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