I have been trying to see if ETM glitches can be reduced by increasing the ~1khz roll off of the ESD, but it seems like this is no the route. I wasn't able to find a filter that would reduce the amplitude of some typical glitches and not adversely affect the phase of DARM at the UGF. The idea was to add a low pass at the output of the L3 drive, but it doesn't seem like this will work.

I got 3 seconds of ESD drive data for one of the locklosses which had several glitches shortly before the lockloss. I then looked at how the time series for those ESD glitches changed when I added various lowpasses using lsim in matlab. First attached bode plot shows 2 filters, red trace (filter2) is a way too aggressive filter, blue (filter1) is something that would probably allow DARM to stay stable but only maybe help a little.

Second plot is a timeseries showing the effects of the 2 filters on one of the glitches in the data I looked at. Yellow is the drive sent to one of the esd quadrants, red is the glitch after applying the unstable filter2 from the first plot, blue is with the less aggressive filter1. Even with the unstable filter the magnitude of the glitch isn't really reduced a ton. The phase margin at ~80hz for DARM is about 30 deg, the blue filter reduces this by about 6 degrees.

Sheila and I have talked about other approaches to make the IFO more robust to these glitches, we'll take a look at that next week.