[sdiy] SSM2164 state variable filter

[Tom Wiltshire]

On 17 Jul 2010, at 09:22, Neil Johnson wrote:

> Hi David,
> 
>> Simply take some of the BP output directly back to the input summer through
>> a fairly large resistor (to give the pendulum a bit of a push, as it were)
>> -- essentially, negative damping -- but cancel this out at low resonance by
>> decreasing the damping feedback input resistor to the summer accordingly.
>> When the resonance gain VCA is cranked down, the damping will disappear, but
>> the "push" will remain.
> 
> Interesting.... do you have a schematic for this you can post or email off-list?
> 
> I think I'll need to think about this.  But at the same time I strongly suggest breadboarding it up and trying it out for real.

I've done it. The basic plan is to apply positive feedback from the BP output as well as using it inverted to damp the feedback from the LP output by sticking a resistor across the Res VCA and associated I-to-V stage.
Putting a 1M in this position works pretty well. The filter won't start to oscillate below several hundred hertz without it, but with it in place it oscillates happily from 50Hz or so. Decreasing the resistor (increasing the 'push') makes this limit go down further. I've tried 470K and 390K, and with the lower of these I can get the circuit to oscillate from below 4Hz.

Altering the V-to-I resistors ahead of the filter stages (the usual 30K input resistor you see on 2164 circuits) also helps. Decreasing this value increases current through the integrators and allows the circuit to work down to lower frequencies. This obviously shifts the overall range, so you can tweak integrator cap values to suit. I've been using 15K and 390pF.

Whilst this solves the limited oscillation range, it doesn't solve the changing oscillation level. At the very lowest frequency, the P-P level is about 3V, whereas at the top end it's about 8V. This isn't a very great variation audibly - what's that in dB? about 8 or 9?

Hanging the zeners off the LP output down to ground (as Neil does in his 3P SVF) gives an extremely flat oscillation level across the whole range (a very nice result, actually) but unfortunately it achieves this by apparently driving the oscillation *much* harder which results in a heavily distorted sinewave. I can see clear spikes of 3rd and 5th harmonic on my scope's FFT display, and I didn't need the display to hear that they were there.

I'd like to experiment using different zener values, but unfortunately I don't have any to hand. Maybe next week. The thought is that if I reduce the zener's limit, I can make sure that the oscillation is always limited by the zeners and not by some other limit of the circuit. At the moment, I don't think that the zeners have any effect at the lowest freequencies. I guess I should test that.

If anyone has any other thoughts, I'm all ears.

T.