There is some instability in the ALS system which drives the IFO Y arm out of lock in a matter of just a few hours, causing the ETMY HEPI to saturate before quickly breaking lock. This is a very low frequency PDH noise source, with a period varying from one to four hours. This long term issue has occurred for all locking periods of substantial length (greater than two hours in length) that I have seen over the last month. Both time ranges with HEPI displacement and VCO frequency are shown in the first two plots attached, and the entire locking period is shown; the lock is lost at the right edge of the plots. In the last plot I include instead the calibrated signal for Y-end green laser frequency, and plot the mean. 

We have checked and ruled out the following as the main source of the unstable displacement noise:
- checked the HEPI calibration (Hugh and Vincent measured driven HEPI displacement on ITMY)
- tidal strain (the discrepancy between HEPI offset and tidal strain is what caused the instability to be discovered)

Possible causes we have not ruled out:
- control systems instability
- unexpected PSL frequency variation
  - most likely to reference cavity temperature fluctuation
- excess VCO noise

Addressing the first hypothesis, I've started to model the PDH slow and fast loop transfer functions to attempt to find their crossover frequency and determine whether this might correspond to the instability. Without a detailed HEPI transfer function, this is not exact, but any crossover frequency looks to be close to 1 Hz, much higher than the instability, so this would appear not to be the most likely cause, as far as crossover instability is concerned, barring an error in the model.

Addressing the second hypothesis, I am comparing HEPI offset and green laser frequency with PSL reference cavity frequency to see if the excess noise is introduced by the PLL as opposed to PDH loops, and looking at the third attached plot this seems to be highly likely. I am computing power spectra for each of these signals to comfirm/reject this hypothesis. So far it does look promising, especially considering the fact that reference cavity temperature was used previously to offload tidal displacement, and thus should have a large effect on the ALS arm stabilization.

The third hypothesis is not currently being investigated as it seems less likely than the second, but should be accounted for at some point regardless.