On the 15th I started a purge on one of the oldest, and least purged ISIs we have. The LN2 boil-off straight from the dewar was measuring ~ -40 td°C, while ambient air was ~ -5 td°C. This unit was taking longer than usual to dry out and wasn't reaching as low of levels as other containers had in the past. I think revisiting Assy. 1 soon would be helpful.

Installed four vertical actuators and four horizontal actuators with the help of Apollo today.  

There is a fault with the file server for /ligo this morning, the file system that contains user account home directories and applications for control room tools.  The file system will be unavailable for use until further notice.

Jeff B reported that the triple test stand watchdogs were being tripped when turning damping loops ON for the SR3 (HLTS) suspension.

This morning I've remotely logged-on to verify that the medm environment is correctly configured for a HLTS suspension (matrices & signs etc) i.e. the correct model is running and the correct environment has been BURT restored.

I also observed watchdogs tripping when enabling damping loops, specifically for the Pitch DOF.

Initially, I suspected some damping loop gain re-tuning maybe required, however the gain factor is consistent with other L1 & H1 HLTSs (-0.002).

Offsets were injected into each OSEM channel, T1, T2, T3, LF, RT, and SD, whilst monitoring the corresponding OSEM sensor readout in dataviewer. This indicated that there was a sign issues with T2 and T3 OSEMs. Temporarily changing the sign in the COIL OUTPUT FILTERS allowed damping loops to close, without tripping watchdogs. Therefore, I would recommend the assembly team double check T2 and T3 magnet polarity.

The IMC locked at undisturbed since 8:00 pm local time (2013/01/22)

[Kiwamu, Lisa, Keita, Matt, Giacomo]

At about UTC 2013-01-22 21:17:40, the IMC_REFL_DC channel suddenly showed and offset of about -8000 counts. In the past week it has been consistently reading between about 50 (IMC locked) and 450 counts (IMC unlocked). The attached plot IMC_REFL_DC_trend shows the trend in the last 12 days, witht he "jump" visible at the extreme right.

The signal still changes by about 400 counts between the IMC locked and unlocked statuses, and the RF part seems to be fine as the IMC locks regularly.

By inserting a breakout board at the input of the interface, we did verify that the power supply readings (+-18 V) were normal and that the offset came in fact from a bias of about 2.2 V on the REFL_DC readout. It does thus appear to be a problem with the head, or anyhow with something downstream of the interface.

It is curious to note how all the WFS channels show something going on at the exact same moment that the offset appeared (see attached plot IMC_REFL_DC_zoom). It is not clear if the problems seen on the WFS (see entry 5207) are related to this, if working on WFS debuging triggered the problem on the REFL_DC or if the two things are completely unrelated.

Summary: In an attempt to better understand the high magnetic coupling to the test masses, I measured magnetic fields at the locations of the 8 PUM magnets for an isolated reaction mass, the fields around ITMY in situ in BSC1, the moments of the 2 types of magnets, and searched for other possible coupling locations. While I have not been able to account for the high coupling, I did find that large parts of the UIMs and the PUM reaction mass were magnetic. I will try to measure the magnetic moments of these masses when I return.

Gradient scales at the PUM magnets

10 Hz

I used a fluxgate magnetometer and measured fields at the positions of the 8 PUM magnets (2 in each flag) for an isolated PUM reaction mass (Figure 1). In the table below the fields B1 and B2 were at sites 2.7 cm apart, close to the separation of magnets in the PUM flag. The measurements were used to estimate how well the paired magnets in each flag could cancel. The last column gives approximate gradients from the field differences for the two sites. The magnets were not in place for the measurements.

At the UL site, I also looked for higher gradients on a smaller scale (smaller movements) and found B/gradB values of 0.012 m for 6mm displacements right at the location of the flag magnet in the AOSEM. Of course the fluxgate’s ability to measure high gradients that extend only over short distances is limited by the scale of the sensor, which is 1.8 cm long. The magnets are 0.6 cm long so they could be subject to localized gradients that would have been averaged down by the fluxgate. For calculations of the magnetic coupling, I halved the 0.12m scale to 0.06 m because of the possibility of  more localized gradients.

100 Hz

I also made a couple of measurements at 100 Hz. For a 2.7 cm difference at UL, my estimate of B/gradB was 0.012 instead of the 0.021 at 10 Hz.

