though i could realize my idea (thanks to cyrille and frank), this was
not the solution for my original problem: modelling the snare-sound of
the snaredrum of tr-808. i could extract the (original) pure snare-sound
by subtracting a soundsample without snare from a sample with both
membrans and snare (here: http://www.netpd.org/~roman/sd_snare.wav).
when looking at the waveform, it doesn't really look like noise, rather
like a kind of filtered dirac, triggered randomly in time.
here is a patch trying to reproduce this.
not really like the original sound, but ...
hehe, cool. this is exactly what i tried once, too. the waveform of this
approach looks similar to the original sound somehow, but it doesn't sound
i think two things could be optimized:
- a minimum delay between pulses would avoid low frequency parts (two or
more subsequently following pulses form one 'fat' pulse).
- the shape of the pulses could be designed using a FIR-filter
- other ideas?