you can, which explores how
the designers of the Deep Web and Bitcoin are at the centre of a battle
for control of how the Internet may be used and its effect on our digital
rights.

341 Single licence - Abingdon School

 8.1 Introduction

Questions for discussion

19 In 2014 Google launched a service in response to a European Court ruling to allow Europeans to ask
for personal data to be removed from online search results.
(a) Why might removal of personal data from search engine results not be sufficient to comply with
the right to be forgotten?
(b) If on the grounds of freedom of speech it is a human right to be able to access information,
why might forcing people to conduct searches with engines that are able to access the Deep Web be
undesirable?

20 “Software such as Tor should be banned because although it can be used for morally sound purposes
it can also be used for criminal and immoral acts.” Do you agree or disagree with this statement and
why?

Questions

1 What is meant by personalised search results?

2 What is meant by personalised advertisements?

3 Some search engines return search results to fit the profile of the person making the query. State two
dangers to individuals and society of providing such personalised search.

4 Social media and search engine companies collect personal data from users to make their services more
useful to users. Explain why it can be beneficial for the user to give up personal data in this way and
why it might not.

5 Software can produce great good but with it comes the ability to cause great harm. Using at least one
example, explain the meaning of this statement.

6 Software and their algorithms embed moral and cultural values. Using at least one example, explain
the meaning of this statement.

7 Why do legislators face difficulty enacting legislation where the Internet is concerned?

8 Developments in computer science and digital technology have dramatically altered the shape of
information flows in society. The labels asymmetry of power and pervasive surveillance have been applied
to these flows. What is the meaning of each of these labels?

9 Processes involving humans communicating with machines have been replaced by machine processes,
i.e. machine to machine communication with machines making decisions and taking actions
according to the code (automated algorithms) that they are programmed with. Machines in these cases
could be embedded computers or nodes in a network such as the Internet, for example. Describe two
examples of applications where machine to machine communication is relied on substantially.
Software engineers and computer scientists have responsibilities in the algorithms that they devise and
the code that they deploy. What precautions should be observed by computer scientists and software
engineers in the development of the two applications that you have described?

Single licence - Abingdon School 342

8 Consequences of uses of computing

In this chapter you have covered:

■■ What it means for a person to act morally
■■ How ethics, a set of principles that applies to a society, can inform a
person’s reasoning when making moral judgements
■■ How the nature of software, data, and information and its global scale
make it difficult to legislate for on the scale of the Internet especially when
it crosses cultural divides, i.e. encounters a different culture
■■ Equally how the nature of software, data, and information and its global
scale creates opportunities through social media, email, blogging, etc. for
people to express themselves freely, to share information and ideas freely,
and to associate freely
■■ The manner in which
• developments in computer science and the digital technologies have
dramatically altered the shape of communications and information
flows in societies, giving rise to an asymmetry of power and pervasive
surveillance
• software is designed so cannot be value free and therefore will inevitably
embed moral and cultural values. This places responsibilities on
computer scientists and software engineers to act morally and ethically
■■ That the ability of computer scientists and software engineers to scale
software and processing quickly for marginal cost offered creates the
potential for great good through access to information whilst software can
also be misused to cause great harm
■■ The difficulty of applying law made in one country to the global Internet
whose content, algorithms and access embed value judgments from
different cultures, societies and legal systems, e.g. online privacy versus
freedom of speech

343 Single licence - Abingdon School

9 Fundamentals of communication and networking

9.1 Communication
Learning objectives:
■■Define serial and parallel
transmission methods and
■■ 9.1.1 Communication methods
discuss the advantages of serial Serial data transmission
over parallel transmission In serial data transmission, single bits (binary digits) are sent one after
another along a single wire by varying the voltage on the wire. Figure 9.1.1.1
■■Define and compare shows a simple electrical circuit for sending single bits coded as 0 volts and 5
synchronous and asynchronous
volts. When the switch is in position A the lamp bulb is connected to 5 volts.
data transmission
When the switch is in position B the lamp bulb is connected to 0 volts. We
■■Describe the purpose of start need to decide what the signal lamp on and the signal lamp off represent.
and stop bits in asynchronous
A
data transmission Signal wire

Key principle 5V
Switch X Lamp bulb

Serial data transmission: B Return wire

In serial data transmission, Figure 9.1.1.1 Simple circuit for sending binary digits serially
single bits (binary digits) are
If single bits are being sent along the wire, then we have one of two possible
sent one after another along a
single wire. binary digit values to represent at any moment, 0 or 1. We may choose to let
the state lamp on represent binary digit 1 and the state lamp off represent binary
digit 0. In which case, the equivalent signals travelling along the signal wire
represent binary digit 1 by 5 volts and binary digit 0 by 0 volts. The binary
digits represent data and the voltages 0 volts and 5 volts their signal equivalent.
Figure 9.1.1.2 shows the transmission of a sequence of data bits using signals
Information
of 0 volts and 5 volts.
In Figure 9.1.1.1 the current
01 1 0 0 1 0 1
travels along a single wire loop.
The moving electric charge picks
5V
up electrical energy in the battery
and delivers this energy to the 0V
lamp bulb.
5V X
Figure 9.1.1.2 Serial data transmission of a sequence of data bits sent as
electrical signals

Questions
1 What is serial data transmission?

Single licence - Abingdon School 344

9 Fundamentals of communication and networking

Key principle Parallel data transmission

In parallel data transmission, bits are sent down several wires simultaneously.
Parallel data transmission: The connecting cable consists of many wires.
In parallel data transmission,
bits are sent down several wires Figure 9.1.1.3 shows two parallel interfaces connected by a parallel connection
simultaneously. The connecting that uses eight data wires labelled 0 to 7, one ground (GND) wire, one clock
cable consists of many wires and signal (CLK) wire. The clock signal wire is set to 5 volts or 0 volts. The data
is called a parallel bus. wires are set to 5 volts or 0 volts.
The receiver reads the data bits in one go by sampling the voltage on each data
wire when it receives the clock signal pulse on the clock wire.

0 bit 0 0
1 bit 1 1
2 bit 2 2
3 bit 3 3
Sending Receiving
parallel 4 bit 4 4 parallel
interface 5 bit 5 5 interface
6 bit 6 6
7 bit 7 7

CLK CLK
GND GND
Figure 9.1.1.3 Parallel data transmission along an 8-bit data bus, controlled
by a clock pulse to signal the arrival of 8 data bits

Questions
2 What is parallel data transmission?

Advantages of serial over parallel

Parallel data transmission has a limited data rate and distance at which it can
Key fact
be reliably operated compared with serial. The limited data rate and distance of
Parallel vs serial data parallel data transmission are caused by skew and crosstalk.
transmission:
Parallel data transmission has a Skew is the phenomenon where the bits travel at slightly different speeds down
limited data rate and distance at each wire in a parallel bus. This includes the clock signal as well. The reading
which it can be reliably operated of the data on the data lines is synchronised with the clock signal. If data and
compared with serial.
clock signal get out of step to such a degree that the data lines are sampled
before the clock signal has appeared or after it has disappeared then data will
not be read correctly. A higher clock rate means narrower clock pulses and
shorter time intervals between pulses. The consequence is a narrower sampling
time window which means less tolerance of skew. The longer the parallel bus

345 Single licence - Abingdon School

 9.1.1 Communication methods

the more data bits on each wire can get out of step with each other and the
clock signal.
Crosstalk is induced signals in adjacent wires of a parallel bus caused when a
signal on one or more wires varies rapidly. The longer a wire and the more rapid
the variation in voltage in adjacent wires the greater the effect.
Serial data transmission doesn’t suffer from skew because it doesn’t use a
separate clock signal and crosstalk is minimised because there are fewer wires
in close proximity and techniques can be applied relatively cheaply to guard
against crosstalk. It is also considerably cheaper over long distances than parallel
would be, simply because fewer wires are used.

A parallel interface is simpler to design than a serial one. The parallel interface
just requires a buffer of the same width as the data to be transmitted or
received. A serial interface must perform a parallel to serial conversion and
vice versa and so is a little more complicated to design than a parallel interface.
However, a parallel interface requires more pins than a serial one.

Questions
3 State two disadvantages of parallel data transmission when compared
with serial data transmission.

Synchronous and asynchronous data transmission Key fact

Serial interfaces are divided into two groups: synchronous or asynchronous.
Serial interfaces:
Synchronous serial data transmission is a form of serial communication in Serial interfaces are divided into
which the communicating endpoints’ interfaces are continuously synchronized two groups: synchronous or
by a common clock. Synchronisation may take the form of special asynchronous.
synchronising bit patterns that are sent periodically or which are attached to
a block of data. It may also take the form of a special clock line as shown in
Figure 9.1.1.4 which is an I2C interface. Key concept
In the Inter-Integrated Circuit (I2C, pronounced I-two-C) synchronous serial Synchronous serial data
transmission:
interface, used by microcontroller-based systems such as the Raspberry Pi, the
Synchronous serial data
clock signal requires its own wire and so adds to the total number of wires that transmission is a form of serial
connect the communicating devices. communication in which the
communicating endpoints’
interfaces are continuously
D7 D6 D5 D4 D3 D2 D1 D0 ACK
SDA Data synchronized by a common
1 2 3 4 5 6 7 8 9 clock.
SCL Clock
GND Ground

Figure 9.1.1.4 Synchronous serial data transmission for an I2C interface

Single licence - Abingdon School 346

9 Fundamentals of communication and networking

Figure 9.1.1.4 shows the three pin and three wire interface for I2C. The data
bits D7 to D0 are individually clocked into the receiving interface by 8 clock
pulses. The receiving I2C interface places an acknowledgement bit onto the
data wire on receipt of the 9th clock pulse.
Whatever the type of synchronous serial interface, a clock signal is associated
with its data
Serial Serial
line(s) and this
interface A interface B
clock signal is RX RX
used by all the
Information
devices connected TX TX
Serial ports:
Desktop PCs used to have at least
to the serial bus Serial
one serial port called the COM to synchronise all bus
port but the advent of USB meant data transfers. GND GND
manufacturers stopped supplying
PCs with serial ports. They are
Another example
absent from tablets and smart of a synchronous Figure 9.1.1.5 Asynchronous serial data transmission
phones where the space for ports interface used between two interfaces supporting transmit (TX) and
is limited. However, the demise of by embedded receive (RX) in both directions simultaneously
serial ports is greatly exaggerated.
microcontroller
Serial ports are everywhere,
in everything from industrial
systems is the Serial Peripheral Interface bus (SPI).
automation systems to scientific Synchronous serial communication is used in telecommunication systems
instrumentation and in Internet of
which the Internet relies upon. These are time-division systems which
Things (IoT) devices. The Arduino
has at least one TTL serial port
continuously send frames of bits between nodes. The nodes are kept in sync
for transmitting and receiving by synchronisation frames which are sent periodically and which distribute the
serial data. common clock signal derived from an atomic clock. This enables very high data
transfer rates not achievable with asynchronous data transfer.
Asynchronous means that data is transferred without support from an external
clock signal. No clock wire is required so reducing the number of wires by one
and the number of connecting pins in each interface by one as well.
Figure 9.1.1.5 shows a serial interface A connected to another serial interface
Key concept B by a serial bus consisting of three wires, one to transmit from interface A to
interface B and one to transmit from interface B to interface A. Both interfaces
Asynchronous data
transmission: share a common ground wire.
Asynchronous means that data Asynchronous serial data transmission sends 7 or 8 data bits at a time and an
is transferred without support optional parity bit framed by start and stop bits in the RS232/RS422 protocol
from an external clock signal.
as shown in Figure 9.1.1.6. These 7 or 8 bits often represent character data
No clock wire is required.
from the ASCII character data set.

Frame: START DATA PARITY STOP

Size (bits): 1 7-8 0-1 1-2
Figure 9.1.1.6 Asynchronous serial data transmission frame

347 Single licence - Abingdon School

 9.1.1 Communication methods

The serial interface is usually a part of an integrated circuit UART

called a UART (Universal Asynchronous Receiver Transmitter)
D0
that buffers a serially received data byte before placing it on an D1

internal 8-bit bus as shown in Figure 9.1.1.7. D2

D3
The UART buffers a byte from the 8-bit data bus before

Parallel
Data bus RX

Serial
D4
clocking it out onto the serial TX wire framed in start and stop TX
D5
bits. The UART performs serial to parallel conversion and vice D6
D7
versa. If the parity bit system is being used, it also generates a
Read/Write
parity bit as well as checking the received data for parity errors. Control bus Clock
Interrupt request

Questions
Figure 9.1.1.7 UART
4 Define the following modes of data transmission
(a) synchronous
Information
(b) asynchronous
Ethernet and USB:
These are serial data transmission
protocols that both encode timing
Comparison of synchronous and asynchronous data information in the communication
transmission symbols, i.e. the bit patterns or
codes that they use. This coding
RS232/RS422 asynchronous serial data transmission is relatively cheap because
therefore delivers both a clock
it requires less hardware than synchronous serial data transmission and is
signal and data, so both can be
appropriate in situations where messages are generated at irregular intervals, for classified as synchronous.
example from embedded systems used in scientific instruments as initiation of
a transfer is relatively quick. However, two in nine or ten bits are control bits
thus a significant proportion of the data transmission conveys no information.

Synchronous data transmissions requires the distribution of a stable clock

signal. In parallel synchronous data transmission this limits the distance and
clock speed. This can also limit serial synchronous data transmission if a
separate clock wire is used. However, in time-division multiplexed synchronous
serial data transmission, this is less of a problem because all clocks are kept
in synchronisation with a master clock by periodically sending synchronising
frames and thus doing away with the need for a separate clock wire. High speed
data transmission is achievable with this method.

Questions
5 Compare synchronous and asynchronous data transmission.

The purpose of start and stop bits

In asynchronous serial data communication such as RS232 the data are
words of a certain word length, for example, a byte, each word is delimited by
start and stop bits as shown in Figure 9.1.1.6. In asynchronous serial data
transmission the transmitter and receiver are not kept synchronised between
transmissions. Instead, the receiver is synchronised with the transmitter only at

Single licence - Abingdon School 348

9 Fundamentals of communication and networking

Key concept the time of transmission. This allows data to be transmitted intermittently such
as when typing characters on a keyboard.
Start and stop bits:
The start bit signals the arrival of Start bit
data at the receiver. This enables The arrival of data at the receiver is signalled by a special bit called a start bit.
the receiver to sample the data
As the arrival of data cannot be predicted by the receiver, the transmission is
correctly by generating clock
pulses synchronised to the bits
called asynchronous.
in the received data. Effectively, The start bit is used to wake up the receiver. The receiver’s clock is set ticking
transmitter and receiver clocks
by the start bit. The clock is just a circuit that is designed to generate a preset
are synchronised by the arrival
number of pulses at fixed intervals which are used to sample the data bits as
of a start bit.
they arrive one after another. The signal changes in the serial data transmission
Both transmitter clock and take place at regular time intervals so the receiver must operate a timing device
receiver clock must have been set at the same rate as the transmitter, so the received bits can be read at the
set up previously to “tick” at the
same regular time intervals.
same rate when running.
Other parameters must agree in The transmitter also operates a timing device, a clock, that is set at a rate
both transmitter and receiver determined by the baud rate (see Chapter 9.1.2). Again this can be as simple a
interface such as the number
circuit to generate a preset number of pulses at fixed intervals of time.
of data bits that will be sent,
the number of stop bits and It is important that the receiver reads each bit during the time that it is not
whether parity is used or not changing, i.e. in the time interval between when changes can take place. This
and if so, what parity, even or
is why the receiver’s timing device needs to be brought in step or synchronism
odd. These parameters are set
with the transmission’s timing.
up once before either interface
is used. Figure 9.1.1.8 shows two computers, A and B, with a TTL serial connection
between their serial ports. TTL is Transistor-Transistor Logic and is a logic that
operates between 0 volts and 5 or 3.3 volts. TTL relies on circuits built from
bipolar transistors to achieve switching and maintain logic states.
For the link from computer A to computer B, the data wire is kept at the
voltage level corresponding to a binary digit 1 when not sending – the
idle state. A data
transmission is started
Computer A Pulse train travelling along single wire Computer B
between A and B by changing the
voltage level
5 volts Idle Stop Parity LSB Idle state

A B to the level for

0 volts MSB Start binary digit 0. This
TX RX
TTL Serial Interface TTL Serial Interface is the start bit. The
GND GND
transmitter (TX) then
follows the start bit
Clock Clock
with 7 or 8 data bits
depending on how the
serial port has been
Figure 9.1.1.8 Asynchronous serial data transmission between the TTL configured. The least
serial interface of two computers, A and B significant data bit
(LSB) is sent first and

349 Single licence - Abingdon School

 9.1.1 Communication methods

the most significant data bit (MSB) last. If parity is enabled, the last data bit is
followed by a parity bit.
Stop bit
Finally, the transmitter attaches a stop bit. The voltage level chosen for the stop
bit is the level for binary digit 1. The time interval for the stop bit allows the
receiver (RX) to deal with the received bits, i.e. transfer them into the RAM of
the computer, before receiving and processing the next serial frame as shown in
Figure 9.1.1.9. Two stop bits are used if the receiver needs more time to deal
with the received bits.
0 Time

1101011001010011100111
t t
le dle tar it 0 it 1 it 2 it 3 it 4 it 5 it 6 it 7 top tar it 0 it 1 it 2 it 3 it 4 it 5 it 6 it 7 top
Id I S b b b b b b b b S S b b b b b b b b S

Figure 9.1.1.9 Asynchronous serial data transmission frame showing

transmission of two bytes without parity

Questions

6 Describe the purpose of start and stop bits in asynchronous data

transmission.

In this chapter you have covered:

■ Serial and parallel transmission methods and the advantages of serial over
parallel transmission
■ Synchronous and asynchronous data transmission and how they compare
with each other
■ The purpose of start and stop bits in asynchronous data transmission

Single licence - Abingdon School 350

9 Fundamentals of communication and networking

9.1 Communication
Learning objectives:
■ Define: ■ 9.1.2 Communication basics
• baud rate
Baud rate
• bit rate The baud rate sets the maximum frequency at which signals may change.
To understand the meaning of baud rate consider the following simplified
• bandwidth
switching system shown in Figure 9.1.2.1. Switch A operates at the baud rate.
• latency For example if the baud rate is 1 baud then the switch remains connected to
one of the four electrical voltages for one second. At the end of each second,
• protocol
the switch can switch the connection to any one of the four possible voltages.
■ Differentiate between baud The changeover happens in a time period very much smaller than one second
rate and bit rate (theoretically, it happens instantaneously). Thus, the output signal that appears
■ Understand the relationship on the signal wire may change every one second but no quicker. If the baud
between bit rate and rate is 10 baud then the switch can change position every tenth of a second.
bandwidth Switch A
7.5 7.5
5.0 5.0
Key concept 2.5 2.5 volts

Baud rate: The maximum rate 0.0

0.0 2.5 5.0 7.5
at which signals on a wire or Signal wire
volts
line may change.

1 baud: One signal change per

Ground/return wire
second.
Figure 9.1.2.1 Simplified switching system illustrating baud rate
For example, a computer’s serial port may be set to send at 1 baud, the signal
sent out by the computer can change only at the end of each elapsed second.
Table 9.1.2.1 shows the rates of signal change for some baud rates.
Time between Rate of signal changes
Baud rate
signal changes (s) (changes per second)
1 1 1
2 0.5 2
4 0.25 4
1000 0.001 1000
10000 0.0001 10000

Table 9.1.2.1 Relationship between baud rate and rate of signal changes

Questions
1 How often may switch A change if the baud rate is 100 baud?

2 How many times a second can the switch change if the baud rate is 1000 baud?

Single licence - Abingdon School 351

9 Fundamentals of communication and networking

Key concept Bit rate

Bit rate is measured in bits per second. It is the number of bits transmitted
Bit rate: The number of bits per second. The bit rate is the same as the baud rate when one bit is sent
transmitted per second.
between consecutive signal changes. However, it is possible to send more than
one bit between signal changes if more than two voltage levels are used to
encode bits. If the voltages 0 volts, 2.5 volts, 5 volts and 7.5 volts are used, then
the decimal numbers in Table 9.1.2.2 can be encoded.
Key fact
bit rate = baud rate x the no Signal level (volts) Decimal number Binary number
of bits per signal (voltage) 0 0 00
2.5 1 01
5 2 10
7.5 3 11

Table 9.1.2.2 Linking signal levels and number of bits they encode

Figure 9.1.2.2 shows how two bits of data can be encoded per time slot on a 1
baud line, giving a bit rate of 2 bits per second.

11 00 10 01 10 11 01 10
7.5
Voltage/V

5.0
2.5
0.0 Time/s
0 1 2 3 4 5 6 7 8
Figure 9.1.2.2 Sending data along a 1 baud line using
four levels of voltage to encode bits

From this we may conclude that the relationship between but rate and baud
rate is
bit rate = baud rate x the no of bits per signal (voltage)

Questions
3 What is meant by (a) baud rate (b) bit rate?
4 Explain the difference between baud rate and bit rate.

5 The following voltage levels expressed in volts are chosen to encode

bits:
-6.0, -4.5, -3.0, -1.5, +1.5, +3.0, +4.5, +6.0

How many bits represent these voltages?

6 For the voltages given in question 5 write down one possible set of
corresponding bit patterns (an example of a bit pattern is 01).

7 If the baud rate of the line is 900 baud what is the bit rate for the
voltage levels given in question 5?

352 Single licence - Abingdon School

 9.1.2 Communication basics

Bandwidth
Key concept
Bandwidth is a measure of how fast the data may be transmitted over the
transmission medium. The greater the bandwidth, the greater the rate at which Bandwidth:
data can be sent. The bandwidth of a
transmission medium, e.g.
The bandwidth of a transmission medium, e.g. copper wire, is the range of copper wire, is the range of
signal frequencies that it may transmit from one end of the wire to the other signal frequencies that it may
without significant reduction in strength. Bandwidth is measured in hertz (Hz), transmit from one end of
the communication link to
e.g. 500Hz. The hertz is a unit of frequency equal to one cycle per second.
the other without significant
Figures 9.1.2.3 and 9.1.2.4 show the effect of the transmission medium on reduction in strength of the
two different frequencies. signal.

A B

Figure 9.1.2.3 Low-frequency signal injected onto wire at A arrives at B

with its strength relatively undiminished

A B

Figure 9.1.2.4 Higher-frequency signal injected onto wire at A arrives at B

with its strength diminished significantly

Although a given signal may contain frequencies over a very broad range,
any medium used to transmit the signal will be able to accommodate only a
limited band of frequencies. This limits the bit rate that can be carried on the
transmission medium. Figure 9.1.2.5 shows the effect of a 500 Hz bandwidth
signal channel, e.g. a copper wire, on a transmission with bit rate of 2000 bits
per second.
Bits 0 1 0 0 0 0 1 0

Pulses before
transmission.
Bit rate 2000 bps

Pulses after
transmission
become the
solid line.
Bandwidth 500 Hz
Figure 9.1.2.5 A 2000 bps transmission over a 500 Hz signal channel
Single licence - Abingdon School 353
9 Fundamentals of communication and networking

Questions

8 What is meant by bandwidth?

Latency
Key concept
Latency is the time delay that can occur between the moment something
Latency: (an action) is initiated and the moment its first effect begins. In a wide area
Latency is the time delay that network involving satellites, significant time delay occurs because of the
can occur between the moment
physical distance between the ground stations and the geostationary satellite.
something (an action) is initi-
ated and the moment its first Requesting and receiving a web page can involve a considerable time delay,
effect begins. even though the bit rate of the uplink and downlink to the satellite is high, i.e.
the bandwidth is large. The speed of microwaves is 3 × 108 m/s. With a round-
trip distance of over 143,200 km, the propagation time delay is approximately
0.4 s.

Questions
9 What is latency in the context of communications?

Key concept Protocol

A communication protocol is a set of pre-agreed signals, codes and rules used
Communication protocol:
A set of pre-agreed signals, codes to ensure successful communication between computers or a computer and a
and rules to be used for data and peripheral device such as a printer.
information exchange between
A protocol will govern such things as:
computers, or a computer and
a peripheral device such as a • Physical connections – e.g. RS423 serial
printer, that ensure that the
• Data format – packet/frame size
communication is successful.
• Error detection and correction

Information • Cabling – Cat 5, Optical fibre

Handshaking protocol: • Speed – baud rate, bit rate
The sending and receiving devices
• Flow control
exchange signals to establish that
the receiving device is connected • How data is to be sent
and ready to receive. Then the C → Are you ready? → P
For example, serial data
sending device coordinates the C ← Yes I am ← P
sending of the data, informing communication, uses a handshaking
Here it is
the receiver that it is sending. protocol to control the flow of data. C → → P
(Start bit)
Finally, the receiver indicates it has In a handshaking protocol, the C ← Busy ← P
received the data and is ready to
sending device checks first to see if That’s it
receive again. C → → P
the receiving device is present. If it (Stop bit)
is present, the sending device then C ← I’m ready again ← P
enquires if the receiving device is Table 9.1.2.3 Handshaking protocol:
ready to receive. The sending device computer = C, printer = P

354 Single licence - Abingdon School

 9.1.2 Communication basics

waits for a response which indicates that the receiving device is ready to receive. On receipt of this signal, the
sending device coordinates the sending of the data and informs the receiver that it is sending the data. The sender
then waits for the receiver to become ready to receive more data.

Questions
10 What is a communications protocol?

Understand the relationship between bit rate and bandwidth

Key principle
There is a direct relationship between bit rate and bandwidth. The greater
the bandwidth of the transmission system, the higher the bit rate that can be Bit rate and bandwidth:
transmitted over that system. If the data rate of the digital signal is W bits per The greater the bandwidth of
the transmission system, the
second (bps) then a very good representation can be achieved with a bandwidth of
higher the bit rate that can be
2W Hz. transmitted over that system.
For example, Figure 9.1.2.6 shows a 4 x 5 chequered board of black and white
squares. Suppose that the information contained within this board of which
squares are black and which white
is sent as a bit stream row by row,
starting from the top left square.
The bits and the corresponding
stream of voltage pulses are 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 W bps
5 volts W Hz
also shown in the figure. The 0 volts 2

fundamental frequency with which

the pulses alternate between 5 volts
and 0 volts is W/ Hz where W Figure 9.1.2.6 A chequered pattern being encoded in bits and sent
2
is the bit rate in bits per second. serially as a bit stream at W bps
We know from section 5.6 that a
square wave is made of harmonics of the fundamental frequency, f. Therefore, the bandwidth of the communication
channel must allow at least the first harmonic which is 3f to travel without significant reduction in amplitude. A
bandwidth of four times the fundamental frequency, f, should therefore be adequate,
i.e. 4 x W/ = 2W Hz.
2 Questions
In this chapter you have covered: 11 What is the relationship between bandwidth and bit rate?

■■ Definitions for
• baud rate
• bit rate
• bandwidth
• latency
• protocol
■■ The difference between baud rate and bit rate
■■ The relationship between bit rate and bandwidth

Single licence - Abingdon School 355

9 Fundamentals of communication and networking

9.2 Networking
Learning objectives:
■ Understand ■ 9.2.1 Network topology
• physical star topology
Topology
• logical bus network The way computers are cabled together or linked to form a network is very
topology important. The term topology is used to describe the layout of a network.
When the computers that are linked are in close proximity, e.g. in a single
■ Differentiate between both
building, the network is called a local area network.
■ Explain their operation The topology in Figure 9.2.1.1(a) is a mesh and that in Figure 9.2.1.1(b) is a
■ Compare each (advantages bus. Computer Computer
and disadvantages)

Bus

Key concept
Network topology:
The shape, layout,
(a) (b)
configuration or structure of the
connections that connect Figure 9.2.1.1 Two different network layouts
devices to the network.
Star topology
The most common physical network topology is the star which is shown in
outline in Figure 9.2.1.2. The centre of the star is either a network switch or a
central computer.

Key concept File server

Local Area Network (LAN):
Linked computers in close
proximity or in a small
geographical area.

Central switch
or central computer

Figure 9.2.1.2 Star network topology

Single licence - Abingdon School 356

9 Fundamentals of communication and networking

Key fact A computer communicates on the network through a network interface card or
network adapter. A network adapter plugs into the motherboard of a computer
MAC address: and into a network cable. Network adapters perform all the functions required
A MAC address is a 48-bit address
to communicate on a network. They convert data between the form stored
expressed in hexadecimal and
separated into 6 bytes, in the computer and the form transmitted or received on the cable (Figure
e.g. 00-02-22-C9-54-13. 9.2.1.3).
It is the physical or hardware
Serial data flows from the network
address of the network adapter adapter card on to the network
and is designed to be unique.
Network
adapter
card
Did you know?
Ethernet uses CSMA/CD Parallel data flows from the
(Carrier Sense Multiple Access/ computer’s motherboard
Collision Detection). Ethernet to the network adapter card

is a very popular bus system and

generally refers to a standard Computer
published in 1982 by Digital motherboard
Equipment Corporation,
Intel Corporation and Xerox
Corporation. Figure 9.2.1.3 Network adapter
It is the predominant form of
local area technology used with A network adapter receives data to be transmitted from the motherboard of a
TCP/IP today. It operates at computer into an area of memory called a buffer. The data in the buffer is then
three speeds:
passed through some electronics that calculates a checksum value for the block
10 Mbps (standard Ethernet),
of data (CRC ) and adds address information, which indicates the address of
100 Mbps (fast Ethernet) and
1000 Mbps (gigabit Ethernet). the destination card and its own address, which indicates where the data is
It uses 48-bit addresses. Data to from; each network adapter card is assigned a permanent unique address at the
be transmitted is broken into time of manufacture. The block is now known as a frame.
variable sized packets called
frames (Figure 9.2.1.4). Ethernet bus protocol uses the frame structure shown in Figure 9.2.1.4.

Destination Source Type CRC

Data
address address
6 bytes 6 bytes 2 bytes 46 - 1500 bytes 4 bytes

Figure 9.2.1.4 Ethernet frame

Did you know? The network adapter then transmits the frame one bit at a time onto the
network cable. The address information is sent first, followed by the data and
The MAC address is written
to EEPROM (Electrically
then the checksum. In the Ethernet protocol, each network card is assigned
Erasable Programmable Read a unique address called its MAC address. MAC stands for Media Access
Only Memory) on the network Control. A MAC address is a 48-bit address normally expressed in hexadecimal
adapter. This means that it is and separated into 6 bytes, e.g. 00-02-22-C9-54-13. Part of the MAC address
possible with the right software
identifies the manufacturer. Each network card manufacturer has been allocated
to alter a card’s MAC address.
Why might a person want to a block of MAC addresses to assign to their cards.
do this? Figure 9.2.1.5 shows how three computers connected via T-pieces to a bus can
send Ethernet frames to each other.
357 Single licence - Abingdon School
 9.2.1 Network topology

Computer Computer Computer

00-02-22-C9-54-13 00-02-22-C9-54-14 00-02-22-C9-54-15

T-piece connector Linear transmission medium or bus

Figure 9.2.1.5 Bus network topology
Figure 9.2.1.6 shows a physical bus network that uses the transmission medium of coaxial cable to interconnect
network adapters and their computers. Since the bus transmission medium is a shared medium only one computer
can send at a time. However, because it is a shared medium, every network adapter is able to “see” each transmitted
Ethernet frame. Each network adapter checks
Network Ethernet frames to see if the destination address
adapter field contains the adapter’s MAC address. If it does
it buffers the frame and reads the data. Collisions
can occur when two adapters try to send at the same
time. Each will stop transmitting and delay sending
again for a random time period. If collisions occur
too frequently the speed of communication over the
network is reduced.
T-piece
Switched Ethernet
Coaxial connector
The solution to the collision problem is to restrict
cable
the communication channel to just each pair of
Figure 9.2.1.6 Physical bus network using coaxial cable sending and receiving computers, at a time. To cope
to interconnect network adapters with more than one computer sending at the same
time, the transmissions are buffered and then sent in
turn to the corresponding receiving computer. A fast
switch is required to make the temporary bus connection between each pair of sending and receiving computers.
This is what an Ethernet switch is
Station Station Station Station
designed to do. A B C D
In switched Ethernet the LAN is
wired in star topology with the
nodes (computers or workstations)
Station Station
connected to a central switch L E
(Figure 9.2.1.7). Backbone
Station Station
Even though the physical layout K F
or topology is a star, the LAN still
behaves as a bus. We say that the
network physically wired in a star
Station Station Station Station
topology can behave logically J I H G
as a bus network by using a
Figure 9.2.1.7 Central switch details and computers or workstations
bus protocol and appropriate
connected in a star configuration to this central switch

Single licence - Abingdon School 358

9 Fundamentals of communication and networking

physical switching. The central switch queues

frames until each frame can be placed onto the Switching Switching
electronics electronics
backbone. The switch ensures that collisions do
not occur. For example, if computer A launches Line card
Input buffer Output buffer
an Ethernet frame for computer H, the switch
creates a temporary exclusive connection Input line Output line
from computer A to computer H. If computer receiver transceiver
B simultaneously launches an Ethernet frame
Unshielded
for computer D, the switch will buffer the twisted pair
frame until the backbone becomes free. Cable
Switched Ethernet eliminates collisions, so its
performance is superior to Ethernet LANs based
on multidrop coaxial cable. From To
Blue workstation workstation
In switched Ethernet a separate cable is run Figure 9.2.1.8 A line card connected to a cable containing
from a central switch to each workstation. If two unshielded twisted pairs
there are n workstations, there are n separate
cables. At the switch end, a cable is connected to a line card. Therefore, for n cables there are n line cards. Figure
9.2.1.8 shows a cable connected to a line card. Figure 9.2.1.9 shows a network cable plugged into the Ethernet
switch and a cutaway of two twisted pairs emerging from CAT5 cable terminated by an RJ45 connector. To enable
bidirectional data transfer, the cable consists of two independent pairs of wires. One pair of wires forms the input
circuit and the other the output circuit. The wires in each pair are twisted together, hence the name twisted pair.

Blue

Orange + White

Orange
Blue + White

Figure 9.2.1.9 Cat 5 network cable showing two twisted pairs:

Orange + White & Orange, Blue + White & Blue
A workstation may be a workstation computer, a server, a dumb terminal or some other device. A workstation
transmits a packet of data to the line card along the input pair. The packet is stored in the input buffer of the line
card. The switching electronics reads the destination address contained in the packet then routes the packet along a
backbone in the switch to the line card connected to the destination. A backbone is a high-speed bus.

359 Single licence - Abingdon School

 9.2.1 Network topology

Figure 9.2.1.10 shows a 10/100 Mbps Ethernet network adapter that connects to the PCI bus of a computer.

Cat5 network
cable plugs
into RJ45
socket RJ45 socket

PCI bus connector

Figure 9.2.1.10 10/100 Mbps Ethernet network adapter that plugs connects to the PCI bus of a computer

Questions
1 In the context of networking, what is a topology?

2 Draw a diagram that illustrates the essentials of a bus network.

3 Draw a diagram that illustrates the essentials of a star network.

4 Explain how a network wired in star topology can behave logically as a bus network.

5 What is a network adapter?

6 What is a MAC address?

7 Explain the collision problem in the context of a bus network.

8 How does switched Ethernet overcome the collision problem?

Single licence - Abingdon School 360

9 Fundamentals of communication and networking

Comparing bus and star networks

Bus and star topologies appear very similar in the way that they are physically wired using the current switch-based
hardware. Even thin-client systems, which can be considered to resemble a traditional star network, use an Ethernet
bus switch to connect a central server to nodes.
In a traditional star network, each link from node to central computer is an independent link. Each link is
therefore secure from eavesdropping by other nodes.
If a link to a node goes down, the other links and nodes are unaffected. However, if the central computer / central
switch goes down, the whole network will fail.
In a true star-based network, the speed of each link to the central computer should remain high, because the links
are not shared. Traffic between nodes in a switch-based bus network will not be adversely affected if a node goes
down, unless the traffic involves the broken node or the node is a domain server that validates users when they
attempt to log in. Unplugging a network cable in a switch-based bus network will not affect the rest of the network.
In a coaxial cable bus network, a break in the cable stops the whole network from working. All connected nodes
are able to read the frames travelling on the coaxial cable bus network. Therefore coaxial cable bus networks are not
secure against eavesdropping. The frames in a coaxial cable Ethernet bus network can collide when multiple nodes
send at the same time, causing a noticeable slowdown. Although collisions between frames in switch-based Ethernet
bus networks cannot occur, performance can be affected when traffic volumes are high, because the buffers in the
switches suffer overflow.
A wireless network is a broadcast network, so it is less secure than a cabled switch-based Ethernet network unless
wireless encryption is enabled. In a wireless network without encryption, it is possible to eavesdrop on traffic
intended for other computers. A wireless network can also suffer congestion because the channels are shared.

Questions
9 Discuss the advantages and disadvantages of operating a logical bus network topology wired as a physical
star topology. You may wish to make reference in your answer to a physical bus topology and a traditional
star topology.

In this chapter you have covered:

■■ The operation of
• physical star topology
• logical bus network topology
■■ Differences between both
■■ The advantages and disadvantages of each

361 Single licence - Abingdon School

9 Fundamentals of communication and networking

9.2 Networking
Learning objectives:
■ Explain the following and
describe situations where they
■ 9.2.2 Types of networking between hosts
might be used Peer-to-peer networking
In a peer-to-peer (P2P) network there is minimal or no reliance on dedicated
• peer-to-peer networking
servers.
• client-server networking All computers are equal hence the name peer.
Each computer can communicate with any other computer on the network to
which it has been granted access rights.
Key concept
Each peer computer acts as a client when initiating requests to another
Peer-to-peer network: computer(s) for resources and as a server when satisfying requests from another
A network in which there is computer(s).
minimal or no reliance on
dedicated servers. All computers Figure 9.2.2.1 shows a wired peer-to-peer local area network and the possible
are equal, and are called peers. communication paths between peers.
Each peer may act as both a
client and server.
.

Peer-to-peer Switch
network

Station A Station B Station C

Peers act as
both clients and
servers
A B C
Peer Peer Peer

Figure 9.2.2.1 Peer-to-peer wired local area network

There is no central control and normally there is no administrator responsible

for the entire network. The user at each computer acts as a user and an
administrator, determining what data, disk space and peripherals on their
computer get shared on the network.
Security control is limited because it has to be set on the computer to which it
applies. The computer user typically sets the computer’s security and they may
choose to have none. It is possible to give password protection to a resource

Single licence - Abingdon School 362

9 Fundamentals of communication and networking

on the computer, e.g. a directory, but there is no central login process where a
user’s access level is protected by a single password. A user logged in at one peer
computer is able to use resources on any other peer computer if the resources
are unprotected by passwords or if the user knows the relevant password.
Peer-to-peer networks can be as small as two computers or as large as thousands
of computers.
Questions
1 Explain peer-to-peer networking.

Peer-to-peer local area networks

A peer-to-peer local area network (LAN) is a good choice for environments
where:
• there are fewer than 10 users
• the users are all located in the same area and the computers will be
located at user desks
• security is not a major concern, so users may act as their own
administrators to plan their own security
• the organisation and the network will have limited growth over the
foreseeable future.
For Windows 7 desktop operating system, the maximum number of peers
permitted in a peer-to-peer local area network is 20 as revealed by the
command NET CONFIG SERVER as shown in Figure 9.2.2.2.

Figure 9.2.2.2 NET CONFIG SERVER command run in Windows 7 console

Questions
2 Describe the circumstances when peer-to-peer is an appropriate choice for local area networks.

363 Single licence - Abingdon School

 9.2.2 Types of networking between hosts

Server-based network Information

A peer-to-peer local area network, with computers acting as both client and
Hosts in networks:
server, is seldom adequate for a system with more than 10 users. Therefore,
A host is a computer or device
most networks use dedicated servers. A server-based local area network is a
that is accessible over a network.
client-server network in which resources, security, administration and other
functions are provided by dedicated servers. Clients request services that are
satisfied by dedicated servers.
A dedicated server is one that functions solely as a server and is not used as
a client or workstation. Servers are usually optimised in both hardware and Key concept
operating system to quickly service requests from network clients and to ensure
Server-based network:
the security of files and directories. Larger networks with a higher volume of
A server-based local area
traffic employ more than one server. network is a client-server
Clients use servers for services such as file storage and printing in a local area network in which resources,
security, administration and
network. Client computers are usually less powerful than server computers. A
other functions are provided
server can also authenticate users attempting to log on at client workstations; by dedicated servers. Clients
it stores the client users’ IDs and passwords for this purpose. Typically, school request services that are satisfied
networks are server-based networks (thick-client networks): a central domain by dedicated servers.
controller stores user accounts and a central file server stores users’ work and
some applications that users download into the client machines they work at.

Questions
3 Explain how a server-based local area network differs from a peer-
to-peer local area network and why it is considered a client-server
network.

Client-server and peer-to-peer networking architectures Key concept

Web browsing and sending email are client applications that rely on an
Client-server architecture:
underlying network architecture to interact with the corresponding server
In a client-server architecture,
applications, a web server and an email server, respectively. there is an always-on host,
Both clients and servers are software applications. They conform to one of two called the server, which services
particular application architectures used in modern networking: requests from many other hosts,
called clients. For example,
1. the client-server architecture when a web server receives a
request from a client host for
2. the peer-to-peer (P2P) architecture
a web page, it responds by
Web servers and email servers are examples of server-based systems. sending the requested web page
to the client host.
Client-server networking architecture
In a client-server architecture, there is an always-on host, called the server,
which services requests from many other hosts, called clients. A classic example
is the web application for which an always-on web server services requests

Single licence - Abingdon School 364

9 Fundamentals of communication and networking

from browsers running on client hosts. When a web server receives a request from a client host for a web page, it
responds by sending the requested web page to the client host.
In the client-server architecture, clients do not directly communicate with each other. For example, two web
browsers do not directly communicate. This is very different from the peer-to-peer architecture where peers can
communicate with each other because peers can act as both client and server (Figure 9.2.2.1).
Another characteristic of the client-server architecture is that a server has a fixed, well-known address in TCP/IP
networks, called an IP address. A client can always contact the server by sending a packet to the server’s IP address
and get a response because the server is always on.
Search engines such as Google and Bing employ more than one server (hundreds of thousands, in fact) in order to
meet demand. However, to the client these servers appear as a single machine, a virtual server.
Questions
A
4 What is meant by client-server architecture
H B and describe a situation where it is used?

Download Figure 9.2.2.3 shows a server, S, connected to

eight clients, A to H. The server may be capable of
uploading (transferring to clients) at a rate of, say,
5000 KiB/s whilst each client may be capable of
G S C
Upload downloading at a rate of, say, 500 KiB/s, i.e. a ratio
of ten to one. If the server application running at
the server is an FTP server delivering eight copies
of a file to an FTP client application running at
F D each of the eight clients then the server is more than
capable of serving each client at their download rate
E of 500 KiB/s. Therefore, a file of size 500 KiB (500
x 1024 bytes) will take one second to download to
Figure 9.2.2.3 Client-server architecture all eight clients. Figure 9.2.2.4 shows an FTP client
downloading a file from an FTP server.

Figure 9.2.2.4 shows an FTP client downloading a file from an FTP server

365 Single licence - Abingdon School

 9.2.2 Types of networking between hosts

Increasing the number of clients to twenty reduces the download rate because
the server’s upload rate of 5000 KiB/s is now shared by 20 clients. This means a
maximum download rate of 5000/ KiB/s per client = 250 KiB/s. This is less
20
than the client’s maximum of 500 KiB/s.
One solution is to add another server so that the demand from twenty clients is
shared now between two servers. This should restore the download rate to 500
KiB/s assuming that each client is able to operate at a download speed of 500
KiB/s.
Peer-to-peer networking architecture
The client-server model works well for many applications. However, if many Key concept
clients are downloading very large files from a server, then download speeds Peer-to-peer architecture:
diminish unless more servers are added. It would be better if uploads could be In a peer-to-peer architecture,
shared amongst both servers and clients. This would lighten the load on servers client hosts not only download
and reduce the amount of server bandwidth that clients have to pay for. In fact, a file, but can upload what they
have obtained to others as well.
web hosting could be eliminated altogether. Client hosts not only download
Client hosts are capable of doing
a file, but can upload what they have obtained to others as well. Client hosts both at the same time. They are
are perfectly capable of doing both at the same time. They are then called then called peers rather than
peers rather than clients. Figure 9.2.2.5 shows how separate parts of a file are clients.
uploaded to three peers which in turn upload what they have to their peers.
Peer distribution can start whilst each peer is still downloading their part of the
file from the server. The server doesn’t have to be a dedicated server but can be
another peer that happens to have the file that other peers want to download.

File Peer that has the file

⅓ Part 1 that others want acts
⅓ Part 2
⅓ Part 3
Server as the seed

S
Part 1
⅓ Part 3

⅓
⅓ Part 2

Part 3

Part 2 Part 2
A B C
Part 1 Part 3
Peer Peer Peer
Part 1

Figure 9.2.2.5 shows peers uploading and downloading files using FTP

BitTorrent came up with a protocol for distributing large files in this way.
BitTorrent doesn’t overload servers that provide the download since it relies on
peers contributing upload capacity. The result is that large files can be received

Single licence - Abingdon School 366

9 Fundamentals of communication and networking

Did you know? faster than would be the case in a client-server architecture where downloads
Jaan Tallinn was one of the co- rely on just a central server.
founders and authors of Skype. Decentralized P2P networks have several advantages over traditional client-
Read an interview with Jaan at
server networks. P2P networks scale well because they don’t rely on costly
http://affairstoday.co.uk/interview-
jaan-tallinn-skype/
centralized resources. Scaling doesn’t lead to a deterioration in download speed
because P2P networks use the processing and networking power of the end-
users’ machines which grows in direct proportion to the network itself.
The costs associated with a large, centralized infrastructure are virtually
eliminated because the processing power and bandwidth reside in the peers
within the network.

Investigation
1 Peer-to-peer (P2P) architecture became widely used and popularized
by file-sharing applications such as Napster and Kazaa. Research the
history of Napster and Kazaa.

2 Skype originally relied on a form of P2P architecture. Research the

history of Skype and why its P2P supernodes were replaced with Linux
boxes hosted by Microsoft.

Questions
5 Explain what is meant by peer-to-peer architecture and describe a
situation where it is used?

In this chapter you have covered:

■■ peer-to-peer networking
■■ client-server networking
■■ situations where they might be used

367 Single licence - Abingdon School

9 Fundamentals of communication and networking

9.2 Networking
Learning objectives:
■ Explain the purpose of WiFi ■ 9.2.3 Wireless networking
■ Be familiar with the The purpose of WiFi
components required for
WiFi was invented to provide a wireless connection between computing devices
wireless networking
and to enable these devices to connect
■ Be familiar with how wireless
Internet
to the Internet via a bridge between a
networks are secured wireless LAN and a wired LAN known Fibre/DSL
Modem

as an access point.
■ Explain the wireless protocol
Router

WiFi or Wi-Fi® is officially called

Carrier Sensing Multiple Wired Ethernet

IEEE 802.11, because of the naming Access Point

Access with Collision
scheme that the IEEE (Institute of
Avoidance (CSMA/CA) with
Electrical and Electronic Engineers)
and without Request to Send/
uses to name their standards. The 802 Basic Service
Set (BSS)
Clear to Send (RTS/CTS)
part means a Local Area Network Figure 9.2.3.1 Wireless Access Point
■ Be familiar with the purpose (LAN) as wireless is short-range, and connection to Internet
of Service Set Identifier (SSID) the .11 part is for wireless.
Figure 9.2.3.1 shows WiFi providing wireless connections to computing
Key concept devices such as laptops, tablets and smartphones that are close to an access
point. The access point is also connected to a wired local area network
WiFi LAN or WLAN:
A wireless local area network (Ethernet) which in turn provides wired access to the Internet via a router and
that is based on international modem (Digital Subscriber Line (DSL) or optical fibre).
standards laid down by the Wireless networking uses radio waves from the electromagnetic spectrum in
organisation known as the two bands of frequencies centred around 2.4 and 5 GHz. Very importantly,
IEEE.
these two bands of frequencies do not require licensing to use them unlike the
It is used to enable devices to
connect to a local area network mobile phone network spectrum which uses expensive licensed frequencies.
wirelessly. Table 9.2.3.1 shows the standards as they have evolved over the years from
802.11 to 802.11ac. Maximum speeds are very rarely achieved for reasons that
Information will be given later.
802.11 mobile and portable 802.11 Frequency Maximum
devices: Year
Standard (GHz) speed (Mbps)
A requirement of the 802.11
- 1997 2.4 2
standard is to handle mobile as
well as portable wireless stations. b 1999 2.4 11
A portable station is one that is a 1999 5 54
moved from location to location, g 2003 2.4 54
but that is only used while at a 2009 2.4 & 2.5 150
n
fixed location. Mobile stations
Typical 800
actually access the LAN while in ac 2013 2.4 & 5
(theoretical1.3 Gbps)
motion.
Table 9.2.3.1 WiFi standards
Single licence - Abingdon School 368
9 Fundamentals of communication and networking

Questions
Key concept
1 What is the purpose of WiFi?
Frame/packet:
2 A frame or packet is a unit
In a wired network, data travels between wired devices by electrical means.
of transfer consisting of
What is the equivalent means by which data travels in a wireless network?
data, addresses and control
information.

Components for wireless networking

Wireless networks typically consist of the following major physical components:
• Stations: computing devices with wireless network interfaces
• Access Points (APs): provide the wireless-to-wired bridging function in which wireless frames are converted
to wired frames - usually Ethernet frames (a frame is a unit of transfer
Information
consisting of data, addresses and control information; sometimes called a
packet); access points are also used to control access to a wireless network Wireless propagation
characteristics:
by authenticating users or devices that wish to join the network
Propagation characteristics are
• Wireless medium: to move frames from station to station, the 802.11 dynamic (not static) and
standard uses a wireless medium consisting of two frequency bands, 2.5 unpredictable.

GHz and 5 GHz, each divided into channels

• Distributing system: used to connect several access points to form a large coverage area of the same LAN,
e.g. a hotel with WiFi access in rooms. The APs are often connected by a wired Ethernet backbone.
In WiFi, devices (stations) belong to what
is called a Basic Service Set (BSS) which is Access Point
simply a group of stations that communicate
with each other wirelessly in an area (green
shading) defined by the propagation
characteristics of the wireless medium and
called the basic service area (Figure 9.3.2.2).
Infrastructure
BSSs are of two types: Independent Basic Service
Basic Service Set (BSS)
• Independent BSS: stations Set (BSS)
communicate with each other
(a) (b)
directly and therefore must be
within direct communication range. Figure 9.2.3.2 (a) independent BSS (b) infrastructure BSS
Independent BSSs are used to create
Information
short-lived networks, e.g. for a meeting in a conference
room. This type of BSS network is commonly known as BSSID address:
A BSSID address is used to identify a wireless LAN.
an ad hoc network because of its limited duration, small
Stations in the same area may be assigned to a
size and focused purpose. Basic Service Set (BSS) to form a LAN. The LAN is
• Infrastructure BSS: infrastructure BSSs always use at then identified by its BSSID.
In infrastructure networks, the BSSID is the MAC
least one access point and access points are used for all
address used by the wireless interface in the access
communications including communication between point. It is a 48-bit identifier for the BSS.
stations. Two hops are therefore used to send frames.
369 Single licence - Abingdon School
 9.2.3 Wireless networking

The originating station transfers the frame to the access point (first hop) which
then relays it to the destination station (second hop). This lifts the restriction
that stations must be in range of each other. They only have to be in range of an
access point belonging to the BSS. Secondly, access points can buffer frames so
that stations which are battery-powered can be powered down until they need to
transmit and receive frames, e.g. wireless sensors.
The stations must have the capability to send and receive over a WiFi connection.
Figure 9.2.3.3 shows an EDIMAX® wireless network adapter that plugs into a
USB port of a computer. This particular adapter has MAC address 001F1FCD5D7A.
All adapters, wireless or wired have a unique MAC address so that they can be
identified.
The tall slim tubular structure is its antenna through which it radiates and receives
radio waves on specific frequencies designated by the IEEE.
A wireless network adapter is also known as a wireless interface.
Desktop computers can have a wireless network adapter installed on their
motherboard in a PCI slot and tablets and smartphones use a built-in wireless
network adapter.
Within each BSS, stations communicate directly with an Access Point (AP),
similar to a mobile phone network base station. The Access Point acts as a bridge
between a wireless and a wired local area network. When a device searches for WiFi
Figure 9.2.3.3 EDIMAX® connectivity, it sends messages to discover which APs are in its transmission range.
USB port wireless adapter
802.11b/g

This results in a list of names

such as shown in Figure 9.3.2.4:
educational-computing and Kevin
Bond’s Guest Network. These are
user-friendly names, commonly
known as Service Set Identifiers
/ Identities (SSIDs), used to
identify a service set to users of the
wireless network.
The user-friendly name, or SSID,
maps to the BSSID which is the
MAC address equivalent. It is
the BSSID that is sent in wireless
frames to identify the access point. Figure 9.2.3.4 IPad WiFi settings screen showing two APs

Single licence - Abingdon School 370

9 Fundamentals of communication and networking

Key concept
SSID:
A user-friendly network name,
e.g. educational-computing,
commonly known as a Service
Set Identity/Identifier/
Identification (SSID). It is used
to identify a Basic Service Set
(LAN) to users of a wireless
network.

Information Figure 9.2.3.5 Wireless radio frequency options for wireless network
with SSID educational-computing
Beacon frames:
In an infrastructure network,
Figure 9.2.3.5 shows the settings for an AP with SSID educational-computing.
the access point is responsible The AP automatically chooses either 2.4 GHz or 5 GHz depending on which
for transmitting beacon frames currently provides the better transmission, and then within the chosen band a
at regular intervals. Wireless particular channel1.
stations in range receive these
beacon frames and use them to Questions
find and identify a network. The
reception area for beacon frames 3 Name and describe four major physical components that may be
defines the basic service area. found in a wireless network.
All communication in an
infrastructure network is done 4 What is the purpose of an SSID?
through an access point, so
stations on the network must be
close enough to the access point Carrier Sense Multiple Access with Collision Avoidance
to receive beacon frames. (CSMA/CA)
Channel selection
Each access point (AP) operates on a given frequency channel, e.g. channel 36.
Both the 2.4 GHz and the 5 GHz frequency bands are divided into a number
of such channels, each with a predetermined width. Figure 9.2.3.6 shows that
the channels for the 5 GHz band are 20 MHz wide. For example, channel 36 is
centred on the frequency 5180 GHz and it covers the range 5170 to 5190 GHz.
5170 5330 5490 5710
MHz MHz MHz MHz
100

104

108

112

116

120

124

128

132

136

140

IEEE channel #
36

40

44

48

52

56

60

64

20 MHz bands

Figure 9.2.3.6 Operating channel bands and frequencies for 5 GHz wireless
transmission in Europe
In order for a station to communicate with its access point, it selects the
channel that its AP is using. This means that all the stations in the same basic

1 Explained in the next section.

371 Single licence - Abingdon School
 9.2.3 Wireless networking

service set, e.g. educational-computing, use the same channel. Therein lies a Did you know?
problem, as more stations join a basic service, the transmission rate usually The late Steve Jobs, former
goes down which is one reason why the maximum transmission rate is rarely CEO of Apple had difficulty
achieved. The AP will actually tell the stations to use a lower transmission rate connecting via WiFi at the
if there is too much traffic from too many devices. conference in 2010 that
launched the iPhone 4
Questions smartphone. Too many people
in the auditorium were already
5 What is meant by channel selection in wireless networking?
using the WiFi channel Steve
needed for his demonstration.
Interference Steve had to ask all the delegates
WiFi is prone to interference from other sources as well as from stations using a to disconnect their WiFi to
enable him to connect.
particular basic service because
• the frequency bands used by WiFi are unlicensed and so WiFi is Information
not the sole user of the spectrum. The spectrum is used by lots of Signal to Interference +
other equipment over which the basic service set has no control, e.g. Noise Ratio (SINR):
microwave ovens used to heat and cook food in the kitchen Radio wave communication
relies on differentiating a signal
• its maximum transmit power is restricted to a very low level because it from background noise and
is unlicensed interference. If the signal power
• it is typically used indoors where there are lots of objects to block the level falls below the noise +
interference power level, it
signal or reflect and echo a transmitted signal into the path of another
becomes more difficult to
Noise extract the signal. The ratio of
The circuits used in WiFi generate unwanted electrical noise (random electrical signal power level to noise +
fluctuations) which can mask signals of too low power. interference power level is a
measure of how easy or difficult it
Collisions is to extract a signal
When two stations, A and B are transmitting at similar times their frames will Signal power
SINR =
collide if they are within interference range of each other which is the case Interference + Noise power
when they both use the same access point (Figure 9.2.3.7). The outcome is
determined by the signal-to-interference + noise ratio (SINR) of
each. Access Point
Figure 9.2.3.8 shows the overlap of two frames one from A and one
from B. The greater the overlap with another frame, the higher the
chance that neither frame will be properly decoded at the receiver.

Signal Frame A
Strength sent by A Frame
sent by B B
A
B
Figure 9.2.3.7 shows the energy of
Overlap of two frames spreading through the
frames air and crossing at the access point
Time
Figure 9.2.3.8 Two frames overlapping in time. A is a stronger signal
than B
Single licence - Abingdon School 372
9 Fundamentals of communication and networking

Questions Key concept

6 What is meant by a collision in wireless networking? Collision:
A collision occurs when
two transmitters are within
interference range of each
Collision avoidance other, and they send at similar
The approach in wireless networking to this problem is to try to avoid collisions times. What actually collides
in the first place by requiring WiFi-enabled devices to be aware of whether other are frames, one from each
devices are currently transmitting, i.e. sensing other device’s transmissions. A transmitter. The result is that
neither frame will be properly
device can transmit only if others are not. It is a bit like a crossroads controlled
decoded at the receiver. A
by Give Way signs. You can proceed across the crossroads as long as you can see frame is a unit of digital data
that your intended path is clear of other traffic (Figure 9.2.3.9). If two road users transmission.
arrive at the crossroads at the same time, each can sense that it is not safe for both
to proceed if it will result in crossing each other’s path. Instead, each delays their
movement until they can interpret what the other’s intentions are. The delay is
a random amount of time whilst some form of mutual coordination for a safe
crossing is achieved.
Multiple access
The WiFi channel through which the WiFi signals travel is a shared medium,
shared between devices on this channel, e.g. channel 36. For this reason, we say
it is multi-access or a multiple access medium. Access must be coordinated and
controlled.
Figure 9.2.3.9 Crossroads
Carrier Sense Multiple Access/Collision Avoidance (CSMA/CA)
explained
The CSMA/CA protocol was designed to allow a station to send as long
as no other station is sending. It is called Carrier Sense Multiple Access
(CSMA) because each station tries to sense the presence of others on
the shared medium. It is called Collision Avoidance (CA) because each A B
station tries to avoid a collision by not sending when another station is
sending.
Figure 9.2.3.10 shows station A transmitting a frame to station B. The C
dotted red circle with A at its centre shows the reach of A’s transmission.
Station C is within this reach and so is able to sense that A is currently Figure 9.2.3.10 Station A transmitting
transmitting. If station C was also transmitting at the same time to, say to station B but with the transmission
a station D, but with enough signal strength to also reach station B, also reaching station C
then B might not be able to decode the transmission from A because
both transmissions interfere with each other in the receiver B’s electronics.
The CSMA/CA protocol requires that the receiving station, for whom the transmission is intended, sends back
an acknowledgement (ACK) signal to the sending station on successfully receiving and decoding the transmission
(Figure 9.2.3.11). This is how the sending station knows that its transmission got through. If an acknowledgement
is not received then the sending station will know that a collision has occurred and its transmission did not get
through.

373 Single licence - Abingdon School

 9.2.3 Wireless networking

The situation shown in Figure Key point

9.2.3.12 can arise when two Acknowledgement
stations, A and C that wish (ACK):
to transmit cannot detect the The ACK signal is
the only mechanism
transmissions of each other Fram
e of indicating that a
because they are not within each
transmission was
Time
other’s wireless reach. Station A successful.
may start transmitting but the Ack
If an acknowledgement
out-of-range C cannot sense doesn’t arrive, the
sender is to conclude
this and so starts transmitting
that the transmission is
as well. C’s transmission to D Figure 9.2.3.11 Timing for the sending of a frame, lost followed by the re-
reaches B at about the same time its processing at the receiver, and the sending of an sending of the frame.
that A’s does. A collision arises acknowledgement signal to the sender Error recovery is thus
at B between A’s transmission the responsibility of the
sending station.
and C’s which results in neither
transmission being decoded by B. B will therefore not
send an ACK to A. The use of an acknowledgement
signal enables stations to detect a collision and to take
remedial action which consists of sending the frame
again after a delay of an appropriate amount of time. A B C
We have simplified the scenario to make a point. In
an infrastructure BSS where everything goes through
an access point, B could be an access point shared by D
A, C and D. A may in fact be sending to C via B and
C could be sending, via B, to D. Figure 9.2.3.12 Station A transmitting to station B
and station C transmitting to D at a similar time with
Before sending, a station has to observe a wait and
Wait & neither A nor C able to sense the other
listen period. If during this period listen
the station does not detect any Frame to B
A
transmissions, it can start transmitting Wait &
listen
at the end of the period as shown for ACK
No ACK
station A in Figure 9.2.3.13. B to A

Wait &
listen Wait &
If a station that wants to send senses listen
Frame to B
that the channel is busy at any time C
during the wait and listen period, then Wait &
listen
the station does not transmit. C starts Frame to E
its wait & listen period just after A’s. It D

senses during this period that A starts

No ACK
sending. It waits a frame + ACK + a E
little bit more before starting another Time
Station
wait & listen period. Figure 9.2.3.13 Shows the timing for the sending of frames by A C, D
It is still possible that a collision can and an ACK from B to A, a collision occurs between frames from C and
occur even though a sending station D which is detected by both not receiving the corresponding ACK

Single licence - Abingdon School 374

9 Fundamentals of communication and networking

waited and found the channel idle before sending. Another station might have begun sending at the same time
because it began its waiting period at the same time and also concluded that the channel was idle. This is shown
in Figure 9.2.3.14 where both C and D start sending at the same time and cause a collision. However, neither
receives an acknowledgement No ACK
signal within the expected
window of time and so both Randomly
pick
conclude that an error has Station Wait & slot
a
occurred. listen 0 15
Frame to B
If both resend at the same time C
the same result ensues.
Wait &
The solution in this listen 0 15
Frame to E
circumstance is for stations C D
and D to employ a contention Randomly
Time
window of say size 15. Such pick
a slot
a window is divided into 15
Figure 9.2.3.14 Shows stations C and D in addition to a wait & listen
equal-sized time slots. Stations
period backing off a randomly determined number of time slots between
C and D then each choose one
0 and 15 after concluding by receiving no ACK that that a collision has
of the time slots at random,
occurred
e.g. C might choose slot 3 and
D slot 7 (there remains a small
chance that they could choose
the same slot). C will now listen Station
C

for the wait & listen period + 3

Station
time slots; D will now listen for A
RTS
fram
e
the wait & listen period + 7 time
Time
slots. If one or the other finds the RTS CTS ACK
Data me
CTS fra
channel idle after its wait then Data
fram
CTS e
each can send. Station
B
ACK
To prevent collisions 802.11 ACK fra
me

allows a station to use Request

To Send (RTS) and Clear To Figure 9.2.3.15 Using RTS and CTS to reserve channel for one station to
Send(CTS) signals to obtain send to another collision-free
exclusive use of the channel for
sending. This is necessary if many collisions are occurring such as when there Information
are many hidden stations. Too many collisions reduces transmission speed. The
Wireless communication
sending station sends an RTS frame to the target station. The target station
symbol for an antenna:
responds by transmitting a CTS frame. The sending station now sends the data
frame to the target. The target responds by returning an ACK frame.
Figure 9.2.3.15 shows the RTS frame, CTS frame, data frame and ACK frame forming a single atomic transaction
between sending and receiving stations.
If the target station receives an RTS it responds with a CTS.

375 Single licence - Abingdon School

 9.2.3 Wireless networking

The RTS silences stations within its range and the CTS silences stations within
its range. In this way collisions that result from the hidden station problem
shown in Figure 9.2.3.12 are avoided. Key concept
In Figure 9.2.3.15 station A is able to reach station B but not station C whilst
Security:
station B is able to reach both. Station A sends an RTS to station B, B responds Security means protecting
with a CTS which reaches both A and C. C now avoids sending until it receives against unauthorised access,
an indication that the transaction between A and B is over. This is the ACK alteration or deletion.
signal. Meanwhile A receives the CTS signal and proceeds to send the data
frame. A finishes sending and then waits for the ACK signal from B.
Key concept
Questions
Authentication:
7 What is meant by multiple access in wireless networking? Proving that the user is who
they say they are.
8 Explain the CSMA/CA protocol used in wireless networking.

9 What is the hidden node problem in wireless networking? Key principle

10 What is the purpose of RTS and CTS in CSMA/CA wireless Access control WPA/WPA2:
networking? To join a WPA/WPA2-secured
personal wireless network, a
user (client) has to successfully
11 RTS and CTS add extra time to a data transmission between two
negotiate an authentication
stations. Under what circumstance would they be used?
stage which checks that the
client knows a pre-shared secret
key (PSK).
Securing wireless networks
WPA/WPA2
Wi-Fi Protected Access (WPA) and Wi-Fi Protected Access II (WPA2) are two Key principle
security protocols developed by the Wi-Fi Alliance to secure wireless computer
Pre-shared secret key (PSK):
networks. WPA was a backwards-compatible temporary measure adopted before In WPA/WPA2 personal, the
WPA2’s development was complete. WPA/WPA2 replaced WEP which is easily access point and stations that
broken because it is a stream cipher which exclusive-ORs the data stream with a are allowed to join the wireless
fixed key stream (see RC4 Chapter 5.6.10). network share a secret key called
the Pairwise Master Key (PMK).
User’s data sent between two devices, e.g. a wireless station and an access point This is a 256-bit key (32 bytes).
needs to be private to those two devices, i.e. kept confidential by securing
against unauthorised access. Unfortunately, radio transmissions over a wireless
network are easily intercepted and read by third parties unless encrypted. Key principle
They are also open to spoofing, i.e. purporting to originate from a genuine user Generating the PMK:
when they don’t. Message authentication lets communicating partners who In WPA/WPA2 personal, the
256-bit Pairwise Master Key
share a secret key verify that a received message originates with the party who
(PMK) is generated from a
claims to have sent it. passphrase/password known
Messages can also be easily intercepted and altered in transit by a third party. to the user and the SSID. The
passphrase is a plaintext string.
Message integrity checks allow such alterations to be detected.

