fred.raab@LIGO.ORG - posted 16:15, Friday 25 December 2015 - last comment - 03:40, Saturday 09 January 2016(24477)
Christmas Day LIGO Fellow Report
Nutsinee had a tough time on OWL shift with H1 behaving badly until approx. 15:10 UTC, when it mysteriously improved. But the bad behavior returned around 17:00 UTC for about 30 minutes and I spent some time investigating its cause during this interval. Microseism was not too bad and wind was calm. There was no indication of excess noise on H1:LSC-MOD_RF45_AM_CTRL_OUT_DQ or coherence of that channel with DARM. Looking at a spectrogram of H1:CAL-DELTAL_EXTERNAL_DQ one can see broadband bursts across the bucket turning on and off suddenly (1-s resolution), with durations verying from a few seconds to about 30-40 seconds. Sometimes these were accompanied by loud bursts in the 10-20 Hz region, other times not. I looked for burstiness in other chnnels that was correlated with this. There was nothing well correlated in seismometers or microphones (either on the floor or in e-bays) or OPLEVs. There were strong similarities on ASC-MICH_P and _Y, weaker on DHARD_P and Y, nothing on DSOFT_P and _Y, nothing on CSOFT, very weak similarities on CHARD and SRC2 P and Y. I am not familiar enough with the LSC/ASC couplings to know if this is just leakage of the length fluctuations onto ASC channels or if alignment is causing the problem. I looked at IMC_F an _L, MC2 TRANS which showed no indication that the cause came into the IFO on the light from the IMC. I tried looking at PRC ASC signals but they looked crazy in frequency with no time dependence (??). However SUS-PRM_M3_NOISEMON has large burst in the 10-30 Hz region on all OSEMS. HVETO finds these good for veto channels. Travis was not finding anything bumping against its limits. At 19:42:12 we lost lock. This coincides with the arrival of the maximum ground shaking in the 0.03-0.1 Hz from an EQ in Afghanistan, although the 1-3 Hz region showed sharp elevated shaking several minutes earlier. We are having a hard time getting back to locking.
Images attached to this report
Comments related to this report
It looks like the RF45 monitor sometimes does not see the RF45 noise. We will consider three times on the 25th. The reference time is 4 UTC, when there was nothing bad in DARM and the RF 45 was quiet. At 7:30, the RF 45 is obviously bad as seen by the monitor channel. At 17:04, there's very similar noise in DARM, but now the RF45 monitor stays at its reference. The first plot is DARM for the three times. The shape and amplitude of the excess noise for the two bad times is very similar. The second plot is the spectra of the RF 45 monitor at these times. It easily sees the problem the first bad time, but at the second bad time (corresponding to a burst of noise in Fred's spectrogram) it stays at the reference. The ASC-MICH channels look to be good witnesses of this noise. I think they're made from RF36, but that should be sensitive to RF45 issues. The third plot is the coherence with h(t) of RF45 and MICH_Y during the reference time. It has a few lines but is otherwise zero. The next plot shows high coherence with both during the first bad time. The last plot is coherence during the second bad time. There's just a tiny bit of coherence with the RF 45 witness. The MICH_Y coherence is basically the same as the other bad time. This is worrisome because the RF 45 monitor is not always a good witness of the noise in DARM (and other channels). But it doesn't seem to be a problem with noise in the witness channel masking the RF 45 junk. Maybe this points to the problem being somewhere that the monitor is not always able to see.
Images attached to this comment
The reason that there's excess noise in PRM_M3_NOISEMON is just because of the control signal from PRCL. The attached spectrum shows that PRCL gets worse during the bad times, and that's just getting fed to PRM_M3. In fact, MICH, PRC, and SRC all have a similar noise shelf up to 30 Hz during the bad time. They all involve some RF 45 signal in their production. The POP_A RF9 and RF45 photodiode signals, in the I and Q quadratures, all have similar shelves except for 9Q (second plot). That's the only one that's not used for control, so I think it's seeing the noise impressed by the MICH/PRC/SRC loops all suppressing the RF 45 noise.
Images attached to this comment
