Discussions with MIT (RichM) and looking closer at the very controllers & filters I plotted showed the problem.
While the developed controller I showed in the SEI Log, looked like it should not have been terrible at dealing with the potential instability at 8hz, the filter I posted in the LHO alog, was in fact a design unable to deal with this problem with strong peaking and marginal phase room at 8hz. Once I realized these two views of the ETMY controller could not be the same, I knew somewhere the problem was twisted. So I painstakingly went through all the filter design steps and sure enough, the design was there but was not saved correctly or prepared properly. The Blue traces are the new X controller in the attached for ETMY while the Red traces are from the ITMX which is how ETMY appeared before.
There is of course no free lunch and if I understand and can explain (mostly for my own benefit). The large phase bump in the 8ish Hz area pulls the controller away from instability and the design inherently has less gain peaking (maybe that is the same thing.) On the other hand, if the zero in the plant (which can move around) moves away from this design point and/or the designer is too aggressive at matching the plant, the controller will develop something... (gain peaking?)
At the moment, ETMY is running under full Guardian Management in FULLY_ISOLATED. Don't want to get ahead of myself but it has been almost 15 minutes now and Rai & Gerardo are still down there working on the chamber. I have had one previous trip of the ISI for L4Cs but HEPI did not trip. So, at least we are in a new epoch for BSC10 HEPI.
Summary : We tested charging/deionizing of the unsuspended, isolated ITMY (Corning ETM02) with a new electrometer, the Alphalab Ultrastable Surface DC Voltmeter. We first measured the field at the HR surface of the optic interleaving a series of timed deionization runs with the TopGun Deionizer. Next we applied and ripped FirstContact, and repeated measurements/TopGun nitrogen deionization. The deionization process appeared to take several minutes, having some effect out to nine minutes - however we have to repeat the measurement to assess charge on the barrel and AR optic surface (G. Moreno, N. Robertson, M. Landry, with C. Torrie and R. Abbott).
Details :
We set ETM02 (recently employed in BSC8/1 as the pilot ITMY for single-arm/HIFO experiments) upright on a teflon V-block in a cleanroom of the west bay of the LVEA ("fiber welding cleanroom"). We made measurements of the electric field 1" from the center of HR surface of the optic, and four measurements halfway between the center and the limb of the optic (right, top, left, bottom). Initially we simply applied the TopGun nitrogen in a series of 3 minute applications. Later we applied FirstContact, let it dry overnight, and the pulled (with concurrent TopGun dionization), and made more field measurements interleaved with de-charging. The data are below.
All values in negative kiloVolts (-kV) as read from the hand-held meter, unless otherwise signed positive. Field values in kV/m can be obtained by scaling by ~30X.
Trial | Vc | Vr | Vt | Vl | Vb |
1 | 2.4 | 2.7 | 3.2 | 2.7 | 2.6 |
2 | 2.2 | 2.2 | 2.4 | 2.3 | 1.8 |
3 | 0.11 | 0.14 | 0.10 | 0.15 | 0.10 |
4 | 0.02 | 0.05 | 0.035 | 0.03 | 0.03 |
5 | 0.02 | 0.035 | 0.01 | 0.03 | 0.035 |
6 | 0.005 | 0.005 | +0.020 | 0.025 | 0.015 |
7 | 0.150 | 0.70 | 0.03 | 0.160 | 0.11 |
8 | 0.108 | 0.130 | 0.045 | 0.23 | 0.05 |
9 | 0.01 | +0.01 | 0.035 | 0.08 | 0.01 |
where
Vc = raw voltage (kV) 1" above center of HR surface
Vr = raw voltage (kV) 1" above center-right (i.e. at a radius of +34cm/2, 0 degrees rotation angle) of HR surface
Vt = raw voltage (kV) 1" above center-top (i.e. at a radius of +34cm/2, 90 degrees rotation angle) of HR surface
Vl = raw voltage (kV) 1" above center-left (i.e. at a radius of +34cm/2, 180 degrees rotation angle) of HR surface
Vb = raw voltage (kV) 1" above center-bottom (i.e. at a radius of +34cm/2, 270 degrees rotation angle) of HR surface
and trial data were taken under the following conditions:
Trial 1 - Initial measure of electric field, HR side of optic, prior to application of FirstContact (FC), or TopGun deionizer (TG)
Trial 2 - same as trial 1, but 24 hours later, to see if there is some discharge in air (there is, ~25% lower than trial 1)
Trial 3 - After 3 min of TG deionization
Trial 4 - After 3 min of additional TG (6 min total)
Trial 5 - After 3 min of additional TG (9 min total)
Trial 6 - No additional TG, but after 3h delay
Trial 7 - FirstContact applied, then removed in a 2 min pull, with concurrent TG (2 min), plus additional 2 min TG thereafter
Trial 8 - After 2 min of additional TG (6 min total)
Trial 9 - After 3 min of additional TG (9 min total)
Individual measurements of the field could take 3-4 minutes to stabilize in the Voltmeter, i.e. values read out on the hand-held unit may have initially read hundreds of volts, falling and stabilizing to a seemingly stable number over the 3-4 minute time frame.