Single licence - Abingdon School 376

9 Fundamentals of communication and networking

Access control and authentication

To join a WPA/WPA2-secured personal wireless network, a user (client) has to successfully negotiate an
authentication stage which checks that the client knows a pre-shared secret key (PSK). This checking is based on a
message authentication code (Message Integrity and Authentication Code or MIAC, abbreviated further to MIC)
generated from the pre-shared secret key. The checking is done at the access point. The access point is responsible
for controlling access to the wireless network.
Pre-shared secret key
In WPA/WPA2 personal, the access point and stations that are allowed to join the wireless network share a secret
key called the Pairwise Master Key (PMK). This is a 256-bit key (32 bytes). Fortunately, users/clients don’t have
to remember this key. Instead, clients share a passphrase/password consisting of up to 133 ASCII characters which
is set up on the access point for a specific SSID. The PMK is generated by combining the SSID and this passphrase.
To join a known SSID network, a user enters the passphrase for this specific SSID at their wireless station. The
wireless station now has everything it needs to calculate its own copy of the PMK for this SSID network. For
example, if the passphrase/password is LetMeIn and the SSID is MyWirelessNetwork then the generated PMK
could be c4f9 400d 1cc7 cc3c 6b68 5b12 13a8 20dc
Pairwise Transient Key (PTK)
The PMK is never transmitted to avoid an unauthorised third party obtaining a copy.
How is it possible then for both access Pairwise Transient Key (384 bits for CCMP)
point and station to demonstrate that
they possess the same pairwise master EAPoL Key EAPoL Key Temporal Key
key (PMK) without sending each their
Confirmation Key Encryption Key (CCMP)
128 bits 128 bits 128 bits
copy of the PMK?
Figure 9.2.3.16 Pairwise Transient Key for Counter Mode with
The solution is for both wireless station CBC-MAC Protocol (CCMP)
and access point to use a Pairwise
Transient Key (PTK) derived from the PMK and to demonstrate to each other knowledge of this PTK.

Extension material beyond A-level

Knowledge is demonstrated using the Key Confirmation Key to produce a message authentication code which the station sends
to the access point. If more than one wireless station has joined the network then each station-access point pairing will have its
own PTK.
The Pairwise Transient Key is a collection of other keys as shown in Figure 9.2.3.16:
• Key Confirmation Key (KCK) – used to prove possession of the PMK
• Key Encryption Key (KEK) – used to encrypt the Group Transient Key (GTK)
• Temporal Key (TK) – used to secure data traffic once connection is established
• The PTK temporal key is used to secure unicast (communication between a single sender and a single receiver over a
network) data transmissions.
The Group Transient Key is used to secure multicast/broadcast transmissions.
The pairwise transient key CCMP uses CCM, a provably secure cipher based on an AES block encryption algorithm.
The particular algorithm used by the transmissions shown in Figures 9.2.3.17-21 is the 128-bit AES block cipher one, a very
secure cipher.

Questions
12 What are WPA and WPA2?
13 State and explain three reasons why wireless networks need to be secured.

377 Single licence - Abingdon School

 9.2.3 Wireless networking

Information (This material is not required for A-level

Four-way handshake Station 1

Communications begin with an Access

point
unauthenticated supplicant (client device, Supplicant

e.g. station 1) attempting to connect with Generate SNonce Generate ANonce

an authenticator (802.11 access point). The Message 1: Authenticator Nonce (ANonce)

client sends an Extensible Authentication Calculate PTK

Message 2:Supplicant Nonce (SNonce), MIC , Supplicant RSN IE
Protocol (EAP)-start message. This begins a
Calculate PTK
series of message exchanges called a four-way and if needed
generate GPK
handshake to authenticate the client. Figure Message 3:ANonce, MIC, RSC, encrypted GTK

9.2.3.17 shows a simplified version of this Install Temporal

Key
exchange. Message 4: Acknowledgement

Install Temporal
Message 1: Key

The authenticator sends an unencrypted

Figure 9.2.3.17 The Pairwise Transient Key (PTK)
message to the supplicant which contains
is computed at the station and the access point it is connecting
the authenticator-generated random
to so that each has a copy
number ANonce (Figure
9.2.3.18).
Nonces are random numbers
which are used once
(Number ONCE).
Message 2:
The supplicant knows ANonce
its own PMK, the value
of ANonce sent to it, its
own MAC address, the
supplicant’s MAC address
and its own nonce, SNonce
which it generates.
Figure 9.2.3.18 Message 1 frame captured with Wireshark®
It now has all it needs to Screenshot reproduced by permission of the Wireshark Foundation
generate its copy of the
pairwise transient key (PTK)
- see Figure 9.2.3.21. It
responds to the authenticator
by sending its SNonce in
unencrypted form across the SNonce

channel (Figure 9.2.3.19).

The authenticator now has all
Message Authentication & Integrity Code (MIC)
it needs to calculate its copy generated using the KCK key
of the PTK. from PTK

Supplicant RSN Information Elements (IE)

Figure 9.2.3.19 Message 2 frame captured with Wireshark

Single licence - Abingdon School 378

9 Fundamentals of communication and networking

Information
Message 2 includes a MIC. This is a message digest that has used the EAPoL Key Confirmation Key (KCK)
from the supplicant’s copy of the PTK. The authenticator now uses its copy of KCK and the received message
to calculate the corresponding
MIC. The two MICs, one received
and the other calculated, are
compared.
Message 3:
ANonce
If the MICs agree, the
authenticator sends an
acknowledgment message to the
Replay Sequence Counter
supplicant confirming that it has
Message Integrity Code
been authenticated and is now to prove data origin
authenticity
allowed to join the network and
Encrypted GTK encrypted
to install the PTK data encryption using KEK key
key (Figure 9.2.3.20). The Figure 9.2.3.20 Message 3 frame captured with Wireshark
authenticator awaits confirmation
from the supplicant that it has installed
the data encryption key (temporal key)
before it installs its copy. The same
GTK is used for all stations.
Message 4:
The supplicant responds with an
SNonce
acknowledgement message (Figure
9.2.3.21) confirming to the authenticator
that it has installed the temporal key
(data encryption key) that should be
used from now on to encrypt data
transmissions as well as to generate the
Figure 9.2.3.21 Message 4 frame captured with Wireshark
MIC to protect the integrity of the data
as well as authenticate its origin. The authenticator now installs its copy of the temporal key.

Task
1 Download and install Wireshark from www.wireshark.org on a computer with a wireless interface. In
Wireshark, select the wireless interface and enable monitor mode. Start capturing wireless frames whilst
at the same time connecting to a wireless network access point. Stop the capture once you are connected.
Set the filter in the main window filter to EAPoL so that you can see four messages similar to those above.
Expand the Authentication part of the frame and examine messages 1 to 4 in turn as above. What you
will see will depend on the wireless protocol that you have chosen and the cipher suite supported by your
wireless interface. Clear the EAPoL filter and search for Beacon, Probe Request and Probe Response frames.

379 Single licence - Abingdon School

 9.2.3 Wireless networking

Extension material (Beyond A-level)

Figure 9.2.3.22 shows a block diagram of the computation of the PTK.
A random number generator at the access point generates the first nonce (ANonce) and another random number
generator at the station generates the second Nonce (SNonce). ANonce is short for Authenticator Nonce and
SNonce is short for Supplicant Nonce. A new and different ANonce and a new and different SNonce are generated
when a station that has disassociated itself from the access point reconnects. This means that a new and different
PTK is generated.
For encrypting and authenticating data after
connection established. Called Temporal Key
Pairwise Master
key (PMK) Data encryption key (128 bits)
ANonce Computation
of EAPoL Key Encryption Key (KEK) 128 bits
SNonce
the PTK
Source MAC
EAPoL Key Confirmation Key (KCK) 128 bits
Destination MAC
For validating that client
Pairwise Transient Key (PTK) knows shared secret key
refers to all three keys For encrypting Group Transient Key
before temporal keys installed
Figure 9.2.3.22 Computing the Pairwise Transient Key (PTK) takes place at the station and the access point it
is connecting to so they both have a copy. EAPoL is Extensible Authentication Protocol over LAN

Information
Figure 9.2.3.23 shows two networked connected stations each with a different
Man-in-the middle attack:
paired temporal key (data encryption and MIC key).
In cryptography and computer
security, a man-in-the-middle
Station 1 Station 3 attack is an attack where the
attacker secretly relays and
pairwise possibly alters the communication
between two parties who believe
temporal they are directly communicating
key 1
temporal
key 1
with each other.
pairwise
Access temporal
point key 2
Information
temporal Replay attack:
key 2
In network security, a replay
attack is a form of network attack
Station 2 Station 4
in which a valid data transmission
A temporal key is used to encrypt data transmissions and to create MICs
to protect and authenticate the data is maliciously or fraudulently
repeated or delayed.
Figure 9.2.3.23 Station 1 and the access point are connected
One way that replay attacks are
with a pair of keys, temporal key 1; station 4 and the access point defeated is to use the sequence
are connected with a different pair of keys, temporal key 2. number of packets. If stations and
access points record the highest
Questions received sequence number then
14 Why is relying on PMK, source and destination addresses alone as they can reject packets with lower
input to the PTK computation not as secure as the method which sequence numbers which occur
with replayed packets.
includes two nonces?

Single licence - Abingdon School 380

9 Fundamentals of communication and networking

Key principle Media Access Control (MAC) address white list filtering
A wireless network could not use any form of encryption for its packets but
Media Access Control (MAC) instead rely on filtering of packets. MAC address white list filtering is one
address white list filtering:
such form of filtering. MAC addresses are 48-bit addresses uniquely assigned
In MAC address white list
filtering, the access point has an to each wireless network interface card. In MAC address white list filtering,
internal table of MAC addresses the access point has an internal table of MAC addresses which it consults to
which it consults to decide decide whether to permit access to the network or not. If the supplicant’s MAC
whether to permit access to the
address is on this list then it may join the wireless network controlled by this
network or not.
access point. If its MAC address is not on the list then the access point will
reject any attempt that the supplicant makes to join the network. Whilst MAC
address white list filtering gives a wireless network some additional protection,
Key principle MAC filtering can be defeated by a spoofer who learns the MAC address of
a valid wireless network interface card, i.e. one on the white list, by scanning
SSID broadcast disabled
protection: wireless traffic and then replacing a validated one with their own MAC. Task 1
Wireless stations require a with Wireshark should have revealed that MAC addresses do not get encrypted
knowledge of the SSID in when travelling over the air between computer and wireless access point. A
order to join the network. If MAC address is “glued” into a network card, but it is possible to command the
broadcast, the SSID appears in
operating system to change information about the MAC address in every data
the network settings window of
stations within range. packet it sends out to the network. In this way a spoofer could gain access to
In this form of protection, the white list protected network.
an access point disables SSID broadcast disabled protection
broadcasting its SSID to wireless
Access points have the option to disable broadcasting their SSID. This
stations. Thus, only clients
who already know the pre- means that the SSID will not appear in the client’s network settings window
configured SSID can establish a (see Figure 9.2.3.4). Clients who already know the pre-configured SSID
connection, others will not be can establish a connection, others will not be able to (without a bit of extra
able to (without a bit of extra
effort). Unfortunately, clients who already know the SSID cause the SSID
effort).
to be revealed to snoopers when establishing a connection with the access
point. Before the authentication stage begins, the client sends a Probe Request
message and receives a Probe Response from the access point in return as
shown in Figure 9.2.3.24. The (unencrypted) SSID is present in these packets,
therefore reducing the effectiveness of disabling broadcasting of the SSID.
To discover the SSID, a snooper might first send a deauthentication message
to the stations that are connected to force them to disconnect and reconnect.
Reconnecting should cause Probe and Response Request messages to be
broadcast which reveal the SSID.

Figure 9.2.3.24 Probe and Response Request frame/packets captured with Wireshark to reveal the SSID

Questions
15 What is (a) MAC address white list filtering? (b) SSID broadcast disabled protection?
16 Explain why both MAC address white list filtering and SSID broadcast disabled protection are insufficient
alone to protect a wireless network.

381 Single licence - Abingdon School

 9.2.3 Wireless networking

In this chapter you have covered:

■■ The purpose of WiFi
■■ The components required for wireless networking
■■ How wireless networks are secured
■■ The wireless protocol Carrier Sensing Multiple Access with Collision
Avoidance (CSMA/CA) with and without Request to Send/Clear to Send
(RTS/CTS)
■■ The purpose of Service Set Identifier (SSID)

Single licence - Abingdon School 382

9 Fundamentals of communication and networking

9.3 The Internet

Learning objectives:
■ 9.3.1 The Internet and how it works
■ Understand the structure of The structure of the Internet
the Internet The term ‘internet’ is a combination of ‘inter’ and ‘net’. The ‘net’ refers to a
■ Understand the role of packet computer network and ‘inter’ refers to interconnections between two or more
switching and routers
computer networks and computers or devices with computing capability on
■ Know the main components of
a packet these networks. This is how the Internet began back in the 1970s. The problem
■ Define: was how to connect packet-switched networks located in North America and
• router Europe. It was solved by Robert Kahn, Vint Cerf and others, and the Internet
• gateway was born. Any network of computer networks is an internet or internetwork.
■ Explain how routing is NHS workers use a secure private internet to access patient records. The
achieved across the Internet network used by the general public for e-mail and web page access is a
■ Describe the term ‘uniform special internet called the Internet. It consists of a network of interconnected
resource locator’ (URL) in the computer networks and computers using a globally unique address space
context of internetworking
(IP addresses) based on Internet Protocol (IP), and Transmission Control
■ Explain the terms ‘domain Protocol (TCP) to support public access to e-mail and web pages, among
name’ and ‘IP address’
other things. Figure 9.3.1.1 shows how computer networks and computers are
■ Describe how domain names
connected by the Internet.
are organised
Global internetwork or the Internet
■ Understand the purpose
and function of the University
Domain Name System campus North Central and
(DNS) and its reliance America South America
on the DNS server
system
■ Explain the service
provided by Internet Home
registries and why they user
International links
are needed. or core backbone

Internet Far East and

Service Pacific
Provider

Middle East Africa

School
network Europe

Continental and
national backbones

Figure 9.3.1.1 General architecture of part of the Internet

Single licence - Abingdon School 383

9 Fundamentals of communication and networking

Key term Europe, Africa, the Middle East, North America, Central and South America,
the Far East and the Pacific are linked by very high-speed connections which
Internet: A network of form the core backbone of the Internet. Each continent has a backbone of very
computer networks, computers
high-speed links which interconnect routers located in each country. Routers
and devices with computing
capability using globally unique are special packet switches that receive incoming packets of data along one link
IP addresses and TCP/IP. and send them as outgoing packets on another link.
Open architecture networking
The Internet uses open architecture networking. Designers are free to design
Key term
networks however they want, but all these different networks can be connected
Packet switching: to and communicate over the Internet because of the way that the Internet has
Messages to be sent are split
been designed. Each network is connected to the Internet through a router (a
into a number of segments
called packets. The packets
special router called a gateway router).
of a message are allowed to Until recently, the end-system devices in these networks that connected to the
travel along independent paths Internet were desktop computers and powerful servers but now a wider range of
through a network of routers.
devices are being connected to the Internet. These end-systems are referred to as
Routers use a packet’s
destination IP address to route hosts because they host (i.e. run) application programs such as a Web browser
the packet, taking account of program, a Web server program, an email client program, etc. Each host has a
how congested particular routes user-friendly memorable hostname, e.g. www.aqa.org.uk.
are.
This network resembles a fishnet Questions
of switching nodes called routers 1 Explain the term internet.
connected by links in a way
that allows multiple pathways 2 What is the Internet?
through the network between
endpoints. The role of packet switching and routers
The role of packet switching is to support end-to-end communication of a
message between two hosts located in different parts of an internet. Messages
split into smaller segments called packets travel along independent paths
Key term
through a network of packet switches called routers. Each packet sent by a host
Role of a router: contains the IP (Internet Protocol) address of the destination host. Each router
Routers are special packet uses the destination IP address to choose from among its outgoing links one
switches that receive incoming
along which the packet can reach its destination.
packets of data along one link
and send them as outgoing Within national boundaries, networks belonging to large businesses and
packets on another link. organisations such as universities are connected directly to the national
backbone. Smaller organisations and home users of the Internet connect to
an Internet Service Provider (ISP), which connects to the national backbone.
Figure 9.3.1.2 shows three networks connected by routers.
The design of the Internet is based on the Catenet (internet) concept of a
network of networks. This concept describes data packets flowing essentially
unaltered throughout an internet with their source and destination addresses
(IP addresses) that of the endpoint systems (now referred to as end-systems)
sending and receiving the packets, respectively.

384 Single licence - Abingdon School

 9.3.1 The Internet and how it works

The key idea of a packet

switched network is built- Endpoint
in redundancy supporting Y
Local Area
multiple pathways between Network
(LAN)
endpoints. The concept of Host

a packet switched network

Router
occurred quite independently
to two researchers, Paul
Link
Baran (USA) and Donald
Endpoint
Davies (UK) around the Link X
Router
same time in the early 1960s.
Both advocated building Router Router
a distributed network that Figure 9.3.1.3 Distributed
looked rather like a fishnet - network of switching
Host Host
Figure 9.3.1.3 - consisting of nodes (routers) resembling
switching nodes connected a ‘fishnet’
Local Area Local Area
by links in such a way that Network Network
(LAN) (LAN)
allowed multiple pathways
through the network between
Figure 9.3.1.2 Connecting three LANs
endpoints, e.g. endpoint X
by routers
and endpoint Y.
Baran’s and Davies’ second proposal was to introduce redundancy when
sending messages. The messages were to be split into a number of fixed-length
“message blocks” which Davies christened packets.
The radical idea was to allow the packets of a message to travel
along independent paths through the network of routers.
Any packet that did not get through was
sent again. The sending station would
wait a certain period of time for an
acknowledgement packet to be sent by the
receiving station. If one was not received
within this wait-time then the sending station
would send a copy of the packet again but in
a different direction through the network.
In 1961 Leonard Kleinrock showed that
packet switching was a better switching
method than circuit switching1. Figure
9.3.1.4 shows Leonard Kleinrock
photographed standing before an early packet
switch called an Interface Message Processor
(IMP). Figure 9.3.1.4 An early packet switch (reproduced with kind
permission of Professor Leonard Kleinrock)
1 Not in AQA A-level specification
Single licence - Abingdon School 385
9 Fundamentals of communication and networking

Circuit switching connects two endpoints with a complete end-to-end electrical circuit for communication by
exclusively allocating all the switches and their links along a single pathway for the duration of the communication.
No other endpoints can use the same pathway if it is already in use.
By contrast, packet switching allows packets from different messages to use the same
nodes and links at the same time. This makes better use of the network’s bandwidth,
Question
defined as its maximum capacity for sending information. 3 Explain the role of
packet switching
The built-in redundancy of packet switching means better congestion handling, e.g. if a
and routers in a
particular route is busy then packets may be rerouted along a different, less busy route.
packet switched
It also means that the network can withstand switching nodes going down. This means
network.
that cheaper, less reliable nodes may be used.
The Internet uses packet switching.
Packet transmission
The packet is the unit of communication in the Internet.

In its simplest form, a packet consists of three parts: Source address Destination Application
address data
When computer X, connected to the Internet, wishes
to send a message or a document to computer Y, also connected to the Internet, computer X splits the message or
document into chunks; Figure 9.3.1.5 shows five chunks: A, B, C, D and E. Computer X then generates as many
packets as there are chunks, placing each chunk in the application data part of the next available packet. The unique
address of the sending computer is placed in the source address part of each packet and the unique address of the
receiving computer is placed in the destination address part of each packet.
Each packet is then dispatched to the Internet through a router. The packets are sent independently through a
series of interconnected routers until they reach their destination. Each router examines the destination address of a
packet it receives to determine what to do with it. Computer Y could reply to computer X by a similar process.
Original
message
EDCBA Network
Message
packets
E D C B A D C A D A
1 3 5

D
Computer
X E B C
Computer
A Y

E E C B D A E C B
2 4 6

EDCBA
Router node Re-assembled
message
Figure 9.3.1.5 Routing of packets A, B, C, D and E through a packet-switched network

386 Single licence - Abingdon School

 9.3.1 The Internet and how it works

The end-to-end principle Key term

Cerf and Kahn proposed that the two communicating computers, X and Y
in our example, should be the endpoints of the communication. The end-to- End-to-end principle:
The end-to-end principle states
end principle states that the two endpoint hosts should be in control of the
that the two endpoint hosts
communication. The role of the Internet is to move packets between these two should be in control of the
endpoints. communication. The role of
the Internet’s packet switched
This has several advantages:
network is to move packets
• The sending application in Computer X and the receiving application between these two endpoints.
in Computer Y are able to survive a partial network failure. The
failure is detected and the packets that did not get through are resent.
• Packets can be rerouted around failures very quickly and sent along
alternative paths.
• The Internet can grow easily because control resides in the endpoints
(end-systems, e.g. Computer X) not in the Internet.
• There is no requirement for Internet routers to notify each other
as endpoint connections are formed or dropped; this simplifies the
design of routers.
• The integrity and security of each packet sent is handled by the
endpoints (end-systems), which simplifies the role of the Internet.
Questions
• Each endpoint need only be aware of the router to which it is directly
connected and, optionally, a name resolution service that converts 4 Explain the end-to-end
user-friendly hostnames (Computer X) into their corresponding IP principle.
addresses. 5 State five advantages
Single logical address space of an internet designed
The end-to-end principle requires that each computer using the Internet should using the end-to-end
be uniquely identified. Cerf and Kahn proposed that each computer be labelled principle.
with a globally unique address known as an IP address. Their numbering
system, called IPv4, is used today and allows 232 different addresses. All these Key term
unique addresses make up a single logical address space. At the binary level, an
IPv4:
IPv4 address consists of 32 bits (4 bytes). Internet numbering system of
Cerf and Kahn split an IP address into two parts (Figure 9.3.1.6): unique IP addresses that make
up a single logical address space.
• bits that identify the network connected to the Internet (NetID) and IPv6 will eventually replace
• bits that identify a host (strictly speaking a network interface) IPv4. IPv6 is also an Internet
numbering system like IPv4 but
connected to the network (HostID).
it consists of 128 bits whereas
The thinking behind this was that since the Internet is made up of networks, IPv4 has only 32 bits.
being able to identify each network would help routers enormously in the task
of routing packets to the correct destination network.
31 0
An Internet address can be expressed in dotted decimal notation,
where each byte of the 32-bit IP address is written in decimal, NetID HostID
separated by a dot:
Figure 9.3.1.6 IPv4 address structure
196.100.11.4
The network ID might be 196.100.11 and the host ID, 4.
Single licence - Abingdon School 387
9 Fundamentals of communication and networking

Router Key point

Routers are used because it is not practical to connect every host directly to every
Difference between a router
other host. Instead a few hosts connect to a router, which connects to other and a switch:
routers, and so on, to form a network (Figure 9.3.1.2). A router receives packets from
a host or router on an incoming
A router receives packets from one host or router and uses the destination IP
link and uses the destination
address that they contain to pass on the packets, correctly formatted, to another IP address that they contain to
host or router. pass on the packets, correctly
formatted, to another host or
Figure 9.3.1.7 shows the hierarchy of routers for a single country. Each router
router connected to an outgoing
in this hierarchy maintains a table of other routers, computers and networks it is link. The software that does this
directly connected to and enough information about the hierarchical structure of is located in the network layer of
the Internet to route a packet to the desired destination. the protocol stack. The router
acts as a kind of packet switch
For example, the IP address range 202.0.0.0 to 203.255.255.255 has been with the ability to determine,
allocated to the Asia-Pacific region. A host on a school network in England on the basis of IP address, which
wishing to communicate with a host on a network in Malaysia will send packets to outgoing link of several to use.
the router on the school network. This router will pass the packets onto the local
A data link switch such as an
router it is connected to. The local router will pass these packets on to the regional
Ethernet switch has several ports
router it is connected to. The regional router will pass packets on to the national but each switch port connects to
router it is connected to, and so on, until the router of the destination network is one device at most. The switch
reached. Each router in the path uses a part of the IP address to make the routing learns the hardware address of

decision. In this example the decision is to route packets up the national hierarchy each connected device and the
port it is connected to. It is then
because 202/203 addresses are outside England.
able to map a packet’s destination

International hardware address to a port to use

that is connected to the device
with this hardware address. The
National software that does this is located
in the link layer of the protocol
Router stack.
Regional

Local

Company
/School
/College/etc
networks
Figure 9.3.1.7 Routing hierarchy for one country

Questions
6 Explain why routers are used.

7 What is a router?
8 How are routers organised in the Internet?

388 Single licence - Abingdon School

 9.3.1 The Internet and how it works

Gateway
Gateways are special routers that provide an interface between two or more dissimilar networks. Gateways translate
one network layer protocol into another, translate link layer protocols, and open sessions between application
programs.
Local Area Network Local Area Network
Gateways allow two or more A B

networks that use different link or

network protocols to be connected Computer 2
Computer 5
so that information can be passed
Computer 3
from one system to another. Computer 4

For example, in Figure 9.3.1.8 Computer 1

Gateway 2
a Local Area Network (LAN) is
Gateway 1
connected to the Internet through
a gateway, Gateway 1, and another
LAN is connected through another
gateway, Gateway 2. The LANs use
a protocol that differs in certain Wide Area Network

respects from the Wide Area Figure 9.3.1.8 Gateway connections to the Internet
Network (WAN) protocols used
on the Internet. The gateway does the job of translating the LAN frame into its equivalent WAN frame and vice
versa. Sometimes gateways are called gateway routers.
Another common use of gateways is to enable LANs that use TCP/IP and Ethernet to communicate with
mainframes that use other protocols.

Question
9 What is the purpose of a gateway?

The main components of a packet

Messages or application data are split into chunks and the chunks together with headers are sent in packets through
the Internet’s packet-switched network from an instance of an application, e.g. Web browser, running on a host one
endpoint, to an instance of an application, e.g. Web server, running on a host at another endpoint.
The two hosts on which the applications are sending and receiving must be identified by numbers called source
and destination IP addresses. Routers will use these IP addresses to route the packets through the packet switched
network.
Two applications sending to and receiving from each other, are identified by numbers called port numbers.
Hosts use the port numbers to allocate the received packets to the corresponding destination application and to
send the required reply to the corresponding sending application.
The packets may be received out of sequence. Therefore, to re-assemble them in the correct order, they each need a
sequence number with the first message packet numbered 1, the second, 2 and so on.
Received packets are acknowledged if the endpoint-to-endpoint connection requires it. For this, each packet
belonging to a message is assigned an acknowledgement number before it is sent. The receiver replies with a packet

Single licence - Abingdon School 389

9 Fundamentals of communication and networking

to the sender constructed with the received acknowledgement number to indicate that the packet with this number
was received successfully.
Packet transmission errors can occur for a variety of reasons but it must be possible for the receiver to detect when
errors have occurred. A checksum (e.g. CRC) attached to each packet is used for error detection.
Packets travel along transmission media consisting of wires, fibre-optic cable and radio frequency links. To launch
a packet onto or to receive a packet from any of these requires a layer of hardware called a network interface.
Each sending/receiving host’s network interface hardware needs to be assigned a “fixed” unique hardware address
called a link layer address. Similarly, each router will need link layer hardware, uniquely identified by a link layer
hardware address, because routers are also connected to transmission media.
A packet is structured into a series of headers which contain all of the above information and the message chunk or
data - Figure 9.3.1.9. The data D and the headers HT, HN, HL make up each packet (and a checksum appended to
the end of the packet).

Web Web
browser server
Application data D Application HTTP HTTP
Port No & Port No &
TCP/IP
protocol
Segment HT D Transport Error control Error control
IP address
stack Datagram HN HT D Network IP
address
Frame HL HN HT D Link hardware
address
hardware
address

Physical Cable/radio
frequency link
Packet Cable/radio
frequency link

Figure 9.3.1.9 The protocol stack for TCP/IP showing how different headers are added to the payload data
- see Chapter 9.4.1

0 15 16 31 TCP header
16-bit source port number 16-bit destination port number
Figure 9.3.1.10 shows some of the detail in the header
32-bit sequence number labelled HT in Figure 9.3.1.9. HT is called the TCP
32-bit acknowledgement number 20 bytes header. The key fields to note are the port numbers and
4-bit header the sequence number fields.
length
16-bit TCP checksum IP header
Figure 9.3.1.11 shows some of the detail in the header
Figure 9.3.1.10 TCP header, HT labelled HN in Figure 9.3.1.9. HN is called the IP
0 15 16 31 header. The time-to-live field is used in order to stop
4-bit header 8-bit type of 16-bit total length (in bytes) of IP
length service (TOS) datagram a lost packet wandering the Internet forever. The
key fields to note are the source and destination IP
8-bit time to 16-bit header checksum addresses and the time to live fields.
20 bytes
live (TTL)

32-bit source IP address Frame header

32-bit destination IP address Figure 9.3.1.12 shows an example of a link layer
header for Ethernet. In Ethernet, a CRC checksum,
Figure 9.3.1.11 IP header, HN
which is not shown in the figure, is added after the

390 Single licence - Abingdon School

 9.3.1 The Internet and how it works

application data. The length field value is for the entire frame including application data but excluding the CRC
checksum.

48-bit destination address 48-bit source address 16-bit length 16-bit type
field field
Figure 9.3.1.12 Link layer header, HL
The type field identifies the type of data, e.g. ARP request which maps an IP address to a hardware address. The
destination and source addresses are hardware addresses usually assigned to the network interface card by its
manufacturer. In the Ethernet protocol, the hardware addresses are known as MAC addresses or Media Access
Control addresses. The key fields to remember are the hardware destination and source address fields.
Figure 9.3.1.13 shows the main components of a packet.
Hardware Hardware Source IP Destination IP Source Dest. Sequence
Data
destination address source address address address port no port no no

Figure 9.3.1.13 Main components of a packet Questions

10 What are the main
components of a packet?
How routing is achieved across the Internet
Figure 9.3.1.14 shows, for a simplified network scenario, the routing of a packet from end-system host X to end-
system host Y.
National ISP
centre

Router R1
Network
End-system host X Link

Application Router R2
Transport Network
Network Link
Link
Home network

Local or
Symbol for a router
regional ISP
centre
Symbol for a switch Network
Link

Symbol for a modem Router R3

Router R4
Network
Switch S1 End-system host Y
Link
Application
Link
Transport
Network
Link

Network
Link

Router R5

Company network
Servers

Figure 9.3.1.14 Routing of a packet from end-system X to end-system Y

through the Internet by network IP address

Single licence - Abingdon School 391

9 Fundamentals of communication and networking

Routing refers to the network-wide process that determines the end-to-end paths that packets take from source to
destination. Routing is necessary when the destination is not directly reachable.
A router receives a packet on an input link and transfers the packet to the appropriate output link. This is known as
forwarding. Every router has a forwarding table. A router forwards a packet by using the destination IP address in
the arriving packet’s header as an index to an entry in the router’s forwarding table. This entry indicates the router’s
outgoing link to use to forward the packet.
Using a driving analogy, routing would be equivalent to an ordered list of roads that need to be travelled to get
from place X to place Y. Forwarding would be equivalent to choosing the exit to take at the junctions/roundabouts
connecting each road segment to the next.
Routers play a crucial role in both the process of forwarding and the process of determining the paths to follow. A
router uses algorithms known as routing algorithms to calculate these paths and perform packet switching when
forwarding a packet. The routing algorithm determines the values that are inserted into the forwarding table of
each router. In the decentralised model, each router executes a part of a distributed routing algorithm relevant to its
location in the hierarchy of routers (see Figure 9.3.1.7).
In Figure 9.3.1.14, host X is sending a message to host Y. The message might be an HTTP GET message to an
HTTP server Y (see Chapter 9.3.2). The transport layer of the TCP/IP protocol stack in host X adds a transport
header to the message packet generated by the application layer to produce a transport layer segment. This segment
then passes to the network layer which adds its own header to it to create an IP datagram which then passes to the
link layer. This layer adds its own header and a checksum trailer to create a link frame. This frame is delivered to the
physical medium where it travels to router R1.
R1 strips away the link layer header and trailer to extract the IP datagram which then passes to the network layer.
The network layer consults its forwarding table to determine that the datagram should be sent onto the outgoing
link that connects R1 to router R2. The datagram is passed to the link layer that adds the correct link layer header
and trailer to create a new frame. The link layer then passes the frame to the physical medium which transfers it to
router R2.
The process that occurred at R1 now repeats at router R2. R2 consults its forwarding table to find which outgoing
link to use, i.e. the one joined to router R3. This process continues until the frame arrives at the end-system host Y.
At host Y, the headers are stripped away progressively as the packet is passed up the protocol stack until only the
message remains. The last header, the transport header, is removed by the transport layer before the message is
delivered to an application layer process, e.g. an HTTP server.
Routers require just two layers for this task: the network and link layers of the protocol stack.
Note that the source and destination IP addresses set by host X remain the packet’s source and destination IP
addresses throughout the journey through the routers from host X to host Y. However, the link layer hardware
source and destination hardware addresses are changed for each hop.
Table 9.3.1.1 shows a simplified example of a forwarding Destination address
Link interface
table for a router with four links, numbered 0 to 3. The router prefix
has IP address 146.97.33.2. It connects to hosts with dotted- 10001110 0
10010000 1
decimal IP addresses beginning with 142, 144, 152, 155 via
10011000 2
other routers.
10011011 3

Table 9.3.1.1 Simplified forwarding table

392 Single licence - Abingdon School

 9.3.1 The Internet and how it works

For example, www.southamption.ac.uk has IP address 152.78.118.52. The router with the forwarding table shown
in Table 9.3.1.1 routes packets it receives with destination IP address 152.78.118.52 to the link interface numbered
2.
Questions
11 Explain how routing is achieved across the Internet.

12 Which addresses remain unaltered throughout routing and which change?

13 Explain why some addresses must change and some must not.

Uniform Resource Locator

A uniform resource locator (URL) is a short string that represents the target of a hyperlink; it was introduced in
early 1990 in Tim Berners-Lee’s proposals for hypertext. A URL specifies which server to access, the access method
and the location in the server. Figure 9.3.1.15 shows that a URL consists of several parts. The simplest version
contains three parts:
The address in this case of
• How: defines which Microsoft’s World Wide Web
protocol is to be used server - the Where
• Where: defines the host
https://www.microsoft.com/en-us/WorldWide.aspx
• What: specifies the name
of the requested object and
the complete path to it. Protocol to be used. In this case The path on the addressed
it is HyperText Transfer Protocol server - the What
Secure - the How
Figure 9.3.1.15 Structure of a URL

Questions
14 What is a URL and what does it specify?

15 Give an example of a URL, different from the one above. Identify its how, where and what parts.

Domain name and IP address

In the early days of the Internet, users of Internet applications such as e-mail were required to enter IP addresses
when they wanted to set the destination and source addresses of an e-mail they were sending.
An IP or Internet Protocol address in the context of the Internet is a globally unique, 32-bit (IPv4) or 128-bit
(IPv6), logical address which identifies a host connected to and directly reachable from the Internet. This wasn’t
a problem while the number of IP addresses in use was very small. However, as the number of networks began to
grow, it became a lot harder to use IP addressing directly.
The Domain Name System (DNS) was invented so that users could use a memorable name called a domain
name to refer to a network and a Fully Qualified Domain Name (FQDN) to refer to a host on that network. For
example, the IP address 144.173.6.226 has the memorable fully qualified domain name emps.exeter.ac.uk. (FQDN
end with a full stop but this is often omitted). For convenience, people often use ‘domain name’ when they mean
fully qualified domain name.
Single licence - Abingdon School 393
9 Fundamentals of communication and networking

Questions
16 Why was the Domain Name System invented?

How domain names are organised

The Domain Name System (DNS), part of which is shown in Figure 9.3.1.16, is a hierarchical system of names
and abbreviations. The root is abbreviated to a full stop. Using this domain name hierarchy, an example of a
domain name is ags.bucks.sch.uk. This domain name was used to identify a network of computers with network
ID 195.112.56 located at a school in Buckinghamshire in the UK when the school hosted its own servers.
mail

emps ags

exeter bucks

in-addr Second-level domains co ac ltd sch plc gov

Generic World Wide Domains Generic U.S.

only Country
First
Arpa Com Edu Net Org Int Gov Mil AD UK or top
level

Root
Figure 9.3.1.16 Domain Name System hierarchy
A particular host on this network was mail.ags.bucks.sch.uk. This name (hostname) is an example of a fully
qualified domain name (FQDN). An FQDN uniquely identifies a host. When the host ID of this computer, 124,
is added to the network ID, the IP address becomes 195.112.56.124.
In mail.ags.bucks.sch.uk. the domain uk includes all hosts that use the top-level domain name suffix ‘uk’.
The second-level domain sch includes all the hosts that use ‘sch’, the third-level domain bucks includes all the hosts
in Buckinghamshire, and the fourth-level domain ags includes all the hosts in the organisation AGS with globally
unique IP addresses. Finally, mail is the public name of the host. Table 9.3.1.2 shows the interpretation of some
top-level domain names. Table 9.3.1.3 shows the interpretation of some second-level domain names.

Domain Type of Domain Type of

Name organisation Name organisation
com Commercial co Commercial
edu Educational ac Academic, higher education or further education
org Non-commercial sch School
uk Located in UK uk Located in UK
Table 9.3.1.2 Some top level Table 9.3.1.3 Some second level domain names
domain names
394 Single licence - Abingdon School
 9.3.1 The Internet and how it works

Questions
17 How is the domain name system organised?
18 Give two examples of top level domain names and two examples of second level domain names.
19 Give the meaning associated with each domain name given in Q18.

Purpose and function of the Domain Name System (DNS) and DNS Servers
People prefer to use the user-friendly memorable form of hostname, i.e. its fully qualified domain name, e.g.
university-of-exeter.ja.net.
However, routers prefer fixed-length IP addresses. In order for the needs of each to be met a directory service is used
that translates hostnames to IP addresses e.g. university-of-exeter.ja.net → 146.97.144.42.
This is the main purpose of the Internet’s Domain Name System (DNS).
Domain Name System servers translate FQDNs into IP addresses.
The DNS is a distributed database implemented in a hierarchy of DNS servers, and an application-layer protocol
that allows hosts to query the distributed database to resolve domain names into IP addresses before connecting to
other hosts on the Internet.
No single DNS server has all the mappings for all the hosts in the Internet. Instead the mappings are distributed
across the DNS servers.
DNS also provides a few other services in addition to translating hostnames into IP addresses:
• Host aliasing - A host with a complicated hostname can have one or more alias names. For example,
www.exeter.ac.uk and admin.ex.ac.uk are alias hostnames for a server with hostname webdata02.
ex.ac.uk and IP address 144.173.6.226 that hosts University of Exeter’s website as well as other sites. The
hostname webdata02.ex.ac.uk is said to be the canonical hostname (CNAME).
• Mail server aliasing - user-friendly memorable email addresses such as [email protected] are aliases.
The hostname of the mail server hotmail.com that is used for Fred’s email will almost certainly be more
elaborate than this, e.g. the canonical hostname of the alias hotmail.com is actually
origin.sn145w.snt145.mail.live.com with server IP address 155.55.152.112. DNS can be used by a mail
application, e.g. Microsoft Outlook, to obtain the canonical hostname for a supplied hostname as well
as the IP address of the host. Aliasing also allows a Web server and a mail server to have identical aliased
hostnames: e.g. company.co.uk.
• Load distribution - servers such as Web servers may be replicated if they host busy sites. Each server
runs on a different end system with each having a different IP address. The set of IP addresses for these
server end-systems is associated with just one canonical hostname. The DNS database contains this set
of addresses. A DNS server responds with the entire set of IP addresses when hit with a client request to
translate a hostname into an IP address. Clients normally pick the first IP address in the set they come
across. The DNS server rotates the order of IP addresses in each reply so that the traffic is distributed
amongst the replicated Web servers (the same is true of mail servers).
Questions
20 What is the main purpose of the Domain Name Service?
21 What services are performed by DNS servers?

Single licence - Abingdon School 395

9 Fundamentals of communication and networking

A very useful tool to use when exploring the DNS system is the Unix DNS tool dig. It is also available to use in
Linux and Mac OS X operating systems. The Raspberry Pi2 computer connected to the Internet may be used to
explore the use of dig but dnsutils will need to be installed first as follows:
sudo aptget install dnsutils
Another useful tool available in both Microsoft Windows and Unix/Linux/Mac OS X systems is nslookup. This
tool maps domain name to IP address.
For example, nslookup www.aqa.org.uk returns the answer 194.34.8.20.
The tool dig may be used to query DNS name servers as shown in Table 9.3.1.4.

Command Description Result

dig NS . Returns the hostnames of the name servers located at the d.root-servers.net.
root of the domain name system. The root is represented e.root-servers.net.
by a full stop. f.root-servers.net.
Note that there are thirteen root servers. g.root-servers.net.
h.root-servers.net.
NS means name server.
i.root-servers.net.
j.root-servers.net.
k.root-servers.net.
l.root-servers.net.
m.root-servers.net.
a.root-servers.net.
c.root-servers.net.

dig NS uk. Returns the hostnames of name servers for the domain dns1.nic.uk.
name uk. located in the top level of the domain name nsd.nic.uk.
system. Nominet UK is responsible for nic.uk. and several others.
dig NS ac.uk. Returns the hostnames of name servers in the domain with auth03.ns.uu.net.
domain name ac.uk. and located in the second level of the ns4.ja.net.
domain name system. and several others.
dig NS bris.ac.uk. Returns the hostnames of name servers for the domain irix.bris.ac.uk.
name bris.ac.uk. located in the third level of the domain ncs.bris.ac.uk.
name system. ns3.ja.net.
dig A irix.bris.ac.uk. Returns the IP address of the DNS server irix.bris.ac.uk.
137.222.8.143
A means address.
dig A snowy.cs.bris.ac.uk. Returns the IP address of the host with hostname snowy.
137.222.103.3
cs.bris.ac.uk.
dig +trace www.aqa.org. This will report all the DNS servers that are consulted in 194.34.8.20
uk @a.root-servers.net. resolving www.aqa.org.uk. NOTE a.root-servers.net (AQA Education
must end with a full stop. 194.34.8.0/24)

Table 9.3.1.4 Using the dig command to query the DNS system
For more examples of the use of the dig command see
www.cyberciti.biz/faq/linux-unix-dig-command-examples-usage-syntax

2 Raspberry Pi is a trademark of the Raspberry Pi Foundation

396 Single licence - Abingdon School
 9.3.1 The Internet and how it works

Tasks
1 Use the URL http://simpledns.com/lookup-dg.aspx to access a DNS delegation trace utility. Use this
utility to trace the DNS server queries for a host snowy.cs.bris.ac.uk. Note that the trace begins with a DNS
root server chosen from the list of possible root servers, and proceeds down the hierarchy of name servers
until the IP address of snowy.cs.bris.ac.uk is obtained.

2 Use the nslookup command to look up the IP address of the following www.exeter.ac.uk, www.
manchester.ac.uk, www.bristol.ac.uk.

3 Use the dig command to find the hostnames of third level DNS servers for the University of Exeter (hint:
use the information in Table 9.3.1.3 to target exeter.ac.uk.). Repeat the exercise for the Computer Science
department, University of Washington in the USA (hint: target cs.washington.edu.). If you do not have
access to the dig command you can use the site referenced in Question 1.

4 Install PingPlotter - https://www.pingplotter.com. Choose the following targets: www.exeter.ac.uk, www.

manchester.ac.uk, www.bristol.ac.uk. In each case observe the route. Does each route have anything in
common? Also make a note of each destination IP address then use each recorded destination IP address in
the url window of a browser in turn. Why might some IP addresses return a web page and others a Request
rejected message?

5 Use ipconfig/all in Windows command line to discover your computer’s IP address, the default gateway
and the physical/hardware address (MAC address) of its network card.
6 Install Wireshark - https://www.wireshark.org. Familiarise yourself with Wireshark by capturing a few
packets and examining their contents.

Internet registries
Private companies and organisations called Internet registrars are responsible for registering Internet domains and
therefore domain names to people, businesses and organisations, domain names such as educational-computing.
co.uk. A registrar is an online retailer where domains (domain names) can be bought.
The world is divided into five geographical regions for the purposes of Internet registries: Canada, USA, and some
Caribbean Islands (ARIN), Africa (AFRINIC), Asia/Pacific Region (APNIC), Europe, the Middle East and Central
Asia (RIPE), Latin America and some Caribbean Islands (LACNIC).
Regional Internet Registrars delegate responsibility for registering domain names to their customers, which include
Internet Service Providers and other organisations. The Regional Internet Registry for Europe is RIPE (Réseaux IP
Européens).
RIPE has delegated to Nominet to hold the official registry for all .UK domain names. Nominet is therefore an
Internet registrar. Nominet provides a
WHOIS tool that can be used to find out Task
if a .UK domain name is registered and 7 Use the WHOIS service of Nominet (www.nominet.uk/whois)
if it is, provide details of the registration to look up the registration of the following: (i) co.uk. (ii) ac.uk,
including the registrant. Nominet sets (iii) org.uk. (iv) plc.org. (v) me.uk. (vi) ags.bucks.sch.uk.
the policies and rules that relate to the (vii) commonweal.co.uk. (viii) lordwilliams.oxon.sch.uk.
management of .UK. It does delegate

Single licence - Abingdon School 397

9 Fundamentals of communication and networking

Question registration to third parties, e.g. RM Education PLC, but third party registrars
22 are required to follow Nominet’s policies and rules in respect of .UK domains.
What is an Internet
registry? Internet registries store registered domain names and the details of the
registrants, e.g. domain name educational-computing.co.uk has been
Key term registered by Educational Computing Services Ltd, the registrant.

Internet registries: The registrant or their ISP will supply the IP address of a DNS server(s) (the
Internet registries store registrant’s or the ISP’s) for this domain name to their Internet registrar. This
registered domain names registrar will then place an entry for this domain name and supplied IP address
together with the details of their at the corresponding level in the Domain Name Server System.
registrants.
Registering a domain name and associating it with a range of IP addresses
Information Suppose that you wanted a domain name for a group of servers, including a
Web server, that you intend managing yourself.
ASN:
The Internet is organised into a You have chosen myowndomain.co.uk as your domain name.
network of autonomous systems
You have obtained Internet connectivity, by contracting with and connecting
each of which manages their own
internal routing. Autonomous to, a local ISP (Internet Service Provider). You have purchased a gateway router
systems are identified by a unique which will be connected via DSL (Digital Subscriber Line) to a router in your
autonomous number (ASN that is local ISP.
stored in the databases maintained
by the Internet address registries. Your local ISP has granted you a block of IP addresses, one of which you will
The web page assign to the Web server (www.myowndomain.co.uk), one to the gateway
https://ipinfo.io/countries/gb router (gateway.myowndomain.co.uk) and another to the DNS server (dns.
shows the ASNs for the UK and
myowndomain.co.uk).
to whom they are assigned, e.g.
AS2856 is assigned to BT public You will need to check with an Internet registrar that your domain name,
Internet Service and covers a myowndomain.co.uk, has not been registered already. If it hasn’t, it may now
block of 11,215,616 IP addresses. be registered with your Internet registrar.
AS204160 is assigned to
SAMKNOWS-LTD and covers a You will also need to provide the IP address of your DNS server to your
block of 16 IP addresses. Internet registrar. Your registrar will then place an entry for your DNS server
(domain name and corresponding IP address) in the .co.uk second level
domain servers. After this is done, the IP address for your domain name and
therefore your DNS server can be obtained via the DNS system on request.
You must provide entries in your DNS server that map the hostname of your Web server, e.g. www.myowndomain.
co.uk to its IP address. You will need entries also for all your other publicly available servers.
Suppose that your ISP provided you with a block of 8 IP addresses expressed as 144.173.6.176/29 (a prefix). This
is interpreted as 29 bits for the network ID and 3 bits for the host ID (0..7) starting with host ID 144.173.6.176.
Therefore, the range of the block of IP addresses is 144.173.6.176 to 144.173.6.183.
Routers need to become aware directly or indirectly of your network ID and range of IP addresses so that packets
can be routed to your network. This is achieved by your ISP which will inform the ISPs it is connected to by
sending them 144.173.6.176/29 and they in turn will propagate your network ID and range of IP addresses to
others. Eventually all Internet routers will know a subset of your network ID and therefore will be able to forward
packets destined for your Web server, etc. Figure 9.3.1.17 shows a hierarchy of routers and their simplified routing
tables. Router 1 routes using the most significant byte of an IP address, Router 2 the next most significant and

398 Single licence - Abingdon School

 9.3.1 The Internet and how it works

so on. Router 5 is the gateway router for domain myowndomain.co.uk. It has the fully qualified domain name
gateway.myowndomain.co.uk.
Router 1

Router 2

144 2 Router 3

173 1 myowndomain.co.uk

Router gateway.myowndomain.co.uk
Router 4
port 144.173.6.176
no 6 3 10110 0 001 1 144.173.6.177
Part of network
ID to match 010 2 144.173.6.178
011 3 144.173.6.179
Routing path to network with prefix 144.173.6.176/29
100 4 144.173.6.180
and network ID in binary (29 bits)
10010000101011010000011010110 101 5 144.173.6.181
110 6 144.173.6.182

Figure 9.3.1.17 Routing path to domain myowndomain.co.uk 111 7 144.173.6.183

Router 5

Tasks
8 Visit the web page https://ipinfo.io/AS786 and discover how many IP addresses with the prefix
129.12.0.0/16 i.e. 16-bit NET ID 129.12 are allocated to the University of of Kent.

9 Visit the web page https://ipinfo.io/AS786 and locate the prefix 192.150.184.0/24 assigned to the University
of Manchester. Explain why for this prefix only 256 IP addresses can be allocated to this University. What is
the network ID for this block of IP addresses?

10 The web page https://ipinfo.io/AS786 shows that 512 IP addresses are allocated to the University of
Northumbria at Newcastle at prefix 192.173.2.0/23. How many bits are allocated for the host ID and how
many for the network ID? What is the network ID?

In this chapter you have covered:

■ The structure of the Internet
■ The role of packet switching and routers
■ The main components of a packet
■ The definition of a
• router
• gateway
■ How routing is achieved across the Internet
■ The term ‘uniform resource locator’ (URL) in the context
of internetworking
■ The terms ‘domain name’ and ‘IP address’
■ How domain names are organised
■ The purpose and function of the Domain Name System (DNS) and
its reliance on the DNS server system
■ The service provided by Internet registries and why they are needed.

Single licence - Abingdon School 399

9 Fundamentals of communication and networking

9.3 The Internet

Learning objectives:
■ 9.3.2 Internet security
Firewalls
■ Understand how a firewall A firewall is a combination of hardware and software that isolates an
works (packet filtering, proxy
organisation’s internal network from the Internet at large, allowing some
server, stateful inspection)
packets to pass and blocking others. Figure 9.3.2.1 shows a firewall located
■ Explain symmetric and between an organisation’s local area network and the router that connects,
asymmetric (private/public via an ISP, the organisation’s network to the
key) encryption and key Internet. With all network traffic entering Key term
exchange and leaving the organisation’s network passing Firewall:
through the firewall, the firewall is able
■ Explain how digital certificates A firewall is a combination of
hardware and software that
to allow authorised traffic through whilst
and digital signatures are isolates an organisation’s internal
blocking unauthorised traffic. Two ways by
obtained and used network from the Internet at
which a firewall can control traffic are large, allowing some packets to
■ Discuss worms, Trojans and • traditional packet filtering pass and blocking others.
viruses, and the vulnerabilites
that they exploit • stateful inspection of packets

■ Discuss how improved code Packet filtering

quality, monitoring and Packet filtering is done by a packet filter acting on a network-layer datagram
protection can be used to (see Figure 9.3.1.9 in Chapter 9.3.1). Network-layer datagrams contain source
address worms, trojans and and destination IP addresses, source and destination port numbers, protocol
viruses. type: TCP, UDP, ICMP (Internet Control
Message Protocol) and so on; Key term
ISP ICMP message type; some flags,
& DNS servers Packet filtering:
SYN, and ACK related to theTCP
Packet filtering is done by
connection three-way handshake
a packet filter acting on a
(Figure 9.3.2.2). The packet
network-layer datagram.
filter is set up to make decisions
Web server according to

Firewall • IP source or destination address

Interface 1
E Internet • Protocol type
Interface 2 gateway
• Source and destination port numbers
• ICMP message type (see Table 9.3.2.1)

A B C D • TCP flag bits: SYN, ACK, etc, (see Figure 9.3.2.2)

Private Local Area Network
• Different rules for datagrams entering and leaving the network
Figure 9.3.2.1 Local Area Network behind
• Different rules for different firewall interfaces.
a firewall

Single licence - Abingdon School 400

9 Fundamentals of communication and networking

Information For example, a datagram arriving at the firewall from the Internet with IP
destination address of the Web server shown in Figure 9.3.2.1 and destination
TCP three-way handshake to
port number 80 will be allowed through the firewall interface connected to
establish a TCP connection:
the Web server. However, if the destination IP address was that of server D
Client Server connected to the firewall’s second interface then the datagram could be blocked
Connect Listening
request
SYN for all ports except those related to an FTP connection.
Accept
ACK A datagram arriving at the firewall from the Internet with IP destination
Connected ACK
address of the Web server shown in Figure 9.3.2.1 and destination port
Connected
number 23 might be blocked by the firewall from passing through the firewall
interface connected to the Web server.
Figure 9.3.2.2 TCP 3-way
handshake Information
ICMP message datagrams have a type and a code field, and contain the header and the first
8 bytes of the IP datagram that caused the ICMP message to be generated.
Information Ping is a software utility used to test the reachability of a host on an Internet Protocol (IP)
network and to measure the round-trip time for messages sent from the originating host to
ICMP message types:
a destination computer and back. Suppose ping sends an ICMP type 8 code 0 message (see
Table 9.3.2.1 - echo request) to the Web server in Figure 9.3.2.1. The ping ICMP type 8
ICMP
Code Description code 0 message is a request to the Web server to reply to the sender with ICMP type 0
type code 0 - echo reply. The reply will contain the originator’s IP address, i.e. the Web server’s
0 0 echo reply IP address. That’s fine because the ping echo request was addressed in the first place to this
destination Web server. However an attacker could do more by guessing the IP address range of the

3 1 organisation hosting the Web server and proceeding to carry out a ping sweep across this
host
IP address range in the hope that other hosts will be discovered. To prevent this, the firewall
unreachable
can be configured to block all ICMP echo request packets.
destination
3 3 port
unreachable Task
echo
8 0 1 Use ping to test if ICMP echo requests are blocked by a firewall:
request
(a) www.ucl.ac.uk (b) www.bristol.ac.uk
TTL
11 0 (c) www.educational-computing.co.uk
expired
Table 9.3.2.1 Some ICMP
Stateless packet filtering
message types
Stateless packet filters do not match return packets with outgoing packet
flow, and therefore ignore whether a connection has been established. Instead
Key term they focus on source and destination IP address, source and destination port
numbers and protocol type. A set of rules are constructed called a ruleset and
Stateless packet filtering:
Information about packets is not then expressed in the syntax of a particular firewall scripting language.
remembered by the firewall. This Suppose we want to allow inbound mail (SMTP, port 25) but only to the
type of firewall can be tricked
Internet gateway shown in Figure 9.3.2.1.
very easily by hackers because
allow/deny decisions are taken on Figure 9.3.2.3 shows how the rule for this could be expressed. Figure
a packet by packet basis and these 9.3.2.3(a) shows the table starting out in the default state which is to block
are not related to the previous
everything (*). Figure 9.3.2.3(b) shows the rule that allows an inbound
allowed/denied packets.
connection to port 25 on the Internet gateway. Port 25 is the Simple Mail

401 Single licence - Abingdon School

 9.3.2 Internet security

Transfer Protocol service for both

sending and receiving electronic mail. Action Source Port Destination Port Comment

If the policy is to allow any host on the Block * * * * Default

local area network to also send email
(a) Incoming
from any one of its ports to the outside
world via the Internet then hosts must
be allowed to connect to port 25 on Action Source Port Destination Port Comment
Connection
an external host. The rule to do this is Allow Gateway to Gateway
* * 25
shown labelled '(a) Outgoing' in Figure SMTP port
9.3.2.4. (b) Incoming
Figure 9.3.2.3 Firewall ruleset tables
The second rule labelled (b)
'Incoming', Figure 9.3.2.4 is Action Source Port Destination Port Flags Comment
LAN host
necessary to allow ACK packets packets to
(a) Outgoing Allow 192.168.1/24 * * 25 * external host
to enter the local area network SMTP port

which is necessary part if a TCP External host

(b) Incoming Allow * 25 192.168.1/24 * ACK
connection is to be established response

with an external host - see Figure 192.168.1/24 means all hosts on network with NetworkID 192.168.1
9.3.2.2. Figure 9.3.2.4 Firewall ruleset for SMTP

Stateful inspection packet filtering Key term

In stateful inspection packet filtering, TCP (and UDP) connections are tracked
Stateful inspection packet
and connection information is used to make filtering decisions. A connection
filtering:
table of current outbound TCP connections is maintained. Suppose an attacker
If the firewall remembers
attempts to send a bogus packet into the local area network behind the firewall connection information for
by sending a datagram with the ACK flag set from TCP source port 80, source previously passed packets, then
address 137.248.8.9 to destination port 49923 and destination IP address the firewall is performing stateful
inspection packet filtering.
95.144.156.56. This masquerades as a genuine HTTP response from an external
Web server at 137.248.8.9 to a Web page request from internal host with IP
address 95.144.156.56. When this packet reaches the firewall, the firewall checks
its access control list in Table 9.3.2.2 which indicates the connection table, Table 9.3.2.3, must also be checked
before permitting this packet to enter the local area network. On checking, the firewall sees that this packet is not
part of an ongoing TCP connection, and rejects the packet.

Check
Source Destination Source Destination Flag
Action Protocol connection
address address port port bit
table
External to
Allow 95.144.156/24 TCP >49152 80 Any
95.144.156/24
External to
Allow 95.144.156/24 TCP 80 >49152 ACK X
95.144.156/24

Deny All All All All All All All

Table 9.3.2.2 Access control list for stateful filter - this is scanned for a match from row 1 downwards

Single licence - Abingdon School 402

9 Fundamentals of communication and networking

Information Source Destination Source Destination

Client-server TCP address address port port
connections: 95.144.156.56 130.88.98.244 49917 80
Server ports are assigned to
numbers < 1023 permanently, 95.144.156.44 144.173.6.226 49918 80
e.g. 20, 21 for FTP, 23 for Telnet
25 for server SMTP, 80 for HTTP. 95.144.156.21 129.11.26.33 49919 80
Client ports are numbers >
49152 allocated dynamically and
Table 9.3.2.3 Connection table for stateful filter
valid only for the duration of the
communication. Proxy server
The boundary values of the Figure 9.3.2.5 shows communication between two computers connected
port number ranges, e.g. 1023, through a third computer acting as a proxy. The proxy acts on behalf of both
are reserved and therefore not
Alice and Bob to allow dialogue to take place indirectly between the two. The
available.
proxy is able to mediate the communication because it is aware from whom
the request comes, to whom the request is directed, and the type of request. It
Key term
might be the case that Alice is not allowed to contact Bob because he is on a
Firewall proxy server: banned list called a blacklist. It might be the case that Alice is not authorised to
A firewall proxy server works at
communicate with Bob in a particular way.
the application level of the TCP
protocol stack and therefore Translating this analogy into proxies in computer networks, a proxy server
is able to filter by application is a server (an application-specific server running on a host) that acts as an
protocol type, e.g. HTTP, and
intermediary for requests from clients seeking resources from other servers.
provide authorisation by user
(stateless and stateful operate at The proxy server would form part of the firewall protection shown in Figure
the level of hosts via IP addresses) 9.3.2.1. A client, say host A in the private local area network, connects to the
and what content users may proxy server, requesting some service, such as a Web page, or other resource
access.
available from a different server. The proxy server evaluates the request and
decides whether to allow it. If it does the proxy
Ask Bob “what is server allows an indirect network connection
What is 4 x 3?
4 x 3?” between the client, host A, and the requested
network service, e.g. a Web page from a remote
Web server located somewhere in the Internet. The
Web server that serves up the Web page will only
be aware of the proxy server not the real source
of the request, host A. As far as the Web server is
Alice Proxy Bob
concerned the client that made the request was the
proxy server acting as a client.
Similarly, host A gets the Web page from the proxy
Bob says 12 The answer is 12
server. If the proxy server has cached this Web page
from a previous occasion then it could, if necessary,
Figure 9.3.2.5 Communication between two
serve up the cached copy. Host A will be unaware
computers connected through a third computer acting
of the origin of Web page it is served, proxy server
as a proxy. Bob engages in a dialogue with the proxy
or remote server.
server but doesn’t know that the dialogue is driven
by Alice with whom he is communicating without Proxy servers work at the application level of the
knowing it. TCP protocol stack and so are aware of the type

403 Single licence - Abingdon School

 9.3.2 Internet security

of request they receive from clients behind the firewall by the application protocol that is used, e.g. SSH, HTTP
request.
A content-filtering web proxy server provides control over the content that may be relayed in one or both directions
by filtering by URL (and DNS) or by keywords in the content.
The proxy server may also scan incoming content in real time for viruses and other malware and block such content
from entering the network.

Question
1 Stateless packet filtering, stateful inspection packet filtering and proxy servers are three kinds of firewall
techniques. Explain by examples how they differ from each other.

Encryption
Encryption is the process of obtaining ciphertext from plaintext. Plaintext is understandable (English) text. The
intention of encryption is to render plaintext incomprehensible to all but those granted the means to reverse the
process, i.e. decrypt the ciphertext.
The encryption process requires two inputs: the plaintext and the encryption key.
Key term
The decryption process also requires two inputs: the ciphertext and the decryption
key. Symmetric encryption:
Symmetric encryption uses the
Encryption is covered in Chapter 5.6.10. same secret key to perform both
the encryption and decryption
Symmetric encryption
processes.
Shared private key
In symmetric encryption the communicating parties use the same key for
encryption and decryption. Symmetric key encryption is also known as private key or secret key encryption
because the key used must be private and known only to the communicating parties. If not, then anyone
intercepting encrypted messages can use a knowledge of the key and the encryption algorithm to decrypt the
messages.
For example, if Alice wanted to use symmetric encryption to send an encrypted message to Bob then Alice would
use the private key k, agreed with Bob, to encrypt the plaintext form of the message. Bob would use the same
private key k to decrypt the received encrypted message. Bob could reply with his own encrypted message which he
has also encrypted with private key k (Figure 9.3.2.6).
Using symmetric encryption, Alice and Bob are able to communicate securely through a communication channel.
However, there is a potential problem with use of a single private key. Let’s suppose that Alice gets to choose the
key. She now has to communicate the key to Bob. They could agree to meet so that Alice could, by whispering into
Bob’s ear, communicate the key securely. However, in a networked world, Alice and Bob may never meet and may
never communicate except over a network, e.g. the Internet. However, to communicate the key securely to Bob via
a network connection, Alice needs
to secure the channel through the Message M Private key, k Private key, k
Message M
network.
Encrypted with key k Encrypted message M Decrypted with key k
How can she do this if she hasn’t
yet distributed the private key? Communication channel

She needs the secure channel Figure 9.3.2.6 Secure communication channel using symmetric encryption

Single licence - Abingdon School 404

9 Fundamentals of communication and networking

to distribute the key. This “chicken and egg” situation is known as the key distribution problem of symmetric
encryption.
Asymmetric encryption (public key/private key encryption)
Key term Secure communication in the world of online transactions
Asymmetric encryption:
The key distribution problem of symmetric encryption becomes an even more
Asymmetric encryption uses two
keys, a public key and a private key.
challenging one in the world of online transactions which the World Wide Web
The public key and the private key has made possible. Many transactions take place online through online stores
are mathematically related but where goods may be bought with a credit or debit card. All these transactions
different. If the public key is used need to take place through secure channels of communication to protect a
for encryption, the private key
purchaser’s payment details. Securing each channel with a separate symmetric
must be used for decryption.
encryption private key shared in advance between seller and customer would be a
Public key encryption: nightmare:
The public key is freely shared • The seller would have to store a large number of private keys, one for each
with any party. Documents/
customer buying goods.
messages destined for the owner
of a public/private key pair are • Each key would have to be distributed to the corresponding customer who
encrypted with the public key. would have to remember the key and for which seller.
This public key is linked A mechanism for secure communication is needed which
mathematically with its paired
private key. The private key is
• reduces the number of keys that need to be remembered
used to decrypt the document/ • allows two parties to exchange information secretly, but with no
message encrypted with the pre-arranged symmetric encryption private key
paired public key. The owner of
the private key must be the only
One solution is public key/private key encryption to secure a communication
one to know this key. channel. This secure channel can then be used to distribute a temporary
If the key pair is chosen well: symmetric encryption key to be used for securing credit/debit card details.
• The private key cannot be This temporary shared key is called a session key because it is only used for the
derived from the public key
duration of the transaction.
• The public key encrypted
messages can only be decrypted Public key encryption
by the private key with which it In public key encryption two mathematically related keys are used:
is paired mathematically.
• a public key p
k
Information • a private or secret key s
k
Key pair:
Amazon EC2 elastic cloud
Public key encryption has two clear advantages over private key encryption:
uses public key cryptography • The online seller has only to store a single private key s rather
k
to encrypt and decrypt login than sharing, storing and managing N different secret symmetric
information. The keys that Amazon
encryption keys (i.e. one for each buyer).
EC2 uses are 2048-bit SSH-2 RSA
keys. • The number and identities of potential buyers need not be known at
the time of key generation.
Public key encryption is an asymmetric encryption scheme, asymmetric because only one of the paired keys, the
private key s is secret. The public key p can be freely shared with any party. Thus an online seller makes their
k k
public key p available to any buyer. A buyer can then use the seller’s public key to establish a secure channel over
k
which they can send their credit card information securely to complete a purchase. The seller uses their private key
s , which only they know, to decrypt the buyer’s credit card information to process the transaction or, if symmetric
k
encryption is used for the latter, to set up a secure means of communicating a shared private session key.

