SpritesMind.Net

Arbee: Indeed they do matter, but I was wondering whether it would be better to discuss them in a separate thread or not.

I defer to the people already in the discussion though, since everyone else here knows a lot more about this stuff than I do. However the discussion has piqued my interest in trying to do an fpga implementation of the ym2612...

LN

I agree with Snake too. By default it should sound like the model 1 Genesis/MD since that's what the games were developed on. Now, that doesn't mean there can't be several modes where you can sound like the other ASICs or knock off YM2612s.

Emulators sound like MD2 mostly, with the nice and clean sound. Real YM2612 has quite bit of the "ladder effect" noise in the sound, where as MD2 ASIC YM2612 does have it much less. Also, real YM2612 has its sine a bit shifted on half periods... one half its 1 step up, other half not... I have a sample somewhere on this forum. Slight tonal change happens because of it. MD2 ASIC has perfect sine to say so. I don't know about the clone chip SE-95... I presume its a slightly modified YM2612 (a decapped one...)

http://www.epicgaming.us/tiido/MD_Sine.rar

Kindly done by Epicenter


I hope this sheds some light on YM2612's internals

Each WAV has 5 or 6 samples in it, First with Total Level at 00h, and each following has it incremented. 0, 1, 2... notice how the wave gets better with lower TLs... and they should be 27.5Hz...

Lord Nightmare wrote:This is why I think for all ym2612 tests on real hardware, data should be captured as soon as it exits the dac on the ym2612 chip

Indeed. But most YM chips output digitally to an external DAC. The YM2612 has the DAC and other things internally. There is some filtering that goes on BEFORE the signal exits the chip, and that's what I was really talking about.

What is it called again where you crack open an integrated circuit and >>dye<< the bits to obtain a 100% accurate representation of what goes on? Expensive, but intriguing?

SmartOne wrote:What is it called again where you crack open an integrated circuit and die the bits to obtain a 100% accurate representation of what goes on? Expensive, but intriguing?

Murder.

Awww, *expletive.* I knew I was spelling that wrong! DYE.

Snake wrote:

SmartOne wrote:What is it called again where you crack open an integrated circuit and die the bits to obtain a 100% accurate representation of what goes on? Expensive, but intriguing?

Murder.

Not quite. The Chinese often do. When the chip must be protected in reading, they whittle away the shell and read under a microscope. And there are real benefits of this technique.
So the method is technically and feasible. And the Chinese are 100% did so. But who would do so now for us?

That's the goal of the "Decapping Project" for MAME.

ElBarto wrote:That's the goal of the "Decapping Project" for MAME.

Isn't goal of this project is to only 'dump' masked ROMs which is undumpable otherwise? Reading masked ROM data from scanned chip surface is much simpler than reconstructing chip logic from same source (if it possible at all).

Not quite. The Chinese often do. When the chip must be protected in reading, they whittle away the shell and read under a microscope. And there are real benefits of this technique.
So the method is technically and feasible. And the Chinese are 100% did so. But who would do so now for us?

It wouldn't be too hard to do decapping at home, but it is a little dangerous. I had a good look at it awhile back. In order to do it properly, you need to use concentrated fuming nitric acid, which is nasty stuff. The method is simple though, it's just the acid that's dangerous. You'd want a proper fume cabinet to use, but it's the kind of thing you could easily do with the equipment you'd find in a highschool science lab.

I had a go at decapping a chip just using a hand-held dremel recently, and the results were surprisingly good on the chips I tried, but you need a really steady hand, and you always end up damaging some parts of the die. Decapping is mainly useful for reading embedded ROM data though. I'm not sure how much the untrained eye could figure out in terms of how an IC actually performs various operations from pictures of the die. Never know until you try though I guess.


Anyway, I've been away on holidays and then working on other areas of my emulator, so I haven't done much on the YM2612 in the last few weeks. Some SSG-EG issues came up in the Regen thread which I need to do some tests for, and I'm definitely going to have to get to the bottom of this track 12 issue in BattleTech. I was hoping that there weren't any YM2612 differences between the various mainstream Mega Drive systems. Looks like we're not quite so lucky. I agree with Snake that the focus should be on emulating the "real" YM2612, but I will do tests on the other models now to figure out what's responsible for this difference at least. I own just about every Mega Drive model, variant, and spinoff ever made, so I can run tests on all the different PAL and NTSC models with what I have on hand.

Still the question is, about the Model2 chip: why the "blip blip" sound is clearly more audible in battetech track 12 but sounds not so different from a Model 1 chip in Comix ZOne track 05 ?

Both use the exact same instrument settings, the only difference being the incriminated operator RR & TL values(0x00 in Battletech, 0x0e in Comix Zone)

I did not succeed in emulating both accurately by simply increasing the former 17-bits register width, this must be something else going on

Indeed. But most YM chips output digitally to an external DAC. The YM2612 has the DAC and other things internally. There is some filtering that goes on BEFORE the signal exits the chip, and that's what I was really talking about

.

I would love to know more about that and no, I'm not tired of asking
One example I think could result from bad filtering is the Konami Logo jingle in some games that sound crackly on emulators but perfectly clear in Kega

http://www.fileden.com/files/2008/4/21/ ... GMPLAY.ogg
Here's Konami logo along with some other things from my MD2. Mike-Stamp's VGMplay is used.

http://d.hatena.ne.jp/video/niconico/sm3495564
This is the only one video i know which compare sound of jp md1 and md2.
I will say md1 better than md2 in this video.
u could easily hear noise in md2's sound.

Are you sure that's the correct Konami logo in the recording? The only one I know that has the crackling issue is the one from Zombies Ate My Neighbors.

Castlevania Bloodlines BGM 26 distorts similarly.