Re: [sc-users] make your digital oscillators sound like a moog!

["Sean Costello" <seancostello2003@xxxxxxxxxxx>]

From: "kernel" <le_kernel@xxxxxxxxxx>

> here, here.  that filter is pretty expensive.  I like the way the sound
> doesn't break up at high cutoff/ressonance - maybe that's the
> oversampling?

Probably. The tanh() spread out between the stages (and inside the filter
stages) does a nice job of producing smooth distortion, while the
oversampling avoids aliasing at high cutoff frequencies. Without aliasing,
filters with nonlinearities can have obvious foldover sounds at high
cutoff/resonance, as the nonlinearities produce harmonics of the resonant
frequency, and these harmonics can fold around Nyquist.

I would also like to try the Antti topology filter (the one I am talking
about, which I saw posted by Bhob Rainey) with different nonlinearities in
the one-pole sections. A 3rd order polynomial would be useful. A 2nd order
parabolic clipping is really cheap if you have fast fabs() and clip
instructions. Does anyone know if fabs() and clip have fast assembly
versions on PowerPC, or for the new Intel?

Of course, this would involve getting the hard drive out of my iBook, which
is making the "click of death" right now.

> there's no control rate on analog obviously.  so that's why I always
> use audio rate LFO's - sounds bigger somehow.

Buchla synths did have control rate. The VCAs and filters were driven by
Vactrols, which are a combination photocell and light source. The Vactrols
have very low noise characteristics, but have a relatively long lag time, so
it is not possible to do audio-rate modulation in circuits that use them. In
addition, the attack and decay characteristics of Vactrols are different,
with the attack being shorter than the decay. The good news for Buchla is
that this resulted in envelope characteristics with very nice sounds.

The control rate in SC3 does not behave like this - it simply takes in
values at the tick rate, and linearly interpolates between them at the
sample rate. This is more useful for most audio applications. Some people
like exponential decay, which can have whatever time constants you want. For
emulating a Buchla, you would want to use exponential decay (a one-pole
filter) running at the sample rate, with different time constants depending
on whether the modulation signal is increasing or decreasing in value. In
other words, yer basic digital AR envelope detector.

Sean Costello