405 Single licence - Abingdon School

 9.3.2 Internet security

If the key pair is chosen well: Information

• The private key cannot be derived from the public key Prime number:
• The public key encrypted messages can only be decrypted by the A prime number is a natural
number greater than 1 that
corresponding private key.
cannot be expressed as the
In practice, encrypting and decrypting with public key encryption is far slower product of two smaller natural
than with symmetric encryption. For this reason, public key encryption is numbers. Note that this means 1
often used to solve the problem of sharing a secret key for use in a symmetric is not a prime number.

encryption scheme. In other words: a number that has

only two factors: 1 and itself, e.g. 2,
One public key encryption scheme that is employed in online transactions 3, 5, 7, 43.
to provide one-way secure communication and authentication is the RSA Relatively prime:
cryptosystem which is named after its inventors Rivest, Shamir and Adleman. Two natural numbers are relatively
prime if they have no common
The following RSA section (included as enrichment material) is beyond A Level. divisor apart from 1.
The RSA public key/private key cryptosystem E.g. 4 and 7 are relatively prime
but 7 and 14 are not.
Let us to suppose that Alice wishes to receive secure messages from other people.
One-way function:
She proceeds as follows: Exponentiation to the power e
modulo N is one-way function:
• Alice selects two distinct prime numbers p and q and then forms her
relatively easy to compute but
public modulus N = pq. hard to invert. The private key d
N must be sufficiently large to ensure that no adversary could is known as a trapdoor because it
factor N = pq except by luck. This means that p and q each need makes inversion possible.
It is easy to multiply two numbers,
to be more than 150 digits long, and not be too close to each other.
p and q but apparently hard to
• Alice then chooses her public exponent e to be relatively prime to factor a number pq into a product
(p -1)(q – 1), with 1 < e < (p – 1)(q – 1). of two others when it is large. Try
factoring 7859112349338149.
• Alice chooses the pair (N, e) as her public key and she publishes this.
Exponentiation:
• Her private key is the unique integer, d such that Exponentiation means raising a
ed mod (p – 1)(q – 1) = 1 and 1 < d < (p – 1)(q – 1) number x to the power y; it is
or in congruence notation as ed = 1 (mod b) where b = (p - 1)(q - 1) written xy.
A useful property of
d is the mod (p - 1)(q - 1) multiplicative inverse of e
exponentiation is that raising a
Bob has a message which he wishes to send to Alice (Figure 9.3.2.7). number x to the power y and
He proceeds as follows: then raising the result to the
power z yields the same result
• Bob encodes the characters of his message as a string of integer codes M, as raising x to the power z and
e.g. 1521112325981720...... raising that result to the power y,
• Bob looks up Alice’s public key (N, e) i.e. (xy)z = xyz = (xz)y
This is true even if the
• Bob splits the integer encoded form of the message M into a sequence
exponentiation uses modular
of blocks M1, M2, M3, ...., Mi where each Mi is an integer that satisfies
arithmetic.
1 ≤ Mi < N.

Message M Public key, pk = (N, e) Private key, sk = d Message M

Encrypted with key pk = (N, e) Encrypted message M Decrypted with key sk

Me mod N (Me mod N)d mod N = Med mod N
Figure 9.3.2.7 RSA Public key/Private key encryption

Single licence - Abingdon School 406

9 Fundamentals of communication and networking

• Bob then encrypts these blocks as Ci = Mie mod N and sends the encrypted blocks to Bob.
Alice decrypts the encrypted message blocks as follows:
• Alice decrypts each Ci to recover Mi using her private key d by calculating Mi = Cid mod N
• Alice then converts the string of integer codes, Mi, back into their equivalent characters to recover the
plaintext form of the message Bob has sent her.
Example
Suppose Alice chooses primes p = 7 and q = 11. So N = 77, (p – 1)(q – 1) = 60 and she chooses e = 7, since 7 and 60
are relatively prime.
Alice then calculates her private key using ed = 1 (mod (p -1)(q - 1) to be d = 43, since 43 x 7 = 301 = 1 mod 60
Hence Alice’s public key is the pair (N, e) = (77, 7) and her private key is d = 43. (Use Windows calculator in
If Bob wants to send the plaintext message M = 4 to Alice he encrypts it as Scientific mode for the mod
ciphertext C = Me mod 77 = 47 mod 77 = 16384 mod 77= 60 (mod 77). operation)
Alice then decrypts C using her private key to recover the message
M = Cd mod 77 = 6043 mod 77 = 4 (mod 77)

Background
Public key/private key encryption schemes rely on the discrete logarithm problem to make it difficult to
discover the private key. A very simple example that illustrates this problem is as follows:
To encrypt integer 19 raise it to the power 43 and then find the remainder after dividing by 77, the result is the
encrypted value 61
1943 mod 77 = 61
To recover the unencrypted value 19, an exponent must be found for which
61d mod 77 = 19
If we try d = 1, 2, 3, 4, 5, 6, 7 we find that 7 does.
Thus, 617 mod 77= (1943)7 mod 77 = 1943x7 mod 77 = 19
by replacing 61 by (1943).
Now suppose, we encrypt 2137 by raising it to the power of 17 and finding the remainder after division by
3233, the result is the encrypted value of 2137
213717 mod 3233 = 166
To recover the unencrypted value 2137, an exponent d must be found for which
166d mod 3233 = 2137
If we try d = 1, 2, 3, ........., 2753 we find that 2753 does.
Now suppose we use exponent 65537 and a 2048-bit modulus such as the one shown in the information panel
on next page labelled Public modulus. We would find that d would be the 2048-bit value labelled Private
exponent in the information panel. Using the public exponent 65537 and the public modulus in the margin
we could encrypt integers up to 256 digits long. To decrypt we would use the private exponent shown in the
margin and the same public modulus. The public key is (public exponent, public modulus) and the private key
is (private exponent).
The discrete logarithm problem is the difficulty with which the private exponent can be discovered given only
the public exponent, the public modulus and the result from encryption. The last example illustrates that this
becomes an exceedingly difficult task when the number of bits used for the public key and the private key is a
large number.

407 Single licence - Abingdon School

 9.3.2 Internet security

Tasks Information
The discrete logarithm problem: Public exponent:
This is the problem of finding x where 65537
ax ≡ b (mod n) Public modulus:
Watch the Khan Academy video “The discrete logarithm problem” 2945766458199006354518714370
https://www.khanacademy.org/computing/computer-science/cryptography/modern- 789937313459038969836551741
crypt/v/discrete-logarithm-problem 474365312554100524284016689
991569699787388856353829206
RSA encryption: 878996051572086976167536159
Watch the Khan Academy video “RSA encryption part 1” 367501426841421545476777770
https://www.khanacademy.org/computing/computer-science/cryptography/modern- 911830025184723218753394504
crypt/v/intro-to-rsa-encryption 845305877537841256481722403
521947540091647129987550526
501191700526487333570141735
Questions 966271760998390300210716322
6848818840015822496173899627
2 Two computers, X and Y, communicate securely using public/private 2694978348391161716080846065
key encryption. X and Y each has a public key and a private key. 8137438912535222664160206809
X encrypts a message that it sends to Y using Y’s public key. 110088175424068694053002994
332201096356829801571379647
Explain why the message should not be encrypted with:
912703834109764508780629994
(a) X’s private key 906036475520264572478205942
(b) X’s public key 172563985777892068763898076
972147981555129254575283743
398768074422492663494096553
654793297152972338693514772
Task 6813268287153607983,
Explore the Wolfram programming lab: Private Exponent:
The Wolfram programming language is a useful language with which to explore 1867954933393873782460117671
encryption, factoring, modular arithmetic and a lot of other concepts. 258803650956417619184085589
https://lab.wolframcloud.com/app/ 395176368450238942707099281
362736371572765095935920688
763253092317817272007727966
Key exchange 049630350743485307336678984
Diffie-Hellman key exchange 444723319447437745471314964
Diffie-Hellman exponential key agreement or key exchange provides a way 559885586137870316567243231
236783738516085134016244637
for Alice and Bob to agree on a key k while communicating over an insecure
080374821711090843470222168
network channel. The key k is used only for one communication session. It is 596156702985658545495780230
then discarded. Hence the name session key. 0706215013490208463305771899
8128856073063404802108286698
A modulus N is made available publicly for network users to use to secure their
8921193324834578936195915247
communications. 363635418311637187885848492
Alice privately selects a large random number sA, and calculates 541980772633601888977561074
s
pA = 2 A (mod N). 349848200284511774738451327
Alice’s private key is sA. Alice sends pA to Bob. 825749779060874980402587688
552346646502447343908134657
An eavesdropper, Eve, who intercepts pA will find it extremely difficult to 726546395530654563666934925
s
discover Alice’s private key sA from a knowledge of 2 A (mod N). 365767118252566387299259415
Bob also privately chooses a large random number sB and calculates 793638059269399354303032695
s 3854631035757879505
pB = 2 B (mod N). Bob sends pB to Alice.

Single licence - Abingdon School 408

9 Fundamentals of communication and networking

Alice has her private key sA and Bob’s public key pB. Task
Bob has his private key sA and Alice’s public key pA.
Diffie-Hellman key exchange:
Bob and Alice can now calculate their shared secret key, k. Watch the Khan Academy video
sA sB sA “Public key encryption: What is
Alice uses her private key sA to calculate kA = pB = (2 (mod N)) mod N
it?”
sB sA sB
Bob uses his private key sB to calculate kB = pA = (2 (mod N)) mod N https://www.khanacademy.org/
s s s s computing/computer-science/
But (2 B (mod N)) A = (2 A (mod N)) B, therefore kA = kB = k, the shared secret key.
cryptography/modern-crypt/v/
The shared secret key k may now be used for symmetric encryption of messages diffie-hellman-key-exchange-
between Alice and Bob. part-1

SSH
SSH, or secure shell, is a secure protocol and the most common way of safely
administering remote servers. Symmetric keys or shared secret keys are used by SSH in order to encrypt the entire
connection between client and server.
The secret key is created through a process known as key exchange. This exchange results in the server and client
both arriving at the same key independently by a process similar to that described in the previous section.
The symmetric encryption key created by this procedure is session-based and is the key used to encrypt the data sent
between server and client, e.g. credit card details.
SSH utilizes asymmetric encryption during the initial key exchange process used to set up the symmetrical
encryption key used to encrypt the session between two parties. The two parties exchange public keys (see pA and pB
in previous section) in order to produce the shared secret key k used for symmetrical encryption.

Question
3 Describe two different methods for communicating a shared secret symmetric encryption key k via an
insecure network channel.

Digital certificate
When a website is visited that supports the Secure HyperText
Cannot guarantee authenticity of the domain
Transfer Protocol HTTPS, to which encrypted connection is established
e.g. https://www.google.co.uk, the identity of the website is
“proved” with public key encryption. It is important that the Application: Google Chrome
URL: www.dingleydell.com
authenticity of the website is checked, i.e. that it really is the Reason: Invalid name on certificate. Either
genuine site and not a man-in-the-middle attacker placed the name is not the allowed list, or
was explicitly excluded.
between a visitor and a website, impersonating both.
With a man-in-the-middle attack, the browser thinks it is Disconnect
talking to the web site on an encrypted channel, and the
website thinks it is talking to the browser, but they are both Continue
talking to the attacker who is sitting in the middle. All traffic
passes through this man-in-the-middle, who is able to read and View certificate
modify any of the data.
Figure 9.3.2.8 Digital certificate warning that
Operating systems and browsers typically have a list of
web site authenticity cannot be guaranteed
certificate authorities that they implicitly trust. If a website

409 Single licence - Abingdon School

 9.3.2 Internet security

presents a certificate that is signed by an untrusted certificate authority, the Key term
browser warns the visitor that something could be amiss - Figure 9.3.2.8. Digital certificate:
Certificates are files containing information about the owner of a website, Digital certificates are files
and the public half of an asymmetric key pair (e.g. RSA). A certificate containing information about the
owner of a website, and the public
authority (CA) digitally signs the certificate to verify that the information in
half of an asymmetric key pair
the certificate is correct. By trusting the certificate, you are trusting that the
(e.g. RSA). A certificate authority
certificate authority has done its due diligence. (CA) digitally signs the certificate
A website that supports HTTPS should have a certificate and a corresponding to verify that the information in
public key. This will enable a connection to be made between a web browser the certificate is correct. By trust-
ing the certificate, you are trusting
and the website using the Transport Layer Security (TLS) protocol. The
that the certificate authority has
browser must also support the TLS protocol. done its due diligence.
When a web browser uses HTTPS to visit a website such as https://www. A website that supports HTTPS
google.co.uk, a TLS connection is established between the web browser and should have a certificate and a
the website. TLS is used to encrypt data sent between both and to prove the corresponding public key. This
will enable a connection to be
identity of the server.
made between a web browser
Question and the website using the
Transport Layer Security (TLS)
4 (a) What is a digital certificate?
protocol.
(b) Explain how it is used to authenticate a website that supports
HTTPS.

Authentication Information
The web browser starts the TLS connection by telling the website which Digital certificate authority:
CAcert.org is a community-driven
ciphersuites it supports i.e. it tells the website which types of encryption it is
Certificate Authority that issues
able to use. The website https://cc.dcsec.uni-hannover.de/ will report what certificates to the public at large
ciphersuites your browser supports, e.g. ECDHE-RSA-AES256-SHA which for free - see www.cacert.org.
means Elliptic Curve Diffie-Hellman Exchange which is used to share a
symmetric key, RSA for authentication, AES256 is the symmetric encryption CAcert’s goal is to promote
awareness and education on
algorithm to use and SHA is the hash function used to create a message digest.
computer security through the
The web browser generates a 128-bit random number called a nonce which it use of encryption, specifically
sends to the website. The website encrypts this nonce using its RSA private key by providing cryptographic
certificates. These certificates
and sends the encrypted nonce to the web browser. The web browser decrypts
can be used to digitally sign
this encrypted nonce using the trusted-certificate-authority-validated public key and encrypt email, authenticate
belonging to the website. A match confirms that the website knew the private and authorize users connecting
key half of the public key/private key pair and therefore is authentic. The same to websites and secure data
nonce is never used twice to prevent a bogus site replaying the encrypted nonce transmission over the Internet.
Any application that supports the
which it had intercepted and recorded on a previous occasion.
Secure Socket Layer Protocol
Digital signature (SSL or TLS) can make use of
A digital signature is a cryptographic technique that can be used in a digital certificates signed by CAcert.
world when you want
• to indicate that you are the owner or creator of a document, or
• to affirm that you agree with a document’s content.

Single licence - Abingdon School 410

9 Fundamentals of communication and networking

For Bob to know that a document genuinely came from Alice, he needs some way to authenticate the document.
This is what Alice does to make this possible:
• Alice digitally signs the Signed document: Enc(M)
document with her private key Document: M Digital signature

s using a public key/private

A Dear Bob, grf54tRy73*c£$jQap
key encryption algorithm as This is the first chance Ms2389@#na./:8109
that I have had to write Encryption 8rt#?<60djeix)eis.......
shown in Figure 9.3.2.9 to to you since.... algorithm Enc
produce Enc(M), the digital
Yours
signature Alice

• Alice then sends Enc(M), the Alice Bob

Alice’s private
encrypted form of M to Bob public key, pA key, sA
• Bob decrypts Enc(M) using private key, sA
Alice’s public key p as shown Figure 9.3.2.9 Digital signing of Alice’s document with Alice’s
A
in Figure 9.3.2.10 and private key
recovers the document M Signed document: Enc(M)
Digital signature Document: M
from Alice.
Does the digital signature Enc(M) grf54tRy73*c£$jQap Dear Bob,
Ms2389@#na./:8109 This is the first chance
meet the requirements of being 8rt#?<60djeix)eis....... Decryption that I have had to write
algorithm to you since....
verifiable and unforgeable?
Yes, because applying Alice’s public Yours
Alice
key to the digital signature, Enc(M)
recovers the original document, M. Bob Alice’s public
Alice’s public key and private key key, pA
are associated mathematically. If Figure 9.3.2.10 Bob verifies that the document came from Alice by
document M had been signed with a decrypting the encrypted document using Alice’s public key
private key different from Alice’s then
Key term
applying Alice’s public key to Enc(M) would not have recovered M.
Digital signature:
The only person who could have signed M is therefore Alice because she is the
A digital signature is a cryptographic
person who generated the pair of keys (pA, sA). This assumes that Alice has not technique that can be used in a digital
given away her private key or had it stolen. world when you want
• to indicate that you are the owner
Also, if Alice or anyone else should alter the original document from M to
or creator of a document, or
M1, the signature that Alice created for M will not be valid for M1 because
• to affirm that you agree with a
Dec(Enc(M)) does not equal M1. Thus digital signatures also provide message/ document’s content or
document integrity, allowing the receiver • to verify that the message/
document was not tampered with
• to verify that the message/document was not tampered with in
in transit or
transit • to verify that the original
• to verify that the original document has not been tampered with at document has not been tampered
with at source.
source.

411 Single licence - Abingdon School

 9.3.2 Internet security

Message digest
Key term
Encryption and decryption are computationally intensive. Message digest:
Digitally signing a document by encrypting the whole document is therefore A message digest is a a fixed-
expensive computationally. length “fingerprint” of a message/

To reduce the computation overhead a hash function H is used. document M of arbitrary length.
It is calculated by applying a hash
This function takes a message/document M, of arbitrary length and computes
function H to message/document
H(M), a fixed-length “fingerprint” of the message/document M called a message M to compute H(M), the
digest. message digest.
Long document: M
Alice now digitally signs the
Alice
message digest rather than the
A VERY LONG REPORT
message/document itself to create a I have so much to tell
you that I am afraid the
digital signature. information in this Fixed-length hash
As H(M) is generally of much report extends to a H(M)
hundred pages. grF54Try73*c£$jqap
Hash function
shorter length than the original ................. Ud4a£z\98m>Dpw2
.................
message/document M, creating the
digital signature consumes much
H(M)
less computational effort.
Alice now sends the document M Send M
digitally signed hash H(M)
Send Digital signature
followed by the digital signature to Digital
signature
Bob as shown in Figure 9.3.2.11. Bob
Hv#(&56rEalpz1+\;9 Encryption
3VxC@=Lqyxa56$^v algorithm
On receipt of M, Bob uses the receives Alice’s private
hashing function H to produce a
M key, sA
Digital signature
hash of document M as shown in Figure 9.3.2.11 Using Alice’s private key to create a digital
Figure 9.3.2.12. signature from the hash of document M
On receipt of the digital Digital
signature
signature, Bob decrypts Bob (Signed hash)
M receives Hv#(&56rEalpz1+\;9 Decryption
it using Alice’s public key Digital from 3VxC@=Lqyxa56$^v algorithm
pA to obtain the message signature Alice Alice’s public
key, pA
digest or hash produced Long document: M
by Alice if she genuinely grF54Try73*c£$jqap Fixed-length hash
Ud4a£z\98m>Dpw2
signed the document. If A VERY LONG REPORT
I have so much to tell
the two hashes match then you that I am afraid the
information in this
Bob can conclude that report extends to a
hundred pages.
Alice was the signer of the .................
.................
document and that the
document has not been
tampered with whilst in Fixed-length hash
Match Therefore
transit. Hash function grF54Try73*c£$jqap Compare document has
Ud4a£z\98m>Dpw2 been genuinely
signed by Alice
Figure 9.3.2.12 Verifying a signed
document on receipt of the document Don’t match Therefore document has been
and its digital signature created from the signed by someone other than
Alice or document tampered with
document’s message digest or hash
Single licence - Abingdon School 412
9 Fundamentals of communication and networking

Question

5 (a) What is a digital signature?

(b) Explain how it is used to authenticate the author of an electronic document.
(c) Explain how a digital signature can be used
(i) to discover that a digitally signed electronic document has been tampered with
(ii) to challenge the author when they say that they did not sign to say that they agree with its contents.

6 Explain how signing a long electronic document or message with a digital signature might differ from
signing a short document or message.

7 Two computers, X and Y, communicate securely using public/private key encryption. X and Y each have
a public key, a private key, and a hash function H that generates a message digest of a plaintext message
of arbitrary length. Computer X sends a plaintext message and a digital signature in encrypted form to
Computer Y as shown in Figure 9.3.2.13. Computer Y processes the received transmission as shown in
Figure 9.3.2.13.
State what processing takes place at each of the stages indicated by a ringed numeral. Where a key is used,
specify which key.

Computer X

Message ② Digital Digital Encrypted

① Signature Signature [Digital
Digest ③
+ Signature
Plaintext Plaintext +
Plaintext
Message Message Message]

Transmission
path

Message Digital Decrypted Encrypted

Digest ⑤ Signature [Digital [Digital
Signature Signature
⑦ + +
Regenerated ④
Plaintext Plaintext Plaintext
Message
⑥ Message Message] Message]
Digest
Computer Y
Figure 9.3.2.13

Worms,Trojans and viruses

The devices that we connect to the Internet enable us to do useful things such as exchange email messages, obtain
search engine results, view Web pages and so on. However, our devices may be subject to infection by malicious
software (malware) which can do all sorts of harmful things such as delete files, take control of our devices, install
spyware that can steal private information such as passwords, debit and credit card details. Examples of types of
malware are worms, Trojans and viruses. Worms and viruses are self-replicating and are thus able to spread from one
computer to another.

413 Single licence - Abingdon School

 9.3.2 Internet security

Worms
Worms are malicious software that can enter a computer from the Internet Key term
without any explicit user interaction by exploiting a vulnerability in a running Worm:
network application program, e.g. the Conficker computer worm that targeted A computer worm is a self-
flaws in Windows OS software to propagate. The worm in the newly infected contained program that attacks
a system and tries to exploit a
computer scans the Internet, searching for other hosts running the same
specific vulnerability in the target.
vulnerable network application. On discovering other vulnerable hosts, it sends a
It replicates itself in order to
copy of itself to those hosts. spread to other computers.
Unlike a computer virus, a worm does not need to attach itself to an existing Unlike a computer virus, it does
program or file because they are self-contained programs. Worms are one way of not need to attach itself to an
existing program or file but
installing a backdoor in the infected computer to allow the creation of a “zombie”
instead it exploits vulnerabilities
computer under control of the worm author. Networks of such machines are often in network application software
referred to as botnets and are very common use is the sending of junk email called running on the infected host.
spam. The aim of a worm is often to
take over a computer for the
Trojans purposes of creating a botnet for
A Trojan horse or Trojan is a type of malicious software that is often disguised as sending spam and Distributed
Denial of Service (DDoS) attacks.
legitimate software. Users are typically tricked into loading and executing Trojans
on their systems. Once activated, Trojans can enable cybercriminals to spy on
you, steal your sensitive data, and gain backdoor access to your system. Unlike
computer viruses and worms, Trojans are not able to self-replicate. Key term
Trojan:
Trojans are classified according to the type of actions that they can perform on the
A Trojan horse or Trojan is a type
infected computer. A selection is shown below. of malicious software that is often
• A Backdoor Trojan enables remote control over the infected computer disguised as legitimate software.
Unlike computer viruses and
by a cybercriminal or hacker to do anything they wish on the infected
worms, Trojans are not able to
computer – including sending, receiving, launching, and deleting files,
self-replicate.
displaying data, and rebooting the computer. Backdoor Trojans are Users are typically tricked into
often used to link a group of victim computers to form a botnet or loading and executing Trojans on
zombie network that can be used for criminal purposes their systems. Once activated,
Trojans can enable cyber-criminals
• Trojan-Banker programs are designed to steal account data for online
to spy on you, steal your sensitive
banking systems, e-payment systems, and credit or debit cards from data, and gain backdoor access to
infected computers your system.
• Trojan DDoS are programs that conduct DDoS (Distributed Denial
of Service) attacks against a targeted web address. They send multiple
requests from your computer and several other infected computers that attack and overwhelm the target
address leading to a denial of service at the target address computer, typically a server
• Trojan-Downloaders can download and install new versions of malicious programs onto your computer
– including Trojans and adware
• Trojan-Ransom can modify data on your computer – so that your computer doesn’t run correctly or you
can no longer use specific data. The criminal will only restore your computer’s performance or unblock
your data, after you have paid them the ransom money that they demand
• Trojan-Mailfinder is a program that can harvest email addresses from your computer.

Single licence - Abingdon School 414

9 Fundamentals of communication and networking

Viruses Key term

A virus is malicious software attached to another file which infects and harms Virus:
a user’s computer when the user is tricked into opening the file, e.g. an email A computer virus is malicious
attachment containing the virus’ executable code. Opening such an attachment software that requires some form

inadvertently runs the malware on their computer. Once executed, the virus is of user interaction to infect a
user’s computer because the user
able to replicate and then spread by sending an identical email with the same
needs to be tricked into opening
malicious attachment to, for example, every recipient in the user’s address book. the file to which the virus is
Viruses piggy-back on seemingly legitimate files but they require some form of attached. Viruses are not stand-
alone programs but are always
user interaction to infect the user’s computer.
embedded in another program or
file. Once the virus is executed
Question it can replicate and infect other
8 Explain the differences between worm, virus and Trojan malware. computers.

Code quality
One way of securing computers against malicious attacks is to improve the quality of the code, from operating
systems to application
programs, which executes on
computers. Attackers exploit
vulnerabilities in code in order
to get a computer to execute
their malicious software.
For example, the command
shell in the Windows
operating system is a separate
software program, CMD.
exe, that executes programs
and displays their output as
individual characters on the
screen. It is known to have
vulnerabilities.
For instance, one of the
commands it can execute is
the echo command which
is shown in Figure 9.3.2.14
being used once to output
the string “Hello World” and
once to output the name of
the current directory stored
in the environment variable
%CD%. On the second
occasion, a new directory has Figure 9.3.2.14 Exploiting a vulnerability in cmd.exe, the command line
been created beforehand using interpreter in the Windows operating system to run the ping program
415 Single licence - Abingdon School
 9.3.2 Internet security

the command md "NewTest&ping 8.8.8.8" and made the current directory. The command shell program CMD.
exe is tricked into executing ping 8.8.8.8 because & is interpreted by the command shell as the separator of multiple
commands on one command line. CMD.exe runs the first command echo c:\test\NewTest1, and then the second
command, ping 8.8.8.8 because NewTest is separated by & from ping 8.8.8.8 in %CD%.
Figure 9.3.2.15 shows the same exploit causing calc.exe to execute. A third example could replace calc.exe with
malware.exe.

Figure 9.3.2.15 Exploiting a vunerability in cmd.exe, the command line interpreter

in the Windows operating system to run Windows calculator calc.exe

CMD.exe can also execute batch files. There are batch files in systems that have been running for years which
contain echo %CD% commands probably with a pipe, e.g. echo %CD% > logfile.txt.
If Microsoft’s recommended fix to change echo %CD% to echo "%CD%" has not been done then running any batch
file that has not been fixed could allow malware to execute.
The contents of environment variable %CD% can be changed with the SET command outside of the batch file
containing echo %CD%,
e.g. SET CD=c:\test\NewTest&malware. If this is the value of %CD% when the batch file is executed the malware
program will be executed.
Buffer overflow
A common vulnerability exploited by worms and viruses is buffer overflow. The Stuxnet worm used buffer overflow
and some other techniques. It was widely suspected of targeting Iran’s nuclear enrichment programme and may
have destroyed 1,000 centrifuges, reduced output and sowed chaos.
Another way that malicious code can be executed is to place this code in the buffer that is being caused to overflow
and to overwrite the return address so it points back into the buffer.
Patching or updating software is usually an effective way to remove vulnerabilities to worms and viruses. Coding to
standards that avoid creating code vulnerabilities in the first place is a better way.

Single licence - Abingdon School 416

9 Fundamentals of communication and networking

Background

Figure 9.3.2.16 shows a simplified case of how buffer overflow can occur when a call is made to a library routine input(Password) that reads
keyboard input from a user one character at a time. The routine copies each character’s byte-value, in turn, into the memory reserved in the stack
frame for the call to IsPasswordValid() labelled Password (the stack frames for the calls to input and stringCompare are not shown). Seven
bytes have been allocated in this current stack frame to Password according to the declaration "chararray[7] Password;" in the program
source code. These seven bytes constitute a buffer. Figure 9.3.2.16 shows the result of the user typing password "1234567" and then on another
attempt to login, password "123456789ABCDEF".

Notice that the library routine input stores the first character of the password in the byte pointed to by Top of Stack Pointer, the next in the
byte above and so on. Notice also that the longer password overwrites the Previous Stack Frame Pointer and the Return Address areas of
the stack frame. This is because the library routine input (Password) does not apply array bounds checking when the 8th byte and subsequent
bytes are written to Password. Buffer overflow occurs and the return address is corrupted. This return address is used to return to the routine
main but this will not happen because the return address now points to somewhere else in the memory of the computer. It is likely that this will
cause an exception (Figure 9.3.2.17). Now suppose the 15 character password is chosen so that the return address on the stack is overwritten
with the address of the else output ("Access allowed"); line of code. The result will be that the user will be granted access to the system.

In a different scenario, buffer overflow could replace the return address on the stack with the address of the malicious code of a worm or a virus.

Top of main memory

Return Address (4 bytes) Operating System

Buffer Overflow
program

Previous Stack Frame Previous Stack Frame Pointer (4 bytes) int main (void)
{
boolean PasswordValid;
output("Enter password: ");
PasswordValid (4 bytes)
PasswordValid <─ IsPasswordValid();
Stack Frame Pointer if (PasswordValid = False)
F then
Return Address (4 bytes) E {output ("Access denied");
Main
D exit(-1);
C
B
}
A else output ("Access allowed");
Previous Stack Frame Pointer (4 bytes)
9 }
Current Stack Frame 8 True or False
7 7 returned boolean IsPasswordValid (void)
6 6
5
{
5
4 4 chararray Password[7];
3 3 input(Password);
Password
2 2 return stringComparison(Password, "LetMeIn");
(7 bytes)
Top of Stack Pointer 1 1 }

Stack grows downwards

Call to stringComparison(Password, "LetMeIn")
towards bottom of main
returns True if password matches "LetMeIn"
memory
otherwise returns False

Each stack frame stores the return address, Buffer Overflow

local variables, previous stack frame pointer, Program Counter program
arguments passed to subroutine that has
been called.
The return address is the address of the Operating system calls
instruction to be executed on return from main which calls
Operating System IsPasswordValid
the subroutine. It is the address of the
instruction immediately following the
instruction that called the subroutine. Bottom of main memory
Figure 9.3.2.16 Buffer overflow

Password validator

Access violation at address 43444546 in module Figure 9.3.2.17 Exception

‘Password.exe’. Read of address 00000000.
caused by Buffer overflow

OK

417 Single licence - Abingdon School

 9.3.2 Internet security

Monitoring and protection

Information
Firewalls are designed to monitor packets and offer protection against attempts
to exploit weaknesses in the TCP/IP protocol suite which worms and Trojans Social Engineering:
This is when attackers set out to
can exploit.
gain the trust of a user so that
Digital certificates enable HTTP and FTP to be reliably secured against attacks. they can steal user information
Digital certificates also offer protection against downloading and installing or dupe them into downloading
malicious software.
malicious software by validating that the software is from a trusted source.
Attackers exploit the lack of
Downloading from untrusted sources has the potential to install a Trojan. knowledge of many users about
Anti-virus and anti-malware software are designed to monitor and protect how technology functions in
order to launch their attacks, e.g.
against attempts to exploit weaknesses in operating systems by hooking deep
a Trojan attack.
into the operating system’s core or kernel and function. Non-technical users can avoid
Any time the operating system accesses a file, the protection software scans the social engineering attacks by
following some simple guidelines
file to check it is a ‘legitimate’ file or not. If the file is identified as malware by
which are intended to offer some
the virus/malware scanner, the access operation will be stopped, the file will be protection against such attacks,
dealt with by the scanner in a predefined way and the issue reported to the user. for example:
The goal of the anti-virus and anti-malware software is to stop any malicious • Never open email
operations on the system before they can occur. Anti-virus and anti-malware attachments unless they
are from a trusted source.
software should be kept up-to-date as new threats emerge on a regular basis.
• Pay heed to warnings that
Some viruses infect files such as MS-Word documents and MS-Excel state that the website or
spreadsheets because the applications allow strings of program commands software you are about to
download does not have a
called “macros” to be stored in document and spreadsheet files. Clicking on the
valid certificate.
document or spreadsheet file launches the corresponding application which will • Disable macros in
run the malicious commands/code/script unless users have disabled execution applications that allow
of macros in these applications. them.
General guidelines:
• Always make sure updates
are installed for both
Information operating system and
application code.
CERT: • Keep anti-virus and anti-
CERT was founded in November 1988 in response to the Morris worm incident, which malware software up-to-date.
brought 10 percent of Internet systems to a halt in November 1988. Since 1988, CERT has • Use a firewall.
handled more than 300, 000 computer security incidents. • Run a security check on
CERT conducts secure coding research and eliminate vulnerabilities in software caused by your system using Microsoft
quality of coding. Baseline Security Analyser.
Analysis indicates that the majority of incidents are caused by Trojans, social engineering,
and the exploitation of software vulnerabilities, including software defects, design decisions,
configuration decisions, and unexpected interactions among systems, e.g. print spooler
exploit in Windows - https://www.youtube.com/watch?v=Fy0S9KMNjnY which was also
implicated in the Stuxnet worm.

Single licence - Abingdon School 418

9 Fundamentals of communication and networking

Questions
9 Worms, viruses and trojans exploit vulnerabilities in computer systems. Describe one vulnerability that is
exploited by a
(a) worm (b) virus (c) Trojan.

10 How might the quality of operating system and application code affect whether a system is vulnerable to
worms, viruses and Trojans?

11 How can monitoring and protection be used to prevent worms, viruses and Trojans from infecting systems?

In this chapter you have covered:

■■ How a firewall works by packet filtering, proxy server, and stateful
inspection
■■ Symmetric and asymmetric (private/public key) encryption and key
exchange
■■ How digital certificates and digital signatures are obtained and used
■■ Worms, Trojans and viruses, and the vulnerabilities that they exploit
■■ How improved code quality, monitoring and protection can be used to
address worms, trojans and viruses.

419 Single licence - Abingdon School

9 Fundamentals of communication and networking

9.4 The Transmission Control Protocol/

Internet Protocol (TCP/IP) protocol
Learning objectives:
■ 9.4.1 TCP/IP
The role of the four layers of TCP/IP protocol stack
■ Describe the role of the four Networking protocols were designed to make possible communication
layers of the TCP/IP stack
between application programs executing on different hosts whilst hiding the
(application, transport,
complexities of the underlying network from these application programs.
network, link)
A host, or host computer, is any computer system that connects to an internet
■ Describe the role of sockets in and runs applications.
the TCP/IP stack The term process is used for an instance of a program in execution, so it
■ Be familiar with the role of is actually processes in different hosts connected by a network that are
MAC (Media Access Control) communicating.
addresses Layered organisation
■ Explain what the well-known Networking protocols are usually developed in layers.
ports and client ports are used Each layer is responsible for a different part of the communication process.
for and the differences between The software that implements a protocol is called protocol software and the
them software that implements a suite of protocols such as TCP/IP, a protocol stack.
The TCP/IP protocol suite consists of four conceptual layers: application,
Key terms
Process:
transport, network or IP (Internet layer), and link layers (Figure 9.4.1.1).
An instance of a program in It is implemented in software as the TCP/IP protocol stack in separate
execution. software modules corresponding to the individual layers of the protocol suite.
Protocol: Each layer and therefore each software module has a different responsibility.
A protocol provides agreed
The protocol stack is installed on each computer either as a part of the
signals, codes and rules for data
operating system or as a software library.
exchange between systems.

Networking protocols: Application Telnet, FTP, email(SMTP, POP3), Web browsing (HTTP)
Networking protocols make
possible communication Transport TCP
between processes executing on
Network IP (Internet layer)
different hosts whilst hiding the
complexities of the underlying Link Ethernet, PPP, WiFi, DOCSIS (cable TV)
network from these processes.

TCP/IP protocol suite: Figure 9.4.1.1 The four layers of the TCP/IP protocol suite and stack
The TCP/IP protocol suite
consists of four conceptual
Application programs interact with the software stack via an Application
layers: application, transport, Programming Interface (API). The de facto standard is the socket API.
network or IP, and link - see RFC
The following Python code snippet shows a client application using the socket
1122 https://tools.ietf.org/html/
rfc1122#page-27.
API to set up a socket to send a message to a server:
clientSocket = socket(socket.AF_INET, socket.SOCK_STREAM)
TCP/IP protocol stack:
message = "Hello Server"
Implements the TCP/IP protocol
suite in software. clientSocket.sendTo(message, (serverName, serverPort))

Single licence - Abingdon School 420

9 Fundamentals of communication and networking

Key term Figure 9.4.1.2 shows a client process and a server process that use the socket
API from the TCP/IP protocol stack to send and receive messages via a TCP/IP
Application layer:
connection pipe established between client and server.
Application layer protocols are
used to exchange data between
programs running on the source Host B
Host A
and destination hosts. It is the
application layer that provides the
interface between these programs
and the underlying network over
which the programs’ messages are Socket API used by both client
Client process and server processes to send Server process
transmitted, e.g. HTTP message
e.g. Web browser
and receive messages e.g. Web server
GET / which fetches the default Application
Web page from a Web server. Messages Messages

Modules
Key term within Client Server
operating socket byte pipe socket
Transmission Control
system or
Protocol (TCP): library of connection pipe
TCP/IP protocol TCP/IP protocol
TCP enables applications routines stack stack
executing on two hosts to
establish a connection and End-system End-system
exchange application-layer
Figure 9.4.1.2 Sending and receiving messages using the socket API
messages through a reliable
byte-stream channel (pipe) for
data flows between the two end Application layer
systems.
A process in one end-system (host) uses the application layer of TCP/IP to
exchange packets of information with a process in another end-system. The
Key term packets of information at the application layer are called messages.
Function of Transport layer: The application layer uses different application-layer protocols for different
The basic function of the
applications. For example, if the application is designed to enable Web pages
transport layer is
• to accept messages/data from
to be fetched from a Web server then the application will use either the HTTP
the layer above it application-layer protocol or the HTTPS application-layer protocol. An
• split these into smaller units application-layer protocol defines the kind of messages to send. In the case of
called segments if necessary HTTP or HTTPS, one such message could be a GET message.
• pass these segments to the
network or IP layer
Transport layer
• ensure that all the segments The transport layer of the protocol stack is a piece of software in each
arrive correctly at the other host. This software implements the transport protocol known by the name
end Transmission Control Protocol (TCP). TCP enables applications executing
• reassemble the received
on two hosts to establish a two-way connection and exchange application-layer
segments, which it gets from
the network layer, in the messages through a reliable byte-stream channel (pipe) for data flows in either
correct order to form the direction between the two end-systems as shown in Figure 9.4.1.2. It also
message/data to pass to the allows the connection to be terminated.
layer above.
TCP breaks long messages into shorter segments which it sends as separate
The transport layer is a true
transport-layer packets known as TCP segments.
end-to-end layer which carries
messages/data all the way from The application at the sending side (e.g. a Web browser using the application-
the source to the destination. layer protocol HTTP) pushes messages (e.g. GET) through a TCP/IP socket.

421 Single licence - Abingdon School

 9.4.1 TCP/IP

The transport-layer protocol TCP has the responsibility of getting the messages Key principle
to the socket of the receiving application process, e.g. a Web server listening on
End-to-end principle:
port 80. Port numbers such as port 80 are 16-bit numbers used for application The end-to-end principle of
and service identification on the Internet. the Internet requires that the
two endpoints, the hosts, are
TCP does everything in its control to guarantee delivery of the application-
responsible for establishing,
layer message and also guarantee that the received TCP segments will be supervising and maintaining
reassembled in the correct order to form the message to be passed to the a connection between two
application-layer and then the corresponding application process. communicating processes,
one on each host. This is done
Once the TCP has established a connection: by a piece of software in each
• it monitors the connection for transmission errors and responds host known by the name
Transmission Control Protocol
when an error is detected by retransmitting the segment that suffered
(TCP).
the error
• it detects when a connection is broken Key term
Network or IP layer:
• it performs flow control by speed matching sender and receiver
The network or IP layer of the
• it provides congestion control when the network is congested. TCP/IP protocol stack in hosts
and routers is responsible for
0 15 16 31 moving IP-layer packets from one
16-bit source port number 16-bit destination port number host to another without regard to
whether these hosts are on the
32-bit sequence number
same network or not.
32-bit acknowledgement number 20 bytes It adds source and destination
4-bit header IP addresses to packets on their
length way from the transport layer to
16-bit TCP checksum the link layer, and removes source
and destination IP addresses from
Figure 9.4.1.3 TCP header packets on their way from the link
layer to the transport layer.

The format of a TCP segment is covered in Chapter 9.3.1 (Figure 9.3.1.10).

Information
The header part of a TCP segment contains the source and destination port
Internetwork or internet:
numbers as shown in Figure 9.4.1.3. Port numbers bind to sockets so that the When two or more networks are
TCP layer can identify the application that data is destined for. interconnected an internetwork
(inter-network) or internet is
Network or IP layer
formed.
The TCP layer uses the IP layer to carry its segments. Each TCP segment is
Internet:
encapsulated in an IP packet before it is sent across the internet. The network The Internet is one such
or IP layer in hosts and routers move these packets known as IP or Internet internetwork or internet. The
Protocol packets from one host to another without regard to whether these Internet is a collection of many
interconnected networks that use
hosts belong to the same network or different networks. The Internet Protocol
TCP/IP. To distinguish the Internet
is a connectionless protocol which just provides a best effort but not guaranteed from other internets the initial
way of delivering packets called datagrams. The reliability of the transmission letter is capitalised.
is left to the layer above, the transport layer. Packet/segment:
Getting to the destination host may require many hops via intermediate routers The term packet is sometimes
used generically to refer to a TCP
along the way.
segment.

Single licence - Abingdon School 422

9 Fundamentals of communication and networking

Both routers and hosts are assigned IP addresses by which they may be identified. An IP
address (Internet Protocol address) is a logical 32-bit (IPv4) or 128-bit quantity (IPv6). The
IP layer in the sending host attaches the source (sending host) and destination (receiving
host) IP addresses to the packets handed to it from the transport layer. The IP layer then
passes these packets to the layer below, the link layer. The IP layer also receives packets from
the link layer, removes the source and destination IP addresses then passes them to the
transport layer.
The format of an IP packet is universal so that all routers recognise it (see Chapter 9.3.1).
This makes it possible for IP packets to pass through almost every network of networks, e.g.
the Internet.
Both hosts and routers need to use the network or IP layer of the TCP/IP protocol stack
but since the job of a router is dedicated to routing packets, a router only requires use
of the network and link layers of the TCP/IP stack. The IP layer in a router must have
sufficient knowledge of other routers and links in its internet to be able to make routing
decisions for packets that pass through it.
Together TCP and IP hide the differences between the underlying networks through
which packets pass when going from source to destination host.
Link layer
The link layer handles all the physical details of interfacing with the network cable or
wireless connection. It includes the network interface card (network adapter) and a device
driver. TCP/IP protocol supports many different types of link layer, depending on the type
of networking hardware being used. One example is Ethernet.
The link layer adds source and destination hardware addresses (e.g. MAC addresses) to
packets that it receives from the IP layer then dispatches the packets onto the local cable or
wireless connection.
If the packet is destined for a host on another network, the link layer destination address
is the hardware address of the gateway (router) to the internet which the other network is
connected to.
In an Ethernet local area network (LAN) these hardware addresses are Ethernet
Key term card addresses, or MAC addresses - see the following section on MAC
Link layer:
addresses. Figure 9.4.1.4 shows a packet despatched by the link layer of a
The link layer handles all the
physical details of interfacing with host with IP address 174.89.0.54 to a remote host with IP address 210.5.0.67.
the network cable or wireless Figure 9.4.1.4 shows the first, second and last hop of many hops.
connection.
Note that the link layer hardware address changes from hop to hop whilst the
The link layer adds source and
destination hardware addresses source and destination IP addresses remain constant. This is because the link
(e.g. MAC addresses) to packets layer’s role is to stream bytes between directly connected machines, hosts and
that it receives from the IP layer routers. It is the link layer that puts bits onto the network cable or wireless
then despatches the packets
connection. Sending to a remote machine is done in hops where each hop is
onto the local cable or wireless
a direct connection (link) between a host and a router, a router and a host, a
connection.
router and another router, or two directly connected hosts.

423 Single licence - Abingdon School

 9.4.1 TCP/IP

Web Web Key term

browser server
TCP/IP socket:
A socket is one endpoint of a
Application layer Application layer
HTTP HTTP two-way communication link
between two programs running
on the network, e.g. a Web
TCP layer TCP layer
browser and a Web server.

IP address IP address
174.89.0.54 IP layer IP layer
210.5.0.67

Link-layer address Link-layer address

00-03-47-C9-69-52 Link layer Link layer
00-04-34-65-07-81

Physical cable Gateway Gateway Physical cable

Link-layer address Link-layer address Key fact
00-03-47-B6-21-46 00-04-34-98-15-21
Router
Identifying a TCP connection:
Link-layer address Link-layer address
00-02-22-E3-54-12 network 00-04-34-A8-19-10 Every TCP connection can be uniquely

First hop
identified by its two endpoints.
For example,
Source Link-layer Destination Source Destination
<174.89.0.54:49717, 210.5.0.67:8080>
address Link-layer address IP address IP address
00-03-47-C9-69-52 00-03-47-B6-21-46 174.89.0.54 210.5.0.67

Second hop

Source Link-layer Destination Source Destination Key fact

address Link-layer address IP address IP address
Binding port and socket:
00-02-22-E3-54-12 00-02-77-A1-88-53 174.89.0.54 210.5.0.67
A socket is bound to a port
number so that the TCP layer can
Last hop
identify the process (instance of
Source Link-layer Destination Source Destination
an executing application) that data
address Link-layer address IP address IP address
00-04-34-98-15-21 00-04-34-65-07-81 174.89.0.54 210.5.0.67 is destined for.
Port numbers are also used:
• as application protocol
identifiers
Link layer address
• for firewall-filtering purposes
00-02-22-E3-54-12
Gateway

Link layer address Link layer address Link layer address

00-03-47-B6-21-46 00-02-77-A1-88-53 00-04-34-A8-19-10
Gateway
Link layer address
Link layer address 00-04-34-65-07-81
IP address
174.89.0.54 00-02-77-A1-45-11
Link layer address
IP address
00-03-47-C9-69-52 Link layer address
210.5.0.67
00-04-34-98-15-21
Local Area Network

Router Network

Figure 9.4.1.4 TCP/IP protocol stack and the role of the link layer in the communicating hosts and
intermediate routers

Single licence - Abingdon School 424

9 Fundamentals of communication and networking

Questions
1 Describe three tasks performed by the transport layer of the TCP/IP protocol stack.

2 Name one application-layer protocol.

3 Explain why the source and destination IP addresses of a packet remain the same whilst link layer addresses
need to change when a packet is sent from a host to a server on a different network.

4 Describe the role of the different layers of the TCP/IP stack in each of the host, the server, and intervening
routers when a Web browser running on a host uses the TCP/IP protocol stack to send an HTTP GET
message to a Web server running on a different network.

The role of sockets in TCP/IP

A socket is one endpoint of a two-way communication link between two programs running on the network,
e.g. a Web browser and a Web server. A socket is bound to a port number so that the TCP layer can identify the
executing application that data is destined for.
An endpoint is a combination of an IP address and a port number.
Every TCP connection can be uniquely identified by its two endpoints. This fact allows multiple connections to
exist between a host and a server as shown in Figure 9.4.1.5.
A socket is created as a two-way resource (capable of both sending and receiving), even if it is only used in one
direction by program code.

Application Service
Endpoint Endpoint

Web browser Port no 49717 Communication link Port no 8080 Web server

Email client Port no 51234 Port no 25 Email server

Communication link

Host IP address Endpoint Endpoint Server IP address

174.89.0.54 210.5.0.67
Figure 9.4.1.5 Shows endpoint pairs for a Web server and a Web browser,
an email client and an email server

A procedure called bind, performed on the client side and the server side, tells the operating system which local
IP address/port no pair to associate with the socket before a connection is established (on the client side the bind
procedure may be called as part of the connect operation). On the server side, the bind procedure establishes the IP
address/port no pair that the server listens on to accept client connections.
Figure 9.4.1.6 shows an HTTP server listening on port 8080 for connection requests from client hosts. The client
shown is called localhost. It has IP address 127.0.0.1 (known as the loopback IP address). A TCP socket has been
created and the bind operation applied to assign the socket to the IP address/port no pair 127.0.0.1:49717 (client
port numbers are in the range 49152-65535) . The colon symbol (:) is used to separate the IP address 127.0.0.1
from the port number 49717. The client host sends a TCP connect request through this socket to the listening

425 Single licence - Abingdon School

 9.4.1 TCP/IP

socket of the HTTP server. This server has been set up on the same machine so Information
also has localhost as its computer name and the same IP address 127.0.0.1. The
Normally, the two IP addresses in
name loopback IP address derives from the fact that TCP/IP stack packets sent
an end-to-end connection will be
to the network adapter of the machine are looped back to another TCP/IP stack
different, but in this example the
within the same machine. The server has created a listening socket that is bound loopback address is being used
to the IP address/port no pair 127.0.0.1:8080. so that students can try the code
On receipt of a connection request from the client on the listening socket, the out using just their machine (with
their machine acting as both client
server decides whether it will accept this request or not. If it accepts, it creates
and server).
a new TCP socket (connection socket in Figure 9.4.1.6) which it then binds to
127.0.0.1:8080. Figure 9.4.1.7 shows a three-way handshake which takes place
Key term
to establish this TCP connection between server and client.
TCP/IP socket:
The established TCP connection between HTTP client and HTTP server
A socket is one endpoint of a
(connection socket to connection socket) is uniquely identified by its two endpoints: two-way communication link
<127.0.0.1:49717, 127.0.0.1:8080>. Having established this TCP connection, between two programs running
the HTTP server now returns to listening on its listening socket for connection on the network, e.g. a Web
browser and a Web server.
requests.
Figure 9.4.1.6 shows Python 3.4 code for both the HTTP client and the HTTP
server. The HTTP server runs (httpd.serve_forever()) until it is shutdown.
200 OK
Client <html><head><title>Title goes here.</title></head>
<body><p>This is a test.</p>
on </body></html>

localhost
200 OK Connection
IP: 127.0.0.1 ......
socket
Connection Port Port GET
socket 49717 byte pipe 8080
Server
Listening on localhost
GET 200 OK TCP 3-way socket IP: 127.0.0.1
......
handshake

import http.client
http_server = "127.0.0.1:8080"
#create a connection to server listening on port 8080
connection = http.client.HTTPConnection('localhost:8080') import http.server
#request HTTP GET to server class MyHandler(http.server.BaseHTTPRequestHandler):
connection.request("GET", http_server) def do_GET(s):
#get response from server #Respond to a GET request
response = connection.getresponse() s.send_response(200)
#print response from server and data s.send_header("Content-type", "text/html")
print(response.status, response.reason) s.end_headers()
data_received = response.read(4096) s.wfile.write(bytes("<html><head><title>A test</title></head>",
print(data_received.decode("utf-8")) "utf-8"))
connection.close() s.wfile.write(bytes("<body><p>This is a test.</p>", "utf-8"))
s.wfile.write(bytes("</body></html>", "utf-8"))

Figure 9.4.1.6 Shows the use of sockets to connect return

if __name__ == '__main__':
Handler = http.server.HTTPServer
an HTTP client to an HTTP server so that a web httpd = Handler(("localhost", 8080), MyHandler)
httpd.serve_forever()
page may be downloaded for viewing as raw text
Client Server
The HTTP client requests a web page from the HTTP server by using the HTTP Connect
SYN Listening
request
GET command. In Python 3.4 code this is connection.request("GET", Accept
ACK
http_server) where http_server is an identifier for "127.0.0.1:8080". This
Connected ACK
request is directed at port 8080 on the server. On receiving this packet, the server
Connected
identifies the TCP connection as <127.0.0.1:49717, 127.0.0.1:8080> and routes
the packet to the corresponding connection socket (Figure 9.4.1.6). The HTTP Figure 9.4.1.7 TCP 3-way
server responds by sending 200 OK followed by a page of HTML through this handshake
Single licence - Abingdon School 426
9 Fundamentals of communication and networking

Key fact TCP connection to the HTTP client. The client-side Python 3.4 code print
Identifying a TCP connection: statements output the response to the client's console.
Every TCP connection can be uniquely Figure 9.4.1.8 shows a web browser using URL localhost:8080 rendering the
identified by its two endpoints. HTML received from localhost:8080 in the browser's window as "This is a
For example,
test". To do this, the web browser has connected to the HTTP server, sent a
<174.89.0.54:49717, 210.5.0.67:8080>
GET request using the established TCP connection, received back 200 OK and
Key fact the HTML with content "This is a test".
Binding port and socket:
A socket is bound to a port
number so that the transport
layer can identify the application
that data is destined for.

Key term
Well-known ports: Figure 9.4.1.8 Shows the same HTML rendered in a browser window
Well-known ports use the port
number range 0-1023 and are
Questions
associated with service names
assigned by the Internet Assigned 5 What is a TCP/IP socket?
Numbers Authority (IANA) such
6 Why is a socket bound to a port number?
as http or www.
7 How is each TCP connection uniquely identified?
Information
8 Describe the role of sockets in the TCP/IP stack when a web browser
System or Well-known ports:
See -
on a host with IP address 195.61 3.4.7 connects with a web server
https://tools.ietf.org/html/rfc6335#page-11 listening on port 80 on a machine with IP address 210.56.78.3 to
download a web page.
Key term
User ports: Well-known ports and client ports
User ports use the port number Port numbers are 16-bit numbers which are also known by their associated
range 1024 - 49151. They are assigned
service names such as "telnet" for port number 23 and "http" (as well as
on request by the Internet Assigned
Numbers Authority (IANA), e.g.
"www") for port number 80. Hosts running services, hosts accessing services
ciscocsdb 43441 is the service on other hosts, and intermediate devices such as firewalls and NATs all need
name and assigned TCP port number to agree on which service corresponds to a particular destination port. Many
of Cisco NetMgmt DB Ports. services have a default port which servers usually listen on. These ports are
For an application form to register a
recorded by the Internet Assigned Numbers Authority (IANA) through the
port number and service, see:
https://www.iana.org/form/ports-services service name and port number registry.
Port numbers are subdivided into three ranges of numbers:
Key term
• the System Ports or Well-known Ports use the range 0-1023
Client ports:
Client ports use the port number • the User Ports or Registered Ports use the range 1024-49151
range 49152-65535 and are assigned
• the Dynamic Ports or Private or Ephemeral Ports use the range
temporarily by the TCP layer so that
it can identify the client application 49152-65535
that data is destined for. The first two ranges are assigned by IANA. The third range is used by clients. A
Port numbers in this range are not
port number from the dynamic ports range (or private port or ephemeral port
controlled or assigned by IANA.

427 Single licence - Abingdon School

 9.4.1 TCP/IP

range) is allocated temporarily to a connection socket requested by a client application. Hence, these are also called
client ports. Client port numbers are never allocated permanently.
Table 9.4.1.1 shows some examples of well-known port numbers and their corresponding service names.

Service name Port number Description

FTP 21 File Transfer Protocol (control)
FTP-data 20 File Transfer Protocol (data)
SSH 22 Secure Shell Protocol
SMTP 25 Simple Mail Transfer Protocol
HTTP 80 World Wide Web HTTP
POP3 110 Post Office Protocol Version 3
HTTPS 443 HTTP protocol over TLS/SSL
Table 9.4.1.1 Some examples of well-known port numbers and their corresponding service name

Questions
9 Explain what the well-known ports and client ports are used for and the differences between them.

10 The following information was captured by packet capture software monitoring the network adapter of a
host when a Web browser sent an HTTP message to a Web server:
50268 192.168.2.22 64.29.1.45.9 80 HTTP GET /books.html
For this captured transmission state
(a) the source IP address (b) the port no associated with the sending application's connection socket
(c) the destination IP address (d) the port no associated with the connection socket in the destination
(e) the application-layer message.

11 In the same capture session, the following information was captured immediately after books.html was
transferred to the client:
50272 192.168.2.22 64.29.1.45.9 80 HTTP GET /img/AQAUnit2.jpg
50268 192.168.2.22 64.29.1.45.9 80 HTTP GET /img/AQAUnit2.jpg
(a) What is the meaning of each line of this capture?
(b) The first number in the second line is the same as the first number in Q10. What explanation can you
give for the two numbers being the same?

The role of MAC addresses Key term

A MAC (Media Access Control) address is a 48-bit address expressed in MAC address:
hexadecimal and separated into 6 bytes, e.g. 00-02-22-C9-54-13 - see Chapter A MAC (Media Access Control)
9.2.1. It is the physical or hardware address of the network adapter and is address is a 48-bit address
expressed in hexadecimal and
designed to be unique. Hosts and routers communicate with the network through
separated into 6 bytes, e.g. 00-02-
a network adapter attached to the host or router. Network adapters perform all 22-C9-54-13.
the functions required to communicate on a network. They convert data from the The MAC address identifies the
form stored in the host/router to the form transmitted or received on the cable or network adapter connected to
the network so that a packet's
wireless link and vice versa. The MAC address of a network packet on the network
destination hardware address can
cable (or received via a wireless link) is read and compared by a network adapter
be matched to a particular host
with this adapter as its address.

Single licence - Abingdon School 428

9 Fundamentals of communication and networking

with its own unique assigned MAC address. If it matches then the network
adapter passes the packet to the link layer software of the TCP/IP stack.

Questions
12 What is the role of Media Access Control (MAC) addresses?

13 The following information was captured by packet capture software monitoring the network adapter of a
host when a Web browser sent an HTTP message to a Web server:
50268 192.168.2.22 64.29.1.45.9 80 74:d4:35:94:ad:53 70:73:cb:b2:f7:d0 HTTP GET /books.html

What do the numbers 74:d4:35:94:ad:53 and 70:73:cb:b2:f7:d0 represent?

In this chapter you have covered:

■■ The role of the four layers of the TCP/IP stack (application, transport,
network, link)
■■ The role of sockets in the TCP/IP stack
■■ The role of MAC (Media Access Control) addresses
■■ What the well-known ports and client ports are used for and their
differences

429 Single licence - Abingdon School

9 Fundamentals of communication and networking

9.4 The Transmission Control Protocol/

Internet Protocol (TCP/IP) protocol
Learning objectives:
■ 9.4.2 Standard application layer protocols
FTP (File Transfer Protocol)
■ Be familiar with the following File Transfer Protocol (FTP) is an application layer protocol that enables files
protocols:
on one host, computer B, to be copied to another host, computer A. One host
• FTP (File Transfer runs an FTP client and the other an FTP server.
Protocol)
FTP servers use two ports: port 21 for commands and port 20 for data.
• HTTP (Hypertext
Transfer Protocol) Figure 9.4.2.1 shows an FTP client connected to an FTP server via TCP so
• HTTPS (Hypertext that it can send a command request for a file Test.txt located on the FTP server.
Transfer Protocol Secure) The FTP response is to send file Test.txt through the TCP connection to the
• POP3 (Post Office FTP client.
Protocol (v3))
Port 57359 is bound to the TCP socket on the client side, whilst on the server
• SMTP (Simple Mail
Transfer Protocol) side, port 21 is bound to the command socket and port 20 to the data socket.
192.168.2.22 64.29.145.9
• SSH (Secure Shell) Computer A Computer B

FTP Client FTP Server

■ Be familiar with FTP client
software and an FTP server FTP protocol Application FTP protocol Application

■ Be familiar with how SSH is

Port 21 Port 20
used for remote management Port 57359
commands data
TCP
■ Know how an SSH client TCP TCP

is used to make a TCP IP layer IP layer

connection to a remote port
Source Destination
for the purpose of sending Link layer
Command Address Address
Link layer

commands to this port using

application level protocols Request Computer Computer
for A B
■ Be familiar with using Test.txt Port 57359 Port 21

SSH to log in securely to a

remote computer and execute
commands Computer Computer
A B Test.txt
■ Explain the role of an email Port 57359 Port 20

server in retrieving and Destination Source Data

sending email Address Address

■ Explain the role of a web Figure 9.4.2.1 FTP transfer of file Test.txt from Computer B to Computer A
server in serving up web pages
in text form The client may need to navigate the directory structure of the server, create new
■ Understand the role of a directories, rename files and directories, delete files and directories. These are
web browser in retrieving sent to the server as command requests.
web pages and web page
resources and rendering these
accordingly

Single licence - Abingdon School 430

9 Fundamentals of communication and networking

FTP client software and FTP server

Figure 9.4.2.2 shows FTP client software (FileZilla) running on a computer with IP address 192.168.2.22
connected to an FTP server running on a computer with IP address 64.29.145.9. This server is located in the USA
whilst the client computer is in the UK.

Figure 9.4.2.2 FTP client using FileZilla FTP client software connected to an FTP server

Anonymous and non-anonymous access

Some FTP servers restrict access to their service and require
users to use a registered user name and password (non-
anonymous access). Other FTP servers do not restrict access
but may prompt users for a user name; then the user name
is normally ‘anonymous’. The user may be prompted for a
password too. If the user name is ‘anonymous’, it is sufficient
to supply an e-mail address as the password. Figure 9.4.2.3
shows the user name Anonymous being set up on a Cerberus
FTP server. Figure 9.4.2.4 shows an FTP client connected
to this FTP server (Cerberus FTP server obtainable from
www.cerberusftp.com).

Figure 9.4.2.3 Setting up Cerberus FTP server

431 Single licence - Abingdon School

 9.4.2 Standard application layer protocols

Figure 9.4.2.4 FTP client using FileZilla FTP client software connected to a Cerberus FTP server

Questions
1 A file Test.txt stored on a computer on one network is to be copied to another machine on a different
network which is reachable from the first computer.
(a) What type of software running on the first computer could enable this to be done?
(b) What type of software running on the second computer could enable this to be done?
(c) Why might it be necessary to send commands as well as file data across the connection between the
two computers? Give an example of one command.

HTTP (Hypertext Transfer Protocol)

Hypertext Transfer Protocol (HTTP) is a very simple application-level protocol. In this protocol, a client
computer sends a request message to the server and the server responds with a response message (Figure 9.4.2.5).
In the example in Figure 9.4.2.5 the file index.html has been requested. The response message may contain
many forms of data. The most popular form of data is text formatted using Hypertext Markup Language (HTML).

WEB BROWSER WEB SERVER MAGNETIC DISK/

BACKING STORE ON
WEB SERVER
Request
Message

Response
Index.html
Message

Figure 9.4.2.5 HTTP request-response messages

Single licence - Abingdon School 432
9 Fundamentals of communication and networking

Other data, such as images or audio files, may also be transmitted. TCP establishes a connection between the client
computer and the server computer so that HTTP has a pathway for its request and response messages.
The simplest request message is
GET / <Return key pressed>
<Return key pressed>
This gets the default web page, index.html, for the given site. HTTP finishes with the connection after the response
message is sent; the TCP connection is broken unless specifically requested to stay connected. A web page returned
by an HTTP GET request is a text file containing content to be displayed together with instructions on how to
style and structure this content when displayed.
Here is what a web browser does:
1. It accepts a URL from a user, e.g. www.educational-computing.co.uk/books.html.
2. It extracts the FQDN (Fully Qualified Domain Name - host name + domain name - e.g. www.educational-
computing.co.uk) and uses a DNS server to translate it into an IP address. DNS is another application layer
protocol.
3. It sends a GET request for the web resource specified in the URL; the request is sent to a web server at this
IP address – port 80 unless another port number is specified.
4. It receives the file returned by the web server.
5. It renders this file’s contents in a web browser window; that means it uses the style and structure
instructions to display the content appropriately.
6. If this file contains other URLs, e.g. a reference to a graphic, then the browser should issue a GET to obtain
this resource from the web server, e.g. GET /images/flower.jpg, and, when received, display it according to
the instructions on style and structure.
To obtain a web page other than the default web page, the web browser sends an HTTP GET request message with
the structure
GET <path to resource> <Return key pressed>
<Return key pressed>
For example,
GET /books/books.html <Return key pressed>
<Return key pressed>

Questions
2 Explain how a web browser and a web server interact via the HTTP protocol when the web browser is used
with URL www.educational-computing.co.uk/books.html.

433 Single licence - Abingdon School

 9.4.2 Standard application layer protocols

HTTPS (Hypertext Transfer Protocol Secure)

Hypertext Transfer Protocol over Secure Sockets (HTTPS) is a web protocol that encrypts and decrypts user
page requests as well as the pages that are returned by the web server. HTTPS uses the Secure Sockets Layer (SSL)
beneath the HTTP application layer. HTTPS uses port 443 instead of port 80 in its interactions with TCP/IP.
Figure 9.4.2.6 shows the SSL sublayer which encrypts the HTTP GET / request before sending it through the
TCP connection to the Web server www.site.co.uk. Both the request and the response are encrypted.
HTTPS has been used for a long time for securing payment transactions on the Web but it is now being more
widely used for general Web access.
CLIENT SERVER
www.site.co.uk/
Web browser https://www.site.co.uk/ Web server <html>.......</html>

<html>.......</html> Application GET / Application

SSL socket SSL socket

SSL sublayer SSL sublayer

TCP socket Secure channel TCP socket

TCP TCP

IP IP

Link Link

Figure 9.4.2.6 Fetching a Web page using HTTPS

SSL provides a simple Application Programmer Interface (API) with sockets similar to TCP's API. When an
application wishes to use SSL, the application includes SSL classes/libraries.
SSL provides encrypted communication but it also embodies support for data integrity, server authentication, and
client authentication. SSL secures TCP. As such it can be used by any application that runs over TCP, e.g. SMTP
and POP3.

Questions
3 Give two reasons why Web browsers have been redesigned to use HTTPS rather than HTTP when
interacting with Web servers.

4 In what way does the use of the TCP/IP stack differ when HTTPS is used instead of HTTP?

SMTP (Simple Mail Transfer Protocol) and POP3 (Post Office Protocol (v3))
Simple Mail Transfer Protocol (SMTP) is used by e-mail clients to send e-mail. It is a relatively simple text-based
protocol. One or more recipients of a message are specified then SMTP is used by the email client to transfer the
message text to a mail server listening on port 25. The mail server takes care of delivering the mail to the ultimate
destination using SMTP. The user who retrieves the message from the destination mail server uses the application-
layer protocol POP3 to retrieve the stored mail. POP3 uses well known port 110. POP3 is Post Office Protocol
version 3. E-mail is stored in a mailbox and a user does not need to be connected for mail to be sent to them. The
server holds incoming mail until the user connects and requests the mail.
There are other email protocols that may be used instead of SMTP and POP3.

Single licence - Abingdon School 434

9 Fundamentals of communication and networking

The POP3 protocol defines commands that can be used to retrieve a mail message, e.g. the command RETR
no, where no is the position number of the message in the mailbox. The command LIST returns a list of these
numbers. DELE no marks an email for deletion.
For creating and sending email the SMTP protocol supports commands such as
• MAIL FROM: - defines the e-mail address of the sender of the message.
• RCPT TO: - defines the e-mail address of a recipient of the message. Repeating this command once for
each recipient means you can send one piece of mail to many users without having to repeat the entire
process over and over again.
• DATA marks the start of the data portion of the message, essentially everything that you would
consider "content", this includes the "To:", "From:", "CC:" etc. as these are not commands but simple
informational components making up a header which the e-mail client picks out of the content and
displays in a far nicer format. Just as a reminder - anything which is in the content can be faked as it is
content and so consequently cannot be validated.

