Saturday 25 December 2010

Tech Report: High-Quality Sound On The Megadrive?

So much for not working on Christmas day. Here’s a small retro-themed piece I just put together a few minutes ago for today. Enjoy!

Sega’s Megadrive (Genesis for all you North American readers) usually gets a pretty bad rap when it comes to sound. It’s not uncommon to hear games with really poor, overly fuzzy voice samples and sound effects, backed up with admittedly solid music composition. But even then, some music on a wide range of titles still tends to sound tinny, like it has been compressed right down to nothing and put in a tin can.

However, it doesn’t have to be that way, and as we will show you today, the MD is actually capable of some really high-quality audio more than matching what the Super NES has to offer.



Take the examples from this video above. Clearly, the first few games showcase the MD’s impressive capabilities in processing good quality voice samples, with clarity and crispness not usually found in most games on the system. Many people tend to confuse the MD’s poor sound with the use of low-fi samples. However, most of the samples used in the video are only 4Hz, 8bit PCM, and as you can see, you can still get great, very clean sound using such low-fi samples.

Instead, it seems like the vast majority of problems stem from the developers capability in writing decent playback code which best represents the samples upon being output by the console. Better playback code will improve the accuracy of how the sample is output to the Audio/Video encoder at the end of the sound/visual output stream. The encoder is another matter for concern, but we’ll look at this later.

Of course, bad samples will sound bad regardless of how good the playback code is. For example, both the muffled voices of SF2:SCE and tinny, grainy FM music of SSF2 cannot really be made to sound any better than they already do. Better playback code may indeed clean them up a little, but the overall effects and music used aren't particularly well implemented to say the least. It also doesn’t help that in SSF2’s case, that the use of FM for music also makes matters slightly worse. The MD’s sound chip is far better suited to PCM playback, and PSG composition, than synthesised FM – although it can do a good job on that too - unlike the Super NES and to a lesser extent the PC Engine, which handle FM sounds in a smoother fashion.



The second demonstration is a little different, and comes from a game called Tempo. Another reason why many sight the MD has having – and wrongfully so - poor sound is down to the actual A/V encoder used in various revisions of the console itself. For example, the earlier systems – we’re talking launch and early model 1 units here – have a far better A/V encoders than later models. The diverse range – and indeed sound quality – from differing encoders varies greatly, with some responsible for outputting incredibly poor sound compared to what the internal sound chipset is actual producing.

Many later Model 1 units have poorer encoders, and the vast majority of Model 2’s suffer from really bad sound in comparison to those, lacking warmth on low level frequencies and delivering tinny higher level stuff in general. Music and sound effects can often sound harsher and fuzzier than they are originally meant to be. There is at least one Model 2 unit that has very good sound – not as good as the best, or a very good Model 1 MD – but one that produces the kind of sound which at least comes fairly close up to what the developers intended us to hear in the first place.

The above video shows just what happens when you run the soundtrack from a 32X game on a Model 1 MD. Tempo uses just the MD sound chip and non of the PWM capabilities of the 32X, so forms the basis for a very good comparison to test this observation. There is also a demonstration in general of the audio capabilities of the console, although the small nature of cart sizes back in the day would prevent such high quality content being available for use in-game as it were.

What we find here, is that like with sound being output via the Model 2 MD, quality is sorely lacking through the 32X’s A/V encoder. Instead, when played back through a stock MD1, the sound takes on a new lease of life, not only sounding cleaner, but also adding depth and scale to the proceedings as well. You can hear the various instrumentals much more clearly, with individual elements being directly discernable whist being tightly intergrated together as a whole.

So, from the videos it is evident that Sega’s Megadrive can indeed produce some great sound which compares directly to some of the best on the Super NES in terms of raw quality. In fact, when using the MD1’s superior output, we can see that the MD can actually handle certain samples with much greater clarity than the Super NES – as can both the NES and PCE, but that’s another avenue entirely.

Essentially, the Super NES distorts all sound produced by its internal hardware via the use of some often unwanted filtering. By contrast, the MD’s sound output is much clearer as a result of both this, and the Yamaha YM2612’s, plus the Texas Instruments SN76489 PSG’s capabilities in combination with the Model 1 MD A/V encoder. Unfiltered Super NES samples sound incredibly clean too, though that isn't possible on a stock console.

In the end, we can see that the MD’s bad rap with regards to sound is perhaps a little more than undeserved – on a sheer technical level at least. On the other hand, it can be argued that Sega should have taken much greater care with its hardware revisions in general using higher quality components which accurately present the capabilities of the underlying hardware. That said, it is possible to perform various audio mods to every revision of the MD, thus getting the same high-quality audio output on any system as you would on and original ‘high-definition graphics’ branded Model 1 console. And that’s just to hear the true output of the internal sound chip, not to improve on it in any way.

Thanks go out to both TmEE and Joe Redifier, who performed the tests and created the videos used as reference in our report.

No comments:

Post a Comment