BS OL Diagonalization
On 11/30 we collected data for calibration and diagonalization of the BS tilt OL and have since analyzed it. See the attached Excel spreadsheet for the raw data and full analysis.
- We scanned the QPD horizontally with the micrometer and noted the readings in nominal yaw (K1:VIS_BS_TM_OPLEV_TILT_YAW_OUTMON). There was a fair amount of physical swing (mostly physical yaw) showing up in both nominal pitch and nominal yaw (due to poor diagonalization), so we used StripTool to look at a representative length of data and eyeball the average. There was a 100x gain (Simon's preliminary guesstimate of the µrad per count at K1:VIS_BS_TM_OPLEV_TILT_YAW_INMON) just before the point we read the signal, so we divided it out in the data analysis and made the _INMON version the reference.
- We scanned the QPD vertically with the micrometer and noted the readings in nominal pitch (K1:VIS_BS_TM_OPLEV_TILT_PIT_OUTMON).
- The calibration factors were -1.53 cts/mm at K1:VIS_BS_TM_OPLEV_TILT_YAW_INMON and +1.45 cts/mm at K1:VIS_BS_TM_OPLEV_TILT_PIT_INMON. These are very close, so apparently the beam is quite round at the QPD.
- We applied yaw actuation at the IM OSEMs (K1:VIS_BS_IM_TEST_Y_OFFSET) and noted the readings in nominal yaw and pitch at the OL, and at the IM OSEMs (K1:VIS_BS_IM_DAMP_Y_INMON and K1:VIS_BS_IM_DAMP_P_INMON). We did the same with pitch actuation (K1:VIS_BS_IM_TEST_P_OFFSET).
- We plotted the responses at the IM as a function of actuation at the IM and found pitch and yaw were already pretty much perfectly independent, so we made no attempt to do any diagonalization of the OSEMs.
- We did similar plots for the nominal yaw and pitch at the TILT OL and found that there was very large cross-coupling. This is due to the odd angle of the first steering mirror, which bends the light from the +X,-Y direction at about 38° to the horizontal, to approximately in +X and level with the OL breadboard. Using the Y and P coefficients of the coupling matrix separately gave estimates of 61.8° and 62.8° for the angle, so the coupling is consistent with a pure rotation.
We derived a net calibration/diagonalization matrix taking into account:
- The measured calibrations from QPD position to nominal yaw and pitch).
- The measured beam path from the BS to the QPD. See klog 3616 for most of the segments of the path. We had to make a new measurement of the last segment of the path to the TILT (as opposed to LEN) QPD: 124 mm.
- The angle of the beam at the BS (38° as estimated from CAD) which reduces the effective length of the OL in yaw by cos(theta).
- The rotation angle introduced by the steering mirror.
- We checked that the new matrix did in fact accurately diagonalize the original data.
- We divided the derived matrix by 100, and then entered it in the K1:VIS-BS_TM_OPLEV_OL2EUL screen, leaving Simon's gains of 100 in place. This makes the raw and diagonalized plots more comparable.
- However we did not get good agreement with the amount of yaw and pitch measured by the IM OSEMs. The diagonalized OL yaw was 2.17 times the IM yaw and the diagonalized OL pitch was 14.3 times the IM pitch.
- We traced some of this to an error in the OSEM2EUL matrices for the IM: the P and Y entries were too small by a factor of 2 due to confusion about specifying the position of the OSEMs relative to the center vs. relative to the opposing OSEM. We give the correct version of the calculation in JGW-T1707205-v4, and we entered the correct matrices around 4:30 pm on Monday 12/11. That brings the OL and IM into reasonable agreement, but leaves a factor around 7 to be accounted for in pitch. A modest amount more pitch response at the IM than at the OL could be explained by the BF and IRM tilting opposite to the IM and BS, but we have the reverse of that, so there may be a dumb mistake still lurking.