WP 4719 Installed updated dataviewer to fix a bug that incorrectly displayed trend mean data as 0 when max and min values are integers.
model restarts logged for Mon 07/Jul/2014
2014_07_07 00:40 h1fw0
2014_07_07 01:38 h1fw0
2014_07_07 02:46 h1fw0
2014_07_07 03:45 h1fw0
2014_07_07 04:39 h1fw0
2014_07_07 05:27 h1fw0
2014_07_07 05:43 h1fw0
2014_07_07 06:39 h1fw0
2014_07_07 07:33 h1fw0
2014_07_07 07:41 h1fw0
2014_07_07 08:40 h1fw0
2014_07_07 09:39 h1fw0
2014_07_07 11:24 h1fw0
2014_07_07 11:53 h1fw0
2014_07_07 13:02 h1fw0
2014_07_07 13:25 h1fw0
2014_07_07 14:43 h1fw0
h1fw0's problem (purple) resolved when solaris QFS box rebooted. No unexpected restarts.
J. Kissel Executive Summary: (1) I believe we DO NOT need to manually alleviate the alignment on the IMs, but it's close. (2) Assuming a worst-case scenario, IM2 would use up ~80% of it range to maintain alignment. (3) The alignment slider calibration I recommend we use is: P: 0.21 +/- 0.015 [urad/ct] Y: 0.39 +/- 0.009 [urad/ct] (4) Calculating the calibration from from a model of the signal chain fails; there's still unknowns in the signal chain that add up to discrepancies of 2 and 4 for P & Y. (5) Given the EUL2OSEM gain, and the usage of the same actuators for P & Y, the actuators can drive 2.96 [mrad] in P and 1.62 [mrad] in Y if alignment offset is needed in both DOFs. (6) We would need to relieve 1 [mrad] of pitch, which is about at the limit of the mechanical alignment precision. I have *not* installed the calibration gain. I'll do it tomorrow. ---------- During the last integrated testing stint, good alignment of IM2 and IM4 was believed to be at the limit of the DAC range of their OSEMs, given their actuation strength has now been decreased by ~10 (see LHO aLOGs 8767, 8759). However, due the confusion of uncalibrated alignment sliders, artificial limitation of the slider range on the MEDM screen, 16-bit vs 18-bit DAC range, coil driver strength changes, computer reboots, electronics being swapped in and out, and a few vent cycles ... I think the saturation had been misreported. Currently, the worst offending with the largest alignment bias request is IM2 with + ~81000 [ct] out of +/- 2^17 = 131072 [ct], or 61% of its range. I've confirmed with the OSEMs that the range of drive is still roughly +/- 6 [mrad] in P and Y, and the calibration of the slider counts into [urad] is P: 0.21 +/- 0.015 Y: 0.39 +/- 0.009 where the uncertainty is the standard deviation across the four IMs. Unfortunately (and the reason why I just measured it), a component by component model of the calibration doesn't make any sense, even after using measured compliance values. More on that later. So the questions are: - Is +/- 6 [mrad] enough to cover the alignment change from in-air to in-vacuum? Yes. Data from prior to the vent and before installation attacked shows that the optics move by at *most* 1 [mrad], where the majority moved of order 100 [urad]. So, assuming IM2 moves the worst amount of distance (which this past vent shows it moved 148 and 174 [urad] in P & Y respectively) that's only an additional ~17% on top of 61%, or 78% (and the damping control signal variation on this is a negligible ~0.1%). - Is it worth trying to mechanically alleviate the alignment for 78% and 2 [mrad]? We have no experience HAUX aligners on site for this remaining week in which we indeed to keep HAM2 open. The precision to which one can mechanically align the SUS is roughly 1 [mrad] (see section 2.1.1 of T1200469). Given Murphy's law, the lack of alignment references, the amount of stuff we've already changed in HAM2 with the IMC bypass, and the uncertainty