<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Cay on Medium]]></title>
        <description><![CDATA[Stories by Cay on Medium]]></description>
        <link>https://medium.com/@caybo?source=rss-8c2db2c6110b------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*d4WtyXanyy960FuF3w0Qhg.jpeg</url>
            <title>Stories by Cay on Medium</title>
            <link>https://medium.com/@caybo?source=rss-8c2db2c6110b------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sun, 22 Nov 2026 20:54:02 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@caybo/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[That Time I Got Reincarnated as a HENNGE GIP Intern]]></title>
            <link>https://medium.com/henngeblog/that-time-i-got-reincarnated-as-a-hennge-gip-intern-ea76d7ddfcd9?source=rss-8c2db2c6110b------2</link>
            <guid isPermaLink="false">https://medium.com/p/ea76d7ddfcd9</guid>
            <category><![CDATA[internships]]></category>
            <category><![CDATA[japan]]></category>
            <category><![CDATA[gip]]></category>
            <category><![CDATA[hennge]]></category>
            <category><![CDATA[software-engineering]]></category>
            <dc:creator><![CDATA[Cay]]></dc:creator>
            <pubDate>Fri, 29 May 2026 02:01:01 GMT</pubDate>
            <atom:updated>2026-05-29T02:01:01.071Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="View from 11F with illustrations of Shobi, Eyad, and Cay" src="https://cdn-images-1.medium.com/max/1024/1*my57E9WJV8NJuv94mzj2zw@2x.jpeg" /><figcaption>The view from HENNGE’s 11th floor office, feat. Shobi, Eyad, and me (Cay!)</figcaption></figure><p>Hi! My name is Cay, pronounced ケイ (or カイas in <em>isekai</em> if you prefer).</p><p>One day I was driving to Walmart through the backwoods of Florida; the next I was hauling two suitcases into the bustling hive of Shibuya, or as it once was called, Bit Valley. The next six weeks would pass in a blur, but they’d be among the best of my life.</p><p>I arrived in Tokyo four days before the internship started and met up with the other interns (make sure to check out Shobi’s and Eyad’s stories too!). During that time, I visited Inokashira Park, a place with an inexplicably magical vibe. I later learned it’s also a popular breakup spot for couples, so maybe don’t take your boo. We also watched <em>Superman </em>dubbed in Japanese (I understood nothing except that the dog was terrifying, highly recommend).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*RGeLFu3dmD40pW1dlB800A.jpeg" /><figcaption>Inokashira Park</figcaption></figure><h3>The Journey Begins</h3><p>When the first day at HENNGE finally arrived, <em>surreal</em> felt like an understatement. Overwhelmed by the number of names, teams, and floors, but I was ecstatic. I finally met the people I had only talked to over Zoom, enjoyed a Dr. Pepper from the world’s only Dr. Pepper vending machine, and even appeared briefly in the background of a television segment filming HENNGE’s “Welcome Café.”</p><p>We were given a bag of essential goodies and got set up right away. Interns don’t have to worry about bringing anything; your phone, laptop, and even umbrella are all provided!</p><p>I quickly learned just how open the office is and the collaborative environment it fostered; no assigned desks and a variety of workspaces ranging from completely quiet (the Forest) to more social and dynamic areas (like the Café). Don’t be surprised if you find yourself working beside Ogura-san, the CEO. My favorite spot was the 4F lounge (sometimes the couch).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*kOwu0tStqQNALgPWCqhW1Q@2x.jpeg" /><figcaption>View from the couch on 4F</figcaption></figure><h3>Bukatsu</h3><p>If you’ve read any other articles by previous GIP interns, you’ve probably heard of the bukatsu. They are company clubs centered around any interest you can think of.</p><p>Easy-board-game-bu is one such bukatsu that happens every Wednesday night. One game we played really tested our Japanese skills: <em>Gentosha Katakanashi</em>. The idea was to describe a word written in katakana without using any other katakana words. Imagine trying to explain a peanut butter and jelly sandwich without mentioning bread, peanut butter, or jelly; that’s what it felt like. Thankfully, I was allowed to sneak in some English when stuck. At one point, someone got the word <em>repertoire</em>. I can’t even describe that properly in English; truly, chapeau!</p><p>Another interesting bukatsu was the diversity-expedition-bu. This club tries new cuisines, and this time, we tried a Czech restaurant: popcorn soup (surprisingly good!), curry, bread dumplings, and more. This somehow sparked a heated (and unresolved) debate on what qualifies as soup, salad, or sandwich.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*LHXl5lnAnk7eLKnEAx-l7w.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*oXcWKk8SrkkczL9yD3v3Ww.