TITLE: 03/07 Eve Shift: 00:00-08:00 UTC (16:00-00:00 PST), all times posted in UTC
STATE of H1: Observing at 63Mpc
OUTGOING OPERATOR: Ed
CURRENT ENVIRONMENT:
    Wind: 16mph Gusts, 13mph 5min avg
    Primary useism: 0.05 μm/s
    Secondary useism: 0.21 μm/s
QUICK SUMMARY: Breezy but otherwise calm. 11 hours at 65Mpc.

TITLE: 03/07 Eve Shift: 00:00-08:00 UTC (16:00-00:00 PST), all times posted in UTC
STATE of H1: Observing at 64Mpc
INCOMING OPERATOR: TJ
SHIFT SUMMARY:

H1 locked for 11hr40min
LOG:

ALL TIMES IN UTC

2:23 Big LF glitch in DARM. waiting for h(t) DMT Omega to update for more details

5:10 Low temp in PSL Verbal alert

5:39 Low temp in PSL Verbal alert

H1 Observing, 64Mpc. Locked for 7hr40min. winds trending below 20mph. One LF glitch.

With a nudge from peterF and mevans, I checked to see how hard it might be to do some time-domain subtraction of the jitter in H1 DARM.  This is similar to what Sheila (alog 34223) and Keita (alog 33650) have done, but now it's in the time domain so that we could actually clean up DARM before sending it to our analysis pipelines.

The punchline: It's pretty easy.  I got pretty good feedforward subtraction (close to matching what Sheila and Keita got with freq-domain subtraction) without too much effort. 

Next steps:  See if the filters are good for times other than the training time, or if they must be re-calculated often (tomorrow).  Implement in GDS before the data goes to the analysis pipelines (farther future?). 

I was finding it difficult to calculate effective Wiener filters with so many lines in the data, since the Wiener filter calculation is just minimizing the RMS of the residual between a desired channel (eg. DARM) and a witness (eg. IMC WFS for jitter).  So, I first removed the calibration lines and most of the 60Hz line.  See the first attached figure for the difference between the original DARM spectrum and my line-subtracted DARM spectrum.  This is "raw" CAL-DELTAL_EXTERNAL, so the y-axis is not in true meters. 

I did not need to use any emphasis filters to reshape DARM or the witnesses for the line removal portion of this work.  The lines are so clear in these witnesses that they don't need any help.  I calculated the Wiener filters for each of the following channels separately, and calculated their estimated contribution to DARM individually, then subtracted all of them at once.  H1:CAL-PCALY_EXC_SUM_DQ has information about the 7Hz line, the middle line in the 36Hz group, the 332Hz line and the 1080Hz line.  H1:LSC-CAL_LINE_SUM_DQ has information about the highest frequency line in the 36Hz group.  Both of those are saved at 16kHz, so required no extra signal processing.  I used H1:SUS-ETMY_L3_CAL_LINE_OUT_DQ for the lowest frequency of the 36Hz group, and H1:PEM-CS_MAINSMON_EBAY_1_DQ for the 60Hz power lines.  Both of these channels are saved slower (ETMY cal at 512Hz and MainsMon at 1kHz), but since they are very clean signals, I felt comfortable interpolating them up to 16kHz.  So, these channels were interpolated using Matlab's spline function before calculating their Wiener filters. Robert or Anamaria may have thoughts on this, but I only used one power line monitor, and only at the corner station for the 60Hz line witness.  I need to re-look at Anamaria's eLIGO 60Hz paper to see what the magical combination of witnesses was back then.

Once I removed the calibration lines, I roughly whitened the DARM spectrum, and calculated filters for IMC WFS A and B, pit and yaw, as well as all 3 bullseye degrees of freedom.  Unfortunately, these are only saved at 2kHz, so I first had to downsample DARM.  If we really want to use offline data to do this kind of subtraction, we may need to save these channels at higher data rates.  See the second attached figure for the difference between the line-cleaned DARM and the line-and-jitter-cleaned DARM spectrum. You can see that I'm injecting a teeny bit of noise in, below 9Hz.  I haven't tried adjusting my emphasis filter (so far just roughly whitening DARM) to minimize this, so it's possible that this can be avoided.  It's interesting to note that the IMC WFS get much of the jitter noise removed around these broad peaks, but it requires the inclusion of the bullseye detector channels to really get the whole jitter floor down. 

Just because it's even more striking when it's all put together, see the third attachment for the difference between the original DARM spectrum and the line-and-jitter-cleaned DARM spectrum.

Sheila, Kiwamu,

This is a followup/update on the beam pointing jitter measurement (34112). We simulated beam pointing jitter with FINESSE.

If we believe the simulation results, we can draw two conclusions:

• The couplings are from differential ITM
• The ITMs are misaligned by several 10 nrad from their optimum. 

The plot below shows a preliminary result of the simulation with H1's O2 specific parameters.

Some details of the simulation can be found in T1700080 (although the results in the document are specifically for a full power interferometer). Here is a brief list of remarks.