Measurements of fields and gradients around the quad in-situ inside BSC1

I measured injected magnetic fields around the ITMY quad in BSC1 (Figure 2) in order to make sure that the field at the ITM during my coupling measurements was similar to my estimates made with a magnetometer mounted outside BSC1, under its center. The average of injected field values inside was 37% lower than I estimated from my outside measurement. Thus the coupling  should be 37% higher (worse) than I previously reported.

The scale of gradients across the back face of the reaction mass chain were measured to be:

(B/gradB: 5.26 m, 1.10 m, 0.58 m, 0.53 m). These measurements were used to estimate how well magnets at different actuators could cancel. The arm cavity baffle was not installed for these measurements, and the BSC door was open.

Magnet moments

I measured the moments of one large (M0, L1) and one small (L2) magnets using a magnetometer at a distance large compared to the magnet size (Figure 3). For the small magnet I measured 0.013 J/T and for the large one, 0.717 J/T (or Am^2).

Estimates of residual moment from magnets

I used the fields measured at the positions of the 2 magnets in the PUM flags to estimate the residual moment for the pairs in each flag. I also used the larger scale gradients from the in-chamber measurements to estimate how well the different locations would cancel (e.g. how well will UL cancel LR). I estimated a residual moment of 1/3 of the moment of a single one of the 8 magnets, if all magnets were oriented properly, and a residual moment of about twice the single magnet value if one magnet was mis-oriented. For the estimates below of coupling at the UIM, I used the same gradient scales as I measured for the PUM.

Magnetized 304 steel in the UIM and PUM reaction mass

Figure 4 demonstrates that magnets stick to many parts of the UIMs and PUM reaction masses (L1, R1 and R2). Many of the large parts of the main and reaction UIM, as well as the reaction PUM are called out as 304 or 316 steel (shops choice). The relative permeability of cold worked 304 steel can be several orders of magnitude higher than 316 steel and is reported to reach 10 or 20. So magnets stick to many parts of the UIM and PUM reaction mass, and which parts are magnetic varies with the particular run of parts. About ½ of the volume of the PUM reaction mass in BSC1 is magnetic and at least ½ of the UIM and reaction UIM are as well.

Estimates of coupling from magnets and magnetized 304 steel

I used the measurements of magnet moments and estimates of  the residual moments along with the estimated field gradients and moment arms to estimate torques on the PUM and UIM. To compare to the angular motions I observed in the magnetic coupling measurements, I used Mark Barton’s functions for force or torque at the various quad levels to displacement, pitch and yaw of the test mass.

To estimate motion produced by the magnetized 304 steel, I assumed that the parts were magnetized by the large DC field from the earth (I used 3e-5T), giving them a magnetic moment and causing them to couple to the small AC fields: m = chi/mu0 BearthV. This does not include the magnetization from the permanent magnets, but estimates that I have made do not suggest that this could increase the moment by more than a factor of a couple. I used a relative permeability of 10, at the high end of permeability for 304 steel. I will try to measure the magnetic moment of parts directly when I return. I have also not examined the connectors.

Attached are plots of dust counts > .3 microns and > .5 microns in particles per cubic foot from approximately 5 PM Jan. 21 to 5 PM Jan. 22. Also attached are plots of the modes to show when they were running/acquiring data. Dust monitor 10 in the H1 PSL enclosure is still indicating a calibration failure. I did not plot the counts in the labs, since this IOC was rebooted and put 'nan' in the data. The IOC was timing out communicating with the dust monitors, but the Comtrol was still on. Rebooting the IOC seemed to fix it.

Giacomo, Lisa, Matt

While yesterday's "MC2-M2 as a low frequency offload path for M3" approach worked for signals well below 100mHz, it did little to save M3 from saturation due to signals around the microseism.  It was also not sufficient for ISI testing.

Through more measurements we found that the MC2 M2-M3 cross-over had a small region of stablity between 10 and 20Hz with only the old 100:1 filter engaged.  The maximum phase margin in this region was about 10dg, which really doesn't sound like enough to be reliable, so I started another filter design cycle.  The result is shown in the attached plot: unconditionally stable up to 20Hz, but not really optimized in terms of gain (we could have a lot more if we are willing to invert the plant features).  This was sufficient to keep the M3 drive RMS below 10k counts (at LOCK filter output, so 2.5k at the DAC), which is about a factor of 10 better than before and about a factor of 40 from saturation.

Giacomo, Keita, Lisa


 This entry is just for reference, now we have  this problem . 

Before the WFS decided to go in a permanent and irreversible bad state we experienced a moment of happiness 
as the calibration of the DC signal path was actually making sense (the difference between what we measure on the ADCs WFS SUM channels and what we expect is within 10%).