We wil repeat these measurements and include barrel and AR surface tests. However, in consultation with C. Torrie and R. Abbott, and in order to be conservative in closing BSCs 1, 3 and 9, we opted to extend nominal few-minute TopGun blowing times for test masses and reaction masses to nine minutes total, i.e. 4.5 min for a given test mass HR face, 4.5 min for the AR gap, 4.5 min for the reaction mass upstream face, 4.5 min for the gap. We cycled between surfaces and gap in 1 minute intervals, summing to 9 min per optic.
While doing cabling for HAM6 in the CER, noticed that the following cables for ISCT1 in ISC-C2 were plugged in backwards. Swapped the following at the EtherCAT Corner 5 Chassis. Cables are connected as listed below and match the pull list E12000408. Cable_ISC_328 connected to Port 9 (DC PD's ISCT1) Cable_ISC_326 connected to Port 10 (Auxiliary ISCT1) Filiberto Clara
(Sheila, Alexa)
After fixing the corner beckhoff, we noticed that the PSL persicope PD was still not reading the correct voltage based on yesterday's measurement alog 13222. It turns out that swapping back cable ISC_326 and ISC_328 fixes this problem.
As desgined from D1100583, E1200077:
The Auxiliary ISCT1 cable ISC_326 from Concentrator 1 --> EtherCat Corner Chassis slot 10
DCPDs ISCT1 cable ISC_328 from Concentrator 2 --> EtherCat Corner Chassis slot 9 --> System Manager L9-10 EL3104 Adapter 9
The System manager is linked according to the specification above. However, with the proper configuration the readback was not connected. Maybe it's an internal swap in the EtherCat Chassis? We have flipped the cables at the concentrator.
Current Configuration:
The Auxiliary ISCT1 cable ISC_328 from Concentrator 1 --> EtherCat Corner Chassis slot 9
DCPDs ISCT1 cable ISC_326 from Concentrator 2 --> EtherCat Corner Chassis slot 10 --> System Manager L9-10 EL3104 Adapter 9
I have updated the MEDM screens which are snapshotted and displayed on the web page:
https://lhocds.ligo-wa.caltech.edu/screens/
Diagnostic breadboard scans were executed; these are to measure the frequency noise, beam pointing, mode content and relative power noise. Some notes in passing: - The higher mode content of the frontend laser looks okay. - Autoalignment did not go green as expected, I will have to sort this out after reading the manual. - The beam pointing seems higher than the requirement but this could be because the calibration is out (plus my operating error). - The relative power noise is higher than the requirement, assuming the calibration is correct.
Reference Cavity Re-alignment ============================= The reference cavity transmission has steadily degraded over time from ~1.1V a month ago down to the ~0.4V when Jeff reported the problem to me. A quick look at the alignment using only the periscope adjustments, did not make things any better and in fact I made it a little worse, leaving the transmission at ~0.22V (as displayed on the MEDM screen). That indicated that either the alignment into the AOM was off or the RF level to the AOM was off. Measured RF on oscilloscope, with 50 ohm input. freq = 79.4 MHz peak-to-peak amplitude = 21.66 V Vrms = 7.965 V So the RF level into the AOM is 1.27 W. This is about the right level for maximum diffraction efficiency. Note that the maximum RF power into the AOM is 1.3 W. zero reading on power meter = 2.7 mW, power meter used is the Ophir thermal calorimeter no measurements have been corrected unless otherwise noted power incident on AOM, measured after PBS = 57 - 62 mW power incident on 21.5 MHz EOM = 21 - 23 mW this represents a double pass efficiency of ~ 22/59.5 = 37% or a single pass diffraction efficiency of ~61% (Sqrt[37%]) After adjusting mostly the height of the AOM, the power before the 21.5 MHz EOM was 26 mW or 66% single pass diffraction efficiency, further tweaks to the height resulted in 34 mW or a single pass diffraction efficiency of ~76% Tweaked alignment into reference cavity using the periscope - mostly height adjustments; measured reference cavity transmission is ~710 mV on DMM, ~1.16V on MEDM display. This is as good as the value from over a month ago.