Questions
5 A TCP connection to port 110 of a POP3 mail server pop3.apm-internet.net was established by an
email client to a user's mailbox stored on the mail server. Give two examples of POP3 commands that
might be sent by the email client over this TCP connection when the user is browsing their emails.

SSH (Secure Shell)

SSH is used for encrypted communication between two computers over a TCP/IP network. It uses port 22.
SSH provides a secure channel over an unsecured network in a client-server architecture, connecting an SSH client
application with an SSH server. Common applications include remote command-line login and remote command
execution, but any network service can be secured with SSH. SSH was developed independently of SSL and doesn't
therefore use SSL but it is similar in that it supports encryption of the communication, both client and server
authentication, and data integrity.
Using SSH for remote management
Managing computers and networks remotely can be done and was done previously using clear-text protocols and
unsecured channels. For this an administrator (or another user) uses Telnet software and the TCP/IP protocol to
establish an unsecured TCP connection through which operating system commands could be sent to the remote
computer. To log in to the remote computer, a Telnet client sends the account name and password in clear-text
form through a TCP connection to a Telnet server running on the remote machine. Any eavesdropper could easily
obtain the transmitted login details by tapping into the TCP connection. Once obtained the eavesdropper could use
these details to log in to the remote computer.
SSH is a secure replacement for Telnet. Both SSH client and SSH server software encrypt messages before they are
sent through a TCP connection.
Figure 9.4.2.7 shows a screenshot of a Windows computer connected to a remote computer, an Apple MacBook
Pro. The account accessed on the Mac belongs to user drbond. The Windows computer is shown logged into the
Apple computer having passed authentication by password. The Windows computer has navigated, using the
command cd, the directory structure on the Mac to a directory named myIOSProjects. A directory listing is
then obtained using the Unix command ls.

435 Single licence - Abingdon School

 9.4.2 Standard application layer protocols

The Windows computer's IP address is 192.168.2.22 and the Apple Macbook Pro computer's IP address is
192.168.2.21.

Figure 9.4.2.7 Using an SSH client on a Windows computer to connect to an SSH server running on a remote
computer so that the remote computer, an Apple Macbook Pro, can be managed
Background
Plink (PuTTY Link) is a command-line connection tool written for Microsoft's Windows operating system and similar to UNIX's SSH
application. Openssh is a fork(version) of SSH which has been released under an open source licence. It is now supported in Microsoft's
Windows PowerShell.

Questions
6 Communication of confidential information such as log in details over an insecure network is a risk. Why
might an SSH client application be used to connect to an SSH server on a remote computer that one
wishes to log in to?

In this chapter you have covered:

■ The following application layer protocols:
• FTP (File Transfer Protocol)
• HTTP (Hypertext Transfer Protocol)
• HTTPS (Hypertext Transfer Protocol Secure)
• POP3 (Post Office Protocol (v3))
• SMTP (Simple Mail Transfer Protocol)
• SSH (Secure Shell)
■ The use of FTP client software and an FTP server
■ Using SSH for remote management of a computer
■ The use of an SSH client to make a TCP connection to a remote port for the purpose of sending
commands to this port using application level protocols
■ Using SSH to log in securely to a remote computer and execute commands
■ The role of an email server in retrieving and sending email
■ The role of a web server in serving up web pages in text form
■ The role of a web browser in retrieving web pages and web page resources and rendering these accordingly.

Single licence - Abingdon School 436

9 Fundamentals of communication and networking

9.4 The Transmission Control Protocol/

Internet Protocol (TCP/IP) protocol
Learning objectives:
■ 9.4.3 IP address structure
A single uniform system
■ Know that an IP address is The physical architecture of Local Area
split into a network identifier An internet
Network A

an internet in which router(s) interconnecting

part and a host identifier part. three LANs
interconnect physical networks Host

(LANs) as shown in Figure 9.4.3.1

was covered in Chapter 9.3.1. TCP/
Key concept Host Router Host

IP systems use a software protocol

Host or host computer:
TCP/IP defines the term host or stack in each host and router to hide
host computer as any computer the physical architecture from hosts
system that connects to an and routers. This software turns Local Area Local Area
Network B Network C
internet and runs applications. an internet into a single uniform Figure 9.4.3.1 An internet consisting of
system or virtual network which three local area networks interconnected
enables users, application programs by a single router
and higher layers of the protocol
Key concept
software stack to communicate seamlessly using a uniform addressing scheme
Uniform addressing scheme:
called the IP address scheme.
A uniform addressing scheme
is a logical addressing scheme,
independent of the underlying Uniform addressing scheme
physical network. Each address
If an internet is viewed as a single, uniform system then all its host computers
conforms to a common format
defined by a standard, e.g. IPv4. must use a uniform addressing scheme in which each address is unique.
Physical network addresses are not good candidates for this uniform addressing
scheme because physical networks can differ in their technologies and because
Key concept each technology defines its own address format.
Interface:
The boundary between a host/ To guarantee uniform addressing for all hosts, the TCP/IP protocol software
router and the physical link that defines an addressing scheme that is independent of the underlying physical
connects the host to an internet network. This scheme is administered by the Internet Protocol (IP) part of
is called an interface.
TCP/IP.
Hosts are connected into a network via a single physical link, e.g. an Ethernet
Key fact cable. When the IP layer of the TCP/IP stack in a host sends an IP datagram
Routers have multiple it does so over this link. The boundary between the host and the physical link
interfaces:
is called an interface. Routers are connected to two or more links because a
Routers have multiple interfaces
because routers are connected to
router's job is to forward a datagram it receives on one link to another one of
two or more links. its connected links. The boundary between a link and the router is also called
an interface. A router thus has multiple interfaces, one for each of its links. The
IP protocol requires that each host and router interface has its own IP address.
The IP version 4 (IPv4) standard specifies that each host interface is assigned a
Single licence - Abingdon School 437
9 Fundamentals of communication and networking

Key fact unique 32-bit number known as the host interface's Internet Protocol address
or IP address or Internet address. The IP version 6 (IPv6) standard specifies
IPv4:
The IPv4 standard specifies that that each host interface is assigned a unique 128-bit number. Users, application
each host/router interface is programs and higher layers of the protocol software stack use these logical
assigned a unique 32-bit number addresses, i.e. IP addresses, to communicate through the Internet with each
known as the host/router
other without regard to whether the sending and receiving hosts are on the
interface's Internet Protocol
address or IP address or Internet
same physical network or a different one and without needing to know physical
address. addresses.
Figure 9.4.3.2 shows IPv4 addresses assigned to host and router interfaces.
Local Area
Key fact Network A

IPv6: 217.1.1.129
The IPv6 standard specifies
that each host/router interface
217.1.1.68
is assigned a unique 128-bit Host
number known as the host/router
217.1.1.34
interface's IP address.
217.1.1.25
217.1.1.1
Host 217.1.2.1 Router 217.1.3.1
Host
217.1.2.26 217.1.3.24

217.1.2.67 217.1.2.38 217.1.3.32 217.1.3.64

217.1.2.130 217.1.3.131

Local Area Local Area

Network B Network C
Figure 9.4.3.2 IP addresses, expressed in dotted decimal notation, assigned to host and router interfaces
for an internet interconnecting three local area networks, A, B and C via a single router

Dotted decimal notation

The binary form of IP addresses was designed for reading/processing by machine. Although IPv4 addresses are
32-bit numbers, users rarely enter or read IP addresses in binary. Therefore, the software that interacts with users
employs a notation called dotted decimal notation which is more convenient for humans to understand. In this
notation, each 8-bit section of a 32-bit number is expressed as a decimal value separated from its neighbour by a full
stop. Table 9.4.3.1 shows some examples of 32-bit binary numbers and their equivalent dotted decimal forms.

32-bit binary number Equivalent dotted decimal

01011100 00010111 00011101 10100010 92.23.29.162
01001110 10010111 11101100 00000001 78.151.236.1
11011001 00000001 00000001 10000001 217.1.1.129
Table 9.4.3.1 Examples of 32-bit binary numbers and their equivalent in dotted decimal notation.The first row
is colour-coded to show the correspondence between the 8-bit sections and their decimal equivalent.
438 Single licence - Abingdon School
 9.4.3 IP address structure

Questions
1 Convert the following IPv4 32-bit addresses from binary to their equivalent dotted decimal form
(a) 11001100 00111111 01010101 00111111 (b) 11011001 10000111 00011001 00010011

Subnet
In Figure 9.4.3.2 one router with three interfaces is used to interconnect twelve
hosts. Key concept
The four hosts in local area network A and the router interface to which Subnet:
they are connected all have an IP address of the form 217.1.1.x with x A network of directly connected
interfaces.
chosen from the range 1 to 254. This means IP addresses in this range are
guaranteed to have the same most significant 24 bits in their IP address, i.e.
Subnet address:
110110010000000100000001 (see last row of Table 9.4.3.1). IP addressing assigns an address
In IP terms, local area network A connecting four host interfaces and one router to a subnet in a format a.b.c.d/x
which enables the network to be
interface forms a subnet.
identified, e.g. 217.1.1.0/24
IP addressing assigns an address to this subnet in the form a.b.c.d/x which where the value 24 indicates
is expressed as 217.1.1.0/24, where the /24 notation indicates that the most that the most significant 24 bits
identify the network.
significant 24 bits of the IPv4 32-bit IP address define the subnet address.
Additional hosts added to local area network A will be required to have
interfaces with an address of the form 217.1.1.x. This means that network A can
use 256 different IP addresses, i.e. 217.1.1.0 to 217.1.1.255.
However, 217.1.1.0 and 217.1.1.255 are both reserved IP addresses not to be used
for host/router interface identification, this leaves 254 IP address for identifying
host/router interfaces.
Figure 9.4.3.2 contains two other subnets:
• 217.1.2.0/24
• 217.1.3.0/24
Questions
2 How many bits identify the network in each of the following subnet addresses:
(a) 129.12.0.0/16 (b) 192.173.2.0/23?

The division of IP address into network ID and host ID

We have learned that an IP address is assigned to a host interface. Conceptually, each 32-bit IP address (IPv4) (and
each 128-bit IP address (IPv6)) is divided into two parts:
• a prefix or network identifier part (Net ID)
• a suffix or host identifier part (Host ID)
The prefix (most significant bits) identifies the physical network to which the host computer is connected. This
physical network is called a subnet. The prefix is more commonly known as the network ID or Net ID.
In Figure 9.4.3.2, the network ID is given by the most significant 24 bits of the IP address, e.g. 217.1.2 in IP
address 217.1.2.3.

Single licence - Abingdon School 439

9 Fundamentals of communication and networking

The suffix identifies a particular host interface connected to the subnet. The
Key concept
suffix is more commonly known as the Host ID. In Figure 9.4.3.2, the Host
Division of an IP address:
Conceptually, each 32-bit IP ID is given by the least significant 8 bits of the IP address, e.g. 3 in 217.1.2.3.
address (IPv4) (and each 128-bit This division is used as the basis of traffic routing between the physical
IP address (IPv6)) is divided
networks making up the Internet.
into two parts:
1. a prefix or network identifier For example, suppose a host with host interface IP address 217.1.1.25 in local
part (Net ID) area network A wishes to send an IP datagram to a host with host interface IP
2. a suffix or host identifier address 217.1.3.64 in local area network C.
part (Host ID) The LAN A host's IP software is able to determine by examining the
The prefix (most significant bits)
destination's IP address that the destination host is on a different subnet
identifies the physical network
to which the host computer (217.1.3.0/24) with network ID 217.1.3 and therefore cannot be reached
is connected. This physical directly.
network is called a subnet. The LAN A host therefore sends the IP datagram to the router interface
The prefix is more commonly
217.1.1.1.
known as the network ID or
Net ID.
This router examines the network ID of the destination's IP address contained
The suffix identifies a particular in the IP datagram and forwards this datagram to its interface with IP address
host interface connected to the 217.1.3.1 since this interface is connected to subnet 213.1.3.0/24. This
subnet. The suffix is more com- interface places the IP datagram onto the link it is connected to and which host
monly known as the Host ID.
interface 217.1.3.64 is also connected.
The IP datagram is is then read by host interface 217.1.3.64.
The Global Internet
Every interface on every host and router in the global Internet must have an IP address that is globally unique
(except for interfaces behind NATs - see Chapter 9.4.8). IP addresses are assigned in a coordinated manner so that
routing is facilitated. The strategy is called Classless InterDomain Routing (CIDR). It generalises subnet addressing
to the global Internet. As with subnet addressing, the 32-bit IP address in IPv4 is divided into two parts using the
same dotted decimal form a.b.c.d/x, where x indicates the number of bits for the prefix. The prefix constitutes the
network part of the IP address.
Every organisation that wishes to send and receive e-mail, or gain access to the Internet, needs at least one globally
unique IP address.
An organisation is typically assigned more than one unique IP address as a block of contiguous addresses with a
common prefix. The IP addresses of all devices within the organisation will share this common prefix.
For example, the organisation Jisc Services Limited, more commonly known as Janet, is a private, UK government-
funded organisation, which provides computer network and related collaborative services to UK research and
education (Further and Higher). To see how blocks of contiguous IP addresses are assigned for the Janet network
visit https://ipinfo.io/AS786. Table 9.4.3.2 shows a sample of these blocks.

Netblock Description No of IP addresses

129.12.0.0/16 University of Kent 65536
192.195.42.0/23 University of Ulster 512

Table 9.4.3.2 Sample of IP address contiguous block allocation within the Janet organisation.

440 Single licence - Abingdon School

 9.4.3 IP address structure

The University of Kent has subdivided its block of IP addresses into 128 subnets. Table 9.4.3.3 shows a sample of
these addresses of these subnets.

129.12.0.0/23 129.12.2.0/23 129.12.4.0/23 1 29.12.6.0/23

● ● ● ●
● ● ● ●
129.12.248.0/23 129.12.250.0/23 129.12.252.0/23 129.12.254.0/23

Table 9.4.3.3 Sample of subnet addresses for University of Kent.

Table 9.4.3.4 shows binary equivalent of the subnet addresses shown in Table 9.4.3.3 and the corresponding
Network ID expressed in binary. Note that the most significant 16 bits of each subnet Network ID is the same. This
corresponds to the prefix for the University of Kent given by 129.12.0.0/16.

Subnet Block Equivalent binary value Network ID (prefix)

129.12.0.0/23 10000001 00001100 00000000 00000000 10000001 00001100 0000000
129.12.2.0/23 10000001 00001100 00000010 00000000 10000001 00001100 0000001
129.12.4.0/23 10000001 00001100 00000100 00000000 10000001 00001100 0000010
129.12.6.0/23 10000001 00001100 00000110 00000000 10000001 00001100 0000011
● ● ●
● ● ●
129.12.248.0/23 10000001 00001100 11111000 00000000 10000001 00001100 1111100
129.12.250.0/23 10000001 00001100 11111010 00000000 10000001 00001100 1111101
129.12.252.0/23 10000001 00001100 11111100 00000000 10000001 00001100 1111110
129.12.254.0/23 10000001 00001100 11111110 00000000 10000001 00001100 1111111

Table 9.4.3.4 Network IDs for the subnets for University of Kent. The university's Network ID is shown in red

Using IP address prefix to route an IP datagram

Suppose that an IP datagram is addressed to a remote destination whose IP address is 129.12.3.237. When an
organisation is allocated a block of IP addresses these are registered and Internet routers are supplied with the
organisation's Network ID expressed in a.b.c.d/x form. For example, Internet routers store the following for the
University of Kent Network ID: 129.12.0.0/16. The most significant 16 bits of 129.12.0.0, the prefix, are used by
Internet routers to relay the IP datagram to the gateway router to the University of Kent's network. Once this IP
datagram has reached this gateway router, this routing table is searched to determine how to route the IP datagram
within the University of Kent. If this gateway router is connected to 128 links (subnets) then the routing table will
contain an entry for each link which corresponds to its subnet address or Network ID (see Table 9.4.3.4), e.g. the
first entry will be 129.12.0.0/23, the second entry 129.12.2.0/23, etc. The binary equivalent of the IP datagram's
destination address 129.12.3.237 is (the most significant 23 bits are shown in red)
10000001 00001100 00000011 11101101

The gateway router's routing table entries indicate that the most significant 23 bits of the IP datagram should be
examined to determine which subnet to forward it to. The 23-bit prefix of 129.12.3.237 in binary is
10000001 00001100 0000001

Writing this in 32 bits by adding nine trailing zeroes we get

10000001 00001100 00000010 00000000

Single licence - Abingdon School 441

9 Fundamentals of communication and networking

In dotted decimal notation, this 32-bit 129.12.0.0

binary value is 129.12.2.0 which is the
Network ID of the second subnet. The .0
.12.2
gateway router will use this result to forward 129

the IP datagram to the second link. The

host with host interface 129.12.3.237 is 129.12.3.237
connected to this link as shown in Figure The Internet Router 128 subnets
9.4.3.3. so the IP datagram has reached its
destination. Gateway router University of Kent
In practice it is more likely that the gateway
router will be connected to a hierarchy of
routers internal to the university with the 129.12.254.0
links to the subnets connected to the routers
Figure 9.4.3.3 Gateway router at University of Kent connected
in the last stage of this hierarchy. In this
to 128 subnets within university.
way, each final stage router will require fewer
interfaces, perhaps just eight and likewise, the gateway router could require only eight interfaces.
Table 9.4.3.5 shows a section of the University of Kent's gateway router subnet Network IDs stored as 32-bit values
assuming that the university's network is organised as shown in Figure 9.4.3.3. Note that nine trailing zeroes have
been added to each 23-bit prefix to create 32-bit values. These nine least significant bits are reserved for the Host
IDs of host interfaces connected to a subnet. Note that the same Host IDs can be used in each subnet because when
combined with the 23-bit subnet address a unique IP address always results.

Network ID Link no
10000001 00001100 00000000 00000000 1
10000001 00001100 00000010 00000000 2
10000001 00001100 00000100 00000000 3
10000001 00001100 00000110 00000000 4
● ●
● ●
10000001 00001100 11111000 00000000 125
10000001 00001100 11111010 00000000 126
10000001 00001100 11111100 00000000 127
10000001 00001100 11111110 00000000 128

Table 9.4.3.5 Network IDs for the subnets for University of Kent expressed
as 32-bit values and the corresponding Link no expressed in decimal.

Questions
3 The following 32-bit value is an entry in the University of Kent's gateway router routing table (see Table
9.4.3.5) 10000001 00001100 00000110 00000000
Which of the following destination IP addresses expressed in dotted decimal notation match this entry
(a) 129.12.7.38 (b) 129.12.6.123 (c) 129.12.5.223 (d) 129.13.6.45?

442 Single licence - Abingdon School

 9.4.3 IP address structure

Questions
4 An IP datagram is addressed to a remote destination whose IP address is 129.12.33.114. Internet routers
forward this datagram to the gateway router at the University of Kent. Explain how this datagram is
forwarded to the host interface with IP address 129.12.33.114 within the University of Kent's network.
You may assume that Figure 9.4.3.3 describes this university's network setup.

In this chapter you have covered:

■■ How an IP address is split into a network identifier part and a host identifier part
■■ The background knowledge needed to understand this splitting of the IP address:
• Uniform addressing scheme
• Dotted decimal notation
• The meaning of the term subnet and subnet address expressed in the form a.b.c.d/x
• The division of an IP address into a prefix and a suffix
• The prefix is the network identifier part and the suffix the host identifier
• Generalisation of subnet addressing to the global Internet
• Using an IP address prefix to route an IP datagram

Single licence - Abingdon School 443

9 Fundamentals of communication and networking

9.4 The Transmission Control Protocol/

Internet Protocol (TCP/IP) protocol
Learning objectives:
■■ 9.4.4 Subnet masking
Subnet mask
■■Know how a subnet mask is We learned in Chapter 9.4.3 that a 32-bit IP address in IPv4 is divided into
used to identify the network
two parts using the dotted decimal notation a.b.c.d/x, where x indicates the
part of the IP address
number of bits for the prefix. The prefix constitutes the network part of the
IP address. For example, we can divide 129.12.7.38/23 into a prefix and suffix
Key concept expressed in binary as follows
23 bits
Subnet mask:
129.12.7.38 = 10000001 00001100 00000111 00100110
The subnet mask is the number
of bits assigned to the prefix 129.12.7.38/23 = 10000001 00001100 0000011 (prefix)
part of an IP address. The prefix
identifies the Network ID or 129.12.7.38 = 1 00100110 (suffix)
subnet address. 9 bits

The suffix is the Host ID. In this example nine bits are allocated to the suffix.
The prefix is the Network ID or subnet address. In IPv4 it is expressed as a
32-bit value by replacing the suffix or Host ID part with zeroes. For example,
Key concept
the Network ID of the host interface with IP address 129.12.7.38 is in binary
Using a subnet mask:
The subnet mask is used by a 10000001 00001100 00000110 00000000
computer/router to obtain the which in dotted decimal form is 129.12.6.0/23.
Network ID or subnet address
of the subnet to which its Therefore to obtain the Network ID for a given IP address we need to know
network interface is connected. how many bits are assigned to the prefix part. This is known as the subnet
This may be done with an mask. The Network ID is also known as the subnet address.
AND operation applied to the
computer interface's IP address Given a 32-bit IP address such as 129.12.7.38 it is possible to obtain its
and a mask of the same number Network ID expressed as a 32-bit value by a bitwise AND operation applied to
of bits constructed from the the 32-bit mask derived from the subnet mask and the IP address. If the subnet
subnet mask.
mask is 23 then the 32-bit mask for the bitwise AND operation will consist of
twenty three ones for the most significant bits followed by nine zeroes.

32-bit mask = 11111111 11111111 11111110 00000000

129.12.7.38 = 10000001 00001100 00000111 00100110

AND
129.12.6.0 = 10000001 00001100 00000110 00000000

The result of the AND operation is the Network ID or subnet address

expressed in 32 bits.

Single licence - Abingdon School 444

9 Fundamentals of communication and networking

Figure 9.4.4.1 shows the Internet Protocol version

4 (TCP/IPv4) Properties window of a Microsoft
Windows 7 computer. The field labelled subnet
mask shows the AND mask that is used by this
computer when it needs to know to which subnet it is
connected.
Why does a computer need to be able to "calculate"
the subnet address of the subnet it is connected to?
When a computer wishes to send an IP datagram to
another computer it needs to know whether or not it
is directly connected to this computer, i.e. connected
to the same subnet as the destination. If linked directly
by wire or wireless, then the hardware address of the
destination computer is obtained and attached to the
IP datagram. The wire or wireless frame encapsulating
the datagram is then sent over the direct link to the
destination computer.
A computer determines to which IPv4 subnet it is Figure 9.4.4.1 IPv4 properties window of a Microsoft
connected by applying a bitwise AND operation to Windows computer showing the subnet mask that is
its IP address and the 32-bit mask derived from the ANDed with the IP address to obtain the subnet address.
subnet mask. The destination's subnet is determined
in a similar way.
If the destination computer is not on the same subnet then the sending computer obtains the hardware address of
the gateway router to which it is directly connected. It then uses this hardware address to send the IP datagram to
this router. The gateway router will use its routing tables to forward the IP datagram to the subnet of the destination
computer so the datagram is able to reach its destination.
Figure 9.4.4.1 shows that the TCP/IP software on the Windows 7 computer has recorded 192.168.1.1 as the IP
address of the default gateway. The subnet address of the Windows 7 computer and the default gateway can be
calculated as follows

255.255.255.0 = 11111111 11111111 11111111 00000000 (mask)

192.168.1.4 = 11000000 10101000 00000001 00000100 (sending computer's IP address)

AND
192.168.1.0 = 11000000 10101000 00000001 00000000 (subnet address /Network ID)

255.255.255.0 = 11111111 11111111 11111111 00000000 (mask)

192.168.1.1 = 11000000 10101000 00000001 00000001 (default gateway IP address)

AND
192.168.1.0 = 11000000 10101000 00000001 00000000 (subnet address /Network ID)

This shows that the Windows 7 computer and the default gateway are on the same subnet (192.168.1.0) and
therefore directly connected.

445 Single licence - Abingdon School

 9.4.4 Subnet masking

Questions
1 Table 9.4.4.1 shows subnet masks and IP addresses expressed using IPv4 dotted decimal notation
Subnet mask IP address Network ID or subnet address
255.255.255.0 192.168.2.15
255.255.0.0 192.168.253.234
255.255.252.0 192.168.253.234
Table 9.4.4.1 Subnet masks and IP address expressed using IPv4 dotted decimal notation
Complete the Network ID column for the given subnet masks and IP addresses.

2 A computer with a host interface IP address 192.168.1.5 sends an IP datagram to a computer with
IP address 192.168.2.3. In each case the subnet mask expressed in IPv4 dotted decimal notation is
255.255.255.0. The default gateway for the computer 192.168.1.5 is 192.168.1.1. Explain why the IP
datagram is sent to the default gateway for forwarding to 192.168.2.3.

In this chapter you have covered:

■■ How a subnet mask is used to identify the network part of the IP address

Single licence - Abingdon School 446

9 Fundamentals of communication and networking

9.4 The Transmission Control Protocol/

Internet Protocol (TCP/IP) protocol
Learning objectives:
■■ 9.4.5 IP standards
Internet addressing 1969
■■Know that there are currently ARPANET, the forerunner to the Internet, consisted initially of four nodes.
two standards of IP address,
Each node was an Interface Message Processor (IMP) - see Chapter 9.3.1 for an
v4 and v6
image of an IMP.
■■Know why v6 was introduced. In 1969, several months before the first of the four original ARPANET nodes
became operational, Stephen Crocker at UCLA authored the first Request For
Comment, RFC 1, "Host Software" - https://tools.ietf.org/html/rfc1.
RFC 1 proposed specifications for the Interface Message Processor software
and host-to-host connections in which 5 bits were allocated to a message's
destination address. The allocation of 5 bits would theoretically provide 25 or
32 destination addresses as follows:

00000 00001 00010 00011 00100 00101 00110 00111

01000 01001 01010 01011 01100 01101 01110 01111

10000 10001 10010 10011 10100 10101 10110 10111

11000 11001 11010 11011 11100 11101 11110 11111

An increase in the total number of addresses could be achieved by adding more

bits to these binary codes. Each additional bit doubles the total number of
addresses. Table 9.4.5.1 shows the total number of addresses for 6, 7 and 8 bits
respectively.

Number of address bits Total number of addresses

6 64
7 128
8 256
Table 9.4.5.1 Total number of addresses for a given number of bits

The emergence of electronic mail

ARPANET was envisaged by its designers to be a network for facilitating
resource and file transfer. They did not forsee the emergence of electronic mail
and the unexpected growth in the ARPANET network that followed as a result.
A radical change took place in 1974 with the design of TCP. TCP makes the
hosts responsible for the reliability of transmission instead of the network as
was the case with ARPANET. RFC 675 "Specification of Internet Transmission
Control Program", December 1974, contains the first attested use of the term
Single licence - Abingdon School 447
9 Fundamentals of communication and networking

internet, as a shorthand for internetworking. TCP/IP followed later in the decade and the Internet was born along
with IP addressing.

Internet addressing 1981 and IPv4

RFC 791 (1981) - https://tools.ietf.org/html/rfc791 - introduced the Internet Protocol standard, later known as
IPv4, expanding the size of each IP address to a 32-bit code divided into a network prefix and a host prefix. The
address length of 32 bits provides a theoretical pool of 232 or approximately 4.3 billion unique Internet addresses.
However, as far back as the early 1990s it was forecast that the Internet would run out of IPv4 addresses and that a
transition to a new version of the protocol offering a larger address space would be required if the problem was to
be avoided. In February 2011, IANA allocated the last remaining pool of unassigned IPv4 addresses to a regional
registry (the regional registries are ARIN, RIPE NCC, APNIC, LACNIC, and AfriNIC). Organizations acquire
IPv4 addresses from their Regional Internet Registry (RIR) or their service providers. Service providers acquire their
IPv4 addresses from their Regional Internet Registry. ARIN (American Registry for Internet Numbers) ran out of
IPv4 addresses on 24th September 2015.

IPv6
In response to the projected problem of running out of IPv4 addresses, the technical community came up with a
specification for IP version 6 (IPv6) in the mid-1990s.
IPv6 allocates 128 bits for IP addresses. This gives a theoretical total number of addresses of 2128 or roughly 3.4 x
1038 addresses. This is more than enough for the foreseeable future. By way of comparison, the number of addresses
required to uniquely label all the grains of sand on planet earth would be 7.5 x 1018. However, adoption of IPv6
is not a straightforward matter because the public Internet is an IPv4 router network designed to work with
32-bit addresses not 128-bit addresses. Whilst new IPv6-capable systems can be made backwards compatible, i.e.
send route and receive IPv4 datagrams, already deployed IPv4-capable systems are not capable of handling IPv6
datagrams.

Questions
1 What is the theoretical total number of IP addresses for
(a) IPv4? (b) IPv6?

2 Why was IPv6 introduced?

In this chapter you have covered:

■■Two standards of IP address, v4 and v6
■■Why v6 was introduced.

448 Single licence - Abingdon School

9 Fundamentals of communication and networking

9.4 The Transmission Control Protocol/

Internet Protocol (TCP/IP) protocol
Learning objectives:
■■ 9.4.6 Public and private IP addresses
Private address spaces
■■Distinguish between routable The address spaces 10.0.0.0/8, 172.16.0.0/12 and 192.168.0.0/16 are the
and non-routable IP addresses
three regions of the IPv4 address space that are reserved (RFC 1918 - https://
Key fact tools.ietf.org/html/rfc1918) for private TCP/IP networks. Addresses in these
Private IPv4 address spaces: address spaces are known as private addresses. The use of private IP addresses
The address spaces 10.0.0.0/8,
requires no coordination by IANA (Internet Assigned Numbers Authority)
172.16.0.0/12 and 192.168.0.0/16
are reserved for private IPv4 TCP/
or an Internet registry because they are only unique within a private TCP/IP
IP networks. network. Private addresses can be reused in any private TCP/IP network which
Private IPv4 address ranges: means that they are not unique across the public, global Internet.
192.168.0.0 - 192.168.255.255
172.16.0.0 - 172.31.255.255
Table 9.4.6.1 shows the range of private IP addresses for each reserved address
10.0.0.0 - 10.255.255.255 space.

Address space Range Total no of addresses

192.168.0.0/16 192.168.0.0 - 192.168.255.255 65,536
172.16.0.0/12 172.16.0.0 - 172.31.255.255 1,048,576
10.0.0.0/8 10.0.0.0 - 10.255.255.255 16,777,216

Table 9.4.6.1 The three private IP address spaces, their range and total number
Key concept Non-routable IP addresses
Non-routable IP address: Hosts within a given private TCP/IP network can send IP datagrams to each
Private IP addresses are non-
other using addresses assigned to each which are chosen from a private address
routable.
Routers in the public, global
space, e.g. 192.168.0.0/24. However, IP datagrams forwarded from the private
Internet will reject IP datagrams network into the larger public, global Internet cannot use these addresses as
with source and/or destination source or destination address because there will be many other connected
IP addresses which fall within networks which use addresses from the same private address space. Routers
a private address range and will
in the public, global Internet will reject such IP datagrams and will not route
therefore not route them.
them. Private IP addresses are therefore said to be non-routable.
Routable IP addresses:
Routable IP addresses
Public IP addresses are routable
IP addresses because their
The assignment to hosts of IP addresses is coordinated by IANA and the
assignment is coordinated by Regional Internet Registries to ensure that each host's IP address is globally
IANA and the Regional Internet unique. This enables routers of the global, public Internet to use the source
registries to ensure that hosts/ and destination public IP addresses of two communicating hosts to route IP
routers are uniquely identified
datagrams through the Internet from one host to the other. Public IP addresses
globally. Routing tables in
routers contain globally unique are routable IP addresses because the coordination provided by IANA and
address information to enable the Regional Internet registries ensures that routing tables in routers contain
successful and unambiguous globally unique address information to enable successful and unambiguous
routing of IP datagrams. routing of IP datagrams.

Single licence - Abingdon School 449

9 Fundamentals of communication and networking

Questions
1 Read sections 1 to 4 of RFC 1918 ( https://tools.ietf.org/html/rfc1918) then answer the following
questions

(a) Give three examples where external connectivity of hosts and routers in an organisation's TCP/IP
network might be unnecessary.

(b) Give one reason why it is good practice to use private IP addresses where possible in IPv4 TCP/IP local
area networks that are connected to the Internet.

2 State the three regions of the IPv4 address space that are reserved for private TCP/IP networks.

3 Distinguish between routable and non-routable IP addresses.

In this chapter you have covered:

■■The three IPv4 address spaces that are reserved for private TCP/IP networks
■■How to distinguish routable from non-routable IP addresses

450 Single licence - Abingdon School

9 Fundamentals of communication and networking

9.4 The Transmission Control Protocol/

Internet Protocol (TCP/IP) protocol
Learning objectives:
■■ 9.4.7 Dynamic Host Configuration Protocol (DHCP)
Obtaining a host address
■■Understand the purpose and How does a host get an IP address for its network interface when it joins a
function of the DHCP system
subnet?
One way of doing this is for the host's user to assign an IP address to the host
interface that connects to the subnet.
This approach assumes
• some expertise on the part of the user
• that the user is able to choose an IP address which is not already in
use
• that the user is able to choose an IP address whose network
identification part matches the subnet address
Assigning an IP address is not all that will need to be done.
A subnet mask will also need to be assigned so that the host interface can
determine the subnet address of the subnet to which it is connected.
The IP address of at least one DNS server will also need to be assigned to the
host interface so that the host can consult this server to convert domain names
into their IP addresses.
The network's administrator will know how to set up hosts and could be called
to the host to configure it. However, this is not always convenient.
For example, the host in question could be a laptop attempting to connect
Key point
wirelessly to a wireless subnet belonging to a cafe offering free WiFi access to its
Functions of the DHCP
customers.
system:
1. To allocate IP addresses to Functions of the DHCP system
hosts via three mechanisms: The Internet Engineering Task Force specified a solution to the above problem
• automatic allocation of a in RFC 2131 (https://www.ietf.org/rfc/rfc2131.txt) .
permanent IP address
The Dynamic Host Configuration Protocol (DHCP) provides configuration
• dynamic allocation of a
temporary IP address parameters to Internet hosts. DHCP consists of two components:
(limited period of time) • a protocol for delivering host-specific configuration parameters from
• conveying a manually
a DHCP server to a host, e.g. subnet mask
assigned IP address to a
host • a mechanism for allocation of IP addresses to hosts.
2. To deliver host-specific
DHCP supports three mechanisms for IP address allocation.
configuration parameters
such as subnet mask to a • In "automatic allocation", DHCP assigns a permanent IP address to
host. a client

Single licence - Abingdon School 451

9 Fundamentals of communication and networking

• In "dynamic allocation", DHCP assigns an IP address to a client for a limited period of time (or until the
client explicitly relinquishes the address)
• In "manual allocation", a client's IP address is assigned by the network administrator, and DHCP is used
simply to convey the assigned address to the client.
A particular network will use one or more of these mechanisms, depending on the policies of the network
administrator.
In addition to host IP address assignment, DCHP also
allows a host to learn additional information, such as
the subnet mask for the subnet it is connected to, the
address of its first hop-router (often called the default
gateway) to which it sends IP datagrams for hosts on
different subnets, and the address of a DNS server.
Figure 9.4.7.1 shows the TCP/IPv4 Properties window
for a Microsoft Windows 7 host. It shows the manually
configured properties for the host interface's IP address,
subnet mask and default gateway. The DNS servers
have not been configured. If the radio buttons "Obtain
an IP address automatically" and "Obtain DNS
server address automatically" were selected then these
property fields would be completed automatically by
preconfigured information obtained from a DHCP
server in a client-server operation.
The four-step DCHP process Figure 9.4.7.1 TCP/IPv4 properties
Step 1 window for a Microsoft Windows 7 host
The first task of a new host (client) that wishes to join
an existing TCP/IP network is to find a DHCP server. It does this by broadcasting a DHCP discover message
over the network. The host at this point does not know the subnet address of the network or the IP address of any
DHCP server. So it uses a special broadcast destination address of 255.255.255.255 and a "this host" source IP
address of 0.0.0.0. This broadcast will reach all nodes attached to the subnet.
Step 2
A DHCP server receiving a DHCP discover message responds to the client with a DHCP offer message that is
broadcast to all nodes on the subnet, again using the IP broadcast address of 255.255.255.255. The DHCP offer
message contains the proposed IP address for the client, the network mask, an IP address lease time (the amount
of time for which the IP address will be valid), and a transaction ID extracted from the DHCP discover message
which links this message with the new client.
Step 3
The new client responds to the offer with a DHCP request message which echoes back the configuration
parameters.
Step 4
The server responds to the DHCP request message with a DHCP ACK message, confirming the requested
parameters.

452 Single licence - Abingdon School

 9.4.7 Dynamic Host Configuration Protocol (DHCP)

Purpose of DHCP system Key point

The primary purpose of DHCP is to automate the setting up of hosts that are
Purpose of the DHCP system:
connecting to a TCP/IP network. This is particularly important where hosts The primary purpose of DHCP
come and go frequently and IP addresses are needed only for a limited period is to automate the setting up of
of time, e.g., in a wireless LAN. Another example would be an ISP that has hosts that are connecting to a
TCP/IP network.
16000 residential customers but no more than 4000 are ever online at the same
time. In this case, rather than needing a block of 16384 addresses, a DHCP
server that assigns addresses dynamically needs only a block of 4096 addresses
(e.g. a block of the form a.b.c.d/22).
Each time a host joins the network, the DHCP server allocates an arbitrarily
chosen IP address from its current pool of available IP addresses. Each time a
host leaves, its address is returned to the pool.

Questions
1 What is the primary purpose of the DHCP system?

2 Give two examples where using DHCP to configure new client hosts is preferable to manual configuration
by other means.

3 Explain the function of the DHCP system.

4 Why is a DHCP server required in a network that uses DHCP?

In this chapter you have covered:

■■ The purpose and function of the DHCP system

Single licence - Abingdon School 453

9 Fundamentals of communication and networking

9.4 The Transmission Control Protocol/

Internet Protocol (TCP/IP) protocol
Learning objectives:
■ 9.4.8 Network Address Translation (NAT)
Using non-routable IPv4 addresses with a NAT-enabled router
■ Explain the basic concepts of Obtaining blocks of routable IPv4 addresses to cover every computing device
NAT and why it is used
in a home, small office or school LAN is problematic because of the shortage
Key Concept of such addresses. However, if such computing devices wish to connect to
Network Address Translation the public Internet and communicate with other Internet-connected TCP/
(NAT) protocol: IP-enabled devices then a solution is required which allows devices to use
The NAT protocol is specified IP addresses drawn from the non-routable IP address spaces 10.0.0.0/8,
in RFC 2663 and RFC 3022.
172.16.0.0/12, 192.168.0.0/16.
Network Address Translation
is a method by which a device,
The current approach to the problem is to use a NAT-enabled router.
e.g. a router, is used to connect The NAT-enabled router has an interface that is part of the LAN as shown in
an isolated address realm Figure 9.4.8.1. The subnet address of this LAN is 192.168.1.0/24 and the IP
assigned private unregistered addresses of the interfaces in this subnet are the private, non-routable addresses.
addresses, e.g. a LAN, to an
192.168.1.1, 192.168.1.2, 192.168.1.3 and 192.168.1.4, respectively. Devices
external realm assigned globally
unique registered addresses, e.g. within the home LAN can send IP datagrams to each other using 192.168.0/24
the Internet. addressing. However, IP datagrams forwarded beyond the home LAN into the
global Internet cannot use these addresses because 192.168.1.0/24 addresses will
be rejected by Internet routers.
NAT translation table Symbol for a router
WAN side LAN side
source =192.168.1.2 : 4236
92.23.29.162 : 5431 192.168.1.2 : 4236
destin. =129.11.26.33 : 80

192.168.1.2
NAT-enabled
① router ②
④
192.168.1.3 source =92.23.29.162 : 5431
192.168.1.1 destin. =129.11.26.33 : 80

92.23.29.162 source =129.11.26.33 : 80

source =129.11.26.33 : 80
destin. =192.168.1.2 : 4236 destin. =92.23.29.162 : 5431
Internet
192.168.1.4 ③

Web server 129.11.26.33

Home LAN 192.168.1.0/24
Figure 9.4.8.1 NAT-enabled router connecting a Home LAN to the Internet

The NAT-enabled router deals with this issue by presenting itself to the outside
world not as a router but as a single device with a single IP address. In Figure
9.4.8.1 all traffic leaving this NAT-enabled router for the global Internet has a

Single licence - Abingdon School 454

9 Fundamentals of communication and networking

source IP address of 92.23.29.162. All traffic entering this router must have a destination address of 92.23.29.162.
The role performed by this NAT-enabled router is to hide the details of the home LAN from the outside world.

NAT translation table

If all IP datagrams arriving from the Internet have the same destination IP address, 92.23.29.162, i.e. the NAT
router's Wide Area Network(WAN) interface address, how does the router know to which LAN host it should
forward the IP datagram? The answer is the NAT Translation table in the NAT-enabled router together with the
use of port numbers. Figure 9.4.8.2 shows the rear of a NAT-enabled router.

Figure 9.4.8.2 NAT-enabled router showing four LAN sockets labelled Ethernet 1, 2, 3 and 4, a WAN
socket, and a broadband socket (an alternative way of connecting to the Internet)
The use of port numbers
To understand how port numbers are used to enable LAN hosts to use the WAN, consider the example shown in
Figure 9.4.8.1. In this figure, LAN host with IP address 192.168.1.2 requests a Web page from a Web server with
IP address 129.11.26.33 listening on port 80. The LAN host assigns an arbitrarily chosen source port number 4236
to the request and sends the IP datagram into the LAN as a link-layer packet addressed to the NAT-enabled router -
see Figure 9.4.8.1 1. Note that the destination IP address is 129.11.26.33 and the destination port number is 80.
The NAT-enabled router receives the IP datagram and makes two changes to this datagram:
• It generates a new source port number 5431 for the datagram and replaces the original source port
number 4236 with this new source port number
• It replaces the source IP address 192.168.1.2 with its WAN-side IP address 92.23.29.162.

455 Single licence - Abingdon School

 9.4.8 Network Address Translation (NAT)

The NAT-enabled router adds the entry 92.23.29.162 : 5431 192.168.1.2 : 4236 to its NAT translation table as
shown in Figure 9.4.8.1. and sends the IP datagram 2 into the public Internet.
When generating a replacement source port number, the NAT router chooses a source port number which is not
currently present in the NAT translation table. Port numbers are 16-bit numbers so the NAT router with a single
WAN-side IP address is able to support 216 simultaneous connections.
The Web server responds with an IP datagram whose destination address is the WAN-side IP address of the router,
and whose destination port number is 5431 - see Figure 9.4.8.1 3. Note that the Web server is not aware that the
Web page request has come from the LAN host with interface IP address 192.168.1.2.
On arrival at the NAT-enabled router, the received datagram's destination IP address 92.23.29.162 together with
the destination port number 5431 is matched to the corresponding entry in the NAT translation table to obtain
the LAN host IP address and port number. The router then constructs an IP datagram using these as destination IP
address and destination port number, respectively, before dispatching this datagram into the home LAN where it
is read by host 192.168.1.2 and passed to the application (identified by port number 4236) that initiated the Web
page request - see Figure 9.4.8.1 4.

Questions
1 Why does an IPv4 TCP/IP LAN need a NAT-enabled router if hosts with private IP addresses are to
connect to the Internet?

2 Explain how port numbers are used to enable LAN hosts using private IPv4 addresses to exchange IP
datagrams with the global Internet.

3 Port numbers are associated in the TCP/IP specifications with layer 4, the Application layer, of the TCP/IP
protocol stack (Chapter 9.4.1). How does the use of port numbers by the NAT protocol differ from their
use by the Application layer protocol?

4 The end-to-end principle is one of the underlying system principles of the Internet (Chapter 9.4.1).
(a) Why is the NAT protocol a violation of this principle?
(b) Why might using IPv6 addressing obviate the need for the NAT protocol?

In this chapter you have covered:

■■ The basic concepts of NAT:
• A method used to connect an isolated address realm assigned private unregistered addresses, e.g. a LAN, to
an external realm assigned globally unique registered addresses, e.g. the Internet.
• A NAT-enabled router enables this connection to the outside world by presenting itself not as a router to the
the world but as a single device with a single IP address.
• The internal hosts are identified by port no and their interaction with the outside world is recorded in a
NAT-enabled router in its NAT translation table which maps the host port no to a router port no.
■■ NAT is necessary because of the scarcity of sufficient blocks of IPV4 addresses

Single licence - Abingdon School 456

9 Fundamentals of communication and networking

9.4 The Transmission Control Protocol/

Internet Protocol (TCP/IP) protocol
Learning objectives:
■ 9.4.9 Port forwarding
Port forwarding is one method by which clients in other LANs connected to
■ Explain the basic concept of
the Internet may reach servers assigned private unregistered IP addresses and
port forwarding and why it is
located behind a NAT router in another LAN. For example, in Figure 9.4.9.1
used
two servers, a Web server listening on port 80 and a multiplayer Minecraft
Key Concept server listening on port 25565, are located in Home LAN 1 behind a NAT-
Port forwarding: enabled router. Both have private, unregistered and therefore non-routable
In port forwarding, specific IPv4 addresses. Clients in other LANs such as Home LAN 2 and Home LAN
router ports (TCP and UDP)
3 can request Web pages from the Web server in Home LAN 1 by sending
are opened up so that the
router is able to direct all traffic their HTTP requests in IP datagrams, marked for the attention of TCP port
arriving at these ports to a 80, to the router interface with routable IP address 92.23.29.162. This router
specific internal IP address. will consult its port mapping table and find that 92.23.29.162 : 80 maps to
192.168.1.4 : 80 on the internal network. The HTTP request IP datagram will
then be forwarded to the Web server in Home LAN 1 at 192.168.1.4.

192.168.1.2 192.168.1.3
Symbol for a router

Home LAN 2
192.168.1.0/24
192.168.1.2
192.168.1.2 NAT-enabled
router
129.11.26.33
192.168.1.1

192.168.1.3 192.168.1.3
192.168.1.1
146.23.18.5
Web server
192.168.1.4 92.23.29.162
192.168.1.4
Port mapping table
Internet
Minecraft WAN side LAN side Another LAN
server
192.168.1.5 92.23.29.162 : 80 192.168.1.4 : 80
Another LAN Another LAN
92.23.29.162 : 25565 192.168.1.5 : 25565
Home LAN 3
Home LAN 1

Figure 9.4.9.1 Port forwarding to enable access from the Internet to a Web server and a Minecraft server
assigned unregistered private addresses in home LAN 1

In a similar manner, clients in other LANs may connect to the Minecraft server
in Home LAN 1 via port forwarding set up in the NAT-enabled router with
public IP address 92.23.29.162.
Figure 9.4.9.2 shows the setup window for port forwarding for a NAT-enabled
router.
Single licence - Abingdon School 457
9 Fundamentals of communication and networking

Figure 9.4.9.2 Setup window for port forwarding on a NAT-enabled router

The port mapping table that enables port forwarding uses private IPv4 address for the internal servers. It is therefore
important that the IP addresses of these servers do not change because if they do the mapping will no longer work.
These private IP addresses must be assigned to the servers as static IP addresses. This may be done in one of two
ways:
• The network administrator sets up these addresses in DHCP so that they are assigned permanently by
DHCP
• DHCP is not used and these addresses are assigned at the servers by using network configuration software
running on the servers.

Questions
1 In what circumstance would port forwarding be used?

2 A Web server connected to a LAN uses the private IP address 172.31.78.4. The Web server listens for
HTTP requests on port 80. A NAT-enabled router with two interfaces, 172.31.78.1 and 146.31.18.97 is
also connected to this LAN. Its 146.31.18.97 interface is connected to the Internet. Explain how the port
mapping table in this router would be set up to allow port forwarding of HTTP requests arriving from the
Internet.
3 The Web server in Question 2 is switched off for maintenance for two days. When it is brought back
online, port forwarding no longer works. State one likely reason why port forwarding no longer works and
suggest a solution.

In this chapter you have covered:

■ The basic concept of port forwarding and why it is used.

458 Single licence - Abingdon School

 9 Fundamentals of communication and networking

9.4 The Transmission Control Protocol/Internet Protocol (TCP/IP)

Learning objectives
■■Be familiar with the
client-server model
■■ 9.4.10 Client-server model
What is it?
■■Be familiar with the Websocket When you order a pizza from a pizza delivery service, called Pizza House,
protocol
you are accessing the service offered by Pizza House, the server, as one client
■■Be familiar with the principles amongst many. Your REQUEST is processed by Pizza House which generates
of Web CRUD Applications a RESPONSE by creating the pizza you ordered before arranging for it to be
and REST: delivered to you. This is an example of a client-server model. In this model, we
expect a server to be available continuously, i.e. Pizza House will continue to
■■CRUD is an acronym for
service orders from clients 24 hours a day, seven days a week – Figure 9.4.10.1.
• C – Create CLIENTS SERVER

• R – Replace
REQUEST
• U – Update
24 hour
• D – Delete Customer 3 Customer 2 Customer 1

■■REST enables CRUD to be Ordered

Caprisiosa
Ordered
Margherita
Ordered
Chicken & Ham Pizza House
mapped to database functions
(SQL) as follows:

• GET → SELECT

• POST → INSERT CONNECTOR

RESPONSE

• DELETE → DELETE Figure 9.4.10.1 The client server model for Pizza House service
• PUT → UPDATE
There is a protocol that client and server use in this scenario that is common to
■■Compare JSON (Javascript many similar scenarios in which a customer (client) orders something over the
Object Notation) with XML
telephone or online from a supplier (server):
1. there is a menu to choose from that defines the identifier for the
resource, e.g. Margherita;
2. a request which begins “I would like to order…” and ends with “for
delivery to 42 Acacia Avenue, Dingley Dell, NeverNeverLand”;
3. the server issues a response to the request addressing it to the client.
The protocol is the “glue” or connector that enables the client-server interaction
to perform as expected.

Single licence - Abingdon School 459

9 Fundamentals of communication and networking

The client-server model is the most commonly employed of the architectural styles for when one application
interacts with another. A server is one application, a client is another. A server application listens for requests, from
client applications, for services which it offers and the client applications consume these services.
A client application, needing a service to be performed, sends a request to the server application via a standard
interface protocol e.g. HTTP.
IP address: 192.168.2.25
USER INTERFACE SERVER
application

100

100 in Fahrenheit is 212

Port
grees
p e r a t u re in de
em
Type a t er
ress Ent
49152

s a n d p
Celsiu

Client application on 192.168.2.12

50

50 in Fahrenheit is 122
Port

s
e r a t u re in degree
mp
Type a te Enter
nd press
53152

C e ls iu s a

Listens on Port 5000

Client application on 192.168.2.15

Figure 9.4.10.2 Client-server operation, server application offers a

temperature conversion service, client applications request this service
The server application either rejects or performs the request and sends a response
Background
back to the client application (possibly invoking a callback function in the client
Callback: application).
A callback function is typically a
Javascript function. A reference to A server application, on the other hand, waits for requests to be made and then
the callback function is passed to reacts to them. A server application is usually a non-terminating process (i.e. on or
the server along with the request. running all the time). It is usually designed to provide a service to more than one
The server invokes this function
client application.
on the client whenever a response
to the request is ready at the The server application shown running in Figure 9.4.10.2 performs a temperature
server. conversion service, e.g. 100°C → 212°F, when it receives a client application
request for this service (server port 5000). Included in the request is the
temperature to be converted, the IP address of the client application, e.g. 192.168.2.15, and the port number on

460 Single licence - Abingdon School

 9.4.10 Client-server model

the client, e.g. 53152, that is connected to the requesting client application. The response of the server application
is to return the temperature in Fahrenheit to the requesting client application, using as return address the client
application’s IP address and port number. The client application connects to the server application for this service.
In this example, the connection uses the Telnet (a less secure alternative to SSH) protocol.
Figure 9.10.4.2 illustrates a very important principle: Separation of Concerns.
Background
The client application is developed quite separately from the server application.
The client application API:
An API is an application
focuses on the user Questions
programming interface. An
interface design and the application programming interface
1 Explain the client-server model.
server application focuses separates the data from the
on the design for the operations that may be performed

processing and formatting of data. Both designs can be changed without regard to on the data. Applications that use
a particular API do not know how
each other whilst preserving a uniform interface between the two, e.g. the Telnet
the data is stored only how to
protocol. access it via API calls.

REST
Web server
The idea of REST is that HTTP
Browser Server application
application data can be queried HTTP Server
GET method
and changed using verbs and POST Server Facebook’s
nouns, represented by HTTP DELETE issues request to
uest graph API
Req
PUT Facebook’s graph
methods and URLs, respectively. A REST API issues
User makes Response
REST request will typically return request to website
data in a machine-readable form, Server sends response
User sees
such as JSON or XML. in format
rendered response
Response returned by API call
Suppose that you wanted to
HTTP Graph
order a book from Amazon, the Client application method Dataset
online bookseller. Do you have
to download a special app to Figure 9.4.10.3 Interaction via an HTTP server between a browser and a
do this or can you use software server application accessing Facebook’s social network graph
that you have on your computer
already? Of course, the answer is that you can use browser software you have on your computer already. Typing the
following URL (uniform resource locator) into the address bar of your browser returns a web page that contains
information about a resource that Amazon uniquely identifies with the ID 0241003008, i.e. the book Very Hungry
Caterpillar Board Book: http://www.amazon.co.uk/Very-Hungry-Caterpillar-Board-Book/dp/0241003008/
Amazon’s web site is a web service based on a design pattern called REST which stands for REpresentational State
Transfer. Central to the concept of REST is the notion of resources. Resources are represented by URIs or Uniform
Resource Identifiers, the Amazon example is one such URI. There are two types of URI: URLs and URNs. We
will focus on URLs (Uniform Resource Locators). Another key aspect of the design of REST web services is that
resources should be linked together and representations of these resources should enable a user to move from one
resource to another by following these links.
Figure 9.4.10.3 show an interaction via an HTTP server between a client application, a browser, and a server
application, that provides a Facebook RESTful web service that accesses Facebook’s social network graph dataset

Single licence - Abingdon School 461

9 Fundamentals of communication and networking

of Facebook users. Facebook’s graph API is an application programming interface (API) that is run on Facebook
servers. If the following URL is typed into the browser’s address bar for the scenario in Figure 9.4.10.3 then what is
returned is a representation that the browser renders as shown in Figure 9.4.10.4:
https://graph.facebook.com/4?oauth_token= CAACE...... See Task 1 below for how to get an access token
(CAACE....) to use with oauth_token.
In the Facebook social network graph, Mark Zuckerberg’s
user node has the identifier “4”, not surprisingly his
identifier was one of the very first to be allocated. The
representation returned is expressed in JSON (JavaScript
Object Notation)
By following the link in the representation shown in
Figure 9.4.10.4, a client application, i.e. a browser,
Figure 9.4.10.4 Mark Zuckerberg’s User node can obtain the next resource representation, Mark
in Facebook’s social network graph, node id is 4. Zuckerberg’s Facebook home page, from the server
Mark’s home page is www.facebook.com/Zuck application. Facebook’s social network graph is, in 2015,
the largest social network dataset in the world.
State 3
State 1 url: https://www.face- Figure 9.4.10.5 shows how the client application’s state
url: empty book.com/Zuck
changes as it interacts with the server application which
in turn accesses Facebook’s social network graph dataset.
Thus, the client application changes (Transfers) to a new
State 2 State when it obtains a new resource REpresentation
url: https://graph.facebook.com/4
from the server application, this is why REST stands for
REpresentational State Transfer!
Figure 9.4.10.5 The client application’s changes of
Questions
state, each transition to a new state causes the server
application to transfer a representation to the client 2 Explain with an example what is meant by REST.
application
The REST architecture is characterised by
• Client-server: separation between server application that offers a service, and the client application that
consumes it
• Stateless: each request from a client application must contain all the information required by the server
application to carry out the request. Session state is kept entirely on the client not at the server
• Uniform interface: the method of communication between a client application and a server application
must be uniform, e.g. via HTTP requests and responses
• Code-on-demand: server applications can provide executable code or scripts for client applications to
execute in their context
• Cacheable: the server application must indicate to the client application requests that can be cached.
• Every resource has a unique ID, e.g. https://graph.facebook.com/4
• Multiple representations of resources provided for different needs, e.g. html, JSON, XML, jpeg, png, gif,
csv
• Resources are linked together, e.g. Facebook’s social network graph.

462 Single licence - Abingdon School

 9.4.10 Client-server model

Uniform interface
Communication between the client application and the server application typically uses HTTP as the uniform
interface. HTTP provides four principle methods for Creating, Retrieving, Updating and Deleting a resource. The
methods are sometimes referred to as CRUD operations:
• C – Create HTTP
Action Examples
• R – Retrieve method
• U – Update GET Retrieve a Retrieve Mark Zuckerberg’s social network graph entry –
• D - Delete resource or GET /4 HTTP/1.1
Host: graph.facebook.com
The HTTP methods that a collection
Retrieve AQA CS Unit 2 main sections
map onto these operations of resources
GET /v1/csunit2s/ HTTP/1.1
are shown in Table Host: cs.apispark.net
9.4.10.1. POST Create Creates a new resource –
The REST API created and a new POST /v1/csunit2s/ HTTP/1.1
Host: cs.apispark.net
run at the server may be resource
{ “id”: “7”, “title”: “Fundamentals of
connected to a relational compter organisation and architecture”,
“link”:[“a test”] }
database such as MySQL.
PUT Update or Updates a resource, “computer” misspelt–
In which case the HTTP
replace an PUT /v1/csunit2s/7 HTTP/1.1
verbs will be converted into Host: cs.apispark.net
existing
SQL equivalent database { “id”: “7”, “title”: “Fundamentals of
resource computer organisation and architecture”,
commands as follows : “link”:[“a test”] }
• GET → SELECT DELETE Delete a Deletes a resource, 8 –
• POST → INSERT resource DELETE /v1/csunit2s/8 HTTP/1.1
Host: cs.apispark.net
• DELETE → DELETE
Table 9.4.10.1 The four HTTP methods for CRUD
• PUT → UPDATE
in order to carry out operations on the database (see Chapter
10.4 for SQL commands). var http = require('http');
var work = require('./mydatabase');
var mysql = require('mysql');
The REST API may also be connected to a NOSQL database var db = mysql.createConnection({
such as MongoDB. host: '127.0.0.1',
user: 'myuser',
Multiple representations of resources password: 'mypassword',
database: 'mydatabase'
What is returned to the client application is not the resource });
but a representation of the resource. The REST service provided
by the server application is usually designed to provide at least Table 9.4.10.2 Server-side Javascript to set up
JSON and XML representations. Figure 9.4.10.6 shows the a MySQL database connection to database,
JSON response and Figure 9.4.10.7 shows the equivalent XML mydatabase. This is part of a larger script that
response. The URL in each case is runs under node.js.

https://cs.apispark.net/v1/csunit2/6
The client application has no knowledge of how the resource is actually stored on the server. The resource sits
behind an interface which hides this information. This means that the resource can be restructured without
affecting any client application which simply continues to use the uniform interface of HTTP CRUD operations.

Single licence - Abingdon School 463

9 Fundamentals of communication and networking

GET /v1/csunit2s/6 HTTP/1.1

Host: cs.apispark.net To access these urls login
Content-Type: application/json
Information
Authorization: Basic
credentials are required. APISpark:
NjY1ZjkyNDUtZTUzNy00MjViLWJhZGI∙∙∙∙∙ (http://restlet.com/products/
Accept: application/json apispark/). It’s an online
Cache-Control: no-cache platform that allows you to
design your APIs following
REST principles and then
{ host them. The data are
"id": "6",
also managed by the
"link": [
"https://cs.apispark.net/v1/csunit2s/subsections/6.1", platform. Users need to
"https://cs.apispark.net/v1/csunit2s/subsections/6.2", have authorisation to use a
"https://cs.apispark.net/v1/csunit2s/subsections/6.3" resource which is obtained
], by just registering to use
"title": "Fundamentals of computer systems" APISpark.
}

Figure 9.4.10.6 JSON representation returned to client application

GET /v1/csunit2s/6 HTTP/1.1

Host: cs.apispark.net
Content-Type: application/xml
Authorization: Basic
NjY1ZjkyNDUtZTUzNy00MjViLWJhZGI∙∙∙∙∙
Accept: application/xml
Cache-Control: no-cache

<Csunit2>
<id>6</id>
<link>
<link>https://cs.apispark.net/v1/csunit2s/subsections/6.1</link>
<link>https://cs.apispark.net/v1/csunit2s/subsections/6.2</link>
<link>https://cs.apispark.net/v1/csunit2s/subsections/6.3</link>
</link>
<title>Fundamentals of computer systems</title>
</Csunit2>

Figure 9.4.10.7 XML representation returned to client application

Every resource has an ID
Resources may be individual resources or collections of resources.
For example, the URL https://cs.apispark.net/v1/csunit2s/subsections/
returns links to all the subsections whereas the URL
• nodes (e.g. a User, a Photo, a Page, a
https://cs.apispark.net/v1/csunit2s/subsection/6.1 Comment)
returns just subsection 6.1. • edges (e.g. connections between Users,
Resource linking Pages, etc)
Facebook’s social network graph consists of • fields (info about Users, such as the
birthday of a User, or the name of a
Page).

464 Single licence - Abingdon School

 9.4.10 Client-server model

So it should be possible to navigate from node to node by following edges, e.g. “likes”. If we start with a user with
Userid 784889091561000 (this is a fictional example), we need our client application, e.g. POSTMAN, to send the
following request to Facebook’s server application to access the open graph(og) part of the social network graph:
https://graph.facebook.com/784889091561000/og.likes?oauth_token= CAACE......
Figure 9.4.10.8 shows a section of the response from the application server. See Task 1 below for how
to get an access token
Questions (CAACE....) to use with
{ oauth_token.
"data": [
3 In the context of REST explain, {
with examples, what is meant by "id": "670741596309143",
"from": {
(a) uniform interface "id": "784889091561000",
(b) multiple representations "name": "Fred Bloggs"
},
(c) every resource has an id
"start_time": "2014-06-17T15:48:54+0000",
(d) resources are linked. "publish_time": "2014-06-17T15:48:54+0000",
"application": {
4 What is meant by CRUD? "name": "Og_likes",
"namespace": "likes",
"id": "193042140809145"
5 A REST API is connected to a },
server-side relational database. "data": {
"object": {
What SQL commands will the "id": "222386191275871",
HTTP verbs, GET, POST, "url": "http://www.needateacher.co.uk/",
DELETE and UPDATE map "type": "website",
"title": "http://www.needateacher.co.uk/"
to? }
},

Figure 9.4.10.8 Section of the response from application server.

The id’s have been altered for privacy reasons
Tasks
1 For this exercise you will need to be on Facebook. Launch Facebook’s social network graph explorer
tool, https://developers.facebook.com/tools/explorer and select the GET method. Enter /me into the
field next to GET. Click on Get Access Token and in the pop-up screen select the items that you wish to
view. Click on submit.
2 Using Facebook’s social network graph explorer tool, enter the following in the GET field:
search?q=secondary school&type=place&center=51.8168, -0.8124&distance=10000
The response should be information about schools in the Aylesbury area –
latitude 51.8168, longitude -0.8124, radius distance 10000 metres.

3 Continuing on from Task 1. Select version v2.2 (Facebook reduced information returned in later
versions). Enter /4 into the address bar next to GET. Click on submit. Copy the link to Mark
Zuckerberg’s home page and visit it in a browser.
4 Google supports several REST web services, one is finding directions. Try the following URL:
https://maps.googleapis.com/maps/api/directions/json?origin=Aylesbury Grammar
School&destination=Tesco Tring Road
5 Experiment with changing the origin and destination parameter values for the URL in Task 4.

Single licence - Abingdon School 465

9 Fundamentals of communication and networking

Tasks

6 Another Google REST service is staticmap. Try the following URL in a browser:
https://maps.googleapis.com/maps/api/staticmap?center=Aylesbury Bucks&zoom=13&size=600x300&
maptype=roadmap&format=png Information
Google Maps Javascript
7 Experiment with the center parameter value for the URL in Task 6 and the
reference:
maptype (roadmap, satellite, terrain, hybrid). https://developers.google.
com/maps/documentation/
8 Install POSTMAN REST client from Google Chrome store. Once installed javascript/reference#Map
run it from outside Chrome web browser. Select the normal page and the
GET method.
Enter http://www.youtypeitwepostit.com/api/ onto the GET line. Click Send. You should get a response
which is a collection of posts to a message board.
9 Use url http://www.youtypeitwepostit.com/api/. Select POST and raw then enter the following below
the raw bar:
{
"template": {
"data": [
{
"prompt":
"Text of message",
"name":
"text",
"value":
"Arfur"
}
]
}
}

Click Send.

10 Using POSTMAN, check that the new message has been posted to the message board using GET.
Click on the hyperlink URL, e.g. http://www.youtypeitwepostit.com/api/25134141254238784, so
that it appears in the GET line. (Id 25134141254238784 must exist for this to work). Change GET
to DELETE and Send. Check that this message has been deleted by doing a GET on http://www.
youtypeitwepostit.com/api/.

11 Explore http://petstore.swagger.io/. Read the instructions very carefully and try

(a) GET (b) POST (c) PUT (d) DELETE

12 Obtain an API key from flickr (Yahoo). Enter the following URL into a browser URL address bar to use
the REST web service of Flickr inserting your API key where indicated:
https://api.flickr.com/services/rest/?method=flickr.photos.search&api_key=this is where you put your
key&format=json&nojsoncallback=1&text=cats&extras=url_o

Find a jpg image in the JSON response and copy its URL into a browser address bar.

13 Create a REST web service with APISpark.

Using Javascript in a browser to connect to a REST API

466 Single licence - Abingdon School

 9.4.10 Client-server model

Table 9.4.10.3 shows Javascript code embedded in an HTML web page, GoogleMap.html. This HTML file uses a
Google map api Javascript library at http://maps.googleapis.com/maps/api/js. The script calls a REST API provided
by Google to pull down a road map. After loading this web page the browser executes the Javascript function
initialize(). This function loads the map data into the web page using a call to an API constructor
google.maps.Map. This creates a new map inside the given HTML container, the Div element, identified by
the identifier “googleMap”. The script section is placed at the end of the web page to ensure that it is executed
after the web page is loaded. The function initialize also places a marker on the map, using the latitude and
longitude coordinates.

Tasks
14 Experiment with the values assigned in mapOptions:
(a) Change zoom from 1 to 20. Note that the map object can be dragged to move location.
(b) Change ROADMAP to SATELLITE.
(c) Change the latitude and longitude to match your location.
(d) Change the marker text that appears when the mouse is over the marker on the map.

