| I have a few reasons for wanting to try direct convolution:
I like to have more than one way of doing things. Csound offers several ways of
doing most things - why not convolution too?
It has educational value (eg reproduce the Moorer experiments)
I'm not convinced that there are no side effects using the FFT approach - how can I
find
out other than by trying both?
maybe I don't just want to apply reverb, but some arbitrary transformation where I
can
have control over sample amplitudes.
Modern PCs are fast enough to make it reasonable.
Sheer, uncompromising crankiness.
Does anyone know what technique is used in the SoundForge Acoustic Modeller plugin?
Richard Dobson
Erik Spjut wrote:
> At 10:56 PM -0800 2/17/98, Hans Pelleboer wrote:
>
> >For very short convolutions, I might even cook up a simple `delay
> >+ multiplication' script, without altering or adding anything to the code.
> >Trouble is that I would like to do LONG convolutions, like recorded
> >reverberation responses, for example. It is possible, of course, to use
> >an external application, solely devoted to that one task, but it would be
> >much more elegant to have it incorporated in csound itself.
> >Any suggestions?
>
> Use convolve. Implementing the direct convolution sum for more than about
> 128 taps is foolish (Csound does fine up to about 20 taps). The convolve
> opcode implements the orders-of-magnitude faster procedure of taking FFT's,
> multiplying spectra, and taking inverse FFT's. You do have to account for
> the delay, and pre-analyze the impulse-response file, but the manual has
> the formula for the delay and the pre-analysis is very fast. If you have
> the memory, you can easily do 10 second recorded reverb responses with
> convolve, and it will run about
> 400,000 times faster than the direct convolution sum. The results, except
> for the delay, are identical. Why would you possibly want to do direct
> convolution? What am I missing?
>
> -------------------------------------------------------------------------------
> Erik Spjut (rhymes with cute) - Acting Director,The Center for Design Education
> and/or Associate Professor of Engineering
> Harvey Mudd College, Claremont, CA 91711 USA
> Erik_Spjut@hmc.edu Ph & Voice mail (909) 607-3890 Fax (909) 621-8967
|