Title: 10/21 OWL Shift: 23:00-7:00UTC (16:00-00:00PDT), all times posted in UTC
State of H1: Observation Mode at 78Mpc for the last 15hrs
Shift Summary: Locked my whole shift, more saturations than usual, microseism is still climbing to dangerous levels, no CW inj.
Incoming Operator: Jeff B
Saturation Log:
Observing @ 78Mpc, locked for 12 hours
LLO is up
Environment calm
Handful of glitches
Kyle will report more but here is a plot of pressure for the last 4 hours. We are not sure how permanent this "fix" might be but the magnitude of recovery suggests that this is the dominant or only leak in this volume.
This morning the pressure is at 1.6e-9 torr. We had been at 4.6e-9 torr. At these pressures the main ion pump may have 1000 l/s pumping speed (guess) for air. Given this the air leak was on the order of 3e-6 tl/s.
The largest signal seen on the helium leak detector was ~ 1e-7 tl/s for helium. However, the leak detector was only sampling a fraction of the helium as we are open to the beam tube on both sides.
Kyle, John Today we applied Apiezon Q putty to that portion of the leak region that we could access, i.e. 2" of seam weld+stitch weld behind the support gusset and 2" of seam weld+stitch weld on the opposite side of the stiffener ring (not blocked by the gusset). Basically the known region of the leak site minus the inaccessible space between the stiffener ring and the spool wall -> No change in Y-mid pressure. Kyle After lunch, I wet the entire region with Isopropyl alcohol -> no change -> Next, I removed the Apiezon Q putty using various scrapers and wire brushes and then used opposite facing wedges to compress a silicone rubber mat against a cut "feature" (that Gerardo had noticed yesterday which is just below the seam weld behind the support gusset and which is ~1.25" away from the stiffener ring) and the support stand. The Y-mid pressure began to respond. This cut had been covered by the putty and should have been sealed. Kyle, Gerardo, John, Bubba We removed the silicone rubber mat but the pressure continued to drop -> We sprayed the entire area with Isopropyl alcohol and the pressure continued to drop -> We wet the region with Acetone and the pressure continued to drop -> We applied VACSEAL (SPI #5052-AB, Lot 1180402) to the cut and the pressure continued to drop CONCLUSION: It may be that the compressed rubber mat was more effective at getting remnants of the initial putty applied into the deep cut feature than was John and I's initial attempt using rigid flat tools. Another possibility is that the brushes used to remove the putty may have embed putty more effectively into the leak. Phase-change freezing of the initially applied alcohol is another possibility except that the leak stayed plugged for longer than the alcohol would have stayed frozen. At this point, I think that we can say that the leak is at the cut in the spool wall. How permanently it stays plugged will be a function of the interaction of all of the substances which were applied to it.
J. Kissel, for R. McCarthy, R. Schofield, V. Roma, J. Warner This has came up recently after I'd asked Duncan Macleod to plot it on the summary pages, but I've found out today, through a grave vine of those mentioned above, that the vault STS2 is busted. It's under semi-active investigation, also by the team mentioned above, but I just want to make sure something is in the aLOG about it. See attached ASD demonstrating the badness. Also for the record, I used a DTT template that lives in the SeiSVN, /ligo/svncommon/SeiSVN/seismic/Common/Data/2014-01-31_0000UTC_GND_STS2_ASD.xml with the calibrations originally from LHO aLOG 9727.
Title: 10/21 Eve Shift: 23:00-7:00UTC (16:00-00:00PDT), all times posted in UTC
State of H1: Observation Mode at 75Mpc for the last 7hrs
Outgoing Operator: Travis
Quick Summary: He seemed to have a good shift after the commissioners got things going again. Been locked since with a few measurements while LLO is down.
Title: 10/21 Day Shift 15:00-23:00 UTC (8:00-16:00 PST). All times in UTC.
State of H1: Observing
Shift Summary: Commissioners were mid-relocking when I arrived this morning. After some hunting down of alignment changes of the IMs, and accepting them in SDF, in addition to letting violin modes rings down, the H1 locked without much of a problem. It has been locked for 7 hours now. Microseism is trending upwards, but wind is calm.
