Reports until 00:08, Friday 07 April 2023
H1 ISC (CAL)
elenna.capote@LIGO.ORG - posted 00:08, Friday 07 April 2023 - last comment - 12:19, Friday 07 April 2023(68484)
IFO Stable at 70W (410kW) for 2 hrs, still counting

[Evan, Kevin, Dan, Elenna]

We have been, for a very long time, operating with large offsets in the REFL WFS 45 yaw signals. This stayed the same even when we moved CHARD Y sensing to WFS 45 yaw and PRC2 off POPX onto REFL. The recent locklosses we experienced when powering up beyond 60W have all been characterized by a rapid loss of alignment in yaw and a sharp increase in the REFL signal. Moving PRC2 to POP and changing DC loop gains has not helped. We concluded the issue was that some DC coupled loop could not hold onto the yaw alignment, and that the source of the problem must be related to this massive 45 yaw misalignment.

One interesting characteristic: REFL WFS pitch signals do not see any misalignment. The pitch loops are sensed as follows:

I have highlighted this last bullet point because this was the large distinction. INP1 yaw uses a combination of REFL 9 signals (+A and -B). This is weird! There is also no particular reason why this should be, as INP1 yaw shows up fairly equivalently in magnitude in all four REFL WFS signals. My crazy idea was this: we should emulate the pitch sensing in the yaw, and see if we still have alignment problems.

At 2W, I tested a +45 A, -45 B combo for INP1 (1.6 and -1.6 respectively). This worked in both magnitude and phase (I tested with the usual 8 Hz injection). I returned PRC2 P to just 9 sensing; this was sensing that had been used before and tested extensively at high power. We did not change CHARD Y, which is sensed on negative 9 with positive 45, all four sensors.

We powered up to 60W successfully with this combination. After moving to lownoise ASC (RPC now off), we powered up further to 70W (no instabilities at all, evidence we can abandon RPC for power up). The REFL WFS 45 signals stayed centered around zero, and the REFL 9 signals drifted to small offsets, but still crossed zero. Even more importantly: the CHARD Y optical gain stayed constant over the first hour of lock. Long readers of the ASC alog will know this has been an issue for this particular loop stability.

This sensing has definitely brought us to new IFO alignment, but I am going to claim that this IFO alignment is better. Whether or not we choose to operate at higher power, staying with this new sensing and alignment is a must. The REFL 9 signals have drifted in a small amount away from zero. We can try to help this alignment shift further by changing the REFL 9 combination used by CHARD Y to the difference of 9s, so we can control that degree of freedom.

The first screenshot is the REFL WFS signals over this lock at 70W. I put the +/- combo of REFL 9 signals into DC7, showing that if we try to control this to zero, we will likely improve the REFL 9 drift.

Obviously, this alignment change will change the calibration. Once we determined we were stable, we took several calibration measurements to check:

All these calibration traces are saved in the CalSVN.

The IFO is very stable so far at 70W. We chose to stop here as a proof of concept. We would like to see how long we can hold this particular lock. We think that the alignment change is very beneficial and might solve many of our power up problems. If this is stable, we can continue to push the power up and see if 75 or 80 W is possible.

Parameters at about 2 hrs 70W lock:

This is exciting!

Images attached to this report
Comments related to this report
elenna.capote@LIGO.ORG - 00:13, Friday 07 April 2023 (68485)

I have added the new input matrix values to the guardian. I have also, in potentially brash confidence, changed the NLN power to 70W in lscparams. This can easily be changed back any time.

jenne.driggers@LIGO.ORG - 11:18, Friday 07 April 2023 (68494)

We're still locked.  At the ~12 hour mark, Dan has begun powering up further.

craig.cahillane@LIGO.ORG - 12:19, Friday 07 April 2023 (68497)
Arm power from the PRG indicates 417 kW,
whereas arm power from the REFL light levels indicate 437 kW.
CARM pole estimate is old and not to be trusted at this level of power.

REFL estimates might be artificially high from alignment offsets, but the REFL values settled to a reasonable level (8.2 mW on LSC REFL A) during this lock, unlike the previous 80 W locks.
Likely due to the INP1 Y sensing matrix fix avoiding a slow walkoff of alignment.
Non-image files attached to this comment