The ITMX and ITMY RH SIM SUB_DEFOCUS and SURF_DEFOCUS filters
were all reporting infinite/NaN outputs. I reset the histories on these to remove these NaNs.
We restarted the psinject CW hardware injections on h1hwinj1 last Wed 11th to see how stable this system is now. All was well until yesterday (Thu) when it stopped running for no discernable reason. Here is a summary:
date | time | event |
wed 11 May | 14:11 PDT | start running psinject |
Tue 17 May | 08:32 - 08:38 PDT | stopped due to h1iscex restart, manually restarted |
Wed 18 May | 10:03 - 10:04 PDT | Jim manually restarted process |
Thr 19 May | 14:20 PDT | psinect stopped for no apparent reason |
We were not running under monit control so that a restart would be easy to spot. Now we have seen a restart, we'll put psinject back under monit control. Remember that the monit script smoothly ramps up the injection to prevent jarring.
Again, it looks like the drift down of the BRSY continues and the trend continues to ease albeit quite slowly. The latest trend extrapolation still puts the 15000ct crossover around the 3rd of June: The upper plot shows most of the BRSY life and the bottom I've zoomed in and eyeballed a line out to rebalance time.
SEI - Jim would like to start locked IFO to test some configurations.
Vac - Ion pump work at the diagonal.
Fac - Chiller showing up today for staging building.
These are the serial numbers for the in-service chillers: P325-18636-1 S/N 44801 (crystal chiller) P625-18637 S/N 44807 (diode chiller) The old chillers had serial numbers: P325-18636-1 S/N 44800 P625-18637 S/N 44806 When pulled from service the chillers had accumulated 30716 hours and 30790 hours respectively. To my knowledge this cannot (as yet) be reset from within TwinCAT, so the "new" chillers will have an artificially high number of hours displayed. To find the real number of operating hours, one needs to interrogate the chiller control panel. All 4 chillers have a date of manufacture of 07/2011.
TITLE: 05/20 Day Shift: 15:00-23:00 UTC (08:00-16:00 PST), all times posted in UTC
STATE of H1: Planned Engineering
OUTGOING OPERATOR: None
CURRENT ENVIRONMENT:
Wind: 10mph Gusts, 4mph 5min avg
Primary useism: 0.02 μm/s
Secondary useism: 0.18 μm/s
QUICK SUMMARY: Unlocked, environment seems calm.
Toward the beginning of the shift, the HAM2 ISI watchdog tripped (as well as all of the optics on this table). After a few hours of locklosses, went ahead and checked/restored all Input Mirrors (IMs) to their OSEM values before the HAM2ISI WD trip (per instructions here).
Some of the mirrors required big slider moves for recovery.
Have run an Initial Alignment on H1 and making attempts at locking. Attached are trends of the IM OSEMs.
[Corey, Jenne, Kiwamu]
The DC Readout state checks whether the ISS first loop is on or not before allowing the IFO to increase power. However, the state_word value for the ISS isn't matching what is in the guardian (it wants 32700), so it doesn't pass this check. Both the integrator and the autolocker are on, which are the primary things that I thought were required.
Previously, the run state would check the ISS, and if it was okay, check ASC convergence, and then turn off the SRC1 loops. However, since we're not passing the ISS check, I had Corey go to manual mode and request IncreasePower, which ended up skipping the SRC1 turn-off step. This caused our friend the SRM-runs-away lockloss.
So, we have modified the run state of DC readout a little bit so that the ISS and ASC convergence checks are separate if statements, not an if/elif combo. If the ASC is converged, it will turn off the SRC1 loops, regardless of what the ISS is up to. If both the ASC and ISS are happy, it'll return True like normal, and carry on.
Operators: If you are stuck at DC Readout and the guardian log says "ASC converged" but you're getting the message that "ISS first loop may be off", then you should check on the ISS first loop*. If it's okay, then go to manual, and select Increase power. Go back to auto, and select whatever state you were trying to get to.
* I think that the only check for ISS first loop at this state that really needs to happen is that the "First Loop" light in the small blue box under the strip chart on the PSL_ISS screen is green (sorry Jim). This indicates that the integrator is on. If it's yellow, click the "On+Int" button on the ISS manual mode sub-screen. Clicking the "On+Int" button to ensure it's on shouldn't cause any trouble, so you can do that if there's any ambiguity.
TITLE: 5/19 EVE Shift: 23:00-07:00UTC (16:00-00:00PDT), all times posted in UTC
STATE of H1: POP AIR PD swap was going on (Evan)
Outgoing Operator: TJ
Quick Summary:
Relanded the signal wires at the break out box, IP11 has been working (despite being valved out), but we were missing the signal for the current from pump A for some time, not properly landed wire.
