By Multiply the amplitude spectra of the two input time series, and compare the result with the amplitude spectrum of the output series. Check the rule "convolution in the time domain is multiplication in the frequency domain". When I subtract the amplitude spectrum of the convolution from the amplitude spectrum obtained by multiplying the two input amplitude spectra, I observed a difference. Where does that come from?
amplitude spectrum (Convolution & Multiplication) difference
Started by ●March 21, 2006
Reply by ●March 21, 20062006-03-21
"Hamoor" <wallody@hotmail.com> wrote in message news:Hc2dnWCSZvpLgb3ZnZ2dnUVZ_tKdnZ2d@giganews.com...> By Multiply the amplitude spectra of the two input time series, and > compare > the result with the amplitude spectrum of the output series. Check the > rule > "convolution in the time domain is multiplication in the frequency > domain". > When I subtract the amplitude spectrum of the convolution from the > amplitude spectrum obtained by multiplying the two input amplitude > spectra, I observed a difference. Where does that come from? > >It could be almost anything.
Reply by ●March 21, 20062006-03-21
Hamoor wrote:> By Multiply the amplitude spectra of the two input time series, and compare > the result with the amplitude spectrum of the output series. Check the rule > "convolution in the time domain is multiplication in the frequency domain". > When I subtract the amplitude spectrum of the convolution from the > amplitude spectrum obtained by multiplying the two input amplitude > spectra, I observed a difference. Where does that come from?A small difference could arise from round-off errors. The calculations are not exact. A large difference would arise from the wrong thing being calculated or a calculation done wrongly. Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
Reply by ●March 21, 20062006-03-21
"Mike Yarwood" <mpyarwood@btopenworld.com> wrote in message news:dvp94a$hv9$1@nwrdmz01.dmz.ncs.ea.ibs-infra.bt.com...> > "Hamoor" <wallody@hotmail.com> wrote in message > news:Hc2dnWCSZvpLgb3ZnZ2dnUVZ_tKdnZ2d@giganews.com... >> By Multiply the amplitude spectra of the two input time series, and >> compare >> the result with the amplitude spectrum of the output series. Check the >> rule >> "convolution in the time domain is multiplication in the frequency >> domain". >> When I subtract the amplitude spectrum of the convolution from the >> amplitude spectrum obtained by multiplying the two input amplitude >> spectra, I observed a difference. Where does that come from? >> >> > It could be almost anything. >Ah! The rest of this note disappeared in my editor - sorry. Most likely differences (not in any particular order) - 1/ your convolution routine isn't doing circular convolution assuming that the signal set repeates indefinitely - so convolve[a with b] gives you an output which is maybe length(a)+length(b)-1 long and you select whichever fraction of it you think is representative before doing your fft. 2/ Your convolution routines and fft have different scaling factors. 3/ Both of the above. Worth checking these before you start panicking about everything else. Best of luck - Mike
