jeffrey.bartlett@LIGO.ORG - posted 09:04, Thursday 13 February 2014 (10054)
HAM4 Payload
Hugh, Mitch, and Justin Adding payload to HAM4. Using roll up door in LVEA 11:10 Done.
Hugh, Mitch, and Justin Adding payload to HAM4. Using roll up door in LVEA 11:10 Done.
Chiller level alarm was heard in shipping and receiving. Added 3/4 cup of water to chiller, alarm cleared. Good for another 2 weeks.
Having previously calibrated the IMC mirror alignment offsets (see aLOG 9870), on the evening of Thursday 6th Feb I proceeded to misalign the IMC in all 4 D.O.F.s in turn and measure the RIN spectra in transmission using MC2 trans QPD. This is the same measurement that was performed for the PSL commissioning mode beam jitter measurement as reported in aLOG 8190, the only differences being the more precise calibration of the applied offsets this time, and of course the PSL being in science mode this time.
The first attached plot shows the newly measured jitter spectra on top of the spectra measured in October with the PSL in commissioning mode. Here it is clear to see an improvement of greater than 1 order of magnitude between about 1Hz and 30Hz or so. This roughly agrees with what we'd expect given the measurements on the PSL table reported for L1 in LIGO-T1300368.
The second attached plot shows the unscaled RIN spectra for the IMC aligned, and misaligned in the 4 orthogonal D.O.F.s. Since the aligned spectrum looks very similar to the IMC misaligned spectra, it's reasonable to assume that even in the aligned case the RIN out of the IMC is dominated by the effects of beam jitter. In the ~f^(-0.5) part of the spectrum above about 5Hz the MC mirror motion is not expected to contribute to the jitter, so we can assume everything here is due to coupling from input beam jitter through RMS or DC misalignment of the IMC into RIN in transmission of the IMC.
From the difference in magnitude of the aligned and misaligned spectra we can make some estimates of the RMS alignment offset in each D.O.F. Since the jitter to RIN coupling is linear with alignment offset, the RMS offset in each D.O.F. can be estimated as follows: RMS offset = aligned RIN / (misaligned RIN - aligned RIN) * misaligned offset. For the D.O.F.s studied this comes out as:
| Mirror D.O.F | IMC waist D.O.F. | Mirror D.O.F. misaligned offset [urad] |
alignedRIN/ (misalignedRIN-alignedRIN) |
RMS Mirror D.O.F. offset [urad] | RMS IMC waist D.O.F. offset | RMS relative HG10 mode amplitude |
| MC2 Pitch | Pitch Shift | 30 | 0.082 | 2.5 | 68um | 0.032 |
| MC1/3 diff. Yaw | Yaw Shift | 40 | 0.15 | 6 | 97um | 0.046 |
| MC1/3 diff. Pitch | Pitch Tilt | 100 | 0.026 | 2.6 | 1.8urad | 0.011 |
| MC2 Yaw | Yaw Tilt | 25 | 0.047 | 1.2 | 3.0urad | 0.019 |
I'll continue with some more analysis of the data, and write it up in LIGO-T1300378.
Daniel, Sheila, Alexa
The TMS Baffle PD 1 220.6 PIT -227.1 YAW 2.24 Volts at 0dB gain. I saved this as the misalignment state.
Baffle PD 4 (channel 3) 289.9 PIT -289.0 YAW 2.47 V 0dB gain.
Aligned position:255.25 PIT -258.05 YAW saved as aligned state. This is a change of -0.6 PIT -0.45 YAW from yesterday and alog 10001
Then I moved PR3 using the single shot beam on the camera, it seems centered now. (PIT -244, YAW -253.98 according to PV info) However, there is something funny about the sliders on PR3: the text entry button says -253, while the number above the slider says -253.
Apparently Epics doesn't know how to round.
To get cavty flashes I moved the ETM to 243.0 PIT, 74 YAW, these are now saved as alinged.
Tha cavity locked and the IAL servos are running now.
Performances of the ETMX and ITMX ISI were consistent between yesterday evening and this evening, cf the first two attached plots. Although, the ground motion was not that different (cf comparison of the seismometers in PEM_CS and PEM_EX).
First two attachements are showing a comparison of oplev Pitch and ST2 GS13 for ITM and ETM
yesterday night in green (gps 1076202015 2014-02-12 01:00:00 UTC)
this evening in red (gps 1076301556 2014-02-13 04:39:01 UTC)
Last two attachements are a comparison of the seismometers in Corner station and End X
yesterday night in red (gps 1076202015 2014-02-12 01:00:00 UTC)
this evening in blue (gps 1076301556 2014-02-13 04:39:01 UTC)
I added to yesterday's plot the motion of the ground and ISI from this morning (gps 107636178), in a noisier environment (between 0.5Hz and 8Hz and below 0.1Hz, due to wind ?), see green curve in the PEM signals (PEM_EX PEM_CS) and pink curve in ISI signals (ITMX_perf and ETMX_perf).