which still swirls around the exact numbers and direction, I say it's a gamble not worth taking. Below is a collection of information needed to assess how much the optics need moving. I've surveyed the (1) alignment sliders, (2) OSEM sensor values, and (3) DAC output values over the past 30 days, to get a feel for how much we need to move. However, three events have occured that complicate the story: (a) On 2014-June-04 16:21 UTC to 2014-June-04 23:13 UTC, Filiberto and Aaron swapped out / investigated AA and AI chassis of H1sush2b, so OSEM sensor signals went dead (no aLOGs, but Fil confirms that's when his work-permit for this swap was open via email). (b) On 2014-Jun-12 17:21 UTC, a power glitch necessitated a restart of the h1sush2b computer (among many others). This reboot reloaded a save.snap value that must have been out of date, because IM4 P & Y alignment value changed by a few hundred counts (see below). (c) On 2014-Jun-24 ~23:00 UTC (+/- 15 [minutes]), the alignment was modified in order to accommodate the IMC bypass. "IM2_P_35DayTimeline.pdf" shows this timeline on IM2 P only for clarity, but "OSEMPosition_2014-06-01_to_2014-07-04.png" shows that all 8 DOFs show similar features over the 35 day period. (1) OSEM sensor values, before the vent (Jun 1), after the vent, once the alignment had asymptotically stabilized just before the (a) AA/AI chassis shutdown on the Jun 4, and after the alignment had been left alone from the IMC bypass during the 4th of July weekend: % Pre-vent P Y osems(1).disp = [ 225.0 1709.5;... % [urad] IM1 1380.6 -45.5;... % [urad] IM2 447.6 620.0;... % [urad] IM3 -3034.0 -2174.8]; % [urad] IM4 % Post-vent P Y osems(2).disp = [ 195.7 1772.9;... % [urad] IM1 1528.8 -220.2;... % [urad] IM2 511.8 599.3;... % [urad] IM3 -3167.0 -2302.0]; % [urad] IM4 % Post-IMC Bypass P Y osems(3).disp = [ 335.0 1851.0;... % IM1 2179.0 165.0;... % IM2 1522.0 816.0;... % IM3 -3201.0 -2212.0]; % IM4 % P Y diff = -29.3 63.4 % [urad] IM1 148.2 -174.7 % [urad] IM2 64.2 -20.7 % [urad] IM3 -133 -127.2 % [urad] IM4 (2) Alignment slider values, before vent and after their change from (c) above. These slider values *have* not been calibrated into physical units, as the alignment bias gain is 1. The only P&Y gain between the sliders and the DAC are the EUL2OSEM matrices compensating for the lever arm, which have a gain of +/- 8.594. Optic DOF Pre-vent After Glitch After Realignment Diff [ct] IM1 P 0.0 (same) (same) n/a Y 1469.0 (same) (same) n/a IM2 P 4600.0 (same) 7600.0 +3000 Y 2000.0 (same) (same) n/a IM3 P -1397.9 (same) 2980.0 +4377 Y 1060.8 (same) 2120.0 +1060 IM4 P -2350.0 -2707.0 (same) -357 Y -3607.0 -3390.0 (same) 217 (3) DAC Output values, before the vent Optic DOFs DAC Output IM1 UL / UR -12631 / -12631 LL / LR +12631 / +12631 IM2 UL / UR +22344 / -56720 LL / LR +56720 / -22344 IM3 UL / UR -21130 / +2897 LL / LR -2897 / -21130 IM4 UL / UR +10802 / +51194 LL / LR -51194 / -10802 ------ What I think the alignment slider gain should be: If I use the same method as used in other susTypes (see e.g. LHO aLOG 4730), where the EUL2OSEM matrix properly compensates for the lever arm an number of OSEMs, OPTICALIGN GAIN [cts/urad] = ( DAC [V/ct] * Coil Driver [A/V] * Coil/Magnet [N/A] * HAUX M1 to M1 [rad/N.m] * 1e6 [urad/rad] )^-1 = ( (20/2^18) * 0.988e-3 * 0.0158 * 74.23 (for Pitch) * 1e6 )^-1 = 11.311 [ct/urad] (for Pitch) or 0.08841 [urad/ct] OPTICALIGN GAIN [cts/urad] = ( DAC [V/ct] * Coil Driver [A/V] * Coil/Magnet [N/A] * HSTS M1 to M1 [rad/N.m] * 1e6 [urad/rad] )^-1 = ( (20/2^18) * 0.988e-3 * 0.0158 * 73.87 (for Yaw) * 1e6 )^-1 = 11.367 [ct/urad] (for Yaw) or 