Since we are pumping down the volume, we decided to postpone restarting the models. So the mission is incomplete.
Completed stripping the Ameristat from the HEPIs and unlocking the platform at 1015pdt. Closed the HEPI loops with no difficulty.
Attached at spectra from overnight (HELI Locked) compared to after unlocking; and, comparing after unlocking to having the position loops closed.
Comparing the three states, basically it depends, some frequencies better, others, not so much.
The Mid-X weather station, which died Saturday evening, was restarted by power-cycling the Comtrol box and the weather station. The EPICS IOC for the weather station was restarted and the weather is now being reported properly. No equipment needed to be replaced.
The h1seiey computer crashed with a kernel panic at 08:41 PDT today. Attempts to remove it from the dolphin network were unsuccessful, so on restart, h1susey was glitched and needed to be restarted as well. Inspection of the computer and I/O chassis showed no obvious reason for the crash.
J. Kissel, K. Venkateswara, E. Shaw Krishna and Eric arrived yesterday, and we've begun installation of the Beam Rotation Sensor at H1 EX. Pictures attached! Some quick notes -- the center of the platform is roughly 41" WEST (+Y), and 9" SOUTH (-X) of the GND T240, and aligned with the IFO's X-axis, parallel to the line formed by the GND T240 and PEM Guralp. I also attach a screen shot of the custom overview screen I created.
BSC1 and 3 gatevalve work in progress The extra viewport on BSC2 will be left in place IOT2 table to be moved to west side of HAM2 The north door will be pulled on HAM6 for ISI rebalancing wipe down and particle assessment beforehand Pumping vertex by noon HAM6: placing of fast shutter component today, remainder of work requires beam Peter K. is checking alignment on PSL table TCS: Started bonding polarizer Start work on X arm table after lunch, request for laser hazard in the afternoon ITMX optical lever work in the morning EE work in racks by HAM6 Hugh will be removing the ameristat at HAM4 Rai W. here to work on mirror charging experiment at end Y end Y will be transitioned to laser safe tiltmeter experiment underway at end X Kiwamu updating and restarting the ASC realtime model Praxair delivery to mid Y upgrade of conlog
model restarts logged for Mon 04/Aug/2014
2014_08_04 12:39 h1lsc
2014_08_04 12:42 h1lsc
2014_08_04 12:45 h1broadcast0
2014_08_04 12:45 h1dc0
2014_08_04 12:45 h1fw0
2014_08_04 12:45 h1fw1
2014_08_04 12:45 h1nds0
2014_08_04 12:45 h1nds1
no unexpected restarts. LSC ipc work plus related DAQ restart.
Sorry, got the dates wrong and skipped Sunday:
Sunday 3rd August: no restarts reported
Monday is actually the 4th of August.
[Jeff Arnaud]
Tonight we balanced the coils of the lower stages of MC2 SRM and SR2. The MC2 notes attached are explaining the procedure in detail. Below is summarized the ASD ratio after/before the exercise for each suspension and each level.
For some reason M3 stage of SR2 was harder to balance than the other ones, so we gave up on it for now and will come back to it later.
