[sdiy] SSM2164 state variable filter

[David G. Dixon]

> Trouble with that approach is that at low frequencies the current
> output from the 2164 gain cell is pretty small.  At this level the
> leakage current through the back-to-back zeners becomes important,
> and in effect you have a leaky integrator.
> That's why in my filter design the limiting zeners are not across any
> of the integrator caps, but sit on the LPF feedback path.


You are probably right.  However, according to simulation (my crutch, I'm
afraid), the sine waves are holding at 6.2V (from 6.8V zeners) at 20Hz.  The
peak leakage through the zeners is about 250pA, compared to a peak current
through the integrator of about 2000 times that much at 20Hz.  And, of
course, one could always decrease the 2164 input resistors and increase the
integrating capacitance to increase the current through the integrators.

Hence, as long as the cutoff frequency is constrained to the audible range,
the loss of sine wave amplitude should be minimal.  Given the general
difficulty of controlling oscillation in SVFs, I don't think this is the
right sort of filter to use as a quadrature-wave LFO.  A 4P cascade filter
like the Mankato is much better suited for that.

Of course, real life may not reflect the simulation results, but Tom's
experiences give me some hope in this regard.