Looking back at the magnetic injections done around SDB1 in October 2019.
Figure 1, shows in blue a calm time, in purple with the magnetic broad band injections next to the SDB1 quadrant and LVDT electronics and in brown next to the south flange of the tower.
The purple curve shows that the magnetic noise injection in the electronics rack increases the noise in the quadrant error signal by 2 orders of magnitude, and in the feed back signal by 1 order of magnitude. But this is still 2 orders of magnitude smaller than the DAC noise of these actuators (see figure 2 taken from VIR-0369A-20). So the coupling path is not the feedback being spoiled.
Figure 3 (from VIR-0369A-20) shows that there are mechanical resonances at 159Hz and 167Hz that can be excited when shaking the bench with the local controls. So the suggestion given in VIR-1029A-19 that magnetic fields are shaking the SDB1 bench by acting directcly on the bench magnets seems plausible. Although it would be important to have a noise projection from the ambient magnetic noise, to see if the amplitude of the noise projection is sufficiently close to the noise curve. The 167Hz peak is actually visible in the noise curve, which cannot be explained by the local control DAC noise, but maybe could be explained by the ambient magnetic noise.
A mitigation could be to replace the magnets on SDB1, and reallocated some of the correction signal to the magnets on the SDB1 marionette. However, first the transfer function between the marionettes actuator and the bench needs to be measured, to see if there is any filtering provided in the vertical degree of freedom by the 3 wire connecting SDB1 to the marionette, which do not provide any vertical angular isolation at the pendulum mode, but might provide some isolation at higher frequencies. This measurement could be done as soon as the SDB1 electronic racks are repositioned and re-connected.