SUS | STAGE | ASD RATIO before/after balancing | |
PITCH | YAW | ||
MC2 | M2 | 19.71 | 2.63 |
M3 | 21.27 | 13.52 | |
SRM | M2 | 9.95 | 38.53 |
M3 | 6.96 | 3.8 | |
SR2 | M2 | ||
M3 | 107.7 | 5.12 |
And the gains which were found optimum for each coils
SUS | STAGE | OSEM | GAIN |
MC2 | M2 | UL | 0.988 |
LL | -0.992 | ||
UR | -1.007 | ||
LR | 1.012 | ||
M3 | UL | 1.082 | |
LL | -0.948 | ||
UR | -1.051 | ||
LR | 0.922 | ||
SRM | M2 | UL | 0.935 |
LL | -1.024 | ||
UR | -0.971 | ||
LR | 1.063 | ||
M3 | UL | 0.972 | |
LL | -1.014 | ||
UR | -0.972 | ||
LR | 0.964 | ||
SR2 | M2 | UL | |
LL | |||
UR | |||
LR | |||
M3 | UL | 1.003 | |
LL | -1.050 | ||
UR | -0.951 | ||
LR | 0.996 |
Measurements will be running overnight to measure the cross couplings transfer functions for SR2 SRM and MC2
There is a discrepancy in the text. While it says that SR2 M3 was not balanced however in the results table it is SR2 M2 the one blank. I can confirm that SR2 M2 is the one not yet balanced.
After BS and ITMY results, attached are itmx tfs ran over the week end, after doors closeout. During the measurement ISI was damped. Results look good. Lower stages show factor of two-ish discrepancy with model as Jeff pointed out.
1. Undamped H1 ITMX M0-M0 tfs compared with fiber model
2. Undamped H1 ITMX R0-R0 tfs compared with thincp model
3. M0 undamp R0 damp H1 ITMX L1-L1 tfs compared with fiber model
4. M0 undamp R0 damp H1 ITMX L2-L2 tfs compared with fiber model
4. H1 ITMX M0-M0 tfs compared with other ITMs
5. H1 ITMX L1-L1 tfs compared with Livingston's ITMX
6. H1 ITMX L2-L2 tfs compared with Livingston's ITMX
Data and scripts were commited to the svn
Today I have been working on porting the ESD Charging measurement automation scripts that Bryan Barr et. al. created at LLO. This has now been completed. I am adding some modifications to make the code more general. I have run the code over all the ESD quadrants on the only mass in vacuum (ETMY) and for several offset values with excitation signal at 4Hz and amplitude of 60000 and 120000 counts. The obtained analysis results did not make any sense so I looked the data more in detail and noticed that the oplev did not seem to be affected by the ESD driving signal. To be sure that I was not making any mistake I used awggui to excite the ESD in pitch and yaw such that they would excite the respective pendulum modes (at 0.56 and 0.6 Hz) identified here: https://awiki.ligo-wa.caltech.edu/aLIGO/Resonances/LHO. Nothing was observed on 'oplev'. Tomorrow I will have to verify that the ESD is actually physically connected.
LVEA laser safe 09:00 – FSS out of lock. Lowered the FSS resonance threshold to 0.3 from 0.4. Peter King will look into PSL alignment. 09:15 Nathan – Working in the optics lab 09:58 Greg – At BSC1 area working on laser barriers for CO2 laser alignment 10:01 Travis – Working on mobility experiment at HAM3 10:06 Betsy – General post chamber close cleanup in the LVEA 10:13 Gerardo – Working on viewport survey at BSC1/BSC2 area 10:55 Jim – Recycle mid-X weather station software 11:12 Peter – Going into H1-PSL enclosure to tweak ISS alignment/power loss 12:00 Power cycle restart of the camera monitor computer in the control room. 12:42 Kiwamu & Dave – Kiwamu update and restart LSC model. Dave restart DAQ 13:00 Bubba – End-Y working on deionizer installation 13:27 Bubba – Drilling bolt holes in floor at End-Y 13:47 Gerardo – Viewport survey in LVEA 13:52 Allister – At TCS-X to examine optics 14:10 Karen – Reported the emergency lights are on in the End-Y VEA 14:37 Doug & Jason – Working at the ITM-X manifold 14:52 Jeff & crew – At End-X 14:55 Alarm on corner station chiller pump #3. Pump tripped off. Informed John W. 15:25 Summer students – PEM work at End-X
J. Kissel, A. Pele Ran transfer functions over the weekend, to assess if the SUS is free after first contact removal and closing doors. While the top stages look great, this is the first time I've been exposed to a factor of two missing between model and measurement for the lower stages. While we don't expect the model to account for the dynamics of the reaction chain (which is why we see a whole bunch of "extra" features in the main-chain dynamics), we don't expect an overall gain mismatch. Arnaud informs me that this is true of every QUAD except one stage of one (including L1 SUS ITMY). No reason to pull doors off, but we should figure out where this discrepancy lies. And also , we should develop a model of the lower stage dynamics which include both chains.
