Xref: utzoo sci.electronics:17815 sci.physics:16832
Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!wuarchive!zaphod.mps.ohio-state.edu!casbah.acns.nwu.edu!ucsd!nosc!manta!north
From: north@manta.NOSC.MIL (Mark H. North)
Newsgroups: sci.electronics,sci.physics
Subject: Re: A question about the Nyquist theorm
Message-ID: <1757@manta.NOSC.MIL>
Date: 17 Feb 91 22:32:09 GMT
References: <91046.095459F0O@psuvm.psu.edu> <1751@manta.NOSC.MIL> <39342@cup.portal.com>
Organization: Naval Ocean Systems Center, San Diego
Lines: 28

In article <39342@cup.portal.com> R_Tim_Coslet@cup.portal.com writes:
>In Article: <1751@manta.NOSC.MIL>
>	north@manta.NOSC.MIL (Mark H. North) writes:
>>
>>This is an incorrect statement of the Nyquist theorem. The sample freq
>>must be *greater* than twice the highest freq component...
>>
>Looks accurate to me, it says "at least twice" and you say "*greater* than 
>twice". Both wordings mean the same to me (although the second is probably
>clearer due to the emphasis).
>

I'm sorry I must be losing it. At least twice implies twice is good enough
which it isn't. No?

>While the Nyquist criteria sets a minimum theoretical sampling rate, practical
>sampling rates are generally at least 5 times max frequency component (and
>more samples may be required if better reproduction is required, 5 times is
>just a "rule of thumb").
>

The reason for requiring a much higher sampling rate than the minimum
theoretical is because no filter is perfect and your nominal 'highest'
frequency is not really the highest. If you knew for sure in advance that
your signal had *no* components greater than nu, say, then you could do
no better a reproduction with a faster sampling rate than 2nu + epsilon.

Mark