Reports 1-1 of 1 Clear search Modify search
AdV-ISC (Commissioning up to first full interferometer lock)
casanueva, bersanetti, nardecchia - 23:51 Tuesday 27 July 2021 (52679) Print this report
CARM offset 0 and DAS ON

We started the shift after some tests, together with the ALS group, on the maximization of the transmitted green power reaching the PDs using the PZTs on the EIB. A more detailed entry on this topic will follow.

Then we started the proper ISC shift, with the first target of reaching CARM offset 0. Before doing so, we solved a few issues.

  • We adjusted the DARM demodulation phase during the whole lock acquisition, based on the analysis done by Madda on logbook entry #52674 and the monitor of the phase tuning, LSC_B1p_56MHz_DPHI. During the rest of the shift, we found that this DPHI is not constant (we had moments where the phase was off by 0.2 for example, depending from lock to lock) between locks, but it is consistant between the different steps in one lock acquisition (that is, if it is wrong by 0.2 at 50mW it is also at 100, 200 etc.). However the DPHI was quite small during the shift.
  • Regarding the gains during the lock acquisition, they were still good. We only had to tune some of them in CARM offset 0, since yesterday there was no time to finish with the fine tuning.
  • We payed special attention to the angular loops, and we noticed that the demodulation phase of the error signals for the PR loop was very mistuned (delta of 0.5rad), final value 4.6 rad.

Once we recovered the robustness of the lock we decided to focus on the DIFFp loop in CARM offset 0.

  • We first tried to simply go to carm offset 0 with DIFFp closed in drift control. As soon as we arrived, there was an oscillation around 5 Hz that was appearing and disappearing from DARM. eventually getting very big and causing an unlock.
  • Then we tried to open it before carm offset 0, and re-close it there. Same 5 Hz oscillation in this case.
  • As a final try, we opened the drift control before carm offset 0 and then we engaged directly the full bandwidth. However, in this occasion we realized that the ITF unlocked almost immediately, after the error signals would drift away from their 0. At this point we convinced ourselves that the sign we were using for the lock acquisition was not the same in carm offset 0.
  • We finally tried to open the drift control before carm offset 0, got there, change the sign of the old gains and close the loops. We did it first in drift control without problems, and finally we were able to close the TX in full bandwidth (CARM null reached at 20:23:20 UTC). However, this not prevented this 5 Hz oscillation to appear in DARM and kill the lock.
  • Another problem that we encountered while studying this loop is the galvos of B1p. We noticed that several times, due to some fluctuations of the power, the galvos would open, making disappear the error signal of the DIFFp, even when we were using it! This problem needs to be solved if we want to have a robust alignment on dark fringe.
  • As a final test we turned off BOTH the drift control and the galvos and we re-closed them at dark fringe. However, this didn't work, and the galvos still opened while in carm offset 0.

As a summary, we encountered two new problems today. The 5 Hz oscillation very well visible in DARM loop, that ends up killing all the loops in CARM offset 0, and the movement of the beam making the galvos to open, removing the possibility of closing the DIFFp using B1p quadrants.

Comments to this report:
mwas - 10:48 Wednesday 28 July 2021 (52683) Print this report

Unfortnuately for the quadrants the problem is bigger than just the galvo loops.

Figure 1 shows one example of the galvo loop opening. Note the B1p QD2 sum signal in the lower left corner, it is clearly saturating at 25V, and then it stops to zero while there is still power fluctuation visible on B1p PD2 DC in the lower right corner. I expect that the Voltage bias of the quadrant is switched by the too large power on the quadrant. Although I am not sure how the internal protection of the quadrant works.

Figure 2 Shows the trend data, the Vbias is turned off, but that time series is sampled at 10mHz. So it is not clear when it was turned off exactly (and if it is automatic or with some manual command in the automation).

So the galvo loop opening is just a consequence of the quadrant switching off due to the power impinging on it being too high.

Images attached to this comment
Search Help