Sines breaking down (was: AW: 35 - the next step)

[Haible Juergen]

	><<Datei: ms35b_vcf.gif>>
	>> To anyone building it: Please ignore my attempts to modify
Juergen's
	>> filter. 
	>
	>I gave it another shot. This solves my real problem with it by
getting that
	>pair of diodes out of the signal path. The sine still breaks down
if you
	>drive it too hard and it's a little noisy but I think it has better
range
	>and it sounds cleaner to me. I didn't show Juergen's new cv scheme 
	>just the changes I made to the core. I did change the output cap to
	>33uf.
	>-tg

It looks like a missing link between Korg's first and second version now  -
nice !

The two Korg versions must have different behaviour on high level input
signals
with the peak pot turned down (no feedback) indeed. The earlier version
would
limit the signal, the second version would not. This should clearly make
a difference in sound. The difference (due to *that* parameter) might not be

that large, though, because the VCAs will limit the signal as well. The old
version (1-transistor-VCA) will limit one half of the signal only, while the
new version (2-transistor VCA) limits more or less symmetrically. But
with the the combination old VCF + old VCA, there would also be
limiting of both half waves. This might or might not mask some of the
VCF's different behaviour.
Now, that was with peak = zero.

One of the real interesting things of any self oscillating VCF is how it
behaves
when overdriven while self oscillating. It's also a matter of taste: Do we
want
the input signal to drown the self oscillation, or do we want both signals 
to sum up more or less independently, or do we want the self oscillation
sine
to partly block out an input signal. All of these options have their
benefits.
I think that was also part of the difference between the Moog ladder and the
ARP ladder filter. (Joachim ?)
The Moog filter does attenuate the input signal very much when in (or near)
self oscillation. There were integrated filters that adjusted the input (or
output) gain to compensate for this. But as long as the signal swing is in
the
linear region of the filter, it's not that exciting. With nonlinearities, we
have
the chance to make both signals interact.
As a side note, my JH-4  4-pole section is designed in a way that with low
signal input level both self oscillation and the signal will be there side
by side,
and with more overdrive the input signal will override the self oscillation
more and
more, without much heterodyning etc. I did this on purpose (though the way
to
go there was trial and error), because for my taste it fits nicely with the
dark sound
of a 4-pole filter.
The MS-20 filter is quite different here, and I like it for being different.
With
a medium level input signal and peak set to self oscillation it's almost
like
a sort of ring modulator - wild interaction of both signals. With more
overdrive,
the input signal dominates, and the self oscillation part starts to sound
almost like hard sync. That's the -35 filter at least (and the 3-transistor
version
as well), but the OTA version is also described as having similar behaviour.
(At least concerning the ring mod / heterodyning effect.)

Against this background, reading your description of the "missing link
version"
	>I gave it another shot. This solves my real problem with it by
getting that
	>pair of diodes out of the signal path. The sine still breaks down
if you
	>drive it too hard
makes me curious for more information. I think the sine "breaking down" with
very high input levels is a feature, asuming that you mean the same effect
I've described above. Like the "ring mod sound" becoming a "sync sound"
(sort of). 
Does that hit the point ? And where is the OTA version in that scope ?
Both "ring mod" and "sync" -like sounds ? Or "ring mod" even when driven
hard ? 

JH.
>  
>