Calibration suite was run after 5 hours of thermalization. Before the measurement was run, work had been done to update the SRCL offset (85362).
Broadband
Start: 2025-06-26 20:35:38 UTC
End: 2025-06-26 20:40:49 UTC
Data: /ligo/groups/cal/H1/measurements/PCALY2DARM_BB/PCALY2DARM_BB_20250626T203538Z.xml
Simulines
Start: 2025-06-26 20:42:09 UTC
End: 2025-06-26 21:05:29 UTC
Data: /ligo/groups/cal/H1/measurements/DARMOLG_SS/DARMOLG_SS_20250626T204210Z.hdf5
/ligo/groups/cal/H1/measurements/PCALY2DARM_SS/PCALY2DARM_SS_20250626T204210Z.hdf5
/ligo/groups/cal/H1/measurements/SUSETMX_L1_SS/SUSETMX_L1_SS_20250626T204210Z.hdf5
/ligo/groups/cal/H1/measurements/SUSETMX_L2_SS/SUSETMX_L2_SS_20250626T204210Z.hdf5
/ligo/groups/cal/H1/measurements/SUSETMX_L3_SS/SUSETMX_L3_SS_20250626T204210Z.hdf5
Current version of the pydarm report can be found at /ligo/groups/cal/H1/reports/20250626T204210Z_prospring/H1_calibration_report_20250626T204210Z.pdf. We are investigating further into why the calibration is so different now.
After inspection on the sensing function from the initially generated report, we changed the model to an anti-spring and to start the fit at 8 Hz, to see if we could get a better calibration from this measurement.
First, we set the is_pro_spring parameter in the H1_pydarm.ini file from True to False. We expected this parameter make the model (orange line) fit better to the measurement (green dots). Overall, there was no noticeable change of the sensing model compared to the initial report, as seen in the first figure (CAL_SENSING_MODEL_ANTISPRING_20250626.png). Additionally, the sensing MCMC corner plots were not gaussian (opposite to what is instructed in T2400215 section 2.4.2), as seen in the second figure (CAL_SENSING_MCMC_CORNER_ANTISPRING_20250626.png).
To improve on the sensing model, we decreased the sensing parameter mcmc_fmin from 10 to 8 Hz. Even though the sensing function is slightly worse at low frequencies compared to the initial report (the one from oli's alog), the sensing corner plot shows more of a gaussian behavior. Additionally, the uncertainty is still within our 10% budget (see snippet of the calibration monitor from grafana CAL_MONITOR_GRAFANA_POST_CALMEAS_20250626.png), so we will pause on this investigation for now.
The updated report is attached as a PDF file. The measurement has not been tagged at the time of posting this comment. We have yet to understand why the model is struggling with the fit parameters.
