J. Kissel

We continue to review all of the identified "times of concern" during O3B where there was an obvious change to the detector configuration (see T1900248 for configuration changes, and the dependency tree of IIET Ticket 12985) that might result in a systematic error that is not accounted for with the standard time-dependent correction factors, and must be specially treated in the hourly budgets. 

Last week, we ruled out the importance of considering the Nov 27 - Dec 03 UIM coil driver configuration switch (see IIET Ticket 14564 and references therein).

Evan and Lilli are working on accounting for the systematic error incurred due to the lack of thermal equilibrium in the first two hours of nominal low noise stretches that occur just after power up (see IIET Ticket 14582 and references therein).

In this aLOG, we check if the issue caused by the rotation stage automation failure -- which resulted in significant drop in PSL power in to the IFO (see IIET Ticket 14726) -- was sufficiently covered by the correction for time-dependence of the IFO optical gain (\kappa_C).

In passing, I also update the trend of all sensing function time-dependent correction factors (TDCFs), using data from the offline, C01, DCS, pipeline for the entire "chunk 1" from Nov 01 2019 to Jan 14 2020.

In short: yes -- the TDCFs cover the power change, and we don't need to do anything different in the hourly estimate of systematic error.

My justification for this is crude, but I believe sufficient:
- The image attachment shows the trend of H1:IMC-IM4_TRANS_NSUM_OUT16, which reports "the power going in to PRM" in Watts (note, that this is different from H1:IMC-PWR_IN_OUT16 which reports "the power going in to the IMC," which is what was discussed in the ticket, when referencing the drop "from 38W to 36W"). This trend shows that, for the week prior and the week after Dec 03 to Dec 10, the power going in to the PRM is *roughly* 33.6 W. During the Dec 03 to Dec 10 week, the power is *roughly* 32.35.

The optical gain of the IFO depends on the square root of the power incident on the beam splitter. If we assume that the power-recycling gain remained constant during this time period, then we can say that the optical gain should have changed by
opticalGain
     change = (opticalGain_nominal - opticalGain_Dec03) / opticalGain_nominal
            = [sqrt(P_BS_nom) - sqrt(P_BS_Dec03)] / sqrt(P_BS_nom)
            = [sqrt(PRG * P_PRM_nom) - sqrt(PRG * P_PRM_Dec03)] / sqrt(PRG * P_PRM_nom)
                 ( assume PRG doesn't change, and thus sqrt(PRG * P_PRM) = sqrt(PRG) * sqrt(P_PRM) and cancels from numerator & denominator)
            = [sqrt(P_PRM_nom) - sqrt(P_PRM_Dec03)] / sqrt(P_PRM_nom)
            = [sqrt(33.6) - sqrt(32.35)] / sqrt(33.6) 
     change = 0.0188
where P_BS is the power on the beam splitter, P_PRM is the power in to the PRM, and PRG is the power recycling gain.

- The trend of \kappa_C shows a change at the same time period (see either page 1 or the span across pages 6 & 7 of the pdf attachment) from 0.995 to 1.01, or
\kappa_C
     change = (\kappa_C_nom - \kappa_C_Dec03) / \kappa_C_nom
            = (0.995 - 1.01) / (0.995)
     change = 0.0151

Given that the noise of \kappa_C is on the order of 0.007, and the PRM input power is wobbling around by 0.2W at either the before or after value, I'm happy with the level of agreement between these two numbers, and consider this issue closed.