additive

[Martin Czech]

> hi guy's.
> theoretcly any acostic sound can be made in a modular
> ether by addtive or FM but this is hard to be made.
> so my qustion is can i play a acostic instrument like a
> pino or violin or other in to a mic and then in to a scope
> and see its waveshape and envelope and then calculaite
> this data and creat the same sound on a modular ?

If I've got you right you mean that if you hear a sound, you'll be not
able to derive a patch and parameters in order to emulate this sound.
And you hope to find these parameters by a scope printout.
Well, for simple sounds, like a recorder (softly blown) this will work,
but in this case you don't need it, because you allready hear what's going on
and it is simple to derive a patch.

For very complex sounds this won't work at all, you will not be able to
even trigger the scope, since the waveform is changing all the time. It can
even be hard to see the envelope. Of course you can use a DSO and
record the instrument in full length, but then you are on the way to
sampling. The waveform does not determine the sound, there are a lot of
waveforms that sound equal, but they don't look equal (simply do
Fourier analysis of, say, a saw wave, and then mix the partials
together again, but with random phase values. Gives very differnent
scope shots, but sounds the same.

No sorry, there is nothing better then listening/patching experience.

> all so is ther any scope or other unit that can take a waveshape
> and do the calculate and give a data of how many sine osc in
> what freq, amplitude and phase do i need to creat the
> same waveshape.

Well there is, it's called Resynthesis and it is usually a software
package for one of these big digital audio workstations (like Fairlight
or Synclavier or Waveterm (or my Atari with Microwave ;->)). You have
to chop the incomming waveform into chunks (windows) of proper size, do
a FFT and voila, you've got an approximation of harmonic content.  This
harmonic data can be used to resynthesize a chunk of waveform (inverse
FFT). For good results you'll need > 128 harmonics, and I think this
amount can't be handled with analog methods any more, this is clearly
for the digital domain. Of course there is no reason why it can't be
done analog, but cost. You also need a clever method to avoid
problems at the ends of the chunks, so overlapping windows etc. might
be neccessary. I've tryed to do this with speech, and it worked
surprisingly good (only 64 harmonics and 63 chunks).  Noise components
are resynthesized by always changing from chunk to chunk, so the method
uses harmonic functions, but is (via windowing) able to cope with
inharmonic stuff as well.

m.c.