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/600/1*JS6ueMu4fpg1nvpUuqNKTA@2x.jpeg" /><figcaption>Karaoke night, popcorn soup, and an *<strong>Epic</strong>* musical theater night</figcaption></figure><p>Bukatsu for trampoline, rock climbing, Japanese study, sweets, karaoke, musical theater, nihonshu, and many, many more also exist. These are wonderful opportunities for not just trying new things but making genuine connections. You can even start a new bukatsu or restart an inactive one.</p><h3>Training Montage</h3><p>The part you’ve (probably) been waiting for: The Training.</p><p><strong>Assignment 1 — Full Stack App</strong></p><p>In this assignment, we were each tasked with creating a social media web app. Stack: TypeScript, Fastify/Express, React/Vue, DynamoDB, S3, and Caddy (or Nginx if you enjoy pain). Coming from an SQL background, DynamoDB required a whole new way of thinking. With plenty of help from docs, the other interns, and my mentor, I eventually got the hang of it, and learned a ton about architecture, prioritization, and teamwork along the way.</p><p><strong>Assignment 2 — DevOps</strong></p><p>My first real experience with AWS. This was a cool opportunity to deploy the application we made in assignment one in a production-ready environment. On the first day at HENNGE, Ogura-san told us to “FAIL, FAIL, FAIL — and sometimes succeed”<em>. </em>Well, I did just that! Pro tip: things will break, and that’s okay.</p><p><strong>Assignment 3 — [Redacted]</strong></p><p>Hint: you get to work on <em>very</em> cool stuff. Shout-out to my mentors for the thorough code reviews, quick debugging skills, and good times.</p><p>Over the six weeks, we were also introduced to various teams within HENNGE, where we learned about their work and had the chance to ask questions. It was inspiring to see the passion the teams had and the daily challenges they face.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/400/1*Lef9m55GF6O5V7jyQdWVDw@2x.jpeg" /><figcaption>Mochi, simultaneously the world’s best motivator and distraction</figcaption></figure><h3>Monthly Technical Session</h3><p>The Monthly Technical Session (MTS) is a time for developers to get together and present technical topics of interest to each other, and interns get to participate. At first, MTS sounded scary (I’ve taken zero public speaking classes). But by the time it comes around, you’ve already done multiple presentations, so it’s pretty chill.</p><p>You get to nerd out on any technical topic you like for about 10–15 minutes, learn about other cool tech and get inspired, and have food and beer afterwards. Seriously, what’s not to like? (Sorry for eating all the Krispy Kreme 🙈)</p><h3>GIP Lunches (a.k.a. Free Food)</h3><p>Interns get to have around twelve free lunches, each with two or three different full-timers. It’s a great chance to get to know people in the company, and I got to try stuff I probably never would have otherwise.</p><p>Also if you’re vegetarian/pescatarian, no problem. Pre-compiled vegetarian friendly options exist, and at konbini/supermarkets just avoid 肉 (meat), 魚 (fish), and 鰹節 (dried fish flakes).</p><p>Also: interns got <strong>one free coffee per day</strong>. Needless to say, I had many hojicha and matcha lattes.</p><h3>Side Quests</h3><p>Weekends and (occasionally) evenings were for exploring Tokyo. Sorry for the dump, but here were some of my favorite adventures:</p><p><strong>A Night in Asakusa: </strong>I tried the best monjayaki, saw Senso-ji temple, and lost a couple thousand yen in a Taito game center (99% of players quit before they win big and I’m not the 1%).</p><p><strong>Day trip to Yokohama:</strong> I got a melon soda float from McDonald’s and laid in a park where I saw more Shiba Inus than I ever have in my life. I highly recommend getting one or all of the bike rental apps to get around (you can use your work-provided phone’s Japanese number to sign up!). Later, I tried mayo corn pizza with Shobi before having a near-miss with the last train. A tip: know when the last train runs, or you might be stuck sleeping on the street (and possibly wake up with ten water bottles around you).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*oQ8djjC4yXKxl8xs8pvpEw.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7WvOynZiGixu8RUpmlGIIw.jpeg" /><figcaption>View of the Yokohama skyline from Zonohana Park (left) and Yamashita Park (right)</figcaption></figure><p><strong>ESCape to Enoshima Island (the Beach Episode):</strong> We tried cheese kare pan (mind blowingly good), survived shrimp crackers (never again), and still regret skipping the denim ice cream (Is it Japanese denim? Levi’s? Just blue?).</p><p><strong>Kamakura</strong>: Big Buddha :) (Did you know you can go inside it?) We also got to see the blood moon rise over the Pacific.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*anDt7IrEkOr7CKJRAMHRdg.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*G6zJCZ2TTSnxZ9H41Sdvmg.jpeg" /><figcaption>Bridge to Enoshima Island (left) and Kamakura (right)</figcaption></figure><p><strong>Catching a Matsuri</strong>: I went to Shimokitazawa for the thrifting and stayed for the odori matsuri<em>, </em>or dance festival. These festivals happen around Obon week, so if you’re in Japan during the summer (apply to GIP hehe), it’s the perfect time to experience one! I also recommend going to see hanabi if you get a chance.