On Fri, Sep 24, 2010 at 09:51:07PM -0700, Patrick Shirkey wrote:
> IIRC Fons suggested the sc4 compressor algorithm was similar to a vocoder.
I did never suggest anything of the sort concerning the sc4 compressor.
More precisely, the * FFT-based equaliser * is not a linear filter but
a vocoder. You just have to inspect the code to see this. The problems
resulting from this are mitigated to a large extent by using a large
overlap factor, at the price of much increased CPU usage. To the point
that a real filter implementation using linear convolution would in fact
be simpler and more efficient.
If this were all I'd have provided a patch years ago. But there are other
problems with this type of filter: it provides way too much detail in the
mid and high frequency regions. There is really no point in having any
significant gain difference between say 4000 Hz and 4020 Hz, the only
result would be a very bad impulse respons that destroys all transients.
Yet the user interface makes it easy to get this sort of thing, even just
accidentally. It's a good example of textbook DSP being applied blindly
and without understanding the consequences.
There are three of them, and Alleline.
Linux-audio-user mailing list