Elenna, Derek
Since at least May 28, there has been a consistent ring-up near 11 Hz starting 1-2 hours before lock loss for locks lasting longer than a few hours. I've attached a spectrogram and glitchgram from May 28 showing the ring up at 11 Hz. I also include glitchgrams from May 27 (that doesn't show this ring up) and May 30 (which does show this). The most recent lock on May 31 also shows evidence of this ring up, as shown in the attached live glitchgram.
Elenna notes that this could be related to PRCL no longer having an appropriate unity gain frequency (UGF) after the most recent ring heater change.
I have attached one particular scope with the LSC trends right before lockloss. During the engineering run, we ran a UGF servo for PRCL for several locks to check how the digital gain needed to be adjusted. I think that trend is now incorrect because of the change to the ring heater. I think the best thing to do would be to recommission the PRCL UGF servo for the next few locks. This will give us enough information to determine the new gain trend for PRCL. Then, we can put that information in the thermalization guardian and turn off the servo
The UGF servo will require the injection of a line that will appear in DARM around 50 Hz.
The PRCL error signal shows a glitchy peak at 12.5 Hz that increases slowly over time into a lock.
This is very likely a low-gain peaking in the PRCL loop that gets wors with time.
Looking at the error signal at the beginning of the lock and about 5 hours in, there seems to be a loss of gain of a factor of about 3, and a large peak / bump at 12 Hz
It's not just PRCL, SRCL also gets the same feature and it seems to actually be bigger. MICH just barely sees the feature. I'm attaching spectrograms of SRCL and PRCL, made with the same settings for comparison. The third plot is REFL_SERVO, which has an increase of a factor of two over the lock across the spectrum.
