Combined report for the weekend
model restarts logged for Fri 04/Apr/2014
2014_04_04 15:07 h1isiham5
model restarts logged for Sat 05/Apr/2014
2014_04_05 07:27 h1fw1
2014_04_05 23:53 h1fw1
model restarts logged for Sun 06/Apr/2014 None reported
h1isiham5 expected, h1fw1 unexpected.
Last we we installed some lower frequency blend filters into the rX and rY filter banks for stage 1 and 2 of the ETMX and ITMX ISIs. This was motivated by seeing a lot of coherence between stage 1 and 2 rY motion to the optical lever pitch signal around the first suspension mode (COH >0.8 from 0.2-0.6Hz more or less).
I've taken some performance plots for these controllers (all data was taken with level 3 controllers)
Units are nm/rHz or nRad/rtHz
I tested three states, 11 DOFs using TCrappy, Stage 1 rZ using Start, - Black and Orange dotted lines T240 and Gs13 signals
9 DOFS using TCrappy, Stage 1 and 2 rY using TBetter (lower blend frequency see plot) Stage 1 rZ using Start Cyab line T240s
10 DOFs using TCrappy, Stage 1 and 2 rY using TBetter Blue, Brown and Pinks curves, T240, Gs13 and L4Cs
The interesting plots are rY which shows lots more isolation at the suspension mode, the GS13 sensor noise is ~3E-10 m/s/rHz @ 0.45Hz. Which says that we are way into the sensor noise (today was a windless sunday), so maybe we can relax the blend a bit and knock down the microseism a bit., I'm a little suspicious that i made a calculation error here
The X plot shows some increase below 0,2Hz
The rZ plot shows a fair amount of microseism which is being increase by the Tcrappy blends, we should try a more aggressive blend for stage 1 rZ
Also ITMY does not have a good set of level 3 controller installed
I did screw up the noise floor, the GS13 spectrum hasn't been taken out of the spetrum yet,that adds a factor of 1/5 and makes the noise floor ~8E-11m/s/rtHz, the spectrum is at ~1E-11m/s/rthz at 0.5Hz
On the test sheet of this unit (SN 2240A), the current was specified at 1.61A. Since the potentiometer for the current was maxed out and the current clamp LED was on, though the current was 1.63 I turned it down to 1.61. No current clamp LED any more.
Sheila, Chris, Stefan In order to tackle the changing sensing matrix problem today we tried to use the soft-hard degree of freedoms. The theory is that the beam overlap is much more sensitive to the hard DoF, thus this should be the one that is easily sensed independent of the sensing-matrix-of-the-day. Then a gain hierarchy could be used to lock down everything. Thus we started with calculation the theoretical hard-soft output matrix. Details can be found at https://gwic.ligo.org/thesisprize/2011/dooley_thesis.pdf For reference, the desin RoCs are ETM: 2245m, ITM: 1934m, giving (with L=3994.5m) g_ETM =-0.779, g_ITM =-1.065. We messed up the initial calculation and worked with ITM = .571 .495 * SOFT ETM -.495 .571 HARD for yaw, and opposite off-diagonal sign for pitch. while we should have used ITM = 1.000 -0.867 * SOFT ETM 0.867 1.000 HARD We verified that the corresponding modes were mostly hard/soft by observing the power drop for a fixed offset in soft/hard. The hard mode has the bigger power drop. Next we engaged DoF's one at a time, starting with the hard modes. For yaw, DoF2 (hard mode) we inverted the last measured sensing matrix and converted it into the soft/hard basis. We got SOFT = 5.14 11.7 * WFS A HARD -.272 1.04 WFS B The hard loop easily closed that way. For the pitch DoF's we first tried the same scheme, but noticed that the hard mode ended up much stronger in what should have been the soft signal. So instead we used the diagonal matrix WFS A -> hard, WFS B -> soft SOFT = 0 1 * WFS A HARD 1 0 WFS B Again closing that loop was straight forward. Next we closed the soft loops with low gain. That seemed straight forward. But then we started noticing that we saturated the PUM with slow signals because the top-mass bleed-off was too slow. (We moved the bleed-off filters to the unused UIM filters to keep the reliev script working.) Before we could fix the saturation issue we noticed that the cavity transmitted spot on the camera did no longer line up with the straight shot beam once all loops were on. So we manually realigned the cavity and tried to reengage the WFS, loop-by-loop. Yaw hard worked fine, but to our surprise pitch hard didn't work at all - the error signal behaviour was completely different. So our tentative conclusion is that our "sensing matrix of the day" doesn't seem provide a clean signal for the supposedly easy hard pitch mode.
I have been following the struggle to get a stable sensing matrix for the WFS and urge you to make sure the WFS head isn't driven into oscillation by the rf on the optical beam hitting the detector. This was a central problem in the WFS heads in initial LIGO and led to the rewiring of input leads in the WFS head. The symptom was that at some critical intensity into a vulnerable quadrant (not all quadrants in a given WFS head did this), the amplifier attached to the offending quadrant would begin to oscillate at the drive frequency and produce an output signal with a different RF phase, the transition from the correct phase to the phase when oscillating was quite sudden as a function of the intensity. The rewiring and the removal of the gain adjust chip in each channel stopped this. Do you suppose this is happening now? A good test is to illuminate the quadrant with RF modulated laser light and display the output of all four quadrants on an oscilloscope while one varies the laser intensity and the amplitude modulation index. One of the traces will shift in phase relative to the others on the screen.
(Sheila, Stefan, Chris)
We again closed 3 WFS loops (the hard pitch loop, and both yaw loops), after decentering the beam on the WFS heads. In this state, I compared Rich's new blend filters (Tbetter) and the standard (Tcrappy), finding a modest but repeatable improvement.
The reason we investigated the centering was that, with the beam spot well centered, the response of individual WFS quadrants to pitch drives on ETMX and ITMX was bogus. All quadrants responded with the same phase, as if it had been a length drive. We then suspected that an ugly beam profile on the WFS could make the DC light levels on the quadrants unreliable for centering. Moving the beam spot down on WFS_A, sure enough, there was a "happy place" where a usable error signal for the hard pitch DOF was recovered. We will check the beam quality on the WFS again to see if it can explain this.
The attached spectra show the ALS COMM noise with the ISI RY blends on Tcrappy (refs 10/11) and Tbetter (refs 8/9). Also, Rich advised us of a problem with the RZ blends, so those were left on Start for all measurements.
Today we finished off measuring the modal/violin frequencies of ITMY. Everything looks fine. These are briefly summarised below: Modal frequency (Hz) bin width (Hz) pendulum 0.65 1.56E-02 transverse 0.656 1.56E-02 pitch 1.109 1.56E-02 yaw 1.094 1.56E-02 bounce (*) 6.87 1.25E-01 roll (*) 9.7 1.25E-01 * Note: bounce and roll will be sqrt(2) higher when both PUM and ITM are free (9.7Hz/13.7Hz) Violin modes frequency (Hz) bin width (Hz) Back Left 501.5 2.50E-01 Back Right 501.3 2.50E-01 Front Left 504.2 2.50E-01 Front Right 502.8 2.50E-01 The average violin mode is 502.4Hz with a standard deviatioin of 1.3Hz (+1.7Hz, -1.1Hz) Fibre guards have now been installed. Giles, Travis, Jason, Gary, Betsy
I have been trying different things to fix the problem of starting of guardmedm from a sitemap which in turn was started from a panel icon. Following Keith's system I configured the icon to run a script to start MEDM instead of running it directly and it seem to have fixed the problem. If you are logged in as controls and need the new configuration you will have to log out and log back in again.
I was not able to print the error of why a directly launched sitemap was unable to run guardmedm. I tried using fie paths in full or redirecting stderr to a file to no avail. If I ran within a terminal to get stderr printed it started working.
