Oli, Edgard, Jeff, Brian
This post is meant to give a clearer explanation of the work in [LHO:84296] and its many comments. We are trying to find a new set of gains for the M1_OSEMINF filter banks for SR3 to calibrate the OSEMs to be in agreement with the HAM5 GS13s.
To do so, we drive the HAM5 ISI in { X , Y , Z } and record the response of the relevant OSEMs between 5 to 15 Hz. At these frequencies, the M1 stage of the suspension should start to become inertial, and the M1_DAMP / SUSPOINT response of each individual OSEM should asymptote to -1, because the OSEMs would be measuring their support point on the cage [barring internal dynamics of the cage itself].
To increase the accuracy of the calibration, we use the MATLAB model of the HLTS and use the full extent of the 5-15 Hz data instead of only the asymptotic behavior.
I post here the code used to generate these results. The code requires the new Common/MatlabTools/ExportedModels folder from the SusSVN [LHO:84458]. The detailed results of the calibrations mentioned is shown below.
_____
- We took a series of transfer functions driving HAM5_ISO_{ X , Y , Z } to measure SR3_M1_DAMP_{L,T,V,R,P,Y}. The templates for this measurement are saved in the sus SVN under [HLTS/SR3/H1/Common/Data/]
- We can see a lot of cross-coupling in some of these transfer functions. For example, the first attachment shows SUSPOINT V to M1 R cross-coupling that ends up being consistent with sensor miscalibration. This miscalibration is also seen in the large V-to-R and R-to-V couplings in the M1 to M1 transfer functions that Oli took in [LHO:83939].
- From the measurements, we reconstruct the following transfer functions:
- SUSPOINT V to SR3_M1 T1, T2, and T3. Which are the OSEMs related to the V, R, and P degrees of freedom of SR3.
- SUSPOINT L to SR3_M1 LF, and RT. Related to the L and Y degrees of freedom.
- SUSPOINT T to SR3_M1 SD. Related to the T degree of freedom
- These transfer functions are shown in the second attachment. Two interesting things to note:
- None of them asymptote to -1. In amplitude, they hover around 0.75 or so.
- The Vertical sensors are backwards. This is not important for what we're doing, we will calibrate by using the absolute value.
- We can get the scaling factors needed to make them asymptote to one by fitting the individual OSEM transfer functions to the model. We fit by using linear regression between 5 and 15 Hz. The scaling factors calculated from the figures in the second attachment are:
- T1 needs to be multiplied by: 2.4539
- T2 needs to be multiplied by: 1.4819
- T3 needs to be multiplied by: 1.4125
- LF needs to be multiplied by: 1.3206
- RT needs to be multiplied by: 1.3705
- SD needs to be multiplied by: 1.3806
- Therefore, after we multiply the OSEMINF gains that were in place at the time of measurement by these scaling factors, the new OSEMINF gains should be:
- T1: 3.627
- T2: 1.396
- T3: 1.345
- LF: 1.719
- RT: 1.490
- SD:1.781
- The T1 gain is egregiously high, which makes all of us a bit nervous. Oli has been in the process of checking if these numbers actually make sense. These results will be posted in another log... This one is already too long.
Adding hera a few bits of information that were missing from the original post, together with an updated version of the script.
_______________________________________________________________________
To compensate for the OSEM gain changes, we estimate that the H1:SUS-SR3_M1_DAMP loops must be changed by a factors of:
L gain = 0.743 * (old L gain)
T gain = 0.724 * (old T gain)
V gain = 0.549 * (old V gain)
R gain = 0.549 * (old R gain)
P gain = 0.691 * (old P gain)
Y gain = 0.743 * (old Y gain)
The calibration will change the apparent alignment of the suspension as seen by the at the M1 OSEMs
NOTE: The actual alignment of the suspension will NOT change as a result of the calibration process
The changes are computed as (osem2eul) * gain * inv(osem2eul).
