The goal of the shift was to implement a drift control om the BS angular control and continue the CARM offset reduction.
In yesterday's turn we already observed that, while closing in on the IR peak, the BS started moving until the beams reflected by the arms would be clearly distincted (Fig.1). We failed to resolve this issue with the local control, and this was probably the cause of the increase on the B1p_DC power that ultimately caused the unlock.
To prevent this from happening again, we decided to implement a drift control loop on the BS and add it to the metatron node that closes the drift control for the PR. Since we had already identified an error signal for the BS alignment (B1p_QD2_6MHz, it just needed a tuning of its gain and demodulation phase), the implementation of the drift control was simple: copying the same instructions already used for the PR was sufficient. We tested this new loop during a lock of the CITF (Fig.2), and it appears to be working as intended by reducing the error signal for the BS alignment and decreasing the DC power on B1p.
We added in the CITF_init.py script the functions BS_TX(), BS_TY() and BS_open() to respectively close the BS drift control on TX and TY, and to open it.
We added in the carm_darm_green_m.py script the function get_carm_darm_set() to compute the average on CARM_FREQ and DARM_FREQ.
We locked attempted a CARM offset reduction at UTC 19:24:00 and got 100 Hz away from the IR resonance. We were stopped again by the two spots appearing on B1p, despite the BS drift control. The error signals for the drift control become progressively worse while the CARM offset is reduced, until they stop behaving like error signals at all. Also, during this attempt we had to increase by a factor 4 the gain of SRCL at the moment of the hand off to the 3f.
This factor has been absorbed into the corresponding 1f/3f ratio in the .ini file of the DRMI_LOCK metatron node.
Since in the previous shifts the actuation on SRCL always tended to be dangerously close to its saturation, we modified its control filter in Acl (UTC: 19:26:00) by adding a new pole @ 65 Hz, Q=0.8, and a zero @ 90 Hz, Q=3 (Fig.3).
After these modifications we worked again on the CARM offset reduction (IR peak @ 19974 Hz).
We managed to arrive again at 100 Hz away from the IR peak (UTC: 20:09:00). During this phase, we only needed to adjust the MICH gain (Fig.4). The drift control implemented for the BS failed again to prevent from showing the same behaviour observed before (Fig.5). We tried to recover with the local controls for the beam splitter after opening its drift control, but regardless of how much we tried to move the mirror on TX, it kept drifting away until we ultimately provoked an unlock of the CITF.
In this occasion, as we tried to act on the BS local control to recombine the reflections of the arms, we noticed that we seemed to be crossing a minimum on B1p: the laser spots would get closer and then again further away from each other. This might suggest that the BS is not the only cause for this behaviour, but we couldn't find other explanations during thsi shift.
During the unlocks we occasionaly saw the photodiodes B7_PD1 and (much less frequently) B2_PD3 go into safe mode, and we had to rearm them and re-open their shutter.
We leave the arms locked on the IR, and all the ALS and DRMI nodes in down.