During this week, the locking of the H1 IMC has been taken care of by the Guardian-based autolocker developed at LLO.
At the moment, it's capable of the following functions:
- Lock the IMC (choose state LOCKED from the IMC screen)
- Unlock the IMC (any other state except MANUAL -- they are all equivalent)
- Do nothing (MANUAL state)
Although 'Guardian' may seem like an overly grandiose term for this level of functionality, it's an improvement over what we had in place before. It runs in a centralized location (h1script0), provides a graphical interface (MEDM), and is pretty adept at locking the IMC.
The most serious limitatation is that the IMC's Guardian is intended to be a "manager", but on H1 it currently has no subordinates. No subsystem guardians (SUS, ISI, et al) exist here, yet. So state verification is not really being done (other than the check of the IMC transmission to tell if the cavity is locked), and switches/gains/etc in the subsystems have to be manipulated directly in the up/down scripts.
What was done
- Move the old IMC Guardian scripts into an archive directory (.../ioo/h1/scripts/imc/guardian_2013-06-20)
- svn up the L1 scripts directory (.../ioo/l1/scripts/imc), copy over the Guardian related ones
- Hack on the scripts to make them ignore the SUS/ISI guardians that aren't running here yet
- Hack on the scripts to accommodate differences in the installed feedback filters (location and quantity of boosts, offloading setup, etc)
- Start the IMC Guardian (commands: ssh h1script0; screen -RAad imcguardian; cd /opt/rtcds/userapps/release/ioo/h1/scripts/imc; ./runIMCGuardian)
- Add a feature to toggle the feedback and kick MC2 while waiting to acquire, to prevent getting stuck on a higher order mode or between fringes.
Performance assessment
The IMC was locked 93% of the time that the refcav was locked during this week. There were 273 lock loss events, and the median time to reacquire lock was six seconds (excluding time spent waiting for the refcav). There were 7 events where it failed to relock within three minutes. I have not checked out all of these, but several could be explained by (1) waiting for the ALS down script (which restores the IMC length feedback that is disabled by the CARM handoff); or (2) slow controls reboots that resulted in a misconfiguration of the MC servo board.
