We have good models on how suspended bench motion couples into h(t) through scattered light. These can be used to calibrate the photodiodes on SNEB and SWEB as sensors of scattered light.
Figure 1 shows the scattered light noise budget (before subtraction in h(t)) during a high wind time (2019-12-12 07:00 UTC). In dark blue is the projection of SWEB scattered light based on the bench motion, and in magenta based on the B8 photodiode spectrum. In both cases the same model (radiation pressure dominated) is used to compute the coupling of scattered light to h(t). The amplitude for the B8 coupling has been roughly adjusted to match the projection compute based on the bench motion. This show us that the scattered light as measured by the B8 photodiode on SWEB, is dominated by the phase modulation due to the fringe wrapped bench motion.
However, for SNEB we don't have such a match. The motion of SNEB predicts that there shouldn't be any significant scattered light. However, the projection of the B7 photodiode shows a broad contribution, that is only a factor 3 below the sensitivity curve at 30Hz.
Figure 2 shows the same noise budget during a calm time (2019-12-11 22:00 UTC). This reveals two bumps, at ~15Hz and ~35Hz in the projection based on B7. This is not new, it has already been there at the beginning of O3a.
Figure 3 shows it is a real effect. There is coherence between B7 and Hrec_hoft_raw and hrec_hoft at both times. In particular the bump at 35Hz in a calm time has a rather significant coherence with Hrec_hoft_raw (B7 is then used in noise subtraction in Hrec_hoft).
The question is where these bumps are coming from. It does not seem to be coming from the bench, broad band injection into the bench actuators do not excite these bumps, see figure 4.
I wonder if any of the past acoustic and/or tapping tests around SNEB could have excited these bumps at 15Hz and 35Hz to help locate them. It could be a problem that is easy to solve, as it is not happening for SWEB.