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*gJcyEACIr-frYQJ0zVSP8Q.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*C12Z-kkWL5L95xNPbs9c5A.jpeg" /><figcaption>Chōfu City Fireworks (left) and Shimokitazawa Bon Odori (right)</figcaption></figure><p><strong>Fun Free Museums</strong>: The Advertisement Museum, Parasitological Museum (if you dare), Toei Animation Museum, Museum of Package Culture.</p><p><strong>Random Food Adventures:</strong> Pufferfish (survived), gokiburi sake (yes, that’s cockroach), ice cream sundaes with questionable flavors (basil was the best), and what might be the best vegan ramen in Tokyo (Jikasei MENSHO, though you didn’t hear it from me)</p><p><strong>The Hunt for Goshuin</strong>: In Japan, you can collect goshuin: stamps from train stations, shrines, and even anime-related spots. My goshuincho (stamp book) is still embarrassingly sparse, but hey, it’s a work in progress.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*-TTYz95DmpC5mLsNInnX6w.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*t9Cdf1pPl9hm_D9aaLz5jQ.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*yMrbitnvGkdZjFoW5lJpKg.jpeg" /><figcaption>Advertisement Museum (left), Bunkamura Gallery Exhibition (center), Shibuya Pixel Art Festival (right)</figcaption></figure><p>For my stay, I somehow scored a room at a Sakura House in Shibuya with only an eight-minute walk to the office. Needless to say, I loved it and met some amazing people. Share houses are their own kind of adventure; people constantly coming and going from all over the world. What struck me most about living in Tokyo was how effortlessly the city brings together people from so many different cultures. HENNGE was no different. Every day felt like a small exchange of perspectives: over lunch, on the train, or in the share house kitchen.</p><p>Shibuya itself is chaos in the best way; you never know who you’ll run into. I actually bumped into one of my favorite musical artists randomly in a Family Mart, so basically I can die happily now. And it never really gets dark. One night I woke up at 2 AM, panicked that it was 2 PM and I had missed work, only to realize my window was just aggressively bright.</p><p>Biggest disappointment? Learning that melon pan does <em>not</em> contain melon.</p><h3>De-incarnation</h3><p>This internship was one of the best experiences of my life and I am incredibly grateful to have had this opportunity. Thank you to my fellow interns for the laughs, debugging help, and adventures; you guys made this experience truly memorable.</p><p>If you’re considering interning at HENNGE, just do it. And if you’re facing a bit of self-doubt, do it anyway. Before I de-reincarnate, I’ll leave you with wisdom from a T-shirt I found at a secondhand shop:</p><figure><img alt="Make good choices so you don’t have waverer tomorrow." src="https://cdn-images-1.medium.com/max/1024/1*1tRncL37ccJEfEijPI9LDQ@2x.jpeg" /><figcaption>“Make good choices so you don’t have waverer tomorrow.”</figcaption></figure><p><em>Cay, from the US, interned with the </em><a href="https://recruit.hennge.com/en/gip/"><em>HENNGE Global Internship Program</em></a><em> from August 1 to September 12, 2025. They are currently a software engineer at HENNGE.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ea76d7ddfcd9" width="1" height="1" alt=""><hr><p><a href="https://medium.com/henngeblog/that-time-i-got-reincarnated-as-a-hennge-gip-intern-ea76d7ddfcd9">That Time I Got Reincarnated as a HENNGE GIP Intern</a> was originally published in <a href="https://medium.com/henngeblog">henngeblog</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Be Kind, Please Rewind: Preserving Memories through VHS Decode]]></title>
            <link>https://medium.com/henngeblog/be-kind-please-rewind-preserving-memories-through-vhs-decode-885acd146fde?source=rss-8c2db2c6110b------2</link>
            <guid isPermaLink="false">https://medium.com/p/885acd146fde</guid>
            <category><![CDATA[open-source]]></category>
            <category><![CDATA[analog]]></category>
            <category><![CDATA[vhs-decode]]></category>
            <category><![CDATA[digitization]]></category>
            <category><![CDATA[monthly-technical-session]]></category>
            <dc:creator><![CDATA[Cay]]></dc:creator>
            <pubDate>Thu, 02 Apr 2026 02:01:06 GMT</pubDate>
