Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!iuvax!rutgers!cmcl2!lanl!opus!ted From: ted@nmsu.edu (Ted Dunning) Newsgroups: comp.dsp Subject: log frequency fft Message-ID:Date: 26 Sep 89 04:48:34 GMT References: <89264.171306P85025@BARILVM.BITNET> <9520001@hpsad.HP.COM> Organization: NMSU Computer Science Lines: 26 In-reply-to: toma@hpsad.HP.COM's message of 25 Sep 89 20:20:30 GMT In article <9520001@hpsad.HP.COM> toma@hpsad.HP.COM (Tom Anderson) writes: It seems like you need a logarithmic frequency axis. I have often wished for an FFT type algorithm with logarithmic frequency spacings. Does anyone know of one? just do the conventional fourier transform at the frequencies of interest. since there will be roughly log n of these frequencies, the overall cost will be n log n just like the fast transform. in addition, you can drop some points from the lower frequency transforms and make them faster still (that is, update the lower frequency points more slowly). what this sort of transform loses is the reconstruction characteristic of the equally spaced transform; you won't be able to recreate the input given the transformed value. it might still be useful for display purposes, but i would guess that you would rather have the power in a number of disjoint frequency bands. -- ted@nmsu.edu remember, when extensions and subsets are outlawed, only outlaws will have extensions or subsets