Incoming operator: TJ
Activity log:
15:30 Sheila and Kiwamu to LVEA to turn on PZT driver
16:27 Joe D to X arm for beam tube sealing
18:09 Kyle to MY leak hunting
19:30 Kyle back
20:13 Joe D back to X arm
20:33 Kyle to MY
21:22 Kyle done
21:38 Richard to roof deck
21:48 Kyle, Gerardo, John, and Bubba to MY
21:50 Richard off roof
22:11 Joe D done
22:44 Kyle, Gerardo, John, and Bubba back from MY
Observing Mode at 21:12 UTC.
We have a few piece of evidence that suggest that anthropegenic noise (probably trucks going to ERDF) couples to DARM through scattered light which is most likely hitting something that is attached to the ground in the corner station.
On monday, Evan and I went to ISCT6 and listened to DARM and watched a spectrum while tapping and knocking on various things. We couldn't get a response in DARM by tapping around ISCT6. We tried knocking fairly hard on the table, the enclosure, tapping aggresively on all the periscope top mirrors, and several mounts on the table and nothing showed up. We did see something in DARM at around 100 Hz when I tapped loudly on the light pipe, but this seemed like an excitation that is much louder than anything that would normaly happen. Lastly we tried knocking on the chamber walls on the side of HAM6 near ISCT6, and this did make some low frequency noise in DARM. Evan has the times of our tapping.
It might be worth revisiting the fringe wrapping measurements we made in April by driving the ISI, the OMC sus, and the OMs. It may also be worth looking at some of the things done at LLO to look accoustic coupling through the HAM5 bellow (19450 and
14:31: tapping on HAM6 table
14:39: tapping on HAM6 chamber (ISCT6 side), in the region underneath AS port viewport
14:40: tapping on HAM6 chamber (ISCT6 side), near OMC REFL light pipe
14:44: with AS beam diverter open, tapping on HAM6 chamber (ISCT6 side)
14:45: with OMC REFL beam diverter open, tapping on HAM6 chamber (ISCT6 side)
14:47: beam diverters closed again, tapping on HAM6 chamber (ISCT6 side)
All times 2015-10-19 local
I've made some plots based on the tap time Evan recorded (the recorded time seems off by half a minute or so compare to what really shows up in the accelerometer and DARM). Not all taps created signals in DARM but every signal that showed up in DARM has the same feature in a spectrogram (visible at ~0-300Hz, 900Hz, 2000Hz, 3000Hz, and 5000Hz. See attachment2). Timeseries also reveal that whether or not the tap would show up in DARM does not seems to depend on the overall amplitude of the tap (seen in HAM6 accelerometer, see attachment 3). PEM spectrum during different taps times doesn't seem to give any clue why one tap shows up in DARM more than the other (attachment 4,5). Apology for the wrong conclusion I drew earlier based on the spectrum I plotted using wrong GPS time (those plots have been deleted).
I zoomed in a little closer at higher frequency and realized this pattern is similiar to the unsolved n*505 glitches. Could this be a clue to figuring out the mechanism that caused the n*505?
Out of Observing Mode for jitter measurements while LLO is down.
Received a verbal GRB alert at 19:00 UTC. Unfortunately, LLO was down at the time.
Keita, Evan, Hugh, Kiwamu,
When clearing the SDF differences, we noticed that the pitch bias offset of IM1, 2 and 3 are very different from the past (by ~ 1000 urad). These new offsets are the results of an alignment recovery effort by Cheryl at around 2 pm PT yesterday.
We checked various trends to understand why IMs needed so much change in the bias, but no clue was found. In the end, we accepted the new IM alignments even though we do not understand why they had moved. See the attached for the SDF difference.
The recovery effort
It looks that yesterday Cheryl changed the bias of IM1, 2 and 3 in order to bring them back to a place where their OSEMs read the same alignment values. As written below, after the interferomter was fully locked in this morning, the relevant QPD signals indicated that the laser light hits the same postion in the HAM2 and HAM3 area. This means her alignment recovery really aligned the beam back to where it should be. Or, in other words, IMs really had moved during the power outage for some reason. We still do not know why IMs had moved.
