I spent the last few days breadboarding the moog filter, and got it mostly working yesterday. But cutoff did not work as expected. On the scope I could see that the wave amplitude got slightly smaller when turning the pot counter clockwise, but at the same time the centering of the wave dropped (adding a negative dc component) and the wave got distorted.
After a lot of debugging I switched what transistors in the CA3046 I used for the bottom pair of the ladder, the one with the common emitter. I had chosen not to use the internally connected pair because of the way I breadboarded the circuit.
This fixed things. Studying the datasheet for the CA3046 I discovered this:
Pin 13 is connected to the substrate, and should be connected to the most negative part of the circuit!
This means that the transistor between pins 12-13-14 CANNOT be used as a normal transistor after all. Switching it for 6-7-8 worked well. Instead, I tied 13 to -15V. I am using two more CA3046s on the breadboard and have to rewire them too to free up pin 13. I am having an issue where the DC component of the signal increases as the cutoff drops, not sure if that's related.
Anyway - I remembered that I have intended to use a different dual transistor for the Xonik VCO. Looking through my parts box tonight I found these:
They are matched dual transistors. They are however in tiny SOT-363 packages. For the 'production' version of the filter they pose no problem but they suck for prototyping. Still, they may be a good replacement for all transistors in the filter.
Update: They also come in SOT-666 and SOT-457/TSOP-6:
TSOP-6 seems a little easier to handle so perhaps I'll look for that instead.