It turns out that ESS Technology (https://www.esstech.com ) is still around and still making audio chips. I'm considering contacting them and asking if they are willing to open source or at least allow me to access the source code for the ES1869 windows driver with the intention of fixing this OPL3 compatibility bug and releasing a new driver build with the fix. Hopefully they still have the source code.

Does anyone here have any suggestions or pointers on how to effectively send this request? I don't want it to just end up in the junk box. 😀

I know some of you here have developed new hardware to emulate old cards and have other DOS / Windows 9x dev experience. Would any of you be interested in pairing up with me on this effort? It would be nice to have this effort be community driven, if there is interest.

I don't know much about VxDs or win9x but I don't think it disables I/O permissions for DOS programs (?) which would suggest that the ESS driver has disabled the OPL3 by way of a configuration feature in the ESS chip itself. The ESS1869 datasheet is online somewhere, and on pg. 41 it shows a power management register which contains a bit to disable the FM interface. It could be something that simple. Maybe you could just open a DOS window and then run a small QBASIC/DEBUG/Asm program to change the register value and re-enable FM. Although it's possible the driver would mess with it again after a period of time. If you need to hack the driver, looking for OUT instructions (or the equivalent API call, if there is such a thing for VxDs) and comparing them with the datasheet would be a good place to start.

Thanks for the input! That makes sense to use the Windows dos box to read / write to the IO ports to reconfigure the chip. I started looking into this with DEBUG, but the power register (Audio_Base+7h i.e. 0x227) had the same value 0x0A with both drivers. I'll keep experimenting.

I should note that the OPL3 still doesn't work after booting to Windows with the ES1869 drivers and then exiting Windows into DOS mode.

After much experimentation, I have a solution!

I have created two utilities that fix this problem: https://github.com/eshaz/essreg?tab=readme-ov-file#essreg
* essreg.exe - Used to set various unsupported registers on the ES1869 chip (3d effect amount, FM digital record, etc.)
* 1869opl3.com - Used to enable the OPL3 support for games in the Windows DOS box.

Notes:

It turns out that reading from the mixer registers (doesn't matter which one) just before running the game unmutes the FM audio. There were a few false positives along the way, since FM shares the same DAC as the wavetable chip ES689 (my machine doesn't have this) and the Zoom video I2S interface, so I experimented with that for a while. The FM volume register on the card was being changed with the IIS (Zoom video) slider on the Windows mixer, which was odd and maybe indicated that the Zoomed video was taking over the FM DAC. My guess is the that FM is being muted by the Windows driver when it's not active, and the game I was trying this wasn't triggering the driver to unmute the DAC. I was testing this with Cybersphere (breakout clone with really good OPL3 soundtrack), and it appears the problem is mostly limited to that game. The other DOS games (Jill of the Jungle, and a few others) I tested with worked just fine without the fix.

I also found a few other interesting features of the card, such as direct digital FM recording at 48kHz 50kHz (see below)! Check out the example here: https://github.com/eshaz/essreg/tree/main/digital_recording. I recorded this with Audacity while playing a midi file with Winamp, and then running Sierra's 3D-Ultra Pinball.

Link to Cybersphere: https://www.dosgamesarchive.com/download/cybersphere. It's a really fun game and has an awesome FM soundtrack!

Thanks for these utilities. The 3-D Ultra Pinball recording sounds pristine.

I added a few more recordings of common midi files, canyon.mid and passport.mid. They sounded a bit slow to me on the recording compared to the actual midi playback. It appears the sample rate of the digital recording is actually 50kHz. After adjusting the files, the recording matched up with the playback speed. This means that the music DAC (ESFM) is running at 50kHz on the ES1869. It's also interesting that after clicking record in audacity, nothing happens until I start the midi playback. I suppose the DAC isn't running when there's no ESFM midi playback.

Information from the ES1869 datasheet (pg. 65) regarding this feature:
Music digital record
1 = Enable direct digital recording of Music DAC
data (including FM, ES689/ES69x, or I2S). In
this mode, the first DMA channel must be
enabled for stereo recording. The sample rate
is determined by the music DAC sample rate
rather than by controller register A1h.