We have transitioned CARM to the TR signals a few times.
We used to leave ALS parked 1 kHz below the arm resonance and just engage the TR CARM path to bring it in. Perhaps because ALS is now quieter this doesn't work very well, so I added a step where the COMM VCO brings the arms with 100 Hz of resonance before we attempt the transition.
First there was some confusion with the order of things in the guardian. In the end I reverted so that the guardian works the same way it did in the version in the SVN for O2. The naming of the guardian states has caused some confusion (because PREP_TR_CARM actually started the transition), so I renamed the states but preserving the old order of events. PREP_TR_CARM now adjusts the QPD offsets and brings the arms to within 100 Hz of resonance. START_TR_CARM engages the REFLBIAS path to the offset the ALS signal (this used to be called PREP_TR_CARM), and CARM_TO_TR ramps up the TR path gain and disengages ALS.
This transition has worked a few times, but is also failing often, so more work will be needed.
Attached is a screenshot and template for the TR_CARM OLG. The loop didn't have much phase margin when we used the same filters as were engaged since 36686. Engaging the VCO compensation filter gives us a better phase margin, although apparently turning it on caused some locklosses in the past. I have added it back in but it now is engaged after the transition has completed so hopefully that will not cause problems.
Other things to note from tonight:
- ALS+ DRMI has been relocking easily tonight, often within 6-7 minutes of lockloss.
- The 3F locking became unreliable when the DRMI alignment wasn't great. Since I started adjusting the alignment by hand each lock it has been doing better.
- When the 3F locking fails, DRMI jumps to PRMI. This is happening when something is going wrong, but it is different than the way that DRMI used to loose lock.
- I tried removing the 27I to MICH element from the input matrix and this made the 3F locking even less reliable when the alignment wasn't great.
- Some guardian states weren't checking for the correct locklosses (ie, states which should be checking for DRMI locklosses were not)
- At around 6 UTC, I made a bad attempt at transitioning to RF Darm, and tripped HEPI and the ISI at EX. TMS was tripped, which I didn't realize, which caused the quad to ring up very slowly. Now things are settling back down.
