🚀 go-pugleaf

"SteveG" <member@adobeforums.com> wrote in message
news:3bbb6938.10@webx.la2eafNXanI

> Nyquist isn't actually wrong - but the direct
> implementation of what he said has always left plenty to
> be desired.

Well Nyquist only told the time-domain part of the story,
Shannon said much of the rest.

>It isn't necessary to go to the extent of
> raising the sample rate to achieve a better performance,
> though - you can remove the nastier artefacts of
> brickwall filters simply by oversampling at the D-A stage
> and noise shaping the resulting signal.

I sense a little confusion here. Oversampling and
noise-shaping are irrelevant to each other. Ovesampling is
an economical means for exploiting the ability of digital
filters to obtain desirable frequency and time-domain
perforamnce.  Noise shaping relates to amplitude domain
performance. It is a means for forcing quantization error
into frequency ranges where the ear is far less sensitive.

> This invariably
> gets away from the phase problems that the brickwall
> filters introduce,

Not at all. Oversampling is just a more economic means for
fabricating brickwall filters with certain desired
properties. The brickwall filtering does not go away, its
implementation just gets moved into the digital domain for
economic reasons.

Repeat: Oversampling does not replace or eliminate brickwall
filtering. Its just another way to implement it.

>and a correctly implemented scheme
> sounds far better at lower sample rates than not
> bothering, and using the brickwall approach.

It turns out that brickwall filtering at 22 KHz or so has
never had to be the sonic disaster universally ascribed to
it by some. If people listened with their ears as opposed to
their eyes...

> A lot of
> these issues are a legacy of the initial approach to
> digital reproduction, which was touted as being 'perfect'
> when it was anything but, and this has unfortunately
> opened the doors for a lot of suspicion about everything
> that's happened to improve the situation since.

Actually, a lot of legacy digital recordings sound great to
this day. As always, some people get it and do it, and
others don't.

>And as a result, also a lot of Snake Oil...

...such as a pathological fear of brick wall filters.

The fact of the matter is that most audio systems have
implemented brick wall filters in the audio band for
decades, even decades before digital was popularized. I'm
referring to the high obvserved roll-offs, and massive
phase-shifts and group delay that are common below 50 or 20
Hz in both modern and legacy systems.

There is a controversy over what's audible at 22 Khz, but
there should be no controversy at all over what's aubible at
20 or 50 Hz. So why do we tolerate and ignore sharp
roll-offs, massive phase shifts and immense group delays in
the bass range while obsessing over often smaller
(numerically speaking) similar effects in ranges that many
think is above audibilty?