The bump in B8_DC disappeared during the big switch-off campaign, but then reappeared during the switch-on that caused the 50 Hz miracle. Since this happened during a lock, it could be seen that the bump re-appeared a few minutes before the 50 Hz was reduced, when the devices in the WEB were switched back on and left on, see fig 4. This narrowed down the possible culprits to the stepper motor drivers of the WE_TCS and SWEB. This evening, we switched the first one off again at 17:36:40 UTC and the second one around 17:38:05 UTC. As shown in fig 5, it was the one driving the SWEB motors that was the guilty one. The improvement is clearly seen in B8_DC, but nothing could be seen in DARM, since we were suffering from some other issues today. To be confirmed in the next days that the 80 Hz bump (which might have cost a Mpc on a bad day) is indeed gone in DARM.
The exact coupling mechanism is still to be understood. We do not use any signals from SWEB in science mode, so it cannot be via the photodiode itself. Electromagnetic coupling to the mirror seems a long shot, which leaves scattered light as an option. That would imply some physical movement, but nothing could be seen in any of the SWEB sensors. It would be good to do a postmortem on the involved devices after the run.
B8_DC gives us a measure of scattered light on SWEB. Using the SWEB longitudinal injection from last Friday, we can calibrate the coefficient to get from scattered light on SWEB to noise on B1.
Figure 1, shows the projection during that SWEB longitudinal injection. In green is the projection based on the SWEB motion during the injection, in red is the projection based on B8_DC power. The two match quite nicely.
Figure 2, shows what happens when this projection is done during a time when the 80Hz bump is high. And the projected noise (as measured by B8_DC) matches well in amplitude the measured noise level in DARM.
Figure 3, to obtain that kind of scattering with motion, I have added a 79Hz line into the measured SWEB motion before doing the projection. That line needs to have an amplitude of 1e-8 m to have a projection comparable to what is observed on B8. This seems to melike a large motion for a suspended bench. The local control noise floor is at 1e-9 m/rtHz, so if the bench was moving coherently at 79Hz then the motion would be visible in the local control signals, which is not the case (Figure 4). So the options are that some individual optic on the bench is vibrating with an amplitude of ~1e-8m, or maybe more likely the window or some scatter attached to the ground (inside the tube?) is having the large motion.
One thing that is not understood yet is the modulation of the amplitude of the bump, which was varying a lot during the evening of the 22th. One possibility is that there are actually two reflecting surfaces, whose separation changes by a few wavelengths due to drifts in temperature. This can cause etalon like modulation of the effective reflectivity. Another possibility might be some small drifts in alignment, which could change the amount of light scattered back into the arm.
