home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!newsstand.cit.cornell.edu!vax5.cit.cornell.edu!cpyy
- From: cpyy@vax5.cit.cornell.edu
- Newsgroups: comp.dsp
- Subject: Re: Numerical Recipes FFT routine bugs ???
- Message-ID: <1992Jul28.093232.14073@vax5.cit.cornell.edu>
- Date: 28 Jul 92 09:32:32 EDT
- References: <1992Jul24.203717.2527@bert.eecs.uic.edu> <$=c1Hsr*p8@atlantis.psu.edu>
- Distribution: comp
- Organization: Cornell University
- Lines: 27
-
- In article <$=c1Hsr*p8@atlantis.psu.edu>,
- mek@guinan.psu.edu (Mark E. Kotanchek) writes:
- > In article <1992Jul24.203717.2527@bert.eecs.uic.edu>
- > eddins@bert.eecs.uic.edu (Dr. Steven L. Eddins) writes:
- >> croteau@ccwf.cc.utexas.edu (Ed Croteau) writes:
- >>
- >> >Does anyone have any information on the "FOUR1.C" routine in the
- >> >Numerical Recipes in C book? I believe that the routine as printed
- >> >(and on a source disk) does not properly compute the FFT (try a
- >> >simple real SIN wave and examine the imaginary). Please let me know
- >> >if I am wrong about this but it seems like the imaginary part is
- >> >frequency reversed or something similar.
-
- I also tried to implement the Numerical Recepie's algorithm for
- speech processing algorithms. Strangely, if you do a forward
- transform followed by a reverse transform, you don't get back
- your original input. I had to go through the Cooley-Tukey
- derivation to produce the correct algorithm. It seems that
- the one in the book is bass-ackwards in such a way that it will
- produce a meaningful magnitude result, with a quasi-meaningless
- reversed phase result.
-
- --
-
- Jon C. Russo
- internet : cpyy@vax5.cit.cornell.edu
- phone : 607.277.3295
-