<!DOCTYPE html>
<html>
<head>
<script
src="http://maps.googleapis.com/maps/api/js">
</script>
</head>
<body>
<div id="googleMap" style="width:600px;height:400px;"></div>
<script>
function initialize()
{
var LatitudeLongitude = new google.maps.LatLng(51.8168, -0.8124);
var mapOptions = {
zoom: 15,
center: LatitudeLongitude,
mapTypeId:google.maps.MapTypeId.ROADMAP
}
var map = new google.maps.Map(document.getElementById('googleMap'),
mapOptions);
var marker = new google.maps.Marker({
position: LatitudeLongitude,
map: map,
title: 'Marker test!'
});
}
initialize();
</script>
</body>
</html>

Table 9.4.10.3 Web page GoogleMap.html that contains a Javascript script to load a Google roadmap centred on
latitude 51.8168 and longitude -0.8124 at zoom level 15.

Single licence - Abingdon School 467

9 Fundamentals of communication and networking

Table 9.4.10.4 shows a Javascript client side script that accesses a REST web service at To access this url login
https://cs.apispark.net/v1/csunit2s/subsections/subsubsections/5.1.1 credentials are required.
The response is then rendered by adding it to the paragraphs according to their class ids, greeting-id, greeting-title,
greeting-link. The script uses jQuery, $ is short for jQuery, e.g. $.ajax is equivalent to jQuery.ajax.
Figure 9.4.10.9 shows the response rendered in a browser window.
<!DOCTYPE html>
<html>
<head>
<title>Hello jQuery</title>
<script src="https://ajax.googleapis.com/ajax/libs/jquery/1.10.2/jquery.min.js">
</script>
</head> Save this script with extension .html. Launch
<body>
<div> in Google Chrome. If an access token is not
<p class="greeting-id">The ID is </p> supplied then the Javascript console window
<p class="greeting-title">The title is </p>
<p class="greeting-link">The link is </p> will show the error message:
</div> Failed to load resource: the server responded
</body>
with a status of 401 (Unauthorized)
<script>
$.ajax({
headers: {"Authorization": "Basic replace with access token here"},
dataType: "json",
url: "https://cs.apispark.net/v1/csunit2s/subsections/subsubsections/5.1.1"
}).then(function(data) {
$('.greeting-id').append(data.id);
You will need to register with
$('.greeting-title').append(data.title);
$('.greeting-link').append(data.link); cs.apispark.net and set up your
}); own REST service. The urls used
</script>
</html> here are the author’s own setup.

Table 9.4.10.4 HTML web page containing a jQuery Javascript script to access a REST web service

Figure 9.4.10.9 Response to jQuery Javascript rendered in a browser window

Comparing JSON and XML

JSON is more compact than XML and therefore quicker to parse but this comes at the price of security compared
with XML. XML is text/parsing, not code execution, whereas JSON is a subset of Javascript and is parsed by Eval()
which is also used to execute Javascript and therefore could execute malicious code inserted into the JSON unless
precautions are taken to prevent this. XML has namespaces whereas JSON does not - so it cannot support multiple
instances of the same field name in a data-interchange. However, JSON does have advantages over XML which the
JSON.org site summarises as:
• JSON (JavaScript Object Notation) is a lightweight data-interchange format. It is easy for humans to read
and write. It is easy for machines to parse and and generate.
• XML is less easy for humans to read and write and less easy for machines to parse and generate.

468 Single licence - Abingdon School

 9.4.10 Client-server model

Questions

6 Give three advantages that JSON has over XML. 7 Give two advantages that XML has over JSON.

Websocket protocol
The use of HTTP is well suited to the applications described so far. However, in applications which require the
server to send data to the client the very moment when it knows that new data is available, HTTP is inadequate
because it is client-driven.
The HTML5 WebSocket specification defines an API that establishes a full-duplex communication channel
operating through a single socket over the Web between a web browser and a server. HTML5 Websocket has been
designed for real-time, event-driven web applications. The Websocket protocol creates a persistent connection
between the client and the server in which both parties can start sending data at any time.
As soon as a connection to the server is established, the client application can start sending data to the server using
the send('a message') method on the connection object.
Equally the server application can send messages to the client application at any time.
The http:// is replaced by ws:// in a url, e.g. ws://echo.websocket.org,
https:// is replaced by wss://, e.g. wss://echo.websocket.org.
The Websocket protocol is used where multiple users need to interact in real time such as in multiplayer games.

Tasks
15 Visit http://demo.kaazing.com/livefeed/ to see a live demo that uses the WebSocket protocol.

16 Try the echo test at https://www.websocket.org/echo.html

Questions
8 Explain why the Websocket protocol has replaced the HTTP protocol for a certain type of client-server
interaction.

In this chapter you have covered:

■■ The client-server model ■■ REST enables CRUD to be mapped to database
■■ The Websocket protocol functions (SQL) as follows:
■■ The principles of Web CRUD Applications and REST: • GET → SELECT
• CRUD is an acronym for • POST → INSERT
▪▪ C – Create • DELETE → DELETE
▪▪ R – Replace • PUT → UPDATE
▪▪ U – Update ■■ Comparing JSON (Javascript Object Notation)
with XML
▪▪ D – Delete

Single licence - Abingdon School 469

9 Fundamentals of communication and networking

9.4 The Transmission Control Protocol/

Internet Protocol (TCP/IP) protocol
Learning objectives:
■■ 9.4.11 Thin- versus thick-client computing
Thick-client
■■Compare and contrast thin- A thick-client is a computer workstation
client computing with thick-
(Figure 9.4.11.1) with non-volatile local
client computing.
storage, e.g. magnetic hard disk or solid
Key Concept state disk. The magnetic disk or solid state
Thick-client: disk stores an operating system, application
A thick-client is a computer programs and local files.
workstation with non-volatile
Thick-client network
local storage, e.g. magnetic
hard disk or solid state disk.
The thick-client is connected in a client-
The magnetic disk or solid state server network to a file server and a domain
disk stores an operating system, controller. The domain controller validates
application programs and local users when they initiate a login session. The
files.
central file server stores users’ files.
Thick-client network:
The thick-client is connected in Thick-client computing
a client-server network to a file The operating system that runs in the thick-
server and a domain controller. client is loaded from the thick-client's local
The domain controller validates
store, i.e. its magnetic hard disk or solid
users when they initiate a login
session. The central file server state disk. Applications that are used by the
thick-client are loaded from its local store. Figure 9.4.11.1 Thick-client
stores users’ files.
Thick-client computing: It may also load some applications from the workstation with 1 TB
The operating system that runs file server or an application server. However, magnetic disk
in the thick-client is loaded
what distinguishes thick-client computing
from the thick-client's local
store. Applications are loaded from thin-client computing is that applications run in the thick-client
from the thick-client's local workstation whereas for the thin-client workstation applications run, by
store and run in the thick-client default, in a central server.
workstation.

Thin-client
Key Concept In a thin-client network, the client is a diskless workstation with a little
Thin-client: RAM, a low specification CPU, a network interface adapter with a boot
A thin-client is a diskless ROM, interfaces for keyboard, mouse, VDU, and possibly interfaces for other
workstation with a little RAM,
peripherals. The boot ROM on the network card is used to load a stripped-
a low specification CPU, a
network interface adapter with down operating system from a central server as well as to obtain an IP address
a boot ROM, interfaces for and other networking configuration information, e.g. a subnet mask and a
keyboard, mouse, VDU, and gateway IP address.
possibly interfaces for other
peripherals.

Single licence - Abingdon School 470

9 Fundamentals of communication and networking

Key Concept Thin-client networking

The central server is an application server, a
Thin-client networking:
The boot ROM on the network file server and a domain controller all rolled
card is used to load an operating into one. The domain controller validates
system from a central server users when they initiate a login session. The
as well as to obtain an IP
central file server stores users’ files. Users
address and other networking
configuration information, e.g.
sit at a thin-client workstation and log into
a subnet mask and a gateway IP the server. Figure 9.4.11.2 shows a diskless
address. Users sit at a thin-client thin-client workstation.
workstation and log into the
Thin-client computing
server.
By default all applications run at the
Thin-client computing:
server. If the server has sufficient RAM,
The central server is an
application server, a file server
a sufficiently powerful multicore CPU
and a domain controller all and the speed of the network is adequate
rolled into one. The central then a user unacquainted with thin client
file server stores users’ files. architecture will think that the applications
The central application server
they are using are running in their client
stores applications. The central
domain controller stores user workstation. Information entered by the
accounts and validates users user at the thin-client workstation gets
when they initiate a login sent to the central server which processes it Figure 9.4.11.2 Jammin LTSP
session. before returning an updated image to the diskless thin-client workstation
By default all applications run
thin-client's screen. with its own VDU, USB
at the server.
Information entered by the keyboard, mouse, and connected
user operating at the thin-client to a client-server network.
workstation gets sent to the
central server which processes Central
Terminal Terminal
Server
it before returning an updated 1 Edit Edit 2
image to the thin-client's screen. commands commands

Words to word processor Words to

Information display display
PiNet: drawing package
Graphic commands Graphic commands
A thin-client network for
Raspberry Pi thin-clients derived Terminal Graphical Graphical Terminal
4 updates updates 3
from LTSP - see http://pinet.org.
uk/. Figure 9.4.11.3 Thin-clients (terminals) connected to a central server
Terminal server: LTSP
Thin-clients behave in a similar
manner to dumb terminals used Figure 9.4.11.2 shows a diskless workstation called a Jammin that is supported
in a previous era of computing by the Linux Terminal Server Project (LTSP1) - http://www.ltsp.org. Linux is
for providing input and receiving very good at disk caching and code sharing. The Linux server requires about
results from application programs
250 MiB before any clients are added and then 50 MiB for each client.
running in mainframes.
The central server in a thin-client
network serves clients which act
like these dumb terminals, hence
the label terminal server.
1 LTSP is a registered trademark of DisklessWorkstations.com, LLC
471 Single licence - Abingdon School
 9.4.11 Thin- versus thick-client computing

Questions Key point

1 Estimate the amount of RAM in MiB required in a Linux Server Thin- versus thick-client:
supporting 100 thin-clients. Compare this with the amount of RAM Software maintenance:
installed in a thick-client workstation that you have access to (or go • Single point of control in thin-
online to a computer retailer and obtain a figure for a typical thick- client for patches/updates and
installations of new applications
client).
• Updates/patches and
installations of new applications
Comparison of thin- and thick-client computing at each thick-client

In a thin-client network there is a single-point of control for patches/updates Hardware maintenance:

• Life expectancy or mean-time-
and for installation of new applications, the terminal server. In contrast
between-failure of thin-client
workstations in a thick-client network require new applications to be installed workstations longer than thick-
and patches/updates to be applied at each workstation. client workstations by about
two years
Software maintenance
Security:
In the thick-client scenario, there is a software maintenance overhead on • Unauthorised software cannot
each client. Patches/updates will need to be applied from time to time to the be installed, virus and spyware
operating system installed on the hard disk drive/SSD of each client. Similarly, infections are not possible
because of absence of local
patches need to be applied to applications installed on the hard disk drive/SSD
storage in a thin-client
of each client. If the number of thick clients is significant, e.g. 50+, this is a
Cost:
time-consuming activity. • Lower specifications for

In a thin-client network, patches/updates are applied to the software stored on CPU, less RAM, no magnetic
hard drive/SSD for thin-client
the terminal server. This is much less time consuming because the maintenance
compared with thick-client
occurs on just one machine. means lower cost to produce
Hardware maintenance Power consumption:
• Lower specifications and less
Fanless thin-clients have no mechanical moving parts which is the principle
hardware means thin-client
cause of failure in thick-clients with magnetic hard disk drives and fans to cool workstations can consume
the CPU. The life expectancy of thin-clients is about two years longer than about one-seventh the power
thick clients with magnetic storage. that PC thick-client does
Reduced licensing costs:
Security • Only as many licences as will be
One of the primary benefits of using thin-client hardware is security. The used, need to be purchased in a
absence of local storage means that users are unable to install unauthorised thin-client system but as many

software onto the system. There is also no local storage to infect with viruses or licences as thick-clients are
needed in a thick-client system
spyware. Virus protection is applied at the terminal server where it is centrally
Apps best run in a thick-client:
managed. • Applications that result in
Cost considerable latency such as
graphic intensive and video
Thin client hardware has a very low price tag compared with thick-client
editing ones. These require a
workstations. The cost of a thin-client station in a Raspberry Pi LTSP thin- lot of processing power, large
client network is £20 to £30. The Jammin series are about £200 but have 2-4 amount of RAM, large file sizes
GiB memory and a fast Gigabit network adapter. and higher network bandwidth

Single licence - Abingdon School 472

9 Fundamentals of communication and networking

Power consumption
Thin client devices tend to consume much less power than thick-client workstations. Power consumption varies
among makes and models, but some estimates indicate that thin-client devices only consume about one-seventh the
power of PC thick-clients.
Reduced licensing costs
In a thick-client environment application software is installed on each client. If the software requires a licence
then the total number of licences that need to be purchased equals the total number of thick-clients, assuming
that users can work at any thick-client. In a thin-client environment only as many licences as will be used, need
to be purchased as the licensed application software runs in the terminal server whilst being accessible from any
thin-client.
Applications that are best run in a thick-client workstation
Thin-client computing works well for typical applications such as email, web browsing, office applications such
as word processing but it doesn't work quite so well for high-level graphics processing and video editing because
there will be considerable latency because of the higher processing power, larger RAM, larger file sizes and network
bandwidth required for these applications. These applications are best suited to running in a thick-client.

Questions
2 Facing tighter budgets and the need to be more responsive to public requirements, local and city councils
are finding it increasingly difficult to meet targets with their existing PC-centric computer infrastructures of
4500+ machines.
(a) Give four reasons why a solution to this problem could be to replace the thick-client systems (PC-
centric) with a thin-client system.
(b) Suggest one reason why a thin-client system might not be the best option for the following
(i) the architects department (ii) computer science students in local council schools

3 In Business Studies lessons at one school, students need to use stand-alone machines. Before each lesson the
teacher has found it is necessary to reset all twenty machines to their base configuration, spending 15 or 20
minutes each time on this task. Explain why replacing the stand-alone machines by a thin-client network
could improve this situation.

4 Explain why clients in a thin-client network have been called dumb terminals.

In this chapter you have covered:

■■ The meaning of
• thin-client
• thick client
■■ and compared and contrasted thin-client computing with thick-client
computing.

473 Single licence - Abingdon School

10 Fundamentals of databases
10.1 Conceptual data models and
Learning objectives: entity relationship modelling
■ Produce a data model from
given data requirements for a
■ Data modelling
simple scenario involving What is a data model?
multiple entities. Just as a designer of a new type of aircraft or building will build models to help
his or her understanding of the task so a database designer will build models
■ Produce entity relationship
to help his or her understanding of the task of creating a database. The models
diagrams representing a data
that the database designer builds are called data models.
model and entity descriptions
in the form: Formally,
Entity1 (Attribute1, A data model is a method of describing the data, its structure, the way it
Attribute2, .... ). is inter-related and the constraints that apply to it for the given system or
organisation.

Key concept Requirements analysis

Data model: Before building a database system a systems analyst or systems analysts will
A data model is a method establish:
of describing the data, its
• the applications the database must support – the processing and user
structure, the way it is inter-
related and the constraints that requirements;
apply to it for the given system • the data that must be stored in order to satisfy the needs of the variety of
or organisation.
users in an organisation – the data requirements;
• the constraints or rules that apply to that data – the data constraints.
From the analysis, a data model is constructed that will provide a conceptual
model of the real-world system from which a database can be logically designed
Background
and then physically built using a particular Database Management System.
The relational database model is
one example of a logical model. Conceptual Model Logical Model

The fact-based model is another

example of a logical model. Analysis of
real-world
system

Figure 10.1.1 Data modelling stages

Data Requirements
Data requirements are usually expressed in English statements, e.g.
‘Each module of study offered by a college has a module code, a
module title and a credit value.’

Single licence - Abingdon School 474

10 Fundamentals of databases

Data Constraints
Data constraints are also usually expressed in English statements, e.g.
‘Students may not study more than three modules per term.’

Key concept Conceptual data model

The conceptual data model is created during the analysis phase of a project
Conceptual data model: from the data requirements. It summarises these requirements in a formal way
A conceptual model is a
that makes it easier for the analyst to check with the customer that the data
representation of the data
requirements of an organisation requirements of their organisation have been covered completely and accurately.
constructed in a way that is The model should contain no references to physical details of database
independent of any software
construction because that would make it difficult for a customer who is not
used to construct a database.
technically minded to understand.

Formally,
A conceptual model is a representation of the data requirements of an
organisation constructed in a way that is independent of any software that is
used to construct a database.

A conceptual model consists of

• a diagram showing the entities and relationships
• a formal description of each entity in terms of its attributes
• descriptions of the meaning of relationships
• descriptions of any constraints on the system and of any assumptions
made
The terms entity, attribute and relationship are defined and explained below.

Entity
An example of an entity is a Student. Students are of interest to an
organisation such as a school or college. The college will need to record the
name of each student currently enrolled, his/her date of birth, home address,
Key concept
and other data about the student.
Entity:
An entity is an object, person,
Formally,
place, concept, activity, event
An entity is an object, person, place, concept, activity, event or thing of interest
or thing of interest to an
to an organisation and about which data are recorded.
organisation and about which
data are recorded.

475 Single licence - Abingdon School

 Data modelling

Questions
1 List three entities for each of the following organisations:
(a) Hospital
(b) Lending library
(c) Athletics club

Attribute
Key concept
The particular items of data such as name, date of birth, home address, etcetera
belonging to an entity such as Student are called attributes. Attribute:
An attribute is a property or
characteristic of an entity.
Formally,
An attribute is a property or characteristic of an entity.

Questions
2 List two attributes of each of the following entities that belong to a
hospital in-patient system:
(a) Patient
(b) Ward
(c) Nurse

Entity occurrence or instance

A college will record the details of all its students.
The details of a particular student are referred to as an instance or occurrence
of the entity Student.
Figure 10.1.2 shows four examples of instances / occurrences of the entity
Student.
Attributes

Student
Date of
Enrolment Surname Forename Address Tel. No
Entity birth
Number
Occurrence 1 19/1/80 433991
Briggs Sarah 42 Benn Av.
or
2 Carter David 22/2/80 10 Acacia Av. 484132
Instance 3 Teng Lee 13/4/80 23 Queens Road 472611
4 Khan Imran 29/5/80 3 Stannier St 447334

Figure 10.1.2 Instances of the entity Student

Single licence - Abingdon School 476

10 Fundamentals of databases

Entity identifier
Key concept An organisation will need to select or identify a particular occurrence of
an entity from among others. It does this with the entity identifier. This is
Entity identifier: sometimes referred to loosely but incorrectly as the primary key.
The entity identifier is an
attribute or combination of
attributes which uniquely Formally,
identifies an instance or The entity identifier is an attribute or combination of attributes which
occurrence of the entity. uniquely identifies an instance or occurrence of the entity.
Sometimes referred to loosely
but incorrectly as the primary
key. Suppose that each student is assigned a number called the student enrolment
number, such that no two students have the same number. This number is
then unique to each student. Therefore, by making this number an attribute of
the entity Student it can be used as this entity’s identifier. An entity identifier
must have a value, it can never be null, i.e. without a value.
If no single attribute possesses the property of uniqueness, then two or
more attributes must be selected to achieve this goal. Such a combination of
attributes is known as a composite entity identifier.
For example, the entity ClassRoomTimeTable has the following attributes:
Class Room Number, Period Of Day, Day Of Week, Class Code,
Subject Code
Some example occurrences are shown in Figure 10.1.3.
Entity ClassRoomTimeTable

Class Room Period Of Day Of

Class Code Subject Code
Entity Number Day Week
Occurrence 1 1 Monday 12C1 CS
or 2 2 Monday 12C1 CS
Instance 1 1 Tuesday 13D2 Phy
1 2 Tuesday 12C1 CS

Figure 10.1.3 Sample occurrences of the entity ClassRoomTimeTable

Attribute Class Room Number alone is not sufficient to uniquely identify a
single occurrence of this entity, but when it is combined with Period Of Day
and Day Of Week then it becomes possible to identify a single occurrence. The
entity identifier is thus the composite identifier Class Room Number, Period
Of Day and Day Of Week.

477 Single licence - Abingdon School

 Data modelling

A composite identifier must be minimal, i.e. it should contain the minimum

number of attributes possible. For example, if attribute Class Code was added
to the composite identifier it would be longer than necessary. Whereas, if
attribute Day Of Week was removed it would no longer be unique so would
not be an entity identifier.

Questions
3 How many occurrences of an entity must a value of an entity
identifier identify?
A. None B. One or more
C. Zero or more D. Exactly one

4 A CarForSale entity has the following attributes:

Manufacturer Name, Model Name, Registration Number, Engine

Size, Body Colour, Trim Colour, Mileage, Year Of Registration,
MOT Date, Price

Which of the following would be suitable as an entity identifier?

(a) Year Of Registration and Model Name
(b) Manufacturer Name and Model Name
(c) Year Of Registration, MOT Date and Model Name
(d) Registration Number
(e) Registration Number, Year Of Registration

5 A Reservation entity in a hotel room booking system has the following

attributes:
Surname Of Guest, Forename of Guest, Home Address,
Contact Telephone Number, Room Number, Date Room
Booked For, Number Of Nights, Date Reservation Made

Which of the following would not be suitable as an entity

identifier and why?
(a) Surname Of Guest
(b) Surname Of Guest, Forename Of Guest and Home
Address
(c) Room Number and Date Room Booked For
(d) Surname Of Guest, Room Number and Date Room
Booked For

Single licence - Abingdon School 478

10 Fundamentals of databases

Questions
6 Which attributes of the following entities would make suitable entity
identifiers?
a) Patient
b) Ward
c) Nurse

■■ Entity relationship modelling

Entity description
Consider a school scenario consisting of students and staff. An analyst has
Key concept recorded two entities Student and Staff and noted the following in Table
Entity description: 10.1.1.
A formal description of an
Entity Example
entity consisting of its name
and attributes with the entity A person who is a enrolled at the
Student
identifier indicated by being school to study one or more courses
underlined, e.g. A person who is employed by the
GP( GPId, GpName). Staff school, e.g. teacher, technician,
secretary, caretaker
Table 10.1.1
The analyst has also recorded the attributes of each entity as follows
Student: StudentEnrolmentNo, Surname, Forenames, DateOfBirth,
Address, TelNo
Staff: StaffNo, Surname, Forenames, DateOfBirth, Address, TelNo,
Qualifications, Salary, TypeOfStaff, Grade, SubjectsTaught,
ContractHours, Permanent
The entities can now be formally described in the following succinct fashion
called an entity description:
Student (StudentEnrolmentNo, Surname, Forenames, DateOfBirth,
Address, TelNo)

Staff (StaffNo, Surname, Forenames, DateOfBirth, Address,

TelNo, Qualifications, Salary, TypeOfStaff, Grade,
SubjectsTaught, ContractHours, Permanent)
The convention is to underline the entity identifier.

479 Single licence - Abingdon School

 Entity relationship modelling

Formally,
An entity description is a formal description of an entity consisting of its name
and attributes with the entity identifier indicated by being underlined, e.g.
GP( GPId, GPName)

Questions

7 A hospital is organised into a number of wards staffed by nurses. The

attributes of the entities Ward and Nurse are listed below
Ward: WardName, NumberOfBeds
Nurse: StaffNumber, Surname, Forename

Write the entity descriptions for these two entities.

8 An airline owns a number of aeroplanes. Each aeroplane has an

aeroplane identification number and type recorded, along with the
number of seats in that aeroplane. Flight staff employed by the airline
have their staff number, name and specialism recorded.

The attributes of the entities Aeroplane and FlightStaff are listed below

Aeroplane: AeroplaneIdNo, Type, NumberOfSeats

FlightStaff: StaffNumber, Name, Specialism

Write the entity descriptions for these two entities.

Key concept
Relationship:
Relationships
A relationship is a two-way
Within an organisation or within a system in an organisation, entities do not association or link between two
exist in isolation but have links with other entities. entities.

Formally,
A relationship is a two-way association or link between two entities.

For example,
• A patient is nursed by nurses the nursed by relationship associates
a patient with nurses who nurse that
patient
• A nurse nurses patients the nurses relationship associates a
nurse with the patients who the nurse
nurses

Single licence - Abingdon School 480

10 Fundamentals of databases

In the school example, a link exists between the entities Student and Staff.
Students are taught by staff whose role is teaching, and staff whose role is
teaching teach students. In the direction from student to staff, the relationship
has the name Is Taught By. In the direction from staff to students the
relationship has the name Teaches.

Degree of a relationship
Key concept When the data requirements are described more precisely as follows:
Degree of a relationship: A member of staff may teach zero or more students (zero because
The degree of a relationship
not all staff teach, e.g. caretaker) and a student is taught by one
between two entities refers
or more members of staff
to the number of entity
occurrences of one entity which It becomes clear that a member of staff who teaches may teach many students
are associated with just one and a student may be taught by many members of staff.
entity occurrence of the other
and vice versa.
Formally,
The degree of a relationship between two entities refers to the number of entity
occurrences of one entity which are associated with just one entity occurrence of
the other and vice versa.

A relationship therefore has a degree as well as a name.

The degree of a relationship may be one of the following:

• one-to-one
• one-to-many
• many-to-one
• many-to-many

The relationships between entities are best represented diagrammatically in an

entity-relationship or E-R diagram.
Entities in an E-R diagram are represented by rectangles containing the name
of the entity.

Rectangle containing Entity name should be

Student
the name of the entity singular

Figure 10.1.4 Entity symbol on an E-R diagram

A relationship is represented by a line drawn between two associated

entities with a shape resembling a crow’s foot drawn at the many end of the
relationship if this exists.

481 Single licence - Abingdon School

 Entity relationship modelling

The four types of relationship are represented diagrammatically in Figure

10.1.5.
One-to-one
1:1
One-to-many
1:n
Many-to-one
n:1
Many-to-many
n:m

Figure 10.1.5 Diagrammatic representations of relationship degrees

It is good practice to label each relationship that appears on an E-R diagram
with its name (two-way relationship, so two names).
Relationship names should be chosen so sentences that are meaningful can
be constructed describing the relationship using the entity names and the
relationship names.
Figure 10.1.6 shows an E-R diagram for the two entities Student and Staff
drawn with two separate one-to-many relationships showing that Staff Teaches
Student and Student Is Taught By Staff.

Teaches
Student Staff
Is Taught By

Figure 10.1.6 Student-Staff E-R Diagram with separate one-to-many rela-

tionships

Figure 10.1.7 shows an E-R diagram in which the two one-to-many

relationships have been replaced by a single many-to-many one.

Teaches
Student Staff
Is Taught By
Figure 10.1.7 Many to many E-R diagram
However, it is not always necessary to label both ends of a relationship. Often,
common sense dictates that a single label describes a relationship adequately.
This label is placed in the middle of the relationship line connecting the two
entities.

Single licence - Abingdon School 482

10 Fundamentals of databases

Questions
9 For each of the following relationship definitions choose the
correct degree classification:

(a) Each hospital ward has zero or more patients whilst

a patient is assigned to a single ward.

Ward Patient
(i) m : n
(ii) 1 : n
(iii) n : 1
(iv) 1 : 1

(b) Each patient is assigned a single hospital consultant,

whilst each hospital consultant is responsible for one or more
patients.
Patient Consultant
(i) m : n
(ii) 1 : n
(iii) n : 1
(iv) 1 : 1

(c) Each lorry driver is assigned their own lorry and
each lorry is driven by one driver.

Driver Lorry
(i) m : n
(ii) 1 : n
(iii) n : 1
(iv) 1 : 1

(d) A computer may contain zero or more application

programs and each may be installed on zero or more
computers.

Computer Program
(i) m : n
(ii) 1 : n
(iii) n : 1
(iv) 1 : 1

483 Single licence - Abingdon School

 Entity relationship modelling

Resolving many-to-many relationships

The entities in many-to-many relationships would record data as multiple
facts not single facts unless this situation is resolved by creating a link entity
and one-to-many relationships as shown below. For example, a firm that
sells computer parts receives orders for parts
from other businesses. A typical official order Order No: 12345
appearing on a form as shown opposite consists
1. 6 x CPU
of several lines, labelled 1, 2, 3, etc. called order
2. 10 x RAM
lines. Each order line lists a specific part and
quantity ordered. If we record these multiple 3. 5 x Video Card
lines in an Order entity we will make it difficult
to query the entity for specific data, e.g. how many CPUs were ordered on the
order with OrderNo 12345? Or, how many CPUs have been ordered in total
this year? Much better to separate out the lines and store them separately but
linked to the order on which they appear.
Two entities for this ordering system are Order and Part, but a third entity
exists that is not so obvious, OrderLine, corresponding to the separate lines of
the order.
Figure 10.1.8 shows that each order is for many parts and that a part can
appear on many orders, a many-to-many relationship but each order is made
up of many order lines each containing an order for a particular part, a one-to-
many relationship.
Each part may appear in an order line in different orders. If all the order lines
over all the orders are grouped together in an entity called OrderLine ( the link
entity) then we have another one-to-many relationship, this time between Part
and OrderLine.

Order(OrderNo, ...) Order Part

Part(PartNo, Description, ...)

Order Part

OrderLine(OrderNo, PartNo, Order

Quantity) Line
Figure 10.1.8 Resolving Many-to-Many relationships
It may be necessary, to re-draft an Entity-Relationship diagram several times
before completing the modelling satisfactorily. The goal is to resolve each
many-to-many relationship into two one-to-many relationships and a link
entity. The process of drafting raises the analyst’s level of understanding of the
data requirements each time it is attempted.

Single licence - Abingdon School 484

10 Fundamentals of databases

Questions
10 The data requirements for a hospital in-patient system are
defined as follows:

A hospital is organised into a number of wards. Each ward has

a ward number and a name recorded, along with a number of
beds in that ward. Each ward is staffed by nurses. Nurses have
their staff number and name recorded and are assigned to a
single ward.
Each patient in the hospital has a patient identification
number, and their name, address and date of birth are
recorded. Each patient is under the care of a single consultant
and is assigned to a single ward. Each consultant is responsible
for a number of patients. Consultants have their staff number,
name and specialism recorded.

(a) In data modelling, what is

(i) an attribute

(ii) a relationship?

(b) State four entities for the hospital in-patient system and
suggest an identifier for each of these entities.

(c) Draw an entity-relationship diagram that shows three

relationships that can be inferred from these data requirements.

In this chapter you have covered:

■■ The meaning of data model
■■ How to produce a data model from given data requirements for a simple
scenario involving multiple entities.
■■ How to produce entity relationship diagrams representing a data model
and entity descriptions in the form:
Entity1 (Attribute1, Attribute2, .... )
■■ What is meant by entity description
■■ The meaning of entity identifier
■■ Using underlining to identify the attribute(s) which form the entity
identifier
■■ Identifying the degree of a relationship
■■ How to resolve many-to-many relationships

485 Single licence - Abingdon School

10 Fundamentals of databases
10.2 Relational databases
Learning objectives:
■■Explain the concept of a
relational database
■■ Relational database model
Logical database model
■■Be able to define the terms
Over the years several different ways of modelling data logically have been
• attribute tried.

• primary key Modelling data logically emphasises that we are still engaged in a stage of
modelling which is independent of a particular database system, e.g. Microsoft
• composite primary key
Access or MySQL. No details of how the data is to be physically stored and
• foreign key accessed will be considered.
The focus of interest in this chapter is the relational model and relational
database. In this model:

Key concept
A relational database is a set of relations or a collection of tables.
Relational database:
A relational database is a set of
relations or collection of tables.
A later chapter covers an alternative model, the fact-based model.
The conceptual model in the previous chapter concentrated on the structure
and meaning of the data for a specific organisation, without answering the
question: “how should the data be structured and interpreted in a software
system?”
A logical database model concentrates on the structure and meaning of the data
in a particular database approach or system, e.g. the relational database model
approach in which relationships between entities are modelled by shared or
common attributes alone.
The goal of the logical model is to create schemas from which a database can
be physically built. A schema is another name for a plan describing what is to
be built.
Relational Data Model
Mathematical relation
Information We focus first on the approach to logical data modelling known as relational
modelling.
The inclusion of the
reference to the concept This type of modelling is based upon a mathematical concept called a relation.
of a mathematical relation For example, if we have two sets named SetOfStudents and SetOfSubjects,
and its link to relational respectively, and populated as follows
database relations is purely
for background purposes. SetOfStudents = {Sarah, Jim, Kevin}
SetOfSubjects = {CS, Physics, Maths}

Single licence - Abingdon School 486

10 Fundamentals of databases

and a relationship between these two sets, called Studies, then we might model
Background this relationship as a set of ordered pairs as follows
Mathematical relation: Studies = {(Sarah, Physics), (Jim, Physics),( Jim, Maths) }
Relation Studies is a subset of
the Cartesian product The ordered pair, (Sarah, Physics), records that Sarah studies Physics.
The set Studies is a subset drawn from the set of all possible ordered pairs
SetOfStudents x SetOfSubjects
formed from the two sets SetOfStudents and SetOfSubjects.
or an element of the power set
Questions
℘ (SetOfStudents x
1 Write down the set of all possible ordered pairs formed by
SetOfSubjects).
the Cartesian product of SetOfStudents and SetOfSubjects of
See Unit 1 4.2.2. which (Sarah, Physics) is one example.

This way of organising these facts and their relationship uses the concept of a
relation – a relation corresponds to a relationship such as Studies used in the
above example.
Key concept We can use the relation to obtain information. For example, we can conclude
Relation: that Kevin does not study any of the subjects in SetOfSubjects, Sarah and Jim
A relation in the relational study Physics and Jim also studies Maths.
data model can be regarded
loosely as a form of table, with Relational data model relation
attributes being named columns A relation in the relational data model can be regarded loosely as a form of
of the table.
table, with attributes being named columns of the table. More precisely, the
table is a depiction of the relation presenting actual values for the relation in
tabular form.
Key concept
Rows of values of the table are called tuples. They correspond to instances of
Attribute:
records in a programming language.
An attribute is a named column
of a table. Relation ClassRoomTimeTable
Attributes
Relation name

Class
relation Period Of Day Of Class Subject
Room
attributes Day Week Code Code
Number
1 1 Monday 12C1 CS
tuples 2 2 Monday 12C1 CS
1 1 Tuesday 13D2 Phy
1 2 Tuesday 12C1 CS
Figure 10.2.1 Table depiction of relation ClassRoomTimeTable

487 Single licence - Abingdon School

 Relational database model

Primary key Key concept

Relations are written in the following format:
Primary key:
Relation name (Attribute1, Attribute2, Attribute3, etc) The attribute or combination
With the primary key, Attribute1, underlined. of attributes which uniquely
identifies a single occurrence
or tuple of the relation. It is
The attribute, or combination of attributes, which uniquely identifies a single indicated by being underlined.
occurrence (tuple) of the relation is underlined. This attribute, or combination
of attributes, is called the primary key of the relation.

If a combination of attributes is required to ensure uniqueness then all the Key concept
attributes in the combination are underlined. In this instance, the primary key Composite primary key:
is called a composite primary key. Minimal combination of
attributes that uniquely
For example, the relation ClassRoomTimeTable is written as follows:
identifies a single occurrence or
ClassRoomTimeTable(ClassRoomNumber, PeriodOfDay, DayOfWeek, tuple of the relation.
ClassCode, SubjectCode)
With the attribute combination ClassRoomNumber, PeriodOfDay and
DayOfWeek chosen as the primary key.
A value of this primary key, such as “2, 2, Monday” selects just one row or
tuple of the table because the combination is unique in the table. It is also
minimal, i.e. has no more attributes than necessary. If we removed one of
ClassRoomNumber, PeriodOfDay, DayOfWeek we forfeit uniqueness.

ClassRoomNumber, PeriodOfDay, DayOfWeek, ClassCode although unique is

not minimal because it has one more attribute than is necessary to ensure
uniqueness, i.e. ClassCode is unnecessary.

Questions
2 A SwimmingGalaRaceResult relation has the following attributes:

GalaNo, RaceNo, StrokeNo, Distance, SwimmerNoOfWinner,

WinningTime
Which of the following would be suitable as a primary key and why?
(a) SwimmerNoOfWinner
(b) RaceNo
(c) GalaNo, RaceNo, StrokeNo
(d) GalaNo, RaceNo
(e) GalaNo, SwimmerNoOfWinner

Single licence - Abingdon School 488

10 Fundamentals of databases

Questions
3 A CDTrack relation in a CD music collection system has the following
attributes:

CompactDiscId, TrackNo, TrackDuration, SongId

(a) Which of the following would not be suitable as a primary key

and why?

(i) CompactDiscId, SongId

(ii) TrackNo, SongId
(iii) SongId
(iv) CompactDiscId, TrackNo

(b) What is the name given to a primary key that consists of more
than one attribute?

Foreign Key
In a relational database, relationships are modelled by the foreign key
Key concept mechanism.
Foreign key: For example, in a hospital system the relationship between the two entities
A foreign key is an attribute
Ward and Patient has degree one-to-many as shown in Figure 10.2.2. Each
in one relation/table which is
also the primary key of another ward is occupied by zero or more patients.
relation/table. It forms a link
between two relations/tables via
this attribute.
Ward OccupiedBy Patient
Occupies

Figure 10.2.2 E-R diagram for the entities Ward and Patient
The entity definitions from the conceptual modelling stage are
Ward (WardName, NoOfBeds)
Patient (PatientNo, Name, HomeAddress, DateOfBirth, Gender)
The equivalent relations are defined as follows:
Ward (WardName, NoOfBeds)
Patient (PatientNo, Name, HomeAddress, DateOfBirth,
Gender, WardName)
The relationship between the two entities is represented in the relation Patient
by the additional attribute WardName. Foreign keys are usually indicated by
italicising them or placing a line over them.

489 Single licence - Abingdon School

 Relational database model

WardName is the primary key of the relation Ward as well as the entity
identifier of the entity Ward and is known as a foreign key in the relation
Patient.
WardName is therefore common to both Patient and Ward.

A foreign key is an attribute in one relation/table that is also the primary key
of another relation/table. It forms a link between two relations/tables via this
attribute.

Link between mathematical relation and relational model relation

If we have sets SetOfStudentForenames and SetOfStudentIds as follows
SetOfStudentForenames = {Sarah, Jim, Kevin}
SetOfStudentIds = {1, 2, 3}
and a relation Student as follows
Student (StudentId, Forename)
where possible values of StudentId are chosen from the SetOfStudentIds and
possible values of Forename are chosen from SetOfStudentForenames.
Then we have the relationship of StudentId value 1 to Forename value “Jim”
modelled by the relation Student as follows
StudentId 1 is Student with Forename Jim
This appears as one tuple in the table depiction (two-dimensional array) of the
relation as shown in Table 10.2.1.

StudentId Forename
1 Jim
2 Kevin
3 Sarah

Table 10.2.1 Relation Student as a table of values

Similarly, if we have sets SetOfSubjectIds and SetOfSubjectNames as follows

SetOfSubjectNames = {CS, Physics, Maths}
SetOfSubjectIds = {1, 2, 3}
and a relation Subject as follows
Subject (SubjectId, SubjectName)
then we can say for example that
SubjectId 1 is a Subject with SubjectName “CS”

as shown in Table 10.2.2.

Single licence - Abingdon School 490

10 Fundamentals of databases

SubjectId SubjectName
1 CS
2 Physics
3 Maths
Table 10.2.2 Relation Subject as a table of values
Figure 10.2.3 shows the entity-relationship (E-R) diagram from the conceptual
modelling stage, the stage before the relational database modelling stage. The
diagram conveys that a student studies zero or more subjects and a subject is
studied by zero or more students (actually the E-R diagram needs an additional
symbol on each crow foot to indicate “zero or more”).

Studies
Student Subject
StudiedBy

Figure 10.2.3 E-R diagram from conceptual modelling stage

In the relational model we use only relations.
Mathematically these are sets of ordered pairs.
Therefore, we need another relation Studies. This is shown below for the
relational database model and as a mathematical relation.
Relational database model
Studies(StudentId, SubjectId)
Primary key is composite, consisting of the combination
StudentId, SubjectId
Mathematical relation model
Studies = {(1,2),(1,3),(2,1),(2,3),(3,2)}
Table 10.2.3 shows a depiction of relation Studies with values corresponding to
the ordered pairs.
StudentId SubjectId
1 2
1 3
2 1
2 3
3 2

Table 10.2.3 Relation Studies as a table of values

491 Single licence - Abingdon School

 Relational database model

Figure 10.2.4 shows the E-R diagram after the relational modelling stage. The
entity Studies has now been included (on an E-R diagram we use the term
entity).
A student is associated with zero of more subjects and a subject is associated
with zero or more students. The relation Studies provides the link that we need
between the relations Student and Subject. That is why we call the entity on
the E-R diagram corresponding to this relation, a link entity.

Studies Key concept

Student Subject
StudiedBy Modelling many-to-many
relationships:
Many-to-many relationships
Studies in the conceptual model must
be modelled in the relational
model by relations that support
Figure 10.2.4 E-R diagram after relational modelling stage one-to-many relationships
The relational database model thus consists of the three relations otherwise problems arise, e.g.
querying can be problematic.
Subject (SubjectId, SubjectName)
Student (StudentId, Forename)
Studies(StudentId, SubjectId)
Many-to-many relationships in the conceptual model must be modelled in the
relational model by relations that support one-to-many relationships otherwise
problems arise. If the one-to-many relationships don’t exist then they must be
created. The previous example illustrates how this can be achieved.

Questions
4 The entities Ward and Nurse are defined below

Ward(WardName, NoOfBeds)

Nurse(StaffNo, Name)

Using the entity-relationship diagram shown below, define the relations

which correspond to these two entities, i.e. add a foreign key to one of
the entities to model the relationship.

Ward StaffedBy Nurse

AssignedTo

Single licence - Abingdon School 492

10 Fundamentals of databases

Questions
5 Each patient is registered with one GP whereas a GP has
many patients. The entity definitions for these data
requirements are

GP(GPId, GPName)
Patient(PatientNo, Name, HomeAddress, DateOfBirth,
Gender, NHSNo)

and E-R diagram is

Treats
GP Patient
RegisteredWith

Define relations for these data requirements.

6 A competition is made up of many events. Each event involves many

teams and a team participates in many events. The entity-relationship
diagram for this competition is shown below

ParticipatesIn
Event Team
Involves

The entities Event and Team are described as follows

Event(EventId, EventDescription, Date, Time)

Team(TeamId, TeamName, ContactTelNo)

(a) Modify this entity-relationship diagram so that its

entities may be modelled in a relational database.

(b) Using the new E-R diagram from (a) write down the
relations.

493 Single licence - Abingdon School

 Relational database model

Questions
7 The entity-relationship diagram for a hospital system is shown below.

OccupiedBy
Ward Patient

StaffedBy Nurses
Nurse Treats

Consultant

The entities and their attributes for this E-R diagram are:

Ward(WardName, NoOfBeds)

Patient(PatientNo, Name, Address, DateOfBirth, Gender)

Nurse (StaffNo, Name, Rank)

Consultant (StaffNo, Name, Specialism)

A patient is nursed by one or more nurses and a nurse nurses one or

more patients but not every patient on a ward.

(a) Modify the entity-relationship (E-R) diagram so that its

entities may be modelled in a relational database.

(b) Using the new E-R diagram from (a) write down the
relations.

Foreign key already present

Sometimes the foreign key attribute is already present in the entity at the many
end of a one-to-many relationship. For example, the entity definitions for the
entities Customer, AuctionItem are
Customer (CustomerId, Name, Address)
AuctionItem (ItemId, AuctionPrice, PaidYN, CustomerId)
A data tuple is never created for AuctionItem without the CustomerId
of the customer who has bought the item at auction. If no items are sold
AuctionItem will be empty.

Foreign Keys in One-To-One Relationships

As an example of a one-to-one relationship consider the E-R diagram of Figure
10.2.5, in which the Drives relationship represents the fact that each taxi driver
drives one car at a time and that each car is driven by only one taxi driver at a

Single licence - Abingdon School 494

10 Fundamentals of databases

time. It might make sense to model with separate entities if drivers are assigned
and re-assigned to different cars, frequently.

Driver Drives Car

Figure 10.2.5 One-to-one relationship

The entities and their attributes are:
Driver(DriverId, DriverName, Gender, ….)
Car(CarId, CarModel, ………)
To convert these entities into relations, first write down the relations ignoring
the relationship Drives.
Then represent the relationship Drives by adding foreign keys. However, unlike
previous examples we now have a choice. We can add the foreign key CarId to
Driver or we can add the foreign key DriverId to Car or we can do both.
If we do both then we get relations as follows
Driver(DriverId, DriverName, Gender, CarId, …..)
Car(CarId, CarModel, DriverId, …..)
Questions
8 Driver and Car could have been combined into one entity.

(a) Why might doing this not always be sensible?

(b) In what circumstances would using just one foreign key
be sensible?

Foreign Keys in Recursive Relationships

Some nurses are in charge of other nurses. Figure 10.2.6 shows how this is
represented in an entity-relationship diagram. The relationship in this case is
said to be recursive.
Manages
Nurse

ManagedBy

Figure 10.2.6 Recursive relationship

The entity Nurse has the attributes:
Nurse(StaffNo, Name, Rank)
The relation Nurse has the attributes:
Nurse(StaffNo, Name, Rank, ManagerStaffNo)
The attribute ManagerStaffNo is a foreign key whose values are taken from the
range of values of the attribute StaffNo.
Some example occurrences of the relation Nurse are listed in Table 10.2.4

495 Single licence - Abingdon School

 Relational database model

Table Nurse
Manager
StaffNo Name Rank
StaffNo
1 Sarah Briggs Staff Nurse 3
2 John Doe Staff Nurse 3
3 Sue Cripps Ward Sister 4
4 Mary Downs Senior Nurse Null
Table 10.2.4 Example occurrences of the relation Nurse
Note that it is possible to have no value at all in a foreign key column, i.e. a
null value, in any relationship, recursive or not. If a value is present then there
must be a matching value present in the primary key column, e.g. 3.

Questions
Manages
Teacher
9 Some teachers manage other teachers.
The entity description for the ManagedBy
Teacher entity is as follows
Teacher (TeacherId, Name, SubjectId)

Write the relation Teacher to model the relationship.

10 A blood donor service uses a relational database to keep track of donors

and their donations of blood. Each donor, on registering to give blood
for the first time, is assigned a unique donor identification number
and has their surname, forename, address, telephone number, date
of birth and blood type recorded. Each time a donor gives blood the
quantity of blood given and the date of giving are recorded together
with the identification number attached to the vessel in which the blood
is stored. Each vessel is assigned a unique identification number and
is used once only. Two entities for this system are BloodDonor and
BloodGiven.
(a) List the attributes of the following entities underlining the
entity identifier attribute(s) in each case

(i) BloodDonor
(ii) BloodGiven

(b) Draw an entity-relationship diagram showing the degree of

the relationship for the two entities BloodDonor and
BloodGiven.
(c)
(i) What is a relational database?
(ii) What is a foreign key in the context of a relational database?
(iii) State how the relationship between BloodDonor and
BloodGiven is represented in a relational database.
(iv) State the relations and their attributes for this blood donor
service scenario. Underline the primary keys.

Single licence - Abingdon School 496

10 Fundamentals of databases

In this chapter you have covered:

■■ Modelling data logically
■■ Relational modelling and the relational database model
■■ Modelling relationships by foreign key mechanism
■■ The link between mathematical relation and relational model relation
■■ Use of link relation to model many-to-many relationship

497 Single licence - Abingdon School

10 Fundamentals of databases
10.3 Database design and
Learning objectives: normalisation techniques
• Normalise relations to third
normal form ■■ Normalisation techniques
• Understand why databases are Repeating groups
normalised Table 10.3.1 shows a fictitious sample of general practitioners (GPs) and their
patients’ names and identification numbers.
In the British National Health Service,
• Each patient is registered at any one time with just one GP
• A GP has zero or more registered patients.
Expressing this table as a relation
GP (GPId, GPName, Patient)
GPId values are drawn from a simple domain or set of values, the natural
numbers. GPName values are also drawn from a simple domain of alphabetic
strings. However, patient values are not drawn from a simple domain of values
but one with structure of array of PatientId, PatientName. If relations are to
be usable, e.g. queried, then their attributes must all be drawn from simple
domains of values. No column or combination of columns must contain
multiple values per row.
Repeating group PatientId PatientName

GPId GPName Patient

718 Bloggs
Key concept 345 Khan
1 Smith
234 Teng
Repeating group: 456 Nunn
Domain for attribute(s) is a
structured type so an instance
1118 Archer
is a group of values, known as
1305 Ali
a repeating group because its 2 Smith
2214 Singh
group-like structure repeats
6541 Nunn
from row to row.

8600 Bloggs
4341 Sorensen
3011 Minns
1678 Ng
2999 Zog

Table 10.3.1 GPs and their patients

The GP relation is said to be unnormalised because of the multiple values in
the Patient column in each row. This state is caused by the Patient’s domain of

Single licence - Abingdon School 498

10 Fundamentals of databases

values being a group of values, known as a repeating group - so named because

its group-like structure repeats from row to row.
Problems occur if a relation is unnormalised
One of the problems with the unnormalised GP relation is that there is no
space available in row 1 of its table for a new patient that registers with Dr
Smith. Row 1 would have to be deleted, the new patient details added to the
repeating group and the new record appended to the end
of the table.

Repeating group
Domain for attribute(s) is a structured type so an instance is a group of values,
known as a repeating group because its group-like structure repeats from row to
row.

To indicate that a relation contains a repeating group a line is placed above the
attribute(s) that contain groups of values
GP (GPId, GPName, Patient)
The solution to the repeating group problem in this example is to remove
the Patient values to their own relation called Patient as shown below, and to
add the primary key of the GP relation to this new relation to model the link
between GP and Patient relations as follows
GP (GPId, GPName)
Patient (PatientId, PatientName, GPId)
There is now no repeating group in either relation – see Table 10.3.2 – and
what’s more
Every non-key attribute is a fact about the key,
the whole key and nothing but the key.
where key means primary key.
Because this is the case we then can say that relations GP and Patient are fully
normalised.
Key concept
Single-Valued Fact (SVF):
When a relation is fully normalised, instances of a single-valued fact (SVF) will be recorded just once, in just one place.

Single-valued facts assert that one thing is associated with just one other thing.

Some examples of a single-valued fact are:

1. A teacher has a name. Given a value of TeacherId, it is a fact that there is just one value of
TeacherName associated with this value.
2. A student has a name. Given a value of StudentId, it is a fact that there is just one value of
StudentName associated with this value.

499 Single licence - Abingdon School

 Normalisation techniques

The instances of single-valued facts for the fully normalised GP and Patient
relations are recorded just once, in just one place:
• GPId determines GPName because, given a specific value of GPId, say 2,
there is just one and only one value of GPName, Smith, associated with
this value of GPId.
• PatientId determines PatientName because, given a specific value of
PatientId, say 1305, there is just one and only one value of PatientName,
Ali, associated with it.

Information
Determinant
Another way of expressing a single-valued fact is to draw a determinancy diagram.
PatientId → PatientName

The diagram asserts that attribute PatientId determines attribute PatientName and therefore PatientId is a
determinant. There is a single value of PatientName for each value of PatientId, i.e. given a value of PatientId
such as 345 the name Khan is returned from a search of the Patient table – Table 10.3.2. However, it is not
true that there is only one value of PatientId associated with each value of PatientName. Two patients can have
the same name but must have different values of PatientId, e.g. given the name Bloggs a search of the Patient
table returns the values of PatientId 718 and 8600.
We say that PatientId determines PatientName or PatientName depends upon or is functionally dependent upon
PatientId or is a fact about PatientId.
Determinants are useful because from a list of determinants drawn up from the data requirements, it is
possible to immediately write down a set of fully normalised relations.

Values in cells must be atomic

Table 10.3.2 shows the Patient relation in table form with each cell occupied
by a single atomic value. Atomic means that the value is as simple as it can be,
i.e. it cannot be broken down any further without losing meaning. The patient
values that have come from Table 10.3.1 have been separately recorded in Key principle
attributes PatientId and PatientName to achieve atomicity. Atomicity:
Not all rows in Table 10.3.2 show actual data. This fact is indicated by ellipses, Values in cells must be atomic.

•••, in cells. Ellipses can also be shorthand for multiple rows.

PatientId PatientName GPId

718 Bloggs 1
345 Khan 1
234 Teng 1
456 Nunn 1
1118 Archer 2

Single licence - Abingdon School 500

10 Fundamentals of databases

PatientId PatientName GPId

1305 Ali 2
2214 Singh 2
6541 Nunn 2
••• ••• •••
••• ••• •••
8600 Bloggs 3011
4341 Sorensen 3011
1678 Ng 3011
2999 Zog 3011

Table 10.3.2 Patient relation table form

Questions
1 Write down the determinants from the data in Table 10.3.1.
Using these determinants can you see how to write down the
normalised relations?

2 The relation Ward is unnormalised

Ward (WardName, WardType, Patient)

Table 10.3.3 shows the Ward relation in table form.

WardName WardType Patient

1456 Smith
Nightingale Orthopaedic
1497 Smart
1461 Berry
Barnard Cardiac 1468 Singh
1478 Alton
1472 Harley
Seacole Medical
1421 Sven
1483 Noggs
Guttman Geriatric 1305 Ali
1678 Ng

Table 10.3.3 Ward relation in table form

(a) What makes the Ward relation unnormalised?

(b) Derive the fully normalised relations corresponding to

the Ward relation.

501 Single licence - Abingdon School

 Normalisation techniques

Normalisation where relationship is many-to-many

Figure 10.3.1 is an E-R diagram for a school scenario in which
• a teacher teaches one subject to one or more students
• a student is taught by one or more teachers

Teacher Student

Figure 10.3.1 E-R diagram for Teacher, Student entities

Table 10.3.4 shows that teacher Mead, for example, teaches students Bond,
Afridi, Smith, Ng and Ali English.

TeacherId TeacherName Student

15898 Bond English
24298 Afridi English
1234 Mead 32145 Smith English
11023 Ng English
18769 Ali English
24298 Afridi Physics
5678 Davies 32145 Smith Physics
11023 Ng Physics
15898 Bond Maths
9123 Younis 32145 Smith Maths
11023 Ng Maths
15898 Bond Maths
4532 Ferris 45910 Singh Maths
19462 Gurung Maths

Table 10.3.4 shows the TeacherTeaches table for a sample of teachers and the
students that they teach
The table contains a repeating group and so its relation is unnormalised.
We indicate the repeating group with an overline as follows
TeacherTeaches (TeacherId, TeacherName, Student)
where the Student repeating group is an array of StudentId, StudentName,
SubjectName.
A first step to resolving the issue is to eliminate the repeating group, Student,
in the TeacherTeaches relation. The simplest way of doing this is to make a
new row in TeacherTeaches for each row of the repeating group and to replace
the Student column by three columns StudentId, StudentName, SubjectName,
respectively. This step is called putting the relation in First Normal Form or
1NF.

Single licence - Abingdon School 502

10 Fundamentals of databases

We have now achieved atomicity.

The TeacherTeaches relation is now
TeacherTeaches (TeacherId, TeacherName, StudentId, StudentName,
SubjectName)

Task
1 Draw up the table for this new form of the relation using the data in
Table 10.3.4.

Clearly it is not true for this new form of the relation TeacherTeaches that
Every non-key attribute is a fact about the key, the whole key and
nothing but the key.
For example, TeacherName is a fact about TeacherId which is only part of the
key.

Questions
3 What is
(a) StudentName a fact about?
(b) SubjectName a fact about?

If we remove these facts from TeacherTeaches to new relations we get

Teacher (TeacherId, TeacherName, SubjectName)
Student (StudentId, StudentName)
leaving
TeacherTeaches (TeacherId, StudentId)
These three relations are fully normalised.
Renaming TeacherTeaches, Teaches we have
Teaches (TeacherId, StudentId)
We can safely delete all tuples in relation Teaches that reference a particular
teacher who leaves the school. At the same, the references in Teaches to
students taught by this teacher are removed but we don’t in the process delete
the corresponding student tuples in the relation Student. The integrity of the
database is preserved.
If we acted instead on the 1NF relation (at top of page) then we would lose
some student names and potentially all references to a student still at the
school. This would be an inconsistency in the database.

503 Single licence - Abingdon School

 Normalisation techniques

E-R modelling approach to normalisation

Can we arrive at a fully normalised set of relations by any other route? The
answer is yes. Often conceptual modelling, i.e. entity-relationship modelling,
leads to relations in the relational model that are already fully normalised.
Figure 10.3.2 arrived at by entity-relationship modelling, has produced a set
of fully normalised entities from which fully normalised relations follow which
correspond to the fully normalised relations in the previous section. The entity
Teaches already has the necessary foreign keys, TeacherId and StudentId.

Teacher Student

Teaches

Figure 10.3.2 E-R diagram for Teacher, Student and Teachers entities

Questions
4
SalesmanId CustomerId Salesman
1
1 Archer
2
3
2 Dent
4
5
3 Rogers
6

(a) Explain why this table is unnormalised.

(b) Normalise the relation for this table into a set of fully normalised
relations.
5 PatientId GPId GPName
1 1 Biggs
2 1 Biggs
3 2 Smith
4 2 Smith
5 3 Timms
6 3 Timms
The table contains redundant data (unnecessary duplication). The
values of GPName, highlighted in red, may be deleted without loss of
information. We say that the table therefore contains redundant data.
For example, PatientId value 2 in row 2 is associated with GPId value 1.
Therefore, GPName of GPId 1 can be looked up in row 1.
Normalise the data in the table into a set of fully normalised
relations.

Single licence - Abingdon School 504

10 Fundamentals of databases

Questions
6 A student is taught by one or more tutors and a tutor teaches one or
more students as shown in the table below.

TutorNo StudentNo TutorName StudentName

1 1 Ainsley Khan
1 2 Ainsley Carter
1 3 Ainsley Chandai
2 4 Svensen Smith
2 2 Svensen Carter
2 1 Svensen Khan

(a) Does this table contain any redundant data?

(b) Is this table fully normalised? Justify your answer.

7
PatientNo PatientName WardName WardType
1456 Smith Nightingale Orthopaedic
1461 Berry Barnard Cardiac
1468 Thomas Barnard Cardiac
1472 Harley Guttman Orthopaedic
1478 Smith Barnard Cardiac
1483 Noggs Spens Geriatric
1497 Smith Nightingale Orthopaedic

(a) In what ways does this table contain redundant data?

(b) Split the relation corresponding to this table into two separate
relations, ensuring that each table for the new relations contains
no redundant data.
(c) Select and indicate a primary key for each new relation.
(d) Are each of the new relations fully normalised? Justify your answer.

Why normalise databases?

It should be clear now why a set of fully normalised relations are desirable. For
fully normalised relations
• All possible relationships between the data are allowed for
• Unnecessary duplication of data is avoided and storage space saved
• Altering data is not unnecessarily time-consuming
• Altering data does not lead to inconsistencies

505 Single licence - Abingdon School

 Normalisation techniques

Information
A fully normalised set of relations contains no redundant data.
Redundant data:
Data is redundant when it
can be deleted without loss of
There may be duplication of data in a fully normalised set of relations but it is
information.
non-redundant and necessary duplication.
Normalising relations to third normal form
Why don’t we just use E-R modelling to arrive at a set of fully normalised
relations?
Key principle
If we have an alternative then we have a means of checking the completeness,
A fully normalised set of
accuracy and consistency of the E-R model.
relations contains no redundant
In this section, we describe a formal technique for arriving at a set of fully data.
normalised relations which is an alternative to E-R modelling.
This formal normalisation technique consists of stages known as Normal Form
(NF).
Starting with the unnormalised relations
Key principle

• Remove repeating groups to place relation in First Normal Form (1NF) Normalising relations to Third
Normal Form:
• Remove partial dependencies to place relation in Second Normal Form
(2NF) 1. Remove
repeating groups
• Remove transitive dependencies to place relation in Third Normal Form
→ First Normal Form
(3NF) (1NF)
A relation which is in 3NF is automatically in 2NF. A relation which is in 2NF 2. Remove
partial dependencies
is automatically in 1NF.
→ Second Normal Form
(2NF)
Information 3. Remove
There are several higher forms of normalisation than 3NF. These forms are transitive dependencies
used in highly specialised cases. In normal cases, a relation that is in 3NF is → Third Normal Form
also in these higher forms. (3NF)

One in particular is worth mentioning because it has its own name,

Boyce-Codd Normal Form (BCNF).

We have already used the informal test for BCNF in earlier sections
Every non-key attribute is a fact about the key, the whole
key and nothing but the key.
It contains within it the tests for 2NF and 3NF

Single licence - Abingdon School 506

10 Fundamentals of databases

Consider the following simple example.

Un-normalised Relation StudentClass

StudentClass (ClassId, ClassDescription,StudentId, StudentName)

ClassId ClassDescription StudentId StudentName

1 Bloggs
A1 Art 1 2 Garcia
3 Begum
4 Mertems
A2 Art 2 5 Becker
6 Zhu
1 Bloggs
B1 Biology 1 3 Begum
5 Becker
2 Garcia
B2 Biology 2 4 Mertems
6 Zhu
Table 10.3.5 shows the StudentClass table in unnormalised form because of
a repeating group
1NF
First Normal Form – remove repeating group to produce a relation in which
all the data values in its table are atomic values.
The easiest way to do this is create a table for each row in the repeating group as
shown in Table 10.3.6.
Note that 1NF is formally defined as
A relation is in 1NF if and only if every non-primary key
attribute is functionally dependent on the primary key – see page
497.

Information
Functional dependency:
Given a relation R, attribute R.B, e.g. PatientName, is functionally
dependent on attribute R.A, e.g. PatientId, if and only if each value of R.A
is associated with precisely one value of R.B, at any one time.

Full Functional dependency:

Given a relation R, attribute R.B, e.g. ProductId , is fully functionally
dependent on attribute R.A, e.g. OrderNo, OrderLineNo if it is functionally
dependent on R.A and not functionally dependent on any subset of R.A
(where A and B may be composite, i.e. consist of more than one attribute).

507 Single licence - Abingdon School

 Normalisation techniques

ClassId ClassDescription StudentId StudentName

A1 Art 1 1 Bloggs Key point
A1 Art 1 2 Garcia
A relation which is in 3NF
A1 Art 1 3 Begum is automatically in 2NF. A
A2 Art 2 4 Mertems relation which is in 2NF is
A2 Art 2 5 Becker automatically in 1NF.
A2 Art 2 6 Zhu
B1 Biology 1 1 In normal cases, a relation that
Bloggs
is in 3NF is also in the higher
B1 Biology 1 3 Begum
forms such as BCNF.
B1 Biology 1 5 Becker
B2 Biology 2 2 Garcia
B2 Biology 2 4 Mertems
B2 Biology 2 6 Zhu

Table 10.3.6 shows the StudentClass table in 1NF Background

StudentClass (ClassId, StudentId, ClassDescription, StudentName) Fourth stage:
Boyce-Codd Normal Form
2NF
(BCNF) reached by removing
Second Normal Form – Remove any non-primary key attributes that are not a remaining anomalies arising
fact about the whole of the primary key to separate relations to get here. from functional dependencies.

ClassDescription is a fact about ClassId and not StudentId. StudentName is a

fact about StudentId and not ClassId.
StudentClass (ClassId, StudentId) Background
Class (ClassId, ClassDescription)
BCNF informally:
Student(StudentId, StudentName) Every non-key attribute is a fact
about the key, the whole key
3NF and nothing but the key.
Third Normal Form - The next stage is to remove any non-primary key
attributes that are both a fact about the key and other non-key attributes to
separate relations to get here.
There aren’t any so the 2NF relations are already in 3NF. Note that this is not
always the case.
StudentClass (ClassID, StudentID)
Class (ClassID, ClassDescription)
Student(StudentId, StudentName)
Check:
Every non-key attribute is a fact about the key, the whole key
and nothing but the key.

Single licence - Abingdon School 508

10 Fundamentals of databases

Questions
8 The relation GP is unnormalised GPId GPName Patient
718 Bloggs
GP (GPId, GPName, Patient) 345 Khan
1 Smith
234 Teng
456 Nunn
The table opposite shows some data for this relation.
1118 Archer
the patient data is a combination of PatientId and 1305 Ali
2 Brown
PatientName. 2214 Singh
6541 Nunn

(a) Place this relation in 1NF (hint: result is a single

relation). Justify your answer.
(b) Place this 1NF relation in 2NF (hint: result is two relations). Justify your answer.
(c) Place the 2NF relations in 3NF. Justify your answer.

Normalising relations to third normal form for a given scenario

We start by considering a simple scenario for which the data requirements are
as follows:
A school offers a number of different courses. Students enrol to study one
or more courses up to a maximum of three. A particular course is taught
by one teacher only but a teacher may teach more than one course.
When students enrol they are each allocated a unique student identifier
and have their name, gender and courses they enrol for recorded. Each
course has a course title and a unique course code. Teachers are assigned
a unique teacher identifier and have their name recorded.
Instead of identifying the entities and their relationships, we will immediately
write down a single relation called StudentCourse consisting of every single
attribute identified in the data requirements. To give flesh to this task we
also construct the table equivalent of this relation and populate it with some
example data. The primary key is StudentId.
The StudentCourse relation is unnormalised because it contains a repeating
group consisting of CourseCode, CourseTitle, TeacherId, TeacherName.

