In case you want to try this yourself, here is a ZIP file containing all the files you need. Load SBFMDRV, and then simply PLAYCMF <filename>. Have fun 😀

For accuracy testing be sure you use the opl-native frequency setting for
both the opl rate as well as the mixer rate (oplrate=49716, [mixer] rate=49716).

That sounds quite a bit better actually, thanks for the pointer! First few chimes still too squeaky (see attached MP3), but if I close my eyes I can almost imagine I am back in the early 90s... 😜

Thanks,
Mikity

Maybe you can post a head-on comparison of a recording from a real sb.

By the way, can I expect any adverse effect from setting the overall mixer rate to the unusual frequency of 49,716 Hz? So far I hear none, tested a series of MOD-based games such as Epic Pinball.

Thanks again!
Mikity

Unfortunately my old 486 with SB 1.5 is at my parents' home in Europe, so any recording will have to wait till my annual trip there later this year. 😀

By the way, can I expect any adverse effect from setting the overall mixer rate to the unusual frequency of 49,716 Hz?

That depends on your soundcard (cheap ones may have problems).

Ideally there'd be some (high quality) resampling in the dosbox opl core.

The cmf files work nicely using adplay as well as loads of other dos era chiptune formats.

The player I use is audacious media player

All open source, just throwing it out there if you're looking to playback music in a dedicated music player. 😀

Using the adplay opl emulation (inferior to dosbox's currently) or real hardware?

Thanks for the pointers. Actually, I have Winamp with Adplug, and sadly the playback of the CMF files I have is less accurate than PLAYCMF under DOSBox...

There should be some unofficial patch for adplug/adplay that has better
opl emulation quality, but any way one of the new opl cores will be integrated
into adplug/adplay thanks to Simon Peter (adplay author).

Okay, it seems like I know some friendly folks with old sound cards... so here is a recording of it using a real SB2.
Can you spot the differences?

Can you spot the differences?

Is this a digital recording or just the final analog output recorded?

I'm not quite sure what you mean? The early SBs didn't have a Digital Out 😉

There are people who can grab the digital information right after the chip's processing.

Anyways, i don't hear anything wrong with the sample, and the minor differences
surely are due to DA conversion features/artefacts of the sb card.

I can hear at least two differences which don't seem to be DA-related. The introduction sounds more delicate with the real SB version and the bass sound seems louder later on.

The waveforms are quite different too, And the FFTs have som subtle differences.

The DOSBox capture has a 32000Hz sample rate, the SB one is at 44100. After bitrate conversion with ffmpg, the DOSBox wav file turned out to be 75976 samples longer (after removing the silence at the beginning of each track). I zero-padded it before applying the FFT.

I guess the high frequency cut-off is MP3 related, and that It appears as a notch on the DOSBox FFT because of the high frequencies introduced after the resampling.

There are some noticeable differences in the diff, as you can see. There are roughly more bass and less treebles on the real SB. Differences related to the length (hence frequency) difference shouldn't appear on these smoothed curves.

It could be nice to have samples at the same initial samplerate, with minimal compression added.

I can hear at least two differences which don't seem to be DA-related.

I hope you'll provide proof that it's not related to any stage after the
digital opl output one by actual digital comparison.

Actually, the graphs weren't there to support that point.

To be more precise, some instruments sound identical, some are different, and the balance between some of them is also different. I don't think that these could be DA-related.

The two "xylophone" notes that end the intro at 00:09 sound one octave lower in DOSBox. I'm sure that this isn't DA related.

The low/high frequency balance could indeed be the result of an analogic post-processing (or more simply related to the impedance of the circuits, without any intention to post process anything), but if it is the case , it would be nice if it were emulated too (you're emulating an SB, not a purely digital OPL chip, after all).

Overall, I find the original more lively.

Can you hear what I'm talking about?

I'm not complaining, btw, I'm very happy with the current state of the SB emulation. This is intended to be constructive criticism.

Can you hear what I'm talking about, though?

Well i don't hear anything that i'd not contribute to the information loss during the recording.

Could you try to tweak the low/high frequency ballance accordingly?

It's not like there's a switch to emphasize some frequencies. If you want to
check if there's really some bug that causes this, you'll have to go to the
digital part first and assure everything is fine there.