Sheila, Camilla.
New SQZ ASC using AS42 signals with feedback to ZM4 and ZM6 tested and implemented. We still need to watch that this can keep a good SQZ alignment during thermalization. In O4a we used a SQZ ASC with ZM5/6, we have not had a SQZ ASC for the majority of O4b.
Prep to improve SQZ:
- No squeezing from 15:31:30 to 15:41:30UTC.
- Adjusted OPO temperature following no observing instructions in 80461. Measured NLG to be 12.8. (0.1165/0.0091) following 76542.
- This improved the squeezing from 2.5dB above 300Hz to 3dB at 300Hz and 4.2bD above 1kHz.
Testing ASC from 80373:
- Ran the "reset as42 nosqz" offsets script while in no sqz. SDF accepted.
- Put the input matrices form 80373 into SQZ ASC and checked they were diagonalized correctly by moving ZM4/6 in turn and watching ASC_POZ/ANG input signals. ZM4 pitch wasn't completely diagonalized but that others looked good.
- Reduced ASC_WFS gain from 0.1 to 0.05, cleared histories, set all ASC_POS/ANG gains to 1 and one at a time turned on the inputs. We had some troubleshooting with POS_Y but there was a minus sign inside one of the filter banks we hadn't seen. In this troubleshooting we found the matrix was mostly diagonal so removed the cross-coupling terms for the YAW input matrix.
- Found WFS gain is better with lower WFS gain to 0.01 so we took a factor of 100 for the input matrix and left H1:SQZ-ASC_WFS_GAIN at 0.5. Accepted all in sdf. And edited sqzparams and reloaded SQZ_MANAGERT to turn on SQZ ASC.
- To test, we tried the "offload AS42 ASC" script with the plan to compare the AS42 alignment with the alignment from the guardians scan SQZ alignment. The script didn't work as expected, it's hard-coded to offload ZM45/6 rather than ZM4/6 that we are now using, we should fix this or go back to using ther guardian states. After turning off the ASC and running again to clear histories, this caused a lockloss, probably as ZM6 P moved 60urad in 0.3s. See attached lockloss plot.
In the first 20 minutes of the lock, the SQZ ASC appears to be working well, plot.
Note to operator team: if the squeezing gets really bad, you should be able to use the SQZ Overview > IFO ASC (black linked button) > "!graceful clear history" script to turn off the SQZ ASC. Then change /opt/rtcds/userapps/release/sqz/h1/guardian/sqzparams.py use_ifo_as42_asc to False and go though NO_SQZEEZING then FREQ_DEP_SQUEEZING in SQZ_MANAGER and accept the sdfs for not using SQZ ASC. If SQZ still looks bad, put ZM4/6 osems (H1:SUS-ZM4/6_M1_DAMP_P/Y_INMON) back to when squeezing was last good and if needed run scan sqz alignment and scan sqz angle with SQZ_MANAGER.
Sheila moved the "0.01:0" integrators from the ASC_POS/ANG_P/Y filters into the ZM4/5/6_M1_LOCK_P/Y filter banks.
This will allow us to more easily adjust the ASC gains and to use the guardian ZM offload states. We turned them on on ZM4/6. Edited OFFLOAD_SQZ_ASC to offload for ZM4,5,6. And tested by putting an offset on ZM4. We put ZM4/6 back to positions they were in in lock via the osesms. SDFs for filters accepted. I removed the "!offload AS42" button from the SQZ > IFO ASC screen (liked to sqz/h1/scripts/ASC/offload_IFO_AS42_ASC.py) as it caused a lockloss yesterday.
Oli tested the SQZ_MANAGER OFFLOAD_SQZ_ASC guardian state today and it worked well. We still need to make the state request-able.
ASC now turns off before SCAN_SQZANG_FDS/FIS and SCAN_ALIGNMENT_FDS/FIS. It wil check if the ASC is on via H1:SQZ-ASC_WFS_SWITCH and turn the asc off before scanning alignment or angle.
We changed the paths so that to get from SCAN_SQZANG_FDS/FIS and SCAN_ALIGNMENT_FDS/FIS back to squeezing, the guardian will go though SQZ_ASC_FDS/FIS to turn back on ASC afterwards.