            <atom:updated>2026-04-02T02:01:06.712Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ybWjhjVY0VrI8GHW_Iq6eA.png" /><figcaption>This was originally a talk given at the Monthly Technical Session #133</figcaption></figure><p>Did you ever try sticking random things into a VCR as a kid to see what happens? If not, that’s probably for the best. But maybe you’re thinking: what is this person even talking about?</p><p>VHS is a video format that preceded the era of DVDs. It was phased out in the mid 2000s with the last VHS tapes to be released commercially being <em>A History of Violence</em> (2006) and <em>Cars </em>(2007). These tapes didn’t just bring the latest blockbusters into your home however; they also served as a way to record and preserve personal home videos.</p><p>On average, VHS tapes degrade 10-20% over 10 to 25 years. Their magnetic oxide layer gradually demagnetizes, while heat, humidity, and physical stress (<em>say, sitting for years in a hot storage container</em>) accelerate the breakdown of both the tape and the data encoded on it.</p><p>For many families and communities, including mine, these fragile tapes hold priceless memories that can’t be replaced. That’s why I set out to preserve them with the highest fidelity possible before they’re lost to time, a process known as <strong>digitization</strong>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/760/1*IbbjjVjuZy_rPnGsX-X4RQ.png" /><figcaption>A shot of Surya Bonaly from my mom’s recording of the 1998 Nagano Winter Olympics, captured and decoded using VHS-Decode.</figcaption></figure><h3>USB capture vs. VHS-Decode</h3><p>The first digitization method I encountered was a USB video capture device that connects a VCR’s analog output to a computer’s USB port.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*NBJA1jJ-aV1UKRf_IG7kFg.png" /><figcaption>Traditional capture workflow (VHS -&gt; VCR -&gt; Capture Device -&gt; PC)</figcaption></figure><p>This workflow is straightforward and affordable, but it relies on decades-old hardware and analog signal processing. Any imperfections in the VCR player will be baked into the final video quality. Since the capture device records the output as it appears on the VCR, you have little control over the raw signal or the ability to apply modern software-based corrections that can significantly improve restoration and preservation.</p><p>If you’re not too concerned about quality (which honestly I’m not sure why you wouldn’t be) then this setup is okay to use. But if you’re looking for the best quality, the best way is to go straight to the source.</p><p>Enter… <a href="https://github.com/oyvindln/vhs-decode"><em>VHS-Decode</em></a>. This project, and FM RF archival in general, represents a fascinating intersection between analog signal processing and modern open-source software development.</p><p>Instead of capturing the video output from a VCR as in a traditional setup, VHS Decode samples the <strong>raw RF signal </strong>and decodes it using software.</p><p>To put it another way, it’s like the VCR is singing a song (making the picture), but it might sing out of tune if it’s old or broken. Instead of recording the song the VCR sings, RF archival records the grooves themselves - the &quot;notes&quot; if you will. Now, imagine ten years from now there’s a brand-new record player with amazing sound quality. Because we recorded those raw grooves (the RF signal) back then, we can always go back and “play” them again using that better player; meaning we can decode the video with improved software, even long after VCRs are gone.</p><p>It&#39;s a lot like saving the <strong>source code </strong>for a program, not the compiled output. You can recompile it later with better compilers (decoders) and optimization settings, instead of being stuck with the old executable forever.</p><p>VHS-Decode has gained traction in the last year through communities on GitHub and YouTube, and there is even an active Discord server (which I highly recommend joining if you get stuck and need help, like me 🙈). Big shout out to the individuals who maintain and contribute to the project.</p><p>Still, the process can be a bit intimidating, so I will attempt to demystify this process as much as I can.</p><p><em>Please note that while the FM RF archival method also works for SVHS/Video8/Hi8/Beta/LaserDisc etc., this article will only cover VHS.</em></p><h3><strong>Why you shouldn’t stick random things into a VCR (or, How VHS Works)</strong></h3><p>Have you ever wondered how a VCR actually works? What kind of magic was hiding behind that little trap door?</p><p>Firstly, what is VHS? VHS stands for “Video Home System”, and it was first released by the Victor Company of Japan in 1976. It did exactly as the name describes: brought video recording and playback capabilities directly into the home.</p><p>VHS tapes store data not as 1s or 0s, but as frequency-modulated (FM) radio-like signals on the magnetic tape surface. Frequency modulation is a just fancy term for changing the frequency of a wave to encode information.</p><p>The FM signal encodes the following components:</p><ul><li><strong>Luma </strong>(Y): Brightness information. Lower frequencies correspond to darker areas, higher frequencies to lighter areas.</li><li><strong>Chroma </strong>(C): Color information.</li><li><strong>Hi-Fi Audio </strong>(Left / Right): Stereo FM audio tracks.</li></ul><p>The VCR then separates these components based on their frequencies, known as <strong>demodulation</strong>.</p><p>We can examine the FM signal’s spectrogram to see these individual frequencies:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*OY4y5RJtzSJoxitZQBcSdg.png" /><figcaption>RF signal spectrogram analysis</figcaption></figure><p>You may be wondering, why is chroma encoded separately from the luma if they are both used to create the picture? There are two main reasons:</p><p>The first has to do with maintaining <strong>backwards compatibility </strong>with older black and white televisions, much like how modern websites are still designed to work in older browsers such as Internet Explorer.