Kevin, Evan
We slightly adjusted the gain of the Michelson feedforward filter to remove some broadband coherence with DARM around 60 Hz. Ideally both MICH and SRCL should have a proper retuning, which we did not do.
We instead wondered if there was any improvement to be had by reshaping the loop to try to reduce the amount of sensor noise reinjection from MICH and we were soon reminded that no, there isn't. We could run without the "LP80" filter in LSC-MICH1, but this reduces the noise reinjection by only about 3 dB around 60 Hz, while removing suppression above 100 Hz. This tradeoff seems to be irrelevant since the amount of residual coherence with DARM in these regions can be brought to less than 0.01 anyway with feedforward.
The real problem is that MICH, which currently has a UGF of around 7 Hz, has only 3 dB of gain margin, about 18 degrees of phase margin, and more than 12 dB of gain peaking (gray traces). We have run stably like this for some time, but the situation is not ideal. As a first step toward remedying this, we have turned off the lag compensator LSC-MICH2 FM3, which is a pole at 1 Hz and zero at 4 Hz (pink traces). The MICH error point rms was 2 pm, so we expect now it is more like 8 pm. This does not yet remedy the low gain margin, but it does improve the phase margin to 36 degrees and lessens the gain peaking somewhat to 9 dB. We may want to re-examine some of the other boosts to see if we can win back a bit more phase in the 5–10 Hz range.
In addition to turning off the 4:1 Hz compensator LSC-MICH2 FM3 last night, I have now addtionally tweaked the 1 Hz integrator LSC-MICH1 FM2 to get some phase back at a few hertz. Overall, the situation now should be that we have 12 dB less gain near and below 1 Hz compared to the status quo. By my reckoning, the loop shape now is roughly 1/f4 between 0.1 Hz and 1 Hz.
I remembered Sheila saying that we probably do not need the BSFM bounce/roll notches anymore in the Michelson loop, and indeed we can run stably without them in the low-noise state (the guardian will still engage these notches during acquisition).
After these interventions, the MICH UGF is about 6 Hz, with 40° of phase margin and 5 dB of gain margin. The gain peaking is about 7 dB and occurs around 4 Hz. The residual displacement is about 10 pm rms, when the microseism right now is about 0.1 μm/s; during high microseism this will likely rise to several tens of picometers. The SRCL residual in the current condition is already about 25 pm, and the PRCL residual is about 0.3 pm.
