I've been researching some YM2612 quirks for several months now, and have noticed some instruments haven't quite been playing correctly in officially licensed games, not the fault of SEGA or the IP creators, as most hardware functions properly, but there are a few outliers, for instance, the Mega Drive/Genesis 2 (which makes perfect sense, as the MD2 doesn't actually have a YM2612, it's a lesser known stripped down version), however, many emulators have a similar problem. So what I've done, is modified this Sonic 1 ROM to auto-correct frequencies by dithering the odd ones. (WARNING: The SEGA sound is loud, it's because of that stupid disassembly bug, so be sure to reduce the volume for this part, sorry, not really had the time to fix these silly things, been more focused on the operator FM stuff). I'm certain I can hear a difference, but I might be biased since I know what I'm listening for. Let me know your thoughts. Why am I dithering the odd frequencies? Here's why... Whilst recording from my MD2, I recorded about 5 hours worth of single operator sinewaves, all with operator MUL 1, unfortunately, I can't exactly post a screenshot of all 3FFF variations, here's the whole thing to get an idea: You can see where each octave is, all 8 of them. ...however, I noticed something strange, all of the odd frequencies are not at the frequency they're supposed to be, they're all slightly slower than they should be. So I've applied dithering by shifting the frequency from odd, to its next even above every 2/3 frames, the slower frequency shift ended up being approximately 23.87%, so it had to be quantised unfortunately. ...I might try and do the same for Sonic 2's driver, I'm sure I can generate the Z80 equivalent of the dithering mechanism. I mean, it could just be one of those things where, the human ear simply can't tell, and I'm more than ready to accept being wrong and dub this all a bit of a failure.
I just spent the last hour staring at my emulator's FM code and Nuked OPN2's extremely cryptic code trying to find the cause of the bug you're describing, to no avail. I didn't think to try the ROM you'd given until now. But you're right - even I can hear the difference. This is mind-blowing.
Can't believe it was such a simple fix. I've always thought the original music sounded a bit discordant.
I dunno... for the most part it is indeed improved, but there seems to be this really strange issue with FM4 where it randomly becomes even more discordant... for like a second. Very curious...
Unfortunately not. At least it gave me a chance to ensure that my emulator's logic was accurate to real hardware.
Playing with the sound IDs... Directly poking FFF009 to value B8 shows some curious behavior with ingame music (although BlastEm did not like it when handled this way). Sound ID B0 has some really strange frequency discrepancies when retriggering it. Wonder what's going on there?
If it makes you feel any better, I'm not lying about the odd frequencies, I just lied about them being MD2 and emulation and having a fix for it. The recording is actually from my MD1, and the odd frequencies are still shifted off, it's very subtle and you won't hear it, but if you graph the frequencies it doesn't flow mathematically. It's what stopped me from emulating my own YM2612, I was aiming for strict accuracy, and this put a wedge in my progress so I shelved it.