Aidan+Alastair (on phone), Bryn, Nutsinee
We are starting to have the same nasty particulates problem that LLO was having up until late April (I assume the problem was somewhat solved there. See LLOalog25658 and many more alogs after this one by Alastair). CO2X is in a better shape (less particulate clumps) compared to CO2Y.
We're going to need a proper procedure on how to resolve this issue. This cannot, and will not, go on until O2.
Both CO2X and Y were restored after the inspection but I don't really like the idea of leaving chillers running after finding these particulates.
Also we don't use DI water for the chillers. We use the same water our PSL is using.
TITLE: 05/19 Day Shift: 15:00-23:00 UTC (08:00-16:00 PST), all times posted in UTC
STATE of H1: Planned Engineering
INCOMING OPERATOR: Corey
SHIFT SUMMARY: PSL shiller took up most of the morning, relocking now.
LOG:
With the LASER off for most of the day, it was a good time to do these long measurements.
See Fabrice's pdf in 8284 for a good discussion of the concept. Basically, one tilts HEPI and measures how much the ISI tilts. Then you drive HEPI vertically and see how much the ISI tilts. The ratio of these results is the magnitude of the correction for the ALIGN matrix, in this case, the Z to RX and Z to RY values. There might be some consistency of the phase giving you sign information but... That is the theory. In practice, my experience is the determination of the tilts is subjective and the measurements are very sensitive to other influences and tend to not be repeatable. Back in Jan2015, using the direct (above) determination method, I found the results somewhat inconclusive and there was not enough time to do more cf 16118.
This morning, I did the skulk about search rather than the direct calc method. I'd have to say still somewhat inconclusive.
More details on the measurement--the ISI Stage1 blends are set to as high as possible. This will use the relative sensors, the CPS, to lock the Stage1 of the ISI to Stage0 and so to HEPI, below a few 100 mHz. Jim made a new blend (Imp[rove]_Hi_Blend) that only uses the L4C on Stage1 so the T240 is completely out of the loop. All sensor correction is turned off. GS13 gains are switched to low. Any tilt produced by the drive will show up as X or Y on the low frequency sensitive T240.
Ideally, we'd see Transfer function magnitude and coherence go down but that does not always seem to be the case. The table below gives the decoupling factors used for the TF's color seen in the attached xml plots. The channel names are all the same except Ref# so they aren't shown to keep them out of the way. The Z to RY channels are ISI-ITMX_ST1_BLND_T240X_IN1/HPI-ITMX_ISO_Z_EXC; the RX plots look at the T240Y.
ZtoRX | ZtoRY | Color | Aves | Comments |
0 | .0015 | Brn Dash | 7-8 | Start values-used since Dec2014 |
.005 | .002 | Blue | 7-8 | RY better RX worse |
-.005 | .001 | Dark Brn | 8 | RY maybe better RX bad |
-.0025 | .0005 | Red | 8 | RY maybe Better RX Mag Improving Coh Bad |
.0025 | .00175 | Green | 8 | RY ehh RX Same |
.001 | .00125 | Lt Green | 9 | RY again about same RX Good Low Mag and Coh! |
.0005 | .00075 | Pink | 9 | RY sameo RX Mag & Coh bad again |
.0015 | .001 | Not Show | 8 | RY repeat was not same RX Indeterminate |
.001 | .00175 | Black | 11 | RY repeat of Green not as repeatable as hoped RX not as good as repeated Lt Grn |
I think these comments and the traces will help the reader understand the difficulty of being objective here...it is not obvious.
I've left these final (Black) value loaded. The RX magnitude is pretty good compared to the starting value and is pretty consistant with the repeat (Lt Grn) measurement although the coherence is not. The RY ... it's so not clear, I don't know, the mag looks better below 25mHz but the coherence...
XML in SeiSVN/seismic/HEPI/H1/ITMX/Misc/Tilt_Decoupling_HEPIZtoISIRXRY_20160519.xml
Similar to LLO a-log 26196 here we compare the broadband injections done using Pcal and Craig's time varying calibration model (26889). The attached plot show the comparison for the broadband injection done during January 2016 (a-log 24784). Except at known lines, such as 60 Hz harmonics, the measurement and model agree within their uncertainties. The code used to make this plot is added to calibration svn at aligocalibration/trunk/Runs/O1/H1/Scripts/PCAL/PcalGDSBroadbandComparison.m .
Last week, TJ created some guardian node to manage the configuration of the endstation ISI's, alog 27141. The intention is for these to be the primary interface for commissioners and operators in changing the Seismic configuration, depending on environment. I'm providing some guidance on when the different states should be used here.
