[perdomot]

I'm quite a noob to the audio world but I've read enough posts complaining about resampling to know its bad. Question is, how does this resampling work and how do we avoid it? I'm using a TBSC card that has a 2.1 PC speaker hook up AND a digital connection to my home stereo system so I would like to get the best possible sound I can. Thanks.

 

[sgrossklass]

Resampling is not bad per se (the reasons for it being used include playback of multiple sounds with different sample rates at once and outputting stuff with non-matching sample rates via old-style AC97 codecs which support 48 kHz only, like those on your card), just not trivial to get right with low distortion (mainly high-frequency IMD, some harmonic too). (Audacity still needs a good bit of work here btw, I was rather disgusted with the RMAA results of a file resampled from 44 to 48 kHz with it. The tricky part: Stuff resampled that way actually sounded better than the original due to the 3rd order harmonics present.) Hardware resampling inside sound card DSPs typically involves a tradeoff between quality and performance (you still want to do other things like 3D sound positioning after all). Creative puts more emphasis on performance (for gaming obviously), while the CS4630 on the "Santa" showed better balance here (see the Digit-Life review). There seem to be multiple ways of doing resampling (synchronous and asynchronous sample rate conversion appear to be main categories), a DSP guy (or resampler programmer) can certainly tell you more. Somewhat nontrivial stuff I guess.
Anyway, taking the good ol' blackbox approach, we only need to know that resampling may cause distortion as mentioned and that the kind applied by sound cards (and their drivers) tends to be of moderate quality only, inviting the use of good-quality software resamplers, e.g. SSRC.

 

[perdomot]

Thanks for the info man. So a card like the 710 using KS would sound "truer" because there is less distortion due to less resampling?

 

[kalzone]

one example of avoiding resampling with your santa cruz would be to get Foobar2000 for playing your audio files and configure it to use kernel streaming and resample it to 48khz. the resampling that would take place inside of foobar2000 would be of much better quality than the santa cruz's.

 

[sgrossklass]

Quote:

Originally Posted by perdomot Thanks for the info man. So a card like the 710 using KS would sound "truer" because there is less distortion due to less resampling?

The 710 isn't the perfect example, I think it only has a matching crystal for 48 and 96 but not 44.1 kHz operation so you'd still need good-quality software resampling (e.g. Foobar with SSRC). (The DAC for ch 7/8 would, however, be better than what the "Santa" has to offer, just provided the external load on the DAC is not too large since there's no output opamp there. And then there's no so-so quality hardware mixing to deteriorate the signal otherwise.) One of the better Envy24HT boards would make a better example, these typically have two crystals for a wider variety of sample rates. The 0404 uses a somewhat different approach, with a PLL based clock generator which only requires one crystal (but may be more difficult to get quiet in terms of phase noise and thus jitter).

 

[breez]

AV-710 does have two crystals, one for 44.1 (and multiples of it) and the other for 48 and multiples. 44.1KHz operation with the card, however, induces a slight HF roll-off (not major, less than NOS dacs for instance).