Path: utzoo!utgpu!attcan!uunet!ginosko!gem.mps.ohio-state.edu!pacific.mps.ohio-state.edu!tut.cis.ohio-state.edu!quanta.eng.ohio-state.edu!raksha.eng.ohio-state.edu!rob From: rob@raksha.eng.ohio-state.edu (Rob Carriere) Newsgroups: comp.dsp Subject: Re: Real-time Fourier Transform Keywords: FFT DFT Message-ID: <3151@quanta.eng.ohio-state.edu> Date: 1 Oct 89 01:34:26 GMT References: <7899@microsoft.UUCP> Sender: news@quanta.eng.ohio-state.edu Reply-To: rob@raksha.eng.ohio-state.edu (Rob Carriere) Organization: Ohio State Univ, College of Engineering Lines: 17 In article <7899@microsoft.UUCP> brianw@microsoft.UUCP (Brian Willoughby) writes: >Its been a while since I studied this in college, anyone care to describe >the DFT vs. FFT in layman's terms, or point to a text which does? I seem >to remember that the inverse FT is basically the same operation as the FT >from time- to frequency-domain... 1) Inverse DFT differs from DFT by the sign of the exponent and (if you're not a mathematician) a factor of 1/N (assuming N data points) 2) FFT is simply a fast method for computing a bunch of DFT values. With DFT, you could compute the spectral value at any frequency you care, but it takes O(N) operations for each frequency. FFT gives you the spectral values at N equidistant frequencies for a total cost of O(N log N) (N has to be a power of 2), significantly cheaper than the O(N^2) sweat the regular DFT works up. SR