</p><p>The second reason is to take advantage of the <a href="https://en.wikipedia.org/wiki/Chroma_subsampling">human eye’s lower sensitivity to color than to brightness</a>.</p><p>And in order to do those things, it uses <strong>layered coding </strong>to create a composite signal. We still use the same idea today in digital systems: data streams, MPEG transport streams, MP4 containers, and even HTTP/2 frames all bundle multiple layers of information into one signal.</p><p>For an awesome explanation on chroma encoding, see this <a href="https://www.youtube.com/watch?v=3evgy0N8Sic">video</a>.</p><p>There’s also a linear, mono audio track recorded along one edge of the tape. Unlike the Hi-Fi stereo audio track, which is mixed with the video’s RF signal, the linear track is a low-frequency analog waveform. It carries the actual frequencies of human hearing, just like an ordinary cassette tape. No special processing is needed to capture it; you can simply connect the VCR’s red and white audio outputs and record it directly. This makes it an easy way to extract audio from a tape without dealing with the complexities of Hi-Fi decoding.</p><p>So, unless it’s carrying a carefully modulated mix of luma, chroma, and audio frequencies, you really shouldn’t stick it in the VCR. It’s not magic, it’s just magnets.</p><h3>Why analog video needs Time Base Correction</h3><p>After the VCR demodulates the signal, it still isn’t ready for stable playback. It needs something called <em>Time Base Correction (TBC).</em></p><p>To understand why, let’s look at how that RF signal actually becomes a picture.</p><p>Analog video is like a sheet of paper put through a shredder. For <a href="https://kodakdigitizing.com/blogs/news/how-do-i-know-if-my-video-is-pal-or-ntsc">NTSC video</a> (the broadcast standard used in North America and Japan), each frame is made up of 525 horizontal scan lines, and each line contains several parts: a <strong>horizontal sync pulse </strong>that signals the start of the line, the <strong>video information</strong> itself, and a <strong>color burst </strong>that gives the system a base reference for color so it can keep each color’s hue and shade consistent (early NTSC lacked the color burst, earning it the moniker “Never Twice Same Colour”). In old CRTs, horizontal sync pulses dictate when the electron beam (or pixels in digital displays) should return from the right side of the screen to the left for a new line.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*OxpKlZdrmwE1eUeNxHSIoQ.jpeg" /><figcaption>Close-up of a CRT Television where you can see the individual scan lines.</figcaption></figure><p>To see where these individual parts of a scan line appear in the signal itself, let’s look at the spectrogram of the RF signal again in Audacity.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/960/1*wG79FuYbUw5U3OH_oR8wkw.png" /><figcaption>Components of a scan line in an RF signal</figcaption></figure><p>These parts also appear in the frame itself. We can see the horizontal sync on the left, the color burst directly after it, and the video portion. Of course, only the video portion of the frame makes it to the final image.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/910/1*LIF94hZbr4YSBgI3zojuNQ.png" /><figcaption>The parts of an NTSC frame</figcaption></figure><p>As mentioned earlier, the NTSC video standard defines a 525-line vertical resolution. However, only 480 of those lines are used for transmitting video information. The extra 45 lines along the top are used to carry control codes (such as interlace information), closed captions, and other similar non-video content. For this system to work, each line must begin at precisely the right time. Now, imagine what happens when the timing begins to drift.</p><p>The VCR is a mechanical system. It relies on a spinning head drum and a moving tape. Over time, motors wear out, belts slip, tape stretches or degrade. All of these moving parts spell out instability. This means the <strong>horizontal sync pulse</strong> that signals the start of the line might not appear at exactly the right moment, resulting in unstable, jittery video. If we think of the tape like a vinyl record again, each groove corresponds to one scan line. If the grooves aren’t read at a consistent timing, the audio will sound off-tempo.</p><p>To correct this, the VCR applies <em>Time Base Correction (TBC)</em>. TBC works like a metronome, re-timing the lines so they align perfectly in the final image. In order to do that, it creates a buffer that temporarily holds each video line’s sync pulse and then releases it at the correct time.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*_BTG8xSl7CcZRh9ppcTFtA.