[sdiy] [OT] Analog synths with 2 pole filters

[Richie Burnett]

Hi Tom and all,

After a bit of Internet searching I found the Jupiter 6 service notes 
detailing the two cascaded SVFs... (31MB download below!)

http://fa.utfs.org/diy/rolandjp6/Roland_Jupiter-6_Service.pdf

The capacitors marked C107 & C108 in the JP6 filter schematic would seem to 
provide phase-lead to the feedback from each SVF's lowpass outputs like the 
idea we were talking about earlier.  Maybe Roland had a similar problem with 
their SVFs squealing as you encountered with yours Tom !?

If you load up the JP6 service notes (Top of PDF page 17) it's also worth 
looking at the arrangement of back-to-back diodes beneath the second SVF 
stage.  This arrangement of diodes implements the "anti-clipping" function 
that I mentioned to stabilise the self-oscillation amplitude.  When the 
signal at the output from the first integrator (SVF's BP output) starts 
getting a bit on the hot side it begins to forward bias the diodes at its 
peaks and troughs.  This results in some damping signal being fed back into 
the mixing op-amp at the front of the SVF and the resonance decreases.  The 
normal voltage-controlled damping for the user's resonance control is 
applied via the OTA marked 48AB that sits above the SVF circuitry.  (Each 
OTA in the IR3109 is also *INVERTING* so the SVF schematic looks a bit funny 
at first.  That's why the voltage-controlled damping signal goes to the 
opposite op-amp input to the "level-stabilising" damping signal, and 
initially appears to be of the opposite phase!  They are actually both of 
the same phase, and both invoke damping action because of the inversion in 
OTA blocks 46A, 48AB etc.)

This Oberheim SEM schematic also shows a similar arrangement of C18 shunt 
capacitor (phase-lead network) in the global feedback path and back-to-back 
diodes (D11 & D12) in the damping feedback path to stabilise the SVF 
self-oscillation amplitude:

http://www.synthfool.com/docs/Oberheim/sem2sch.pdf

Going back to the Jupiter 6 SVF, the addition of the capacitors C107 and 
C108 in the feedback paths will have the side-effect of making the gain from 
the non-inverting input of the mixing op-amps turn upwards at 
high-frequency!  Not good.  So, caps C106 and C109 were likely added to 
compensate for this HF boost that would otherwise effect the filter's dry 
audio input.

Compensating for compensation circuits.  Yeah, its definitely classed as 
fannying around but if it's good enough for Roland and Oberheim....

-Richie,