QPDs do not indicate a significant change
We checked the IM4 trans and POP_B QPDs in full lock to see if there is a significant difference in the actual laser beam. It seems that they are back to where they were in the past. This means that the interferomter alignment is not significantly different from the past with the new IM alignment. Good.
Also we looked at the HAM2 ISI oplev. Even though the oplev beam was not centered, we did not see a major change in PIT and YAW signals. They are about the same as before the power outage within a 1 urad.
Note that since POP_A is a part of the ASC control loop, it stayed centered.
HEPI and ISI
With help from Hugh, we checked HAM2 HEPI and ISI to see if they are back to nominal. And they are the same as before except for two HEPI prignle modes. See the attached for a trend of HEPI position locations. The HP mode moved by 5e5 (nm?) and VP mode changed by 3.5e3 (nm?). We are not sure how significant these numbers are in terms of the ISI table alginement.
Note that since the pringle modes are not DC-controlled and therefore it is sort of natural that they did not come back to where they used to be.
The power outage of yesterday caused the EPICS IOC to not exit completely, so the instructions to restart the process did not work. I logged in to the computer running the process and manually killed the dewpoint process. At this point the instructions to restart (in the CDS wiki) worked. Data is now being collected from the dewpoint sensors.
We have made it back to Observing Mode at 16:59 UTC. The violin modes are a bit rung up, but they are coming down in their characteristically slow fashion.
I ran the A2L script before going to Observing. It gave the following error:
IOError: [Errno 13] Permission denied: 'LinFit.png'
The script creates a temporary file (LinFit.png) that I had forgotten to chmod, so the Ops account couldn't write to it.
Travis just successfully ran the a2l from the Ops account, so I *think* all the bugs are fixed. Note that the script won't return the command line prompt to you until you hit Enter on the keyboard, so it's a bit confusing as to when it's finished, but if all the SDF diffs are gone, it's over and you can go to Observe.
Things that Jeff B and I have done:
Jeff was having difficulty locking the X arm in IR for inital alignment. Jeff moved SR3 pit to bring witness and oplev back to where they were before the power outage (both indicated that it lost 1-2urad of pit in the outage) (Still could not lock X arm in IR) Jeff re-ran the green WFS alignment, and moved PR3 by 0.5 urad in yaw.
checked all shutters to retore them to the way they were before the outage. (opened ISCT1 spare, IOT2L spare, and X end fiber.) This uncovered an interesting bug, which is that if theX green beam shutter is open, and the fiber shutter is closed, opening the fiber shutter will close the green beam shutter. In most other situations both of these shutters seem to work fine.
After this the X arm locked. Jeff went through the rest of inital alingment without incident until SRC align. Since SR3 was closer to the alignment from before the outage, we re-engaged the cage servo, which was turned off earlier probably because SR3 had moved. the SRC alignment was way off when we started the SRC align step and Jeff aligned by hand to get us close.
We then were able to lock DRMI several times, but had difficulty engaging DRMI ASC as people described last night. We tried doing things by hand, it seemed like we were OK engaging the MICH and INP1 loops, but PRC2 was a problem. In august, Evan and I changed the input matrix for this loop to no longer use refl 45 (alog 20811.) I found the old input matrix in the svn and tried it. This worked once so I've reverted the matrix in the guardian.
Input matrix for PRC2 since august 24th: refl91A-refl9IB
input matrix before august 24th and now:
# PRC2 REFLA45I - REFLA9I
asc_intrix_pit['PRC2', 'REFL_A_RF45_I'] = asc_intrix_yaw['PRC2', 'REFL_A_RF45_I'] = 0.83
asc_intrix_pit['PRC2', 'REFL_A_RF9_I'] = asc_intrix_yaw['PRC2', 'REFL_A_RF9_I'] = 0.5
asc_intrix_pit['PRC2', 'REFL_B_RF9_I'] = asc_intrix_yaw['PRC2', 'REFL_B_RF9_I'] = 0.5
asc_intrix_pit['PRC2', 'REFL_B_RF45_I'] = asc_intrix_yaw['PRC2', 'REFL_B_RF45_I'] = 0.83
This matrix gets reset in the full lock ASC engage states, so this is not a change to the full lock configuration.