The guardian code is still being worked on, configurations still need to be tested. Thats why there are fields with options, or without a full description of the state. The wind speeds are pretty much guesses, but the 10% and 90% microseism are indicated on with dashed lines on the .1 - .3 hz BLRMS on the seismic foms.
Wind speed | 0-10 mph | 10-25 mph | 25-35+ mph |
Microseism | |||
10% | Quite_90 No SC |
Quite_90 No SC
|
Some windy configuration |
10-90% |
Quite_90 No SC, or Quite 90 useism depending on microseism |
Quite90 useism Quite_90 BRS |
Quite_90 BRS Some windy config |
90% + |
45 No SC, 45 Useism 45 BRS SC depending on microseism |
Quite 90 BRS | There be dragons here. |
For your viewing pleasure, I translate between the states indicated by Jim in his table above and the actual name of the states in the Guardian Configuration nodes created by TJ (as Jim cites, see LHO aLOG 27141), and other Jargon that we'cve been tossing around in the SEI group: State Name Jim's Table Name Sensor Correction Description BLEND_Quite_90_SC_BRS Quite_90 BRS BRS-corrected STS GNDX to ST1X (or GNDY to ST1Y), "broad"-band filter (0.08 - 0.5 Hz) BLEND_Quite_90_SC_useim Quite 90 useism STS Alone, using "narrow"-band focused at the microseism (0.1-0.3 Hz) BLEND_Quite_90_SC_None Quite_90 No SC BLEND_45mHz_SC_BRS 45 BRS SC BLEND_45mHz_SC_useim 45 Useism Runs in the corner station (but there's no node) BLEND_45mHz_SC_None 45 No SC Before BRSs existed, this was the nominal configuration for the End Stations Note that I put "narrow" and "broad" band in quotes, because there's about 10 sensor correction filter options and none of them are really truly narrow or broad in their modelled / expected performance, and any real performance depends heavily on the ground motion. E.g. while one might expect to get great performance out of a broad-band sensor correction filter, if it's windy, that ideal performance may be completely washed out by other noise.
C. Cahillane I have looked at the actuation strengths to check the covariance between the actuation stages for each of the four measurements. The four measurements are the three calibration week actuation measurements on August 26, 28, and 29, and the Post O1 January 7 measurement. The covariance was calculated by taking the residuals of the fitted actuation strengths and comparing points between L1 and L2, L1 and L3, and L2 and L3 at the same frequency for each day's measurement. The different dots in the covariance plots correspond to frequency. The dark blue is low frequency, and the light orange is high frequency. The covariance is on the same order as the overall variance for each actuation stage. Worse still, it is positively correlated, meaning we may have underestimated uncertainty by ignoring these terms. My results also indicate the measurements may be inconsistent between days. The August 28 measurement inexplicably displays no covariance between actuation stages, while all of the other measurements do. Jeff has tried looking at the optical gain fluctuations during each day's measurements, but this doesn't appear to be the cause.
J. Kissel, C. Cahillane, K. Izumi Some more details on this. For context about why the discovery of this ~1-2%-level, positive sqrt(covariance) matters, check out G1601084. In this aLOG, we're discussing scenario (4)(c), where because we use the same PCAL to DARM transfer function to provide the absolute calibration for each ith Stage EXC to DARM transfer function, there may potentially be covariance between the ith and jth stage's absolute calibration result. We would expect this to happen if the PCAL2DARM transfer function has some very-low-frequency, statistical-in-nature fluctuation over the course of the entire ~30 minute measurement that is larger than any similar fluctuation in the ithStage2DARM transfer function. Craig's results above show that for 3 out the 4 days, the above scenario must be true -- that there is some slow stochastic process that is influencing the PCAL2DARM transfer function that's larger than whatever process (either the same or different) in the ith Stage EXC 2 DARM transfer function, so there is covariance between our estimate of the statistical uncertainty in the ith and jith stage. In the right 4 columns of the 3x5 array of subplots, which show this covariance between the i and jth stage, and the 4 columns are the four measurement days. We can see by the color-coding of each frequency point [as Craig says, blue to orange is denoting low to high frequency] that it's *not* that, say, the low frequency points are more covariant than the high frequency points, or vice versa. What's most intriguing is that the covariance seems to be a function of measurement day, and the *least* covariant day is the 2015-08-28 measurement. *A* theory for why the covariance may vary from day to day is the optical gain of the IFO (i.e. the DARM part of the PCAL2DARM TF) is slowly fluctuating by different amplitudes from