When the new boards come from OSHPark, probably tomorrow, I already have two tests on design changes queued up: adding another LC filter on the buck converter output, and using a separate high precision regulator for the DAC reference.

There's another that I've just considered: replacing the inverting compression OP amp with a purely passive resistor based attenuator. From the DAC output a 1K resistor goes to a pair of 10k resistors to Vref and ground. The result, ostensibly, is the same output voltage range, but without an OP amp.

My original justification for the amp was Connor Winfield's application note, but the output of the DAC is fairly low impedance, and the CV input of the oscillator is high impedance. I don't see a good reason why it can't just be removed.

So that'll be tested too.

UPDATE:

I took an opportunity last night before the new boards came to test this change in isolation. With everything between the DAC and the oscillator replaced with a 1k series resistor followed by a 10k voltage divider from the 3.3 volt rail, the overnight ADEV (after letting the oscillator warm back up) was at worst a tie and perhaps very, very slightly better. So that amp is coming out. For the boards that are arriving today, a sort of mosaic of 0Ω resistors yields the equivalent circuit.