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*_hDKxMqs38ulIGixzWEhEA.png" /><figcaption>Out of sync Analog Frame (left); Time Base Corrected Frame (right). Source: VHS Decode Wiki</figcaption></figure><p>Unfortunately, most built-in TBC units in VCRs aren’t very reliable, and dedicated full-frame TBC units are aging and becoming harder to find. The most popular model, the Datavideo TBC-1000, has been out of production for years. At the time of writing, a used one can cost around $1,300 on eBay (that’s like roughly fifteen Shinkansen rides between Tokyo and Kyoto!).</p><p>Sure, you can use a cheaper VCR to capture your videos, but while it might save you money upfront, the real cost shows up in the quality you lose during the process; in 2021, software-based TBC solutions surpassed the performance of traditional in-deck hardware.</p><p>If you want to know more about how a VCR itself works, check out <a href="https://youtu.be/lRNv4uM6auI">this video</a>!</p><h3>Enter VHS-Decode: a different way to capture VHS</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/470/1*A4ulrXxog7C9F9zwkNYfRw.png" /></figure><p>VHS-Decode doesn’t require you to have a top-notch VCR or TBC. It just needs <a href="https://github.com/oyvindln/vhs-decode/wiki/004-The-Tap-List">a VCR that lets you “tap” the RF signal</a>. I managed to buy one off of Ebay for just $30 USD.</p><p>So, how does it actually work?</p><p>VHS Decode “taps”, or samples, the RF signal and decodes it using software. To recap, the RF (radio frequency) signal refers to the raw, un-demodulated video and audio signals directly recorded onto magnetic tape.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*jdBMgbeewi7naITym57LoQ.png" /><figcaption>Steps for archiving a VHS tape using VHS Decode</figcaption></figure><p>Due to my limited soldering capabilities, the VCR I chose had RF test points that I could plug into easily.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/487/1*DCi5REJ4RQ4xQxl0AaL32g.png" /><figcaption>The RF “Tap” points in my Sony-678HF VCR Player</figcaption></figure><p>That RF signal goes into a <a href="https://github.com/oyvindln/vhs-decode/wiki/CX-Cards">CX Video Capture Card</a> that with a bit of modification turns into an Analog to Digital Converter (ADC). The best one to use right now is the CX White CX258000. We do some minimal soldering and tweaks, plug that bad boy into a PCIe x4 slot in your motherboard, install the driver called <a href="https://github.com/happycube/cxadc-linux3/">CXADC</a>, and you’re off to the races.</p><p>There’s also a fancier, a bit trickier way where you use two CX Cards with something called a <a href="https://github.com/oyvindln/vhs-decode/wiki/clockgen-mod">Clockgen Mod</a>. This helps keep the audio and video signals all lined up perfectly so syncing them later is just one click instead of a headache.</p><p>How this works is one CX card records the video signal at 40 million samples per second, and the other records the Hi-Fi audio at around 10 million samples per second, while a PCM1802 ADC captures the linear audio. Because I’m using multiple parts, the tiny differences in their built-in clocks (called quartz oscillators) can cause problems. Think of a quartz oscillator like a clock that keeps everything running on time. But sometimes, the clock on one card might tick just a tiny bit faster or slower than the other. Normally that’s no big deal, but when you want everything to stay perfectly synced, it matters.</p><p>The Clockgen Mod replaces the crystals on the cards with one master clock signal that all the cards listen to. That way, everything stays in perfect sync.</p><p>Of all the parts of this project, building the Clockgen Mod was by far the hardest. I decided to make it myself, and since it was my very first time ordering PCB boards and electrical parts, of course I ordered a wrong capacitor. That cost me a solid 2 to 4 weeks waiting for the new one to show up.</p><p>Thankfully, you have the option of ordering a <a href="https://ko-fi.com/s/ef0ed6126c">premade kit</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*gONTTVmj_cnQTA2pRT3PRw.png" /><figcaption>My modified CX258000 Card (Not pictured: C31 capacitor removal)</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/943/1*gZxyd8oTCsZfKbI4dfoE3g.png" /><figcaption>My final assembled Clockgen Mod!</figcaption></figure><p><em>Why do we sample the video at 40 million samples per second? Isn’t that a bit much?</em></p><p>Our RF video signal contains frequencies up to 8 MHz, or 8 million samples per second. According to the <a href="https://en.wikipedia.org/wiki/Nyquist%E2%80%93Shannon_sampling_theorem">Nyquist-Shannon sampling theorem</a>, to sample an analog signal without introducing errors called <em>aliasing</em>, you must sample at at least twice the highest frequency present in the signal. So, we need to sample at a minimum of 16 million samples per second (16 MSPS).</p><p>However, we actually sample at 40 MSPS because that’s the rate of the crystal oscillator on the Clockgen Mod. Sampling at 40 MSPS gives us plenty of extra detail and allows the software to perform better filtering for when we eventually downsample to 20 MSPS, resulting in a cleaner, more accurate video capture.