After this change we made it past DRMI, to the point where DHARD WFS are engaged durring the CARM offset reduction. The bounce and roll modes are rung up, we spent a few minutes damping them but probably moved on too soon and lost lock probably due to roll modes in the final stages of CARM offset reduction.
Kiwamu and I spent about 15 minutes locked at CARM 10 picometers to damp bounce and roll. We lost lock after that, possibly because the IFO got misaligned as we were sitting at 10pm damping.
In the next lock, we saw that ETMX violin modes are also rung up, Kiwamu lowered the damping gains to stop PUM saturations.
We made it through engaging the ASC in full lock, and found that we couldn't lock the OMC because the Kepco power supply was off.
After we made it to low noise, I cleared 82 diffs in the ASC SDF. Most of these were due to the dark offset script that Jenne and Jeff ran last night. I accidentally accepted all of these with one button click (I hit accept all assuming that was only the first page which I could read, but accept all really means accept all.) Betsy pointed me to the last version to be accepted in SDF, so I was ble to check on the things I had inadvertently accepted. There were a few oddball things, like ADS SIG DEMOD TRAMPS, (I accepted the new 3 second TRAMPs), and offsets in the SRC1 loops (now set to 0).
During ER8, there was a calibration artifact around 508 Hz - a non-stationary peak with a width of about 5 Hz. The peak went away on Sep 14 16 UTC probably due to an update of the calibration filters which was documented in this alog. When re-calibrated data is produced, it's worth having a look at some of this ER8 time to check that the peak is removed. I made a comparison spectrum a bit before and after the change of the filters (plot 1). The wide peak is removed and the violin modes that it covers (ETMY modes, maybe some others) appears. I did the same thing for a longer span of time, comparing Sep 11 and Oct 17 (plot 2). The artifact manifests itself also as an incoherence between GDS-CALIB_STRAIN and OMC-DCPD_SUM (plot 3). The only other frequency where these channels aren't coherent is at the DARM_CTRL calibration line at 37.3 Hz. I've also made a spectrogram (plot 4) of the artifact. It has blobs of power every several seconds. The data now looks more even (plot 5), though it's more noisy because the calibration lines are lower.
I agree that it was due to bad digital filters in CAL-CS. See my recent investigation on this issue at alog 22738.
Executive summary:
A matlab file (37 MB) containing the averaged inverse-noise-weighted spectrum from the first week can be found here: https://ldas-jobs.ligo.caltech.edu/~keithr/spectra/O1/H1_O1_week1_0-2000_Hz.mat Because of the way multiple epochs are handled, the matlab variable structure is non-obvious. Here is how to plot the full spectrum after loading the file: semilogy(freqcommon,amppsdwt{1,1})
Keith has found: "There is a sporadic comb-on-comb with 0.088425-Hz fine spacing that appears with limited spans in three places near harmonics of 77, 154 and 231 Hz (ambiguity in precise fundamental frequency)" Using the coherence tool, we have seen coherence between h(t) and a number of auxiliary channels that shows this comb around 77 Hz. Seems to be around the input optics, in channels: H1:PEM-CS_MAG_LVEA_INPUTOPTICS_Z_DQ H1_SUS-ITMY_L1_WIT_L_DQ H1:SUS-BS_M1_DAMP_L_IN1_DQ H1_SUS-ITMY_L1_WIT_P_DQ H1:SUS-BS_M1_DAMP_T_IN1_DQ H1_SUS-ITMY_L1_WIT_Y_DQ H1:SUS-BS_M1_DAMP_V_IN1_DQ H1_SUS-ITMY_L2_WIT_L_DQ H1:SUS-BS_M1_DAMP_Y_IN1_DQ H1_SUS-ITMY_L2_WIT_Y_DQ See the attached figures. Nelson, Soren Schlassa, Nathaniel Strauss, Michael Coughlin, Eric Coughlin, Pat Meyers