day to day based on the seismic environment. As a test of this theory, I've gathered the times of all of the measurements used to compose these data sets, and plotted the optical gain (as measured by the intra-cavity arm power, as reported by the average of the TRANSMON QPD SUM). From this time-series, I assume that the optical gain is fluctuating stochastically over the entire ~15 minute excitation period, take a histogram and report the standard deviation as a single-number, quantitative estimate of the amplitude of said fluctuation. This way, I can compare the amplitudes during the 4 measurement times. Sadly, though we do see a change in amplitude of optical gain fluctuations between measurement days -- as large as a factor of two between the best and worst -- the pattern of optical gain fluctuation amplitude does not match the covariance amplitude pattern from measurement day to measurement day. The covariance is comparable on days 1, 3, and 4, the optical gain fluctuations are smallest during the first day of measurements, and steadily get worse in time. Further, the amplitude of optical gain fluctuation is comparable between a given measurement day's PCAL2DARM transfer function and each ith stage of EXC2DARM transfer function. -------------- Details: The last page of the attachment shows my attempt to use several versions of the TRANSMON QPDs in order to arrive at the intra-cavity arm power. I'm quite sad that the different methods of calibration don't agree, so I've chose to use the one the bast agrees with about what we think the intra-cavity power should be. The channels I've used are H1:LSC-TR_X_NORM_OUT16 H1:LSC-TR_Y_NORM_OUT16 H1:ASC-X_TR_A_NSUM_OUT16 H1:ASC-X_TR_B_NSUM_OUT16 H1:ASC-Y_TR_A_NSUM_OUT16 H1:ASC-Y_TR_B_NSUM_OUT16 where I've used the EPICS channels sampled at 16 Hz, because I'm looking for minute-time-scale fluctuations (and there aren't so many versions stored as fast channels). All signals are normalized to the input power. For the LSC versions, at the advice of Jenne, I've taken the average of the two channels, and scaled the raw counts by the input power, PRM transmission, and arm-cavity gain, accounting for the beam splitter, Circulating Arm Cavity Power [ct] = 0.5 * (H1:LSC-TR_X_NORM_OUT16 + H1:LSC-TR_Y_NORM_OUT16) Circulating Arm Cavity Power [W] = (Circulating Arm Cavity Power [ct]) * 22.5 [W] * 0.03 [W/W] * 280 [W/W] * 0.5 [W/W] where I've trended the PSL input power to find the power into PRM, and 0.03 and 280 are well-known numbers. For the ASC version of the signals, I'd tried following the prescription from Dan's LHO aLOG (15431), for which Sheila suggest should be the same for all four QPDs: Circulating Arm Cavity Power [W] = H1:ASC-X_TR_A_NSUM_OUT16 [ct] * 1/3.98 [V/V] % QPD Whitening Gain * 1/2621440 [W_atQPDs / ct] % Electronics Chain Calibration (sans whitening gain) * 1/0.025 [W_afterETMS / W_atQPDs] % Amount of light exiting the ETM than gets to QPDs * 1/3.6e-6 [W_inArm / W_afterETMs] % ETM transmission where I've trended the whitening gain to see that it was no longer the 18 [dB(V/V)] = 7.9 [V/V] quoted in Dan's aLOG, but 12 [dB(V/V)] = 3.98 [V/V] at the time of the first three measurements. Since getting a consistent calibration for the intra-cavity power was not the point of this study, I moved on sticking with the metric that gave the answer closest to what was expected, as mentioned above. Script to produce these results: /ligo/svncommon/CalSVN/aligocalibration/trunk/Runs/O1/H1/Scripts/Uncertainty/showopticalgainfluctuation_during_actuationfunctionmeasts_O1.m
Sheila, Jenne, Evan
We have suspected for some time that the POPAIR B diode (which we use to extract POPAIR18 and POPAIR90 signals) is saturated in full lock, because these signals do not scale appropriately when increasing the PSL power. Now that we are seriously trying to use POPAIR90 to diagnose DRMI angular behavior during power up, we would instead like to use a signal that we can trust.
We tried a few different things today:
The last of these is the configuration that we have now. Whitening and dark offsets were set to give appropriate signal levels.
Careful examination will actually show that there appears to be a bit of saturation with this POPAIR45 signal as the power is increased. According to the POPAIR_A_LF channel, there is 83 mW (!) of dc power on this PD at 25 W. Probably we should be closing the POP beam diverter at this power, or else we need to swap some optics on the table.
Ideas for how to improve:
BBPD modifications are detailed in G1500595 by Koji. Page 5 shows the redlined drawing. Parts are available.
Unmodified BBPD S1200236 (which was POPAIR B) has been replaced with modified BBPD S1200239.
POP90 and POP18 are now supplied by this new diode.
actually the psinject process was still running on h1hwinj1, but it was not actually exciting the test point H1:CAL-PINJX_CW_EXC. I started monit, and then used monit to stop the running process and perform a clean start.