home *** CD-ROM | disk | FTP | other *** search
- From: myers@hpfcso.FC.HP.COM (Bob Myers)
- Date: Thu, 3 Sep 1992 16:37:27 GMT
- Subject: Re: Where can I find the Nyquest theorem
- Message-ID: <7490220@hpfcso.FC.HP.COM>
- Organization: Hewlett-Packard, Fort Collins, CO, USA
- Path: sparky!uunet!usc!sdd.hp.com!scd.hp.com!hplextra!hpfcso!myers
- Newsgroups: rec.audio
- References: <1992Aug25.200552.12909@fcom.cc.utah.edu>
- Lines: 27
-
- > Fs = 2Fa would be the minimum Nyquist sample rate
- > ^^^^^^^
- >
- > anything > 2Fa would improve accuracy of signal reproduction
-
- Slight quibble; Fs = 2Fa (assuming, as I think was done in the original
- posting that used these variable names, that the signal to be sampled is
- strictly band-limited to Fa) sets the lower bound on acceptable sampling
- rates, but increasing much beyond this point does NOT improve the
- accuracy of reproduction. Let's take, for example, a signal which has
- no components above 20 kHz. What Nyquist (actually, Shannon) said was that
- sampling this at even 40.001 kHz would allow you to reconstruct the original
- signal *perfectly* (more precisely put, the phase, frequency, and amplitude
- of all signal components up to the original 20 kHz limit can be exactly
- recovered). Increasing the sampling rate to, say, 80 kHz won't buy you
- anything unless you're also raising the bandwidth of the original signal.
- (This may be counter-intuitive - everyone wants to think that 10 samples of
- a 20 kHz sinusoid just HAS to be better at reproducing that signal than 3
- samples - but it's true. The "trick" is that imposing the strict
- band-limiting lets you safely assume that any signal component at 20 kHz
- IS a sinusoid, and so the additional samples wouldn't tell you anything you
- don't already know.)
-
-
- Bob Myers KC0EW Hewlett-Packard Co. |Opinions expressed here are not
- User Interface Tech. Div.|those of my employer or any other
- myers@fc.hp.com Fort Collins, Colorado |sentient life-form on this planet.
-