The structure at 76.4Hz Nelson listed some channels for above shows up in at least 50 other channels. Greatest coherence is consistently at 76.766 Hz, second greatest is (mostly) consistently at 76.854Hz. Spacing between the two combs is about 0.0013Hz. The epicenter seems to be the INPUTOPTICS/the SUS-BS and SUS-ITM* channels, like Nelson said (see below for fuller list). The plots above are pretty typical, but I have plots for all channels listed and can post any more that are useful. Most or all channels showing the comb with max coherence greater than 0.1 are listed below. Max coherences over 0.2 are marked below as strong, and max coherences under 0.15 as weak. Those marked strongest are around 0.22. I haven't included anything of max coherence <0.1 but I'm sure there are many. H1:ASC-AS_A_RF36_I_PIT_OUT_DQ (weak) H1:ASC-AS_A_RF36_I_YAW_OUT_DQ H1:ASC-AS_A_RF36_Q_PIT_OUT_DQ H1:ASC-AS_A_RF36_Q_YAW_OUT_DQ (weak) H1:ASC-AS_B_RF36_I_YAW_OUT_DQ H1:ASC-AS_B_RF36_Q_YAW_OUT_DQ (strong) H1:ISI-BS_ST2_BLND_RZ_GS13_CUR_IN1_DQ (strong) H1:ISI-BS_ST2_BLND_Z_GS13_CUR_IN1_DQ (strong) H1:ISI-HAM2_BLND_GS13RZ_IN1_DQ H1:ISI-HAM2_BLND_GS13Z_IN1_DQ H1:ISI-HAM3_BLND_GS13Z_IN1_DQ (strong) H1:ISI-HAM5_BLND_GS13RZ_IN1_DQ H1:ISI-HAM5_BLND_GS13Z_IN1_DQ H1:ISI-HAM6_BLND_GS13RZ_IN1_DQ H1:ISI-ITMX_ST2_BLND_RX_GS13_CUR_IN1_DQ (weak) H1:ISI-ITMX_ST2_BLND_Z_GS13_CUR_IN1_DQ (strong) H1:ISI-ITMY_ST1_BLND_RZ_T240_CUR_IN1_DQ (weak) H1:ISI-ITMY_ST1_BLND_Y_T240_CUR_IN1_DQ (weak) H1:ISI-ITMY_ST2_BLND_RZ_GS13_CUR_IN1_DQ (strong) H1:ISI-ITMY_ST2_BLND_Z_GS13_CUR_IN1_DQ (strong) H1:LSC-PRCL_IN1_DQ H1:PEM-CS_LOWFMIC_LVEA_VERTEX_DQ (strong) H1:PEM-CS_MAG_LVEA_INPUTOPTICS_Y_DQ (strongest) H1:PEM-CS_MAG_LVEA_INPUTOPTICS_Z_DQ (strong) H1:SUS-BS_M1_DAMP_L_IN1_DQ (strongest) H1:SUS-BS_M1_DAMP_T_IN1_DQ (strong) H1:SUS-BS_M1_DAMP_V_IN1_DQ (strong) H1:SUS-BS_M1_DAMP_Y_IN1_DQ (strong) H1:SUS-ITMX_M0_DAMP_R_IN1_DQ (strong) H1:SUS-ITMX_M0_DAMP_V_IN1_DQ (strong) H1:SUS-ITMY_L1_WIT_L_DQ (strong) H1:SUS-ITMY_L1_WIT_P_DQ (strong) H1:SUS-ITMY_L1_WIT_Y_DQ (strong) H1:SUS-ITMY_L2_WIT_L_DQ (strong) H1:SUS-ITMY_L2_WIT_P_DQ (strong) H1:SUS-ITMY_L2_WIT_Y_DQ (strong) H1:SUS-MC1_M3_WIT_L_DQ H1:SUS-MC1_M3_WIT_P_DQ (weak) H1:SUS-MC2_M1_DAMP_L_IN1_DQ H1:SUS-MC2_M1_DAMP_T_IN1_DQ H1:SUS-MC2_M1_DAMP_Y_IN1_DQ H1:SUS-PR2_M1_DAMP_P_IN1_DQ H1:SUS-PR2_M1_DAMP_R_IN1_DQ H1:SUS-PR2_M1_DAMP_V_IN1_DQ H1:SUS-PR2_M3_WIT_L_DQ H1:SUS-PR2_M3_WIT_P_DQ (weak) H1:SUS-PR2_M3_WIT_Y_DQ (weak) H1:SUS-PR3_M1_DAMP_P_IN1_DQ H1:SUS-PR3_M1_DAMP_V_IN1_DQ H1:SUS-PRM_M1_DAMP_L_IN1_DQ (strongest) H1:SUS-PRM_M1_DAMP_T_IN1_DQ H1:SUS-PRM_M1_DAMP_Y_IN1_DQ (strong)
The 99.9989Hz comb Keith found (designated H) appears in 109 channels (list is attached). Coherence is uniformly greatest at the ~500Hz harmonic, with many channels approaching .7 and greater, drops off sharply at the ~600Hz and ~700Hz, and is invisible after 700. (See spreadsheet titled "comb_H_sigcohs_wk1.xslx" for a list of cohering channels by line, with coherence value.) At all harmonics except the ~300Hz, the structure manifests in the signal and the coherences as two lines .001Hz apart, but if I recall correctly .001Hz is the resolution of the frequency series, so it's safer to say that this is a bulge with .001Hz < width < .002Hz. At ~300Hz, almost all the cohering channels with data in that range show a bulge of width about 0.5Hz (see attached "disjoint_plots" for a comparison of typical channels by harmonic). This bulge, and the fact that it appears in all the same channels associated with the rest of the comb, makes me think that the fundamental may be the bulge at ~300Hz and not the line at 99.9989Hz. An interesting feature of the bulge is that in many cases, it has a prominent upward or downward spike at 299.96Hz, which is just the place the line would be if it were there (see "bulge_w_spike.jpg"). More to come re: changes in week 4 data, patterns in cohering channels, and the spike.
