Stefan, Matt, Kiwamu,
The online calibration, aka CAL-CS, is more accurate now; the sign of the simulated ETMY L3 stage (CAL-CSDARM_ANALOG_ETMY) was found to be wrong and we fixed it. Fixing the error resulted in an improved noise level at around 100 Hz in CAL CS. This should not affect the GDS pipe line calibration. The attached shows a comparison of CAL-CS before the fix and an offline calibration using DARM IN1 and the full DARM model (28179). It is clear something wrong was going on in 40 - 200 Hz.
What we changed:
-
FM4 (N_per_cnts) in CAL-CS_DARM_ANALOG_ETMY_L3
- zpk([],[],-0.000000000004239369201660156,"n") -> zpk([],[],0.000000000004239369201660156,"n")
This fix gave us an actuator model which is consistent with the measurement (28179) in the sense that the relative phase between the PUM and TST have a relative phase of 180 deg at high frequencies. Also, traditionally, when ETMY had a positive bias, the gain of the L3 stage used to be set to -1 in the O1 era (see for example 25575). Therefore today's fix is consistent with the O1 era too. One thing I still don't understand is the relative calibration difference between GDS and CAL-CS (summary page). The relative magnitude should show of a factor of 2 difference or so around 100 Hz assuming the sign error was only in CAL-CS, but it does not show such a big difference. Not sure why.
The online GDS (called C00 during O1) calculation uses CAL_DELTAL_CTRL and CAL_DELTAL_RESIDUAL to produce h(t). Compared to the front-end, It applies better relative timing between the two signals and other high freqency corrections. Since CAL_DELTAL_CTRL is obtained after the application ANALOG_ETMY_L3 filter, the online GDS will also have the same problem as front-end DARM signal. Only the offline GDS (called C01, C02 during O1), uses CAL-DARM_ERR and CAL-DARM_CTRL along with actuation and sensing models to produce h(t) and hence would have been different. I am not sure whether we have produced that at this point.
Thanks, Shivaraj.
You are right. I misinterpreted the subway diagram (G1501518-v10) last night 
. I agree that C00 must have the same sign error and therefore what we saw in the summary page is correct.
The scipt, which produced the comparison plot, is saved into svn, so that one can use the code in some future when it is needed. The code lives at
/ligo/svncommon/CalSVN/aligocalibration/trunk/Runs/PreER9/H1/Scripts/ControlRoomCalib/H1CalibDoubleCheck.m