Scott L. Ed P. Chris S.

Today we were able to clean 50 meters of tube moving towards X-1-5 double doors. The afternoon was spent traveling to town to fill the diesel tank and fueling the support vehicles.
Continuous monitoring of beam tube pressures by the control room operator.  

07:00 - 08:00 Cris M. in LVEA
~ 8:00 Reverse Osmosis in alarm, notified John W.
08:10 Hugh transitioning BS ISI stage 2 to isolated, running measurements
08:54 Ed to H2 building
09:06 - 9:11 Sudarshan at end Y blocking PCAL beam
09:22 Elli to LVEA to align ITMX and ITMY spool cameras
09:56 Elli back
11:08 Jodi to mid station
12:30 Jodi back
13:33 Ed and Peter to H2 PSL enclosure to work on Livingston PMC
15:58 Ed and Peter done

SudarshanK and RickS

Since first noticed during the LLO Yarm commissioning effort two weeks ago, we have been chasing a subtle problem that is confusing the Pcal calibrations.

Today, we finally found the source of the problem, and fortunately it has a relatively simple remedy.

The Pcal transmitter modules use uncoated, 3 deg., BK7 glass wedges to generate the sample beams for the Optical Follower Servo (OFS) and Transmitter (Tx) Monitor photodetectors

The wedge is operated near Brewster's angle to reflect only about 0.1% for the OFS PD (first surface) and 0.5% for the Tx PD (second surface).  We have been observing slow (few to tens or even hundreds of seconds) transients in the OFS and Tx PD signals that have apparently resulted from fast, large-amplitude changes in the OFS signals, e.g. from changes in the loop gain or offset.

We have been making measurements to eliminate potential offenders and today we tracked the source of the slow variations we have been seeing to small, time-varying depolarization of the Pcal beams caused by the acousto-optic modulator (AOM) that generates the Pcal beams that are directed to the ETMs.  This depolarized light, while only a few tenths of a percent of the incident light, is in S-pol when incident on the wedge beamsplitter, so about 10% of it is reflected to the photodetectors.  By installing a polarizing beamsplitter cube downstream of the AOM the slow transients in the Pcal beams were eliminated (or at least significantly reduced - we will investigate further).

The attachment below shows StripTool trends of the H1 Y-end Pcal OFS, Tx, Rx (receiver), and AOM drive monitor signals when the AOM drive level was changed (left half of figure, up to 450 seconds), when the loop switch was closed but the loop had not locked (490-910 dseconds), then when the loop locked after the offset was adjusted (910 seconds).  Note that the OFS PD signal (blue trace) goes flat when the loop closes, but the drifts in the OFS PD signal due to the AOM-generated S-pol light is imposed on the other "out of loop" signals.

The second attachment is an oscilloscope screen photo taken after installing a polarizing beamsplitter (PBS) cube downstream of the AOM.  The green trace is from a temporary photodetector that was installed in the beam reflected from the PBS (s-pol light), yellow is the OFS PD, the blue trace is the Tx PD, and the magenta trace is the monitor of the drive to the AOM modulation input (modulates the amplitude of the 80 MHz RF).  Note that the s-pol light exhibits the long (10s of seconds) transient observed at LLO and at LHO.

The third attachment is a photo of the lab setup with the PBS and temporary photodetectors installed.

We have packaged two complete PBS hardware setups along with beam dumps for the reflected beams.  They will go out to LLO to Shivaraj's attention via overnight delivery tomorrow for installation ASAP at LLO.  We plan to install an identical setup at LHO Yend later today or tomorrow morning.  If we find that this is a suitable remedy, we will procure more hardware for LHO Xarm and the 3rd. interferometer.

One of the working standards is currently at LLO.  The Pcal calibrations will have to be repeated.  This should take about two hours at each end station.  We plan to repeat the LHO Yend calibration tomorrow.

Guardian state indices have now been added to all of the requestable, and some of the "static" non-requestable, states in the seismic guardians.

The SEI guardian state index mapping can be found in LLO Log 15748.

The indices have been added to the code, and the code has been committed to the USERAPPS SVN, but the SEI nodes have not been restarted/reloaded yet, to pull the change.  This will happen during maintenance period tomorrow.

J. Kissel, A. Staley, S. Dwyer

While investigating the MICH OLTF on resonance, I noticed that the bounce mode notch in the oplev damping loop (ellip("BandStop",4,1.5,40,19.3,19.9)) was not the same as the M2_LOCKING_L notch FM10 (ellip("BandStop",4,1.5,40,17.7,17.9)). However, the roll modes were idenitcal (ellip("BandStop",4,3,40,25.7,26.4)). Using the oplev, I measured the bounce mode to be at 19.636 Hz and the roll mode to be at 26.211 Hz. So, I added another Bounce Roll M2 lock filter FM8 that matches the filter in the oplev damping. Using FM8 instead of FM10, seemed to have improved the MICH OLTF. However, later when we lost locked and tried re-acquiring DRMI the bounce mode rung up and we immediatly lost lock. I have reverted back to the original FM10, and we can now lock DRMI again. Unclear ...