The reaction chain will displace about the same amount as the main chain for the lower stage measurements. Could this explain the factor of two gain relative to the model, since the OSEMs measure the difference between chains?
HA !
Plausible. Let's model it to find out!
I added a feature to .../SusSVN/sus/trunk/QUAD/Common/MatlabTools/QuadModel_Production/generate_QUAD_Model_Production.m to compile a quad model with both main and reaction dynamics to predict what the L1 (UIM) and L2 (PUM) OSEMs will measure. See the results in the attached pdf. The overall gain is still not 100%, but is definitely closer. Pitch has some obvious mismatch due to known extra stiffness on the reaction chain (from OSEM cables I think). The update to generate_QUAD_Model_Production.m has instructions commented into the header for using the new combined main/reaction model. I copy those instructions here. " You can compile multiple chains simultaneously, to predict what the UIM (L1) and PUM (L2) OSEMs will measure. To use this feature, sandwich the string '_with_' between the two build types you want to use, main chain 1st, reaction chain 2nd. For example, for a monolithic ITM use buildType = 'fiber_with_thincp' The combined model will only combine the outputs for the UIM (L1) and PUM (L2) L, P, and Y DOFs, to reflect the OSEM measurements. The inputs however will be shared for all the seismic inputs as well as the UIM and PUM L, P, and Y DOFs. The indices for inputs and outputs are unchanged. Caution: there is no error or warning for non-physical combinations. For example, 'fiber_with_fiber' will compile OK, but the results will be inconsistent with reality. "
For future reference, attached are osems spectra from before the swap comparing top mass damping on / off
This plot is interesting. It seems the only moderately pronounced peak in the horizontal ground spectrum wanders, seen from 8 to 11hz changing quite rapidly.
The attached plot's current traces are with HEPI loops closed and the ISI Damped. While the middle graph shows all four local IPSs have the strong 8hz peak (the dashed REFs are with the HEPI loops open), the peak only shows in the X & Y and not the RZ cartesian traces (top graph). Also, note the bottom graph where the ground sensor pretty much has nothing at 8hz but does have a minor peak just above 9hz that is also peaking up in the local sensors.
The second plot shows just the ground Seismometer with X & Y traces from last night (Dashed) and the others from this morning. The peaks in that area come and go and wander around so it may or may not be a problem or just a red herring (are those edible?)
I am attaching the spectrograms for the X, Y, and Z directions of the STS Seismometer in ETMY. I used 19 continuing hours, starting 2014-07-22 01:00:00 UTC. There is a feature that is present all the 19 hours: Around 10.5 Hz in X direction. Around 9 Hz in Y direction. Around 10.5 Hz in Z direction. More features can be seen wandering along the 19 hours around 3-5 Hz and 8-10 Hz, for X, Y, and Z directions. In the spectrograms, each count is 1 nm/s /sqrt(Hz).
Continuing with the investigation on the 8 Hz, I am attaching the spectrograms for the X, Y, and Z directions of the GND STS in EY. Comments on the spectrograms: - Feature wanders between 7.5 to 12.5 Hz. - Depends on the time of the day. - It is present every 15 to 30 minutes. - Displacement amplitude higher than 2 nm. 6 hrs were used for each spectrogram, starting: - August 01, 2014 00:00:00 UTC (Figure 1) - August 01, 2014 06:00:00 UTC (Figure 2)
The 8 hz motion seen here is most likely related to the "pier resonance" Laura Nutall has a nice set of plots in the DCC https://dcc.ligo.org//LIGO-G1400820 which show this motion. Likely you are seeing the floor component of this motion. Rich M. has data showing that (at MIT) the slab bending is an important component of this motion
I investigated the GND SDS and PEM MIC channels using the coherence tool at 100mHz bandwidth and found several peaks between 5 and 9 Hz. I've attached some slides with zoomed-in plots.