[Rich, Johannes]

Following the previous observations and Rich's suspicion that the observed resonance comes from an oscillation in the HF modulation, we opened up a spare current driver that wasn't in use (but supposed to be identical to the ones used for the cryo cavities) and found that it didn't come as advertised. The attached schematic shows it's shortcomings. The capacitor C39 had been modified (had a test point attached and capacitance added with a secondary cap.), and R40 was still at 1kOhm as in the original version (Note 1 for version 6 on the document says it has been replaced with 0 Ohm due to high frequency oscillations - duh). We placed 100 pF and 0 Ohm as supposed, which removed the 8MHz feature in the transfer function.

I proceeded to apply this fix to both in-use current drivers, the resulting amplitude modulation transfer function is attached. The >1MHz region is still not feature-less, but the ripples do not exceed 5dB above the DC response, as opposed to 40-50 dB before, and the phase looks smooth.

04/19 edit: Along with the change to the current driver, I also changed the lowpass on the demodulation to the less aggressive mini circuits SLP-21.4+, which achieves stopband attenuation at ~60MHz and is therefore sufficient to filter the 2omega terms. With these changes I was able to bring the UGF of the cavity locks up to 300 kHz.