Keita, Jeff K, Sheila, TJ

Some of the locklosses that Betsy points out here: 51706  (especially the ones where there is a glitch in DARM that causes the ADS signals to move the IFO) look similar to some of the locklosses over the weekend durring engage soft loops and increase power (eg  51677) In these locklosses the SRC1+SRC2 error signals aren't held to 0, and neither is INP1, and the PRC2 offloading which is slow is very behind.  Yesterday I moved top mass integrators in SRM and SR2 so that we would have the option of adding gain upstream of the intergrators, which was implemented when Ed reloaded the ISC_DRMI guardian last night. 

This morning we increased the gain of the top mass offloading for SR2 +SRM by 10 dB.  (I caused one lockloss with a mistake while doing this.)  We also moved the integrator for PR2, and made the appropriate changes in ALIGN_IFO so that we could add 20dB more offloading gain to the PRC2 P+Y loops.  These changes are all in the ISC_DRMI guardian, and we have locked twice with them implemented as is.  This seems to do a better job of offloading, and has spend up the DRMI ASC.  It looks like we would benefit from increasing the SRC offloading gains by another 10dB, and from increasing the overall loop gain of INP1. 

We lost lock once today because of a mistake I made, a few other times with symptoms simliar to the INCREASE power and soft loop locklosses over th weekend (before implementing these changes and with them partially implemented), and our last lockloss happened while the soft loops were steering the beam around the point absorpber, as we passed through a point where all the build ups get low.  We decided to try in the next lock not to steer the beam around, but manually let the ADS converge with the spots centered, then moved directly to the final position.  This worked and didn't bring the buildups as low as they were in the previous lock (see 1st screenshot, time cursors are at times when the spot position was moved, the first one was going around the test mass, the second was going directly to the final position).  We've implemented this in the guardian, which should help to speed up the long wait for ADS convergence, and hopefully avoid some locklosses while the spots are moving around to places where build ups are bad.