Today I verified the power recycling gain channel.  
We wanted to make sure the PRG channel was correct because the new PRG is so much higher.

The PRG is 55 at 2W.
Our old PRG calibration was correct.  I've adjusted the calib filter in H1:LSC-PR_GAIN FM10 up by 1%.

If this result holds, this means our new ITMY is much, much better.  Our O3 PRG at full lock was 43.  Our O3 PRG at 2W was 47.5 (looking at e.g. September 09 2020 22:35:00 UTC).

----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Review of the PRG check procedure:
- I went back to when Sheila and Varun locked the Y-arm by itself at GPS 1299600400: alog 58209.
I grabbed the input power, (H1:IMC-IM4_TRANS_NSUM_OUT16)
and Y-arm transmission at this time, (H1:LSC-TR_Y_NORM_OUT16).

Y-arm trans = 0.932
Input power = 1.672 W

- I went to the end of ENGAGE_ASC_FOR_FULL_IFO in the last lock at GPS 1300503766.

Y-arm trans = 1634
Input power = 1.655 W

- The PRM transmission is Tp = 0.031.

From these numbers, we can calculate the PRG:
PRG = Tp * Y-arm trans (full ifo) / Y-arm trans (yarm only) * Input power (yarm only) / Input power (full ifo)

This yields a PRG = 54.9.
The PRG channel H1:LSC-PR_GAIN_OUT16 = 54.2 during the locking time, so it was correct the whole time.

----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

This calculation relies on some assumptions:
1. The Y-arm gain is the same when locking Y-arm by itself and in full lock. (The calculated PRG will be biased up if the full IFO Y-arm gain is lower than the Y-arm gain locked by itself)
2. The TRY photodetector is well-aligned for both IFO configurations.  

The code for verifying the prg calculation is in an aligo labutils git repo: prg_calibration.py.
This is also in /ligo/gitcommon/labutils/prg_calibration/prg_calibration.py on Hanford CDS machines.