As a follow up to my other post about ETMX glitches, I looked at using a strategy used in DRMI locking to try to help the IFO ride through the high frequency glitches that have been causing locklosses. On the BS suspension the ISC path includes filters that whiten the high frequency ISC signal, a limit is applied to that whitened signal then it's dewhitened. Think the filters used for this are a zpk(1,200,1) and it's inverse, with a limit of 50000 in the ISC input to the BS.

I attempted to look at the impact of doing that on a couple glitches leading up to a lockloss. I got data for the ESD drive from one of the lock losses and used lsim in matlab to model the change in the ESD timeseries. The attached image shows the timeseries for each step of whitening, limit and dewhiten compared to the original glitch. It's not a proper model of the DARM loop, Sheila and I might talk about doing that, I just wanted to see if doing the simplest estimate would blow up before digging deeper into it. 

Thick blue is the original ESD drive from one of the quadrants, Red is the whitened siganl, yellow is the whitened signal limited to below the saturation level for the ESD, and the thin purple line is the dewhitened, final timeseries. The thin line doesn't show crazy behavior and stays below the saturation threshold (2^19*275, which comes from some adjustments made to accomodate the new dac on ETMX). 

I expect if this worked, it would inject some higher frequency junk into this segment, but the glitches already do that. The hope is this would reduce the drive from these high frequency saturations and let the IFO ride them out.