Quote:

Originally Posted by

**nuthinupmysleeve**
How? According to Nyquist, the answer is no. If you disagree, how does it impact the audible range?

Nyquist said nothing of the sort. I've said it before and I'll say it again: I wish that people (regardless of their position on the subject) would stop conflating the mathematics of the sampling theorem with the psychoacoustics of listening.

What Nyquist, Shannon, and Fourier did was to lay the groundwork of the sampling theorem. The sampling theorem itself says nothing about human hearing per se - indeed, it isn't even applicable only to audio, but to signal processing in general. Some people around here have been making the implicit assumption that the sampling theorem itself defines the range of human hearing, which is most assuredly does not. The mathematics of the sampling theorem obtain whether the sampling rate is 22.05 kHz, 44.1 kHz, 88.2 kHz, 176.4, or 45097 kHz.

The conclusion that 22 kHz bandwidth was sufficient for audibility was determined through psychoacoustic research quite independent of the information theory research undertaken by Nyquist, Shannon, Fourier, etc. I profess to being ignorant of whom you should cite if you want to make the case that 20 kHz or so is the absolute limit of human perception (Helmholtz? Fletcher?) - but Nyquist et al. did not deal with frequency perception directly.

Again, there are perfectly good reasons why one would prefer a higher sampling rate that have nothing to do with misunderstanding sampling theory or with claims of ultrasonic hearing. (Note that I take no position on ultrasonic hearing.) The simple fact is that there are no ideal converters, and real-world implementations, while improving continuously, are still not perfect. Personally, I feel that it makes sense to get the analog filter as far out of the audio band as humanly possible. Paradoxically, I think this makes even more sense in a consumer-level device (which is apt to have an inferior analog section and a less accurate clock) than in a mastering-grade converter, with its superior analog filters and clock.