Summary--Changing Guardian code to not switch the GS13 Gains during the Isolation removes the RZ twist seen previously.

See attached, At T=1 the guardian zeros the CPS offset and Channel6, the GS13 SWSTAT switches the GS13s to low gain w/o whitening.  This is a coding error we are fixing, it should be going to high gain w/o whitening.  On channel9 at T=1 the DAMP path glitches which is seen on the RZ_INMON. After the Z loop closes, which glitches, the RZ comes on shortly thereafter and the big excursion on RZ_OUT16 swings the table.  Fortunately, the WFS loops now keeps the DRMI riding through the switch.  At T=2, the Isolation is taken down, evidenced by the OUT16s going quiet and the INMONs going noisy.  Immediately I switch the GS13 to whitened low gain and again this glitches the damping out and the ISO INMONs.  At T=3, Guardian does the isolation this time with switching the GS13s and no glitches are seen on the ISO INMONs nor OUT16s, everyone is happy.  I don't know from what the VHF spikes seen on the GS13INF (Channel7, lower right) originate but they do correlate to a transition to reduced inertial sensor noise... maybe it is the complettion of the Isolation and the BOOST filters.  At T=4, the guardian test is repeated with similar results.

The ITMX and ITMY gige cameras have been repositioned so that the IR beam is centered on the camera and so that the image is focused in IR.  The lens apperature is fully open for both cameras.  The lens is zoomed the out as far as possible while still being able to focus on the IR beam.  Note that the camera mounts are very sensitive to being touched; if the camera, camera housing or camera cable is touched even gently, the location of the spot position on the camera will change.

Laser:
Output power ~ 32 W (should be ~ 30 W)
Watchdog is active
No warnings other than 'VB program online'

PMC:
Locked for ~ 5 days, 22 hours
Reflected power is ~ 8.7% of transmitted power (should be 7% or less)

FSS:
Locked for ~ 11 hours
Trans PD threshold is .4 V (should be .9 V?)

ISS:
Diffracted power is ~ 7.6% (should be 7%)
Last saturation event was ~ 13 hours ago

SEI:
Hugh running measurements on beam splitter ISI for ~ 1 hour

3IFO:
Jodi and Gary may be in mid stations this afternoon

CDS:
tomorrow: Vacuum cabling and ESD investigations

Here is what I did tonight:

- I found an input matrix for sensing PRC2 (PR2) in full lock, by requiring that it doesn't see CHARD:
    0.5*(REFL 9I A + REFL 9I B) + 0.83*(REFL 45I A + REFL 45I B) --> PRC2_P (PR2)
  This was the crucial missing piece that was hurting us for CHARD engagement. Now PR2 is servoed faster than CHARD using this CHARD-independent error signal.
  The REFL I input matrix is therefore
        #REFL 9I A - REFL 9I B ->  INP1 (IM4)
        ezca['ASC-INMATRIX_P_1_9'] = 1
        ezca['ASC-INMATRIX_P_1_13'] = -1
        ezca['ASC-INMATRIX_Y_1_9'] = 1
        ezca['ASC-INMATRIX_Y_1_13'] = -1
        # 0.5*(REFL 9I A + REFL 9I B) + 0.83*(REFL 45I A + REFL 45I B) --> PRC2_P (PR2)
        ezca['ASC-INMATRIX_P_4_9'] = 0.5
        ezca['ASC-INMATRIX_P_4_13'] = 0.5
        ezca['ASC-INMATRIX_P_4_11'] = 0.83
        ezca['ASC-INMATRIX_P_4_15'] = 0.83
        ezca['ASC-INMATRIX_Y_4_9'] = 0.5
        ezca['ASC-INMATRIX_Y_4_13'] = 0.5
        ezca['ASC-INMATRIX_Y_4_11'] = 0.83
        ezca['ASC-INMATRIX_Y_4_15'] = 0.83
        #REFL 9I A +REFL 9I B -> CHARD
        ezca['ASC-INMATRIX_P_10_9'] = 1
        ezca['ASC-INMATRIX_P_10_13'] = 1
        ezca['ASC-INMATRIX_Y_10_9'] = 1
        ezca['ASC-INMATRIX_Y_10_13'] = 1
- Also set the PRC2 input matrix back to (1,0,0,0) in the DRMI ASC setup to make sure the DRMI angular control guardian still works.
- Noticed that the PR pointing loop output matrix denationalization for yesterday was not in the guardian. (These were found in DRMI configuration, by requiring that none of the WFS loops need to do any work for pointing changes). Added the following to the DRMI guardian ASC setup:
        # PRC1 -> PRM
        ezca['ASC-OUTMATRIX_P_1_3'] = 1
        ezca['ASC-OUTMATRIX_P_2_3'] = -0.160 # to remove coupling to PR2
        ezca['ASC-OUTMATRIX_P_9_3'] = -0.456 # to remove coupling to SRM
        ezca['ASC-OUTMATRIX_P_10_3'] = 0.060 # to remove coupling to SR2
        ezca['ASC-OUTMATRIX_P_15_3'] = 0.044 # to remove coupling to IM4
        ezca['ASC-OUTMATRIX_Y_1_3'] = 1
        ezca['ASC-OUTMATRIX_Y_2_3'] =  0.170 # to remove coupling to PR2
        ezca['ASC-OUTMATRIX_Y_9_3'] =  0.284 # to remove coupling to SRM
        ezca['ASC-OUTMATRIX_Y_10_3'] = 0.040 # to remove coupling to SR2
        ezca['ASC-OUTMATRIX_Y_15_3'] = 0.008 # to remove coupling to IM4

