[PlanetCCRMA] Audiuphile 192 and others

Gregory Maxwell gmaxwell@gmail.com
Tue Nov 15 11:27:01 2005

On 11/15/05, Frode Petersen <fropeter@online.no> wrote:
> Thanks for the warning! I'll check that out.
> Do you have an idea of how big an impact on sound quality the difference
> in rate has? (Supposing I listen to the music on a high-end hifi system)

Fundamentally there is no reason why 192K should produce a perceptible
improvement over 96K.  However, most D/As and ADCs suffer from various
vulgarities which can be ameliorated by moving the nyquest frequency
further away from the audio spectrum and downsampling.

Modern audio ADCs internally operate at a much higher frequency than
their output (because it's nearly impossible to make a direct 24bit
dac due to matching problems) but their downsampling engines are not
optimal (and good they are not because an optimal process is going to
add latency).  For example, I use a RME multiface.. when I operate it
at 44 or 48K I see a excess of phase skew and an attention of
frequency response right near the top of the pass band. If I operate
the converter at 96k and use libsample rate to downsample, this is
gone.  I find it highly doubtful that the distortion is audible, but
since it is both in passband of human hearing and is measurable, I
prefer to eliminate it.

I expect you'd see the same thing with 96k vs 192k, but in that case
the problems will already be well outside of the realm of hearing, so
I wouldn't worry about it. Certantly you shouldn't expect to hear it
on a hifi system... have you ever seen the unsmoothed time and
frequency domain response of a (very good) system in a real room,

Now on the processing side of things... it's quite possible that there
are advantages of running many sorts of DSP algos at a larger multiple
of the output sample rate because they are insufficiently super
sampled internally. For this reason it would be good to see jack with
192k support.