• In order to get a f-shape rise above several 10 Hz, the coupling has to be through differential ITM.
• To get the measured level of couplings, the ITMs have to be misaligned such that Delta Theta_IX = Delta Theta_IY = 10-40 nrad.
  • Note that in T0900142, the test masses are tilted by 1 nrad, an order of magnitude smaller than the ones shown above.
• With such ITM misalignment, jitter components which act as lateral beam shift at IM4 (or translation) couple the most to DARM above 40 Hz.

I think that the excitement found in the power plots and the chiller plots have been best described in Jason's alog

See CSWG log 11204

TITLE: 03/06 Eve Shift: 00:00-08:00 UTC (16:00-00:00 PST), all times posted in UTC
STATE of H1: Observing at 64Mpc
OUTGOING OPERATOR: Travis subbing for Corey
CURRENT ENVIRONMENT:
    Wind: 32mph Gusts, 27mph 5min avg
    Primary useism: 0.22 μm/s
    Secondary useism: 0.25 μm/s 
QUICK SUMMARY:  

Travis reports nothing notable for handoff

TITLE: 03/06 Day Shift: 16:00-00:00 UTC (08:00-16:00 PST), all times posted in UTC
STATE of H1: Observing at 63Mpc
INCOMING OPERATOR: Ed
SHIFT SUMMARY:  Observing for the entire shift, except for the Commissioning break from 18:00-22:09 UTC.  Windy all day (~30 mph), but not effecting locking.
LOG:

16:05 Bubba to MX

18:00 Start Commissioning period

18:06 Kissel starting calibration sweeps

18:35 Kissel done, Jenne starting arm cavity scans

18:51 Richard to MX for fire panel work

19:19 Richard done

21:20 Gerardo to EY

21:38 Gerardo back

22:09 Commissioning period end, back to Observing

Vaishali, Kiwamu,

We have implemented the calibration coefficients for the bullseye sensor in the front end today. The PIT, YAW and WID signals are now calibrated in fractional amplitudes. They are defined as

(fractional amplitude) = | HOM amplitude (e.g. E_{10} )| / | E_{00} |

Using this calibration, we virtually propagated pointing jitter (a.k.a PIT and YAW) from the bullseye to the IMC WFSs. We were able to get somewhat quantitative agreement with the measured jitter spectra there for the PIT and YAW degrees of freedom.

Next things to do:

• measure/check the IMC WFS calibration.
• measure the pointing jitter coupling transfer functions (again).

[New calibration settings]

Since Vaishali is writing up a document describing the calibration method, we skip the explanation here and just show the results.

• PIT_GAIN = 1.0
• YAW_GAIN = 1.1
• WID_GAIN = 0.54

Once they were set, we then accepted the SDF, although they are NOT monitored.

[Some noise spectra]

Below, the calibrated spectra are shown in the lower right panel.

The jitter level was about 3x10^-5 Hz^-1/2 for PIT, YAW and WID at around 100 Hz. If these fields go through the PMC, their amplitude should be suppressed by a factor of 63 in amplitude (T0900649-v4). So their level should be something like 5x10^-7 Hz^-1/2 when they arrive at the IMC. By the way, for some reason the intensity fluctuation and beam size jitter are positively correlated: as the laser power becomes larger, the beam size becomes larger at the same time.

Here are plots showing the noise projection for the IMC WFS which qualitatively agree with what Daniel has measured in the past using the DBB QPDs instead (31631) -- IMC WFS noise are superposition of acoustic peaks and HPO jitter.

Finally, if one let these jitter fields propagate through the IMC, they should get an attenuation of about 3x10^-3 in their amplitudes. Therefore the amplitudes of the HOMs after the IMC should be roughly 1.5x10^-9 Hz^-1/2. This number is consistent with what Sheila has estimated for pointing jitter at IM4 (34112). However, a funny thing is that, in order to explain a high coherence of 0.1 for pitch when the interferometer is locked in low noise (see for example, blue curves in the first attachment in 34502), the coupling must currently be about 1x10^-11 m / fractional amplitude, which is more than 10 times larger than what Sheila measured back in February (34112). What is going on?

Sheila Dwyer, Heather Fong

During today's commissioning period, we wanted to investigate how the fire pump from March 1 affected DARM (alog entry 34525). To do so, we injected signals into the HEPI loops: HAM1-HAM6, BS, ITMX, and ITMY, for x-, y-, and z-directions. The injected signal had a frequency bandwidth of approximately 10-40 Hz, which is similar to the bandwidth that was affected by the fire pump.

The data for the injected channels are available here. We did not observe coupling between the HEPI motion and the DARM spectrum. For an example, please see the attached figure "HEPI_excitation_ITMY_Z.png".

We also compared our z-direction results to the PEM injections from the beginning of O2, and we observed coupling in HAM1 that is consistent from what was observed before. The plot with which we compared can be found here, and our plot is attached to this log entry ("HEPI_excitation_HAM1_Z.png").