 - With this setting I was able to engage the following WFS:
  - MICH was never turned off
  - DHARD was already on
  - INP1
  - PRC2
  - CHARD
  - PRC1
  - SRC2
  - SRC1
- The order above is what I expect to work best for engaging. 
- However, the PRC2 loop works so well, and has high enough gain, that directly engaging all 12 loops (6 pit and 6 yaw) at the same time seems to work.
- I tested that twice - with a rather bad initial alignment.

- I left the IFO locked in the engage ASC state - I want to see how much it drifts over night.

When I came in today the ALS kept killing the MC, presumably to a bad alignment. So I tried to take the IFO through an initial alignment using the ISC_DOF guardian.
 - I had trouble with YARM_ALIGN_IR - The REFL_WFS_CENTERING didn't work, but the guardian didn't notice and pulled the ITMY into nirvana.
 - Also, SRM align didn't lock properly.
I manually aligned the stuff that didn't work, and got the back to full lock.

model restarts logged for Sat 07/Mar/2015
2015_03_07 00:26 h1fw1

one unexpected restart.

During the last stretches of full lock,  powers have been remarkably stable, but still showing some residual fluctuation. Here I'm analyzing the correlation of those power fluctuation with residual angular motion. I compute the best linear combination of all the ASC error signals (WFS and QPD) and their squared value, to predict the power fluctuation. The results fit very well the power time variation. One general comment is that the largest contributions to the power fluctuations come from pitch residual motions. The MATLAB code is attached.

AS RF90

The first plot shows the real AS_RF90 power (in blue) and the best fit obtained from the ASC signals (red trace). The second plot is a zoom in time. Using the same algorithm that I developed in the non stationary noise studies, I obtain a ranking of the most relevant ASC signals. The full result in shown in the third figure. The most important channels are AS_B_RF36_I_PIT, AS_A_RF45_I_PIT and REFL_B_DC_PIT.  The first signal is used for SRM, the second one is the other quadrature of the DETM signal (maybe the demodulation phase is not well tuned for DETM?) . One important remark is that most of the dependencies are with the linear channels and not with the squared values, meaning that the actual alignment working point is not the optimal one (in that case one would expect only quadratic dependencies).

AS LF

As above, the plots 4 and 5 show the fit to the power fluctuation. Plot 6 gives the channel ranking. In this case the largest contributions come from: AS_A_RF45_I_PIT^2, AS_A_RF45_Q_PIT and AS_B_RF36_I_PIT.  In this case we have mostly a quadratic contribution, meaning that the alignment is better for this particular port.

POP LF

As above,  plot 7 shows the fit to the power fluctuation. Plot 8 gives the channel ranking. In this case the largest contributions come from POP_A_PIT and AS_B_RF36_PIT. The first channel is highly correlated with the PRM motion, and the second one is used for SRM.

XARM

As above, plot 9 shows the fit to the power fluctuation. Plot 10 gives the channel ranking. The largest contribution is the same as for POP LF: POP_A_PIT. This seems to indicate that most of the power fluctuations in the arms comes from fluctuations in the PRC.

Stefan, Dan, Gabriele, Evan

Summary

Tonight we worked on making the ASC more robust, and the signals more decoupled.

