Here are the estimates of the overall systematic error in the response function due to systematic errors in 1) PUM actuator and 2) Sensing function. Firstly I copied the status summary from Jeff's alog 47941

1) PUM actuator systematic error

- The PUM actuator systematic error remains below 20 Hz. We've identified that it's parasitic length drive not accounted for through L2A / A2L. This is prominent because the IFO spot positions on ETMX are -18 +/- 0.2 mm in pitch (-vertical) and 18 mm +/- 0.2 in yaw (+transverse) off from the center. (note for PCAL team -- ETMY spot positions: -17 +/- 0.2 mm (-vertical) and 7.5 +/- 0.2 mm (+transverse))

The PUM results are briefly discussed in 47926. See the first pdf attachment. Page 1 and 2 show the measurements taken from drivealign bank and test bank, respectively, against the model. The fitting through the test bank is better, but the measurements from the drivealign bank are the reality. Page 3 shows the comparison of the contributions from the residuals to the overall response function, which is produced in the following way:
1) Create two separate reference models for drivealign and test scenarios, based on the Mar 16 model, by writing back the MCMC PUM MAP values.
2) Compute the error residuals for two scenarios separately.
3) Plot Response_drivealign_with_error/Response_drivealign_ref, and Response_test_with_error/Response_test_ref.

To improve this, we need to either make the physical actuator better by removing the existing features in the measurements, or update the model to compensate the features. Since it only impacts freq lower than ~20Hz, we decide to work on this improvement at a later time.

Measurements were taken on Mar 26 (in /ligo/svncommon/CalSVN/aligocalibration/trunk/Runs/O3/H1/Measurements/FullIFOActuationTFs)

Script used to generate the plots: /ligo/svncommon/CalSVN/aligocalibration/trunk/Runs/O3/H1/Scripts/FullIFOActuationTFs/compare_drivealign_vs_test.py

2) Sensing systematic error

- We have a prominent detuning below 20 Hz, and it's pro-spring. The fit to and implementation of compensation for a pro-spring is still in it's infancy, so it is under-reporting the frequency -- it's measured to be about 6.5 Hz, but the fit claims 3.7 Hz. The systematic error induced by this flaw is small above 20 Hz. 

See the second pdf attachment. The first plot shows the measurement and the reference model. The second plot shows the residual contribution in the overall response function. Generally it is okay above 20Hz. One data point at about 150Hz produces a larger error. We probably will add more points around that freq for next measurement. Since the main contribution is also below ~20Hz, it does not prevent us from moving forward to update CAL-CS, etc at the moment.

Sensing measurements were taken on Mar 27 (in /ligo/svncommon/CalSVN/aligocalibration/trunk/Runs/O3/H1/Measurements/FullIFOSensingTFs)

Script used to generate the plots: /ligo/svncommon/CalSVN/aligocalibration/trunk/Runs/O3/H1/Scripts/FullIFOSensingTFs/process_sensingmeas_20190327_contribution.py

Note: To generate these contribution plots, I need the intermediate data (measurements and the TFs computed from ref models) from processing actuation and sensing. Hence I have modified actuation.py and sensing.py, adding the intermediate data into the results structure returned from these functions.