I looked at MICH_BRIGHT_ALIGN today, since it's been a problem recently for initial alignment.

It seems that even with just length locked (no ASC yet) we are occcasionally saturating the M2 stage of the suspension.  When this gets really bad (especially if the BS oplev loops start sending out very large signals) we start to constantly saturate the BS, and we are unable to lock or run the alignment loops.  If we were able to catch and hold the length lock, turning on the ADS lines at their former values would start this saturation cycle, and we lose the length lock.

I checked the MICH length OLG, and the loop looks fine. 

I was unable to measure the BS oplev loops, mostly because I wasn't patient enough to tune a template (do we have SUS oplev loop templates?) and my first few guesses of excitation amplitude were too high and caused the BS to saturate.  So, I don't know that it is an oplev loop stability problem, although that would be surprising since it's been fine for years now, and the typical oplev loop shapes are nice and stable.  Turning down the BS oplev loop gains to 100 for pitch (from  300) and 200 for yaw (from 650) seems to be a good compromise of having it on so it damps the BS when we lose lock and having it off so we don't saturate.  We ran through initial alignment with this, and it seems okay.  The DRMI DOWN state already sets these gains back to their nominal values.

Even with the oplev loops turned down a bit, the old ADS dither amplitudes were still causing the BS to saturate and we'd lose lock.  So, I've turned the ADS dither amplitudes down in the PREP_MICH_BRIGHT_ALIGN state to 1000 for pitch (from 30,000) and 300 for yaw (from 3000).  I have not yet done so, but we should probably increase the gain of the ADS loops by a factor of 10 so that these loops don't take so long to converge. 

Separate from the ASC part of MICH, if we were failing to acquire the length lock, guardian wasn't clearing the history of the BS M2 LOCK L filters, which have some high Q features.  So, the align IFO node's down state now clears that history.  This should solve the problem of continuing to saturate the BS when we're starting to try acquiring again (thus guaranteeing that we're going to fail to catch the lock). 

We've run through MICH_BRIGHT_ALIGN succesfully twice now, so it seems like it's okay, although we should increase the ADS loop gain next time we have a moment to test this.

The lowering of the BS oplev loop gains is *not* currently in the guardian, although perhaps we should make it so.