Scandy wrote on 2021-07-31, 13:52:

This core includes a Covox clone ("Stereo-on-1"?) compatible with the Disney Sound Source.

That in itself sounds a bit fishy, because Covox Speech Thing and Disney Sound Source are quite different, internally.
The Covox Speech Thing is a pure DAC, whereas the Disney Sound Source has an integrated 16-position FIFO buffer.
IIRC, the standard detection procedure for the Disney Sound Source involves observing the parallel port status bit that is supposed to signalize "FIFO full" while filling the FIFO.

Thank you so much for your kind reply. It seems that it's DSS, but with a slightly increased sampling rate (44100 Hz/6 = 7350 Hz), this could affect the detection in Hocus Pocus?
It works well even in Alone in the Dark (floppy version).

I wonder if its a Timing issue that screws up the detection. I Guess you could try to debug

First interesting piece of info: You can run HOCUS.EXE (the shareware version) with the parameters ss1, ss2 or ss3 to force sound source detection at ports 3BC, 378 and 278 respectively, circumventing the auto-detection mechanism.

Full algorithm in the shareware version (installed) as distributed DOS games archive:

1. Deassert /INIT on the printer port (write 04 to the control register)
2. Wait for four timer ticks at 140Hz (I guess to drain the FIFO)
3. Test that ACK reads as zero
4. Strobe 32736 data bytes as fast as possible into the DSS (likely a bug, and intended to be 32 bytes). Strobing is done by temporarily pulsing SELECT IN high (writing 0C to the control register, than writing 04 again). If you supply the parameter "sst" (sound source stereo?), not only SELECT IN, but AUTOFEED is pulsed too.
5. Test that ACK reads as one

If both tests for ACK succeed, the sound source is detected. Note that the initialization part depends on timing, whereas the actual FIFO testing part does not.

Thank you so much!!!
I'll check with ss1, ss2 or ss3 parameters first.

The /ss2 parameter fixed it, thank you!