kool kitty89 wrote on 2020-05-11, 00:33:

Further, it seems to have stereo when set to Sound Blaster or Sound Blaster 2.0 clone mode and sounds heavily filtered as it does in SB-16 mode.

This is doubtful.

Genuine SB16 cards should be FIR filtering at the Nyquist limit in all circumstances for the given playback mode. This is correct from an audio engineering standpoint although it will sound different to the output of a CT1600 or other boards with fixed frequency filters.

I can't recall what the output characteristics were of the Opti card he tested, but IIRC Audiodrive and YMF ISA cards switch in a fixed filter for SB Pro modes and disable it for 16 bit/44.1 khz playback.

YMF71x cards have separate DAC for SB output, with a fixed filter. WSS output has a dedicated DAC+ADC with typical brickwall FIR and OPL3 output uses a dedicated DAC also with its own brickwall filter (YAC516 equivalent most probably).

SBpro cards have single 8bit DAC in them, with sample & hold multiplexing done in the mixer+DAC chip in stereo mode, similar to many low cost designs such as Yamaha FM chip DACs and MT32 sound output. SB16 doesn't support SBpro stereo due to not trying to emulate this mixer chip functionality on the main chip.

Tiido wrote on 2020-05-11, 05:39:

YMF71x cards have separate DAC for SB output, with a fixed filter. WSS output has a dedicated DAC+ADC with typical brickwall FIR and OPL3 output uses a dedicated DAC also with its own brickwall filter (YAC516 equivalent most probably).

Interesting. I'm surprised they didn't make the DAC re-clockable to 44.1k and do the easy bit rate conversion.

Tiido wrote on 2020-05-11, 05:39:

SBpro cards have single 8bit DAC in them, with sample & hold multiplexing done in the mixer+DAC chip in stereo mode, similar to many low cost designs such as Yamaha FM chip DACs and MT32 sound output. SB16 doesn't support SBpro stereo due to not trying to emulate this mixer chip functionality on the main chip.

As the SB16 implements a full dual channel DAC (thus not requiring multiplexing), presumably this would have been straightforward given more DSP grunt. In a lot of ways, sticking with an 8051 seemed to cause a lot of problems.

The sound engine in Tie Fighter might just be doing odd things and that might be part of why stereo works on that Opti card for it and not many other games Phil tested. (he didn't try using it in SB-16 mode though)

Things also get really bright/noisy unfiltered sounding when actual SB Pro mode is enabled, similar to my experience. That's true for cases where it's not actually playing back in stereo either, but still gets really bright/noisy compared to the mono modes. Or in the Descent sound test with center, right, and left: center is filtered and right and left are super bright/unfiltered. (all are still in MONO, though)

see:
https://youtu.be/okxN4GVLJRI?t=515

And on second look, the Tie Fighter sound test does sound filtered. (Descent does not except for the center channel)

I'm going to re-check my Vibra-16 tests with headphones this time to make sure I'm actually getting stereo output. (I swear I was getting proper left to right panning in tie fighter's setmuse sound test, though)

I'll also see if Jazz Jackrabbit plays in stereo when set to SB Pro.

Bear in mind this isn't an original Sound Blaster 16 chipset, but the Vibra 16S CT2504-TCQ. (card model CT2800) So the feature sets might differ from the multi-chip implementations.

No plug and play chip and not the later single-chip vibra 16 or waveffects SB-16 compatible chip with CQM.

Edit:

OK, my ears were fooling me ... or that and the combination of the tie fighter flyby sound effect making it sound like panning when it's not (there's better stereo tests, but this one was handy).

SB-16 stereo works, all the others are mono, but the SB-Pro option is still very noisy and bright sounding compared to the 16 and SB and SB2 settings.

Maybe the SB Pro setting's noise has something to do with how the SB pro stereo works. Unless I'm mistaken, the stereo stream has to be mixed as interleaved bytes (LRLRLRLRLRLRLR) and not as two separate byte streams (ie there aren't two separate buffers being read and then sifted into left and sifted together for the single, toggled DAC ... which similar DAC hardware could do: ie discrete left and right or L+R+Center buffers or even separate DMA channels then output through a single, multiplexed DAC ... potentially a PWM DAC or oversampling type DAC or some hybrid type)

I don't think it's the interleaved stereo panning that's doing it either, though. If that was the case there's be more signal noise when panned left or right than center (when panned hard left or hard right, you should effectively end up with a high frequency square wave, which could be audible depending on the sample rate used ... so sort of like PWM squeal, or even more like PCM playback done by volume modulation of a square wave PSG channel set to an audible frequency: say if you played back 4-bit PCM on the Atari POKEY or NES sound, but had the tone generator set to something like 4 or 8 kHz, or even somewhat above 10 kHz).

