Methodology

Something rougher and quicker would have been easier. For most people, it would have been good enough. But the Mega Drive Codex is not designed to be good enough.

Don't mistake this for a vanity project. It is preservation. And when you treat video game history with the same discipline that a museum treats centuries-old manuscripts, or a sound archivist treats master tapes, the choices you make connect to form something much larger than the sum of their parts.

One day, these games will be centuries old. The instruction manuals will be ancient manuscripts. At the time, the first episode of Doctor Who was just another television programme. Now it is an important piece of cultural history. Sonic the Hedgehog was, in 1991, just the next upcoming platform game. Now it is one of the most significant artifacts in gaming folklore. The Mega Drive Codex treats every game in the PAL catalogue with the weight that history will eventually demand of all of them.

Physical media is actively dying. Paper yellow-fades, ink cracks, and staples turn to rust. The capacitors inside original consoles eventually leak and corrode. The original physical artifacts of the Mega Drive era have a shelf life. If you make a quick, rough scan or a rushed audio capture, you are recording a degraded artifact, and someone else will have to do it all over again in ten years because the data was not dense enough to survive a technological leap.

We do it the hard way, with absolute precision, and that means it only ever has to be done once.

Audio: why hardware capture

Back in 1993, emulators did not exist. If you wanted to listen to the music in Streets of Rage II, you put the cartridge in your console, turned up the volume on your television, and listened. Hardware audio capture recreates that experience with fidelity that no software can match.

VGM files (Video Game Music files), are logged data streams of sound chip commands. They tell a sound chip what notes to play, but they do not capture how the original console's analogue output stage actually handled those signals. Every emulator that replays a VGM file interprets those commands through its own software model of the hardware. The analogue output stage, the specific characteristics of the DAC, the capacitors, the headphone amplifier circuit, the subtle interactions between components, none of that is in the log. It is approximated, and every emulator approximates it differently.

A hardware capture records what actually comes out of the console. The real analogue signal, shaped by the real components, at the real moment of playback. This makes sure it's not a model of the sound, unlike other audiophile projects that use VGM's as their base. It IS the sound.

The hardware

The Mega Drive used for all recordings is a PAL Model 1 VA6, and we use that specific revision for a reason. The VA6 board uses a YM3438, a variant of the YM2612 FM synthesis chip that produces less DAC ladder distortion than later revisions. The Model 1 retains the original headphone amplifier circuit that Sega removed in the Model 2 as a cost-cutting measure, a decision that demonstrably degraded audio output quality. Later revisions also introduced a high-frequency noise characteristic that is entirely absent on the Model 1.

So it isn't just a personal preference, but one that is documented in the hardware.

The console has been recapped, meaning its original electrolytic capacitors have been replaced with new ones. Capacitors degrade over time. A capacitor that is over 35 years old is not performing to its original specification. Recapping restores the console's audio circuitry to the condition it was in when it left the factory. Without it, you are capturing the sound of an ageing circuit board, not the sound of the hardware as it was designed to perform.

All recordings are captured at 96kHz and 24-bit through a Motu M2 audio interface into a MacBook Pro running Adobe Audition.

The mastering chain

Recording at 96kHz/24-bit gives a capture window that extends well beyond the frequency range of human hearing. When spectral de-noise, filtering, and normalisation are applied during mastering, the work is done on a signal with significantly more information than the final output requires. That excess provides precision. Every processing decision affects only what it is intended to affect.

Some argue that any processing compromises a hardware capture. This is a narrow view that overlooks how audio hardware and human hearing actually work. The goal of a hardware capture isn't to faithfully document the limitations of a 35-year-old circuit board. It's to recreate what the original developers intended you to hear.

Hardware captures pick up ambient electrical noise, line hum, and ground loops that were never part of the game's composition. They would not have come through your television speakers in 1992. Spectral de-noise removes hiss and hum that is an artifact of analogue cables and ageing components, not the audio data. If a composer wrote a silent pause into a track, they wanted silence, not a 50Hz hum.

A high-pass filter cuts sub-audio frequencies below 20Hz, which humans cannot hear. Sub-audio rumble is wasted energy. It forces speakers to reproduce inaudible movement, which distorts the frequencies you can actually hear. Filtering it out protects playback equipment and returns that headroom to the music.

The mastered recordings are resampled to 48kHz, dithered, and encoded to 24-bit FLAC using XLD. Every frequency within human hearing is present and accounted for.

The processing chain does not alter the music. It removes what time and physics added to it. The distinction matters.

Think of it this way - the Mega Drive Codex is not altering the painting, but it is cleaning the glass of its frame.

Loop points

Loop points, the moments where a track repeats, are identified by ear from the hardware capture. VGM files are not used as a reference.

VGM files treat a loop point as a hard reset because that is what the data log says to do. But instruments with heavy modulation, reverb, or decaying echo do not reset to silence at the moment a track loops. A note playing at the end of one loop is still sounding at the start of the next. A loop point that aligns perfectly on a visual waveform grid might look correct but sound jarring. The human ear catches what the data does not record.

Working from the hardware capture means working with the definitive physical reality of the machine. The Codex does not claim that its loop points are the only valid interpretation. It claims that they are made honestly, from the best possible source, by a set of ears that has listened to every track in the catalogue more times than is reasonable.

Scanning: why 600 pixels per inch

The public-facing scans on the Mega Drive Codex website are compressed JPEGs, optimised for web delivery. The archival masters are TIF files scanned at 600 pixels per inch.