0.087974 [urad/ct] -- a factor of 2 and 4 off the measured values -- where the DAC gain is taken from T1200311, the coil driver transconductance from T1200264, and coil magnet strength from T1000164, and the compliance from the collection of H1 HAUX measurements in LHO aLOG 8344, since the transfer function model still does not well-match the transfer function data. -------- The final attachment "2014-07-08_H1HAUX_AlignmentOffset_Calibration.pdf" shows the measurement results of simply taking the alignment sliders at several data points along its range, and pulling the OSEM value off of a data viewer trace. Because I left the opposing DOFs offsets ON during the range walk through, some of the SUS which already had appreciable offsets reached saturation more quickly (i.e. IM4P, IM2Y, and IM2Y). For those data sets, I removed the end points from the fitting routine (which was just a first order polyfit, nothing fancy). Data was reduced and plots were created by ${SusSVN}/sus/trunk/HAUX/H1/Common/alignmentrelief_Summer2014.m
I've installed the above calibration gains, adjusted the alignment offsets to yield the same displacement, and captured a new safe.snap. The new alignment offsets are H1:SUS-IM1_M1_OPTICALIGN_P_OFFSET 0 H1:SUS-IM1_M1_OPTICALIGN_Y_OFFSET 3786.1 H1:SUS-IM2_M1_OPTICALIGN_P_OFFSET 35849.3 H1:SUS-IM2_M1_OPTICALIGN_Y_OFFSET 5154.7 H1:SUS-IM3_M1_OPTICALIGN_P_OFFSET 14057 H1:SUS-IM3_M1_OPTICALIGN_Y_OFFSET 5464 H1:SUS-IM4_M1_OPTICALIGN_P_OFFSET -12769.1 H1:SUS-IM4_M1_OPTICALIGN_Y_OFFSET -8737.1
Currently, the bottom of a search results page includes the "Displaying reports X-Y of Z. Go to page Start 5 6 ... End" links. Conspicuously absent from the links are links for the Next and Previous pages of results. Navigating through the results therefore always requires more energy than should be needed, to identify the number of the current results page, and select the next number. If a Next link was provided, it would be much easier to scroll through search results.
Better yet, the subsequent results should just automatically show up at the bottom of the page when scrolling down. This can be achieved fairly simply with the jQuery jScroll plugin.
This morning, I finished the cable routing through the reaction chain to the lower stage OSEMs. Then, Travis and I worked on fully suspending both chains. After adjusting out some course pitch observed at the test mass and CP, we noticed we have a large pitch at the top mass. Both chains are also hanging about 1-2mm high (we have shims to remove which will correct this later when IAS sorts any SUS+HEPI height issues). We'll continue the coarse pitch work tomorrow which is a bit of a head scratcher. Both chains have been left ~fully suspended.
J. Kissel, T. Sadecki, B. Weaver, Because it was free, we took some new open light current values for H1 SUS ITMY, the have been installed. No values have changed more than 10-15%, but all have gone down since we last took these (see LHO aLOG 4549). OSEM Open Normalizing Compensating Light Gain Offset Current [ct] M0F1 27894 1.075 -13947 M0F2 27280 1.100 -13640 M0F3 27059 1.109 -13530 M0LF 25600 1.172 -12800 M0RT 23555 1.274 -11778 M0SD 27538 1.089 -13769 R0F1 23855 1.258 -11927 R0F2 24148 1.242 -12074 R0F3 24696 1.215 -12348 R0LF 25745 1.165 -12872 R0RT 23838 1.258 -11919 R0SD 24543 1.222 -12271 L1UL 27800 1.079 -13900 L1LL 28750 1.043 -14375 L1UR 27706 1.083 -13853 L1LR 29654 1.012 -14827 L2UL 23978 1.251 -11989 L2LL 20272 1.480 -10136 L2UR 16251 1.846 -8125 L2LR 17254 1.739 -8627 Note that these values still need to be captured in a safe.snap, and the OSEM Chart needs updating.