But I don't hear it getting less noisy (or louder in volume) when the sound test should be at the center, so it shouldn't be that either.

Maybe the SB Pro option has the sample/stream being scaled up to a higher playback rate (on the software end) to avoid any squeal/tone from the interleaved mixing at a lower sample rate, and thus my Vibra 16 is filtering less heavily than it should be. (say if they're using something like 8 kHz samples scaled up to 22 or 44 kHz ... though I think the SB Pro was limited to 22 kHz for stereo mode)

If the filtering threshold is set based on the sample rate selected (hardware DMA sample rate) than scaling up samples would effectively bypass the filter. (and given how noisy these samples are, I'm going to assume thy're in the 8 kHz or less range: though most of the sampled speech, especially in the cutscenes, sounds much cleaner or identical when SB Pro mode is set here, so they're probably already 22 kHz or close to it, but most of the sound effects are much, much noisier or heavily filtered)

Hmm, the 2D/cutscene portions could also be mixed at a higher sample rate where the in-game 3D engine drops that down more. (except for SB Pro everything would still be getting scaled up after being mixed)

Also, Tiido, I believe the 50 kHz stereo PCM output of the Atari STe also uses that same type of multiplexing, though it seems to use 2 discrete 8-bit DACs as well. I think it actually has 2 channel 8-bit stereo outputting through separate DACs. (and you can set each channel to LRLRLRLR interleaved stereo or to mono)
There's also an external stereo panning/mixing chip (LMC1992) but I think it's just used for volume control on the 2 channels from what I can see from the sound routing. (it could be used for panning, but would need to have the PCM channels fed into more inputs than they appear to be: they also don't look like they can be combined to a single 16-bit channel, even though that volume/mixer chip has the right volume range/steps to allow that)

On a side-note, this Vibra 16S (CT-2800, and also a CT-2900) isn't a Plug n' Play Sound Blaster 16, but also isn't manually jumper configured. It works in some cases without drives (especially as Adlib), but needs the Vibra 16 installation software to really work properly.

I'm pretty sure these are functionally similar to the Vibra 16C cards with similar looking Creative ASIC onboard. (16C has the CT2505 with embedded OPL3 clone, the 16S uses the CT2504 with discrete OPL3 ... or potentially a clone like some OPTi cards use, but I've only seen ones with Yamaha marked chips there)

And I also shouldn't have lumped the SB16 Waveffects in with the Plug n Play SB16 or Vibra 16. The chip used on the CT4170 Sb16 waveffects cards is sometimes labeled Vibra16XV (CT2511) but I think it may have some differences other than form factor compared to other Vibra models and certainly seems to be different as far as card pin-out goes. (it's only got the 8-bit data bus lines and a very limited number of extended I/O lines present)

This thing:
https://upload.wikimedia.org/wikipedia/en/7/7 … 6waveffects.jpg
https://i0.wp.com/www.toughdev.com/content/wp … 7/11/ct4170.jpg

I've seen some comments towards the CT4170 being 8-bit slot compatible, but haven't tried it myself and I think it's worse than the ESS Audiodrive for that purpose unless it manages to be more compatible with Turbo XT boards or something. (but worse OPL3 clone and having SB-16 compatibility is probably not very useful on XT compatible software ... but I guess it might be handy in a free 8-bit slot in some 286/386/486 boards)

Hmm, and the bright, unfiltered PCM output seems to be common to the ESSAudiodrive as well, so it very well may be an artifact of the SB Pro stereo functionality. I don't have a real SB Pro available anymore (Apolloboy sold his and I never had one of my own) but I suspect the mono (44 kHz SB 2.0 compatible) mode on the SB Pro and Pro 2.0 have the normal frequency cut-off and filtering, too.

https://youtu.be/fWOdIuRO5fI?t=425

The ESS Audiodrive native mode for tomb raider might be doing something else, too, but the SB Pro option certainly has that very raw sound with high frequency aliasing.