Looking at the dat from the night between Jul 9 and Jul 10, the issue was that once OMC1 is locked, the DARM control is transfered to DC read-out, and the first reduction of dark offset is done, then after 10-20s the DARM loop start oscillating at ~20Hz (the lower UGF of the DARM loop) and the ITF unlocks.
The issue was likely to be that the dark fringe offset of that first reduction was changed from 0.00055 to 0.00035 without adjusting the gain of the DARM loop, which reduced the DARM loop gain.
On July 10 in the morning, after getting confused by the messy automation logic of dark fringe reduction, I have restored the dark fringe offset to be 0.00055 after OMC1 locks. This worked once, then I have tried to lock OMC2, which unlocked the ITF as the B1 weight into the DARM signal had not been changed for the switch from 2 B1 PDs to 1 B1 PD, which caused the dark fringe offset to change by a factor 2 each time OMC2 went through a resonance (and likely caused the unlock).
After this the reduction of dark fringe offset to 0.00055 after OMC1 lock stopped working, with again an oscillation at 20Hz of the DARM loop, and the alignment running away (B1p power increasing steadily over a few seconds and then unlock). Several changes in the AA control had been tried with no success, one issue being that with many unlocks in the row the alignment was not good and the WE LC opened and excited the suspension several times.
At the end, we changed tactics and went the slow route to lock. Waiting 10min in PRITF_DF, then ~30 mins in LN1 for ITF and tunig the alignment, the locking OMC1 without a dark fringe reduction (DARM set 0.004). Then reducing slowly (over the course of 1 hour) the dark fringe offset to 0.00003 by hand while increasing the DARM gain from 2.1 to 15.0. Then locked OMC2 without issues and reached LN3.
We left the automation with a dark fringe offset of 0.00055 and a gain of 10.5 just after OMC1 locks (the value that had been used for the past 3 months), and then a further reduction to 0.0003 and gain 15.0 before starting OMC2 lock, but this automation hasn't been tested (and we didn't dare unlocking to test it).
We have seen during the dark fringe offset reduction (during the ramp) that the DARM loop start oscillating at ~150Hz (the upper DARM loop UGF). This is likely due to the linear ramp used for the gain. The gain adjustment for the DARM loop should scale as sqrt(P_B1), whereas we are making the ramp gain proportional to P_B1, which means that in the middle of the offset/gain ramp the gain is too high. Changing this may help with dark fringe offset reduction transiton.
Figur 1. Also even though SDB2 was controlled and fixed in terms of local control postion, looking at the beam position on the bench camera it has drifted during the first night from July 9 and July 10 by 50-100um. It is possible that a thermal transient is affecting the bench local controls causing this drift.