</p><p>The way I think about it is like this: <a href="https://www.youtube.com/watch?v=hjWSRTYV8e0&amp;t=109s">playing a game at 200fps on a 60hz monitor</a> will often feel smoother than playing at 60fps on a 60hz monitor. Even though the monitor only refreshes 60 times per second, having more frames available ensures that a more up-to-date frame is always available at each refresh, reducing input lag and improving motion smoothness.</p><p>In the same way, sampling at 40 MSPS and then downsampling to 20 MSPS doesn’t increase the final output rate, but it gives the decoder more precise timing and signal information to choose from, resulting in a smoother, cleaner reconstruction of the video signal.</p><p><strong>TL/DR</strong>: Capture more than you think you need because you can always down-sample, but you can’t invent missing information.</p><h3>But… won’t oversampling mean large files?</h3><p>One of the downsides of capturing more than you need is the sheer amount of space that it requires. For example, both raw video RF .u8 file for a 34 min and 20 second video sampled at 40msps and 16 bit was 76.8GB — and that’s just the video. The accompanying Hi-Fi audio RF file clocks in at another 77.6 GB, meaning you’re looking at roughly 150 GB for just half an hour of footage. While keeping these raw files is ideal for archival purposes, allowing us to revisit and improve decoding as technology advances, it quickly becomes impractical if you have a large VHS collection.</p><p>Fortunately, there’s a way to compress the file without losing any data. Since the RF signal behaves like an audio waveform, we can apply lossless compression using FLAC. This method removes redundant noise without sacrificing any data, shrinking file sizes to about one-third of the original.</p><p>Another thing we can do is down-sample from 40msps to 20msps. As mentioned previously, the RF signal is sampled at 40msps to ensure all data is captured, but the actual signal range for NTSC VHS is 16msps.</p><p>Downsampling effectively reduces file sizes to around one-fifth of the original, bringing that 34-minute video from 76.8GB down to just under 15 GB, making long-term storage and processing far more manageable.</p><p>At this point in time, it’s also important to generate and save checksums of the RF file to maintain data integrity.</p><h3>We have the RF data; now let’s decode it</h3><p>Running the decode suite is as simple as one command. Behind the scenes, VHS decode is performing demodulation and time base correction.</p><p>That said, decoding large files can take up to a full day. For example, decoding <em>The Spongebob Movie</em> (2004) took overnight because I was working with the full 40 MSPS sample rate. If I had sampled at 20 MSPS and used more CPU threads, the process would have been much faster.</p><p>But here’s the tricky part: how do you know if you’re doing it right? Sure, you end up with a video and synced audio, but how can you be sure it’s the best quality possible?</p><p>That’s where tools like <strong>ld-analyze</strong> come in. After decoding, you can check the <em>Signal to Noise Ratio </em>(SNR) to get a feel for your video quality. An SNR above 40 is excellent, above 30 is still pretty good, but below 30 means the signal is weak, and you’ll start seeing a lot of visual artifacts. If it drops below 20, well… your video might start looking more like glitch art than a clean picture.</p><p>Audacity is great for analyzing the RF signal itself. The spectrogram of the RF signal will help you know whether you should adjust certain parameters on the CX Card, such as the gain level.</p><p>You can also use ld-analyze to visually inspect your video, tweak settings like black and white levels, adjust chroma gain, and fine-tune other parameters to get the cleanest, most accurate picture before the final export. I’ve included screenshots of some of the software’s features below.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/776/1*-hVhO7ERD9X7vy93S1WB-w.png" /><figcaption>Inspecting the luma frequencies using a scan-line oscilloscope in ld-analyze. Notice how the dark area inside Patrick’s mouth corresponds to a lower frequency.</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/939/1*IpILUyH0A_TEJPG3MHn4zA.png" /><figcaption>Inspecting the chroma using the scan-line oscilloscope in ld-analyze. The envelope (how the wave’s amplitude changes) corresponds to color saturation. The phase (how the waveform shifts left or right) corresponds to hue (color tint).</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/857/1*qqorqkaBMDc3G70kboulFg.png" /><figcaption>Inspecting the Signal-to-Noise Ratio (SNR) for my Spongebob Movie (2004) capture inside ld-analyze. A range of 40–50dB indicates a great signal.</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/687/1*BdNMwKEuL_dMRUkNVQvOIA.png" /><figcaption>Inspecting the dropouts of my Spongebob Movie (2004) capture inside ld-analyze.</figcaption></figure><p>After analyzing, we export the video to an .mkv file and align the audio and video using a script.</p><p>VHS Decode also provides a tool to process the <strong>Vertical Blanking Interval (VBI)</strong> data to extract closed captions or other metadata embedded in the tape.