Title: 9/19 Day Shift 15:00-23:00 UTC (8:00-16:00 PDT). All time posted in UTC.
State of H1: Lock acquistion.
Outgoing Operator: Jeff Bartlett
Summary: A series of EQs in Chile have had the IFO down for most of the OWL shift. They have rung down enough now that we should be able to lock. Alignment looks good, so I will attempt to relock without initial alignment first.
Activity Log: All Times in UTC (PT) 07:00 (00:00) Take over from TJ 07:00 (00:00) IFO relocking after EQ lock loss 07:16 (00:16) IFO locked at NOMINAL_LOW_NOISE, 22.9W, 72Mpc 07:22 (00:22) Set Intent Bit to OBSERVE 08:05 (01:05) ETMY saturation 09:56 (02:56) ITMX saturation 09:56 (02:56) Lockloss – 6.1 Mag earthquake near Chile 10:50 (03:50) IFO not locking – Run initial alignment 12:31 (05:31) While at LOCK_DRMI_1F BS ISI stage 2 WD trip – Reset watchdog 13:30 (06:30) BS ISI stage 2 WD trip – Reset watchdog 13:34 (06:34) Set IFO to DOWN while ring down from last EQ 14:34 (07:34) Received a GRB alert. IFO is currently down 15:05 (08:05) Turn over to Travis End of Shift Summary: Title: 09/19/2015, Evening Shift 07:00 – 15:00 (00:00 – 08:00) All times in UTC Support: Incoming Operatror: Travis Shift Summary: - 09:56 (02:56) – Lockloss 6.1 mag EQ near Chile – Seismic activity rung up, but coming down rapidly. Will attempt to relock. - 13:25 (06:25) – Another 6.2 mag EQ near Chile – Seismic activity has rung up again. Continue with relocking attempt. - 13:34 (06:34) – IFO repeatedly coming up to LOCK_DRMI_1F and higher before losing lock. Put IFO in DOWN state until seismic activity calms down. Spoke with LLO they are also holding for things to settle down. - 13:45 (06:45) – Another 5.4 mag EQ near Chile - 14:34 (07:34) – GRB alert. IFO down at this time
Recovering from 09:56 (02:56) lockloss. Relocking failing at various points along the Guardian path. Had DRMI_1F catch in split mode a couple of times. A tweak of the BS corrected one, went down on the other before I could catch it.
Transition Summary: Title: 09/19/2015, Owl Shift 07:00 – 15:00 (00:00 – 08:00) All times in UTC State of H1: Unlocked Outgoing Operator: TJ Quick Summary: IFO is relocking after lock loss due to earthquake. LLO also recovering at this time.
All times posted in UTC
STATE Of H1: Relocking after lockloss at 5:55, currently have DRMI locked.
SHIFT SUMMARY: Most of the shift was quiet except for a few ETMY saturations and one OMC DCPD.
INCOMING OPERATOR: Jeff B.
ACTIVITY LOG:
ETMY Saturations:
OMC DCPD Saturation - 2:10
Lockloss - 5:55
5:55 - Lockloss
As I was typing up a report saying how well things have been going, the ASC Striptool started looking bad. After a few minutes of this, it lost it.
I had a look at factors that impact our duty cycle, starting at midnight UTC Sept 10th until 2:43 UTC Sept 18th (the latest time available when I started downloading data.) I choose this time because there were fewer of the commisioning activities that maade it hard to get a good picture of the duty cycle in the earlier weeks of ER8.
Ground motion
It seem that we've taken much of the low hanging fruit in making our locking sequence more robust, while there are still parts that could be speed easily up the gains to be had from doing that are pretty small. Probably the most important thing we can do to improve the duty cycle now is to figure out why we are loosing lock durring relatively small earthquakes. Another alog about that comming soon.
