An unattended and not working ITF gave me the opportunity to make some investigation concerning the alignment working point and the impact of an excess SIB1 motion on the locking stability.
While the ITF was systematically losing the lock at the beginning of CARM null or before, I moved SIB1 TY and TZ to the less noisy local control. The difficulty to acquire the lock did not change (maybe a CITF lock acquisition a bit faster?).
A stable CARM NULL was achieved when I tried to push manually NE_TX in the good direction, the one which lower the position of the resonant beam on NI. Maybe this has been a chance, but I think the way the NE alignment is managed at the very beginning of the lock (jumped / not jumped state) can be reviewed.
The fact that the ITF is more robust when NI_MIR_Y_AA is negative has been observed several time. It corresponds, as usual, to the fact that the sidebands are higher in that region. It could be due to a better alignment of COMMp TX on a shifted beam, or it could be due to a better NI map in that region. Anyway, I have never seen the sidebands crossing a maximum when the optical axis has been moved lower and lower.
Tonight it was possible to stay stable with an offset -3 mm or -4 mm on NI Y, and all the other positions at zero (position on WI not controlled). This is not a news: the same situation was experienced last Sunday.
An interesting result has been obtained turning on SIB1_TY linear alignment full band. The beam tilt registered by B2 quadrant, in loop with PR transversal, was really relevant, and too fast to be compensated by PR loop. A worsening of ITF stability was not evident soon, but in a while a couple of sidebands drops occurred, well correlated to a SIB1 motion. The third drop caused the unlock. This problem requires a solution, but a temporary patch could be staying in drift control. A stable improvement could be achieved passing the control of B2 from PR translation to INJ. Among the two options: BPC_TY or SIB1_TY, now the best seems SIB1_TY, provided that the noise comes from its control. Anyway I think that noise should not be there.
It would be also useful to quantify how much this beam tilt (around SIB1?) is seen by B4 quadrant, and if it is compatible with the use we want to do of that sensor.
During a second lock similar to the first, I have turned on SIB1_TZ linear alignment full band. The effect on the angular sensors is very similar. On the ITF stability it seems less invasive. I leave everything like that: the not reverted actions, even in case of unlock, are SIB1 in drift control and pointing on END TM centers. Tomorrow morning the standar working condition will be restored.