Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP Path: utzoo!mnetor!uunet!seismo!husc6!bloom-beacon!gatech!emory!platt From: platt@emory.uucp (Dan Platt) Newsgroups: comp.lang.c Subject: Re: fast fourier transform Message-ID: <2176@emory.uucp> Date: Thu, 16-Jul-87 19:06:50 EDT Article-I.D.: emory.2176 Posted: Thu Jul 16 19:06:50 1987 Date-Received: Sat, 18-Jul-87 08:57:25 EDT References: <1065@bloom-beacon.MIT.EDU> <3251@cit-vax.Caltech.Edu> Reply-To: platt@emory.UUCP (Dan Platt) Distribution: world Organization: Math & Computer Science, Emory University, Atlanta Lines: 14 Keywords: fft, general radix, C In article <3251@cit-vax.Caltech.Edu> alee@tybalt.caltech.edu.UUCP (Andrew Lee) writes: >... What's wrong with simply padding the array with zeroes >to a power of 2? > The problem with padding with zero's is that it introduces a window in the data which the transform tries to fit with a superposition of frequencies producing a pattern which would be the convolution of your data with a sin(omega*T)/(omega*T) like function, where omega is the frequency, and T is the size of the window. This generaly produces ringing (false oscilations) which wouldn't be inherent in your data at the start. Hope this is a help... Dan