TITLE: 09/18 [DAY Shift]: 15:00-23:00 UTC (08:00-16:00 PDT), all times posted in UTC STATE Of H1: Observing, ~70 Mpc SHIFT SUMMARY: Remaining half of shift quiet INCOMING OPERATOR: TJ ACTIVITY LOG: 19:36 - 19:41 UTC Stepped out of control room 20:52 UTC Turned off test points I hadn't realized I had open 21:04 UTC Large glitch in range, but did not see corresponding glitch in spectrum or any indication of saturation from the verbal alarm log. Shortly afterward there was another smaller glitch in range, but this time with an ETMY saturation and glitch in spectrum. 21:48 UTC Peter and Jason done taking premodecleaner in car from OSB to LSB. They put it on a cart in the OSB receiving and wheeled it down the hallway to avoid opening the rollup door. I did not note the start time. 22:20 UTC Kyle and Gerardo to mid Y to valve out aux cart from ion pump. 22:30 UTC Kyle reports helicopter flying over end Y 23:01 UTC Kyle and Gerardo back. The ion pump was able to maintain the pressure on its own, so they turned off and disconnected the aux cart. All should be back to normal.
Sometime around 21:20 UTC (14:20 PDT) a truck from Brockman office furniture arrived at the gate. I redirected them to the LSB.
State of H1: Observation Mode at 76Mpc for the last 6 hours.
Outgoing Operator: Patrick Thomas
Quick Summary: Patrick had one lockloss most likely due to an earthquake, he also encountered a very large glitch around 21:00. Other than that it seemed to be smooth sailing.
Kyle, Gerardo New pump running unassisted -> shut-down and removed aux. cart.
Daniel, Evan
We looked at the DARM spectrum in the vicinity of the OMC dither line (at 4100 Hz) and its second harmonic (at 8200 Hz) using the OMC DCPD IOP channels.
The spectrum around the line itself appears to be bilinear, perhaps as one would expect (if the OMC is exactly on resonance, there should be no first-order modulation of the transmitted power). The second harmonic of the line is quite clean, with no obvious upconversion. Is this what we should expect?
That's what you expect (when OMC transmission is quadratic to the length).
Power propto L^2 = (L0+dL+dither)^2 = L0^2 + 2L0(dL+dither) + dL^2+ 2dL*dither + dither^2
where L0 and dL are the DC- and AC-ish component of the OMC length and dither is the dither.
dL*dither^2 will only appear in L^3 response, and (dL*dither)^2 in L^4.
Summary:
I was baffled to find that each and every omega scan I have seen so far shows some wandering lines in H1:IMC-PZT_PIT_OUT_DQ and YAW that extends to hundreds of Hz, e.g.
https://ldas-jobs.ligo-wa.caltech.edu/~nairwita/wdq/H1_1126294545/#H1:IMC
These wandaring lines seem to be irrelevant of any of the glitches, and they're there even when there is no fast feedback to the PZT (indeed, since Sept. 09 the fast feedback has been disabled https://alog.ligo-wa.caltech.edu/aLOG/index.php?callRep=21300).
This is an artefact of single precision rounding error that happens when the data is saved on frame files (in the frontend the data is double).
That tells us that we need to be careful when looking at signals with big DC and small AC.
Analysis 1: Measured long data spectrum VS predicted single precision rounding error spectrum agree very well.
When you look at the first attachment, you'll see that PZT_OUT and PZT_IN1 are different now (blue and green) though the only difference should be that the former has a fixed offset added in. Also, the ratio of the noise level above 20Hz is approximately equal to the ratio of the DC value.
