It seems that the PRM position for DRMI ASC needs to be reset to better match initial alignment. Right now, the initial alignment sets the PRM in a place that is pretty close to where it should be for full IFO (this is about -0.6 on the POP_A QPD pit_inmon signals, which looks like -0.8 at the PRC1 error signal since the POP_A QPD has a digital offset of -0.2). But, the DRMI ASC wants to put the PRM to a place where the PRC1 error signal is zero. This makes DRMI ASC a little hairy, since having the PRM move so much so fast is difficult for other dofs to follow (last night we decreased the PRC1 P gain by a factor of 10 by hand, today I've put into the guardian to reduce the gain by only a factor of 5). This then makes the full IFO ASC (including soft ADS loops) slow, because they have to move the PRM back to where the initial alignment wanted it.
When we had the DRMI ASC on, I moved the PRC1 offset to see if the buildups were okay if the PRM were closer to its initial alignment location, but the DRMI buildups didn't seem so happy. I didn't wait for full ASC convergence, and I didn't go all the way to where the initial alignment location for PRM was, so we probably need to revisit this next time we are at DRMI ASC (especially now that I see that the full IFO PRM place is quite close to where initial alignment leaves it). For now, so that we survive the DRMI ASC, I've lowered the PRC1 pit gain in the guardian to 2, from its typical value of 10. We'll probably want to speed it back up once we reset the setpoint.
Separately, for PRX initial alignment: Last night Corey had to increase the gain of PRCL by hand, but I think that is just because I had forgotten to hit 'load' on the ALIGN_IFO guardian after last night (alog 51044) re-reverting the gain to the high value. After loading the guardian today, PRX locked. However, the ALIGN_IFO guardian didn't think it was locked, since the PRM position (which had been set most recently by offloading DRMI ASC during the previous lock) was so far wrong. The guardian wanted POP_A_LF to be greater than 1.0, but it was hovering around 0.85. I lowered the 'locked' threshold for PRX in lscparams.py to be 0.5, and then the ALIGN_IFO guardian was able to align PRX as normal. Probably once we fix the DRMI PRM setpoint to match the initial alignment and full IFO positions, POP_A_LF will start out much closer to its usual locked and aligned values, so we can put the locked threshold back to its old value of 1.
I again did a teeny bit of exploring, but again it seems that if I put the POP A QPD offset to the place where the full IFO wants it, the POP18 buildup dropped a lot. So, for now I've put the setpoint 25% of the way closer to where the full IFO wants it (offset of 0.0, from old offset of -0.2, when the full IFO wants the offset to be +0.6), and the POP18 value is still pretty good. Once we reached PREP_ASC_FOR_FULL_IFO, it was also clear that the PRM was closer to the final location that the ADS system wants it, which is good.
I have accepted this new POP A QPD offset in both the safe and observe ASC sdf files.
In the end, I put the POP A QPD offset back to -0.2 as it was last night, since we've been failing at acquiring lock with other values. This acquisition we let DRMI ASC go with the old POP QPD offset, but after we offloaded the DRMI ASC Cheryl touched a few of the optics (PR2, PRM and SR2, I believe) to further improve POP18 and POP90 buildups. After she did that, the POP18 and PR gain traces looked a lot smoother than they have all day. With this alignment, we have (so far) been able to get to DC readout for the first time in several hours, and are currently increasing power.
So, I'm not sure why, but our DRMI ASC is leaving the optics in a place that isn't the most optimal alignment. When this is happening, we lose POP18 buildup as we go through the CARM offset reduction sequence. With Cheryl's nice hand tuning, we dind't really see this happen. So, we need to understand what's going on here, but for tonight we'll just let the IFO observe.