26 Aug
2007
26 Aug
'07
9:32 p.m.
Something I've been wondering about recently is this - if you implement
a digital filter, how do you maintain stability when it's pushed into
self-oscillation? Most "real" synths seem to put out a slightly clipped
sinewave when in self-oscillation. Might it be as simple as an atan()
clipper in the feedback loop?
Gordon
27 Aug
27 Aug
7:27 a.m.
On Mon, Aug 27, 2007 at 12:32:26AM +0100, Gordon JC Pearce wrote:
Something I've been wondering about recently is
this - if you implement
a digital filter, how do you maintain stability when it's pushed into
self-oscillation? Most "real" synths seem to put out a slightly clipped
sinewave when in self-oscillation. Might it be as simple as an atan()
clipper in the feedback loop?
Yes.
You can use atanf(), tanh(), x / sqrtf (1 + x * x), ...
--
FA
Follie! Follie! Delirio vano è questo !
8:12 a.m.
On Mon, 2007-08-27 at 11:35 +0200, Fons Adriaensen wrote:
On Mon, Aug 27, 2007 at 12:32:26AM +0100, Gordon JC
Pearce wrote:
So that would hold it around unity gain, maintaining the oscillation
without going hugely offscale?
I did briefly consider a simple "compressor" in the feedback loop with a
shortish (a few cycle's worth) time constant, the idea being for it to
work like the lightbulb or thermistor in a Wien Bridge oscillator.
Gordon
Something I've been wondering about recently
is this - if you implement
a digital filter, how do you maintain stability when it's pushed into
self-oscillation? Most "real" synths seem to put out a slightly clipped
sinewave when in self-oscillation. Might it be as simple as an atan()
clipper in the feedback loop?
Yes.
You can use atanf(), tanh(), x / sqrtf (1 + x * x), ...
11:18 a.m.
On Mon, Aug 27, 2007 at 11:11:39AM +0100, Gordon JC Pearce wrote:
On Mon, 2007-08-27 at 11:35 +0200, Fons Adriaensen
wrote:
Yes. Once the oscillation is sustained the amplitude will settle at a
level that corresponds to exactly unity gain.
You can use atanf(), tanh(), x / sqrtf (1 + x *
x), ...
So that would hold it around unity gain, maintaining the oscillation
without going hugely offscale? I did briefly consider a simple "compressor"
in the feedback loop with a
shortish (a few cycle's worth) time constant, the idea being for it to
work like the lightbulb or thermistor in a Wien Bridge oscillator.
That would probably be much more difficult - 'a few cycles' depends on
the frequency, and depending on the type of filter you have in the level
detector, the whole thing may start behaving like a pulsar rather than
as an oscillator (can be fun as well). But it can be done.
Anyway, maintaining a constant Q (and a fixed feedback level at which
oscillation will be sustained) is not obvious at higher frequencies,
say above Fs/8. I remeber it was one of the more difficult aspects
when I wrote the Moog Filter plugin years ago.
--
FA
Follie! Follie! Delirio vano è questo !
6234
days inactive
6235
days old
linux-audio-dev@lists.linuxaudio.org
3 comments
2 participants
participants (2)
-
Fons Adriaensen -
Gordon JC Pearce