[Ardour-Users] Anyone using a HDSP9632 at 96K ?

[John Rigg]

On Sat, Jun 07, 2008 at 09:24:48PM +0200, Ralf Mardorf wrote:
> 
> > Strangely enough though - both utilities seem to be perfectly happy to
> > switch automatically between 48K and 44.1K.  In other words, if I set up
> > Ardour for 48K or 44.1K I don't need to faff around with anything else.
> > So what stops them from switching automatically to 96K?
> 
> Just a thought:
> 
> 44,1 we can hear the less quality, but it's CD standard
> 
> 48 we only be able to miss analogue tape compression etc., but the sound
> seems to be like the original and it's DAT standard
> 
> 96 might be the best quality for many sound chips, but it might be
> needles, because it's to good in most cases

There are a couple of reasons why 96kHz can sound better.

Most audio ADC chips use an incorrectly implemented decimation
filter to convert the single bit oversampled data from
the sampling stage to multibit PCM. If you check any of the
common ADC data sheets you'll see a stopband spec for the
filter of about 0.55 x sample rate, a clear violation of
sampling theory. This allows a small amount of aliasing to occur
on high frequency signals, and increasing the sample rate shifts
the problem to a higher, less audible frequency range. Correctly
designed converters like those from Lavry or Prism don't suffer
from this, but are very expensive.

The other reason is that non-linear DSP like compression
or limiting introduces harmonics, which will also produce aliasing
of high frequency signals. A higher sample rate reduces (but doesn't
entirely eliminate) the aliasing.

In practice, these effects are often not noticeable enough to
put up with the disadvantages of doubling the data rate, but YMMV.

John