This evening Lisa pointed out that the low carrier buildup can be explained by misalignment in the Y arm.

I spent an hour to carefully adjust the alignment. Indeed, after the alignment, I got a cavity flash of 45 (normalized for a single arm build up) which is the highest we ever obtained. Also, the funny REFL9I offset disappeared and I could see a PDH-looking signal there when CARM passed through a resonance.

What I did:

• Aligned ETMY to maximize the IR in Y. (CAREFUL: This is an opposite way of the proposed global alignment procedure)
• Then steered TMSY and added an offset in the TMSY QPDs to recover a high buildup for the green light.
  • The resultant green buildup is now 10% smaller than the usual, but optimized for the infrared. Not terrible.
  • ALS-Y_QPD_A_PIT now has an intentional offset of 0.8. This gave a high buildup for the green light.
  • I did not steer ITMY or BS at all in this process. In other words, I relied on the PRMI local alignment.
  • Then I did a mode scan of the Y arm to see how good the alignment is. The spatial HOMs are now small. Their power are roughly 10-20 times smaller than the carrier 00. Good.
• I tried the locking process again, starting from a CARM offset of 700 Hz with REFL45 for PRMI. This worked fine.
  • The funny offset in REFL9I was much smaller this time. It was about 10 times smaller than yesterday. Good.
• I reduced the offset to let CARM resonate while holding PRMI with 135I and Q.
  • I didn't have to re-adjust the demod phase at all this time. It was robust until the arm cavity started building up the carrier light.
  • Apparently PRMI was not able to survive through the arm cavity resonances.
  • However if the sweep through a resonance is fast enough, PRMI does not drop the lock since the carrier does not build up so high.
• I disconnected the REFL_DC_BIAS cable from the common mode board and connected the REFL9I signal in order to try self-lcoking.
  • It seemed that a good control sign is positive at the input2 of the common mode board because this brought CARM closer to the resonance.
• I tried self-locking a couple of times with no success.
  • I was not able to approach the resonance as close as 100 Hz because of the fragile PRMI before introducing REFL9I. So I tried to add REFL9I when CARM was as far as 150 Hz or so.
  • Self-locking tends to push CARM to the resonance as expected, but IMC unlocked every time. The arm power went up to 3 or 4 before IMC unlocked.
  • I tried different gains (almost randomly) in input2, but no luck. I guess that CARM is basically too far from the resonance and therefore self-locking can not squash such a big offset by its gain.