Klarinet Archive - Posting 000719.txt from 1997/11

From: Jonathan Cohler <cohler@-----.net>
Subj: Re: Nyquist
Date: Thu, 20 Nov 1997 11:18:13 -0500

Roger Shilcock wrote:

>Perhaps we should also take heed of the term "highest frequency
>component", which could be of arbitrarily high frequency. This makes
>Nyquist-compliant reconstruction physically
>impossible to implement for a
>faithful analog(ue) of a real audio signal. Isn't this the
>*real*
>point?
>roger Shilcock
>

Certainly, a pressure wave in the air could conceivably contain higher
frequency components than 20KHz. In fact, that's what ultrasound is, for
example. That's also why we have dog whistles.

But we human beings can't hear anything above 20KHz (that's why we don't
hear dog whistles, but dogs do--their hearing goes higher than ours).

So a "Nyquist" reconstruction is possible for things that we humans hear,
by sampling at the 44.1KHz that is available on modern day CDs.

By the way, I would agree with Jordan Selburn's comment that there is some
marginal benefit to making the sample size 20-bits instead of the present
16-bits. This has the effect of increasing the available dynamic range of
the recording media from its current 96dB to 120dB. 24-bits corresponds to
a dynamic range of 144dB, which I believe is excessive (and an ear doctor
could back me up on this).

Here's why. If we set our stereo system so that the softest sounds
(corresponding to the lowest or LSB of the 24-bits) is just audible, then
the loudest sounds on the recording will be at a level that causes severe
physical harm to one's ears. If, on the other hand, we set the volume so
that the softest sounds cannot be heard, then we are not using the dynamic
range anyhow, and therefore it is not useful. (I believe that 120dB is
already at the edge of what can cause damage to the ears.)

----------------------
Jonathan Cohler
cohler@-----.net