Mass-produced video game packaging was printed using offset lithography, a process that creates images from thousands of tiny dots of cyan, magenta, yellow, and black. At resolutions below 600ppi, the scanner's pixel grid aligns awkwardly with the printed dot grid and creates a distorted interference pattern across the image called Moiré. At 600ppi, the scanner resolves the individual print dots cleanly, capturing the image as it was actually printed rather than as an approximation of it.

A compressed JPEG created from a 600ppi master will almost always look better than one created from a lower-resolution scan, even at identical output dimensions. When a large, detailed file is scaled down for the web, the software averages pixel data to compress it. Because the master contains a massive amount of accurate information, the result retains sharper edges, better text legibility, and smoother colour transitions. The compression algorithm is only as good as what you feed it.

If you use poor ingredients, do not be surprised if the meal does not taste very good.

Restoration also demands resolution. Covers get scratched, spines fade, staples rust and leave stains, manuals get creased by careless owners. If you attempt to repair damage on a low-resolution file, you quickly run out of pixels, resulting in blurry patches. At 600ppi, there is sufficient data density to work at the pixel level, isolating a blemish, repairing it, and blending it back into the surrounding texture without destroying it. Museums do this. The Mega Drive Codex does this.

And technology does not stand still. The displays of today will look primitive in twenty years, as the 800x600 CRT monitors of the 1990s look primitive now. A 600ppi master contains enough raw data to be reprinted at full scale on modern commercial presses without loss of quality. The Codex is not scanning for today's web browsers. It is creating a digital double of a fragile paper artifact before the original degrades or disappears entirely.

The restoration process

Every scan is restored in Photoshop before it appears on the site. Nothing is altered. Nothing is added that was not in the original. The goal is to return each item to how it looked the day someone went to Dixons or Argos and bought it off the shelf.

In practice this means removing fingerprints, dust, and rust stains left by decaying staples. It means mending tears, smoothing creases, and filling spiderweb cracks in ink where paper was folded carelessly. It means correcting the colour shift caused by decades of sunlight and oxidisation, restoring paper to its original white and bringing ink saturation back to what the printer intended.

The line between restoration and alteration is intent. Restoration fixes what time did to the document. Alteration changes what the original creator did. The role here is invisible repair technician, not editor.

Sourcing

Every game in the catalogue has been physically owned and processed personally. Where multiple copies were sourced to find the best condition example, the decision about which copy to scan comes down to one question: which requires the least reconstruction?

Manufacturing consistency in the 1980s and 1990s was variable. One copy of a game might have a pristine cover but a manual with someone's handwriting in it. Another might have a dirty cover but a perfect manual. The best elements of each can be combined to produce a definitive package representing what a pristine example looked like when you sold your NES for a Mega Drive and made a very wise decision.

Where a choice must be made between two imperfect copies, colour shift is always preferred over blurry print registration. A colour shift across a page can be corrected globally. Blurry factory print is baked into the pixels. Detail that was never captured cannot be recovered.

The in-period voice

Games in the Mega Drive Codex are written about in the present tense, as if the game has just been released. This is a concious decision because the aim of the Mega Drive Codex is to take you back to the day the game was released.

Writing in the present tense removes the distorting lens of hindsight. It judges a game on what it actually achieved at the moment of its release, not how it compares to decades of subsequent technology. Rise of the Robots, for example, was genuinely believed to be a Street Fighter-killer before anyone had played it. The magazines were quick to bring people back to Earth. We know how that turned out. But the in-period voice captures why people believed it, which is historically more honest and considerably more interesting than simply laughing at it from a distance of 35 years.

It also transports the reader. Instead of a dry historical overview, the Codex puts you in the shoes of a teenager standing in John Menzies, reading Mean Machines Sega while skipping school. It makes preservation feel alive rather than academic.

Writing this way requires active discipline. You have to forget the future, because in the period being written about, it did not exist. You cannot hint at Sega's eventual exit from the hardware market. You cannot reference games that had not yet been made. You must treat the Mega Drive's position in the market with the weight it actually carried at the time of writing, not the weight history later assigned to it.

The retrospective

Games in the Mega Drive Codex include a retrospective section, explicitly separated from the in-period writing by a horizontal rule, that steps outside the original frame and examines the game from the present day.

The retrospective does not undermine the in-period voice. It gives it value.

Without the retrospective, the review is a novelty time capsule. With it, the in-period writing becomes evidence, a primary source that the retrospective then analyses. If the 1992 review predicted that a particular mechanic was a gimmick that would not catch on, the retrospective is where that prediction is tested against what actually happened. The gap between period consensus and historical reality is where the most interesting history lives.

The retrospective serves the reader who is living in the present and wants to know how a 1991 artifact fits into the broader timeline of gaming. It does not weaken the in-period voice.

Why all of this, together

When the last instruction manual falls apart, when the last CR2032 battery stops saving your Sonic 3 progress, and when the last Mega Drive stops powering on, the Mega Drive Codex will still be here.

By demanding 600ppi scans to resolve print dots, restoring documents to factory specification, mastering audio to remove the noise of ageing copper, and writing reviews that capture what it felt like to be there, the Codex creates a digital twin of something that is actively disappearing. A permanent, immutable record at the highest quality baseline achievable, so that whoever comes looking for it next, in ten years or in two hundred, finds something worth finding.

The Codex does not exist because it was easy an easy task, no. It exists because the alternative is letting the history fade away like the Streets of Rage end credits.

And if we did that, we would become the bad guys.