I had to use the 48000 mixer rate in DOSbox, due to instability when specifying 49716.

What instability? Try high fixed cycles and record stuff then, even if it sounds
choppy the recording should be ok.

On those samples, the old one sounds more authentic. If that's true, is there a reason why we went with the new method?

Use it if you like it better.

wd wrote:

I had to use the 48000 mixer rate in DOSbox, due to instability when specifying 49716.

What instability?

I'm not sure if instability is the right word for it, rather, random pitchy "bleeps" occur during real-time playback. I wasn't able to clear this up with any amount of changing either the CPU cycles or the mixer blocksize, but apparently, it may not bear any relevance anyway - the actual 49716 captures seem to be error-free, as you mentioned.

[Edit: Bah, I'd posted some re-captures, but will need to do so again later. My Audacity settings were affecting the exported Bit Rate...]

Thanks, i'll check them out by times.

Vlanar: The reason? Free (GPL) code, faster (better code base?).

I agree with you that for Zeliar, the MAME(0.72) code is closer for some parts. For the rhythm, I find dbopl(0.73) colser to the real thing than MAME.

These files are wierd... If you remove 9 samples from the 0.73 rhythm file, the baseline noise is sample for sample identical to the 0.72 one. But the onset of the drum samples are slightly different, making it impossible do diff them in one run.

The output of both emulators seem extremely similar at first sight.

Please don't use 0.72 for comparison but only 0.73 with the =old setting.
Several bugs have been fixed in 0.73.

By 0.72, I meant 0.73+old. Cloudschatze uses 0.72 in the filenames for describing that config.

Could you please add oplemu=compat to the comparison, if possible?
The generating app might be interesting as well, or dro captures.

wd wrote:

Could you please add oplemu=compat to the comparison, if possible?
The generating app might be interesting as well, or dro captures.

Done. The parent post is now modified with the new and updated files.

You'll have to forgive the inconsistent start/stop points between captures; I kept out of Audacity this time around.

I hope this is of some help. If there's anything else I can do, just ask. I realize that a digital capture from the OPL2 might be ideal - I'll see what I can do. If the digital output from an AWE32 is usable, that's easily done, however, it may not be the "purest" of examples, given some of the factors involved...

but I'd like to point out some differences in the playback of the OPL rhythm sounds between an actual OPL2, and the DOSbox 0.73 emulation cores.

There's some bug (Harekietish overoptimization) in the default opl core in
two of the rhythm instruments. Thanks for putting up the samples (the other
cores are fine, up to unknown inaccuracies/differences compared to the real opl).

Cloudschatze wrote:

wd wrote:

I had to use the 48000 mixer rate in DOSbox, due to instability when specifying 49716.

What instability?

I'm not sure if instability is the right word for it, rather, random pitchy "bleeps" occur during real-time playback. I wasn't able to clear this up with any amount of changing either the CPU cycles or the mixer blocksize, but apparently, it may not bear any relevance anyway - the actual 49716 captures seem to be error-free, as you mentioned.

[Edit: Bah, I'd posted some re-captures, but will need to do so again later. My Audacity settings were affecting the exported Bit Rate...]

The latest CVS no longer presents these bleeps or whatnot during real time playback. The actual output is still incorrect.

Feel free to post comparative recording of the pre-analog output of the chip
or, at your choice, a patch against the emulation.

wd wrote:

Feel free to post comparative recording of the pre-analog output of the chip
or, at your choice, a patch against the emulation.

Since I am not able to do the latter, I may soon be able to accomplish the former, unless another beats me to it first.

What I have available is a Yamaha YMF-724 PCI card with digital optical output on an older computer. I can record that by connecting it to a coaxial digital input on a newer computer (with a converter box). This is a genuine OPL3, so it should be faithful to the OPL2 that the game would have expected.

Thanks, that would be helpful.

Great Hierophant wrote:

What I have available is a Yamaha YMF-724 PCI card with digital optical output on an older computer. I can record that by connecting it to a coaxial digital input on a newer computer (with a converter box). This is a genuine OPL3, so it should be faithful to the OPL2 that the game would have expected.

The MAME emulation of top cymbal and high hat is a far approximation, and has considerable differences with the real OPL2 output of these two instruments. No reason to blame them, since the rhythm documentation is really poor, and the effect of parameters in these instruments has some chaotic behaviours.

I recorded outputs from both a real OPL3 and the MAME emulation back in 2008, and it showed just that. Unfortunately, I did it in Cubase, and doesn't have it installed in my PC anymore.

So, I guess if you focus on the top cymbal and high hat, you will easily find that.

I will record as soon as my optical to coaxical converter box arrives.

My converter box has arrived and works well. Here is a link to the recording I made of Zeliard's opening music:

http://www.mediafire.com/file/5rtnlxkmyqi/zeliard spdif.wav

I hope it will be helpful. I will post a recording of the percussion sounds once I master Adlib's Visual Composer program.

And here is the naked rhythm sounds, thanks to Adlib Visual Composer:

http://www.mediafire.com/file/mwmndzcyzym/rhythm spdif.wav

I hope these will help.