Allow me to join this information rich thread, as i'm playing with
convolution also.

> Well, according to all the dsp and math books I have read, convolution in
> the time domain is equal to multiplication in the frequency domain. Thus,
> when you multiply one fft window ( representing a bit of sound) by
another
> (eg a filter response shape) you get a filtered sound out the other end
(in

Is there a way in Csound to make a table that Csound interprets as a
filter-response shape ?

"specscal" perhaps ?


> the time domain, a filter is doing a convolution with the input samples).
> This is one way to do vocoding - multiplying the fft of some sound with
the
> fft of a vocal sound.

Damn....the manuals says that the FFT spectral datatypes are NOT
processable by other
Csound units.....How can one multiply an FFT channel from one sound with an
FFT channel
from another sound then ?


David.