I checked the polarization of the CC beam by looking at its transmission and reflection through the first polarized beam splitter on the injection path. It is almost completely reflected with very low transmission, so basically is s-pol. I guess the polarization wouldn't suddenly change for no reason but might as well check anyway.

I replaced the battery for the fiber PD and it turns out that fixed the issue with the disappearing CC PLL. I thought this fiber PD was outputting an electronic signal based on what I saw connecting it to the oscilloscope, but it seems that in fact the battery had run out. I used a Panasonic 12V A23 size battery from Yodobashi Camera, though Thorlabs only recommends Energizer ones from their website. Also the optimal laser temperature is about 37.96 C, somewhat less than what is listed on the wiki. If this remains consistent I will update the wiki. 

A scanning amplitude is sent to the CC1 error signal via the knob on the control servo to make it a sinusoidal shape, and from there we see the peak to peak level that is representative of the total squeezing (once the scan amplitude knob on the servo is turned past a certain number, the Vpk of the error signal is independent of the scan amplitude). I measured 560 mVpk at 25 mW green injection, which is about twice as much as what it was last time, the difference being that I optimized green alignment to the OPO using nonlinear gain of BAB last week.

I attempted to measure squeezing at 195 (39*5) MHz ppol but the glitch noise is extremely bad. The shot noise level while blocking squeezing is stable (-132 dBm, maybe slightly unbalanced homodyne but not important right now), so the noise has nothing to do with the homodyne path (we thought it might be an issue with homodyne power supply). I could barely make out that I was in a squeezed quadrature before the noise level glitched upwards by about 15 dBm.

After some time, the CC loops were behaving strangely, so I unlocked and relocked and noticed the green phase shifter HVD was sending 0.0 V and that the "LOCK" and "RAZ" lights on the GRMC/MZ servo boards were flashing. I saw that the green mode cleaner was *again* misaligned, which has become a repeating issue whenever large changes in PZT voltage get sent to the green phase shifter. I quickly recovered MZ/GRMC alignment to 92% mode matching but now the CC1 error signal was making some weird non sinusoidal shape. It was getting late so I didn't confirm, but I suspect that once again the green has misaligned to the OPO.

This CC glitch noise and green path hysterisis is very annoying and is severely impeding the ability to do priority tasks (ML data taking and OPO replacement). There are 4 points of suspect performance in my opinion - the CC PLL, green phase shifter, CC loop stability and CC loop signal to noise ratio. I personally do not believe the CC PLL is the issue since we measured the phase noise recently and it didn't have any of the glitch behaviour. We also tried to swap the boards ADF4001/ADF4002 used for CC and ppol, although now that I think about it I didn't do it correctly, but now I know the correct procedure - when the PLL software is loaded and the board is selected, I thought the initial screen was the correct settings, but in fact it is wrong and you have to load the correct settings for CC/ppol. Also, the signal frequency is set in the DDS software, but we have to input f*3 for CC frequency and f/5 for ppol frequency (i.e. above I input 39 MHz for a 195 MHz ppol lock). Last time we tried, I applied "division by 5" when switching the CC to the PLL board, but in fact the correct loaded settings also do the multiplication/division of the number coming from DDS, so I should have put 21 MHz (multiplication by 3). But even so, like I said, I want to check the CC control loop properties since I am more inclined to believe that these are the problem.