In the morning I saw the violin modes got rung up, might be due to that the IFO kept trying to go to NLN without letting the violin modes to settle down sufficiently first. This kept causing OMC DCPD saturation, which then led to locklosses, and rang up the violin modes further. To avoid such situations in the future we might add a checking state in the guardian (maybe at the ENGAGE_DC_VIOLINS state) to monitor the violin modes and if their rms contribution is too high, the guardian should stop keep going further.

======================================================================

While waiting the violin modes to be damped, we tried to restore the O2 longitudinal MICH setup to reduce its rms value (see LHO:44693).

Specifically, we turned on FM1 in MICH1 and FM3 in MICH2 to get more gain at < 4 Hz. The boosts ate some phase at the UGF~10 Hz, therefore we also modified the lp50 filter in MICH2 to a higher cutoff freq ~ 80 Hz to get more phase. I left FM2 in MICH2 unchanged but put the new LP filter to FM5 in MICH1 (originally a 2nd order butter lp at 80 Hz).

For the attached figures:

Fig. 1 is the comparison between the MICH error signal between the new (red; O2-like) and old (blue; past few months) filter setups.

Fig. 2 shows the original MICH2 FM2 (lp50) in blue and new MICH1 FM5 (lp80) in red. The new, higher cutoff filter should give about 10 more deg of phase compared to the original one.

Fig. 3 is the OLTF of the MICH loop. The blue trace is the original OLTF used for the past few months and the red is the new one. Both are measured at 2W at the ENGAGE_DC_VIOLINS state. For the new setup we have 10 Hz BW and a phase margin of 28 deg.

Fig. 4 is the filter bank setup for the new, high-LF-gain config.