[sdiy] SSM2164 state variable filter

[cheater cheater]

Hi David,
very interesting post, thanks.

On Wed, Jul 14, 2010 at 20:50, David G. Dixon <dixon at interchange.ubc.ca> wrote:
> Tom: sorry I didn't put a clipping network on the SVF schematic in my little
> analysis document.  It would have saved you a bunch of grief.  I was too
> anxious to get on to the Korg MS-20 emulation.  Anyway, better late than
> never; see below...
>
>> I'm playing with a SSM2164 SVF circuit again. I'm still not happy with it,
>> or perhaps still not happy with my understanding of it.
>>
>> There are several things that I've been told/learned which I have no
>> particular reason to doubt:
>>
>> 1) The circuit is essentially an oscillator, given the feedback from the
>> LP output.
>
> Without damping, the circuit is an oscillator -- a very crappy one which
> won't give anything even resembling a sine wave, since the oscillations
> aren't filtered, as they would be in a 4P cascade filter.  The oscillations
> will look more like a square wave with slanty sides.
>
>> 2) The "resonance" path from the BP output actually controls the damping
>> by cancelling the oscillation. This is why the control works "back to
>> front".
>
> Yes.  And this is why 2164 makes such a nice control element here, since it
> naturally gives a "reverse exponential" response and a gentle approach to
> the oscillation point.
>
>> 3) Some type of limiting or clipping is required for the oscillation to be
>> stable and low distortion.
>
> Absolutely.
>
>> 4) The best place for such a network would be in the resonance/damping BP
>> feedback path.
>
> No.  The limiting network (back-to-back zeners) must be placed across the
> first integrator cap (which generates the BP output).  This will give very
> nice sine waves during oscillation, and still allow nice resonance during
> filtering.  Placing the zeners across the second integrator cap (which
> generates the LP output) would still give sine waves during oscillation, but
> the filter won't work properly since the resonance will clip very close to
> the peak signal level.  Putting the zeners across the input summer's
> feedback resistor doesn't limit the oscillations.
>
> For a signal input of +/-5V, I would recommend two 1N4736 6.8V zeners.
> These will give sine waves with only about 0.01% THD from the BP output,

How do you get this number? Is that based on Spice or theory? I'm
interested in what's involved in generating it.

> and
> virtually perfect sine waves from the HP and LP outputs, and will limit the
> degree of noticeable clipping on the BP output during resonant filtering.
> Of course, when the input signal and the filter are tuned to the same
> frequency, the two will reinforce each other and then the BP output will
> clip fairly severely, but otherwise not.

Would it be feasible to have an auto-gain compressor around that which
would lessen the huge differences in pass-band volume that happen when
you're sweeping the harmonics of a wave? Could make for an interesting
sound.

> No clipping whatsoever should be
> apparent on the HP or LP outputs.  Lower-voltage zeners could be used, but
> then clipping of the BP signal becomes more noticeable (but perhaps this
> adds a bit of desirable "fuzz" to the sound...?).

I really like distorted resonance myself - I can confirm that it's
fairly useful! :)

What happens if you put clipping around the BP and the LP as well
(with enough headroom on the LP output for 'nothing to really
happen')?

> That's my take on the situation, based on simulation.  Others may wish to
> chime in with differing viewpoints...?