The EX susrack magnetometer (all directions) is unsafe, meaning it picks up signal directly from DARM. Presumably it's picking up the electrical signal from the feedback of DARM to EX.

This was determined by Sheila's suggestion of looking at a calibration sweep. We're also analysing the detchar hardware injections, which should show the same thing. We see the calibration sweep appear in the magnetometers below about 100 Hz. The Y direction also sees something at double the frequency. I'm only attaching that spectrogram.

Edited to add: The calibration sweep is going through the Y end, so it's not just picking up that.