[linux-audio-user] Sample Rates / Sample Sizes

[Steve Harris]

On Sat, Jun 19, 2004 at 04:30:26 -0400, Chris Pickett wrote:
> At first glance that seems fairly honest.  It does also suggest that you 
> wouldn't appreciate the benefits with a normal audio CD, so in that 
> respect it seems pointless, but maybe I'm missing something.  As for the 
> Nyquist frequency I read some discussion that some people can hear up to 
> 23 kHz, and that there may even be psychoacoustic effects up to 30 kHz, 
> but I didn't try to find any references on this.

I vaguely remember being shown a paper where a group of people were played
sine and saw waves at 15+ kHz and they could tell them apart. If I
rembered it correctly that does mean that humans can detect the presence
of partials at 30ish kHz, but my memory of it is vague.
 
> As noted, the effect processing argument says that it's better to 
> process everything internally at as high a resolution as possible (bit- 
> and frequency-wise); does this by itself mean we should also record 
> everything as high as possible, to get the best input to the effect 
> functions?  I have no idea.

Internally many processes upsample, process and then downsample.

- Steve