Though the rounding error spectrum depends on the wave form before the casting, if you take a long (relative to the frequency analyzed) data and make an average, like I did here, you can expect that it becomes white-ish. Since the RMS of the signals are totally dominated by the DC, and since I know that the exponents did not change during the measurement, you can calculate the expected white-ish noise spectrum as
2^(floor(log2(DC level))) / 2^23 / sqrt(12) / sqrt(1024),
i.e. 1.1E-6 /sqrtHz and 2.2E-6 for PZT PIT and YAW OUT respectively, which perfectly agree with the measured noise floor. 1/2^23 is a "1-bit" fraction, and 1/sqrt(12) because the RMS of +-0.5 bit fraction is 1/sqrt(12) bits when you assume a uniform distribution, and sqrt(1024) because this is first down sampled into 2kHz and then cast into single.
Analysis 2: Short data spectrum shows wandering structures
Since what we really want to know is short term variability at around 100Hz, I looked at the 6Hz BW spectrum.
In the video attachment, on the top is IMC-DOF_3_Y_IN1_DQ signal, which is the error point of one of ASC sensors and that's the only signal that goes to PZT YAW. On the bottom is the IMC-PZT_YAW_OUT_DQ.
Red are short data spectrum (2 averages exponential), blue are the long spectrum (0.047HZ, 10 averages fixed). This is from around the same GPS time as the example omega scan time.
Look how variable the bottom is, you can easily see the same thing as Omega scan!
This is not the case with the top plot. This is because the DOF3 signal is NOT limited by the single precision casting due to small DC value relative to the noise floor. DOF3 signal is zero on average, and RMS is almost always smaller than 0.05, so the expected rounding error noise floor is in the vicinity of 10^-10 to 10^-11/sqrtHz.
Analysis 3: How large is the "true" noise going to PZTs?
Third attachment shows DOF_3_PIT_IN1_DQ and YAW (these are the only source of PZT feedback after Sept. 09) at the example omega scan time, projected on the PZT output using the DOF3 filter. This can be regarded as the "true" noise as the high frequency noise of DOF3 is not limited by the rounding error.
Anyway, as you can see, PZT_OUT spectrum is at least 5 orders of magnitude larger than the "true" noise level at 100Hz, and there's nothing to worry about.
I just realized that I have had two LSC test point channels open since the beginning of this lock acquisition and in Observation Mode. I just turned them off. 20:52 UTC Stopped running H1:LSC-TR_X_NORM_OUT and H1:LSC-TX_Y_NORM_OUT test points in dataviewer. Checked that they are no longer present on the H1LSC_GDS_TP medm.
Whoa. How do we know if we have test points up? Heck, I might have been looking at these channels, too (or did I have non-test point channel versions of these channels?). Is it bad to look at Test Points? (if it is, is there a monitor to show us test points are being looked at?)
This is ok as long as we don't open way too many.
ER8 Day 32. No restarts reported.
ER8 Day 31. No restarts reported.
15:41 UTC Joe D. called from VPW. Turning on a couple of water pumps and a water tap. 16:32 UTC lockloss, earthquake, either 5.4 in Northern Mid-Atlantic Ridge (15:59:42 UTC) or 4.5 63km W. of Ovalle, Chile (15:58:37 UTC) 16:36 UTC Reset OMC SUS WD ALS is not staying locked, still riding out earthquake 16:46 UTC Kyle and Gerardo leaving control room to start replacement of ion pump at mid Y (WP 5500) 16:52 UTC Attempting to lock DRMI 16:56 UTC Richard reports he turned on AC for LDAS room in staging building, it had not been on 16:58 UTC Locked DRMI, lost DRMI lock, but guardian did not go to down, continued to attempt to lock DRMI 17:00 UTC Kyle and Gerardo at mid Y 17:16 UTC Searching for ALS DIFF IR resonance by hand 17:18 UTC Done searching for IR resonance by hand, the peak went by on the strip tool, but guardian says it is moving on. I guess it caught it. 17:32 UTC Back to Low Noise. Seismic 0.03-0.1 Hz band is around 10^-1 um/s. 17:38 UTC Back to Observing. Range is low ~ 54 MPc. 17:43 UTC Kyle reports seeing a helicopter flying low over mid Y about 5 minutes ago. Range seems to be rising as seismic comes down. 18:52 UTC Kyle and Gerardo back. Ion pump has been replaced. Aux cart attached to new ion pump left running. They left mid Y at 18:40 UTC.