 Direct power measurement on IOT2 

- IMC unlocked

Total power in the WFS path combined (before the first BS IO_MCR_BS2, nominally 50/50: 2.25 mW)
Power in front of WFS_A (in reflection from BS2): 0.95 mW
Power transmitted after BS2: 1.30 mW
Power in front of WFS_B = 1 mW (after the second BS IO_MCR_BS3)
Power going to the camera (transmitted by the second BS IO_MCR_BS3): 0.27 mW

Power reflected by WFS_A: 0.13 mW
Power reflected by WFS_B: 0.1 mW

- IMC locked 
  Total power in the WFS path combined: 250 uW


 Expected number of ADC counts with IMC unlocked and WFS interface in high gain mode 

WFS DC transimpedance = 1000 V/A
Diode Responsivity = 0.8 A/W
ctsPerVolts = 1638  counts/V
DC WFS interface gain = 10

==> 13104 counts/mW 

Expected 
ctsWFS_A [counts] = 13104 [counts/mW] x (0.95-0.13)mW = 10745 counts
ctsWFS_B [counts] = 13104 [counts/mW] x (1.0-0.13)mW = 11794 counts 

 Measured number of ADC counts with IMC unlocked and WFS interface in high gain mode

WFS_A_SUM = 9272;
WFS_B_SUM = 11014;

Dark offset
WFS_A_SUM = -1215;
WFS_B_SUM = 25;

Measured
ctsWFS_A_SUM = 9272 - (-1215) = 10487 counts
ctsWFS_B_SUM = 11014 - 25 = 10989 counts 

 Laser Calibrator D1201258 on WFS_B, high gain mode 

4mW @ 980 nm 

Measured ctsWFS_B_SUM = 35037;

We don't know the exact responsivity of the Q3000 diodes at 980nm, but it should be between 0.5 and 0.6 A/W.

==> This measurement gives us 0.54 A/W 

Something is really only marginaly stable.

Attached is an example of low frequency oscillation of IMC WFS DC measured by using a DB15 break out board on the front panel of the DC interface (i.e. the breakout board is inserted between WFS head and the WFS DC interface).

This is easily triggered by disconnecting the WFS DC cable from the front panel of the WFS DC interface and then plugging in again. In this case it was WFSA, segment 4 (i.e. between pin1 and pin9, why is this not the segment 1, I don't know), and it's oscillating at about 100Hz. When this happens, all the quadrants show similar symptom, and it doesn't go away spontaneously. Sometimes this goes away by wiggling the cable or touching some of the pins on the breakout board.

It's not exactly easy to make things accidentally oscillate at such a low frequency, I'm suspicious about power problem, maybe weak grounding of big capacitors e.g. power capacitor on the WFS head.

Interestingly, when the breakout board is not there, when you look at the fact channels for WFS DC, the oscillation dies down on its own within 10 or 20 seconds for WFSA.

I was curious to do the same measurement for WFSB, and with the breakout board it does something similar (second picture, in this specific  case it was slower). Without breakout board, I didn't see any craziness. Also, when this was going on, the +-18V power that are passed through to the WFSB head were also showing something similar (third picture, in this case the WFSB was oscillating faster than when the second picture was taken) . I haven't measured WFSA power (see the entry attached to this one).

When they're not oscillating, things look OK-ish in that nothing is outrageously wrong.

The fourth BSC-ISI assembly was recently tested. It looks good. The testing report can be found at E1100297-V1 - aLIGO ISI-Unit 4 - LHO - Phase I report.pdf.

Manually moved the model running on the tripleteststand front-end to the HLTS 05 model, for some reason the script to change this failed almost completely.  Started the model, copied the master.05 file to master and restarted the daqd, verified the medm screen for the 05 model showed appropriate activity.

The network switch for vacuum controls in the MSR as well as several other devices were powered by normal AC power.  The power strip for all devices was moved to a UPS powered outlet to allow operation during brief power failures.  This did not cause an interruption for the vacuum network since the switch had dual power supplies, but other services were interrupted for about 2 seconds.  Nobody noticed.

The SATABOY RAID unit associated with the h1fw1 system suffered a controller board failure at 1:30am Sunday. Dan (LDAS) diagnosed the problem and for now removed the failed controller card and moved everything over to the second controller card. A replacement card is on order, we will schedule a h1fw1 outage when this arrives.

h1fw1 resumed writing frames at 15:30 this afternoon local time.

Mark B.

Commencing another round of Matlab TFs on PR3.