getting ahead: convolution

[Martin Czech]

After a few days of silence...  (the mail address was changed, I could
listen, but not talk, and I had absolutely no time, anyway).

This is about artificial non recursive revere and filtering, AKA
convolution.

Due to the friendly support of Magnus Danielsen and Ingo Debus (and may be
others I have forgotten, sorry; Ingo, the spelling checker suggests that
you change your name to Bingo Debugs ;->)  I'm now able to deconvolute
impulse responses taken with a balloon.  The interesting point will be how
much noise and different balloon variations will affect the measurements.
I'm also able to get the transfer function/impulse response of a linear
system by cyclical white noise excitation (be it a room, or a filter or
whatever). I noticed the random numbers are not quite white, so I used
uniform magnitude, but random phase inverse Fourier transformed waves
instead. They are white by definition.


It is funny that such a wave really sounds like noise.
It is , after all, a harmonic wave with very low fundamental frequency and
rich harmonic content.

Next steps will be proper programming the bits and pieces I have
so far, and first tests with reverbs etc.

An impulse invariant filter design of a SAMPLED filter should
have no alias problems, because of the sampling guard filter.

I hope that the results will be interesting.

Of course, this all works only for time invariant systems,
no sweeps, or Leslie.

It all started when I didn't want to use outboard effects any more due 
to the additional ADC/DAC conversion, and realizind the a good reverb
plugin costs as much as a hardware unit...


m.c.