This is a budget for the DARM noise on 2015-09-12. Compared to the previous budget, the only major difference (that we know of) is the elimination of the excess 45 MHz AM.
For the DARM estimation I use the DCPD sum, which is calibrated into milliamps. Calibration into freerunning displacement requires the DARM OLTF (I used the CalSVN measurement from 2015-09-10) and an estimate of the optical plant. I use Kiwamu's measurement of the optical plant along with Sudarshan's corrections to arrive at an optical gain of 3.39 mA/pm, and a DARM pole of 349 Hz.
Quantum noise is taken from GWINC, using the same parameters as previously: recycling gain 37 W/W, 107 kW circulating power, 755 kW BS power, 87 % quantum efficiency, 14 % other readout losses. The dark noise is taken from a measurement made several months ago.
The null stream is higher than this shot noise prediction by about 7 %. This was the case previously, but this is almost certainly a coincidence, since the previous budget used data with two stages of whitening on (which we already know has a small calibration discrepancy). Previously I had used pcal directly to estimate the optical plant (without the necessary pcal correction factors), so this could explain part of the issue.
The attachment shows sum, null, and quantum noise curves, along with 20.0 mA shot noise. It could be that we are missing some readout loss, or the GWINC curve could still require some more tuning of the power.
The seismic and Newtonian curves are the vanilla ones from GWINC, as is the coating Brownian noise.
DAC noise (using Peter's model) is propagated forward to displacement noise at the quads. The elevated noise from Chris's pringle measurements is not yet included.
Intensity noise and frequency noise are as described previously. Intensity noise was measured with the ISS outer loop off. I injected into the ISS inner loop error point and used the outer loop's out-of-loop PD array (sum 5–8) to estimate the RIN. Frequency noise is a lower bound, including only the noise of the first stages of the CARM loop (PD dark noise, PD shot noise, and electronics noise of the CARM board).
LSC control noises include PRCL, MICH, and SRCL. ASC control noises include dHard, BS, and SR2 loops (these are our high-bandwidth loops). I injected broadband noise for each of these and estimated the coupling via ratio of excess power (not coherent TFs).
Compared to the last budget, the SRCL coupling has increased slightly (see attached injection spectra), which is responsible for the slightly increased LSC noise around 50 Hz. Unclear whether this was a short-term excursion or a long-term drift of the SRCL coupling.
Squeeze film damping and residual gas are calculated as before. End station pressures are assumed to be 1×10−8 torr of molecular hydrogen. For residual gas, I have used 5×10−9 torr of molecular hydrogen.
IMC input jtter was measured by driving the IMC PZT in pitch and in yaw. On Keita's suggestion, as a sensor for the PZT motion I use WFS B I for pitch and WFS A I for yaw. Again the coupling is estimated by excess power ratio. Jitter at the OMC is not yet incorporated.
Based on Bruco, there is still some coherence with the endstation QPDs around 78 Hz, as Gabriele has already noted. But other than that, no channel seems to jump out as the culprit for the noise from 50 Hz to 150 Hz. So either we are looking at some nonlinear process, or we are seeing some kind of displacement noise that is not captured by the digital system.
In the above entry I have added a zoomed plot in the region where we have unexplained noise.
In this entry I have attached the residual of the measured noise and the budgeted noise, in two versions. The first uses the total noise trace as given in the above entry. The second uses a total noise trace in which the quantum noise has been increased by 7 %, since (as mentioned above) the quantum noise curve seems to underestimate the shot noise by this amount.
[Of course, simply scaling the overall quantum noise is not correct, but since the radiation pressure part of the curve is buried below the total noise, it doesn't matter too much here.]
Injection times, 2015-09-12 Z: