We have in operation, since the end of August 2018, a sensor that reads the voltage drop between two distant points connected to different electric grounds. This sensor is simply a differential amplifier (Stanford SR-560 used with x10 gain) sent to a 20kHz sampled ADC.

On December 4th, starting from 10:00 UTC, this sensor has been moved and is now connected between the West Input Tower metallic body and the West Input Suspension Rack metallic body (namely Sc-WI crate).

Note: the West Input Suspension rack (as well as all the racks on the CEB platform) is connected to the safety earth (PE) using the standard yellow/green cable; the metallic vacuum chamber of the tower is connected to ground as well, although with another path.     It seems that there is no direct connection between towers and racks (to be confirmed).

Having a look at this new signal, there is a clear correlation between some spikes and the META_ITF_LOCK_Index when this last has the value of 40, that is "LOCKING_ARMS" (see plot 1 for a general view, plots 2, 3 and 4 for zoomed events)

This is a bit surprising to me, because the current sent to the coils is not negligible (hundreds of mA), but "simmetrical" (differential driving), so no current is drawn from ground.    We could argue that, having the Coil Driver boards a simmetrical power supply (+/- V) that is derived from a single supply rail (+12 V in common for the whole crate), this could explain the path and the correlation.

Nevertheless having a path from the ground differential potential and the coils acting on the mirror could also be a path for a "common mode noise" (50Hz ?) acting on a charged mirror.
Indeed the 50Hz line (and harmonics) read by this electric sensor is low (mV pp) but of the same order of the spikes seen by the sensor when the ITF is in Lock_Index 40 (locking arms).

It is worth noting that in the long term (days) there are variations in this electrical sensor that should be analyzed with some care (see plot 5).

Note: the sensor use a x10 amplification factor, but for the time being the value has not yet been scaled in the acquisition system; the Time plots here shown has been manually rescaled, so values are correct.