joseph.betzwieser@LIGO.ORG - posted 12:31, Thursday 20 February 2025 - last comment - 10:15, Thursday 27 February 2025(82935)
Calibration debugging
[Vlad, Louis, Jeff, Joe B] So while exercising the latest pydarm code with an eye towards correcting the issues noted in LHO alog 82804, we ran into a few issues which we are still trying to resolve. First, Vlad was able to recover all but the last data point from the simulines run on Feb 15th, which lost lock at the very end of the sweeps. See his LHO alog 82904 on that process. I updated the pydarm_H1.ini file to account for the current drive align gains and point at the current L1SUSETMX foton file (saved in the calibration svn as /ligo/svncommon/CalSVN/aligocalibration/trunk/Common/H1CalFilterArchive/h1susetmx/H1SUSETMX_1421694933.txt ). However, I had to also merge some changes for it submitted from offsite. Specifically https://git.ligo.org/Calibration/ifo/H1/-/commit/05c153b1f8dc7234ec4d710bd23eed425cfe4d95, which is associated with MR 10 which was intended to add some improvements to the FIR filter generation. Next, Louis updated the pydarm install at LHO from tag 20240821.0 to tag 202502220.0. We then generated the report and associated GDS FIR filters. This is /ligo/groups/cal/H1/reports/20250215T193653Z. The report and fits to the sweeps looked reasonable, however, the FIR generation did not look good. The combination of the newly updated pydarm and ini changes was producing some nonsensical filter fits. This is the first attachment. We reverted the .ini file changes, and this helped to recover more expected GDS filters, however, there's still a bit of a small ~1% change visible around 10 Hz (instead of being a flat 1.0 ratio) in the filter response using the new pydarm tag versus the old pydarm tag which we don't understand quite yet, and would like to before updating the calibration. I'm hoping we can do this in the next day or two after going through the code changes between the versions. Given the measured ~6 Hz SRC detuning spring frequency (as seen in the reports), we will need to include that effect in the calcs front end to eliminate a non-trivial error when we do get around to updating the calibration. I created a quick plot based off the 20250215T193653Z measured parameters, comparing the full model over a model without the SRC detuning included. This is the attached New_over_nosrc.png image.
Images attached to this report
Non-image files attached to this report
Comments related to this report
The slight difference we were seeing in old report GDS filters vs new report GDS filters was actually due to a mcmc fitting change. We had changed the pydarm_cmd_H1.yaml to fit down to 10 or 15 Hz instead 40 Hz, which means it is in fact properly fitting the SRC detuning, which in turn means the model the FIR filter generation is correcting has changed significantly at low frequencies. We have decided to use the FIR filter fitting configuration settings we've been using for the entire run for the planned export today. Louis has pushed to LHO a pydarm code which we expect will properly install the SRC detuning into the h1calcs model. I attach a text file of the diff of the FIR filter fitting configuration settings for the pydarm_H1.ini file between Aaron's new proposal (which seems to work better for DCS offline filters based only looking at ~6 reports) and the one's we've been using this run so far to fit the GDS online filters. The report we are proposing to push today is: 20250222T193656Z
Non-image files attached to this comment
