I will test some cards later. I have real opl 2 /3 / L and clones .
Why do you need this data? Want to make a program, a game or hardware? Im just curious.

DerBaum wrote on 2022-07-31, 11:53:

I will test some cards later. I have real opl 2 /3 / L and clones .
Why do you need this data? Want to make a program, a game or hardware? Im just curious.

That would be really helpful, thanks.
I'm working on the 86Box emulator, and this would allow to accurately emulate the real hardware's behavior.

I posted 5x results into the attachments. I have no OPL2 to play with for the time being. Timer 1 with 255 is highly variable with its result on my test machine (1.4GHz PIII machine with VIA chipset)

Anyway for completeness sake :
The timer circuitry runs off YM's sample rate which for
OPL2 is MCLK / 72 = 3579545 / 72 = 49715.90277....Hz (20.11428... µs)
OPL3 has 4x higher MCLK but it is divided down by 4x so end result is same as OPL2
OPL3L is a little trickier, MCLK there is 33868800Hz and internal sample rate slightly different from OPL2/3 : 33868800 / 684 = 49515.78947...Hz (20.19557...µs), which is then decimated into standard 44100Hz rate (33868800 / 768) for final audio output.

OPL2/3 Timer 1 counts every 4x YM output cycles (80.457153...µs), Timer 2 every 16x YM output cycles (321.82861...µs).
OPL3L Timer 1 has (80.78231...µs), Timer 2 every 16x YM output cycles (323.12925...µs).

I'm pretty sure the 255 is a typo, though it is difficult to gauge from the results I got as 255 result on Timer 1 is all over the place.
Timer 1 period is supposed to be (72 * 4) * (256 - T1val) clock cycles
Timer 2 period is (72 * 16) * (256 - T2val) clock cycles

Timer value of 0 should be longest possible count and using that will determine if 255/256 difference is actually there, if it is 255, 0 will show much different result. (it is not something I have personally tested) so perhaps you should make a test with 0 value aswell.

Tiido wrote on 2022-07-31, 13:55:

I posted 5x results into the attachments. I have no OPL2 to play with for the time being. Timer 1 with 255 is highly variable wi […]
Show full quote

I posted 5x results into the attachments. I have no OPL2 to play with for the time being. Timer 1 with 255 is highly variable with its result on my test machine (1.4GHz PIII machine with VIA chipset)

Anyway for completeness sake :
The timer circuitry runs off YM's sample rate which for
OPL2 is MCLK / 72 = 3579545 / 72 = 49715.90277....Hz (20.11428... µs)
OPL3 has 4x higher MCLK but it is divided down by 4x so end result is same as OPL2
OPL3L is a little trickier, MCLK there is 33868800Hz and internal sample rate slightly different from OPL2/3 : 33868800 / 684 = 49515.78947...Hz (20.19557...µs), which is then decimated into standard 44100Hz rate (33868800 / 768) for final audio output.

OPL2/3 Timer 1 counts every 4x YM output cycles (80.457153...µs), Timer 2 every 16x YM output cycles (321.82861...µs).
OPL3L Timer 1 has (80.78231...µs), Timer 2 every 16x YM output cycles (323.12925...µs).

I'm pretty sure the 255 is a typo, though it is difficult to gauge from the results I got as 255 result on Timer 1 is all over the place.
Timer 1 period is supposed to be (72 * 4) * (256 - T1val) clock cycles
Timer 2 period is (72 * 16) * (256 - T2val) clock cycles

Timer value of 0 should be longest possible count and using that will determine if 255/256 difference is actually there, if it is 255, 0 will show much different result. (it is not something I have personally tested) so perhaps you should make a test with 0 value aswell.

Thanks for testing ! My measuring of the time is only accurate to around 10-20 µs, so that would account for the variability you see with timer 1 at 255.
I agree with you that the "255" in the OPL3 datasheet is most likely a typo.
What is much stranger to me is your results with "YMF278B 1.jpg " where you basically get the same period no matter what.. is that the same sound card as in "YMF278B 2.jpg" or a different one ?

YAMAHA OPL2 YM3812-F (Aztech (I38-MMSD801) Sound Galaxy NX II)

YAMAHA OPL3L YMF289B (Aztech (I38-SN96103) MM Pro IIB-3D)

CRYSTAL OPL3L (clone) CS4289-KQ (Unknown Manufacturer CS32-3DI)

System Pentium 1 200MMX.

elyosh wrote on 2022-07-31, 14:07:

What is much stranger to me is your results with "YMF278B 1.jpg " where you basically get the same period no matter what.. is that the same sound card as in "YMF278B 2.jpg" or a different one ?

It is a different sound card, and yeah, it was weird to see that happen. I'm not sure why it was so either, since all tests were done in pure DOS with exact same environment so maybe the card has some hidden fault on it.

Quite interesting findings. Would anyone be able to run these tests on cards with OPL clones (like Crystal FM, ESFM, and CQM for example)?

DerBaum wrote on 2022-07-31, 14:42:

YAMAHA OPL2 YM3812-F (Aztech (I38-MMSD801) Sound Galaxy NX II) 2022-07-31 16.05.56.jpg […]
Show full quote

YAMAHA OPL2 YM3812-F (Aztech (I38-MMSD801) Sound Galaxy NX II)
2022-07-31 16.05.56.jpg

YAMAHA OPL3L YMF289B (Aztech (I38-SN96103) MM Pro IIB-3D)
2022-07-31 16.09.21.jpg

CRYSTAL OPL3L (clone) CS4289-KQ (Unknown Manufacturer CS32-3DI)
2022-07-31 16.35.58.jpg

System Pentium 1 200MMX.

Thank you for testing. It's interesting to see that the only card that actually implemented the typo from the OPL3 datasheet that causes the timer to expire immediately if it is set to 255 is your Crystal CS4289 clone !

elyosh wrote on 2022-07-31, 15:55:

Thank you for testing. It's interesting to see that the only card that actually implemented the typo from the OPL3 datasheet that causes the timer to expire immediately if it is set to 255 is your Crystal CS4289 clone !

If somebody wants to see the card...

on the top left you see the "CS4289-KQ" OPL3L clone. The Chip right next to it on the right side "4333-KS" is the DAC.
(The unpopulated pads below the chips are for a real YMF262 and YAC512.)

What does OPLProbe use as a timing reference under DOS?

georgel wrote on 2022-07-31, 16:54:

What does OPLProbe use as a timing reference under DOS?

A 100KHz PIT counter is used for timing.

elyosh wrote on 2022-07-31, 17:02:

georgel wrote on 2022-07-31, 16:54:

What does OPLProbe use as a timing reference under DOS?

A 100KHz PIT counter is used for timing.

Which channel and which mode?