I have update the violin mode filters in CAL CS in order to make them more accurate. This will impact on the calibration at sub-percent level at around 30 Hz.
Joe B pointed me out that the way my matlab script (alog 21322) handles violin's zpk was not ideal (i.e. I was implicitly assuming certain ordering in the zeros and poles in zpk data format). I corrected the script as was already done in Livingston (LLO alog 20512). This resulted in somewhat better accuracy for PUM in 1-100 Hz . The attached screenshots are the new filters and discrepancy between the full ss model and the installed discrete filters. Compared with the one I previously reported in alog 21322, the magnitude of PUM is now somewhat better. The magnitude of PUM at 30 Hz is now more accurate with a very small discrepancy of 0.08 % (which used to be 0.2 % discrepancy ), and it is also more accurate at 100 Hz with a small discrepancy of 0.65 % (which used to be 2.4 %). I do not expect any noticeable change in the binary range with this update.
I have installed the new filters and loaded the coefficients in CAL-CS.
This is a follow up on the change we made on the L1 and L2 stage violin mode filters for calibration.
As Andy reported in alog 22631, there had been a prominent peak at 508 Hz before the change on the violin calibration filters. This was due to the fact that both ETMY L1 and L2 stages of CAL-CS had a too high violin mode by mistake (see the original entry above) at 508 Hz. The spectral shape of the violin modes that he posted looks very similar to what we mistakenly had in CAL-CS before Sep. 14th.
I am concluding that the 508 Hz nonstationary behavior seen in the calibrated signals before Sep. 14th are indeed artifact due to too-high response in the violin calibration filters.
I made a comparison between the violin calibration filters before and after my fix on the matlab code. See the attached screenshots below:
Fig.1 L1 stage violin calibration filter. (Blue) before the bug fix, (red) after the bug fix.
Fig.2 L2 stage violin calibration filter. (Blue) before the bug fix, (red) after the bug fix.
It is clear in the plots that the previous filters had a high peak at 508 Hz and they were as tall as 120 dB ! Therefore the ETMY suspension calibration filters must have been unnecessarily sensitive to any small signals in DARM_CTRL before Sep. 14th. In fact, this was exactly the thing I was worried and was the main motivation to decrease the violin Qs down to 1e3 (alog 21322). Note that, according to the suspension model, the Q-factor of the violin modes can be as high as 1x109. However, we decided to artificially decrease the violin Qs for the actuator calibration filters in order to maintain the IIR filters reasonably stable. Otherwise, the modes would be easily rung up by numerical precision errors, a small step in the actual signal or anything.
I also attach the difference of the filters in zpk formt. See the third and fourth attachements. Since their violin Qs are chopped off to be 1e3, both L1 and L2 stages have the same frequency response.