I checked again the PLLs in TAMA. I don't know why but the CC fiber detector output voltages changed to 9.4 mV for ML and 4.4 mV for CC compared to last entry. Because I suspected that I connected the polarization maintaining fiber poorly before, I double checked again that they were inserted correctly. I then saw that CC PLL beat signal Vpp (output from fiber PD to oscilloscope) could be increased from 4.40 mV (176.5 degree on HWP to CC fiber) to 11.4 mV (206 degree on HWP to CC fiber). The signal to CC BEAT PLL servo input is -0.5 dBm and the CC PLL MON channel is -27 dBm (same as ppol PLL MON). I could stably lock CC for a few minutes using the settings of fig 1 and CC PLL servo set to non-INV.
I left both PLLs locked before leaving and will check if they are still locked in the morning. Maybe not, because I turned the lasers from off->on before doing this test and then they drift a bit for about an hour.
I checked again. Both PLLs unlocked overnight. The laser temperature drifted down for the ppol (maybe 50 MHz equivalent) and also down for the CC (again maybe 50 MHz equivalent).
I relocked both of them. The ppol one had a bit of weird behaviour where it does some kind of twitchy unlock relock behaviour - it skips maybe one MHz away from the lock point but then is quickly caught again. I reloaded the PLL settings (Analog Devices ADF4002 SDP board black) and turned on the Charge Pump Gain setting. I'm not entirely sure if that fixed it but I left for lunch, came back to TAMA 90 minutes later and both PLLs were still locked. Maybe I can turn up the timeout delay on the ppol too. I think fiddling with these settings too much may change the phase noise but I will check that after I recover squeezing. Perhaps there could also be a bit more optimization with ppol mutual polarization matching and laser to fiber alignmnet but for now it's serviceable. Anyway at the very least I think I have confirmed that there is no major stability issue with either PLL loop on its own. Now we see if that is maintained for CC1 and CC2 on squeezing.
Both PLLs maintained lock for 5+ hours. The weird ppol issue went away (maybe laser temperature stabilizing). Next up - checking stability of squezing coherent control.