StudentCourse (StudentId, StudentName, Gender,

CourseCode,CourseTitle,TeacherId, TeacherName)

509 Single licence - Abingdon School

 Normalisation techniques

Table StudentCourse
StudentId StudentName Gender CourseCode CourseTitle TeacherId TeacherName
AQA0643 A Level CS 1234 Mead
15898 Bond M UCL0675 A Level Maths 5678 Davies
EDE0187 A Level Art 9123 Milsom
UCL0675 A Level Maths 5678 Davies
24298 Smith F AQA0643 A Level CS 1234 Mead
AQA0432 A Level ICT 1234 Mead
EDE0187 A Level Art 9123 Milsom
10598 Robert M UOC0987 A Level French 4567 Crapper
AQA0432 A Level ICT 1234 Mead
13497 Nixon F UOC0987 A Level French 4567 Crapper
Table 10.3.7 Table StudentCourse unnormalised
Repeating group shown in red.
Stages of normalisation
1NF: Remove repeating groups to place relation in First Normal Form.
The First Normal Form of the table, Table 10.3.8 is shown below.
StudentId StudentName Gender CourseCode CourseTitle TeacherId TeacherName
15898 Bond M AQA0643 A Level CS 1234 Mead
15898 Bond M UCL0675 A Level Maths 5678 Davies
15898 Bond M EDE0187 A Level Art 9123 Milsom
24298 Smith F UCL0675 A Level Maths 5678 Davies
24298 Smith F AQA0643 A Level CS 1234 Mead
24298 Smith F AQA0432 A Level ICT 1234 Mead
10598 Robert M EDE0187 A Level Art 9123 Milsom
10598 Robert M UOC0987 A Level French 4567 Crapper
10598 Robert M AQA0432 A Level ICT 1234 Mead
13497 Nixon F UOC0987 A Level French 4567 Crapper
Table 10.3.8 Table StudentCourse in First Normal Form
StudentId on its own no longer satisfies the criterion of uniqueness and so a
new primary key must be found. StudentId together with CourseCode becomes
the new primary key.
The relation in first normal form is
StudentCourse(StudentId, CourseCode, StudentName,
Gender, CourseTitle,TeacherId, TeacherName)
2NF:
Second Normal Form - Every non-primary key attribute is functionally
dependent on the whole of the primary key.
The next stage is to remove any non-primary key attributes that depend only
upon part of the primary key to separate relations to achieve Second Normal
Form.

Single licence - Abingdon School 510

10 Fundamentals of databases

This step will transform a relation that is in 1NF to two or more

relations that are in 2NF form. Note that a relation that is in 1NF
may already be in 2NF if the condition for 2NF is also satisfied.
Attributes StudentName, Gender depend upon (are facts about) StudentId not
StudentId, CourseCode, the new primary key. Attributes CourseTitle, TeacherId
and TeacherName depend upon (are facts about) CourseCode not StudentId,
CourseCode.
Therefore, we create two new relations called Course and Student leaving a
reduced StudentCourse relation as shown below:
Course(CourseCode, CourseTitle, TeacherId, TeacherName)
Student(StudentId, StudentName, Gender)
StudentCourse(StudentId, CourseCode)
3NF:
Third Normal Form - The next stage is to remove any non-primary key
attributes that are not solely directly dependent on the key to separate
relations.
This step will transform a relation that is in 2NF to two or more relations that
are in 3NF. Note that a relation that is in 2NF may already be in 3NF if the
condition for 3NF is also satisfied.
TeacherName is dependent on the primary key CourseCode of relation Course
but it is also dependent upon TeacherId. So relation Course is not in 3NF.
Relation Student and relation StudentCourse are already in 3NF.

Task
2 Discuss why TeacherName is dependent on the primary key
CourseCode of relation Course. Look at the data requirements.

Therefore, we create a new relation called Teacher. TeacherName is removed

from relation Course together with a copy of attribute TeacherId as shown
below. A copy of TeacherId must be left behind in relation Course otherwise
it will not be possible to determine the teacher assigned to teach a particular
course.
Course(CourseCode, CourseTitle, TeacherId)
Teacher(TeacherId, TeacherName)
StudentCourse(StudentId, CourseCode)
Student(StudentId, StudentName, Gender)

511 Single licence - Abingdon School

 Normalisation techniques

We have arrived at a set of relations in which there is no unnecessary

duplication, i.e. no redundancy. Updating the database is straightforward
and avoids potentially inconsistent results. For example, if a teacher marries
and changes their surname, the change is made in one place only. In the
unnormalised table, if the teacher name to be changed is Mead, this requires
changing in three places. If, let’s say, the changes are carried out in only two of
the three places accidentally, then the database will inconsistently reflect the
new status of teacher Mead. The changes will also take longer to make in the
case of the unnormalised database.

Questions
9 Table 10.3.9 shows some data for a TeacherTeaches relation

TeacherTeaches (TeacherId, TeacherName, Student)

The data requirements are summarised as follows

• a teacher teaches one or more students
• a student is taught by one or more teachers
• a teacher may teach more than one subject
The student data is composed of StudentId, StudentName, SubjectName.

TeacherId TeacherName Student

15898 Bond English
24298 Afridi English
1234 Mead 32145 Smith English
15898 Bond French
24298 Afridi French
24298 Afridi Physics
5678 Davies 32145 Smith Physics
11023 Ng Physics
15898 Bond Maths
9123 Younis 32145 Smith Maths
32145 Smith RS
15898 Bond Maths
4532 Ferris 45910 Singh Maths
45910 Singh GS
Table 10.3.9 TeacherTeaches table
(a) Place this relation in 1NF.
(b) Place this relation in 2NF.
(c) Place this relation in 3NF.

Single licence - Abingdon School 512

10 Fundamentals of databases

Questions
10 Swimming galas take place at different venues during the course of
the swimming season. Each gala consists of several races. Each gala
race consists of several swimmers and is for one particular swimming
stroke, e.g. breast stroke and one particular distance, e.g. 50m.
Races are numbered starting from 1 for each gala. All galas code
the strokes in the same way: strokes are numbered from 1 and each
has a stroke name, e.g. 1 is always breast stroke. Each venue has a
unique name, e.g. Leeds. Each swimmer is assigned a country-wide
unique swimmer number so that no two swimmers will use the
same swimmer number at any gala. The swimmer’s time for a race is
recorded.

The unnormalised relation corresponding to Table 10.3.10

showing swimming gala data (note for space reasons, distance is not
included) is
Gala (SwimmerNo, Name, GalaNo, Venue, RaceNo
StrokeNo, StrokeName, Time)
Normalise this relation to produce a set of relations in 3NF.

SwimmerNo Name GalaNo Venue RaceNo StrokeNo StrokeName Time

1 Leeds 1 1 Breast 1.30
3 3 Crawl 0.30
2 Derby 1 3 Crawl 0.32
1 Bond 6 1 Breast 1.33
3 Oxford 1 5 Medley 2.14
3 4 Back 0.59
10 Hove 10 2 Fly 1.15
10 Hove 1 2 Fly 1.16
13 4 Back 0.58
21 Bristol 1 4 Back 0.57
123 Teng
26 2 Fly 1.17
45 Swindon 1 1 Breast 1.34
31 5 Medley 2.16

Table 10.3.10 shows a sample of swimming gala data

513 Single licence - Abingdon School

 Normalisation techniques

Questions

11 An agency arranges bookings of live bands for a number of venues.

The data requirements are defined as follows.

• Each band is registered with the agency and is assigned a unique

BandId
• Each band is managed by a manager
• A manager may manage several bands
• Each manager is registered with the agency and is assigned a
unique ManagerId by the agency
• Each venue is registered with the agency and is assigned a unique
VenueId
• The agency records the following details
űű Manager name
űű Band name
űű Venue name
űű Date of a booking, band booked and venue for booking
The constraints are

A band will never have more than one booking on any particular date.

Agency (ManagerId, ManagerName,

BandId,BandName,BookingDate,VenueId,VenueName)

Normalise this relation to produce a set of relations in 3NF.

Table 10.3.11 shows a sample of agency data.
ManagerId ManagerName BandId BandName BookingDate VenueId VenueName
1 Nice 3/10/2015 1 Arch
10/10/2015 2 Ten
17/10/2015 3 Macs
24/10/2015 4 Friars
1 Bloggs
2 Loud 3/10/2015 5 Oak
10/10/2015 6 Elm
17/10/2015 7 Caesars
24/10/2015 8 Locarno
3 Riff 3/10/2015 9 Mill
10/10/2015 10 Floss
17/10/2015 11 Riverside
24/10/2015 12 Cult
2 Ramases
4 Moss 3/10/2015 13 Goth
10/10/2015 14 Cairns
17/10/2015 15 Boot
24/10/2015 16 Mod
Table 10.3.11 shows a sample of the agency’s data

Single licence - Abingdon School 514

10 Fundamentals of databases

Problems occur if a relation is not fully normalised

Now we have a little more experience working with relations/tables that are
not fully normalised we can revisit the issues that arise with such relations and
describe these using the examples that we have considered. The issues are:
1. Redundant data:
Clearly, the table StudentClass, Table 10.3.6, contains unnecessary
repetition of data, e.g. ClassDescription data such as Art 1 corresponds to
ClassId A1 but the same information is repeated unnecessarily on rows 2
and 3. We don’t need Art 1 to be repeated on rows 2 and 3. We say that
the table contains redundant data.
2. Data anomaly on deletion:
Clearly the table StudentCourse, Table 10.3.8, contains unnecessary
repetition of data. Let’s suppose, student 13497 Nixon leaves the school.
We delete Nixon’s tuple from table StudentCourse. We lose value 4567
Crapper but it still exists in the table in tuple with StudentId 10598.
But let’s suppose that this student now leaves the school. We delete
this student’s tuple but in the process we also lose the information that
teacher with TeacherId 4567 has name Crapper and that they can teach
course UOC0987.
3. Data inconsistency on update:
Also, if Smith in Table 10.3.8 were to change her surname, it will need
to be changed in three places which again could lead to an inconsistency
if this is done incompletely. Also, the changes in each case will consume
more time than would be necessary if the data to be changed were
recorded just once.
4. Data anomaly on insert:
Suppose a new student, Bunter, joins the school and is assigned
StudentId value 7. If this student is inserted into Table 10.3.6 before
being allocated to any classes then the primary key, ClassId, StudentId
will be null, 7 respectively. This is not allowed because no part of a
primary key may be null, i.e. without a value.

515 Single licence - Abingdon School

 Normalisation techniques

Extension Material
Boyce-Codd Normal Form
We have already stated that Boyce-Codd Normal Form is the next stage after 3NF.

Informally, when a relation is in BCNF, every attribute which is not part of the primary key, is a fact
about the key, the whole key and nothing but the key (“so help me Codd”).

Formally,
BCNF a relation is in Boyce-Codd Normal Form (BCNF) if and only if every determinant is a
candidate key.

We have already seen that determinants are useful. From a list of determinants drawn up from the data
requirements, it is possible to immediately write down a set of fully normalised relations in BCNF.

We have also seen that

The BCNF test will immediately reveal if there are any redundant data in a relation or table.

BCNF &
3NF and BCNF are equivalent for a relation with only one candidate key.
3NF

Candidate A candidate key is an attribute or combination of attributes that has the

key property of uniqueness and minimality. Such a key distinguishes one row of
a table (one tuple of a relation) from another.

Minimality The minimum number of attributes which guarantee uniqueness are used.

A relation may have more than one candidate key. One is chosen as the primary key and the others are
alternate keys.
For example, if each student is allocated a locker to store their things and pays a refundable deposit, the
Student relation becomes
Student(StudentId, StudentName, LockerNo, Deposit)
Both StudentId and LockerNo each determine StudentName
and both StudentId and LockerNo are unique and candidate keys for the Student relation but only one can be
chosen as primary key.

Single licence - Abingdon School 516

10 Fundamentals of databases

Extension Questions
12 The table below shows some sample data for relation Patient.
Patient is not fully normalised
PatientId GPId GPName PatientName
1 1 Biggs Singh
2 1 Biggs Brown
3 2 Smith Mian
4 2 Smith Fadhil
5 3 Biggs Fadhil
6 3 Biggs Brown
(a) State which of the following are candidate keys for relation Patient.
(i) PatientId
(ii) GPId
(iii) GPName
(iv) PatientName

(b) Draw a determinancy diagram – see page 500 - for the

Patient table.

(c) Normalise the relation Patient for the data in the table into
a set of relations in BCNF.

(d) Why is the set of BCNF relations identical to the 3NF set?

13 For the scenario in Q10

(a) Draw a determinancy diagram.

(b) List the determinants in this scenario.
(c) List the candidate keys.
(d) Produce a set of normalised relations in BCNF directly from
the list of determinants.

517 Single licence - Abingdon School

 Normalisation techniques

In this chapter you have covered:

■■ Normalising database relations to third normal form (3NF)
• remove repeating groups → First Normal Form (1NF)
• remove partial dependencies → Second Normal Form (2NF)
• remove transitive dependencies → Third Normal Form (3NF)
■■ A database is normalised so that
• all possible relationships between the data are allowed for
• unnecessary duplication of data is avoided and storage space saved
• altering data is not unnecessarily time-consuming
• altering data does not lead to inconsistencies.

Single licence - Abingdon School 518

10 Fundamentals of databases
10.4 Structured Query Language (SQL)
Learning objectives:
■■Be able to use SQL to
■■ Using SQL to retrieve, update, insert and delete
• retrieve data from multiple data
tables of a relational
Querying a database
database
The main purpose of storing data in a database is to enable applications to
• update data in tables of a interrogate the database for information. This interrogation is called querying
relational database the database.
• insert data into tables of a
Structured Query Language (SQL)
relational database
Structured Query Language (SQL) can be used to query a database. It is a
• delete data in tables of a simplified programming language.
relational database
Although SQL is an ANSI (American National Standards Institute) and ISO/
■■Be able to use SQL to IEC JTC1 (Joint Technical Committee 1) standard, there are different versions
define a database table. of the SQL language. However, to be compliant with the ANSI/ISO standard,
they all support at least the major commands (such as SELECT, UPDATE,
DELETE, INSERT, WHERE) in a similar manner.

Retrieving data from a single table

Table 10.4.1 shows data for the Student relation
Student (StudentId, StudentName, Gender)
The following query, expressed in SQL, will retrieve all of the data in the
Student relation
SELECT *
FROM Student;
The wildcard character * matches the attribute list
StudentId, StudentName, Gender

Student
StudentId Gender
Name
1 Ames M
2 Baloch F
3 Cheng F
4 Dodds M
5 Groos M
6 Smith F
Table 10.4.1 Relation Student in table form

Single licence - Abingdon School 519

10 Fundamentals of databases

The ANSI/ISO SQL standard requires that a semicolon is used at the end of the SQL statement but some systems
relax this requirement. When writing SQL the convention is to use upper case for the SQL commands.
If we wanted just the data for StudentName we would refine the query as follows
SELECT StudentName
FROM Student;
We could refine the search even further by adding a WHERE clause that applies a search condition as follows
SELECT StudentName
FROM Student
WHERE Gender = 'F';
The result set that would be returned when this query is applied to relation Student would be as follows
Baloch
Cheng
Smith
because only these rows of the table/tuples of the relation match the search condition Gender = 'F'.
Gender = 'F' is actually called a predicate because it evaluates to either TRUE or FALSE.
If we also wanted the values of StudentId returned then the query would be
SELECT StudentId, StudentName Questions
FROM Student 1 Write an SQL query that returns the names of all students in
WHERE Gender = 'F'; Table 10.4.1 who are male.

Retrieving data from multiple tables

WardName NurseInCharge NoOfBeds
Table 10.4.2 shows data in table form for the Ward relation
Victoria Sister Bunn 30
Ward (WardName, NurseInCharge, NoOfBeds) Aylesbury Sister Moon 40

Table 10.4.3 shows data in table form for the Patient relation Table 10.4.2 Relation Ward in table form
Patient (PatientId, Surname, WardName) PatientId Surname WardName
The two relations are linked via a shared or common attribute 1 Bond Aylesbury
WardName. The existence of an attribute common to both relations 2 Smith Victoria
is not enough to join data from the corresponding tables correctly, as 3 Jones Aylesbury
the following SQL query demonstrates 4 Biggs Victoria
SELECT Ward.WardName, Ward.NurseInCharge, Table 10.4.3 Relation Patient in table form
Patient.PatientId 1
Victoria Sister Bunn
FROM Ward, Patient; Victoria Sister Bunn 2
The part of the query Ward.WardName references the WardName Victoria Sister Bunn 3
attribute in relation Ward and the part Patient.PatientId Victoria Sister Bunn 4
references PatientId attribute in relation Patient. Aylesbury Sister Moon 1
Aylesbury Sister Moon 2
The FROM Ward,Patient part joins both relations without regard
Aylesbury Sister Moon 3
for the way that the data is actually linked via matching values of the
Aylesbury Sister Moon 4
shared attribute, WardName. The result set returned by the query is
shown in Table 10.4.4. Table 10.4.4 Result set ignoring
relationship between Ward and Patient
520 Single licence - Abingdon School
 Using SQL to retrieve, update, insert and delete data

The problem is caused by the fact that our query does not make use of the common attribute that models the real
world relationship between ward and patient, WardName.
When the search condition
WHERE Ward.WardName = Patient.WardName

is added to the SQL query we are able to exclude values that are not linked by the attribute WardName and to
include only those that are. This SQL query will return the result set that corresponds to the real world situation
shown in Table 10.4.5.
Aylesbury Sister Moon 1
SELECT Ward.WardName, Ward.NurseInCharge, Patient.PatientId
Victoria Sister Bunn 2
FROM Ward, Patient Aylesbury Sister Moon 3
WHERE Ward.WardName = Patient.WardName; Victoria Sister Bunn 4

The two relations have been joined on their common attribute, WardName, i.e. Table 10.4.5 Result set taking
where the value of WardName is the same in both relations. account of relationship between
Ward and Patient
Writing the query as follows would return the same result set because dropping
the relation name prefix before NurseInCharge and PatientId in the SELECT part of the SQL query is allowed
where there is no ambiguity as to what is intended.
SELECT Ward.WardName , NurseInCharge, PatientId
FROM Ward, Patient Background
WHERE Ward.WardName = Patient.WardName;
Integrity:
Means maintaining the accuracy
Questions and consistency of the data and
relationships between data.
2 Write the SQL query that returns from Tables 10.4.2 and 10.4.3 the
name of the nurse in charge of the ward, surnames of all patients in
Referential integrity:
their ward and the ward name. Each foreign key value must
have a matching primary key
Role of a foreign key value in a related table, or be
Null.
What purpose does a foreign key serve if it doesn’t automatically join the
corresponding two relations? Null:

The answer is so that the relational database system can perform a referential Special value that is used to
indicate the absence of any data
integrity check, i.e. check that the referenced primary key value exists before it
value.
is used as a foreign key in another relation. For example, if an attempt is made
to add a new patient to the Patient table, either 'Victoria' or 'Aylesbury' must
be chosen for WardName because these are the only two values present in the linked-to table Ward. The database
system would use a referential integrity check to trap any other entered value for WardName and report an error.
Similarly, if an attempt was made to delete the row in the Ward table for ‘Victoria’ ward, an error would occur
because this value is used in the Patient table. In both of the cases described, preventing these changes occurring
maintains the integrity of the data in the database of those attributes which make references to other attributes.

Single licence - Abingdon School 521

10 Fundamentals of databases

Ordering the result set returned by a query

We can order a result set returned by a query in SELECT Ward.WardName, NurseInCharge, PatientId
ascending or descending order with the keyword FROM Ward, Patient
ORDER BY qualified by one of the keywords ASC WHERE Ward.WardName = Patient.WardName
or DESC. If the qualifier is omitted then ASC is ORDER BY Ward.WardName ASC;
assumed. For example, we can place the result set
returned in ascending order on WardName by the Aylesbury Sister Moon 1
query opposite. Aylesbury Sister Moon 3
Victoria Sister Bunn 2
Table 10.4.6 shows the outcome of applying this query to the Ward and
Victoria Sister Bunn 4
Patient relations.
Table 10.4.6 Result set ordered
Questions on WardName in ascending
alphabetic order
3 Write the SQL query that returns the names of both nurses and their
patients, from Tables 10.4.2 and 10.4.3, ordered in descending patient
name order.

Using more than one search condition

We may refine the query to return only those matches for which the number of beds is more than 30. We can do
this by adding the search condition Ward.NoOfBeds > 30
and connecting it logically to the search condition Ward.WardName = Patient.WardName
with the logical connective AND as follows
SELECT Ward.WardName, NurseInCharge, Patient.PatientId
FROM Ward, Patient Aylesbury Sister Moon 1
WHERE Ward.WardName = Patient.WardName Aylesbury Sister Moon 3
AND Ward.NoOfBeds > 30
ORDER BY WardName ASC;
The result set returned when this query is applied is as shown above right. Ward Aylesbury has 40 beds. Ward
Victoria does not appear because this ward has only 30 beds and so is ruled out by the search condition
Comparison Ward.NoOfBeds > 30.
Description
Operator
= Equal to
Questions
< Less than 4 Write the SQL query that returns the ward names
> Greater than and patient surnames, from Tables 10.4.2 and
<= Less than or equal to 10.4.3, for which the patient identifier is greater
>= Greater than or equal to than 1. Order the result set in ascending order of
<> Not equal to
patient name.
The IN operator is used to compare
a value to a list of literal values that
Relational or comparison operators for search
have been specified or returned
IN condition
from another SQL query, e.g.
Table 10.4.7 shows comparison operators that may be
IN (SELECT PatientId FROM
used in SQL queries.
Patient).
Table 10.4.7 Comparison operators for SQL queries
522 Single licence - Abingdon School
 Using SQL to retrieve, update, insert and delete data

Questions
5 Write the SQL query that returns the ward names and patient surnames , from Tables 10.4.2 and 10.4.3,
for which the patient identifier is less than or equal to 3. Order the result set in descending order of patient
identifier.

Customer
Ordering on more than one attribute CustNo Location
Name
Table 10.4.8 shows two database tables, Customer and Order,
Kauai Dive Kapaa Kauai
respectively. 1221
Shoppe
The effect of applying the following SQL query to this database 1351 Sight Diver Kato Paphos
SELECT Customer.CustomerName , Order.OrderNo
Tom Sawyer Christiansted
FROM Customer, Order 1356
Diving Centre
WHERE Customer.CustNo = Order.CustNo
Blue Jack Waipahu
ORDER BY Customer.CustomerName, Order.OrderNo; 1380
Aqua Center
is to return the following result Blue Jack Aqua Center 1001 Davy Jones’ Vancouver
2156
set with columns CustomerName, Blue Jack Aqua Center 1009 Locker
OrderNo. Sight Diver 1005
OrderNo CustNo SaleDate
Note that, the result set is Sight Diver 1007
1007 1351 20140412
ordered on the first column, Tom Sawyer Diving Centre 1003
1009 1380 20140417
CustomerName and then the Tom Sawyer Diving Centre 1004
1004 1356 20140420
second OrderNo. 1005 1351 20141106
In the Order table, the order of the values of OrderNo for the 1003 1356 20141113
1007 1001 1380 20141202
1009 orders placed by these customers is shown on the left. The data
1004 is extracted in this order. It is next sorted on the first column, Table 10.4.8 Tables Customer and
1005 CustomerName to produce the order of values shown on the Order
1003 right abbreviated. The rows with the same CustNo values are
Blue… 1009
1001 then sorted on OrderNo to produce the result set that is returned by the SQL
Blue… 1001
query.
Sight… 1007
Logical operators for search condition Sight… 1005
Table 10.4.9 shows some of the logical operators that can be used in SQL queries. Tom… 1004
Tom… 1003
Logical Operator Description
The AND operator connects multiple search conditions in an SQL statement’s WHERE
AND
clause that must all be true for a successful match.
The OR operator allows multiple conditions to be combined in an SQL statement’s
OR
WHERE clause any of which can be true for a successful match.
The NOT operator reverses the meaning of the logical operation with which it is used e.g.
NOT
NOT (age > 14 AND age < 19).
Table 10.4.9 Logical operators for SQL queries
When more than one logical operator is used in a statement, NOT is evaluated first, then AND, and finally OR.

Single licence - Abingdon School 523

10 Fundamentals of databases

To impose a different order of evaluation parentheses can be added to the query. The logical conditions inside
parentheses are evaluated independently before logical operators outside of the parentheses are applied.
Table 10.4.10 and Table 10.4.11 show data for relations Borrower and BooksOnLoan.

BorrowerId Surname Initial

1 Smith K ISBN CopyNo BorrowerId DateDueBack
2 Barnes W 9781907982514 1 2 10/9/2014
3 Minns M 9781907982514 2 1 4/9/2014

Table 10.4.10 Table showing some Table 10.4.11 Table showing some values for the relation
values for the relation Borrower BooksOnLoan
The result set returned when the following SQL query SELECT Borrower.Surname, Borrower.Initial

is applied to these tables is Barnes, W FROM Borrower

Smith, K WHERE BorrowerId IN BooksOnLoan;

The search condition “WHERE BorrowerId IN BooksOnLoan”

matches BorrowerId 2 and 1 because these BorrowerIds are present in the BooksOnLoan table.
The result set returned when the following SQL query SELECT Borrower.Surname, Borrower.Initial
is applied to these tables is Minns, M FROM Borrower
WHERE BorrowerId NOT IN BooksOnLoan;

The search condition "WHERE BorrowerId NOT IN BooksOnLoan" matches BorrowerId 3 because this
BorrowerId is not present in the BooksOnLoan table.
The result set returned when the following SQL SELECT Borrower.Surname, Borrower.Initial
query FROM Borrower
WHERE BorrowerId IN (SELECT BorrowerId
is applied to these tables is shown below
FROM BooksOnLoan
Smith, K
WHERE BooksOnLoan.DateDueBack < '5/9/2014');
The search condition
WHERE BorrowerId IN (SELECT BorrowerId
FROM BooksOnLoan
WHERE BooksOnLoan.DateDueBack < '5/9/2014')

matches BorrowerId 1 because this BorrowerId has a DateDueBack value of ‘4/9/2014’ in the BooksOnLoan table
and this value is less than ‘5/9/2014’.
The result set returned when the following SQL SELECT Borrower.Surname, Borrower.Initials
query is applied to these tables is shown below FROM Borrower
WHERE BorrowerId =
Barnes W (SELECT BorrowerId
FROM BooksOnLoan
WHERE BooksOnLoan.ISBN = 9781907982514
AND BooksOnLoan.CopyNo = 1);

524 Single licence - Abingdon School

 Using SQL to retrieve, update, insert and delete data

Relation Country is Name Capital Population Area

Country (Name, Capital, Population, Area) Argentina Buenos Aires 32 300 003 2777815
Bolivia La Paz 7 300 000 1098575
Table 10.4.12 shows some data in table form for
Brazil Brasilia 150 400 000 8511196
relation Country.
Canada Ottawa 26 500 000 9976147
The result set returned when the following SQL Chile Santiago 13 200 000 756943
query Colombia Bagota 33 000 000 1138907
SELECT Name, Capital, Population Cuba Havana 10 600 000 114524
FROM Country Ecuador Quito 10 600 000 455502
WHERE (Population < 7000000) El Salvador San Salvador 5 300 000 20865
OR (Population > 30000000); Guyana Georgetown 800 000 214969
is applied to this Country relation with attributes Table 10.4.12 Table for relation Country showing some
Name, Capital, Population, Area is shown below values
Argentina Buenos Aires 32300003 Questions
Brazil Brasilia 150400000
Colombia Bagota 33000000 6 SELECT Capital, Population, Area
El Salvador San Salvador 5300000 FROM Country
Guyana Georgetown 800000 WHERE (Area < 900000
AND Population < 11000000)
Questions AND NOT Name = 'Cuba';

7 SELECT Capital, Population What result set is returned when this SQL query is
FROM Country applied to the data in Table 10.4.12?
WHERE Name IN ('Chile',
'Cuba', 'Guyana');
Questions
What result set is returned when this
SQL query is applied to the data in Table 8 Rewrite the query in Q7 so that it uses
10.4.12? OR instead of IN.

Deleting data in a single table

The DELETE statement is used to delete rows of a table.
Questions
DELETE FROM table_name
WHERE some_column = some_value; 9 Write the SQL statement to delete the row
with BorrowerId 3 in the Borrower table
The WHERE clause specifies which row or rows should be
shown in Table 10.4.10.
deleted. If the WHERE clause is omitted, all rows will be
deleted!
10 Write the SQL statement to delete the row(s)
For example referencing Table 4.10.12, with Population > 15000000 in the Country
DELETE FROM Country table shown in Table 10.4.12.
WHERE Capital = 'Brasilia';

deletes the row Brazil, Brasilia, 150400000, 8511196.

Single licence - Abingdon School 525

10 Fundamentals of databases

Inserting data in a single table

The INSERT INTO statement inserts a new row into a table. It is possible to write this statement in two forms.
The first form does not specify the column names where the data will be inserted, only their values:
INSERT INTO table_name
VALUES (value1, value2, value3, ...);

The second form specifies both the column names and the values to be inserted:
INSERT INTO table_name (column1, column2, column3, ...)
VALUES (value1, value2, value3, ...);

In the first form, a value of the correct data type must be supplied for every attribute of the relation and the order of
the supplied values must be the same as the corresponding columns.
In the second form, a value for every specified column must be supplied and each value must match in data type the
corresponding specified column, i.e. value1 corresponds to column1, value2 to column2, etc. The value Null will be
inserted for any columns not referenced.
WardName NurseInCharge NoOfBeds
For example, for relation Ward, Table 10.4.2, reproduced here Victoria Sister Bunn 30

First form: Aylesbury Sister Moon 40

INSERT INTO Ward VALUES ('Gresham', 'Mr Oonga', 20);

Table 10.4.2 Relation Ward in table form

Second form:
INSERT INTO Ward (WardName, NurseInCharge) VALUES ('Savernake', 'Sister Teng');

This second form creates a new row in Table 10.4.2 with values 'Savernake', 'Sister Teng', Null

Questions

11 Write the SQL statement to add a new row to the Ward table (Table 10.4.2) for ward 'Amersham',
containing 25 beds. The nurse in charge is 'Sister Brody'.

12 Write the SQL statement to add a new row to the Country table (Table 10.4.12) for 'UK', 'London'.

Updating data in a single table

The UPDATE statement is used to update an existing row of a table.
UPDATE table_name
SET column1 = value1, column2 = value2, ... Questions
WHERE some_column=some_value;
For example, 13 Write the SQL statement to update the
row of the Country table (Table 10.4.12)
UPDATE Ward
for 'UK' to add population 64100000, area
SET NurseInCharge = 'Mr Ali',
243610. Assume that an insert statement has
NoOfBeds = 25
inserted 'UK', 'London' already as in Q12.
WHERE WardName = 'Victoria';

526 Single licence - Abingdon School

 Using SQL to create a database

Data Manipulation Language (DML)

SQL qualifies as a data manipulation language because it can be used to retrieve, insert, delete and update the
data content of databases. It is possible to practise using and writing SQL to manipulate data in a database online at
http://www.w3schools.com/sql/default.asp.
The tutorials on this site use the Northwind sample database (included in MS Access and MS SQL Server).
This site also contains a handy SQL quick reference that can be consulted at
http://www.w3schools.com/sql/sql_quickref.asp. Online SQL tests are also available at this site.

■■ Using SQL to create a database

Data Definition Language (DDL)
Data Definition Language is a standard for commands that define the different structures in a database. DDL
statements create, modify and remove database objects such as tables, indexes, and users.

CREATE TABLE command

The CREATE TABLE command is used to create a table in a database.
Tables are organized into rows and columns. Each table must have a name. The structure and name of a table is
created with a statement of the form
CREATE TABLE table_name The column_name parameters specify the names
( of the columns of the table.
column_name1 data_type(size), The data_type parameter specifies what type
column_name2 data_type(size), of data the column can hold (e.g. VARCHAR,
column_name3 data_type(size), INTEGER, DECIMAL, DATE, etc...).
....
);
The size parameter sets a limit on the length of values of the data type, e.g. VARCHAR(10) sets a limit of 10
characters. The size parameter may be omitted.
For example,
CREATE TABLE Person
(
The PersonId column is of type SMALLINT and will
PersonId SMALLINT,
hold an integer containing up to 5 decimal digits
Surname VARCHAR(25),
(SQL : 1999 standard).
FirstName VARCHAR(25), The Surname, FirstName, Address, and City columns
Address VARCHAR(25), are of type VARCHAR and will hold characters up to
City VARCHAR(25) a maximum length of 25 characters.
);
This DDL statement when executed will create an empty Person table with structure as follows
PersonId Surname FirstName Address City

The empty table can be filled with data using the INSERT INTO statement.

Single licence - Abingdon School 527

10 Fundamentals of databases

SQL Constraints
SQL constraints are used to specify rules for the data in a table and are applied either at the level of column or at
the level of the table.
Column level:
CREATE TABLE table_name
(
column_name1 data_type(size) constraint_name,
column_name2 data_type(size) constraint_name,
column_name3 data_type(size) constraint_name,
....
);

In SQL, we have the following constraints:

• NOT NULL - Indicates that a column cannot store NULL value
• UNIQUE - Ensures that each row for a column must have a unique value
• PRIMARY KEY - A combination of a NOT NULL and UNIQUE. Ensures that a column (or combination
of two or more columns) has a unique identity which helps to find a particular record in a table more easily
and quickly
• FOREIGN KEY - Ensure the referential integrity of the data in one table to match values in another table
• CHECK - Ensures that the value in a column meets a specific condition
• DEFAULT - Specifies a default value when none specified for this column
To see an application of these we will use the Ward, Patient scenario from Table 10.4.2 and Table 10.4.3 suitably
modified and with a Gender attribute to illustrate use of the CHECK constraint.
Table level: Column level

CREATE TABLE Patient

(
PatientId SMALLINT PRIMARY KEY,
Surname VARCHAR(25) NOT NULL,
FirstName VARCHAR(25),
Gender CHAR(1) CHECK (Gender = ‘M’ OR Gender=’F’),
WardName VARCHAR(25) FOREIGN KEY
REFERENCES Ward(WardName)
);

CREATE TABLE Ward

(
WardName VARCHAR(25) PRIMARY KEY,
NurseInCharge VARCHAR(25) NOT NULL,
NoOfBeds SMALLINT
);

528 Single licence - Abingdon School

 Using SQL to create a database

There is a variation on syntax for specifying primary key. It can also be defined at table level as shown below. You
will need to check which syntax is accepted by the database system that you use. A foreign key is expressed at table
level as shown below.
Table level:
CREATE TABLE Patient
(
PatientId SMALLINT,
Surname VARCHAR(25) NOT NULL,
FirstName VARCHAR(25),
Gender CHAR(1) CHECK (Gender = ‘M’ OR Gender=’F’),
WardName VARCHAR(25),
PRIMARY KEY(PatientId),
FOREIGN KEY(WardName)
REFERENCES Ward(WardName)
);

Composite primary key

A composite primary key can only be defined by a named constraint at the table level as follows. The example
below also shows how foreign keys can be defined by a named constraint as well, necessarily so if the foreign key
is composite. In the example below, EnrolClassNoPK is the name of a named constraint that defines a composite
primary key, YearNo, ClassNo, StudentId for the EnrolForClass relation. EnrolClassFK is the name of a named
constraint that defines the foreign key YearNo, ClassNo that references the primary key YearNo, ClassNo in the Class
relation (not shown).
CREATE TABLE EnrolForClass
(
StudentId CHAR(6),
YearNo CHAR(4)
ClassNo CHAR(8),
Grade CHAR(2),
CONSTRAINT EnrolClassPK PRIMARY KEY (YearNo, ClassNo, StudentId),
CONSTRAINT EnrolClassFK FOREIGN KEY (YearNo, ClassNo)
REFERENCES Class (YearNo, ClassNo),
CONSTRAINT EnrolStudentFK FOREIGN KEY (StudentId)
REFERENCES Student (StudentId)
);

Single licence - Abingdon School 529

10 Fundamentals of databases

Data types
Table 10.4.13 shows some general data types used in SQL (SQL:1999 standard) when defining tables using
CREATE TABLE.
CHARACTER(n) or CHAR(n) Character string. Fixed-length n. Default length is 1
when n is omitted. Use CHAR when the sizes of the
column data entries don’t vary much.
VARCHAR(n) or CHARACTER Character string. Variable length. Maximum length n.
VARYING(n) or CHAR Default length is 1 when n is omitted. Use VARCHAR
VARYING(n) when the sizes of the column data entries vary
considerably.
NCHAR(n) National character string type that uses fixed width
Unicode string. Width specified by n.
NCHAR VARYING(n) National character string type that uses variable width
Unicode string. Maximum width specified by n.
BOOLEAN Stores TRUE or FALSE values.
SMALLINT Integer numerical (no decimal). Signed unless keyword
UNSIGNED present. Precision 5. Exact numeric type.
INTEGER Integer numerical (no decimal). Signed unless keyword
UNSIGNED present. Precision 10. Exact numeric
type.
DECIMAL(p, s) or DECIMAL Exact numerical, precision p, scale s. Example: decimal
(5, 2) is a number that has 3 digits before the decimal
point and 2 digits after the decimal point. Exact
numeric type so can be used to represent money. Can
omit (p, s) to obtain default precision and scale.
FLOAT(p) or FLOAT Approximate numerical, mantissa precision p. A
floating number in base 10 exponential notation. The
size argument for this type consists of a single number
specifying the minimum precision. Approximate
numeric type. Can omit (p) to obtain default precision.
REAL Approximate numerical, mantissa precision 7.
DATE Stores year, month, and day values.
TIME Stores hour, minute, and second values.
Table 10.4.13 Table for some general data types used in SQL:1999 (ISO/IEC 9075-2:1999 (E))

Database Engine to practise with

SQLite is a self-contained, server-less, zero-configuration, transactional SQL database engine. The code for SQLite
is in the public domain and is thus free for use for any purpose, commercial or private. It is obtainable from
http://www.sqlite.org/. There are several other database engines that can be used instead of SQLite, some come with
a graphical user interface manager.

530 Single licence - Abingdon School

 Using SQL to create a database

Command line operation

To use SQLite requires the sqlite3 program, “sqlite3.exe”. Typing sqlite3 at the command prompt of a
console or terminal window, optionally followed by the name of the file that holds the SQLite database, starts the
SQLite database engine. If the file holding the database does not exist, a new database file with the given name
will be created automatically. If no database file is specified, a temporary database is created, then deleted when the
sqlite3 program exits unless saved with the .save filename command. The engine can also be started by clicking
on sqlite3.exe in the usual manner. That was how the window shown in Figure 10.4.1 was opened.
When started, the sqlite3 program shows a brief banner message as shown in Figure 10.4.1, before prompting
to receive SQL. SQL statements (terminated by a semicolon) can be entered at the command prompt followed by
pressing the “Enter” key to execute the SQL.

Figure 10.4.1 SQLite running in a console window

Figure 10.4.1 also shows a database, Ward.db being created (it didn’t previously exist) with the command .open
and command line argument Ward.db.
.open Ward.db

If it does exist then it will be opened by the same command and command line argument.
Next, the database’s first table is created with the Data Definition Language (DDL) command

CREATE TABLE Ward (WardName VARCHAR(20) PRIMARY KEY, NurseInCharge VARCHAR(20),

NoOfBeds SMALLINT);

The data types applied to the attributes are VARCHAR(20) and SMALLINT.
The command .schema when executed at the command prompt returns the DDL script that was used to create
tables and any other structures such as indices, triggers, etc.
The DML command to add to table Ward the tuple ('Victoria', 'Sister Bunn', 30)
is as follows INSERT INTO Ward VALUES ('Victoria', 'Sister Bunn', 30);

Confirmation of the success of this insert is shown in Figure 10.4.1 using the Data Manipulation Language
(DML) query SELECT * FROM Ward;

Single licence - Abingdon School 531

10 Fundamentals of databases

Next, the database’s second table, Patient, is created with the Data Definition Language (DDL) command below,
also shown in Figure 10.4.2.
CREATE TABLE Patient
(PatientId SMALLINT PRIMARY KEY, Surname VARCHAR(20), WardName VARCHAR(20),
FOREIGN KEY(WardName) REFERENCES Ward(WardName));

This table is linked to the Ward table by a foreign key which is created in table Patient by

FOREIGN KEY(WardName) REFERENCES Ward(WardName)

Figure 10.4.2 Creating table Patient with one row of values

The DML command to add to table Patient the tuple (2, 'Smith', 'Victoria')
is as follows INSERT INTO Patient VALUES (2, 'Smith', 'Victoria');

The value 'Victoria' used for the attribute WardName of Patient satisfies the referential integrity condition of
existence of this value in the Ward table. If it didn’t the INSERT action would be aborted and the reason for the
failure to carry out the command reported in an explanatory error message.
Confirmation of the success of this insert is shown in Figure 10.4.2 using the Data Manipulation Language
(DML) query SELECT * FROM Patient;

Questions

14 Write the DDL statements to create tables in database Library for the relations
Borrower (BorrowerId, Surname, Initial)
BooksOnLoan (ISBN, CopyNo, BorrowerId, DateDueBack)
as shown in Tables 10.4.10 and 10.4.11.

In this chapter you have covered:

■ How to use SQL to
• retrieve data from multiple tables of a relational database using SELECT FROM WHERE
• update data in tables of a relational database using UPDATE SET WHERE
• insert data into tables of a relational database using INSERT INTO VALUES
• delete data in tables of a relational database DELETE FROM WHERE
■ How to use SQL to define a database table using CREATE TABLE.

532 Single licence - Abingdon School

10 Fundamentals of databases
10.5 Client server databases
Learning objectives: ■ 10.5 Client server databases
■ Know that a client server Client server database system
database system provides
Simultaneous access
simultaneous access to the
In a multi-user client server database system, stored data items in a database
database for multiple clients
on a server may be accessed simultaneously by programs running on client
■ Know how concurrent access workstations as shown in Figure 10.5.1.
can be controlled to preserve The database server allows users at client workstations to retrieve information
the integrity of the database (read operation) stored in the database and modify this information (update/
insert/delete operations).
User Typically, a user reads data items from one or
1 Client
more database tables before modifying one or
more of these data items. A database transaction
User is a group of these operations. Each operation
2
Client
must succeed before the entire database
Server Database
transaction is considered successful. If an
User
3
Client operation fails, a database transaction allows the
program to back out from all previous operations
User and leave the database in its original state.
4
Client

Figure 10.5.1 Client server database system with users 1 to 4

Concurrent access
Key term Transactions submitted by users at client workstations can sometimes target
Client server database: the same data items in the database. Access to the same data items by different
In a multi-user client server users can occur over the same time interval, i.e. concurrently. This can result in
database system, stored data
an inconsistent database if not carefully controlled.
items in a database on a server
An inconsistent database is one in which the state of the database does not
may be accessed simultaneously
by programs running on client truly reflect reality, e.g. the number of airline seats sold should be 120 but
workstations. the database has recorded only 118 seats sold. The integrity of the data in the
database will be affected. Data integrity refers to the accuracy and consistency
of the data in the database.
Key term Lost update
To illustrate this potential problem consider an airline flight reservation
Concurrent access:
Means access occurs at the same
database system which stores information about each flight such as:
time or over the same time • flight number
interval.
• date of flight
• number of seats sold
• the number of seats left to sell

Single licence - Abingdon School 533

10 Fundamentals of databases

Suppose that a computer terminal located in a travel agent’s office in Birmingham attempts to book three seats
about the same time as a terminal located in a travel agent’s in Swindon attempts to book five seats on the same
flight. They each request copies of data for this flight from the server located in London. Figure 10.5.2 illustrates
what could happen. The problem that ensues is known as the lost update problem. This occurs when two
transactions that access the same database items have their operations interleaved in such a way that makes the value
of some database item incorrect.
Key term
Flight Code AY67
Lost update problem:
An update is lost when two Flight Date 21/11/2016
transactions that access the
same database item have their No of seats sold 120
operations interleaved in such
No of seats unsold 140
a way that makes the value of
the database item incorrect or
inconsistent.
Each office gets a copy
of the same information

Flight Code AY67 Flight Code AY67

Flight Date 21/11/2016 Flight Date 21/11/2016

No of seats sold 120 No of seats sold 120

No of seats unsold 140 No of seats unsold 140

Birmingham office Swindon office

sells 3 seats and updates their copy sells 5 seats and updates their copy

Flight Code AY67 Flight Code AY67

Flight Date 21/11/2016 Flight Date 21/11/2016

No of seats sold 123 No of seats sold 125

No of seats unsold 137 No of seats unsold 135

Birmingham office writes their updated

copy back to London just before Swindon office writes their updated
Swindon office writes their copy copy back to London

Birmingham office’s change is

Flight Code AY67
lost. The database is now
An UPDATE Flight Date 21/11/2016
inconsistent. The actual number
of seats sold is 128 but the
has been lost! No of seats sold 125
database indicates 125. This is
incorrect! The actual number of
unsold seats is 132 not 135. This is
No of seats unsold 135 also incorrect!

Figure 10.5.2 Lost update problem

534 Single licence - Abingdon School

 10.5 Client server databases

If we consider the detail of how individual seats are booked then we see that it is a two-stage process as shown in
Figure 10.5.3:
1. The database is first queried to obtain the seating plan for a specific flight, e.g. flight AY67 on 21/11/2016,
using an SQL Select statement sent from the client to the database server
2. The SQL query results returned are used to select a seat which is not already reserved (SeatBooked =
FALSE) and then an SQL update operation is sent to the database server to set the chosen seat to reserved
(SeatBooked = TRUE) for the given flight and date.

Select * From Flights

Where FlightNo = 'AY67'
And Date = '21/11/2016'
User Client And SeatBooked = FALSE
1
Results of Query
Database
Server
Results of Query
User Client
2 Select * From Flights A B C D E F
Where FlightNo = 'AY67'
And Date = '21/11/2016'
1
And SeatBooked = FALSE 2
3
4 AY67
5 21/11/2016
6
7
8

Update Flights
Set SeatBooked = TRUE, CustID = 2003
Where FlightNo = 'AY67'
And Date = '21/11/2016'
User Client And SeatID = '5B',
1

Database
Server

User
Client A B C D E F
2 Update Flights
Set SeatBooked = TRUE, CustID = 2165 1
Where FlightNo = 'AY67'
2
And Date = '21/11/2016'
And SeatID = '5B', 3
4 AY67
5 21/11/2016
Which customer is allocated seat 5B? 6
7
8

Figure 10.5.3 Two users, User 1 and User 2, attempting to reserve the same seat at the same
time and in the process allowing one user to overwrite the other’s reservation
Single licence - Abingdon School 535
10 Fundamentals of databases

A problem arises if the SQL Update operations proceed as shown in Figure 10.5.3. Both User 1 and User 2 believe
that seat 5B is available at the time when they each send an SQL Update operation to the database server to reserve
seat 5B. The database can only record this seat as being booked by one customer, either the customer with CustID
2003 or the customer with CustID 2165. If a database server allows uncontrolled access to a specific record in
the manner described then it will become inconsistent, i.e. its record of the bookings will not agree with its users’
record. To avoid this problem the database server must control concurrent access to its database.
Record locking
Key concept
One approach relies upon a locking mechanism. Consider a simplified version
Record lock: of the scenario shown in Figure 10.5.3 in which relation Flight consists of
A record lock is a concurrency attributes, FlightNo and NoOfSeatsSold. Its table depiction consisting
control method which can
of two columns is shown in Table 10.5.1.
prevent a lost update. The lock
ensures exclusive-access to a FlightNo NoOfSeatsSold
record when it is being updated. 1 50
Other transactions are blocked 2 120
from accessing the record until 3 87
after it has been updated and the 4 156
change permanently recorded.
The blocked transactions then Table 10.5.1 Relation Flight consisting of two columns in table form
see the updated record. Now suppose that two clients, User 1 and User 2, attempt to access the same
row of the Flight table, e.g. row 1.
Let User 1 retrieve row 1 first. After that, assume that User 2 retrieves the same
row.
Time User 1 User 2
postgres=# BEGIN; USER 1 starts transaction postgres=# BEGIN; USER 2 starts transaction
BEGIN BEGIN
postgres=# SELECT * FROM Flight WHERE
FlightNo = 1 FOR UPDATE; FlightNo | NoOfSeatsSold
-------------------------
1 | 50
postgres=# SELECT * FROM Flight WHERE
FlightNo = 1;
postgres=# UPDATE Flight SET NoOfSeatsSold BLOCKED
= 52 WHERE FlightNo = 1;USER 1 sells 2 seats
UPDATE 1

postgres=# COMMIT; USER 1 ends transaction

COMMIT UNBLOCKED - Select query now executes
postgres=#
FlightNo | NoOfSeatsSold
-------------------------
1 | 52

postgres=# COMMIT; USER 2 ends transaction

COMMIT
postgres=#
Table 10.5.2 User 1 and User 2 interacting concurrently with the Flight table

Table 10.5.2 shows User 1 using the BEGIN command to start a transaction on the PostgreSQL database containing
the table Flight. User 1 wishes to update the NoOfSeatsSold field of the record in the first row.

536 Single licence - Abingdon School

 10.5 Client server databases

It locks this record with the SQL command

SELECT * FROM Flight WHERE FlightNo = 1 FOR UPDATE;
The "SELECT * FROM Flight WHERE FlightNo = 1" part queries the database. The database server
returns the query result
FlightNo | NoOfSeatsSold
--------------------------------
1 | 50

The "FOR UPDATE" part locks the record to prevent other users from accessing this record until the lock is
released. This lock is known as an exclusive-lock for this reason.
User 2 sends the following SELECT query to the database server just after User 1 gets the query result shown
above
SELECT * FROM Flight WHERE FlightNo = 1
User 2’s query is blocked because the record with FlightNo = 1 is locked.
User 1 then sends the following UPDATE command to the database server
UPDATE Flight SET NoOfSeatsSold = 52 WHERE FlightNo = 1;
To commit this update to the database and end the transaction, User 1 sends the COMMIT command.
This update is applied and the lock is removed. Whereupon, the pending SELECT query from User 2 is executed
by the database server and the following query results are returned to User 2
FlightNo | NoOfSeatsSold
--------------------------------
1 | 52

Note that the NoOfSeatsSold reflects the update applied by User 1.

If User 2 had attempted a "SELECT * FROM Flight WHERE FlightNo = 1 FOR UPDATE;" just after
User 1 had set an exclusive-lock then it would have been blocked. Only when User 1’s transaction was finished
would User 2 been granted an exclusive-lock. User 2 could then have applied an update to the NoOfSeatsSold
field of record 1. This is how locking can be used to prevent the Lost Update problem occurring.
In the example, the lock applied to the Flight record/row of the table where FlightNo = 1. If User 2 had
chosen to query a different row then User 2 would not have been blocked. For a similar reason, User 2 would not
have been blocked from updating a different row of the Flight table.
Serialisation
The isolation level of a transaction determines what data the transaction can see and whether it can modify a data
item or not when other transactions are executing concurrently. The highest level of isolation is SERIALIZABLE.
The default level is often READ COMMITTED. This is a lower level of isolation. READ COMMITTED does not
prevent the LOST UPDATE problem whereas SERIALIZABLE does.
The isolation provided by each is as follows:
READ COMMITTED (Not in AQA specification but included because it is relevant to understanding serialisation)
An SQL operation can only see rows committed before the transaction began. It does not prevent the
situation when a modification by one transaction is overwritten by another executing concurrently.
SERIALIZABLE
1. A transaction cannot read data written by a concurrent uncommitted transaction.
2. A transaction can only see data that was committed before it began even when this data has been
modified and committed by another transaction.
3. A transaction cannot modify a data item that has been modified by a concurrent transaction.

Single licence - Abingdon School 537

10 Fundamentals of databases

If the isolation mode is set to SERIALIZABLE then the database server can detect and prevent the LOST
UPDATE problem occurring. A characteristic of the LOST UPDATE problem is that the transactions that cause
this conflict are not serialisable, i.e. do not behave as if they were executed one right after another in a serial fashion.
The isolation level can be set using the SET TRANSACTION CHARACTERISTICS command or the SET
SESSION CHARACTERISTICS. These set the default transaction characteristics for a transaction and for
subsequent transactions of a session, respectively.

Time User 1 User 2

SET SESSION CHARACTERISTICS AS TRANSACTION
ISOLATION LEVEL READ COMMITTED;
SET
postgres=# BEGIN; USER 1 starts transaction
BEGIN
SET SESSION CHARACTERISTICS AS
TRANSACTION ISOLATION LEVEL READ
COMMITTED;
SET
postgres=# BEGIN; USER 2 starts
BEGIN transaction
postgres=# SELECT * FROM Flight WHERE
FlightNo = 1; FlightNo | NoOfSeatsSold
-------------------------
1 | 50
postgres=# SELECT * FROM Flight WHERE
FlightNo = 1; FlightNo | NoOfSeatsSold
-------------------------
1 | 50
postgres=# UPDATE Flight SET NoOfSeatsSold
= 52 WHERE FlightNo = 1;
UPDATE 1 USER 1 sells 2 seats
postgres=# UPDATE Flight SET
NoOfSeatsSold = 51 WHERE FlightNo = 1;
BLOCKED TEMPORARILY
postgres=# COMMIT;
COMMIT UNBLOCKED
postgres=#
UPDATE 1 USER 2 sells 1 seat
postgres=# COMMIT;
COMMIT
postgres=#
postgres=# SELECT * FROM Flight WHERE SHOULD BE 53, 2 SOLD BY USER 1 and 1
FlightNo = 1; FlightNo | NoOfSeatsSold
------------------------- SOLD BY USER 2
LOST UPDATE 1 | 51

Table 10.5.3 User 1 and User 2 interacting concurrently with the Flight table
when isolation level is READ COMMITTED

Table 10.5.3 shows User 1 and User 2 interacting concurrently with the Flight table when the isolation level
is READ COMMITTED. This setting fails to prevent the Lost Update problem. What is needed is a way of
detecting and preventing concurrent transactions that would not execute serially, if attempted. Figure 10.5.4 shows
User 2’s attempt to update the same record as User 1 but later in time than User 1. The database server forces User
2 to abort its update because the server has concluded that it would not be possible to serialise these updates. When
538 Single licence - Abingdon School
 10.5 Client server databases

User 2 restarts the transaction it finds that the NoOfSeatsSold field has been updated. User 2 uses the new value
52 recorded in the database, adding one to calculate the value 53 which it uses to update the database.
NoOfSeatsSold = 50 Make change permanent NoOfSeatsSold = 52
Figure 10.5.4 The effect of serialisation on
BEGIN SELECT UPDATE COMMIT USER 1 two transactions interleaved in time when
both attempt to update the same data item
NoOfSeatsSold ← 52 Update record
Could not serialize access NoOfSeatsSold = 53
NoOfSeatsSold = 50 due to concurrent update NoOfSeatsSold = 52

USER 2 BEGIN SELECT UPDATE ROLLBACK BEGIN SELECT UPDATE COMMIT

NoOfSeatsSold ← 51 NoOfSeatsSold ← 53

ROLLBACK
cancel change
Table 10.5.4 shows the sequence of events for two users interacting concurrently with a postgreSQL database when
the transaction isolation is set to SERIALIZABLE.

Time User 1 User 2

SET SESSION CHARACTERISTICS AS TRANSACTION
ISOLATION LEVEL SERIALIZABLE;
SET
postgres=# BEGIN; USER 1 starts transaction
BEGIN
SET SESSION CHARACTERISTICS AS TRANSACTION
ISOLATION LEVEL SERIALIZABLE;
SET
postgres=# BEGIN; USER 2 starts transaction
BEGIN
postgres=# SELECT * FROM Flight
WHERE FlightNo = 1; FlightNo | NoOfSeatsSold
-------------------------
1 | 50
postgres=# SELECT * FROM Flight
WHERE FlightNo = 1; FlightNo | NoOfSeatsSold
-------------------------
1 | 50
postgres=# UPDATE Flight
SET NoOfSeatsSold = 52 WHERE FlightNo = 1;
UPDATE 1 USER 1 sells 2 seats

postgres=# COMMIT; USER 1 ends transaction

COMMIT Change made permanent
Postgres=#
USER 2 cannot see the change even though it is now permanent postgres=# SELECT * FROM Flight
WHERE FlightNo = 1; FlightNo | NoOfSeatsSold
-------------------------
1 | 50
USER 2 is stopped from updating the database because its data is postgres=# UPDATE Flight SET NoOfSeatsSold
out of date. If allowed it would overwrite the change that USER 1 = 51 WHERE FlightNo = 1;
made and lead to a LOST UPDATE. Could not serialize access due to concurrent update
postgres=# SELECT * FROM Flight WHERE postgres=# ROLLBACK; USER 2 ends transaction
FlightNo = 1; ROLLBACK Update aborted
FlightNo | NoOfSeatsSold
Database is consistent ------------------------- postgres=#
1 | 52
USER 2 must try again to update the database adding 1 to 52 not postgres=# SELECT * FROM Flight WHERE
1 to 50: UPDATE Flight SET NoOfSeatsSold = 53 FlightNo = 1; FlightNo | NoOfSeatsSold
Database is consistent
-------------------------
WHERE FlightNo = 1; 1 | 52

Table 10.5.4 Effect of serialisation on two transactions interleaved in time when both update the same data item
and the isolation level is SERIALIZABLE
Single licence - Abingdon School 539
10 Fundamentals of databases

Key concept Timestamp ordering

A transaction is a unit of work. When a transaction is committed its work
Serialisation: is done and any change (update, insert, delete) is made permanent. Until a
Serialisation attempts to serialise
transaction is committed changes are registered with the database server but are
access to a data item in order
to detect and prevent the lost not actually applied to the database. They are considered temporary changes
update problem occurring. which means that they can be undone or rolled back (ROLLBACK).
Transactions attempting to alter
Temporary changes are made first to allow the database system to check for
a data item that is currently the
subject of another transaction concurrency violations. If none have occurred then the changes are applied
are detected and aborted. Any permanently to the database. However, if the database system detects a
temporary changes cancelled. concurrency violation then the transaction is aborted and no permanent
change is made to the database. Any temporary change must be undone. This
is called ROLLBACK.

Key concept Typically, timestamp values are assigned in the order in which transactions are
submitted to the system, so a transaction timestamp can be thought of as the
Timestamp ordering:
transaction start time. We refer to the timestamp of transaction T as TS(T). A
A timestamp is a unique
identifier created by a database simple counter which starts at zero can be used as the source of transaction
server that indicates the relative number. Each new transaction causes the counter to be incremented and the
starting time of a transaction. resulting new counter value is assigned to the transaction.
The database records the
transaction timestamp of the last The database server records the largest timestamp for a data item, X, when
transaction to read data item X a read transaction is performed on the data item, X - ReadTS(X). It records
and the transaction timestamp separately, the largest timestamp for a data item, X, when a write transaction
of the last transaction to write is performed on the data item, X - WriteTS(X). The counter starts initially at
data item X.
zero.
The database server applies rules
using these to determine if a Let’s suppose User 1 begins a transaction, T1, that contains a Read request
transaction’s actions will result followed by a Write request. The timestamp counter is incremented and its new
in the integrity of the database
value 1 (it started at 0) assigned to transaction T1.
being compromised. If it will
the server aborts the transaction. User 2 begins a transaction T2 and is assigned transaction time stamp value
2. T2 also contains a Read request and a Write request. Now suppose the time
ordering of these Read and Write requests is as shown in Table 10.5.5. User
2 completes before User 1. Therefore, User 1’s Read data is potentially out of
date. Therefore, User 1’s transaction, T1 must be aborted.

User 1 TS(T) User 2 Counter ReadTS(X) WriteTS(X)

0 0 0
Starts a transaction 1 1 0 0
Read request 1 1 1 0
2 Starts a transaction 2 1 0
2 Read request 2 2 0
2 Write request 2 2 2
Write request 1

Table 10.5.5 Timestamp ordering protocol ensures serializability among transactions in their conflicting
read and write operations

540 Single licence - Abingdon School

 10.5 Client server databases

The following rules are used with timestamps to determine whether a transaction operation is allowed or not.
Rules: Another transaction has read
1. Transaction T1 issues a WriteItem(X) operation: X since T1 started or another
transaction has modified X
If ReadTS(X) > TS(T1) Or WriteTS(X) > TS(T1)
since T1 started
Then abort and rollback T1 and reject the operation.
Else
Another transaction modified
If WriteTS(X) <= TS(T1)
X but before T1 started
Then execute the WriteItem(X) operation of T1
and set WriteTS(X) to TS(T1)

2. Transaction T1 issues a ReadItem(X) operation:

Another transaction modified
If WriteTS(x) > TS(T1)
X after T1 started
Then abort and rollback T1 and reject the operation.

Else
Another transaction modified
If WriteTS(x) <= TS(T1)
X but before T1 started
Then execute the ReadItem(x) operation of T1

and set ReadTS(x) to the larger of TS(T1)

and the current ReadTS(x)

If we revisit the User 1 and User 2 lost update problem, then an attempt by User 1 to update the Flight record
for which FlightNo = 1 will result in User 1’s transaction being rescheduled because User 2’s transaction
timestamp has a larger counter value than User 1’s.
Commitment ordering
Commitment ordering is used in client server database systems such as mobile banking, and e-commerce.
A transaction is an “all or nothing” unit of work. A transaction is either committed, i.e. its effects on all the
resources (data items) involved become permanent, or it is aborted (rolled back), i.e. its effects on all the resources
are undone. The term that summarises this is atomicity:
In database systems, an atomic transaction is an indivisible and irreducible series of database operations such
that either all occur, or nothing occurs.
When a transaction is in conflict with another transaction, i.e. committing Key concept
both would lead to database inconsistency such as the lost update problem, one
Commitment ordering:
of the transactions is aborted and started again. This is a form of concurrency
Used in a distributed database
control. system such as mobile device
Concurrency control applied to a client server database system where clients are client server systems where
mobile devices presents its own special problems. Mobile devices move from transactions are created locally
on the mobile device before
base station to base station and the connection with the database server can be
being sent to the server for
intermittent. What happens if a transaction is started on a mobile device which execution. The transactions are
fails to complete in a reasonable time because the connection is lost? There scheduled at the server so that
is a danger, if locking is used, of a data item being locked in exclusive-access they are committed in an order
that avoids concurrency conflict.
mode for a lengthy period of time thus preventing other transactions from
completing.

Single licence - Abingdon School 541

 10.5 Client server databases

One solution is for transactions to be done locally on the mobile device before they are sent to the server for the
final stage. In a non-distributed system, the transactions are created centrally at the server. However, mobile device
database systems uses a distributed model in which the transactions are generated locally. Relative timestamps
are assigned locally to operations (Read, Update, Insert, Delete). The list of operations with timestamps is then
transferred to the server. The absolute time of the execution of the operations is calculated in the server using
the relative timestamps. The server uses a scheduling algorithm to produce a commit order which avoids conflict
between transactions. This is what is meant by commitment ordering.

Questions
1 What is a client server database system?

2 What is meant by concurrent access?

3 What is the Lost Update problem?

4 Describe three methods that are used in non-distributed client server database systems to prevent the Lost
Update problem occurring.

5 Describe a method used in distributed client server database systems such as mobile banking to prevent the
Lost Update problem occurring.

In this chapter you have covered:

■■ Client server database systems which provide simultaneous access to a database for multiple clients
■■ How concurrent access can be controlled to preserve the integrity of the database
• Record locks
• Timestamp ordering
• Serialisation
• Commitment ordering

Single licence - Abingdon School 542

11 Big Data
11.1 Big Data
Learning objectives:
Section 1
■■Know that 'Big Data' is a
catch-all term for data that
■■ 11.1 Big Data
won’t fit the usual containers What is Big Data
The term Big Data is described as data that can’t be processed or analysed
■■Know that Big Data can be using traditional processes or tools because it falls into one or more of the
described in terms of
following categories:
• volume - too big to fit into ■■ too big to fit into a single server
a single server
■■ too heterogeneous (diverse in character or content) - structured, semi-
• velocity - streaming data, structured or totally unstructured
milliseconds to seconds to ■■ its production can occur at very high rates.
respond
To give you an example of the challenges of Big Data, on a typical day in
• variety - data in many 2015, 500 million tweets were sent with an average of 5,700 tweets per second
forms such as structured, (TPS). During the 2010 World Cup, the influx of tweets - from every shot on
unstructured, text, goal, penalty kick and yellow or red card - repeatedly took its toll and made
multimedia. Twitter unavailable for short periods of time. Twitter had reached the limit
Did you know? of throughput on their storage systems. The MySQL storage system Twitter
employed was having trouble processing tweets at the rate that they were
On Saturday, August 3, 2013
showing up, and the solution of throwing more machines at the problem, each
in Japan, people watched an
airing of “Castle in the Sky”, with a new MySQL database, was deemed a sticking plaster solution whilst
and at one moment they took a complete redesign of their system was undertaken. The bottlenecks in the
to Twitter so much that a one- system were the system’s inability to handle concurrency effectively - multiple
second peak of 143,199 tweets
tweets arriving simultaneously - and the system’s inability to handle the storage
per second was recorded. This
was a world record.
requirements.
Twitter’s solution was to create a sharded and fault-tolerant distributed database
Key concept system. A database shard is a horizontal partition of data in a database, e.g.
the first 10,000 rows. Each shard is then held on a separate database server
Big Data:
instance, to spread load.
Big Data is the term applied to
datasets whose size is beyond Volume,Velocity and Variety
the ability of traditional
The three defining characteristics of Big Data are
software tools to capture,
manage and process; or whose 1. Volume - data to be analysed is too big to fit into a single server
production rate exceeds the
2. Velocity - speed at which data must be processed to keep up, if the
ability of this software to
respond in a timely manner; or data is at rest it can be batch processed but if it is data in motion, i.e.
the dataset lacks the necessary streamed data, then processing needs to take place in real time.
structure to be stored and
3. Variety - data can appear in many forms from structured through
processed in the relational
model. semi-structured to unstructured.
A dataset is a collection of data.

Single licence - Abingdon School 543

11 Big Data

Information This is summarised pictorially in Figure 11.1.1.

Big Data is measured in terabytes
(1012 bytes) or even petabytes ity Var
loc iet
(1015 bytes), and is rapidly heading Ve Streaming Data Structured Data y
towards exabytes (1018 bytes). Big
In 2015, Facebook generates 10 Data Unstructured
Batch
terabytes (TB) of data every day, Data
Zetabytes Terabytes
Twitter 7 terabytes. One terabyte is
1012 bytes.
Volume

Information
Google recognised in 2000 that Figure 11.1.1 Schematic of the three V’s, Volume, Velocity and Variety
the custom-built infrastructure
it had been relying on to provide Questions
its search engine service was
no longer fit for purpose. 1 What is meant by Big Data?
This infrastructure could not
be scaled to cope with the
2 Name the three defining characteristics of Big Data.
increasing volume of data being
collected nor the demands its
users were placing upon it.
Volume
Google were unable to buy
Volume in Big Data refers to the size of the data to be processed. Data-analysis
a solution off the shelf, at the
time, because no commercially
scenarios that process hundreds of terabytes or petabytes fall into the Big Data
available software existed that category if that data must be analysed as a single dataset.
could handle this volume and
A physical data centre that hosts an exabyte of data is not necessarily dealing
level of demand.
with Big Data. This data might be nightly backups stretching back over ten
Google’s solution was to design
and build their own new data years. However, if, say, a petabyte of data needs to be analysed to answer a given
processing infrastructure to solve question, then this is definitely a Big Data problem. Thus the “Big” in Big Data
this problem. is more than an assessment of the size of the data to be stored, it really relates to
The two key services in this the processing of this size of data.
system were
• a distributed file system, When the data volumes that needed to be processed outgrew the storage and
the Google File System, or
processing capabilities of a single host an alternative to the traditional data
GFS, which provided fault-
tolerant, reliable, and
processing approach was required.
scalable storage In a traditional data processing system,
• MapReduce, a data Information
processing system based ■■ file blocks belonging to a file of
on a functional data would be stored on a single A file is divided into blocks of a
programming paradigm server fixed, pre-determined size. The
that allowed work to be split chosen block size corresponds
among large numbers of ■■ The program written to execute to the block size written to a
servers and carried out in this traditional data processing physical disk.
parallel. activity would also execute in a
Both GFS and MapReduce single server.
were designed to run on the
commodity server hardware In the Big Data era, the blocks of a file have to be distributed across more
that Google used throughout its
than one server simply because there are too many of them to fit one server. In
data centres.

