Archive of the former Yahoo!Groups mailing list: MOTM

previous by date index next by date
  topic list next in topic

Subject: Nyquist was a wombat

From: "Paul Schreiber" <synth1@...>
Date: 2002-10-30

> > The Nyquist criterion states that any sample rate of more than twice
> > the highest frequency is all thats needed for accurate reproduction,
>

OK, this is pet peeve time!

NO, it DOES NOT say this. Because this is 1/2 of the statement. No one quotes the OTHER half
because few people take graduate-level DSP and calculus. The second half can be ∗simplified∗ to
say:

"for accurate reproduction.....assuming an IDEAL lowpass filter."

The ideal filter needs to have sin (x)/x response, which is a PHYSICAL impossibility. So, you
have to approximate it.

Also, there is a mathematical "assumption" made about the sampling part: that the "jitter" is
zero. Meaning, the sample period is PERFECTLY periodic, not within say 1ns but better than
0.001ps! That is also PHYSICALLY impossible. In fact, having low-jitter sample rate clock is MORE
AUDIBLE than just about anything else. That's the ONE point I agree with Stereophile: the better
CD players have extremely low sample-rate jitter. This is something you can plot as a histogram.
That's why Yamaha and Crystal Semi sell boatloads of jitter PLLs.

Paul S.