Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!cs.utexas.edu!uunet!microsoft!brianw From: brianw@microsoft.UUCP (Brian Willoughby) Newsgroups: comp.dsp Subject: Sliding FFT == freq. spectra in real time Message-ID: <7879@microsoft.UUCP> Date: 28 Sep 89 23:26:40 GMT Reply-To: brianw@microsoft.UUCP (Brian Willoughby) Distribution: na Organization: Microsoft Corp., Redmond WA Lines: 19 I looked up the EDN article that I mentioned in an earlier post: Electronic Design News, Sept. 29, 1988, page 161, "Sliding FFT computes frequency spectra in real time", by Tom Springer of Pinson Associates Inc. I don't have anything to try this out with (yet), but it appears to offer more than a tenfold increase in speed when computing FFT's. It updates at the sample rate, provided that the system can compute N*4 multiply and N*6 add/subtract operations N times per sample, where N is the number of points. I would be curious as to how fast the 56000 could execute this algorithm. BTW, it uses a pre-computed exponential phase shift table as a constant factor, which eliminates the need for trig functions during the real-time computations. Brian Willoughby UUCP: ...!{tikal, sun, uunet, elwood}!microsoft!brianw InterNet: microsoft!brianw@uunet.UU.NET or: microsoft!brianw@Sun.COM Bitnet brianw@microsoft.UUCP