We continue to have 13 Hz ringups in the LSC that cause locklosses, so I took at look at what's going on with PRCL (it's the usual culprit). I measured the OLG every few minutes after we reached NLN. I found that PRCL loses a factor of 2 gain over the first 40 minutes of NLN. The current design is set to have PRCL with a UGF of 30 Hz once the gain settles after thermalization. However, this leads to >10 dB gain peaking at the start of the lock. There's overall not a lot of phase in the loop design, and if the gain gets too low this is causing the locklosses.

I looked at the loop design, and reminded myself of Gabriele's intentions with the PRCL redesign, here. His design applied the Dcntrl filter, which means to go back to the old design we only have to disengage this filter. You can see from the design comparison in his alog that the original design has a much broader phase. We might inject more noise by going back to the old design, but I'm not sure that will be an issue, especially if we continue to run the PRCL feedforward.

Turning off the Dcntrl filter gives more phase at high frequency, meaning we can increase the PRCL UGF. I increased the PRCL2 gain to 1.9 and got a UGF of 50 Hz, with about 35 deg phase. If we don't change this gain at all, PRCL will start in the lock with a UGF of 100 Hz and a phase margin of 18 degrees if the digital gain is always set to 1.9. This will then thermalize down to 50 Hz.

I also disengaged the 3.4 Hz resgain filter, because we no longer need to suppress motion at that frequency, and hopefully we'll get back a little more phase at low frequency as well.

Oli is reworking the thermalization guardian to ramp the PRCL2 gain from 1 to 1.9 over the first hour, starting after we reach max power.

I commented out the lines in lownoise_length_control that change the PRCL shape and loop gain.

I checked the PRCL noise injection with the feedforward template, and there slightly more noise injection from 20-30 Hz, but I'm not sure that's a problem.

I updated the SDF for the filter changes, and to unmonitor H1:LSC-PRCL2_GAIN since the guardian will now update that gain.

Plot 1 shows the evolution of PRCL olg with the 30 Hz UGF design, Plot 2 shows the evolution of the loop suppression

Plot 3 and Plot 4 show the OLG and loop suppression respectively with the 50 Hz UGF design

Plot 5 compares a PRCL coupling measurement from last week with today.