544 Single licence - Abingdon School

 11.1 Big Data

traditional data processing systems the block size is typically 512 bytes whereas
Information
in a Big Data processing system, the block size is typically 64 or 128 megabytes.
Growth in data:
Secondly, the program that does the processing must now be written so that it
The seventh annual Digital Universe
can execute on more than one machine at the same time.
study (2014)
Summarising, the Big Data approach to processing a dataset that has already (http://www.emc.com/about/news/

been collected and stored (data at rest) relies on press/2014/20140409-01.htm)

stated that the digital universe is
■■ a distributed file system doubling in size every two years and
will multiply 10-fold between 2013
■■ a way to parallelise and execute programs
and 2020 – from 4.4 zetabytes to 44
zetaabytes.
Questions
3 Explain the meaning of the Big Data characteristic known as volume.
Information
4 Explain how Big Data processing system differs from a traditional data
Commodity hardware:
processing system. Commodity hardware is a device or
device component that is relatively
inexpensive, widely available and
Distributed File System
more or less interchangeable with
A distributed file system is one in which the blocks of individual files are spread other hardware of its type.
across more than one server. Google’s distributed file system is GFS. Yahoo,
Facebook, and Twitter use HDFS, the Hadoop Distributed File System. Both
Information
systems use racks of servers with network switches interconnecting servers in
a rack and servers in other racks. Figure 11.1.2 shows a schematic consisting Yahoo:
To support research for ad-systems
of several servers connected via a network switch within a rack, Rack 1 and to
and Web search, Yahoo, in 2015
another rack, Rack 2. uses more than 100, 000 CPUs in
approximately 20,000 computers
Network
Network switch running Hadoop; the biggest cluster
Figure 11.1.2 Schematic switch has 2000 nodes (2∗4 cpu boxes
of several servers Rack 2 each with a 4TB disk).
connected via a network
Facebook:
switch
As a source for reporting analytics
and machine learning, Facebook in
2015 uses a 320-machine cluster
Rack 1 with 2560 cores and 1.3 PB of
storage to handle processing of
unstructured data.
Figure 11.1.3 shows racks of commodity servers at one of Google’s data
centres. The data centre houses some 100,000 commodity servers consuming
Information
a total power of 40 Megawatts. This is roughly the total power output of
Coolkeeragh power station in Northern Ireland. Mean Time To Failure (MTTF):
A particular device or device
Both GFS and HDFS are fault-tolerant. They need to be because of their
component will fail at some time.
reliance on commodity magnetic disk hard drives and servers. Commodity MTTF specifies on average how long
magnetic disk hard drives have a Mean-Time-To-Failure of about 300,000 a device or device component will
hours or a probability of failing within an hour of 1 in 300,000. With one last before failing.

Single licence - Abingdon School 545

11 Big Data

hundred thousand servers in a data centre this means a probability that a disk
drive will fail in the next hour of about 33%.

Figure 11.1.3 Racks of commodity servers at one of Google’s

data centres (image Google/Connie Zhou)
Figure 11.1.4 shows Google’s
first server rack. Their first Figure 11.1.5 shows in HDFS that each block belonging to a particular file is
search engine ran on this written three times with at least one block written to a different server rack, e.g.
system Block 2, Block 2' and Block 2''.

Information
Figure 11.1.4 shows Google’s first
Block 1 Block 3’
server rack circa 1999. Their first
search engine ran on this. It was
retired from service after five years Block 3 Block 1’
to the Computer History museum,
Mountain View, California, USA.
Block 2’ Block 2’’

Block 2
Information
HDFS stores files across a cluster.
A Hadoop cluster is a special type Block 1’’
of computational unit designed
specifically for storing and analysing Server +
Disk Drive Block 3’’
large datasets across a distributed
computing environment. One
Rack 1 Rack 2
machine in the cluster is designated
as the NameNode and another
machine as JobTracker - these are Figure 11.1.5 In HDFS each block is written three times and at least one
the masters. The rest of the machines of the three is written to a different server rack
in the cluster act as both DataNode
and TaskTracker; - these are the Both GFS and HDFS are scalable. More servers (known as nodes) can be
slaves. Hadoop clusters are often added if the file grows in size because all the racks of servers are interconnected
referred to as “shared nothing”
and the file system is able to keep track of which blocks belong to which file
systems because the only thing
that is shared between nodes is the
wherever the blocks are stored whether Rack 1 or Rack 10,000.
network that connects them.

546 Single licence - Abingdon School

 11.1 Big Data

Questions Tasks
5 What is meant by a distributed file system? 1 Watch a YouTube video
about a Facebook data
6 What is meant by fault-tolerant? centre
https://www.youtube.com/
7 Explain how fault-tolerance is achieved in HDFS. watch?v=0pB9falsA9k

8 What is meant by the term scalable in the context of HDFS and GFS?

Information
Function-to-data model Vertical scaling:
The second aspect of the Big Data approach is parallelising the execution of Takes place in a single server
programs. This is done with the function-to-data model. In the function-to- when more CPUs, memory, hard
drives are added.
data model because there is so much data, the analysis program is sent to the
data, i.e. a copy is sent to each server. This model is used in Hadoop which is Horizontal scaling:
an open source highly scalable, distributed batch processing system for large Takes place when more servers
datasets. At the heart of of the function-to-data model is a technique borrowed are added. It is like adding another

from functional programming called MapReduce. lane to a two-line highway to

make a three-lane highway in
All MapReduce programs that run natively under Hadoop are written in Java, order to accommodate the
and it is the Java Archive file (jar) that’s distributed by the JobTracker to the increase in traffic.
various Hadoop cluster nodes to execute map and reduce tasks.

Information
Information Traditional database systems
use special database servers
MapReduce
of high reliability. Such systems
MapReduce is a programming model which uses a parallel, distributed
attempt to solve the scaling
algorithm to process large datasets “at rest” in a cluster. MapReduce takes
problem by vertical scaling. This
an input, splits it into smaller parts, executes the code of the mapper on
is an expensive option because
every part, then gives all the results to one or more reducers that merge
highly-reliable components are
all the results into one.
expensive.
For example, suppose that we wanted to demonstrate determining the
number of times each word appears throughout a collection of texts, e.g.
Google’s solution was to
the novels War and Peace, Les Miserables, Three Men in a Boat and Fireside
scale horizontally using much
Stories.
cheaper commodity hardware.
First, the files containing the text of these novels would be copied from
Commodity hardware is cheaper
where they are stored to Hadoop’s distributed file system HDFS as shown
but less reliable. Google’s solution
in Figure 11.1.6.
was to build in fault-tolerance so
that the failure of a component
Next a copy of a Map function would be distributed to each HDFS server.
would not bring down the system.
In this example, the Map function would be something along the lines

function wordcountMap (lineoftext)

{for (word in lineoftext.split(" ")) {emit (word, 1)}}

Single licence - Abingdon School 547

11 Big Data

The collection of key-value pairs emitted by each copy of map would be stored temporarily before being
grouped and sorted by shuffle and sort operations, e.g. all <A, 1> key-value pairs would be grouped together
and placed before all <About, 1> key-value pairs. Finally, copies of a Reduce function count all <A, 1>s, all
<About, 1>s and so on to produce a total count for each word.
function wordcountReduce (word, value)
{sum = 0
for (nextvalue in value) {sum += nextvalue}
emit(word, sum)
} Map code sent to After shuffling key-value
servers hosting pairs sorted alphabetically
the input files’
blocks so net-
work traffic Sort
HDFS across the cluster Shuffle
Blocks minimised Phase
Finally, Reduce
<A 1>
<A 1> <A 1> produces a total
Map <cat 1> Shuffle
Linux File A cat sat on a mat....
<sat 1> <About 1> Reduce for each word
System <About 1>

Well Prince Genoa and Lucca <Well 1>

are now family estates of the Map <Prince 1>
<Genoa 1> Shuffle Sort
Buonapartes....
<epoch 1> <A 600>
<epoch 1> <Be 400>
<About 1>
About the epoch of the Map <the 1>
<epoch 1> Shuffle <epoch 1> Reduce <THREE 200>
coronation.... <epoch 1> <Well 350>

<THREE 1>
THREE MEN IN A BOAT (TO
Map Sort
SAY NOTHING OF THE DOG)....
<MEN 1>
Output file
<IN 1> Shuffle <THREE 1>
<THREE 1> spread across
several
<Well 1> Reduce
<Well 1> HDFS
Linux File System Blocks
block size Map Shuffle
4kB
Map emits key-value pairs, Sort
HDFS e.g. <cat 1>
block size for each word it encounters
64MB Running MapReduce on a
hduser@ubunutu: hadoop fs -put /etc/gutenberg/ /user/hduser/gutenberg Ubuntu
hduser@ubuntu: hadoop jar hadoop-mapreduce-examples-2.7.0.jar wordcount /user/hduser/gutenberg /user/hduser/gutenberg-output

Figure 11.1.6 MapReduce applied to text of several novels. The text is copied from the host operating system to
Hadoop’s HDFS file block system which sits on top of the host file system

Extension Question
9 Describe MapReduce in the context of counting word occurrences in a large dataset consisting of text.

548 Single licence - Abingdon School

 11.1 Big Data

Variety Information
Data can appear in many forms from structured through semi-structured to
unstructured. The simple, structured data
managed by legacy Relational
Structured data database management systems
Structured data is data that can be represented in tabular form. Such data is (RDBMS) invented in the 70s
modelled in the relational model as described in Chapter 10.2. Structured data continue to be used for simple,
lends itself to formal modelling of the data before building a database because structured data. They are not
suitable for complex data
the data has a clear and identifiable structure. In the relational model, the data
because such data does not
is structured into tuples that record values for a fixed number of predefined easily fit the row and column
attributes, format required by relational
<memo>
e.g. attributes databases. Data that fits the
<to>Sarah</to> latter is highly structured and
StudentId,
<from>Fred</from> can be pre-modelled prior to
StudentSurname,
<about>Milk for tea and coffee </about> collection and storing.
etc. However, a different kind
<status>Urgent</status>
Semi-structured of data classified as semi-
<message>
data structured or unstructured
Don’t forget to buy a pint of milk! now predominates.
This is data such
</message> Unstructured data can be
as XML- or
</memo> difficult to pre-model because
JSON-formatted of its lack of structure. It is
files that do not also characterised by sheer
<memo> volume and frequency of
have a formal
<to>Fred</to> generation. Examples are Web
structure but
<from>Sarah</from> logs of events such as mouse
nevertheless have
<about>Milk for tea and coffee </about> clicks on Web pages; text,
some structure, images, video, audio, sensor
<moan>I am always buying the milk</moan>
albeit of a variable readings, output from scientific
<reply>Bought one pint!</reply> modeling, and other complex
kind. Figure
</memo> data types streaming into data
11.1.7 shows
centres today.
Figure 11.1.7 Two related examples of XML two related New data processing
data showing the semi-structured nature of examples that are techniques must be applied to
this type of data XML-formatted. extract the value contained
The structure within this data.
Algorithms tailored to both
of each memo is clearly identified by tags that could be predefined but the
the data and the questions that
actual structure of each memo cannot be formally modelled because it is not can be answered are applied to
consistent. No schema is enforced to ensure consistency. each of the data types such as
images, video, or audio.
Unstructured data The Big Data approach is then
This is data such as text whose content is so variable that able, by linking the results, to
deliver insights that would not
■■ it cannot be modelled in advance or
be possible otherwise.
■■ it cannot be fitted to a column-row structure required by relational
database modelling or
■■ its (key) elements are not identified with tags.

Single licence - Abingdon School 549

11 Big Data

Information Examples include, the body of e-mail messages, web pages, text files, video,
Spring XD can work with
images, data from sensors. Even though each of these examples does have
both the Twitter Search API structure it is not of the semantic kind, e.g. a Web page contains HTML
(twittersearch) and data from mark-up such as paragraph tags, but this is solely for the purpose of
Twitter’s Streaming API. rendering the Web page and not for specifying its meaning (semantics) and
therefore what information it conveys.
Information
Before you can start consuming Questions
data from Twitter into Spring XD,
10 Explain the meaning of the Big Data characteristic known as variety.
you need to set up a development
account and install the Twitter
development environment. If
Velocity
you don’t have a Twitter account,
you can sign up for one at http:// Velocity refers to whether the data is in motion and with what frequency
www.twitter.com. The developer or at rest.
website enables users to create
Data at rest is data that has been stored on some permanent data storage
keys for their applications to
access Twitter’s APIs. These device. The data may be processed at any time because it is permanently
keys need to be placed in a stored and speed of processing is not critical as the rate of arrival and
configuration file - xd/config/ processing of this data can be controlled. Big Data at rest is usually batch
modules/modules.yml:
processed. In batch processing, processing, once started, is carried out to
twitter:
completion without user interaction.
consumerKey:
consumerSecret: Data in motion is data that is streamed at some frequency
accessToken: continuously, e.g. 1000 events per second. There is an expectation that it
accessTokenSecret:
will be collected at the rate it arrives. In stream processing, data is processed
as it arrives and before it is stored permanently.
Information In the Big Data scenario, this data is likely to arrive at a high rate, and
Spring XD can be downloaded
from multiple sources simultaneously - e.g. log data, Twitter streams,
as a zip file from https://spring.io/
and RSS feeds. The site http://developer.usa.gov/1usagov sends event
blog/2015/04/30/spring-xd-1-2-m1-
and-1-1-2-released for the version notifications at the rate of less than 10 per second. The stream is composed
available at time of writing this of JSON events, each of which is generated every time someone clicks
book. on any US .gov or .mil URL that has been shortened. When this site is
Unzip and install. To start the
launched in a browser a simple long-lived HTTP connection is established
server on a Windows machine
click on
and data is subsequently streamed back to the browser until the HTTP
\spring-xd-1.2.0.RELEASE\xd\bin\ connection is ended. 10 events or less per second is manageable by one
xd-singlenode.bat for single node machine but what if the velocity was increased to about 46,000 events per
operation. second - see www.internetlivestats.com. This equates to the rate at which
Launch the command shell by
Google receives search hits.
clicking on
\spring-xd-1.2.0.RELEASE\shell\ Imagine that Google streamed these search request events to the world.
bin\xd-shell.bat. Any system receiving these would have to cope with collecting the events
See www.educational-computing.
at a velocity of 46,000 per second. This would require scaling the collection
co.uk/CS/Book/SpringXD/
SpringXDExamples.txt for
part of the system so that it would cope, i.e. the equivalent of adding some
examples to try. more lanes to a highway.

550 Single licence - Abingdon School

 11.1 Big Data

Information
Figure 11.1.8 shows a basic stream of event-driven data passing from source to sink in Spring XD, a stream processing system applied to
Big Data. The source converts data arriving at unpredictable times into an internal message format. The messages travel along channels
between any number of processors. One kind, a “wire tap”, is shown in the figure. The data travelling in the stream is tapped into and
some processing carried out. In this example it is a simple count of tweets arriving from Twitter. The main stream is set up and deployed
with the shell command
stream create tweetcount --definition “tap:stream:tweets>aggregate-counter” --deploy true
The wire tap is set up with
stream create tweets --definition “twitterstream | file” --deploy true

Using Spring XD a unified, distributed, and extensible system for data ingestion, real time analytics, batch processing, and data export

sensors xd>stream create tweets --definition "twitterstream | file " --deploy true
(e.g. temperature)
data feed
(e.g. stock change, message Wire tap message
credit card transaction)
Source Sink
mobile

tweets
tweets tweets
social ts
ee message
(tweets, tw HDFS
channel
count of tweets

blogs)
message
tweets.out

storage

sink for
Spring XD count
xd>stream create tweetcount --definition "tap:stream:tweets> aggregate-counter" --deploy true

{“created_at”:”Wed Jul 08 12:26:35 +0000 2015”,”id”:618758072752832512,”text”:”RJ is going to use \”big data..

{“created_at”:”Wed Jul 08 12:26:36 +0000 2015”,”id”:618758078813573120,”text”:”@ukieandyt @uk_ie @doct...
{“created_at”:”Wed Jul 08 12:26:39 +0000 2015”,”id”:618758090146607104,”text”:”Big Data: Genomical .........
{“created_at”:”Wed Jul 08 12:26:44 +0000 2015”,”id”:618758110690308096,”text”:”RT @BigDataBlogs: ............

Figure 11.1.8 Streaming using Spring XD tool to pull tweets from Twitter, perform some processing on the
stream as it is in motion with atap into the stream and storing all tweets received in tweets.out for later batch
processing

The role of the sink is to store the stream data permanently. This could be as a file using the file system of the operating system or it
could be HDFS, for example, a distributed file system built on top of the native file system of the operating system. In addition to the
stream data being “tapped into”. It could be processed in transit before being stored, e.g. case converted from lower case to upper case.
Figure 11.1.9 shows the visualisation of tweets filtered by the “lang” field in tweets at a wire tap. The visualisation updates in real time.
To see this in action download the video from www.educational-computing.co.uk/CS/Book/Videos/TweetLang.avi.

A stream can be thought of as a series of connected operators. The initial set of operators (or single operator)
are typically referred to as source operators. These operators read the input stream and in turn send the data
downstream. The intermediate steps comprise various operators that perform specific actions. Finally, for every
way into the in-motion analytics platform, there are multiple ways out, and in streams these are outputs called sink
operators - see Figure 11.1.8. In a more technical sense, a stream is a graph of nodes connected by edges. Each
node is an operator.

Single licence - Abingdon School 551

11 Big Data

Figure 11.1.9 Spring XD

analytics showing
visualisation in real time of
tweets broken down by the
“lang” field and piped via
the following tap:
stream create tweetlang --definition “tap:stream:tweets > field-value-counter --fieldName=lang” --deploy true

Questions
11 Explain the meaning of the Big Data characteristic known as velocity.

Machine learning
Machine learning techniques are needed to discern patterns in data and
to extract useful information. Big Data datasets are analysed with machine
learning techniques which enable the value in the datasets to be extracted. This
can take the form of a predictive model that can then be used in the algorithm
that processes streaming data to extract the value from the data in the stream.
The value of data in large datasets was revealed in 2009 when a new flu virus,
H1N1 hit the streets. Public Health authorities feared a pandemic on the scale
of the 1918 Spanish flu that killed millions.
Doctors were requested to inform health authorities of new flu cases. However,
people might feel ill for days but wait before seeing a doctor. With other delays
in the reporting system and a rapidly spreading disease the authorities were
unable to get a clear picture.
About the same time, Google engineers had reported in the science publication
Nature how Google could “predict” the spread of the winter flu in the United
States, not just nationally, but down to specific areas of the States by what
552 Single licence - Abingdon School
 11.1 Big Data

people searched for on the Internet using the three billion search queries a day that they were receiving and saving.
This led to Google identifying the areas infected by the flu virus.
Google processed some 450 million different mathematical models in order to test search terms, comparing their
predictions against actual flu cases from 2007 and 2008. The result was their software found a combination of
45 search terms that, when used together in a mathematical model, strongly correlated their predictions with the
official historical figures. The result: Google could tell where the flu had spread in near real time, not a week or two
after the fact. Google’s result was built on Big Data. It is Big Data and machine learning applied to Big Data that
enables the value in data to be extracted and used to provide useful insights of significant value.
To understand how machine learning applied to a dataset or datasets can produce a predictive model, consider a
much simpler case of trying to find the relationship between growth rate of something of interest and temperature.
Figure 11.1.10 shows some data plotted on an x-y graph together with a best-fit straight line generated using linear
regression, a common machine learning technique. The programming language used was R, a strongly functional
programming language for statistical computing. The code to produce this graph and to generate the model and
plot the best-fit straight line is shown below.

R programming language
Information
R programming language:
> temperature <- c(10,20,30,40,50,60,70,80,90,100,110,120,130,140
R is a free software environment
,150)
for statistical computing and
> growthrate <- c(20,26,38,49,45,55,55,55,67,74,74,77,83,95,108) graphics.
R can be downloaded from
> plot(temperature, growthrate)
https://cran.rstudio.com/.
> model <- lm(growthrate~temperature) It can also be run online at
www.getdatajoy.com
> abline(model)

The model generated by the program was

growthrate = 0.5371 x temperature + 21
This model could be used as shown in Figure 11.1.11
in a coded algorithm applied to the data stream of
temperature data to predict growth rate and take action
if it was too slow.

Figure 11.1.10
Plot of growth rate
against temperature
generated in the
programming
language R

Single licence - Abingdon School 553

11 Big Data

Information
Alarm HDFS:
system Hadoop Distributed Filing System.

Data stream

Predictive Model → algorithm → code

growthrate = 0.5371 x temperature + 21

Machine Learning HDFS

Figure 11.1.11 Predictive model algorithm applied to stream data from a temperature sensor

Generating a predictive model is done with the data at rest. The model is then
applied to streaming data. A classic example is fraud detection on credit card
transactions - see http://www.fico.com/en/wp-content/secure_upload/FICO_
Real_Time_Fraud_3095IN.pdf.
Recommendation systems also fall into this category - see http://techblog.
netflix.com/2012/04/netflix-recommendations-beyond-5-stars.html.

Questions
12 Explain how machine learning is used in Big Data systems to leverage
value in stored datasets. Give real examples of two different types of
machine learning Big Data systems.

Programming tasks
1 Gain access to the R programming system and launch the GUI.
Create your own example dataset of temperature and growth rate.

2 Plot these with growth rate on the y-axis and temperature on the x-axis.
> plot(temperature, growthrate)

3 Create a linear regression model to fit a best-fit straight line to your

data, e.g. model <- lm(growthrate~temperature)

4 Plot the best-fit line from the given model using

> abline(model)

5 View this model with command model.

554 Single licence - Abingdon School

 11.1 Big Data
Section 2
Functional programming is a solution to the Big Data processing problem Learning objectives:
The functional programming paradigm (see Chapters 12.1.1 to 12.3.1) makes ■■Know that when data size
it easier to write correct and efficient distributed code because functional are so big as not to fit on a
programming languages support single server:
■■ immutable data structures
• the processing must be
■■ statelessness distributed across more
■■ higher-order functions than one machine

Using functional programming, a developer composes a program by assembling • functional programming

a series of functions. is a solution, because it
An immutable data structure is a data structure whose state cannot be modified makes it easier to write
after it is created. correct and efficient
There are no assignment statements in functional programming, so once distributed code.
something is given a value of 5, let’s say, it continues to have that value.
■■Know what features of
This is quite important because when we call a function, say one that squares its functional programming
input, we always expect square (3) to be 9, for example. If, however, we called makes it easier to write:
a different function, say, one that returned the state of the balance of a bank
account, we wouldn’t expect to get the same result everytime because bank
• correct code
account balances fluctuate. Whereas the square function operates in a stateless • code that can be distributed
manner, the function to discover a bank balance operates on a system which across more than one server
is not stateless. The result returned depends on both the input and the current
state of the system in a system with state. If there is state to be affected then it is Principle
possible to write functions that in addition to returning a result also alter state. Functional programming (FP)
This is called a side-effect of calling the function. If a function is labelled pure, is based on a simple premise:
Programs are constructed using
it has no side-effect. Functions in a pure functional programming languages are
only pure functions—in other
pure. words, functions that have no
Therefore programs written in a pure functional programming language consist side effects. A function has a
only of pure functions. This means the programs can be reasoned about side effect if it does something
other than simply return a
mathematically to check that they do what they are expected to do even when
result, e.g. update a counter
run in parallel, i.e. that they are correct. There is no state to consider as well. that the function relies on when
calculating the result to return.
Higher order functions compose, which means that a function can be passed
as an argument to another function, e.g.
reduce (map (list of things)) Questions
It follows that it is very easy to see which parts of 13 Explain what is meant by
code are independent and arrange for these to run (a) immutability (b) statelessness
in parallel. The independence is a consequence of (c) higher-order functions.
immutability and statelessness.
14 What features of functional programming make it
If it is possible to execute code in parallel on a easier to
different part of a dataset then we can scale the (a) write correct code
hardware horizontally, i.e add more servers if more (b) distribute code to run across more than one
data needs to be processed. The code can be just machine?
copied, distributed and run on the additional servers.
Single licence - Abingdon School 555
11 Big Data

Section 3
Fact-based model
Learning objectives:
The fact-based model is a conceptual model for modelling data. Physical
■■Be familiar with the: models based on the fact-based model are Bigtable used by Google, Cassandra
• fact-based model for used by Apple, Facebook, EBay, Twitter, Instagram. It models data in a
representing data completely different way from the relational model and it fits the Big Data
approach better than the relational model does. Each fact within a fact-based
• graph schema for capturing
model captures a single piece of information.
the structure of the dataset
Data is immutable
• nodes, edges and properties In the fact-based model data is immutable, i.e. it cannot be altered except to
in graph schema delete data which has been entered erroneously as a result of human error. This
is completely different from the relational model in which data is mutable, i.e.
Principle
Fact-based model:
can be overwritten with new values in an update operation. In the relational
• Raw data stored as atomic model, an erroneous entry caused by human error cannot be undone once
facts committed, i.e. it is not possible to go back to a previous value. Table 11.1.1
• Each fact captures a single shows a section of student data from a fact-based model system. As students
piece of information
move from year to year they move classes. This information is recorded as a
(i.e. atomic)
• Facts are kept immutable historical record, e.g. student 1 was assigned to class Year 11 D on 03/06/2014
and eternally true by using and then roughly a year later to class Year 12 B on 27/08/2015. The historical
timestamps record for 03/06/2104 is not removed because in the fact-based model, data is
• Each fact is made immutable.
identifiable so that query
processing can identify StudentId YearClass TimeStamp
duplicates (facts with same 1 Year 11 D 03/06/2014 10:35:16
identity) 2 Year 11 A 03/06/2014 10:38:05
• A nonce is used to make 3 03/06/2014 10:44:45
Year 11 D
identical facts identifiable (a
1 Year 12 B 27/08/2015 12:45:51
nonce is a randomly
generated 64-bit number) 2 Year 12 E 26/08/2015 11:15:31
Table 11.1.1 Student data for a fact-based model

If data had been modelled using the relational model then student 1’s data
Information would have been updated, and in the process the value Year 11 D would have
Immutability in the fact- been overwritten with the value Year 12 B.
based model:
The ability of a Big Data system to store vast quantities of data in a single
The immutability concept applied
to datasets in the fact-based dataset lends itself well to the fact-based model which requires a change to
model was borrowed from be recorded as a new fact and not applied as an update - the master dataset
functional programming. continually grows with the addition of immutable, timestamped data.
Human fault tolerance
People will make mistakes. The impact of such mistakes is minimised in the
fact-based model because it is an immutable data model, no data can be lost. If
bad data is written, earlier (good) data units still exist.
In comparison, in the relational model, a mistake can cause data to be lost
because values are overwritten in the database.
Data is true forever
The key consequence from immutability is that each piece of data is true in

556 Single licence - Abingdon School

 11.1 Big Data

perpetuity, i.e. forever: a piece of data, once true, must always be true. Key fact
Simplicity
Advantages over relational
Databases based on the relational model rely on indexes to retrieve and update
model:
data. These must be built and subsequently modified as data is updated/altered/ • Simplicity - no indexing
deleted. • Append - new data units just
added to end of dataset
In the fact-based model, data is immutable so all that is required is the ability
• Perpetuity - data is true for
to append new data units to the master dataset. This does not require an index ever
for the data. This is a significant simplification. • Human fault-tolerant -
Atomic facts immutable facts means
human errors can be
In the fact-based model, data is broken down into fundamental units that are
corrected by returning to
called facts. A fact, for example, is student 1 is in class Year 10 D. Another fact
earlier “good” facts
is, student 1 is in class Year 12 B. What differentiates these two facts in time are • Historical queries - queries
their timestamps, 03/06/2014 10:35:16 and 27/08/2015 12:45:51. can be historical because
facts immutable
A fact possesses two core properties:
• Partial information - null
■■ it is atomic value not required, simply
omit missing information
■■ it is timestamped.
Facts are atomic because they can’t be subdivided further into anything
Key fact
meaningful.
Fact-based model:
Collective data such as the classes for a student and their subjects studied are
With a fact-based model, the
represented as multiple, independent facts as shown in Figure 11.1.12. master dataset will be an ever-
growing list of immutable,
1 in Year 10 D atomic facts.
(02/06/2013 09:05:26)
Big data techniques are required
1 in Year 11 D 1 in Year 9 A in order to accommodate the
(03/06/2014 10:35:16) (04/06/2013 13:43:05) growing list of immutable atomic
facts, i.e. more than one server
will typically be required for the
master dataset.
1 in Year 12 B Raw data about 1 studies maths
(27/08/2015 12:45:51) student 1 (02/06/2013 09:33:03)

Concept
1 studies Physics Atomic fact:
(02/06/2013 09:38:29) 1 studies English Facts are atomic because they
(02/06/2013 09:35:13)
can’t be subdivided further into
anything meaningful.
1 studies CS
(02/06/2013 09:36:18)

Figure 11.1.12 Collection of independent facts for student 1

As a consequence of being atomic, there is no redundancy of information Principle
Core properties of facts:
across distinct facts. The timestamps make each fact immutable and eternally
• atomic
true. Also, storing atomic facts makes it easy to handle partial information • timestamped.
about an entity without introducing NULL values by simply leaving out the
missing information.

Single licence - Abingdon School 557

11 Big Data

Principle Graph schemas

Facts in graph schema: Let’s suppose that in addition to storing facts about students our dataset also
Each fact modelled in the graph stores facts about which intranet web pages students visit and for each page
schema represents either a piece the total visits from all students. Each fact for this dataset represents either a
of information about an entity, piece of information about a student or a relationship between a student and an
e.g. entity student has surname
intranet web page. A graph schema can be used to describe this scenario.
Alex, or a relationship between
entities, e.g. student 1 visits A graph schema captures the structure of a dataset stored using the fact-based
intranet page model. Structure means a description of the types of facts contained in the
http://192.168.0.32/physics
dataset, e.g. student has surname of data type string, and a description of the
student Alex is friends with
student Kevin. relationships between the entities. In Figure 11.1.13 student and intranet
page are entities because students visit intranet pages.

Firstname: Surname:
Student (StudentId): Kevin Bond
2

PageView Class:
Nodes are the entities Student (StudentId) Year 11 D

in the system. 1
PageView
Figure 11.1.13 The graph
Page(url): PageView Class:
schema for students and http://192.168.0.32/physics Year 12 B
intranet page access.
There are two node types: Timestamps are not
Page(url):
students and pages. http://192.168.0.32
shown to avoid cluttering
Student nodes have Total views: the diagram.
150
properties Firstname,
Properties are information about
Surname and Class. Page Edges are the
entities, e.g. student or page.
nodes have one property, relationships between
Total views:
Totalviews nodes. 50

Concept The graph in Figure 11.1.13 represents the facts for the student and intranet
Graph schema: page access dataset. The graph highlights the three core components of a graph
A graph schema captures the schema—nodes, edges, and properties:
structure of a dataset stored
■■ Nodes are the entities in the system. In this example, the nodes are
using the fact-based model.
Structure means a description students and pages.
of the types of facts contained ■■ Edges are relationships between nodes. In this example an edge
in the dataset, e.g. student
between a student and a page represents the relationship between a
has surname of data type
string, and a description of the page and a user who has visited this page.
relationships between entities, ■■ Properties are information about entities. In this example, surname,
e.g. student visits intranet page,
firstname, class, total views.
student and intranet page are
entities Edges are strictly between nodes and are shown visually connected
by a solid line. Dashed lines are used for connecting properties to the
corresponding node.

558 Single licence - Abingdon School

 11.1 Big Data

In the example an intranet page is able to identify a student by StudentId, e.g. Information
student 1. A relationship labelled “PageView” exists between a student and a page. Graph schema:
Another way that an intranet page could identify students is by using a cookie. The graph schema provides a
complete description of all the
The intranet site places a cookie on the student’s computer and when a student
data contained within a dataset.
returns to this intranet site, the student is identified by the stored cookie, e.g.
cookie abs123. At some point, the intranet site might be able to associate
student 1 with this cookie if it is stored on student 1’s computer. A node labelled Principle
Student(CookieId) abs123 could then be connected to node labelled Student Nodes, edges, properties:

(StudentId) 1 by the relationship labelled “Equivalent”. The three core components of a

graph schema are nodes, edges,
One of the attractive features of the fact-based model is that it is very easy to and properties.
add new types of information to the schema just by defining new node, edge,
and property types. Existing fact types are unaffected because facts are atomic.

Tasks
2 Research Google’s data storage system, Bigtable (http://static.googleusercontent.com/media/research.google.
com/en//archive/bigtable-osdi06.pdf ) and Google’s Cloud Bigtable (https://cloud.google.com/bigtable/).

3 Research Amazon’s Cloud Search system (http://aws.amazon.com/cloudsearch/).

Questions
15 Describe the fact-based model.

16 What are the two core properties that a fact possesses?

17 What is meant by atomic fact?

18 What does each fact modelled in a graph schema represent?

19 In the fact-based model, what is (a) a node (b) an edge (c) a property?

20 Complete the graph schema in Figure 11.1.14 to represent Location : Name : Alex
the following facts: Reading
(a) Alex, Kevin, and Sue are friends Age : 23
(b) Alex’s age is 23, Kevin’s is 24 and Sue’s is 22
(c) Alex lives in Reading, Kevin lives in Aylesbury, Sue lives in Bristol
Person ID:1
21 Alex and Sue each post a message on an electronic message board.
Extend your graph schema to include these facts.
What use would be made of timestamps in this graph schema? Figure 11.1.14
22 Why does the fact-based model require Big Data techniques?

23 Give three advantages of the fact-based model over the relational model.

Single licence - Abingdon School 559

11 Big Data

In this chapter you have covered:

■■ ‘Big Data’ is a catch-all term for data that can’t be processed or analysed
using traditional processes or tools because it is
• too big to fit into a single server
• too heterogeneous
• its production can occur at high rates.
■■ Big Data is described in terms of
• volume - data to be analysed is too big to fit into a single server
• velocity - refers to whether the data is in motion and with what
frequency or at rest. Data in motion is data that is streamed at some
frequency continuously, e.g. 1000 events per second. Data at rest is
data that has been stored on some permanent data storage device.
• variety - data in many forms such as structured, unstructured, text,
multimedia. The data is said to be heterogeneous.
■■ When dataset size is so big as not to fit on a single server:
• the processing must be distributed across more than one machine
• functional programming is a solution, because it makes it easier to
write correct and efficient distributed code.
■■ The features of functional programming that make it easier to write:
• correct code - immutable data structures, statelessness: pure functions
• code that can be distributed across more than one server - immutable
data structures, statelessness, higher-order functions.
■■ The fact-based model for representing data is a conceptual model for
modelling data - data is broken down into fundamental units that are
called facts. Each fact captures a single piece of information. Data is
immutable, human fault-tolerant and true forever. Data is modelled in a
completely different way from the relational model. Better fit with Big
Data approach than relational model.
■■ Graph schema - captures the structure of a dataset stored using the
fact-based model. Structure:
• the types of facts contained in the dataset
• the relationships between the entities.
■■ Graph schema drawn in a graph-like structure with
• Nodes - the entities in the system
• Edges - the relationships between nodes
• Properties - information about entities.

560 Single licence - Abingdon School

12 Fundamentals of functional programming
12.1 Functional programming paradigm
Learning objectives:
■■Function as process ■■ 12.1.1 Function type
■■Function as object What is a function?
■■Function, f, has a function Loosely speaking, a function is a rule that, for each element in some set A of
type, f : A → B where the type inputs, assigns an output chosen from set B but without necessarily using every
is A → B. member of B.

■■A is the argument type, For example, the function f

and B is the result type. f : {0,1,2,3} → {0,1,2,3,4,5,6,7,8,9}

■■The set A is called the domain maps 0 to 0, 1 to 1, 2 to 4 and 3 to 9 when the rule is: output the square of the
and the set B is called the input.
co-domain.
Function as process
■■The domain and co-domain In function as process, a function is a rule that tells us how to transform some
are always subsets of objects in information into some other information, e.g. the integer 2 into its square 4.
some data type.
Function as object
Key principle In function as object, the function is a thing in its own right.
Function as process: For example, a pencil sharpener is an object. If the focus of attention is a pencil
A function is a rule that tells then the pencil sharpener just represents a process - sharpening pencils, input:
us how to transform some
unsharpened pencil; output: sharpened pencil.
information into some other
information.
In the function as process view, we are applying the function sharpen to
Function as object: pencils; it’s the pencil that counts. But we can also think about the pencil
The function is a thing in its sharpener as a thing in its own right, when we empty it of pencil shavings, or
own right. worry about whether its blade is sharp enough. This is the function as object
view.

Questions
A function f
f : {0,1,2,3} → {0,1,2,3, ..., 25, 26, 27}
maps 0 to 0, 1 to 1, 2 to 8, 3 to 27.
1 What is the rule?
A function f
f : {0,1,2,3} → {0,1,2,3, 4, 5, 6}
maps 0 to 0, 1 to 2, 2 to 4, 3 to 6.
2 What is the rule?

Single licence - Abingdon School 561

12 Fundamentals of functional programming

Questions
3 For each of the following what is the function as process and
what is the function as object?
(a) A single sheet of A4 paper containing text is placed in
the machine whose action is to produce a
printed copy of the sheet.
(b) A kitchen tool is used to remove skin from
potatoes.
Key principle
What is a function type?
Function type:
Just as data values (e.g. 6, 9.1, True) have types (integer, real, Boolean
A function f which takes an
argument of type A and returns respectively) so do functions. Function types are important because they state
a result of type B has a function what type of argument a function requires and what type of result it will return.
type which is written
A→B
A function f which takes an argument of type A and returns a result of type B
has a function type which is written
A →B
To state that f has this type, we write
f:A→B
For example,
1) squareroot : real → real
2) square : integer → integer

The function named squareroot applied to an argument of data type real

produces a result of data type real, e.g.

squareroot (4.0) → 2.0

The function named square applied to an argument of data type integer

produces a result of data type integer, e.g.

square (2) → 4

562 Single licence - Abingdon School

 12.1.1 Function type

Domain and co-domain Key concept

If f : A → B is a function from A to B we call the set A, the domain of f, and
the set B the co-domain of f. The domain and co-domain are always subsets Domain and co-domain:
If f : A → B is a function from
of objects in some data type. For example, if A is a subset of domain data
A to B, we call the set A, the
type integer then its values might be 0, 1, 2, 3, ..., 149, 150. Often it is just domain of f, and the set B the
convenient to use the data type directly, co-domain of f.

square : integer → integer

The function square then has an argument type, integer and a result type, integer
even though in practice a subset of integers only will be used.

Practical Activity
Use a text editor such as NotePad++ to write Haskell programs. Save these
Haskell programs using extension .hs.
Figure 12.1.1.1 shows NotePad++ being used to create a function named
square with one parameter x of data type Integer and a body x∗x. This file has
been saved with filename square.hs in folder c:\book\haskell.

Figure 12.1.1.1 NotePad++ editor showing square.hs

The :: operator (read as has type) is used in Haskell to express what type an
expression has.
Integer is the type of mathematical integers (int could have been used and is
the type of integers that fit into a word on the computer - this will vary from
computer to computer).
Launch WinGHci if you are using a machine running the Windows operating
system (ghci on Linux-based machines). The WinGHci window is shown in
Figure 12.1.1.2.

Single licence - Abingdon School 563

12 Fundamentals of functional programming

Figure 12.1.1.2 WinGHCi showing square.hs loaded, compiled and run

At the Prelude prompt (Prelude>) type the command to change to a
specified folder.
:cd c:\book\haskell followed by <return>.

Commands begin with a colon, i.e. :

Now load the file containing the program defining the function square.
At the Prelude prompt type
:load square.hs followed by <return>.

WinGHCi will perform a compilation of a module called Main in order to

run square.hs interactively.
If there are no errors loading and compiling the Prelude prompt will be
replaced by the prompt *Main.
At the *Main prompt, type
square 4 followed by <return>.

The correct answer, 16, is displayed.

To return to the Prelude prompt, type :module or :m

In this chapter you have covered:

■ Function as process
■ Function as object
■ Function, f, has a function type, f : A → B where the type is A → B.
■ A is the argument type, and B is the result type.
■ A is called the domain and B is called the co-domain.
■ The domain and co-domain are always subsets of objects in some
data type.

564 Single licence - Abingdon School

12 Fundamentals of functional programming
12.1 Functional programming paradigm
Learning objectives:
■ Know that a function is a
first-class object in functional
■ 12.1.2 First-class object
programming languages and First-class objects (or values)
in imperative languages that First-class objects (or values) are objects which may
support such objects. • appear in expressions (expressions such as 5 + 3 x y3)

■ Know that a function can be • be assigned to a variable

an argument to another • be assigned as arguments to functions
function as well as the result of
• be returned by function calls
a function call
Typical first-class objects in many programming languages are integers,
floating-point values, characters and strings.
Key principle For example,

First-class object: x := 5 is an assignment statement in which the first-class value

First-class objects (or values) are 5 is assigned to the variable x.
objects which may:
MyStringVar := Uppercase('Hello World!')
• appear in expressions
• be assigned to a variable is an assignment statement containing a function call Uppercase
• be assigned as arguments in with argument
function calls
'Hello World!'
• be returned as a function
call result. which is a first-class value of type string. This function returns a first-class
value of type string
‘HELLO WORLD!’
Functions as first-class objects
In functional programming languages and in some imperative programming
languages a function is a first-class object. This means that it can
• appear in expressions
• be assigned to a variable
Figure 12.1.2.1 function as
• be passed as an argument to another function
argument
• be returned as the result of a function call.

Information For example, in Python a function may be created and assigned to a variable v
Alternatively, using the keyword lambda as follows x**2 means x2
v = lambda x: x**2
def square(x):
return x*x It may be called in a print procedure to square the value 5 as follows
v = square print (v(5))
print(v(5))
Figure 12.1.2.1 shows the above coded and executed in IPython using Python
3.4.1.

Single licence - Abingdon School 565

12 Fundamentals of functional programming

The value 25 is returned by the call v(5) and the procedure print then
outputs this value to the screen.
The function is defined with
lambda x: x**2
This is an anonymous function with one argument x and a body x**2. It
is a first-class object because it can be assigned to a variable and passed as an
argument to another function/procedure.
The following Python code defines a function exp with one argument n that
returns a function also with one argument x and a body x**n.
def exp (n): return lambda x: x**n

exp(3) returns a function with body x**3

exp(10) returns a function with body x**10

Figure 12.1.2.2 Another function

as argument f = exp(3)

print (f(4))

If exp(3)is assigned to f then f references a function with body x**3, i.e if

f is called, it will return a function with body x**3.

For example, if f is called with argument 4 then the function with body x**3
is returned with x replaced by 4. The function call f(4) will therefore return
43 which is 64. We can then display this value on a screen with the print
procedure.

Questions
1 What requirements must be satisfied for an object or value to be
classified as first class?

2 A function f takes one argument n and returns a function with body

xn. f is a first-class object. It is assigned to variable g as follows
g = f(10)
What is returned by the call
(a) g(2) (b) g(3) (c) g(5)

566 Single licence - Abingdon School

 12.1.2 First-class object

Functions as arguments
Functions as arguments is particularly useful when a common underlying
pattern can be identified. For example, suppose we need to double the result of
squaring, cubing, etc. Although trivial, it does illustrate how this generalisation
could be used for more complicated cases.
We first define a function double which takes a single argument
afunction. The argument afunction is a function type. The function
double returns 2*afunction.
Key principle
In Python 3.4, double would be defined as follows
def double (afunction) Functions as arguments:
return 2*afunction Functions as arguments is
particularly useful when a
We can now define maths functions square, cube and so on in Python 3 as common underlying pattern
follows can be identified. The common
def square(x): programming patterns that
recur in code, but which are
return x*x
used with a number of different
def cube(x) functions can be abstracted and
return x*x*x then given a general name, e.g.
sum where summing might be
Now we can use double, square and cube as follows a sequence of natural numbers,
double(square(4)) numbers squared, numbers
cubed, terms in an infinite
The function call double(square(4))returns 32. series, etc.
double(cube(4))
The function call double(cube(4))returns 128.

In Haskell, for the same problem, we have an argument to function double:

afunction x of type (Integer -> Integer)
afunction x (2*afunction) x
When function double is called it returns
a function of type
(Integer -> Integer)
with body
(2*afunction) x

Figure 12.1.2.3 Haskell version of Python code

double(square) is applied to
argument 5 to produce 50.

Figure 12.1.2.4 Execution in WinGHCi

Single licence - Abingdon School 567
12 Fundamentals of functional programming

Programming task
1 Using a text editor such as NotePad++ enter the following Haskell
program and save as summy.hs.

summy :: (Integer -> Integer) -> (Integer -> Integer)

summy afunction 0 = 0

summy afunction x = (afunction x) + summy afunction (x - 1)

square :: Integer -> Integer

square n = n*n

cube :: Integer -> Integer

cube n = n*n*n

identity :: Integer -> Integer

identity n = n

Launch WinGHCi and at the Prelude prompt type

:load summy.hs (your working directory must contain summy.hs).

At the *Main prompt try the following

(a) summy(square) 5
(b) summy(cube) 5
(c) summy(identity) 5
Now
(d) Explain how summy in each case arrives at the observed
output.

In this chapter you have covered:

■■ What is meant by first-class object or value
■■ Functions as first-class objects in functional programming languages
and in imperative programming languages that support such objects
■■ A function as an argument to another function
■■ Returning a function as a function call result

568 Single licence - Abingdon School

12 Fundamentals of functional programming
12.1 Functional programming paradigm
Learning objectives:
■■Know that function
application means a function
■■ 12.1.3 Function application
applied to its arguments. Computations
The arithmetic expression 3 × 4 + 2 represents a computation which can be
described with reductions as follows
3 × 4 + 2 → 12 + 2 → 14

The given expression describes a single computation.

Formula
A formula is an expression containing variables. It represents a whole class of
computations, since the variable(s) can assume different values.
For example, the formula corresponding to the computation above
a x b + c
uses the variables a, b, c.
Substituting integers for the variables, a, b, c, produces an expression
whose value can be computed. For example, a = 3, b = 4, c = 2
3 × 4 + 2 → 12 + 2 → 14

The formula a x b + c describes one computation for each possible

combination of values for a, b, c.
Since there are an infinite number of integers, the formula represents an infinite
number of computations.

Procedural and Functional Abstraction

The formula a x b + c is an abstraction because it omits the actual numbers
to be used in the computation. The formula a x b + c represents a
computation method, a procedure. Such an abstraction is called a procedural
abstraction, since the result of the abstraction is a procedure, a method.
In general, there are many methods for obtaining a desired result.
The result of a procedural abstraction is a procedure, not a function. To get
a function another abstraction, which disregards the particular computation
method, must be performed.
The result of this abstraction is a function and the abstraction is called
functional abstraction. The focus is then on the input(s) and the output.

Single licence - Abingdon School 569

12 Fundamentals of functional programming

For example, suppose we wished to calculate the sum of the first n natural
numbers. We have two choices of method or formula for calculating this sum:
n × (n + 1)
2
and
1 + 2 + 3 + 4 + ….+ (n – 1) + n

The above two methods are examples of procedural abstraction.

The black box in Figure 12.1.3.1 hides the particular method used to calculate
the sum for a particular n and so is an example of functional abstraction.
n sum

Figure 12.1.3.1 Calculation of sum of first n natural numbers

All that a user needs to know are the number and order of the inputs and
the name of the function in order to be able to apply the function to these
inputs.

Parameters
The formula used to define a function is sometimes called the body of the
function. The name used for the quantity that can vary is called the parameter
of the function. Therefore in the example above n is the parameter.
n × (n + 1)
The function associated with the formula is given a name, e.g.
2
sum.
To use the sum function, we apply it to an argument.
For instance, to find the sum of the first 6 natural numbers, the function sum
is applied to the argument 6.
Thus the parameter, n, is the name used in the function body to refer to the
argument, 6.
To compute the value of the function for some argument, replace the parameter
in the body of the function by the argument and compute the expression.
For example,
n × (n + 1)
n=6
2

6 × (6 + 1)
=
2

= 21

570 Single licence - Abingdon School

 12.1.3 Function application

Function Application Key principle

Note that the name given to a function does not involve the parameter. The
name of the function above is sum not sum(n) or sum n which is the Function application:
Function application is the
notation for sum applied to n. We call a function applied to its argument(s) a
process of giving particular
function application. A function application is when the function is applied inputs to a function, e.g.
to a particular argument, e.g. 6 in sum(6) or sum 6. add (3, 4) represents the
application of the function add
A function add takes two integer arguments and returns their sum. In Haskell
to integer arguments 3 and 4.
this function could be written as follows:
add :: Integer -> Integer -> Integer
add x y = x + y Information
Application of this function to arguments 3 and 4 would be written as follows Function type:
Function add has function type
add: integer x integer → integer
which is expressed in Haskell as
add :: Integer -> Integer

-> Integer
See Unit 1 Chapter 4.2.2 for
an explanation of the Cartesian
product integer x integer,

Questions
1 In Haskell a function square is defined as follows
square :: Integer -> Integer
square n = n*n
What is the result of the function application
(a) square 3 (b) square 5?

2 What is the parameter in the function square defined in question 1?

In this chapter you have covered: 3 What are the values of the arguments used in question 1?
■ What is meant by
• a computation 4 What is meant by function application?
• a formula
• procedural and functional abstraction
• parameter of a function
• argument of a function
■ Function application is when a function is applied to a particular
argument, e.g. 6 in Sum(6) or Sum 6.

Single licence - Abingdon School 571

12 Fundamentals of functional programming
12.1 Functional programming paradigm
Learning objectives:
■■Know what is meant by partial
function application for two
■■ 12.1.4 Partial function application
and three argument functions. Some “partially applicable” devices
A radio can be viewed as a partially applicable function. Its main function is to
■■Be able to use the notations transform electromagnetic(radio) waves into sound waves, but before listening
function: data type → (data type to a radio broadcast a radio station is selected by frequency, e.g. 101 MHz.
→ data type)
Hence, the radio first takes a radio station frequency as argument and then the
which is equivalent to
radio waves to be transformed.
function: data type → data type
→ data type In itself, the radio is quite a general function with many different broadcast
for example, frequencies to choose from. By selecting a specific broadcast frequency,
add: integer → (integer → the radio’s function is specialised into one dedicated to transforming
integer) electromagnetic waves for this particular frequency. This is still a function.
which is equivalent to General function:
add: integer → integer → Station frequency
Radio Sound waves
integer Electromagnetic waves
■■Know that for an example such as
Specialised function:
function add
Radio on
add: integer → (integer → Electromagnetic waves Sound waves
101 MHz
integer)
means add can take a single In conclusion, for this example, a partially applicable function is a function that
argument of type integer and given its first argument returns a new, more specialised, function. If you supply
return a more specialised this new function with an argument, you get the final result.
function (integer → integer)
This is what is meant by partial function application.
which takes a single integer
In this example, a function moves from being a two-argument function to
argument and returns an integer.
being a one-argument, more-specialised function. The function application of
the original function is applied to just one of its arguments but not both.
In general, in partial function application, the function application of the
Information
original function is applied to just some of its arguments but not all.
Remember:
The example shown in Figure 12.1.4.1 illustrates how a more -specialised
Functions are not partial, it is
function add2 with one argument is created in Haskell from an addition
their application which is partial.
function, add, with two arguments.

Single licence - Abingdon School 572

12 Fundamentals of functional programming

Three input device

If the radio has more than one band ( e.g. FM and AM), then it will have three
inputs:
General function:
Frequency band
Station frequency Radio Sound waves
Electromagnetic waves

First, a band is selected, e.g. FM, creating a two argument function.

Specialised function:

Station frequency
Electromagnetic waves FM Radio Sound waves

Then, a station frequency is selected to produce an even more specialised

function:
FM Radio on
Electromagnetic waves Sound waves
101 MHz
Partial application for two-argument functions
Consider the function add which adds together two integer arguments and
Figure 12.1.4.1 Example of returns an integer result.
partial function application for a add x y = x + y
two-argument function add x y This function can be viewed as a box, with two input arrows and an output
arrow.
x
add
y
This function takes a pair of integers and returns an integer, i.e.
Integer -> Integer -> Integer

Input Output
If we apply the function to two arguments, the result is a number; so that, for
instance, add 2 3 equals 5.
2 add 5
3
What happens if add is applied to one argument 2? Pictorially we have
2
y add

What is the output of this function application to one argument?

It is actually another function of a more specialised nature, this time a function
add2 with one argument y.
y add2
This new function will return 2 + y.
Figure 12.1.4.2 Shows bracket So for function application add2 3 the value 5 is returned.
notation Function add2 maps an integer to an integer, i.e. in Haskell
Integer -> (Integer -> Integer) Integer -> Integer

Input Output
573 Single licence - Abingdon School
 12.1.4 Partial function application

The function add2 is a function. It is function add partially applied to one Key principle
argument, the argument that substitutes for input x, i.e. in Haskell
Partial function application:
Integer -> (Integer -> Integer) Any function taking two
Input Output or more arguments can be
partially applied to one or more
We feed in the particular value, 2, for input x and get as output a function arguments.
add2 which takes an integer and returns an integer.
The function add2 is denoted by output (Integer -> Integer)
y output
as shown in Figure 12.1.4.2.

This is an example of a general principle: any function taking two or more Information
arguments can be partially applied to one or more arguments. Partial function application makes
It also means that sense for two or more argument

Integer -> Integer -> Integer functions but not for one
argument functions.
is equivalent to
Integer -> (Integer -> Integer)

as shown in Figure 12.1.4.1 and Figure 12.1.4.2.

Questions
1 A computer can be considered a “partially applicable” device.
A computer is used to calculate some output from a given input, but before performing
the calculation, it is provided with a program – its first argument.
program output
computer
input
Redraw the figure to show the computer as a “partially applicable” device executing a
program with the name prog1.

2 A function multiply calculates x ∗ y for integer arguments.

x
multiply
y

Redraw the figure to show this function partially applied to the argument 10.

3 In Haskell the partially applied multiply function in question 2 has the function type
Integer -> (Integer -> Integer)

Give the meaning of this notation.

4 Explain why a radio can be viewed as supporting partial function application.

5 Explain why function type
Integer -> Integer -> Integer

can be interpreted as function type

Integer -> (Integer -> Integer)

Single licence - Abingdon School 574

12 Fundamentals of functional programming

Partial application for three-argument functions

Now consider the function addxyz which adds together three integer
arguments and returns an integer result.
add x y z = x + y + z
This function can be viewed as a box, with three inputs and an output.
x
y addxyz
z
This takes three integers and returns an integer, i.e. in Haskell
Integer -> Integer -> Integer -> Integer

Input Output
If we apply the function to three arguments, the result is a number; so that, for
instance, addxyz 2 3 4 equals 9.
2
3 addxyz 9
4
What happens if addxyz is partially applied to one argument 2? Pictorially we
2 have
2
Integer -> (Integer -> Integer -> Integer) y addxyz
Input Output z

This function application computes 2 + y + z.

If we treat 2 + y + z as a new function it represents a function add2yz,
with two inputs, y, z.
The function addxyz partially applied to argument x = 2 returns a function
add2yz of function type Integer -> Integer -> Integer

Input Output
y add2yz
z
Function add2yz is thus a two argument function.
What happens if addxyz is partially applied to two arguments 2 and 3?
2 3
Integer -> Integer -> (Integer -> Integer) 2
3 addxyz
z
Input Output
This function application computes 2 + 3 + z. If we treat 2 + 3 + z
as a new function it represents a function add23z, with one input z. The
function addxyz applied to arguments x = 2, y = 3 returns a function
add23z of function type Integer -> Integer

z add23z Input Output

For example, if argument z = 4,

function application add23z 4 = 2 + 3 + 4 = 9

575 Single licence - Abingdon School

 12.1.4 Partial function application

Figure 12.1.4.3 Shows partial function application for three-argument function addxyz

Programming task
1 Using a text editor such as NotePad++ write a Haskell program for functions multiplyxy which
takes two integer arguments and returns x times y. Function multiplyxy has function type
Integer -> Integer -> Integer

Using the existing definition of multiplyxy, add the function multiply2y which takes one integer
argument y and returns 2 times y.
Function multiply2y has function type
Integer -> Integer
Save as multiplyxy.hs.
Launch WinGHCi and at the Prelude prompt type
:load multiplyxy.hs (your working directory must contain multiplyxy.hs).
Test function multiplyxy with x = 2 and y = 3
Test function multiply2y with y = 3

Single licence - Abingdon School 576

12 Fundamentals of functional programming

Questions
6 A function multiplyxyz takes three integer arguments x, y, z and
returns x ∗ y ∗ z. The function is applied to arguments 2 and 3 as
shown below.
2
3 multiplyxyz
z

Which of the following represent(s) the type of multiplyxyz with

x = 2 and y = 3?
1. Integer -> Integer -> (Integer -> Integer)
2. Integer -> (Integer -> Integer -> Integer)
3. Integer -> Integer -> Integer -> Integer
4. Integer -> (Integer -> Integer)

7 The function multiplyxyz is applied to argument 2 as shown below.

2
y multiplyxyz
z
Which of the following represent(s) the type of multiplyxyz with
x = 2?
1. Integer -> (Integer -> Integer -> Integer)
2. Integer -> (Integer -> (Integer -> Integer))
3. Integer -> Integer -> Integer -> Integer
4. Integer -> (Integer -> Integer)
5. Integer -> Integer -> Integer

In this chapter you have covered:

■■ Partial function application means any function taking two or more
arguments can be partially applied to one or more arguments.
■■ The use of notations
function: data type → (data type → data type)
which is equivalent to
function: data type → data type → data type
for example,
add: integer → (integer → integer)
which is equivalent to
add: integer → integer → integer
■■ add: integer → (integer → integer) means add can take a single argument
of type integer and return a more specialised function (integer → integer)
which takes a single integer argument and returns an integer.

577 Single licence - Abingdon School

12 Fundamentals of functional programming
12.1 Functional programming paradigm
Learning objectives:
■ Know what is meant by
composition of functions
■ 12.1.5 Composition of functions
Function composition
Key principle
The operation function composition combines
Function composition: domain of f is A
two functions to get a new function.
Function composition or
Given two functions f: A → B co-domain of f is B
functional composition
g: B → C domain of g is B
combines two functions to get a
new function. function g ∘ f, called function composition of g co-domain of g is C
and f, is a function whose domain (See Chapter 12.1.1)
is A and co-domain is C.
g ∘ f: A → C
Note that the co-domain of f must be the same as the domain
of g.
For example,
suppose f (x) = x + 3
and g (y) = y2
then g ∘ f = g (f(x)) = g (x + 3) = (x + 3)2
Function g is applied to the result of applying function f, i.e.
Figure 12.1.5.2 Defining g is applied to x + 3. Since function g squares its argument, g
functions f and g applied to x + 3 is (x + 3)2.
Applying the composition g ∘ f to argument 4, we get f applied
to 4 first then g applied to the result as follows
g ∘ f 4 = (4 + 3)2 = 72 = 49
Function composition is one of the simplest ways of structuring a
program. In function composition a number of things are done one
after another. This allows a separation of concerns so that each part
can be designed and implemented independently.
In Haskell the function composition operator is '.' In function
composition the operator is placed between the two functions of the
composition. The operator has the effect of passing the output of one
function as input to the other function as shown in Figure 12.1.5.1.
g. f
x y g z
f

Figure 12.1.5.3 Application of Figure 12.1.5.1 Function composition for functions f and g, g.f
the function composition Figure 12.1.5.3 shows (g.f) applied to argument 4 two ways. Figure
(g.f) to argument 4 12.1.5.2 shows g and f being defined.

Single licence - Abingdon School 578

12 Fundamentals of functional programming

Sometimes it is not possible to use Haskell’s function composition operator.

For example, if we wanted to find the sum of the squares of two integers, say 4
and 5 then we would use the expression
sum (square 4) (square 5)

This is still function composition.

Programming tasks
1 Using a text editor such as Notepad++, enter the following Haskell
code and save as doublesquare.hs
square :: Integer -> Integer
square x = x∗x
double :: Integer -> Integer
double w = 2∗w
Launch WinGHCi then at the Prelude prompt type
:load doublesquare.hs
At the *Main prompt write function compositions using the function
composition operator '.' to
(a) double the square of 6
(b) square the result of doubling 6
(c) repeat (a) and (b) without using the function composition operator.

2 Define a function add which takes two integer arguments and returns
their sum. Incorporate this definition into doublesquare.hs and save as
addsquare.hs.
Launch WinGHCi then at the Prelude prompt type
:load addsquare.hs
At the *Main prompt write a function composition which adds the
squares of 2 and 3. Note you will not be able to use the function
composition operator for this.

3 Change the argument type from Integer to Float in both functions add
and square. Save as addsquarefloat.hs.
Launch WinGHCi then at the Prelude prompt type
:load addsquarefloat.hs
Haskell has a built-in function sqrt which takes one argument of type
Float and returns a Float.
At the *Main prompt write a function composition which calculates the
square root of the sum of the squares of 3 and 4. Note you will not be
able to use the function composition operator for this.

579 Single licence - Abingdon School

 12.1.5 Composition of functions

Questions
1 What is function composition?

2 Function f applied to x and function g applied to y are defined as follows

f(x) = x - 3
g(y) = y3

The domains and co-domains of f and g are ℕ

What is g ∘ f applied to x?

3 Two functions f and g have the following function types

f: A → B
g: C → D
Explain why it is not possible to have the function composition g ∘ f.

4 In Haskell, a function add has the following function type

add :: Integer -> Integer -> Integer
A function square has the following function type
cube :: Integer -> Integer
Function add is defined as follows
add x y = x + y
Function cube is defined as follows
cube z = z*z*z
What is the result of the function composition
add (cube 3) (cube 4)?

In this chapter you have covered:

■■ The meaning of function composition:
• function composition combines two functions to get a new function
• function g ∘ f is called function composition of g and f
• g ∘ f is a function whose domain is A and co-domain is C if f: A → B
g: B → C
• f is applied first and then g is applied to the result

Single licence - Abingdon School 580

12 Fundamentals of functional programming
12.2 Writing functional programs
Learning objectives:
■ Have experience of ■ 12.2.1 Functional language programs
constructing simple programs
Haskell
in a functional programming WinGHCi and GHCi
language. After installing Haskell (www.haskell.
■ Use higher-order functions. org), launch WinGHCi (assuming the
■ Have experience of using the installation is for Microsoft® Windows®)
following in a functional - see Figure 12.2.1.1. A prompt with
programming language. Prelude> at the beginning should appear Figure 12.2.1.1 Start menu
in the window - see Figure 12.2.1.2. This tells
• map
you that the interpreter is ready to accept commands, and that the only loaded
• filter
module at this moment is Prelude, which contains the most basic functions
• reduce or fold and data types in Haskell. The application is running an instance of a Read
Eval Print Loop (REPL), or in other words an interpreter.
In WinGHCi, you input an expression and press the Enter key. The
expression gets evaluated, and the result is shown in the screen.
For example, 7 + 2 as shown in Figure 12.2.1.3.
Haskell can also evaluate expressions that WinGHCi Prelude> 7 + 2
contain rational numbers, 9
Figure 12.2.1.2 Initial screen
Figure 12.2.1.3 REPL in
e.g. Prelude> 1/2 + 1/3
action
0.8333333333333333
Haskell also has a command line application called GHCi. If you launch this you may look up the standard
functions built in
to the language. For
example, typing s into
GHCi at the Prelude
prompt and pressing
the Tab key, causes
a list of all possible
functions beginning
with the letter s to
appear as shown in
Figure 12.2.1.4.

Figure 12.2.1.4 GHCi command line interpreter

In WinGHCi , to see all the definitions in a module: Prelude> :browse Prelude

Single licence - Abingdon School 581

12 Fundamentals of functional programming

If you then type q and press Tab again, only one possibility is left, sqrt, which
is automatically displayed ready to be used. To find the square root of 2, just
write at the Prelude prompt:
GHCi Prelude> sqrt 2
Prelude> sqrt 2
1.4142135623730951

Programming tasks
1 Try (a) sqrt 3 (b) div 5 2 (c) mod 5 2 (d) cos 0
(e) (^) 2 4 (f ) (^) 2 8 (g) gcd 128 32 (h) lcm 4 5
(i) take 5 "Hello World!" (j) drop 6 "Hello World!"
(k) splitAt 6 "Hello World!" (l) words "The black cat sat on the mat"

Working with characters, numbers, strings and Booleans

Character values can be created in two different ways:
• Writing the character itself between single quotes,
like'a'
• Writing in decimal between '\ and ', or in
hexadecimal between '\x and ' using the Unicode standard, e.g. the
character 'a' can be written as '\97 ' or '\x61'.
Using WinGHCi, the actual type of an expression can be checked by using the
:t command, followed by the expression itself. For example,
WinGHCi Prelude> :t 'a'
typing :t 'a' .
'a' :: Char
The :: symbol means “is of type”. Char is a pre-defined type
in Haskell.
Only a few functions are loaded by default. The import
WinGHCi Prelude> import Data.Char
command is used to add more functions, e.g. to import the
Prelude Data.Char> :t toUpper
Data.Char module use import Data.Char. The prompt
toUpper :: Char -> Char
of the interpreter changes to reflect the fact that now two
different modules are loaded, Prelude and Data.Char.
Click on Help and select Libraries documentation to see other modules that
are available.
In Haskell, everything has its own type including functions. We see that the
type of the function toUpper is Char -> Char.
The arrow -> syntax is used to specify the type of a function. In this case,
WinGHCi Prelude Data.Char> chr 98 toUpper is a function taking a character (the Char on the
'b' left side) and returning another one (because of the Char on
Prelude Data.Char> :t chr the right side).
chr :: Int -> Char Function types may be specified with a different data type
Figure 12.2.1.5 Function type with different before and after the arrow as shown in Figure 12.2.1.5.
argument and result data types

582 Single licence - Abingdon School

 12.2.1 Functional language programs

For functions with more than one argument, each argument type is separated
from the next with
WinGHCi Prelude Data.Char> max 5 3 a single arrow. For
5 example, the max
Information
Prelude Data.Char> :t max function takes two
=>:
max :: Ord a => a -> a -> a ordinal arguments, e.g. Everything before => is a class
of type Integer, and constraint. Ord a => means that
Figure 12.2.1.6 Function type of a function with
returns the smallest the type of the arguments and
two arguments return value must be members of
one.
the Ord class.
Its function type is expressed as a -> a -> a because this function is Ord contains all data types that
polymorphic, i.e. it will work with any ordinal type - Figure 12.2.1.6. allow data values to be put into
an order e.g. numerical order or
Ord is the language-defined name for an ordinal type.
alphabetical order.

