Oscillator log/antilog amplifier conformity

[ilya]

I'm restoring an old Russian synth and one of the oscillator modules is puzzling me. The problem is scaling. For some reason it has non uniform scaling that differs for different octaves. If I tune it for top octave, the lower octaves scaling is different (that is, when register input summer has -3.5 or -4V at its input).
At first I though that transistor array is to blame and changed it. But this didn't help. I'm also using low Vos op-amps for register and CV summers, as well as for log amp.

This looks like a lack of the log/antilog converter conformity, but what part of the circuit exactly should I look at? My first though was that I had mismatched transistors in the array, but which ones should be matched and what parameters?

I'm at a loss as to where to look for the solution of this particular problem. Any thoughts?

[mediatechnology]

I'm still thinking about this one.

Regarding the low-value compensation and other capacitors and others: Are any by chance old ceramic discs?

I had a customer once that got some bad ones that actually became "leaky" enough to actually, when used as bypass, get warm.

That would be the absolute last place anyone would look to solve a log conformity problem.
I'm thinking C1-C4 in the expo converter or maybe even C5 in the VCO.

[ilya]

Hmm.... That's an interesting idea.

C1 and C2 are indeed ceramic discs (or rather "flags" - square red caps). C3 and C4 are films, and C5 is, I believe, also film or mica.

C4 is a compensation cap for A5, which is a precision op-amp that works as a thermostat controller (one of the transistors is used as a heater to keep the array at an elevated temperature), so I doubt that changing this cap will cause any effect on scaling.