I've captured a new safe.snap, and committed it to the svn repo.
The north door on ham 5 was removed this morning. We replaced the adapter spool at GV 7 and installed the annulus piping. The cleanroom was transported from the north bay to north / west bay main crane access area.
Dan and Dave.
h1fw0 was down for about an hour today, Dan performed diagnostics on the /ldas-h1-frames file system. He first tried a samfsck on the file system, but the problem reappeared within 5 minutes. After many other investigations we ended up rebooting the h1ldasgw0 machine and h1fw0 has been stable for over an hour.
905 - Manny working on LVEA electronics along X-manifold
933 – Jeff B monitoring particle counts as HAM 5 door removed, then replacing fuses in all LVEA dust monitors
959 – Betsy working in Beer Garden @ ITMY
1033 – JimB and DaveB @EY checking AA chassis and restarting IOP WDs
1130 – Aaron to EY to pull shorted AA chassis
1136 – Dave B shutting down fw0
1151 – Dave B restarting fw0
1313 – JustinB recovering ETMY trip (reset all watchdogs incl IOP, requested fully_isolated via guardian)
1301 – Betsy returning to ITMY work
1424 – JeffB and Andres prepping for magnet gluing at HAM5
1530 - Thomas and Greg at EY working on HWS
Lights in laser and ante room are ON. Cover missing on HPO and DBB. Unsure about HVAC fan status; will check. Laser Status: SysStat: Frontend online Output power is 36.2W High Power Oscillator offline. PMC: It has been locked 3 days 6 hours Reflected power is 2.8 Watts and PowerSum = 25.6 Watts. FSS: It has been locked 3 days 6 hours Threshold on transmitted photo-detector PD = 0.9 V ISS: The diffracted power is around 7.5 %
While trying to untrip the ETMY ISI through guardian the ISI stage 2 watchdogs tripped (again). JeffK suggested manually turning on the Gain filters for the GS-13 inputs (to reduce the analog gain by 10...or LOW GAIN mode), resetting the watchdogs again, then requesting Fully_Isolated from guardian.
if it is necessary to reduce the gain by hand, shouldn't guardian handle this?
> if it is necessary to reduce the gain by hand, shouldn't guardian handle this?
Yes it should, if it is determined that this is necessary to achieve full isolation. However, this wasn't previously necessary, so something in the system has clearly changed.
But either way, more information is needed. Did you try letting guardian handle the isolation with the default procedure more than once? If so, did it trip the GS13 in the same way every time? Presumably lowering the GS13 input gain did allow the isolation to succeed? Was there other activity going on around ETMY at the time? Have any other parameters of the system changed, such as cart bias offsets on any of the HPI or ISI dofs?
All good questions. Mostly I just wanted to get a trip report logged as a data point to see if this is happening with some frequency and not getting reported.
There was no other work going on at EY at the time and the seismic background was not particularly choppy. I did not let guardian try a second time before changing the gain. As for changes to the DOFs or bias, I don't know. Guess we'll just see if it happens again. Thanks.
