Path: utzoo!utgpu!news-server.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!cs.utexas.edu!yale!mintaka!mit-eddie!bbn.com!moyarzun
From: moyarzun@bbn.com (Miguel Oyarzun)
Newsgroups: comp.dsp
Subject: Re: Mystery FFT Code...
Message-ID: <57353@bbn.BBN.COM>
Date: 12 Jun 90 21:05:42 GMT
References: <30678@cup.portal.com> <2846@syma.sussex.ac.uk>
Sender: news@bbn.com
Reply-To: moyarzun@labs-n.bbn.com (Miguel Oyarzun)
Organization: Bolt Beranek and Newman Inc., Cambridge MA
Lines: 17

In article <2846@syma.sussex.ac.uk> paulr@syma.sussex.ac.uk (Paul T Russell) writes:
>This looks very much like the code from "Numerical Recipes in Pascal".
>(Vetterling et al ?). The arrays consist of alternate real, complex
>values and start at an indexs of 1, so array[1] = x[0].re, array[2] =
>x[0].im, array[3] = x[1].re, etc...
>
>Incidentally, I think they have the signs for the FT and inverse FT
>the wrong way round - can anyone confirm this ?
>

If you look at the Numerical Recipes' definition of the Fourier integral,
you will see that their algorithm is consitent with their definition.
Unfortunately, they selected the opposite sign convention on the complex
exponential from that commonly used in engineering work :-(.  I guess
it's the old scientist-vs-engineer's way of looking at the world.  

Miguel