After today's changes to the PMC servo, which included increasing the loop gain, we noticed a familiar bump around 4.2kHz in DARM.  Lowering the PMC gain from 16dB to 0dB made the bump smaller in DARM (see images).  Given the coherence with REFL9, it seems that this is coupling through frequency noise.

We also made a small exploration around the already good PMC input beam alignment.  This didn't have much impact, but it revealed that small changes in the alignment can change the input offset to the PMC servo.  That is, the offset which maximizes the PMC throughput can be changed with small alignment tweaks. Using this effect, we made a 0mV input close to optimal (see second image).