Ultrasonic sampling devices: why?

[janesconference]

If an human ear can hear no frequency greater than 20kHz, and, by Nyquist's theory we need no more than 40 kHz sampling rates, why does exist equipment that plays and records at 96 or 128 kHz sample rate? I mean, other than to records the bats hunting?

[metamorphosis]

Because

(a) Most cheaper DACs do a very poor job of sampling at lower frequencies (not a very steep rolloff approaching nyquist).
(b) Plugins can do a better job at frequencies above 60khz (can hide aliasing in unhearable frequencies).
(c) Synths do a better job at freq > 60khz for similar but different reasons
(d) It sounds better.

[janesconference]

Because

(a) Most cheaper DACs do a very poor job of sampling at lower frequencies (not a very steep rolloff approaching nyquist).
(b) Plugins can do a better job at frequencies above 60khz (can hide aliasing in unhearable frequencies).
(c) Synths do a better job at freq > 60khz for similar but different reasons
(d) It sounds better.

Mmmmh ok. I'm with you on a), could you explain b) and c) ? d) is subjective and needs at least a serious double-blind, don't you think?

[jupiter8]

could you explain b) and c)

Plugins and synths (not all of them but some) generate aliasing that disappears when ran at higher sample rates.
http://www.box.net/shared/vek1fl81vq
Here's 3 audio files recorded at different sample rates and downsampled to 44.1 kHz. The improvement is pretty obvious.

EDIT: And the most obvious reason they do it is because most people don't understand the sampling theorem and naturally assumes a higher number is better.
I can't even count the times i've read nonsense like analog has infinite resolution and sample rate implying a higher sample rate is more analog.

[metamorphosis]

Mmmmh ok. I'm with you on a), could you explain b) and c) ? d) is subjective and needs at least a serious double-blind, don't you think?

Nope! *wink*

[janesconference]

could you explain b) and c)

Plugins and synths (not all of them but some) generate aliasing that disappears when ran at higher sample rates.

So you're saying that plugins or synths generate frequencies higher than 44.1 kHz that are "wrapped back" at the beginning of the spectrum when "played" @ 44.1.
Btw, I'm curious. How is it possible to generate higher frequencies with a digital plugin? Can you make some practical example?

http://www.box.net/shared/vek1fl81vq
Here's 3 audio files recorded at different sample rates and downsampled to 44.1 kHz. The improvement is pretty obvious.

Yeah, but can't the degradation be due to downsampling? I know that when you downsample to a non-divisor frequency, the sample does not "fall" in the right bin, so to speak, and this causes degradation.

What you should have done to prove your point was to generate the same sound with the same synth at 44.1, 192 and 128 kHz. If you were right, the sound @44 would have been the worst due to aliasing, without introducing downsampling.

EDIT: And the most obvious reason they do it is because most people don't understand the sampling theorem and naturally assumes a higher number is better.
I can't even count the times i've read nonsense like analog has infinite resolution and sample rate implying a higher sample rate is more analog.

Yeah, like the Monster Cables vs. Coat Hangers article: http://www.engadget.com/2008/03/03/audi ... cable-and/

[jupiter8]

could you explain b) and c)

Plugins and synths (not all of them but some) generate aliasing that disappears when ran at higher sample rates.

So you're saying that plugins or synths generate frequencies higher than 44.1 kHz that are "wrapped back" at the beginning of the spectrum when "played" @ 44.1.
Btw, I'm curious. How is it possible to generate higher frequencies with a digital plugin? Can you make some practical example?
/

Any non linearity will do for example out = tanh(in);
Or just generate a "naive" waveform like a "perfect" square.

http://www.box.net/shared/vek1fl81vq
Here's 3 audio files recorded at different sample rates and downsampled to 44.1 kHz. The improvement is pretty obvious.

Yeah, but can't the degradation be due to downsampling?

If the degradation were due to downsampling why would it manifest itself only at the top frequencies ? The degradation would be uniform thru the audiofile.

I know that when you downsample to a non-divisor frequency, the sample does not "fall" in the right bin, so to speak, and this causes degradation.

Well what can i tell you ? You know wrong,that isn't even remotely a problem.

What you should have done to prove your point was to generate the same sound with the same synth at 44.1, 192 and 128 kHz. If you were right, the sound @44 would have been the worst due to aliasing, without introducing downsampling.

The problem with that is that few people have a soundcard that is capable of playing 192 kHz audio files and i certainly don't.
The original purpose of those files were to show that running at higher sample rates results in clearly better audio even when resampled to 44.1 kHz.
I used Voxengo R8Brain which is one of the absolute best resamplers there is to minimize the resampling artifacts,i can assure they're as close to 0 as makes no difference.

[janesconference]

Any non linearity will do for example out = tanh(in);
Or just generate a "naive" waveform like a "perfect" square.

Ok.

Well what can i tell you ? You know wrong,that isn't even remotely a problem.

Great, you got me, thank you.

[warrior545]

everything is good until a customer comes into your studio and ask the project at 96KHz. I will play that game but not beyond that rate.