Sheila, Evan

We looked again at the situation with the 45 MHz REFL WFSs, which are used as sensors for the PR3 ASC loop. In the end, we didn't implement any changes to the sensors or the associated loops.

We were motivated by the following:

1. The current demod phases for these WFSs simply do not make sense; we expect that they should be similar (since the analog delays should be fairly well matched), but instead they are scattered by 100+ degrees.
2. The cHard and PR3 loops are strongly cross-coupled. We would like to be able to turn up the gain of cHard without impressing extra motion from other optics into DARM. [20523, 20497]
3. Any perturbation in the PR3 seems to couple into the SRC loops. We already know that the SRM loop is quite delicate, and we would like to get rid of this coupling if possible.

We tried the following:

• We changed the demod phases of the 45 MHz REFL WFSs, and then closed the PR3 loops.
  • We chose phases so that cHard pitch appears entirely in the Q phase (thereby leaving I dominated mostly by PR3 motion, hopefully). This was done previously, but loop retuning was not investigated in detail. Also, the phases of A3, A4, B3, and B4 seem to need an extra 180° phase shift in order to produce correctly phased signals (see screenshot).
  • We then found that only REFL45B gave a sensible error signal in response to PR3 motion (both pitch and yaw), so we closed the loops around B only. Additionally, the required digital gain went from 2 ct/ct (with the old phasing) to 40 ct/ct, perhaps indicating that the "old" PR3 configuration is actually dominated by cHard motion.
• This seemed to produce a constellation of bad things:
  • The sideband buildups (POP18 and AS90) were slightly suboptimal (i.e., fuzzy and not maximized).
  • The soft loops reduced the recycling gain (to 36 W/W or so) when they came on.
  • Switching the SRM yaw loop input matrix elements at 3 W seemed to make POP90 worse.
• We tried using new trans QPD offsets for the soft loops, but this led to the familiar angular instability when going to 20+ W.
• We also tried to find a new SRM yaw matrix element, but this didn't work either.
• We also turned off the dc oplev coupling on SR3, since we didn't want the optic position to be dragged away by the ground.
• In the end, we simply reverted all the changes we made.

After that, we couldn't get to full power because of an oscillation that showed up in dHard pitch when going to 20+ W. (We've seen this before, and it seems to be distinct from the angular instability mentioned above.) To get around this, we had to slightly beef up the resonant gain that gets turned on in dHard pitch when the power is increased.

Also, there were some small maintenance tasks that we did:

• Improved the filter hygiene of the DC centering loops (there were some integrators that didn't have ac gain of 1, some low-pass filters that didn't have dc gain of 1, some filters with ambiguous names, etc.)
• Added an optional ISC_LOCK guardian state called REDUCE_MODULATION_DEPTH, which implements the 3 dB modulation depth reduction that Stefan and I have carried out before.