[Clam-devel] Spectral Analysis/Synthesis

Pau Arumí parumi at iua.upf.edu
Thu Feb 7 13:39:24 PST 2008


On dj, 2008-02-07 at 22:25 +0100, Karsten Krispin wrote:
> Hi Pau, 
> 
> and thanks for the quick answer! :)

> 
> > The code doesn't rings a bell on me about the problem. But it is likely
> > that it's is related to the fact that you are not using ports.
> > SpectralAnalysis ports consumes a given HopSize of samples while access
> > a given WindowSize.
> > So my advice is to take a look at this test, and maybe start from this
> > code:
> > CLAM/test/FunctionalTests/ProcessingTests/TestStreamingSMSAnalysisSynthesis
> >.cxx
> >
> > In case the problem is because of that: you can still use the
> > processings without ports (as you do now) but buffers would require some
> > extra cooking
> 
> That's unfortunate, because using buffers seems to be more flexible to me.
> 
> How ever, might it be sufficient to just spend my classes a private port for 
> each,  for pushing audiodata into SpectralAnal and getting the spectrum out 
> of it and vice versa for SpectralSynth?

Exactly, as it is done in the test I pointed before, if I remember
correctly.

[...]
> Am Donnerstag, 7. Februar 2008 21:09:52 schrieb Pau Arumí:
> > your system seems fairly complex and I think that having an isolated
> > example of spectralAnalysis/Synthesis working would help a lot.
> Yep, especially for the SpectralAnalysis/Synthesis part... There is actually 
> nothing. But I don't understand, because SMS consumes too much CPU power and 
> just having the spectrum should be quite enough for the most purposes.
> 

I think you mean SpectralAnalysis not SMSAnalysis (which uses SA and
does more things)

SpectralAnalysis is appropriate when you need to do hard spectral
transformations. If you don't, don't use it, but use the FFT (and IFFT)
directly.

If this is the case, you might also want to use the "new spectral
processing" plugin (right now at CLAM/plugins/spacialization/) with new,
much simpler and more efficient FFT and Spectrum classes (but
spectral-analysis is not there)

Pau








More information about the clam-devel mailing list