Sprayed "around" each new joint with an audible flow of helium for ~25 seconds duration -> Access to several joints was impeded by various obstructions and, in many cases, I could not apply the helium to the test slots etc. -> No flanges were bagged and this exercise should not be considered a definitive test but should have revealed any significant offenders. LD indicated baseline signal was "offscale" < 10-11 torr*L/sec for the 1st half of testing -> I guessed the time constant to be ~100 seconds for this V/S -> to validate the less than one time constant application duration I verified that the baseline helium indication was legitimate by spraying the O-ring at the inlet to the LD and observing the permeation. The signal rose to a high of 8 x 10-9 torr*L/sec over the course of 60 minutes -> I let the signal decay to < 8 x 10-10 torr*L/sec before resuming and testing the remaining joints
End Y PLL Open Loop TF results:
Servo Board Settings:
The UGF is at 23kHz with a phase margin of 52 deg (consistent with HIFOX). Using the AG spectrum analyzer, I tried finding the PZT resonance. The data is a bit ratty, but there appears to be a resonance a little below 200kHz (see the picture -- the two pictures were collected with different BWs; the first at 300Hz and the second at 2Hz; this notch-looking feature seemed to be repeatable...). Meanwhile, the board has a notch at around 280kHz, which also seems to appear in the pictures.
I have also uploaded and plotted the data I collected.
Attached are plots of SEI watchdog trips.
It would be useful if there was some context for each of these. Do we know what was going on at the time of each trip? It looks like there are four distinct events:
It would be useful to know what was going on around each of these times.
8:15 morning meeting: HAM4 HEPI replacement of Parker valve Jim W. working on HAM4/5 SEI model update Jim W. troubleshooting isolation filters at end Y Grout cleaning at end Y Kyle helium leak testing at end Y 08:36 Jeff B. to end Y for grout removal 09:00 David H. working on TCSX 09:00 Keita transitioned end Y to laser safe 09:12 Filiberto getting TCS electronics from mid Y 09:32 Karen to end Y to clean grout removal area 09:46 Filiberto and Jim W. to end Y, troubleshooting seismic rack components for misbehaving seismometer 10:19 Grouting removal finished, end Y transitioned to laser hazard 10:55 Karen back from end Y 12:04 David H. and Alastair stopping work on TCSX for lunch 12:47 Thomas restarting h1ecatc1 Beckhoff system manager 13:08 David H. and Alastair working on TCSX 14:09 Filiberto and Aaron to end Y to replace chassis 15:06 h1isiham5 model restarted I removed the WD SUM and DC components from the end Y SUS IOP watchdog alarm handler. These were no longer valid.
WP4548
Jim, Rich M and Dave.
We installed a new h1isiham5 model on h1seih45. Jim corrected a cut-and-paste error with the BIO parts.
No DAQ restart was required, restart went smoothly. This closes WP4548.
Opened the PLL CM board chassis (S1002641) and Installed a notch filter daughter board S1300952 in the fast option slot. The notch frequency is about 290kHz now.
Put a timing comparator in the remote rack.
Recalibrated the DC diodes (fiber power, fiber wrong pol monitor, BBPD).
Auto locker worked with these things in place.
One odd thing was, when we started working the beat note was -40dBm or something, a big decrease from yesterday.
Then we started calibrating things, touched some steering mirrors to center the beam on DCPDs and such.
Then something happened, suddenly we're back to -10dBm beat note. We couldn't find out any obvious cause for this.
Power levels:
Fiber output is about 60uW.
About 3uW is sent to the fiber power monitor.
About 55uW goes to the PBS.
This morning 31uW was on the wrong polarization monitor PD (it was less yesterday).
About 24uW goes to 50:50 splitter and is combined with Prometheus IR.
From the Prometheus, about 39mW is coming to the PBS.
HWP was adjusted to send 1mW to the IR monitor diode path and 38mW to the 50:50 splitter that combines Prometheus and fiber.
In the Prometheus IR monitor path that receives 1mW, there is a pick off mirror that throws away most of the power and send 10% or something to the actual PD, but I didn't measure the splitting ratio there.
The BBPD receives 19mW.
The excess grout found on the Op-Lev receiver pier at End-Y has been removed. The grout was kept damp during the removal process to hold down dust contamination. Dust counts were monitored during the process and although elevated no readings over 9000 0.3 micron particles per sample time were recorded. The surrounding floor area has been recleaned.