Commissioners have wrapped up what they wanted to do today, so we are back to Observing.

Starting CP3 fill. LLCV enabled. LLCV set to manual control. LLCV set to 50% open. Fill completed in 2143 seconds. LLCV set back to 17.0% open.
Starting CP4 fill. LLCV enabled. LLCV set to manual control. LLCV set to 70% open. Fill completed in 1961 seconds. TC A did not register fill. LLCV set back to 39.0% open.

A major question for the H1 interferometer is whether one of our arm cavities has anomalously high absorption.  To look into this, Aidan asked me to run mode scans of the arm cavities immediately after lockloss, so that we can watch the frequency separation of the higher order modes change as the interferometer cools down.  Prep work for this was done during last week's commissioning window (alog 34512), following the technique that Kiwamu used in 2014 (LLO alog 13768).

Prior to breaking the lock, Sheila and I went to the LVEA and bypassed the /10 frequency difference divider in the ALS COMM electronics chain, such that I can scan several FSR. 

Travis got us back to ALS locked about 3.5 minutes after I broke the 9.5 hour lock, and the scan was started immediately after that.  After the first scan or two, I remembered to slightly misalign the IR beam into the cavities, which I did by moving PR3 0.8urad in pitch (which was enough that I started seeing the TEM10 mode in the arm cavity transmission powers).

The scan ran for 30 minutes. 

I put the FDD back in place, and Travis is now re-locking the IFO.  The attached screenshot shows my striptool of the sweeps, mostly just that indeed it went on for about 30min.

The PSL tripped 4 times last night, 3 of them in quick succession.  I've done some preliminary forensics, and all 4 were trips of the "Head 1-4 Flow" watchdog, with the Head 3 flow sensor being the culprit in every trip.

Trip #1: 5:07 UTC (21:07 PST 2017-3-5)

PSL tripped due to a trip of the "Head 1-4 Flow" interlock, caused by the flow reading from the Head 3 flow sensor dropping below the trip point.  While restarting the laser after this trip, it took a couple attempts to get the crystal chiller to start and stay running.  After the first attempt, the reported flow from the Head 3 flow sensor did not get above the trip point, so the chiller shut off.  The chiller restarted without issue on the 2nd attempt and the PSL restart went smooth from there.  Once the system was up and running, Jim had to reset the noise eater.

The first attachment shows the flow readings from the active laser head flow sensors along with the "Head 1-4 Flow" interlock signal.  The second attachment shows a slightly zoomed in view of the reading from the Head 3 flow sensor at the time of the trip.  As can be seen, the flow signal from Head 3 became ragged and dropped below the trip point, shutting down the system.

Trip #2: 5:47 UTC (21:47 PST 2017-3-5)

The PSL tripped again, for the same reason as trip #1.  This was also caused by the flow reading from the Head 3 flow sensor dropping below the trip point.  The third and fourth attachments show info for this trip, showing that again Head 3 was the cause.  The restart went smooth, and Jim had gone to the LVEA to reset the noise eater when...

Trip #3: 6:02 UTC (22:02 PST 2017-3-5)

The third PSL trip occurred while Jim was in the LVEA resetting the noise eater.  This trip was identical to the last 2.  The fifth and sixth attachments show that once again the flow reading from the Head 3 flow sensor was the cause of the trip; I also included the power output from the HPO in the fifth attachment.  As can be seen in the fifth attachment (PSL_Trip3_Head1-3_Flow_2017-3-6_06:02:06.png) there was another issue with restarting the crystal chiller; just over 30 seconds after restarting the chiller it shut down due to the flow reading from the Head 3 flow sensor.  The second chiller restart attempt was successful.  There were then zero issues in restarting the PSL; Jim had to reset the NPRO noise eater again once the rest of the system was up and running.

Trip #4: 7:59 UTC (23:59 PST 2017-3-5)

The fourth and final trip occurred just as TJ was taking over for the OWL shift.  Once again this was a trip of the "Head 1-4 Flow" interlock caused by the flow reading from the Head 3 flow sensor dropping below the trip point; this is shown in the final 2 attachments.  There were no issues restarting the PSL after this trip.  Once everything was up and running, TJ had to reset the noise eater twice and was then on to locking the IFO.

J. Kissel, E. Goetz

I've gathered our "bi-weekly" calibration suite of measurements to track the sensing function, and ensure all calibration is within reasonable uncertainty and to have corroborating evidence for a time-dependent detuning spring frequency & Q. Evan will post analysis results later.

The data have been saved and committed to:
/ligo/svncommon/CalSVN/aligocalibration/trunk/Runs/O2/H1/Measurements/SensingFunctionTFs
    2017-03-06_H1DARM_OLGTF_4to1200Hz_25min.xml
    2017-03-06_H1_PCAL2DARMTF_4to1200Hz_8min.xml

    2017-03-06_H1_PCAL2DARMTF_BB_5to1000Hz.xml

This suite, again for time-tracking, took ~35 minutes from 2017-03-06, 18:02 - 18:35 UTC.