Using the alignments from 2025-05-07_0000 (UTC) as a reference, the new apparent alingments are:
DOF Previous value New value Apparent change
---------------------------------------------------------------------------------
L -4.3 um -2.6 um +1.7 um
T -19.7 um -14.2 um +5.4 um
V -24.0 um -8.4 um +15.6 um
R -490.6 urad -219.3 urad +271.4 urad
P -300.2 urad -203.8 urad +96.5 urad
Y -569.3 urad -422.5 urad +146.8 urad
Here I post the coherence and transfer functions between the excitations of HAM5 in X, Y, and Z to the SUSPOINT degrees of freedom.
The band from 5-10 Hz seems to be low enough amplitude that I think we can claim that the drives are clean enough in the SUSPOINT basis to perform the calibration.
Note that the high coherence between L/T and HAM5_ISO X/Y is expected, since the SR3 euler basis does not perfectly align with the cartesian basis of HAM5.
J. Kissel (for O. Patane and E. Bonilla)
Also during this barrage of measurements, Oli and Edgard gathered Open Loop Gain (IN1/IN2), Loop Suppression (IN2/EXC), and Closed Loop Gain (IN1/EXC) tfs under the presence of DAMP_EXC.
Within the templates mentioned below, there're two data sets.
The "New OSEMINF gains" data is with with the above mentioned
H1:SUS-SR3_M1_OSEMINF_T1_GAIN 3.627
H1:SUS-SR3_M1_OSEMINF_T2_GAIN 1.396
H1:SUS-SR3_M1_OSEMINF_T3_GAIN 1.345
H1:SUS-SR3_M1_OSEMINF_LF_GAIN 1.719
H1:SUS-SR3_M1_OSEMINF_RT_GAIN 1.490
H1:SUS-SR3_M1_OSEMINF_SD_GAIN 1.781
The "OG OSEMINF gains" data is with the OSEMINF gains that have been present throughout O4 (as they were reverted post-measurement)
H1:SUS-SR3_M1_OSEMINF_T1_GAIN 1.478
H1:SUS-SR3_M1_OSEMINF_T2_GAIN 0.942
H1:SUS-SR3_M1_OSEMINF_T3_GAIN 0.952
H1:SUS-SR3_M1_OSEMINF_LF_GAIN 1.302
H1:SUS-SR3_M1_OSEMINF_RT_GAIN 1.087
H1:SUS-SR3_M1_OSEMINF_SD_GAIN 1.29
The raw .xmls for both these data is
/ligo/svncommon/SusSVN/sus/trunk/HLTS/H1/SR3/SAGM1/Data/
2025-05-21_1800_H1SUSSR3_M1_WhiteNoise_L_0p02to50Hz_OpenLoopGainTF.xml
2025-05-21_1800_H1SUSSR3_M1_WhiteNoise_P_0p02to50Hz_OpenLoopGainTF.xml
2025-05-21_1800_H1SUSSR3_M1_WhiteNoise_R_0p02to50Hz_OpenLoopGainTF.xml
2025-05-21_1800_H1SUSSR3_M1_WhiteNoise_T_0p02to50Hz_OpenLoopGainTF.xml
2025-05-21_1800_H1SUSSR3_M1_WhiteNoise_V_0p02to50Hz_OpenLoopGainTF.xml
2025-05-21_1800_H1SUSSR3_M1_WhiteNoise_Y_0p02to50Hz_OpenLoopGainTF.xml
The open loop gain transfer functions (IN1/IN2) have already been exported
/ligo/svncommon/SusSVN/sus/trunk/HLTS/H1/SR3/SAGM1/Data/
2025-05-21_1800_H1SUSSR3_M1_WhiteNoise_?_0p02to50Hz_OpenLoopGainTF_tf.txt << exports of "OG OSEMINF gains" data
2025-05-21_2000_H1SUSSR3_M1_WhiteNoise_L_0p02to50Hz_OpenLoopGainTF_tf.txt << exports of "New OSEMINF gains" data
Also, I've exported the loop suppression of the "OG OSEMINF gains" data as
/ligo/svncommon/SusSVN/sus/trunk/HLTS/H1/SR3/SAGM1/Data/
2025-05-21_1800_H1SUSSR3_M1_WhiteNoise_?_0p02to50Hz_OLGTF_LoopSuppression_tf.txt << exports of "OG OSEMINF gains" data