[Joe B, Vlad B, You-Ru Lee]
We've been trying to understand a ~5% systematic error in the LHO calibration for awhile now.  After eliminating possibilities one by one, Vlad found that the H1:SUS-ETMX_L1_LOCK_L, FM6 aL1L3 filter definition in our calibration model (as well as in the h1calcs front end model) did not match the actual interferometer's definition.  Specifically, a 50 Hz pole had been changed to a 30 Hz complex pole pair on March 20th, 2024.  See LHO alog 76546.

Even though the UIM is rolled off relative to other stages quite a bit by 30 Hz, it still has a several percent contribution to the overall interferometer response there, so a ~50% error at 30 Hz due to the filter shape can contribute significantly to the overall error.

The other contributing factor is the SRC detuning effects, which up until this point have been neglected when pushing updated calibration models to the actual pipeline.

We've been looking at early September, as that is part of the time we need to regenerate due to poor 60Hz subtraction injecting noise, and specifically wanted to eliminate this small but noticeable systematic.

When I combine the fit September SRC detuning (specifically report 20240905T200854Z), along with an updated foton file (specifically ligo/svncommon/CalSVN/aligocalibration/trunk/Common/H1CalFilterArchive/h1susetmx/H1SUSETMX_1394948578.txt instead of H1SUSETMX_1394415894.txt), I get the attached plot, which seems to agree with the PCAL monitoring lines we run all the time still.  Still some residual mismatch, but I didn't do a particularly good job of lining up the average kappa values I applied and the grafana page, but it certainly will flatten out the large bump and phase wiggles at 30Hz and below.  The monitoring line data comes from Maddie's monitoring line grafana pages.

Still a bit more work to be done, and comparisons at other times, but we should aim to fix this next week.