I recovered bright alignment beam and local oscillator alignment to the homodyne detector. The homodyne detector shot noise is ok (-132 dBVrms/rtHz at 2.2 mW LO injection) but has some 50 Hz line presence so power cables should be checked.
BAB and ppol to OPO
I realigned BAB and ppol to OPO to about 87% mode matching. ppol is aligned from OPO TRANSMISS IN on the servo, BAB from the "stick" power meter in transmission of the OPO set to HI BW and range 8.8 mW. Then I optimized the ppol frequency and OPO temperature by locking OPO for ppol and maximizing the transmission of BAB with the power meter still in place. The optimal settings without green are 7.120 kOhm OPO temperature (from 7.118 kOhm before) and 245-250 MHz ppol PLL frequency (same as before, OPO generally is not hugely sensitive to this parameter). The transmitted BAB power through the OPO is 346 µW (79.2 mV on oscilloscope when power meter is set to 8.8 mW range).
Homodyne balance
I balanced the homodyne using the usual method - lock IRMC, block IR transmission through OPO. Tweak the lens near to IRMC to maximize HOM SUB DC on oscilloscope and tweak the lens further from IRMC to minimize HOM SUB DC. These steps are just to prevent clipping of the beam on the homodyne aperture. Then, adjust the input beam splitter in yaw to balance the homodyne (average level 0 V).
Homodyne noise
I checked the homodyne shot noise. At 2.2 mW IRMC transmission (LO power) I see -132 dBVrms/rtHz which is basically the nominal level. In full span (102.4 kHz) it looks flat, but at 6 kHz span there is a lot of 50 Hz peaks. Checking homodyne shot noise with all of the lights off gives not a huge amount of difference for LO shot noise so it is probably a power supply issue. I should check the plugs (next time).
Homodyne dark noise floor is about -155 dBVrms/rtHz, so more than 20 dB clearance below shot noise, but has a lot of 50 Hz peaks as well. The spectrum analyzer noise (put 50 Ohm terminator in the measurement channel) is about -164 dBm.
Misc
ppol PLL lock still bounces a lot - it will unlock then quickly relock. I noticed that it's less "bouncy" if you place the peak (by ppol laser temperature adjustment) about 15-20 MHz above the target frequency. I guess this has something to do with the "gain/bandwidth" (technically charge pump and timeout counter) of the PLL phase frequency detector.