Programming tasks
2 At the Prelude prompt in WinGHCi type import Data.Char.
Try (a) chr 48 (b) chr 49 (c) chr 50 (d) chr 57 (e) chr 65
(f ) chr 127 (g) ord '0' (h) ord '1' (i) ord '2' (j) ord '9'
(k) ord 'A' (l) min 45 32

Numbers
Haskell supports a great variety of number types.
• Int is a fixed-precision integer type with at least range [-229 .. 229-1].
These are bounded, machine integers, represented by 29-bit signed binary
at least. The exact range for a given implementation can be determined by
using minBound and maxBound.
• Integer is an unbounded integer type (mathematical integers, also
known as “bignums”): it can represent any value without a fractional part
without underflow or overflow. This is very useful for writing code without
caring about the range.
• Exact rational numbers using the Ratio type. Rational values are
created using n % m.
• Float and Double are floating-point Function Module
truncate
types of single and double precision, round
respectively. In module Prelude
floor
Haskell provides functions to convert between ceiling
toRational
types. Table 12.2.1.1 shows a sample of these. In module Data.Ratio
fromRational
%

Table 12.2.1.1 Some functions that convert between number types

Single licence - Abingdon School 583

12 Fundamentals of functional programming

WinGHCi Prelude> truncate 6.7

6
Prelude> round 6.7
Programming tasks
7
3 Try (a) truncate 45.34 (b) round 45.6 Prelude> ceiling 6.2
(c) ceiling 5.3 (d) floor 5.3 7
Prelude> floor 6.7
4 Type import Data.Ratio to add this library
6
to Prelude. Try
Prelude Data.Ratio> import Data.Ratio
(a) toRational 0.5
(b) toRational 0.25 Prelude Data.Ratio> 1%2 + 1%4
(c) toRational 0.125 3%4
Prelude Data.Ratio> toRational 6.7
7543529375845581 % 1125899906842624
Prelude Data.Ratio> fromRational (6 % 3)
2.0
Prelude> import Data.Ratio
Prelude Data.Ratio> :m - Data.Ratio
Prelude>

Infix operators
Key point
The infix operators +, −, ∗, /, ^ are used as follows
Infix operators as functions:
Prelude> 4 + 5
Functions whose name is
9
built entirely by symbols, like
+, must be called using infix They may also be used in a similar way that a function is applied by placing
syntax: writing them the operator in brackets as follows
between the arguments, instead
of in front of them. So you Prelude> (+) 4 5
write x + y, not + x y. To use 9
a symbol function in the Care must be taken with negative numbers, e.g. −4 in the vicinity of an
normal fashion, you must use infix operator or any of the functions (+), (−), (∗), (/), (^). We must
parenthesis as follows wrap in parentheses the expression it applies to, e.g. 4 ∗ (−3) = −12 where
(+) x y.
the parentheses have been used to surround (−3).

Programming tasks
Key point
5 Try (a) (+) 9 7 (b) (∗) 3 4 (c) (/) 4 3 (d) (−) 4 3
^:
(e) (−) 4 (−3)
^ is the exponentiation
operator, e.g. 4^2 = 16 Relational operators
because it is 4 raised to the
power of 2. Op Description Op Description Op Description
We can also write this in == Equal to > Greater than < Less than
Haskell as (^) 4 2.
Greater than or Less than or
>= <= /= Not equal to
equal to equal to

584 Single licence - Abingdon School

 12.2.1 Functional language programs

Strings
If characters are grouped together then we have what is called a string. In
Haskell a string is enclosed in double quotes, e.g. "Hello World!".
If the WinHGCi interpreter is asked what type is a string, it responds WinGHCi Prelude> :t "Hello World!"
with [Char]. The square brackets, [ ]surrounding Char indicate
"Hello World!" :: [Char]
that "Hello World!" is not Char but a list of characters, each of
type Char. Lists are the most commonly used data structure in functional
programming. We study lists in greater detail in the next chapter. Information
WinGHCi Joining two strings together is ++ operator:
++ can also be used to
Prelude> "Hello " ++ "World!" called concatenation.
concatenate two lists, e.g.
"Hello World!" The ++ function is used for this, [1,2,3,4] ++ [5,6,7,8]

Prelude> (++) "Hello " "World!" = [1,2,3,4,5,6,7,8]

e.g."Hello " ++ "World!"
"Hello World!" or (++) "Hello "" World!" WinGHCi
Prelude> True && True
Booleans
True
The truth values, true and false, are represented in Haskell by the literal
Prelude> True && False
constants True and False respectively.
False
Boolean operators
Prelude> True || True
Op Description Op Desccription Op Description
&& and || or not not True

Anonymous function Prelude> False || False

An anonymous function is one that is not given a name but instead is defined False
as follows Prelude> not True

(\x -> x + 2)
False
Prelude> not False
This means given a value for x add 2 to it. The argument to this anonymous
True
function is x and the result returned is the value of the expression x + 2. To
apply this anonymous function we supply an actual argument e.g. 4 as follows
(\x -> x + 2) 4 WinGHCi Prelude> (\x-> x + 2) 4
A shorthand form of this is possible if the function is an operator: 6

(+2) 4
Information
Programming tasks Anonymous functions:
Anonymous functions are useful
6 Write anonymous functions and apply these to the integer 5
where there is no requirement
(a) double a given integer (b) double a given integer and add 3
to make the function visible
to other parts of the program
Conditional expressions code or to reuse the function.

We can write general conditional expressions by means of the They are often used for callback
functions in event-driven
if ... then ... else construct, e.g. the value of the following expression
programming.
if condition then x else y

is x if the condition is True and is y if the condition is False.

Single licence - Abingdon School 585

12 Fundamentals of functional programming

Using the two-argument anonymous function, arguments x and y

Programming tasks (\x y -> if x >= y then x else y)

7 Write an anonymous We can apply this function to x = 4 and y = 5 as follows

function that outputs WinGHCi (\x y -> if x >= y then x else y) 4 5
True if x = 2∗y (\x y -> if x >= y then x else y) 4 5
otherwise False 5

Using let to make temporary definitions

Key point
It is possible to make temporary definitions in WinGHCi or GHCi using let
let: as follows:
It is possible to make temporary let x = 5
definitions in WinGHCi or
GHCi using let as follows:
We can use let to name and define a function square as follows
let x = 5 WinGHCi Prelude> let square x = x*x
Prelude> square 4 let square x = x ∗ x

16
Programming tasks
Pattern matching
8 Write a named function Pattern matching is an important aspect of Haskell and functional programming
using let to languages in general.
(a) double a given
The left-hand side of equations such as fac 0 = 1 and
integer
(b) cube a given integer fac n = n ∗ fac (n - 1) contain the patterns 0 and n, respectively.

When a function is applied these patterns are matched against argument values.
If the match succeeds, the right-hand side of the equation is evaluated and
returned as the result of the application. If it fails, the next equation is tried.
If all equations fail, an error results. All the equations that define a particular
function must appear together, one after another. The definition of the fac
function is a recursive one because fac refers to itself on the right-hand side of
the second equation.

Using a text editor to create Haskell functions

It is often more convenient to use a text editor such as NotePad++ or WordPad
to write Haskell programs. These Haskell programs should be saved with
extension .hs.
Switch WinGHCi to the directory in which the Haskell program file was saved,
e.g. c:\book\haskell, by typing at the Prelude prompt the following
command,
:cd c:\book\haskell followed by the Enter key

586 Single licence - Abingdon School

 12.2.1 Functional language programs

The Haskell program, e.g. factorial.hs may then be loaded into WinGHCi
at the Prelude prompt as follows
:load factorial.hs followed by the Enter key

WinGHCi will perform a compilation of a module called Main in order to

run factorial.hs interactively.
If there are no errors loading and compiling, the Prelude prompt will be
replaced by the prompt *Main.
To return to Prelude at any time, type :m.
WinGHCi
We are now ready to run the factorial program.
Prelude> :load factorial.hs
At the *Main prompt type [1 of 1] Compiling Main ( factorial.hs, interpreted )
fac 6 followed by the Enter key Ok, modules loaded: Main.
The correct answer, 720, is displayed. *Main> fac 6
720

Programming tasks

9 Use a text editor to create the following two functions, one

after another in the same file, cubedouble.hs: cube x and
double x where x is an Integer.
Try:
(a) cube 3 (b) double 6 (c) double (cube 4)
Key concept
Higher-order function
Higher-order functions A function that takes a function
as an argument or returns a
Much of the power of a functional language comes from advanced use of
function as a result (or does
functions. In particular, a function that takes a function as an argument both) is said to be a higher-
or returns a function as a result (or does both) is a higher-order function. order function.
Higher-order functions make it possible to define very general functions that
are useful in a variety of applications.
Key principle
Map
Map:
Our first example of a higher-order function is the map function. This function
Map is a higher-order function
applies a given function to each element of a list, returning a list of results. For that applies a given function to
example, to apply the abs function to every element of the integer list each element of a list, the results
in a new list.
[-1,-2,-3,-4] we do the following

map abs [-1,-2,-3,-4]

WinGHCi Prelude> map abs [-1,-2,-3,-4]
The abs function returns the absolute value of its argument, e.g.
[1,2,3,4]
abs −3 is +3. The map function applies the abs function to each
value in the list, i.e.
abs −1 abs −2 abs −3 abs −4 map f [x1, x2, ..., xn] == [f x1, f x2, ..., f xn]

Single licence - Abingdon School 587

12 Fundamentals of functional programming

To add 3 to every element of the integer list [1,2,3,4] do the

WinGHCi Prelude> map (+3) [1,2,3,4]
following
[4,5,6,7] map (+3) [1,2,3,4]

To multiply every element of the integer list [1,2,3,4] by 3 do the following

map (∗3) [1,2,3,4] WinGHCi Prelude> map (∗3) [1,2,3,4]
[3,6,9,12]
Programming tasks

10 Try the following in WinGHCi or GHCi

Information
(a) map (/2) [2,4,6,8]
Function application
(b) map (\x -> x*x) [1,2,3]
operator $:
(c) map (\x -> x*x) [1..100] Can be used with map to map
(d) map (\x -> x^2) [1..100] a function application over a list
(e) map (\x -> x^2) [1..1000] of functions, i.e. $ is replaced in
turn by each function. In Q10(i)
(f ) map sqrt [1..16]
($ 4) gets mapped over the list.
(g) :m Data.Char The list happens to be a list of
map toUpper ['a'..'z'] functions. So every function in the
(h) map words ["hello world", "the sun has got list gets applied to 4.

its hat on"]

(i) map ($ 4) [(10∗),(5+), sqrt, (^2)]

Filter
Key principle
The filter function is a higher-order function that processes a data structure,
Filter:
The filter function is a higher- e.g. a list, in some order to produce a new data structure containing exactly
order function that processes those elements of the original data structure that match a given condition.
a data structure, e.g. a list, For example, filter can apply the even function to every element of the
in some order to produce a integer list [1,2,3,4,5,6,7,8] and return a list containing integers that
new data structure containing
possess the property of evenness WinGHCi
exactly those elements of the
original data structure that filter even [1,2,3,4,5,6,7,8] Prelude> filter even [1,2,3,4,5,6,7,8]
match a given condition. Similarly, filter can also apply the odd [2,4,6,8]
function to every element of the integer
list [1,2,3,4,5,6,7,8] and return a list containing integers that possess the
property of oddness WinGHCi
filter odd [1,2,3,4,5,6,7,8] Prelude> filter odd [1,2,3,4,5,6,7,8]
[1,3,5,7]
Programming tasks
Information
11 Try the following in WinGHCi or GHCi
Function application operator $:
(a) filter (\x -> x > 3) [0..10] When a $ is encountered, the expression on its
(b) filter (\x -> (mod x 3)== 0) [1..100] right is applied as the argument to the function
(c) :m Data.Char on its left, e.g. in 11(d) filter odd [1..100]
filter isUpper ['!'..'z'] is evaluated first, the result then becomes the
argument to map (⋆2).
(d) map (⋆2) $ filter odd [1..100]

588 Single licence - Abingdon School

 12.2.1 Functional language programs

Reduce or fold
Reduce or fold is the name of a higher-order function which reduces a list of values to a single value by repeatedly
applying a combining function to the list of
values.
r o
u lat H
In the folding or reduction process, a m e
cu
A c a
function, e.g. sum, is applied to the list 0 d
1
element by element, returning something such
2 3
as the total sum of all elements. A fold takes a
binary function, a starting value (often called
4
an accumulator), and a list to fold up. The
12.2.1.7 Folding from the left
fold reduces the entire list down to a single
accumulator value. In Haskell folding from
the left is done with foldl. For example to sum a list of integers [1,2,3,4] using foldl do the following
foldl (+) 0 [1,2,3,4]
WinGHCi Prelude> foldl (+) 0 [1,2,3,4]
foldl takes three arguments: the binary function (+), the
10
accumulator which has value 0, and the list [1,2,3,4].
Binary means the function (+) takes two
operands. The first of these is the accumulator.
The second operand starts with the first element
of the list (the head). (+) returns their sum and T
1 2 a
this becomes the new accumulator. foldl or
il u lat
3 4 c um
then applies (+) to the new accumulator Ac
value and the second element in the list to 0
produce a new value of the accumulator
and so on until all the integers in the list are 12.2.1.7 Folding from the right
summed.

To start the traversal from the opposite end of the list, the tail, we use foldr as follows

foldr (+) 0 [1,2,3,4] WinGHCi Prelude> foldr (+) 0 [1,2,3,4]

foldr differs from foldl in the order of the 10
arguments which the function is applied to.
For example, if the function applied is (^), exponentiation, then
the new value of accumulator = current list value ^ old value of accumulator.
With foldr (^) 2 [1,2,3] in WinGHCi, the accumulator is 2. WinGHCi
The list is folded from the right as follows Prelude> foldr (^) 2 [1, 2, 3]
list value^ accumulator 1

3^2 = 9

2^9 = 512

1^512 = 1

Single licence - Abingdon School 589

12 Fundamentals of functional programming

WinGHCi With foldl (^) 2 [1,2,3] in WinGHCi, the accumulator is 2. The list is
folded from the left as follows
Prelude> foldl (^) 2 [1,2,3]
accumulator ^ list value
64
2^3 = 8

8^2 = 64

64^1 = 64

Programming tasks

12 Try the following in WinGHCi or GHCi

(a) foldl (+) 0 [1..1000]
(b) foldl (∗)1 [1..3]
(c) map(^2)[1..3]
(d) foldl (+)0 $ map (^2) [1..3]
(e) foldr (++) "!" ["Hello ", "World"]
(f ) foldl (++) "!" ["Hello ", "World"]

Questions

1 In a functional programming language, a higher-order function map takes two arguments, a function f,
and a list and applies the function f to each element of the list returning the results in a new list.
(a) A function named square is defined below
square x = x ∗ x
The result of making the function call square 2 is 4.
Calculate the result of making the function call
Function call Result
map square [1,2,3,4]

(b) Another function defined as (++ "!") takes a single string argument and returns a string which is the
concatenation of the string argument with !
++ is the concatenation operator. For example,
(++ "!") "Wow" returns "Wow!"
Calculate the result of making the function call
Function call Result
map (++ "!") ["SLAP", "BAM",
"WALLOP"]

590 Single licence - Abingdon School

 12.2.1 Functional language programs

Questions

2 In a functional programming language, a higher-order function filter takes two arguments, a function f
that returns a Boolean value, and a list and applies the function f to each element of the list returning the
results in a new list.
A function defined as (>4) takes a single integer argument and returns true or false depending upon
whether the integer is greater than 4 or not.
Calculate the result of making the function call
Function call Result
filter (>4) [5,3,6,2,-1]

3 In a functional programming language, a higher-order function foldl takes as arguments a binary

function, f, e.g. (+), a starting value called the accumulator, and a list, and returns a single value. A new
accumulator value is calculated as follows if function f is (+)
new accumulator = old accumulator + a list value
Calculate the result of making the function call where f is (∗) and ∗ is the multiplication operator.
Function call Result
foldl (∗) 1 [5,3,6,2]

4 Using the higher-order functions map, filter and foldl.

Calculate the result of making the function call

Function call Result
foldl (+) 0 $ filter (>14) $ map (∗3) [3..8]

$ means evaluate the rightmost expression first.

In this chapter you have covered:

■■ Constructing simple programs in a functional programming language.
■■ Using higher-order functions.
■■ Using the following functional programming language higher-order
functions:
• map
• filter
• reduce or fold.

Single licence - Abingdon School 591

12 Fundamentals of functional programming
12.3 Lists in functional programming
Learning objectives: languages
■■Be familiar with representing
a list as a concatenation of a
■■ 12.3.1 List processing
head and a tail What is a list?
A list is a collection of data items stored in no particular order, having the
■■Know that the head is an following properties:
element of the list and the tail
• data items may be inserted or deleted at any point in the list
is a list
• data items may be repeated in the list
■■Know that a list can be empty
• lists may contain any type of object
■■Describe and apply the
• a particular list may contain different object types
following operations:
A list is represented by square brackets enclosing list values separated by
• return head of list
commas. For example:
• return tail of list
• test for empty list [Emma, John, Fred, Janet]
• return length of list is a list consisting of the items Emma, John, Fred, Janet.
• construct an empty list
A more complex example of a list containing different types of item (including
• prepend an item to a list
another list) is:
• append an item to a list
[Mick, 46, 5.15, London, [Spurs, Chelsea]]
Key concept This type of list of mixed-type items is not supported by all programming
List: languages that support list processing.
A list is a collection of data Element of a list
items stored in no particular
We refer to a data item in a list as an element or item of a list, e.g. Emma is an
order, having the following
properties: element or item of the first example list above.
• data items may be inserted Empty list
or deleted at any point in If all the elements of a list are removed we still have a list but it is now an
the list
empty list. The empty list is denoted by square brackets as follows [].
• data items may be repeated
in the list Thus every list is either empty or non-empty.
• lists may contain any type
of object Representing a list as a concatenation of a head and a tail
• a particular list may contain A non-empty list may also be written in the form h:t where h is the first item
different object types
in the list and t the remainder of the list. t is itself a list.
A list is represented by square
brackets enclosing list values For example,
separated by commas. For
example:
the list [1,2,3] = 1 : [2,3]
[Emma, John, Fred, Janet] We call the element 1 the head of the list and [2,3] the tail of the list. : is
called the list constructor operator.

Single licence - Abingdon School 592

12 Fundamentals of functional programming

Key concept Return head of list

In an imperative programming language, the function call
List element or item:
head(list)
We refer to a data item in a list
as an element or item of a list. returns the element at the head of list if list is non-empty, otherwise an
error is reported.
Key principle
e.g. for list [Emma, John, Fred, Janet]
Every list is either empty, [], or
head([Emma, John, Fred, Janet]) will return Emma.
non-empty.
In the functional programming language Haskell, a function that returns the
Key principle item at the head of a list, h : t is defined as follows

head(list):
head :: [a] -> a

The operation head (list) head (h : _) = h

returns the first element in the
where h is the head. The underscore represents an anonymous variable which is
list.
the mechanism used by Haskell to ignore a value.
WinGHCi Prelude> head [1,2,3,4]
1 Using WinGHCi, the pre-defined function head applied
Prelude> head [] to [1,2,3,4] returns 1. Applying head to the empty list
*** Exception: Prelude.head: empty list [] causes an exception.

Return tail of list

In an imperative programming language, the function call
tail(list)

returns a new list containing all but the first element of the original list, list.
Key principle
e.g. for list [Emma, John, Fred, Janet]
tail(list):
tail([Emma, John, Fred, Janet]) returns [John, Fred, Janet]
The operation tail (list)
returns a list formed by In Haskell, a function that returns the tail of a list, h : t is defined as follows
removing the first element of
the original list.
tail :: [a] -> [a] WinGHCi prelude> tail [1,2,3,4]
tail(_: t) = t [2,3,4]
where t is the tail.

Test for empty list

In an imperative programming language, a function call
empty(list)

Key principle returns True if list is an empty list or False otherwise.

empty(list):
In Haskell, a function null that returns True if list is an empty list or
The operation empty (list) False otherwise is defined as follows
returns True if list is empty or null :: [a] -> Bool WinGHCi prelude> null []
False otherwise. null[] = True True
null (_ : _) = False
prelude> null [1,2,3,4]
where [] is the empty list.
False
593 Single licence - Abingdon School
 12.3.1 List processing

Questions Information
1 What result is returned by the following functions applied to the list Convention for representing
lists:
Towns where
In Haskell it is an informal
Towns = [Swindon, Aylesbury, Banbury, Stevenage, Slough] convention to write variables
(a) empty(Towns) (b) head(Towns) (c) tail(Towns) over lists in the form xs, ys
(d) head(tail(Towns)) (pronounced ‘exes’, ‘whyes’) and
(e) tail(tail(tail(tail(Towns)))) so on, with variables x, y, ... for list
elements. For example, x : xs is
a list with head element x and
Return length of list tail xs.

In Haskell, a function length that returns the length of a list h : t is defined as

follows
Key principle
length :: [a] -> Integer
WinGHCi prelude> length [1,2,3,4] length(list):
length [] = 0
The operation
length (h:t) = 1 + length t 4 length(list) returns the
number of elements in the lists,
The length function uses recursion to calculate the length of list h : t. i.e. its length.

h has a length of 1 to which we add length of the list t. The list t is the original
list minus its head.

Construct a list from the empty list

Every list can be built from the empty list by repeatedly applying the list
constructor operator ':'. For example, the list [1,2,3,4] can be created as
follows WinGHCi Prelude> 4:[]
[] 4:[] = [4] 3:[4] = [3,4] 2:[3,4] = [2,3,4] [4]
1:[2,3,4] = [1,2,3,4] Prelude> 1:2:3:4:[]
OR
[1,2,3,4]
1:2:3:4:[]
The operator : serves a special role: it is a constructor of lists, since every list Key principle
can be built up in a unique way from [] and ':'. The form 1:2:3:4:[] is
The : list constructor operator:
called the list constructor form of the list [1,2,3,4]. Every list can be built from
the empty list by repeatedly
Questions applying the list constructor
operator ':', e.g.
2 Write the list constructor form of a list for the following lists 1:2:3:4:[] = [1,2,3,4]
(a) [6,3,8,1] (b) [2]

Single licence - Abingdon School 594

12 Fundamentals of functional programming

Prepend an item to a list

Prepend is when an item is added to the beginning of a list. For example, in
Haskell the following code prepends 0 to the list [1,2,3,4]
0 : [1,2,3,4] WinGHCi Prelude> 0:[1,2,3,4]
Key principle [0,1,2,3,4]
Prepend: Questions
To prepend an item to a list we
add the item to the beginning 3 What lists result from the following prepend operations
of the list using the':' operator.
(a) 6 : [5,4] (b) 4 : [] (c) 4 : [1, 2, 3]?
E.g. 0 : [1,2,3,4] =
[0,1,2,3,4]
Append an item to a list
Append is when an item is added to the end of a list. For example, item 5 is
appended to list [1,2,3,4] to produce the new list [1,2,3,4,5].
(a ,[a]) -> [a]
y A user can define a function append in Haskell that appends an item y
append z:zs
x:xs to a list x : xs as follows using the function reverse which is built-in to
Haskell
Two argument input function
is first treated as a function append :: a ->[a] -> [a]
with a single argument y which append y (x : xs) = reverse(y : reverse(x : xs))
returns as result a new function
For example, append 5
x:xs -> z:zs. WinGHCi *Main> append 5 [1,2,3,4]
a ->([a] -> [a])
[1,2,3,4] returns
[1,2,3,4,5]
[1,2,3,4,5]
y yappend x:xs -> z:zs

Append is an example of the principle that any function taking

Therefore if for y, the value 5 two or more arguments can be partially applied to one or more arguments.
is input and then for x:xs the
value [1,2,3,4] is input, the (a ,[a]) -> [a] => a ->([a] -> [a]) => a -> [a] -> [a]
final result is [1,2,3,4,5]. The function first reverses the list x : xs. It then prepends the item y to the
This step by step process is
reversed list and then finally reverses the result. For example, if we wish to
expressed in Haskell as
append 5 to [1,2,3,4]
a -> [a] -> [a]
Reverse: [4,3,2,1]
Key principle Prepend: 5 : [4,3,2,1] = [5,4,3,2,1]
Append: Reverse: [1,2,3,4,5]
To append an item to a list we
add the item to the end of the Questions
list.
E.g. append 5 [1,2,3,4] 4 What is returned when the function append y (x:xs) is called as follows
returns [1,2,3,4,5] (a) append 6 [1] (b) append 5 [3,2]?

5 Define the function

listproduct :: [Integer] -> Integer
which returns the product of a list of integers or 1 if the list is empty.

595 Single licence - Abingdon School

 12.3.1 List processing

Programming tasks
1 The function max is built-in to Haskell. It takes two arguments of an
ordinal type and returns the larger of the two. Use WinGHCi and enter
the following at the Prelude prompt: WinGHCi Prelude> max 9 3
Prelude> max 9 3
9
Now try (a) max 46 57 (b) max 46 43

2 Create the following Haskell program in a text editor such as

Notepad++ and save as maxy.hs.
maxy :: (Ord a) => [a] -> a
maxy [] = error "no maximum of empty list!"
maxy [x] = x
maxy (x:xs) = max x (maxy xs)

Ord is Haskell’s ordinal data type, (Ord a) => specifies that the values
of a must belong to an ordinal type.
Load this program into WinGHCi using the command
:load maxy.hs
Try (a) maxy [] (b) maxy [4]
(c) maxy [34,23,7,67,31]

In this chapter you have covered:

■■ Representing a list as a concatenation of a head and a tail, e.g. the list
[1,2,3] = 1 : [2,3] where : is the list constructor operator, 1 is
the head element and [2,3] is the tail
■■ The empty list []
■■ The following operations on lists:
• The operation head of list returns the element at the head of the list
• The operation tail of list returns a list which is the original list minus
the first element, the head
• The operation empty list tests for an empty list returning either True
or False
• The operation length of list returns the length of the given list
• Every list can be built starting from the empty list by repeatedly
applying the construction operator ':',
e.g. [] 4:[] = [4] 3:[4] = [3,4]
• Prepend operation which is when an item is added to the beginning of
a list
• Append operation which is when an item is added to the end of a list.

Single licence - Abingdon School 596

 Index
Symbols Address bus 258, 262
Address mode 281
^ 584
Address space 262
: 594
Advantages and disadvantages of lossless and lossy
[] 593
compression 178
< 522
Advantages of programming in machine code and assembly
<= 522
language compared with HLL programming 223
<> 522
Advantages of the vector graphics 162
> 522
Advantages of using MIDI files for representing music 176
>= 522
AES256 410
1NF 502, 503, 506, 507, 510
Aliasing 169
2NF 506, 508, 510
ALU 270, 276
3NF 506, 508, 511
ALU performing an ADD operation 276
10/100 Mbps Ethernet network adapter 360
Amplitude 165
802.11 mobile and portable devices 368
802.11 Analogue data 123, 129
Standard 368 Analogue/digital conversion 129
:: operator 563 Analogue electrical signal 126
[R0] 289 Analogue sensors 136
Analogue signal 126
A
Analogue to digital converter (ADC) 129, 317
Absolute error 75 AND 298, 522, 523
Abstraction 569 AND Rd, Rn, <operand2> 282
Access control and authentication 377 AND truth table 234
Access control WPA/WPA2 376 Anonymous access 431
Access Point (AP) 368, 370 Anonymous function 585
Access time 325 Anti-malware software 418
Accumulator 285, 589 Anti-virus software 418
Acknowledgement (ACK) 374 AP 371
Acknowledgement (ACK) signal 373 Aperiodic waveform 164
Acknowledgement number 389 API 420, 461
Active RFID tags 320 Append 595
Adaptive dictionary scheme 180 Application-layer 421
ADC 317, 318 Application layer protocol 433
ADCs and analogue sensors 135 Application Programming Interface (API) 420, 434
ADD 292 Application programs 216
Adding two binary numbers with carry 241 Application software 216
Addition of two unsigned binary integers 48 Approximating a number 75
ADD Rd, Rn, <operand2> 282 Archiving data 329
Addressable memory locations 262 Argument 562

Single licence - Abingdon School 597

 

Arithmetic and Logic Unit (ALU) 270, 276 BCNF informally 508
Arithmetic expression 569 B <condition> <label> 282
ARM 283 Beacon frame 371
ARM Cortex-A7 quad-core processor 310 BEQ <label> 297
ARM Cortex processor 289 Bespoke software 217
ARM processor 282 BGT<label> 297
ARM processor architecture 284 Big Data 333, 543
ARM μVision 289 Bigtable 559
ARM μVision V 5.17 Debugger in single-step mode 289 Binary 24
ARPANET 447 Binary addition with Sum and Carry 240
ASC 522 Binary data 325
ASCII 96, 316 Binary digit 36
ASMTutor 276, 288 Binary digits 344
ASN 398 Binary function 589
Assembler 218, 222, 225 Binary, the language of the machine 106
Assembly language 220 Binary to hexadecimal conversion 31
Associative law 249 Binding port and socket 424, 427
Asymmetric encryption 405 BIOS 260
Asymmetry of power 332, 333 Bit 36
Asynchronous data transmission 346, 347 Bitcoin 341
Asynchronous serial data transmission 347 Bit depth of bitmapped image 143
Asynchronous serial data transmission frame 350 Bit depth of sampled sound 166
Atomic facts 556, 557 Bit is the fundamental unit of information 36
Atomicity 500 Bitmap 141, 319
Atomic transaction 541 Bitmapped graphics 139
Attribute 476, 487 Bitmapped graphics uses 163
Attribute domain 498 Bitmapped image 141
Audacity 116 Bitmap resolution in number of pixels per inch (ppi) 145
Authentication 376, 410 Bitmap size in pixels 143
Average no of clock cycles per instruction 308 Bit patterns 106, 221
Bit rate 352, 354
B
Bits 325
Backing up data 329 BitTorrent 366
Bandwidth 168, 353 Bitwise AND 445
Barcode reader 316 B <label> 282
Barcode scanning versus RFID scanning 322 Block-oriented storage device 324
Barcode symbols 316 BLT <label 297
Base 2 24 Blue-ray disc 326
Base 10 24 BNE <label> 297
Base 16 25 BOOLEAN 530
Basic Service Set (BSS) 369 Boolean algebra 247
Baud rate 351, 354 Boolean expression equivalent of a logic gate circuit 238
Bayer filter 318 Boolean function 233

Single licence - Abingdon School 598

 

Boolean functions 247 CCD camera 317

Boolean identities 248 CCD light sensor array 139
Boolean operators 247, 585 CCD-type camera sensor 318
Boolean variables 231, 247 CCD vs CMOS 318
Boot ROM 470 CD-R 326
Boyce-Codd Normal Form 516 CD-ROM 325
Boyce-Codd Normal Form (BCNF) 506, 508, 516 CD-RW 326
Branching (conditional and unconditional) 295 Central Processing Unit (CPU) 257, 270
Branch instructions and effect on PC 279 Central server 361
Breaking the Caesar cipher 194 Central switch 359
Brute force approach 194 CERT 418
BSSID address 369 Channel selection 371
Buffer overflow 416 Character form of a decimal digit 99
Built-environment 338 Charge-Coupled Device (CCD) 317
Built-in redundancy 385 CHAR(n) 530
Bus 358 Check digit 103
Bus Grant 261 Checksum 102, 357, 390
Bus line 258 Chuck 174
Bus masters 258 Cipher 184
Bus Request 261 Ciphertext 185
Bus subsystem 258 Cipher wheel 185
Bus topology 361 CIR 270, 275
Bus width 258 Classification of software 216
Bus width effect on processor performance 311 Classification of waveforms 164
Byte 37, 348 Class of computations 569
Bytecode 228 Clear To Send (CTS) 375
Byte quantities 40, 42 Client computer 364
Client ports 427, 428
C
Client-server 462
Cache hierarchy and role in Fetch-Execute cycle 312 Client-server architecture 364
Cache memory 312, 313 Client server database system 533
Caesar cipher 185 Client-server model 459
Caesar cipher weaknesses 197 Client-server TCP connections 403
Calculating storage requirements for bitmapped images 152 Clock 261
Callback function 460 Clock cycle time 309, 314
Camera resolution 147 Clock frequency 272
Cancelling NOTs 252 Clock period 272
Capacity of storage media 328 Clock rate 309
Carrier Sense Multiple Access/Collision Avoidance (CSMA/ Clock speed 272, 309
CA) 371, 373
Cloud computing 339
Carry 275
CMOS 317
Carry flag (C) 276, 293
CMOS light sensor array 139
CAT5 cable 359
CMOS sensor 318
CCD 317

Single licence - Abingdon School 599

 

CMP R0, R1 297 Connection socket 426

CMP Rn, <operand2> 282 Contiguous 179, 328
CMYK colour model 319 Control bits 275
Code quality 415 Control bus 258, 261
Codes versus ciphers 185 Control instructions 281
Co-domain 563, 578 Control Unit 270, 271
Collision avoidance 373 Converting an integer from decimal to two’s complement
Collisions 372 binary 54
Colour depth of a bitmap 143 Converting an integer from two’s complement binary to

Colour laser printer 319 decimal 55

Converting digital audio signals to analogue
Combining Boolean functions 247
using a DAC 137
Command line 531
Converting from binary floating point to decimal 66
COMMIT 537
Converting from decimal to binary floating point form 65
Commitment ordering 541
Converting from decimal to fixed point binary 62
Commodity hardware 545
Converting signed decimal to signed two’s complement
Commodity server 544
fixed point binary 64
Communication protocol 354
Cookie 337
Compact Disc (CD) 323
Cookie Law 337
COMPARE 295
Core 309
Comparing absolute and relative errors 76
Cosine waveform 165
Comparing bus and star networks 361
Counting 20
Comparing JSON and XML 468
CPI 309, 314
Comparison of synchronous and asynchronous
CPU 265, 270
data transmission 348
CPU time 309
Comparison of the adv. and disadv. of fixed pt.
CR2 files 318
and floating pt. 83
CREATE TABLE 527, 531
Compiler 218, 223, 225
Creating digital sound files 117
Compiling 226
Crosstalk 346
Complementary Metal-Oxide Semiconductor 317
Crow’s foot 481
Components for wireless networking 369
CRUD operations 463
Composite entity identifier 477
Cryptanalyst 184
Composite primary key 488, 529
Cryptographer 184
Computation 569
Cryptography 184
Computational secrecy 214
Cryptosystem 184
Computational security 213
CSMA/CA 371
Computation method 569
CSMA/CD 357
Conceptual data model 474, 475
CTS 375
Concurrent access 533
Cultural attitudes 338
Conditional branch 297
Cultural issues and opportunities 330
Conditional branch instructions 297
Current Instruction Register 271, 275
Conditional expressions 585
Current Instruction Register (CIR) 270
Condition codes 275, 293, 297
Cycle of waveform 167
Congruence 191
Cycles per instruction (CPI) 308

Single licence - Abingdon School 600

 

D Decryption 184
Dedicated registers 270, 274
Dark Web 341
Dedicated server 364
Data 123
DEFAULT 528
Data anomaly on deletion 515
Degree of a relationship 481
Data anomaly on insert 515
DELETE 463, 519, 525
Data at rest 543, 550
De Morgan’s laws 252
Database design 498
DESC 522
Database inconsistency 541
Destination hardware address 423
Database normalisation techniques 498
Destination IP address 389
Database transaction 533
Destination register 281
Data bus 258, 259
Determinancy diagram 500
Data compression 177
Determinant 500
Data Constraints 475
DHCP 451, 458
Data Definition Language (DDL) 527
Dictionary-based compression methods 180
Data exhaust 333
Difference between a router and a switch 388
Datagrams 422
Diffie-Hellman key exchange 408, 410
Data inconsistency on update 515
Digit 2
Data in motion 543, 550
Digital camera 317
Data integrity 533
Digital certificate 409, 410, 418
Data link layer 390
Digital certificate authority 410
Data link layer address 390
Digital data 125
Data Manipulation Language (DML) 527
Digital ethics 332
Data model 474
Digital image file 109
Data processing instructions 281
Digital representation of sound 164
Data Protection Act 332
Digital signal 127, 326
Data Requirements 474
Digital Signal Processor 266
Dataset 543
Digital signature 410, 411, 412
Data transfer instructions 281
Digital single-lens reflex cameras 139, 317
Data transfer time 325
Digital Subscriber Line (DSL) 368
Data type 562
Digital to analogue converter (DAC) 134
Data types 530, 531
Digital Versatile Disc (DVD) 323
DATE 530
Direct access to registers 224
Datum 35, 123
Direct addressing 288
DCIM (Digital Camera IMages) 318
Direct Memory Access (DMA) 264
DDL 527, 531, 532
Disadvantages of programming in machine code
Decentralized P2P network 367
and assembly language compared with HLL
Decimal 24
programming 224
DECIMAL 530
Disadvantages of the vector graphic approach 163
Decimal to binary conversion 26, 27
Discrete 126
Decimal to hexadecimal conversion 28
Discrete data 124
Decoder 180
Discrete logarithm problem 407, 408
Decoding 180
Disk block 324
Decrypting with cipher wheel 187

Single licence - Abingdon School 601

 

Disk buffer 324 Encoding in compression 180

Diskless workstation 470 Encrypting long messages 212
Display and print resolution 144 Encrypting with cipher wheel 186
Distributed file system 544, 545 Encryption 184, 404
Distributive law 249 Endpoint 425, 426
Division of an IP address 440 End-system 384
Division of IP address into network ID and host ID 439 End-to-end principle 387, 422
DML 527, 531 ENIAC 267
DNS 393, 394 Entity 475
DNS server 433 Entity description 479
Domain 563, 578 Entity identifier 477
Domain name 393 Entity occurrence or instance 476
Domain Name System (DNS) 393 Entity-relationship diagram 481
Domain Name System servers 395 Entity relationship modelling 474, 479
Dotted decimal notation 387, 438, 442 Environmental information 35
DRAM 311 EOR Rd, Rn, <operand2> 282
Drawing and interpreting logic gate circuit diagrams 235 Ephemeral port 427
DSP 266 Epiphany multicore chips 310
D-type flip-flop 243 Equal 275
Dumb terminal 471 E-R diagram 481
DVD-R 326 E-R modelling approach to normalisation 504
DVD-RAM 326 Error checking 100
DVD-ROM 326 Error correction 100
DVD-RW 326 ESynth 165
DVD+RW 326 Ethernet 348, 357
Dynamic Host Configuration Protocol 451 Ethernet bus 358
Dynamic Port 427 Ethernet bus switch 361
Dynamic random access memory 311 Ethernet frame 357
Ethernet switch 358
E
Ethical dimension of data processing 337
Edges 558 Ethics 330
Edge-triggered D-type flip-flop 243 Even parity 101
EDSAC 267 Exclusive-lock 537
EEPROM (Electrically Erasable Programmable Read Only Exclusive-OR (XOR) 101, 206
Memory) 327, 357 Executable binary codes 221
Effect of address bus width on processor performance 314 Exponent 61, 82
Effect of cache on processor performance 312 Exponentiation 406
Effect of interrupts on the fetch-execute cycle 306 External hardware devices 316, 323
Effect of word length on processor performance 313
Element of a list 592
Embedded systems 283, 348
F
Empty list 592, 593 Facebook RESTful web service 461
Encoder 180 Facebook’s graph API 462
Encoding 180
Single licence - Abingdon School 602
 

Facebook’s social network graph dataset 461 Full-adder 241

Fact 557 Fully normalised set of relations 504
Fact-based model 556, 557 Fully qualified domain name (FQDN) 394
Factors affecting processor performance 308 Function 561
Factual information 35 Functional abstraction 569
Fault-tolerant storage 544 Functional language programs 581
Fetch-Execute cycle 278 Functional programming 547, 555
File blocks 544 Functional programming paradigm 565, 569, 572, 578
File-sharing application 367 Function application 571
File Transfer Protocol 430 Function argument 570
Filter 588 Function as object 561
Firewall 400 Function as process 561
Firewall proxy server 403 Function body 570
First-class object 565 Function composition 578
First-class values 565 Function parameter 570
First Normal Form 502, 506 Functions as arguments 567
Fixed point form of signed numbers 59 Functions as first-class objects 565
Fixed point underflow 91 Function-to-data model 547
Flip-flop 243 Function type 562, 571
FLOAT 530 Fundamental frequency 167
Floating point form 60
G
Floating point overflow 94
Floating point underflow 91 Gateway 389
Fold 589 Gateway router 442
Foreign Key 489 General purpose application software 217
FOREIGN KEY 528 General purpose registers 270, 274
Foreign key already present 494 GET 463
Foreign Keys in One-To-One Relationships 494 GET / 433
Foreign Keys in Recursive Relationships 495 GFS 544
Format of a machine code instruction 283 Ghostscript 155
Formula 569 Global Internet 440
FOR UPDATE 537 Globally unique address space 383
Forwarding table 392 Google data centre 546
Four-way handshake 378 Google File System 544
FQDN 393, 433 Graphics 107
Frame header 390 Graph schema 558, 559
Fraud detection on credit card transactions 554 Grayscale bitmap 179
Frequency 165 H
Frequency of a sound 164
Hadoop 547
FTP 366, 430
Half-adder 240
FTP client 430
Half-adder Sum function 241
FTP client software 431
HALT 282, 302
FTP server 430, 431
Handshaking protocol 354
Single licence - Abingdon School 603
 

Hardware 216 I
Hardware interface 264
I2C 346
Hardware maintenance 472
IANA 427, 448, 449
Hardware random number generators 202
Identifying a TCP connection 424
Harmonic 167
Idle state 349
Harvard architecture 265, 268
Image quality 154
Hash function 412
Image resolution 144
Haskell 581
Image sensing and acquisition 139
HDFS 546, 554
Immoral action 330
head(list) 593
Immutable data 556
hertz (Hz) 353
Immutable data structures 555
Hexadecimal 25
IMP 447
Hexadecimal as shorthand for binary 31
Imperative high-level language (HLL) 223
Hexadecimal to binary conversion 30
Imperative programming language 593
Higher-order function 555, 587
Improving processor performance 309
High-level language 222
IN 522
High-level programming language 222
Inconsistent database 533
High-level programming language classification 222
Individual (moral) issues and opportunities 330
Horizontal scaling 547
Inferred data 339
Host 384, 437
Infix operators 584
Host aliasing 395
Infix operators as functions 584
Host computer 437
Information 34, 333
Hostname 384
Information carrier 34
Hosts in networks 364
Information = data + meaning 35
How domain names are organised 394
Information-theoretically secure 209
How routing is achieved across the Internet 391
Information types 35
HTML 432
Inkscape 159
HTML5 WebSocket 469
Input and output devices 316
HTTP 432
INSERT 463, 519
HTTP client 426
INSERT INTO 526, 531
HTTP GET 392, 426, 433
Instructional information 35
HTTP GET / 434
Instruction count 309
HTTP GET / request 434
Instruction Register (IR) 275
HTTP methods 461, 463
Instruction set 228, 283
HTTPS 410, 434
Instruction set is processor specific 283
HTTP server 426, 461
Instructions per second 273
Human fault tolerance 556
INTEGER 530
Hyperscale computing 336
Integer numbers 4
Hypertext Markup Language 432
Integer overflow 93
Hypertext Transfer Protocol 432
Integrity 521
Hypertext Transfer Protocol over Secure Sockets 434
Interface 437
Hypertext Transfer Protocol Secure 434
Interface Message Processor (IMP) 385
Hypertransport bus 312
Interference 372

Single licence - Abingdon School 604

 

Inter-Integrated Circuit (I2C) 346 Isolation mode 538

Intermediate code 218 ISR 306
Intermediate language 228
J
Internal hardware components of a computer 257
Internal structure of a processor/central processing unit 270 Jammin LTSP diskless thin-client workstation 471
internet 422 Janet 440
Internet 336, 383, 384 Java 547
Internet Assigned Numbers Authority (IANA) 427 Java Archive file (jar) 547
Internet Protocol address 423 Java bytecodes 228
Internet Protocol (IP) 383 Java Virtual Machine 228
Internet registrars 397 JES 116
Internet registries 397 JPEG 107, 318
Internet security 400 JSON 461, 462, 463, 468, 549
Internetwork 422 K
Interpreter 218, 225
Kazaa 367
Interpreter vs compiler 226, 227
Key 185
Interrupt Enable/Disable 275
Key exchange 408, 409
Interrupt Enable flag 276
Key pair 405
Interrupt request 261
Key-value pairs 548
Interrupts 304
kibi Ki 42
Interrupt service routine 306
kilo k 42
Interrupt signal 304
Knowledge 333
I/O controller 257, 260, 263
I/O device controller 263 L
I/O port 257, 264
lambda 565
I/O Read 261
LAN 356, 358
I/O transfer instructions 281
Laser printer 319
I/O Write 261
Latency 312, 354, 473
IP address 383, 387, 393, 423
Laws of Boolean algebra 249
IP datagram 401, 455
Layered organisation 420
IP header 390
LDR Rd, <memory ref> 282
IP layer 422
Legal issues and opportunities 330
IP standards 447
length(list) 594
IPv4 387, 437, 438, 445, 448
Letter frequency attack 195
IPv4 address structure 387
Library programs 218
IPv6 423, 438, 448
Limitations of the one-time pad 213
IP version 4 437
Link between mathematical relation and relational model
IP version 6 438, 448
relation 490
IPython 565
Linking of information 335
Irrational numbers 11
Link layer 423
Irrelevancy 178
Linux Terminal Server Project (LTSP) 471
ISBN 105
List 592
Isolation level 537
List concatenation 592

Single licence - Abingdon School 605

 

List constructor operator 594 M

List element 592
MAC address 357, 423, 428
Listening socket 426
Machine code 220
List head 592
Machine code instruction 221, 284
List processing 592
Machine code instruction format 282
Lists in functional programming languages 592
Machine code language program 221
List tail 592
Machine code program 278
Load 288
Machine dependent 224
Load operation 271
Machine learning 333, 552
Load-Store architecture 284, 288
Magnetic hard disk drive (HDD) 323
Local Area Network (LAN) 356
Mail server aliasing 395
Localhost 426
Main components of a packet 389
Logical AND operation 233
Main memory 257, 260, 263, 311
Logical bitwise operators 298
Main memory RAM chips 263
Logical database model 486
Majority voting 100
Logical NAND operation 234
Man-in-the middle attack 380
Logical NOR operation 234
Manipulating digital images 110
Logical NOT operation 234
Manipulating digital recordings of sounds 118
Logical operators 523
Mantissa 61, 82
Logical OR operation 232
many-to-many 481
Logical OR Truth Table 233
many-to-one 481
Logical shift left operation 300
Map 587
Logical shift operations 300
MapReduce, 544, 547
Logical shift right 301
MAR 270, 275
Logical XOR operation 235
Mathematical relation 487
Logic gate 235
MatLab 107, 110
Logic gate circuit equivalent of a given Boolean expression
MBR 270, 274
238
Mean Time To Failure (MTTF) 545
Logic gates 231, 235
Measurement with real numbers 20
Logic of accumulation 333
mebi Mi 42
Logic of action 333
Media Access Control (MAC) address
Logic of information reflection 333
white list filtering 381
Loopback IP address 425, 426
mega M 42
Lossless compression 178
Memories for life 336
Lossy compression 178
Memory 271
Lost Update 537, 538
Memory Address Register (MAR) 270, 272, 275
Lost update problem 533, 534, 538
Memory bandwidth 311
Low-level programming language 220
Memory Buffer Register (MBR) 270, 271, 274
Low-powered laser diode 316
Memory controller 311
LSL Rd, Rn, <operand2> 282
Memory Data Register (MDR) 274
LSR Rd, Rn, <operand2> 282
Memory-mapped peripherals 264
LTSP 471
Memory Read 261
Memory Write 261

Single licence - Abingdon School 606

 

MEMS 136 NAT router 457

MEMS 3-axis gyroscope 136 NAT translation table 455
Message authentication 376 Natural numbers 1, 2
Message blocks 385 Negative flag 293
Message digest 412 Negative flag condition code (N) 276
Message integrity checks 376 Net ID 439
Messages 421 Network adapter 357
Metadata of bitmap 148 Network Address Translation (NAT) protocol 454
Microelectromechanical systems 136 Network ID 439, 444
MIDI 171 Networking protocols 420
MIDI Data bytes 172 Network interface card (NIC) 357
MIDI messages 172 Network layer 422
MIDI Status byte 172 Network topology 356
MIDI Voice Channel messages 172 Nodes 558
Missing rational numbers 71 Noise 164, 372
Mnemonic 221 Nominet 397
Modelling many-to-many relationships 492 Non-anonymous access 431
Modular arithmetic 189 Nonce 380, 410, 556
Modular arithmetic in daily life 190 Non-routable IP addresses 449
Modulus 189 Non-routable IP address space 454
Monitoring and protection 418 Non-routable IPv4 addresses 454, 457
Monochrome laser printer 319 Non-volatile 323
Moral action 330 Non-volatile storage 326
Most significant bit 61 No of clock cycles per instruction 308
Most significant data bit (MSB) 350 No of different arrangements of n bits 38
Motherboard 260 No of possible opcodes 282
MOVE 291 Normalisation 498
MOV Rd, <operand2> 282 Normalisation algorithm 88
MSB 350 Normalisation algorithm for negative mantissa 89
Multiple access 373 Normalisation in decimal 85
Multiple cores 309 Normalisation of floating point form 85
Multiplication of two unsigned binary integers 49 Normalisation where relationship is many-to-many 502
Musical Instrument Digital Interface(MIDI) 171 Normalising an un-normalised floating point binary
MVN Rd, <operand2> 282 representation 87
MVN Rd, <operand2> 299 Normalising relations to third normal form 506
NOR truth table 234
N
NOSQL database 463
NAND flash memory 327 NOT 299, 523
NAND logic gate wired as a NOT gate 255 NOT NULL 528
NAND truth table 234 NOT truth table 234
Napster 367 nslookup 396
NAT 454 Null 521
NAT-enabled router 454, 455, 457 Null value 496

Single licence - Abingdon School 607

 

Number 1 Packet forwarding 392

Number base 24 Packet switching 384, 392
Numbers with a fractional part 58 Packet transmission 386
Numeral 1 Packet transmission errors 390
Numeral systems 2 Pairwise Master Key (PMK) 377
Nyquist’s theorem 168, 169 Pairwise Transient Key (PTK) 377
Nyquist’s theorem and recording sound 170 Parallel bus 258
Parallel data transmission 345
O
Parallel interface 346
Object code 225 Parallella computer platform 310
Observed data 339 Parallel vs serial data transmission: 345
Octave 107, 110 Parameters 570
Odd parity 101 Parity bit 101, 347
One bit of information 36 Parity error 348
One-dimensional barcode 316 Partial application for three-argument functions 575
One-time pad 198, 201 Partial application for two-argument functions 573
one-to-many 481 Partial function application 572, 574
one-to-one 481 Passive RFID device 320
ONE-to-ONE mapping 222 Passive RFID tags 321
One-way function 406 Pattern matching 586
Opcode 279, 282 PC 270, 274
Opcode field size 282 PCI bus 360
Open architecture networking 384 PCI bus connector 360
Operand field 279 PCI slots 260
Operands 279, 281 Peak amplitude 165
Operating system software 218 Peer-to-peer architecture 366
Operation code 282 Peer-to-peer local area networks 363
Optical disc 325, 326 Peer-to-peer networking 362
Optical fibre 368 Perfect secrecy 208, 210
OR 299, 523 Perfect security 208, 211, 212
ORDER BY 522 Periodic oscillation 164
Ordering on more than one attribute 523 Periodic system 165
Ordering the result set 522 Peripheral devices 257
Ordinal number 18 Personal data 332, 333, 337, 339
ORR Rd, Rn, <operand2> 282 Pervasive surveillance 332, 333
Overflow 93, 275 Phase 165
Overflow flag (O) 276, 293 Photodiode 316, 317, 318
Oyster card 321 Photoelectric detector 316

P Photosensor 317
Physical bus network 358
P2P 362
Physical dimensions of printed images 145
Packet 385
Physical phenomenon 35
Packet based serial bus 312
PiNet 471
Packet filtering 400
Single licence - Abingdon School 608
 

Ping 401 Process 420

Pitch 164 Processor 257, 265, 270
Pixel 140 Processor and its components 270
Pixel-based graphics 140 Processor instruction set 280, 283
Pixels per inch (ppi) 145 Processor word length 271
Plaintext 185 Product 251
Platter 323 Product of sums 251
POP3 434 Program code and data are the same 268
Port forwarding 457 Program Counter (PC) 270, 271, 274, 282
Port mapping table 457, 458 Programmer-accessible registers 282
Port number 389, 425, 455 Properties 558
Possible binary codes for registers 284 Properties of vector graphic objects 159
POST 463 Protection of personal data 332
Post Office Protocol (v3) 434 Protocol 354
Postscript interpreter 155 Pseudorandom 203
Power consumption 473 Pseudorandom number generators (PRNGs) 203
Powers of 2 41 Pseudorandom numbers 203
PPI 144 Public exponent 407, 408
Practical obscurity 336 Public IP addresses 449
Precision 81, 83 Public key 411
Precision of the measurement 81 Public key encryption 405
Precision versus range 83 Public key/private key encryption 405
Predictive model 554 Public key/private key encryption algorithm 411
Prepend 595 Public modulus 407, 408
Pre-shared secret key 377 Pulse Amplitude Modulation (PAM) signal 129
Pre-shared secret key (PSK) 376 Pulse Code Modulation (PCM) signal 130
Primary key 477, 488 Pure function 555
PRIMARY KEY 528 Purpose and function of DNS 395
Prime number 406 Purpose and function of DNS Servers 395
Printer resolution and Dots per inch (DPI) 147 Purpose of DHCP system 453
Printing a bitmapped image on paper 147 Purpose of start and stop bits 348
Privacy 340 PUT 463
Private address spaces 449
Q
Private exponent 407
Private Exponent 408 QR code 317
Private IP addresses 449 Quantisation 125, 133
Private IPv4 address ranges 449 Quantisation distortion 133
Private IPv4 address spaces: 449 Quantisation error 133
Private key 411 Quantisation noise 134
Private key encryption 404 Quartz crystal oscillator 260, 273
Private port 427 Querying a database 519
Problems occur if a relation is not fully normalised 515 Quotient 193
Procedural abstraction 569, 570 Quotient and remainder theorem 193

Single licence - Abingdon School 609

 

R Relationship name 481

Relative error 75
Radio frequencies (RF) 320
Relatively prime 406
Radio frequency identification (RFID) 320
Remainder 193
RAM 257, 260, 263, 311, 323
Repeating group 498, 499, 502
Random Access Memory 263
REPL 581
Randomness 202
Replay attack 380
Range for a given number of bits 57
Representational State Transfer 461
Range of numbers in unsigned binary in n bits 46
Representation of two’s complement
Rational number approximation to a real number 22
floating point binary 61
Rational numbers 6
Representation problem 70
Rational numbers as recurring decimals 9
Request message 432
Rationals as terminating decimals 8
Request To Send (RTS) 375
Raw data 556
Requirements analysis 474
RAW image file 318
Reset 261, 293
RC4 207
Resolution of a bitmap 144
Read Eval Print Loop (REPL) 581
Resolution of an ADC 132
Reading file contents byte by byte 114
Resolution of computer displays 145
Read-write head 323
Resolving many-to-many relationships 484
Real number line 14
Response message 432
Real numbers 14
REST 461, 462
Receiver 349
REST request 461
Record lock 536
REST web services 461
Reduce 589
Retina displays 145
Reduced Instruction Set Computer 283
Retrieving data from a single table 519
Reduced licensing costs 473
Retrieving data from multiple tables 520
Reduce function 548
RFC 447
Redundancy 100, 178
RFID 320
Redundancy theorem 251
RFID device 320
Redundant data 506, 515
RFID price smart tag 321
REFERENCES 532
RFID reader 320
Referential integrity 521
RFID system 320
Regional Internet Registry 448
RFID tag 320
Regional registries 448
RFID tag characteristics 321
Registers 270, 274
RFID transponder 320
Registers always involved in the Fetch-Execute cycle 278
RISC 283
Relation 487, 502, 519
RJ45 connector 359
Relational database 486
RJ45 socket 360
Relational database model 486
RLE 179
Relational model 486
RLE packet 179
Relational modelling 486
Role of a compiler 225
Relational operators 522, 584
Role of a foreign key 521
Relationship 480, 558, 559
Role of an assembler 225
Relationship between bit rate and bandwidth 355

Single licence - Abingdon School 610

 

Role of an interpreter 226 Sector 324

Role of an operating system 219 Sector address 324
Role of a router 384 Secure shell 409
Role of interrupts 304 Secure Shell 435
Role of interrupt service routines (ISRs) 306 Secure Sockets Layer (SSL) 434
Role of MAC addresses 428 Securing wireless networks 376
Role of packet switching 384 Security 376, 472
Role of sockets in TCP/IP 425 Seed 203
Role of the four layers of TCP/IP protocol stack 420 Seek time 325
ROLLBACK 539, 540 SELECT 463, 519
Rotational delay 325 Semi-structured data 549
Rounding errors 69 Sensor 135
Rounding errors in floating point 74 Sensor platforms 137
Rounding errors in signed fixed point 74 Separation of Concerns 461
Rounding off 22, 73 Sequence number 389
Routable IP addresses 449 Serial bus 258, 259
Router 388, 423, 437 Serial data transmission 344
Routing 392 Serial interface 346, 348
Routing algorithms 392 Serialisation 537, 540
R programming language 553 SERIALIZABLE 537, 539
RS232/RS422 347, 348 Serial Peripheral Interface bus (SPI) 347
RSA 410 Serial ports 347
RSA encryption 408 Serial to parallel conversion 348
RSA public key/private key cryptosystem 406 Server application 461
RTS 375 Server-based local area network 364
Rules of significant figures or digits 82 Server-based network 364
Run Length Encoding (RLE) 179 Service name 428
RX 350 Service Set Identifier 370
Service Set Identity 370
S
Set 293
Sample resolution 166 SET 526
Sampling 139, 166 SET SESSION CHARACTERISTICS 538
Sampling and quantisation 139 SET TRANSACTION CHARACTERISTICS 538
Sampling a waveform 166 SHA 410
Sampling rate 166 Shared private key 404
Scalable 546 Shared transmission medium 258
Scalable storage 544 Shift cipher 185
Scanned and digital camera images 145 Side-effect 555
Scheduling algorithm 542 Sign 275
Screen resolution 146 Signal 125
SDRAM 311 Signal to Interference + Noise Ratio (SINR) 372
Secondary storage devices 323 Sign bit 61
Second Normal Form 506 Signed binary using two’s complement 52

Single licence - Abingdon School 611

 

Significant digits in floating point representation 82 SPI 347

Significant figures or digits 81 Spinnaker Human Brain project 260
Simple Mail Transfer Protocol 434 Spyder 117
Simplicity 557 Spyder python 114
Simplifying Boolean expressions 248 SQL 519
Simultaneous access 533 SQL Constraints 528
Sine waveform 165 SQLite 530
Single Instruction Multiple Data (SIMD) 309 SSD vs other flash-based devices 327
Single logical address space 387 SSEM 267
Single-Valued Fact (SVF) 499 SSH (Secure Shell) 409, 435
Sinusoidal 164 SSID 370, 371

Sinusoids 165 SSID broadcast disabled protection 381

Skew 312, 345 SSL 434

Skype 367 Stack machine 228

SMALLINT 530 Stack Pointer register 282

Smart labels 320 Stages of normalisation 510

Smart sensors 136 Standard application layer protocols 430

SMTP 434 Start bit 347, 348, 349

Star topology 356
Social Engineering 418
Stateful inspection packet filtering 402
Social (ethical) issues and opportunities 330
Stateless 462
Social processes 338
Statelessness 555
Socket 425
Stateless packet filtering 401
Socket API 420
Software 216 Status register 270, 275, 293

Software and their algorithms embed moral Stop bit 347, 348, 350

and cultural values 338 Storage media and typical applications to which

Software can produce great good but with it comes they can be put 329

the ability to cause great harm 340 STORE 290

Software maintenance 472 Stored program computer 267

Solid-state disk (SSD) 323, 326 Stored program concept 267, 268

Sound 116, 164 Store operation 271

Sound and text files 120 Stream cipher 207

Sound waves 129 Streamed data 543

Source code 225 STR Rd, <memory ref> 282

Source hardware address 423 Structured data 549

Source IP address 389 Structured Query Language (SQL) 519

Sources of interrupt 305 Structure of a machine code instruction 281, 283, 284

Spatial resolution 145 Structure of a simple computer 257

Special purpose applications software 217 Stuxnet worm 416

Special purpose registers 270, 274 Subnet 439

Speech 126 Subnet address 439, 454

Speed of access of storage media 328 Subnet mask 444

Speed of calculation 84 Subnet masking 444

SUB Rd, Rn, <operand2> 282

Single licence - Abingdon School 612

 

SUBTRACT 293 Thick-client 470

Subtraction in two’s complement 56 Thick-client computing 470
Sum 251 Thick-client network 470
Sum of products 251 Thin-client 470
Supervisor 275 Thin-client computing 471
Supervisor mode flag (S) 276 Thin-client networking 471
Surface address 324 Thin-client system 361
SVG image 157 Thin- versus thick-client 472
Switched Ethernet 358 Third Normal Form 506
Symbol as information carrier 34 Timbre 167
Symbolic name 221 TIME 530
Symmetric encryption 404 Time-division multiplexed synchronous serial
Synchronisation 125 data transmission 348
Synchronous data transmission 346 Timestamp ordering 540
Synchronous dynamic random Token 180
access memory (SDRAM) 311 Topology 356
Synchronous serial data transmission 346 Track 324
System bus 258 Track address 324
System clock 270, 272
Traditional database systems 547
System On a Chip (SOC) 283 Traditional (von Neumann) computer system
System or Well-known ports 427 address bus 262
System programs 216 Traditional (von Neumann) computer system
System software 216 control bus 261
Traditional (von Neumann) computer system
T
data bus 259
tail(list) 593 Transaction 540
TCP connection 426 Transfer ACK 261
TCP header 390 Transistor-Transistor Logic 349
TCP/IP 420 Transmission Control Protocol/Internet Protocol
TCP/IP protocol stack 420 (TCP/IP) protocol 420
TCP/IP protocol suite 420 Transmission Control Protocol (TCP) 383, 421
TCP/IP socket 421, 424, 426 Transmitter clock 349
TCP/IP stack 423 Transport layer 421
TCP segments 421 Transport Layer Security (TLS) 207
TCP socket 430 Transport Layer Security (TLS) protocol 410
TCP three-way handshake 401 Trie 181
Telnet 435, 461 Trojans 413, 414
Terminal server 471 Truncation or rounding down 73
Text, digitised sound and images: 121 Truth table equivalent of a logic gate circuit 236
The differences between compilation and interpretation 226 TTL 349
The issue of scale 336 tuple 494
The Transmission Control Protocol/Internet Protocol Twisted pair 359
(TCP/IP) protocol 430, 437, 444, 447, 449, 451, 457, 470 Twitter Search API 550
The Web means the end of forgetting 336 Twitter’s Streaming API 550

Single licence - Abingdon School 613

 

Two’s complement 52 Vector graphic primitives 156

TX 348 Vector graphics 153
Types of networking between hosts 362 Vector graphics uses 163
Types of program translator 225 Vector graphics versus bitmapped graphics 160
Velocity 543, 550
U
Vernam cipher 205, 206
UART (Universal Asynchronous Receiver Transmitter) 348 Vertical scaling 547
Unconditional branch 296 Virtual machine 228
Underflow 91 Viruses 413, 415
Unicode 96, 98 Visual X - Toy 276
Uniform addressing scheme 437 Volatile 263, 323
Uniform interface 462, 463 Volatile environment 306
Uniform Resource Locator 393 Volume 543, 544
UNIQUE 528 Volunteered data 339
Uniqueness of representation 85 von Neumann architecture 258, 263, 265
Units of storage 40 von Neumann computer 267, 268
Units of the decimal and binary numeral systems 69
W
Universal Asynchronous Receiver Transmitter 348
Universality of NAND gates 255 Web server 457
Universality of NOR gates 256 Websocket protocol 469
Universal Product Code (UPC) 316 Well-known ports 427
Unix DNS tool dig 396 WEP 207
Unnormalised 498, 499 WHERE 519, 521
Unnormalised relations 506 Whole number 5
Unsigned binary 45 Why normalise databases? 505
Unsigned binary arithmetic 48 WiFi 368
Unstructured data 549 Wireless interface 370
UPDATE 463, 519, 526 Wireless LAN 368
URI 461 Wireless network adapter 370
URL 393, 433, 461 Wireless networking 368
USB 348 Wireless propagation characteristics 369
User ports 427 Wireshark 379
Using a subnet mask 444 WLAN 368
Using IP address prefix to route an IP datagram 441 Word length 271
Using SSH for remote management 435 Workstation 359
Utility programs 218 Worms 413, 414
WPA/WPA2 376
V
Writing functional programs 581
Value free 338 ws:// 469
VALUES 526 wss:// 469
VARCHAR(n) 530
X
Variable 569
XML 461, 463, 468, 549
Variety 543, 549
XOR 299
Vector graphic image 153
Single licence - Abingdon School 614
 

XOR truth table 235

Zero flag 293, 299

Zero flag (Z) 275, 276

Single licence - Abingdon School 615

 

Single licence - Abingdon School

A Level
Computer Science
FOR AQA Unit 2

This textbook covers Unit 2 of AQA A

Level Computer Science in an accessible
and student-friendly way.
Additional resources to accompany the
text will be available from our web site Cover picture © Dr K R Bond
www.educational-computing.co.uk. Blue Mountains, New South Wales,
Please note that these resources have not Australia
been entered into the AQA approval
process. Only the student textbook has About the author
been approved. Kevin Bond has many years of A
Level Computing/Computer Science
teaching and examining experience.
He has worked at the interface
between Science, Computer
This book has been Science and Engineering, first as a
approved by AQA Research Scientist, then as a Senior
Development Engineer for a major
Defence Contractor and then as a
Senior Systems Analysis for a major
Telecommunications company.
He has written several best-selling
textbooks supporting A Level and AS
Level Computing.

Educational Computing Services Ltd

42 Mellstock Road
Aylesbury
Bucks
HP21 7NU
ISBN 978-0-9927536-6-5
Tel: 01296 433004
www.educational-computing.co.uk

Single licence - Abingdon School