[linux-audio-dev] Re: Cheby amp code

[Steve Harris]

On Mon, Nov 04, 2002 at 01:54:56 +0100, Tim Goetze wrote:
> >Thats interesting, cos liiing at your waterfall plot it doesn;t seem liek
> >that should be enough.
> 
> afai understand the paper, they've been doing this as an academic 
> exercise. ivory towers and so on ...

[sorry for the bad-even-for-me typing it was very cold in my studio]

Well, IMHO the point of ivory towers is that you dont have to be connected
to reality, so you'd think they would just go for the most accurate
process. NB I haven't read the paper.

> but pod distortion waveforms don't look like they use chebyshev
> polynomials at all.

I agree. I wonder if they've just worked out transfer functions for
different envelope amplitudes...

You could feed in sinewaves (or maybe saws, I'd have to think about the
consequences of that though), and construct a transfer function table by
mapping input amplitudes to output - you'd need to correct for the delay
caused by the filters in the circuit, but otherwise it should be fairly
easy.

The highpass filter might make things interesting, but you could probably
correct for it by averaging the output from the upcurve of the sine with
the output at the downcurve.

It still wouldn't be frequency dependent, but its another possible route.
It will be much cheaper than our other approaches, probably cheap enough
that you could carve up the input signal into frequency bands and process
them seperatly without it being slow... which would also make the alising
easier to deal with.

- Steve, feeling entusiastic again :)