Indeed, thanks for reporting @wbcbz7! DOSBox-X has the 0x36 command implemented, so I missed that when I ported the code over.

I found it strange that the tunes kept playing for a few more seconds when you switched to a different one; now I understand the reason, that was the fade out time 😄

Enjoy Staging 😎

I've been recently checking out DOSBox Staging, and noticed that ESS AudioDrive emulation does not support extended ESS mixer registers

@wbcbz7 I've fixed this in the linked ticket, should be in the next dev build. Are you aware of any other programs that exercise the ESS specific mixer registers?

I've just checked the latest dev build, and seems like fadeout is almost there, but it sounds like volume curve being too steep; also volume adjustment (using -/+ keys) doesn't work properly (volume drops by ~12dB and locks at that level). Looking at mixer-related commit, I've noticed there is a typo in ctmixer_read() function - it uses register index 0x34 instead of 0x36 for extended FM volume read. As volume control in musicdisk reads current volume from the mixer, this can be the issue for now. Nice progress, anyway :)

As for other ESS-enabled applications that are using ESS-specific mixers, apart from Windows 3.x/9x driveres, I am not aware of those.

@wbcbz7 Thanks for the pointers, this is helpful.

I'll fix the wrong register index for the mixer read port, sure, but fixing the volume ramp would require some reversing (DOSBox-X uses the same wrong ramp, just by looking at the code).

I have a bunch of ESS sound cards so I could reverse it, but frankly I have better things to do 😄 Maybe I'll try a hack to use the SB16 ramp for ESS, maybe that curve is a bit closer.

Overall, the DOSBox-X ESS mixer seems to be shoehorned into the SB mixer implementation, so it's kinda expected to be somewhat wrong if there was no extra validation of the actual hardware behaviour.

EDIT: Seems like the attenuation is constant 2db per step according to this:
https://github.com/86Box/86Box/blob/e61c6205584e4414fc4a64494be5c6f1aab3e6ff/src/sound/snd_sb.c#L93

I'll give that a go.

@wbcbz7 Fixed, and this is the extent I'm willing to go. The "proper" fix would involve using "mixed" SB Pro volume curves for the master volume and the DAC, and the different ESS volume curve for the FM output (as we only support ESFM at the moment, but not the rest of the ESS card), and the complication is not worth it. It would involve significant refactors for not much benefit.

Our supported use case for the ESS card is a hybrid mode: ESFM for FM ("extended mode"), and SB Pro compatibility for the rest, including the DAC ("legacy mode").

yup, with the recent dev build FM volume mixer seems to behave as intended, and the difference between emulated and real mixer curve seems to be negligible enough, so I'll leave it as now. thanks for fixing! :)