new Formant book arrived

[jh.]

>>Maybe it doesn't lock. In OTA polygon filters with 1/1+s it does lock
>>because of the "short" feedback loop that turns intopositive feedback.
>
>Ok, but the same feedback path ensures that there is only a very small
>input differential voltage for the OTA. I've been very surprised by this,
>seeing that in this configuration the input divider doesn't play a big role
>(if any), and that the circuit is quite linear, more than what I had
>expected. I have to mention that it was Jörgen who pointed me at this.

That refers to MS-20 type filters and Polygon filters now ? (just to
be sure). For low frequencies, yes. Feedback loop linearizes the
OTA's input stage. But for higher frequencies, loop gain decreases
and nonlinearities become valid again. In LPF polygon filters, the
amplitudes for later stages are lower at these critical frequencies,
because it's filtered by previous stages, so you won't notice much
unpleasant artefacts. The one case where you *really* have to fight
nonlinearities is the all pass filter - there the soft clipping sounds
nasty IMO.
SVF (which started the thread) is different. SEM and Wasp and ARP
have these diodes that increase damping of the filter - that's the main
limiting factor there.
The (old and new) Formant SVF doesn't have these diodes. Speaking
for the old one, it sounds remarkably good nevertheless. Overdrive
*is* more on the harsh side than a SEM for instance, though.

>Overdriving results in harsher clipping, rather than the soft tanh curve
>which one might expect. (Leaving reversal aside.)
>The circuit in question is my MS20 filter "clone", which is actually very
>sensitive to amplitude, but merely because of that nonlinear feedback loop.

Yes.

JH.