You got a brand new pair of roller skates...

[Harry Bissell]

I got a Sallen-Key...

Tjanks everyone who explained the MS-20 being a Sallen-Key filter. So
I've done
a little study of an Ideal MS-20 (spice simulation) and learned a couple
of things...

The two OTA that end in capacitors are the two resistor-capacitor pairs.
To get the
best performance they should be similar, but slight changes in the time
constant don't matter too much. Yesterday I posted that adding a
"balance" control didn't work as I hoped. Thats because varying these
resistors just changes the "Q" slightly.

The third OTA (this is based on the TomG VCF8e version) is a gain stage,
affecting the
overall "Q". But it differs from the traditional Sallen-Key architecture
in that the output is taken from the INPUT to this gain stage (which is
always at a nominal "unity" gain, rather than from the output of the
gain stage... as it is in the traditional LPF. The thing that is funny
about this is that the gain stage can overload VERY EASILY... which in
the case of this filter is a "Q" killer.  So the original MS-20 uses
three diodes as a limiter here. (VCF8e does not). I've tried some
various clipping stages, including a diode-breakpoint limiter in the
output of the Gain OTA (on the current output itself, before the buffer)

I had varying success with this idea.

The simulation revealed a couple of strange things...

1) The optimum input amplitude is about 700mV-1V.  Even with perfect
op-amps and just using resistors (NO OTA...) higher input levels kill
the maximum resonance. Why is this... Any comments ???

2) The three diode clipper is just about perfect as a limiter... I tried
to scale the whole thing up... but it didn't work...

Read a paper today on Sallen Key for loudspeaker crossover... They said
most people use the state variable because the Sallen Key is "prone to
oscillation" and it can be hard to stop.  No shit Sherlock !!!

Rene mentionied the oscillation at the zero-cross as opposed to at the
waveform peaks...
in most cases the spice simulation shows that effect too...

The MS-20 (VCF8e) is a much nicer filter now that I understand its
quirks. It still seems more like a syncronised oscillator in most high
"Q" cases than it does a filter... It also gives very strange waveforms
below resonance... as it will lock subharmonics.

Everybody go to the studio/workshop and play with this puppy a little
bit tonight.
That's where I'm going...

H^)