</p><h3>Post processing improvements</h3><p>The raw output you get is an <strong>interlaced</strong> video file. Interlacing was a technique used by old CRT TVs to display video efficiently by splitting each full frame into two separate fields. Each field contains only half the image (either the odd or even horizontal lines) and these two fields are shown one after the other very quickly. When combined, they create the illusion of a full frame. This is known as <em>deinterlacing</em>. There are several deinterlacing methods that I won’t get into here, but you can check them out <a href="https://github.com/oyvindln/vhs-decode/wiki/Deinterlacing">here</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/900/1*lY56B_NpalY6dZAR-agPiQ.png" /><figcaption>Interlaced (left), De-interlaced using FFmpeg (right)</figcaption></figure><p>It is also recommended to upscale the video to 4K if you plan to upload the video to YouTube, since YouTube will run its own compression on your video.</p><p>And that’s it!</p><h3>Why (or why not) VHS-Decode?</h3><p>VHS-Decode is the highest-fidelity method available, but also the most complex and time-intensive. There’s a lot of reasons not to use it, and it’s important to take into consideration your own time and resources:</p><ul><li>Ordering the necessary parts could take time.</li><li>There’s a bit of a steep learning curve, especially if you’re unfamiliar with the command line or soldering.</li><li>It requires a good amount of long-term storage space.</li><li>Decoding is slower than real-time (depending on your CPU and memory).</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/470/1*u0PKgqzSYJw2Ry7RC7F3Xw.png" /></figure><p>However, there’s a lot of reasons <em>to</em> do it.</p><ul><li>You’re getting the best quality.</li><li>It’s less expensive than traditional methods.</li><li>The decode suite is completely open source, so anyone can expand the features or port it to new operating systems. That means your video could potentially look even better a year from now. As the VHS-Decode’s motto says, “<em>Capture today, Decode forever</em>!”</li><li>Digitized RF data is header-less, meaning it is very resilient to bit-flips and cosmic ray events even if not stored on solid optical archival mediums.</li><li>Nothing beats the feeling of sending the digitized tape to your friends and family.</li><li>You can proudly say, <em>“I digitized and decoded this myself.”</em></li></ul><p>Thanks for tuning in! Read on to learn about what creates the visual artifacts in VHS video :)</p><h3>Bonus — The “VHS Effect”</h3><p>Tutorials on how to recreate the VHS look are everywhere. The filters are usually over the top, but they capture that unmistakable, nostalgic feel. But have you ever wondered why old VHS tapes sometimes make the picture appear grainy and distorted?</p><p>One common reason is <strong>TBC failure</strong>, and we’ve already seen an example. The image appears skewed, lines are misaligned, and there’s a lot of grainy noise in the image.</p><p>Another reason is <strong>dropout</strong>.<strong> </strong>And no, not the school kind or neural network kind, the <em>scan line</em> kind. Dropout occurs when the physical tape is physically missing information (like a scratched vinyl record). This generally occurs due to disc rot.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/760/1*IMCdJMjYmMlWuiW1EF6Yyg.png" /><figcaption>Example of dropout in my Spongebob Movie (2004) capture</figcaption></figure><p>If you’ve ever owned a VCR, you might have been told to “clean the heads”. The heads are what read the VHS tape. If they get dirty, then the picture can have a wavy or cross hatch (diamond) pattern. VHS “snowing” can also occur from dirty or misaligned video heads.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/910/1*2y2PTGGCt9SCwrV_XyhYDA.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/939/1*8CtiECC_cZ3q0d2QBRzDDQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/939/1*0ZR7aXY-6MisR2tySPS4fw.png" /><figcaption>Cross hatch pattern (left), wavy grain pattern (center), snow (right). Source: <a href="https://github.com/oyvindln/vhs-decode/wiki/Visual-Video-Signal-Errors#dirty-head">VHS-Decode Wiki</a></figcaption></figure><p>Finally, when an interlaced video is played on a progressive display (like the screen you’re most likely using to read this), the entire frame is displayed at once, causing the two fields to not align perfectly, especially if there’s motion. This misalignment results in the “combing” or “jagged line” effect, where you see distinct lines where the fields separate.</p><p>To see more video signal errors, check out <a href="https://github.com/oyvindln/vhs-decode/wiki/Visual-Video-Signal-Errors">this page</a> in the wiki.</p><p><em>This was originally a talk given at MTS.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=885acd146fde" width="1" height="1" alt=""><hr><p><a href="https://medium.com/henngeblog/be-kind-please-rewind-preserving-memories-through-vhs-decode-885acd146fde">Be Kind, Please Rewind: Preserving Memories through VHS Decode</a> was originally published in <a href="https://medium.com/henngeblog">henngeblog</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>