Before Removal In Chamber 0.3 = 0 0.3 = 0 0.5 = 0 0.5 = 0 1.0 = 0 1.0 = 0 During Door Removal Breaking/Removing the bolts 0.3 = 168 0.5 = 78 1.0 = 30 Installing the Gasket Protectors In Chamber 0.3 = 230 0.3 = 130 0.5 = 150 0.5 = 50 1.0 = 90 1.0 = 20 Putting on the C3 Door Cover 0.3 = 410 0.5 = 260 1.0 = 130 After Removal In Chamber 0.3 = 0 0.3 = 0 0.5 = 0 0.5 = 0 1.0 = 0 1.0 = 0
This morning I went into BSC1 to re-resume work on the ITMy SUS. Particle counts taken this morning are as follows:
After repairing wall, in the beer garden
All zero
In BSC3 near flooring, before I entered
0.3um 70
0.5um 40
1.0um 20
In BSC3 near flooring, after wiping flooring and walking around a few minutes
0.3um 540
0.5um 330
1.0um 210
In BSC3 near flooring, just after I left the chamber
0.3um 270
0.5um 100
1.0um 80
I had to repair the false CR wall around the BSC3 door this morning because I found it had falled down sometime since last Wed. Note, if doing this type of false wall in the future, tape needs to be added perpendicular to the joint tape, to make reinforcing "stitches" across the sheet joints.
J. Kissel, A. Pele, T. Sadecki, B. Weaver,
After another day's worth of fumbling and hand-waving, we've finally traced the discrepant transfer function magnitude to a loose cable connection at the air-side of the feedthrough. In the course of attempted fixes, we measured a few more things, moved a few more things, replaced the R0 F2 and F3 OSEMs, but the only thing that fixed it was the cable connection. This flaw with the electronics chain should have been found earlier with our cable swap on Friday, with remeasuring the open light current, and with Arnaud's DC actuation test below, but frustratingly, it did not. But we found it.
I'm in the process of getting formal transfer function measurements now, but I'm confident from what I'm seeing thus far that this SUS is read for close-out. Will post comparison TFs once they're finished.
We have exORSIIIIIIIZED the demons. This SUS is clee-uh.
Details (In rough chronological order)
-------
- Arnaud compared the ETMY and ITMX R0 F2 to F2 and F3 to F3 drive at DC using the alignment offsets (the only QUADs available, so take ERM and TCP chain comparison with a grain of salt), and concluded that the chains are driving with comparable and within a chain F2 and F3 are of comparable strength with such a quick study.
- Betsy retook an L2L transfer function (just for sanity's), saw the same discrepant factor of two -- and some more Yaw resonant cross-coupling. Still F2 shows the majority of the badness.
- Decided just to swap the OSEMs, and hopefully move on. Below are the new serial numbers, open light current, and compensation gains and offsets. These *have* been installed. We'll take a new safe.snap this evening after all the commotion has died down.
- Betsy & Travis poked at this, looked at that, iterated with measurements several times on the floor.
- At last, Betsy went a long the signal chain, and made sure all connections were secure, and finally found this bad connection at the in-air connection at the feedthrough.
- Simultaneously, Travis loosened up the the cable loop from the optical table to the top mass. This may have alleviated the extra Yaw peaks that had shown up today.
Open light current values and serial numbers for the new OSEMs.
Raw New New S/N
ADC Gain Offset
R0F2 25646 1.170 -12823 077
R0F3 26554 1.130 -13277 147
We'll also update the OSEM chart and grab a new safe.snap once transfer functions are finished.
TFs look great. For some reason the comparison script is failing on my laptop; will post comparisons tomorrow. But, for all intents and purposes, H1 SUS ITMX is ready for close-out. A victory for team SUS!
Comparison plots. All looks well, as expected. To demonstrate the problem, I also include the prior *bad* measurements of the reaction chain (2014-06-27_2117).
More transfer function measurements were taken on the lower stages of ITMX last tuesday.
Interesting things to notice :
1. The measurements with no top mass damping (1st and 3rd attachments) appear to show resonnances which are not predicted by the model. Those are most certainly resonnances of the reaction chain moving because of the reaction force (e.g. 3rd page of 1st attachment : resonnance at 0.66Hz is not predicted by the model, but matches with the frequency of the first yaw mode measured on the top mass reaction chain, cf this measurement from alog 12555).
2. There is a factor of 2ish discrepancy between the model and the measurement for both UIM and PUM