I reassembled the south FSS in the experiment and re optimized the gain settings.
Looking at the transfer functions, there is some gain peaking at 224 kHz and 427 kHz in both paths. This limits the stability of the loop once the common gain is turned above around 2.4 V (applied acromag voltage). This limits the maximum North FSS bandwidth to 195 kHz. South FSS bandwidth is limited to around 150 kHz.
However, we don't really need to push loop bandwidth very high for now. I've backed off the common gain and moved the crossover frequencies to ~10 kHz in both paths. By simultaneously looking at the crossover (on SR785), common TF (on Agilent 4395A) and the spectrum of error signal tapped off directly after the mixer (so can see any DC offsets along with spectrum), I found some settings with >100 kHz UGF, and clear of instability caused by HF PZT resonances and actuator railing.
I forgot to note down the phase margin but from last time I did this my hand notes say I got bode phase of -102 deg @ 123 kHz (south) and -98 deg @ 158 kHz (north). The phase is pretty flat around this region so should be close to this for the above measurements.
We shouldn't need to optimize beyond this until we have done some scatter noise diagnostics and reduction.
I also tweaked the polarization alignment into the ISS BBEOMs. For some reason I had chosen horizontal polarization into the south BBEOM used for intensity stabilization, I changed this to vertical. Both should work but vertical is consistent with the north configuration.
I checked some powers before and after the refcavs.
Incident on cavities: South = 1.212 mW, North = 1.247 mW
Reflected back to PDs (off resonance): South = 0.970 mW, North = 0.960 mW
Power promptly after transmission: South = 575 µW, North = 615 µW
ISS PD: South = 122 µW, North = 288 µW
Power to BN detector after BS: South = 92.8 µW, North = 193 µW
The light from the final combining BS to the PD is also attenuated, I didn't get to measuring this.
Light incident from north and south on the BN detector combining beam splitter is about the same from north and south (~300 µW), I discovered that a p-polarized BS (Newport 10Q20HBS.33P) was used instead of s-polarized (all the PLL BN board is S after the cavities). This explains the discrepancy in power from the south path. We don't really need to boost the size of the BN for now, but this BS in the PLL path should be replaced with an s-pol 50:50 at some stage.
The discrepancies is in the ISS are a similar thing, a p-pol optic was used before the north Thorlabs ISS detector, giving the wrong splitting ratio. We would like to replace this eventually with the M2 ISS and its not critical to have much more power for this at the moment anyway.