[sdiy] SSM2040 filter question

[Antti Huovilainen]

On Wed, 12 Nov 2003, Czech Martin wrote:

> Recently there was a discussion about the Synthacon
> filter. Another structure.
> If I remember right.

Does someone have URL for the schematic?
Also, where can I find schematics for SVF filter designs that don't use
the linearizing diodes of an OTA.

> better amplitude response, but this can
> make phase things worse.

Or better. The goal is to get total phase shift of 180 degrees at cutoff
frequency for the feedback loop. This includes unit-delay that introduces
phase shift that is linearly dependant on frequency (180 degrees at Fs/2).
Using a combination of impulse-invariant and other sections can compensate
for this quite effectively, so that the amount of tuning needed is not too
much. I don't use any tuning compensation and at 10khz the
self-oscillation frequency is about one semitone off.

> For feedback structures this is "poison", because the stability
> properties of the circuit will be modelled wrong.
> IMHO the sample rate must be raised so that the phase properties
> inside the bandwidth of interest are not affected.

For general case, yes. For first order filter with not-too-severe
nonlinearity (I'd say almost all practical filter circuits), this is not a
problem.

This DOES represent a problem for resonance, but there are techniques to
compensate for that (that don't care whether the filter is linear or not).
In most 4-pole filters, resonance happens at the frequency where filter
phase is 180 degrees. For self oscillation, the total amplification of the
filter at this frequency must be 0.25. So you need to compensate for both
tuning (phase not 180 degrees) and feedback amplitude.

Stilson & Smith's Moogvcf paper deals with this although the approach
taken is unnecessarily complicated IMO.

> IMHO the Wiener-Voltera approach allows for a black box analysis
> of any nonlinear systems, as long as no "hysteresis" is involved.
> I.e. no ambiguity.

Does it allow for time-variance?
I had an argument about this when I did some tube simulation (using my
"own" method). There the biasing is time-varying and depends on the
(average) input amplitude.

Antti

Give a man a fire, and he'll be warm that day,
Set him alight and he'll be warm for the rest of his life