Several special calibration lines have been running for the past couple of weeks so that the calibration group can assess how IFO thermalization impacts the response function (see LHO aLOG 69284). We observe the coupled-cavity pole frequency and optical gain evolving over the first ~2 h of lock when the IFO is "cold" and still thermalizing to normal high-power operations as measured by the calibration lines.

These additional lines, together with the standard set of calibration lines, have been processed offline to measure the transfer function of PCAL / DARM_IN1 = ( 1 + G) / C = R, and DARM_EXC / DARM_IN1 = (1 + G) so that we can actively track the changing response function (R) and the sensing function, derived from the combination of the measurements to isolate C [(DARM_EXC/DARM_IN1) x (DARM_IN1/PCAL)]. The DARM_EXC lines are placed close in frequency to additional PCALY lines so that is is straightforward to remove the effect of the closed servo loop.

At present the script to do this is still in "draft" form (see merge request), but we expect further improvements to analysis, plotting, and general robustness for future use cases.

I downloaded this script to my home directory (~evan.goetz/lscrepos), activated Louis' environment (conda activate cds-pydarm-test), and ran:
$ python process_sensing_darm_comb.py -i H1 -s 'may 2 2023 11:00:00' -e 'may 2 2023 11:30:00' --pydarm_ini=/ligo/groups/cal/H1/ifo/pydarm_H1.ini --pydarm_unc_ini=/ligo/groups/cal/H1/ifo/pydarm_uncertainty_H1.ini

This queries for the data and saves it so that re-running the script does not require re-querying for the data. It processes the data to determine R(t,f) and C(t,f). Attached is an initial series of plots showing the evolution from the start of the analysis period (data points in blue) to the end of the analysis period (data points in yellow). The pyDARM modeled values are shown in blue lines for comparison. This initial measurement shows the sensing function evolving from a pro-spring response to nearly no spring over the course of 30 minutes.

Future work should be to go back and run this program on all of the lock stretches for at least the last few days. Going back to the initial locks may not prove as useful because the front-end TDCF values were not correct, owing to incorrect "reference transfer function" EPICS records. Nevertheless, we can collect a few more of these "thermalization" measurements over the next several lock stretches will help inform the additional uncertainty needed for start of lock stretches.