AWE32 - Analog/Digital Output Comparison

[Cloudschatze]

CT2760 TDA1543T 16-bit DAC → SPK (amplified) output, default EQ:
Onboard DAC, analog

Tyrian - The Level, OPL3 (dry)
Tyrian - The Level, OPL3 (reverb 30)
Tyrian - Asteroid Dance 1, EMU8000


CT2760 S/PDIF → CS4226 20-bit DAC, default EQ (none):
External DAC, analog

Tyrian - The Level, OPL3 (dry)
Tyrian - The Level, OPL3 (reverb 30)
Tyrian - Asteroid Dance 1, EMU8000


CT2760 S/PDIF
Digital

Tyrian - The Level, OPL3 (dry)
Tyrian - The Level, OPL3 (reverb 30)
Tyrian - Asteroid Dance 1, EMU8000
(If you're wondering why it starts off with a portion of Asteroid Dance 2... it has to do with Tyrian resetting the EMU8000, loss and subsequent re-synchronization of the digital signal, and me, too tired to bother doing anything else with it. Enjoy!)

[Rhizome]

The OPL3 recordings sound much more crisper on the S/PDIF output, but can't really tell much difference on the EMU8000.

[Great Hierophant]

I hope my ears aren't going bad, but I can hear no difference between the analog and digital versions of any sample.  Maybe I would if I had some good headphones, but alas. 

[Ari]

I must be losing my hearing as well. I could hardly hear any difference at all.

[Rhizome]

I think I'm the youngest on this forum - could explain it 

[5u3]

There is a difference, it is most noticeable in the "dry" OPL3 recordings. For example, the instrument used for the beat has very high-pitched "ringing", and this is more apparent on the digital recording than on the analog one. I can only make out the difference with headphones and the treble on my amp cranked all the way up though.
Both EMU8K recordings sound exactly the same to me.

[NewRisingSUn]

I think I'm the youngest on this forum - could explain it

Placebo - that could explain it, too.

CT2760 S/PDIF → 20-Bit DAC, default EQ (none):

What do you need the DAC for? S/PDIF is a digital signal already. Are you routing the S/PDIF signal to a 20-Bit DAC, then recording the analog output? Because if so, I don't know what you want to tell us with the recordings, other than that an outside DAC will sound different from the AWE32's own DAC. I notice that the short silence at the start is not perfectly digital zero in the second recording, suggesting an analog source.

Another thing to consider: For a digital recording, S/PDIF would operate at 44.1 kHz in your case. The OPL3 however outputs at 49.7 kHz, according to its data sheet. Either the AWE32 operates it at a different clock speed so that it does output at 44.1 kHz contrary to the data sheet, which would be impractical since the pitch would also be lower given the same frequency register writes by the software, OR it somehow resamples the 49.7 kHz to 44.1 kHz for the S/PDIF output.

[Cloudschatze]

CT2760 S/PDIF → 20-Bit DAC, default EQ (none):

Are you routing the S/PDIF signal to a 20-Bit DAC, then recording the analog output?

Yes.

Because if so, I don't know what you want to tell us with the recordings, other than that an outside DAC will sound different from the AWE32's own DAC.

That's basically it - A comparison of the analog output from the AWE32, versus the same from an external DAC. I see how the subject was a bit misleading.

What may not be completely obvious in the recordings is that the noise floor is greatly reduced with the external DAC, in addition to the brighter highs.

I do plan on recording the pure digital output; it's just more time-consuming, given the equipment I have, and I've not had the chance to do so.

Either the AWE32 operates it at a different clock speed so that it does output at 44.1 kHz contrary to the data sheet, which would be impractical since the pitch would also be lower given the same frequency register writes by the software, OR it somehow resamples the 49.7 kHz to 44.1 kHz for the S/PDIF output.

I'd wondered this as well. I suppose the EMU8000 might resample the input, but I have no idea really.

[Cloudschatze]

I'd wondered this as well. I suppose the EMU8000 might resample the input, but I have no idea really.

Different theory: The OPL3 implementation in the CT1747, like the OPL3-L, may only output at 44.1kHz.

[NewRisingSUn]

More likely, yes.

I just read the OPL3-L (YMF289B) data sheet, and it seems that unlike the OPL2 (YM3812) or OPL3 (YMF262), the registers can actually be read, which may be used to tell a OPL3 from a OPL3-L.

[Great Hierophant]

What, if any, audible effect would changing the output frequency from 49.7 kHz to 44.1 kHz have on the sound's character?

[NewRisingSUn]

Theoretically none, since both rates have a Nyquist limit above the threshold of human hearing. In practice however, the chip generates a mathematically non-perfect signal, which results in distortions in some rare cases, and those could sound different on a chip with a different rate.
There was a sound effect in 3D Construction Kit that sounded like "yayayaya" , which is usually a particular kind of aliasing; that might sound different. The PIEP2.BMF song included in the Easy Adlib 1.0 tracker uses a similar sound. I have no OPL3-L to check, only a OPL3 and a CQM, so I cannot find out if my speculations are true.