Yesterday and today we observed a x-arm power build-up loss when locked over about 10min. The buildup recovered when not locking the arm on IR.
- We now think that the length feed-back drives a 5-ish Hz mode in MCE (most likely pitch), which then grows so strong that it pulls the arm cavity off resonance, resulting in a build-up loss.
- We can observe that mode grow in AOS-TMSX_IR_PD_OUT_DQ and VIS-MCE_TB_OPLEV_PIT_OUT_DQ at the same time as the arm power drops.
- Of course, the question is why the arm gain isn't high enough to completely squash this mode in the first place. It looks like sometime last week the gain in the CARM loop was significantly dropped (more than 10x), because the loop started oscillating at higher frequencies. Yutaro and Valera are now in the Mountain to further diagnose the problem.
== Turned out to be due to a failure of CARM servo board ==
Valera, Keiko, Yutaro
= Summary =
To investigate this issue, we went inside the mine and looked into the CARM (X arm) loop.
As a result, we found that CARM servo seems to be broken somehow and X arm can be locked in a more or less nominal setting (a few kHz UGF) by using another common mode servo.
We have to replace CARM servo with a spare one.
= What we did =
- measured the OLTF of CARM loop, and found that the gain was too low (by ~20dB). We tried to increase the gain, but that lost the build-up somehow.
- measured all the TF around CARM loop, including IMC loop gain, MCE actuation gain, and servo gain itself. Everything looked fine.
- found that the switch for Common boost1 in CARM servo is not working right now.
- tried to lock the X arm with PLLY board instead of CARM board. We were able to lock the X arm with high gain (a few kHz UGF).
- went back to the original configuration just to make sure, but again we could not lock it with high gain using CARM board. So we concluded that CARM servo is broken.
We guess that the noise of CARM servo is too high to go to high bandwidth lock.
- left CARM servo as is, because we barely lock the arm with presumably-broken servo. We have to replace it ASAP.
== Replacing the CARM servo did not solve the issue ==
As entitled.
It remains to be a mystery.
See the attachment. Before June 20, we were able to lock it with more or less nominal gain setting, but we could not since July 2.
Something should have happened between Jun 20 and Jul 2.
