Summary: For whatever reason, RM motion in HAM1 is shows up more strongly in REFL WFS A RF9 than the other ASC/LSC diodes in HAM1. Also, the appearance of the motion is not consistent among the different quadrants, suggesting that quadrants 1, 2, and 3 are seeing some HAM1 motion that quadrant 4 is not, particularly between 10 to 20 Hz. This is from a lock in May 2021.
Fig 1: before talking about RF9, this shows the LF sum and pitch signals from the WFS, and a comparison with the LF signals from the LSC diodes. Nothing too surprising here; evidently WFS A and B see similar amounts of RM-induced spot motion. The fact that WFS A LF sum is higher than WFS B LF sum above 5 Hz is slightly suspicious (electronics noise?).
Fig 2: the RF9 quadrants on the WFS are summed to make 9 MHz length signals, and compared with the 9 MHz length signal from REFL B. We see that the RM motion shows up as a length signal much more strongly in WFS A than in WFS B or in LSC REFL B. Since these three signals are supposed to be out-of-loop sensors of the REFL 9 MHz field, I'd think that except for possible differences in I/Q phase rotation, they should all show the same thing. Calibration into watts is based on shot noise level at high frequency.
Figs 3 and 4: spectra of the individual RF9 quadrants from the WFS. It becomes obvious that the different quadrants of WFS A are recording different motions above 6 Hz, with quadrant 4 generally having lower noise than the other three quadrants, and with a different shape (no shelf around 18 Hz). Could it be clipping on one of the steering mirrors or beamsplitters on the WFS sled? (This seems hard to reconcile with the fact that according to the HAM1 layout, the beam for WFS A is picked off before the beam for WFS B, so I'd think any clipping on A should also appear on B.) Or is there an explanation for this that invokes the mode content of the reflected 9 MHz beam?
Figs 5 and 6: complex coherences of the individual RF9I quadrants against their respective sum signals. The signal above 6 Hz in quadrant 4 of WFS A is neither 0 nor 180 degrees relative to the sum channel. Compare to the situation below 1 Hz, where all four quadrants appear with a phase relationship (0 or 180 degrees) one would expect from some pitch-dominated motion.
Figs 7 and 8: complex coherences of the individual RF9I quadrants against HAM1 TT Z motion, essentially confirming the picture from Figs 5 and 6. The magnitude of the coherence with the WFS A segments seems to indicate that quadrants 1, 2, and 3 of WFS A are more sensitive to HAM1 motion than quadrant 4 from 10 to 20 Hz. WFS B may also have some of these same problems, though to a smaller extent; its quadrant 3 is less coherent with HAM1 motion between 10 to 20 Hz compared to the other three.
Threre is a long history with WFS_A: FRS 4526, even so this seems to effect the DC segment 4 only (coincidence?). Also, WFSA and B are swapped , see alog 45369.
I looked at the same channels during this weekend's recent lock. Generally the sum channels (LF and RF9) of WFS A look more similar to WFS B and to LSC REFL B than they did last year. In the case of the LF sum, WFS A seems to have improved (RM motion no longer shows up). In the case of the RF9 sum, WFS B and LSC B seem to have gotten worse, now matching the noise level in WFS A.
It is still true that WFS A segment 4 is less sensitive than the other three to RM motion from 10–20 Hz. This is complicated by the fact that unlike last year, the RM motion now mostly shows up in Q rather than I. As far as I can tell the digital phase rotations have not changed and they still seem to be appropriate for putting most of the sub-5-Hz RF signal into the I quadrature.
Overall this seems like something having to do with the optical field rather than an obvious problem with the WFS A chain. To pick one question of several — why is there a ~10× variation in the total RF9 sum power at 10 Hz from last year to now on these diodes, even though the dc pointing channels indicate a similar level of RM motion both times?
One more iteration of the above plots — this time from the end of O3b. Compared to the above lock, RM motion does not appear as strongly in the rf signals, even though the dc pointing signals indicate the RM-induced beam motion on the photodiodes is similar in all cases.