One of the more common manual interventions that operators have been doing is that when PRMI won't lock they use check_MICH_FRINGES to see the dark michelson and manually adjust the beam splitter.  (This seems to account for most of the DRMI interventions lately).  Jenne and I had talked about automating this, today I took some time during maintence to implement it.  

For operators:

• The guardian will (as before) attempt to lock DRMI for 5 minutes, if it hasn't locked after 5 minutes it will try to lock PRMI.  If that hasn't locked after 5 minutes, It will change to CHECK_MICH_FRINGES.
• CHECK_MICH_FRINGES is now using the ALIGN_IFO node instead of directly locking things itself, it will increase the power into the IMC to 10W, then lock the bright Michelson, and use the dither alignment to lock it.  This is the same guardian code which is used in initial alignment.  
• There is a new state called MICH_OFFLOADED in ISC_LOCK, which will happen right after CHECK_MICH_FRINGES, which turns off the AS centering loops, turns off the MICH loop, and turns the laser power back to 2W for PRMI and DRMI acquisition. 
• The gaurdian should then move on to PRMI and then DRMI as normal. 

Details about the guardian change:

• I wanted to avoid copying and pasting from the inital alignment guardians into ISC_LOCK, so I added an additional path in ALIGN_IFO which is nearly a copy of the MICH_BRIGHT_ALIGN path that we currently use.  To do this I made most of the states in this path into generators, with two instances, one with the old state names and numbers which we still use for inital alignment.  The intial alignment has been tested and is working as it used to.
• I added a state in the inital alignment mich branch called set_sus_for_MICH, this used to be done in PREP_FOR_MICH but I needed to separate it out so that is can be different when we are locking FPMI than when we are locking just the michelson.  
• I made two changes to the DOWN state in ALIGN_IFO, to try to preserve the ALS lock when we run this down.  We set ezca['IMC-MCL_GAIN'] = 1, I added an if statement so that this is not done if ALS_COMM is in the IR_FOUND state.  We also set the MICH gain to 0 with 0 ramp time, I changed this to a 3 second ramp.  
• We have been able to keep ALS locked while this state runs down several times now.  
• We might loose the ALS lock if the michelson looses lock here, but it has stayed locked several times as I tried this. 
• We did this on the way to relocking today, and saw that the resulting PRMI alignment wasn't very good, but it was good enough for PRMI to lock and the ASC to work work.  We have been suspicous that we aren't getting great michelson alignments from initial alignment, see Jenne's alog and my comment from earlier today: 54906
• These changes are revision 20978

If for some reason this change is a problem, it can be avoided by setting the timer of how long to wait in ACQUIRE_PRMI before trying CHECK_MICH to be very long.  This timer is on line 966 of ISC_LOCK, and is currently 300 seconds, but you could set it to 3000 and reload ISC_LOCK if there is a problem with these states.