Details

We are able to close the same loops as last night (LHO#17124), with the same issues:

1. In order to engage cETM or PR2, these optics must be manually adjusted beforehand to bring the error signals close to zero.
2. The PR2 and PRM loops seem to make the sideband buildups less stable.
3. No matter what loops we close or what optics we touch, we cannot seem to improve the recycling gain (as measured by the arm buildups). Over the last few days, we seem to get about 26 W/W. However, we can seemingly improve the visibility (mainly by adjusting cETM) as seen in LSC-REFL_A_LF. So it seems that we can make more power enter the interferometer, but we cannot get it to circulate in the arms.

The goal for tonight was mainly to try fixing (1) and (2). For (2), Stefan worked on decoupling the POPA → PRM loop by applying a pitch step to PRC1, and then recording the resulting steps in the control signals of INP1, PRC1, PRC2, SRC1, and SRC2. These step sizes (normalized by the step in PRC1) are then put into the output matrix. This gave a noticeable decoupling of the loops.

For (1), we noticed that for every 0.1 µrad step in ETMX (pitch or yaw), PR2 required a step of 0.1 µrad in the same direction. Therefore, we are now experimenting with a new loop: REFLA9I + REFLB9I → ETMX + ETMY + 2×PR2. The factor of 2 was established empirically using the same method as above. The full cETM drive vector is −5.2 for PRM, 2.0 for PR2, 1 for ETMX, 1 for ETMY, −9.5 for SRM, 1.8 for SR2, and −0.15 for IM4, but we have not yet employed it successfully. It causes a very sudden unlock of the interferometer, whose cause we haven't pinned down yet.

Some notes about housekeeping:

1. We rephased REFLA9I and REFLB9I by driving an 8 Hz line into PR2 pitch and putting all of that signal into the I phase.
2. The turning off of the DRMI WFS and the top-stage suspension offloading is now handled in PREP_TR_CARM instead of CARM_ON_TR.

Evan, Gabriele

We plugged the LSC-EXTRA_AO_2_EXC output into the excitation 1 of the MC board. In this way we could inject a 100 Hz line (100 counts amplitude) into the laser frequency and check the coupling to the IMC transmitted power.

As described in 17129, we used a script to compute in real time the transfer function from our laser frequency excitation to the IMC transmitted RIN. The attached plot shows that we got the expected linear dependency of the TF on the IMC locking servo. Unfortunately we don't have enough range in the MC offset to go to zero. We need about -15 V to go to zero coupling.

For the moment being, we left the offset adjusted to -10 V. This reduces the coupling of frequency to intensity noise by a factor 2.6

This morning I measured the IMC angular sensing matrix, using the same script we used last night in full lock. The script is attached. I added to the list of signal the IMC transmitted power, just to see how large the coupling to RIN is. It turns out that for all angular lines, there is always good coherence of IMC transmission with all angular lines. This seems to indicate that the IMC is not very well aligned in the present configuration, intoducing an excess of RIN due to input jitter.

1109796196 1109796316 H1:SUS-MC1_M3_LOCK_P_EXC
1109796339 1109796459 H1:SUS-MC2_M3_LOCK_P_EXC
1109796482 1109796602 H1:SUS-MC3_M3_LOCK_P_EXC
1109796623 1109796743 H1:IMC-PZT_PIT_EXC

1109796832 1109796952 H1:SUS-MC1_M3_LOCK_P_EXC
1109796975 1109797095 H1:SUS-MC2_M3_LOCK_P_EXC
1109797117 1109797238 H1:SUS-MC3_M3_LOCK_P_EXC
1109797259 1109797379 H1:IMC-PZT_YAW_EXC

Pitch sensing matrix (abs)

Pitch coherence matrix

Pitch sensing matrix (complex)

Yaw sensing matrix (abs)

Yaw coherence matrix

Yaw sensing matrix (complex)

I added aline at 20 Hz on MC2_M3_L, amplitude 1000 counts. The line is very well visible in the IMC transmitted power. I used a modified version of my phase tuning script (attached) to plot a time series of the transfer function PSL-ISS_SECONDLOOP_SUM14_REL_OUT_DQ over SUS-MC2_M3_LOCK_L_EXC, which should be zero when the IMC is locked on top of the resonace.

However, as visible in the attached plot (both real and imaginary part of the TF over time, it's ot possible to zero the transfer function, but basically only to rotate the phase. My guess is that the longitudinal actuation has a too large coupling to angles, that in turn generate intensity noise. We should repeat this measurement with a frequency line.

model restarts logged for Fri 06/Mar/2015
2015_03_06 02:14 h1fw1
2015_03_06 15:43 h1fw0

both unexpected restarts.

Sudarshan, Rick

We saw some transient in Pcal laser power in our recent observation. The power variation in these transients are about 20% at its max. We are working on to find where the issue is. Until then, donot trust Pcal calibartion to more than 50% (over-estimation) of what it reports.