And so it came to pass, on 12-15-96 22:14,
   that Gordon Gilbert spake unto Matt Ion:
 GG> -=> Quoting Matt Ion to T Owen <=-
 MI>> This is a very good point that don't think I've ever considered
 MI>> before (although I have always agreed that there are good reasons
 MI>> NOT to limit bandwith to exactly within audible limits).  Given
 MI>> than a CD's sampling rate of 44.1kHz allows a theoretical maximum
 MI>> frequency of ~22.05kHz, that means you need a "brick wall" filter
 MI>> (a few dozen orders for at least a couple hundred dB per octave)
 MI>> at around 20kHz, and this causes hideous amounts of phase shift
 MI>> and other ugliness.  If one can increase the sample rate by, say,
 MI>> a factor of four (to 176kHz), the "steepness" of the filter can be
 MI>> reduced dramatically, with a matching decrease in the problems it
 MI>> causes. 
 GG>         You might have a good point if you'd made it back in the
 GG> mid eighties.  The fact is you can do these things digitally now
 GG> with digital filters and oversampling.  Brick-wall filters are a
 GG> thing of the past.
The problem is, sampling aliasing is sampling aliasing... how are the digital 
filters to separate it from the signal?  And on the D/A output, there's still 
significant filtering necessary to remove artifacts.  Use of oversampling, 
though, was exactly my point.
-------------------------------------------
Love, luck, and lollipops...
Matt


--- Sqed/32 1.10/unreg
---------------