I swept the south laser with a triangle wave and optimized the mode-matching as best I could using the periscope mirrors and the translation stages. I got to a visibility of 0.3, which stinks (the maximum is 0.7 with these birefringent coatings).
I took a beat spectrum (attached) and noticed that the noise around 0.1–1 kHz is improved. Indeed, by reducing the visibility south I find the beat gets worse.
I decided some more involved mode-matching (involving beam profiling and alm simulation) is needed.
Before setting up the beam profiler, I noticed that the beam entering the cavity does not appear Gaussian, as seen on an IR card.
The laser beam entering the first Faraday isolator appears to be 1–2 mm too low. It is clipping on the input aperture, and the transmitted beam looks like crap.
Neither the Faraday nor the laser itself have any alignment adjustment knobs. I therefore had to choose between two evils: shim up the laser mount (and thereby risk having to realign the entire optical path, as well as possibly reducing the heatsinking of the mount to the table) or reinsert the PMC and move the BB EOM.
I opted for the latter: I reinserted the PMC, removed the EOAM (+QWP+PBS), placed the resonant EOM where the EOAM used to be, and then placed the BB EOM where the resonant EOM used to be.
I will optimize the alignment through these components, check the polarization, and then take a new beat.