[sdiy] 36dB VCF output drop while increasing emphases

[David G Dixon]

> -----Original Message-----
> From: Olivier Gillet [mailto:ol.gillet at gmail.com] 
> Sent: Sunday, April 19, 2015 2:03 AM
> To: David G Dixon
> Cc: Harald; Chris McDowell; synth-diy diy
> Subject: Re: [sdiy] 36dB VCF output drop while increasing emphases
> 
> It would be more accurate to say that we got our inspiration 
> from the same sources...
> 
> Thomas Henry, Osamu Hoshuyama and Roman Sowa had all used the 
> 2164 as a 1-pole LPF cell before.
> 
> As for the resonance trick, I got it at the source while looking at
> IR3109 filters (it's straight from the Jupiter-8 schematics, 25A OTA :
> http://fa.utfs.org/diy/roland_filters/JP8.jpg). My previous 
> attempt was to sum the output of the resonance VCA with the 
> 4th stage, and let people customize the soft-clipping they 
> wanted by installing back-to-back diodes there.
> 
> Actually I use none of the design choices unique to the Dr 
> Octature (switchable gain for each stage, 2164 for the 
> feedback VCA). I see two benefits in using an OTA over a VCA 
> for the resonance control here :
> differential inputs save you an op-amp, but more importantly 
> the OTA non-linearity is what stabilizes the filter and 
> provides the "right"
> flavour of soft-clipping at high-resonance.

Fair enough.  In fact, there are four resonance gain stages in the Dr.
Octature, and they are all 2164 VCAs.  You are right that the 2164 is not
the best choice for feedback in this type of filter, because the exponential
response is exactly the opposite of what you want.  However, I overcame this
by actually using the 2164 to subtract signal from the chain, in what I call
the "reverse exponential response", and I believe that this gives a much
smoother approach to high resonance than even linear amps.  In other words,
the Doc Oc is hardwired to be in self-oscillation, and the resonance amps
decrease the overall gain from 4 to 1.

As far as soft clipping, the latest incarnation of the Doc Oc uses zener
clippers on the input.  These conduct when the overall gain of the input
signal is pushed above unity.  The filter itself softens this clipping.
There are zener limiters on the feedback loop, but these are to limit the
magnitude of the sine waves in self-oscillation to 10Vpp -- they really have
little or no effect on the filter when it is used as a filter.  The flat
tops of the resulting fed-back sine waves are filtered out by the filter
stages.  Of course, the sine wave from the fourth stage is the cleanest of
all, but they're all pretty clean nonetheless.