J. Kissel, A. Staley, S. Dwyer In order to trouble shoot her ALS noise model, Alexa took an equivalent measure of the MCL / MCF crossover model using the fast path, where as I chose the slow path. She discovered that the crossover transfer function was a factor of two higher than mine (see 10064). This was notably before I changed the MC2 compensation filters (see 10043). After I changed the filters, I've taken the slow-path version of the IMC_L / IMC_F crossover transfer function again, and can confirm Alexa's result. This confirms that the crossover is border-line stable up at 25 [Hz], uncomfortably close to the 29 [Hz] Roll notch. This may explain why ALS CARM handoff has been so difficult last night and today. Suspecting utter debauchery between the two measurements only two days apart, Sheila and I confirmed the following: - All digital settings are identical to what I'd described in LHO aLOG 10042 - The 9 MHz modulation depth is set at the nominal +10dBm - The overall open loop gain has not changed We tried to check if the demodulation phase had rotated, but the test was inconclusive because we couldn't get any I signal from the demodulator for unknown reasons. We also measured the raw, peak-to-peak voltage of the RF output of the photodiode, and got ~250 mV. Clearly a more thorough investigation is in order. :-(. In the mean time the mode cleaner remains locked...
Blue Team
We ramped up the effort to get the geen ALS WFS used for the One Arm Test on ISCT EX, to help us with the alignment-induced noise between red and green.
Here is where we stand at the moment:
- Electronics: mostly installed, see entry by Filiberto and Daniel
- WFSs + cables + patch panel: found, in the EX VEA
- Picomotors, breakout box and cables : found (breakout box already at EX), picomotors mirrors on Evan's desk
- New WFS layout designed by Jax and Evan: lenses, mirrors, beam splitters -- all in hand
- Cabling: Mostly done by Filiberto between the field rack and the remote rack; the cables between the optics table and the field rack need to be made - list sent to Filiberto
Plan for tomorrow/Friday (in parallel with Filiberto's work)
- Get new optics, WFS and picomotors on the table (when the green team is not working)
- Install patch panels and cables as they become available
- Update electronics drawings
- RCG model update
A quick outline of the proposed mode-matching solution:
We are working from D1100607. This is the layout for ISCTEY, but is (I am told) a pretty good description of the current state of ISCTEX (moreso than the ISCTEX document, anyway).
According to Sheila, mirror M16 is some sort of beamsplitter (labeled as 90/10, but measurements show more like 50/50). The transmission of M16 goes onto a high reflector, which then goes onto a QPD. This is meant to be temporary, and we intend to remove it when laying down the WFS. Apparently there is 6 mW going through M16. In order to get < 1 mW on the WFS, we will probably replace the high reflector with a 90/10 BS and a dump in transmission.
We will propagate the reflection of the 90/10 BS through a lens and then onto a 50/50 BS. The reflection of the 50/50 BS will go onto a picomotor-mounted high reflector, and then onto a WFS. The transmission of the 50/50 BS will then go onto two folding mirrors, then a picomotor-mounted high reflector, and then the other WFS (or maybe we will make one of the picomotor mirrors do double duty as a folding mirror).
An a la mode optimization plot is attached. The position of M16 is taken to be 0 m. The 200 µm waist at -0.41 m is taken from the ISCTEY document; there is no guarantee that it's actually there. alm says to place an f = 544 mm lens at 0.27 m. This gives a 400 µm waist at 1.35 m. The Rayleigh range is then 0.94 m. For optimal Gouy phase separation and spot size, we want to place each WFS one Rayleigh range away from the waist; i.e., one at 0.41 m and another at 2.29 m.
The spot size on each WFS should be 560 µm, giving a diameter of a little over 1 mm.
J. Kissel I've found a window to stick in the proper coil driver compensation filters into MC2. As described in LHO aLOG 9956, the compensation now correctly compensates the modified M2 triple acquisition driver, FM1 zpk([13],[65],1,"n") FM2 zpk([11;21],[1;210],1,"n") FM6 zpk([65],[13],1,"n") FM7 zpk([1;210],[11;21],1,"n") The Binary IO filter logic had been changed long ago, so everything works as designed now. Also, I've cleaned up and removed what was in FMs 3 and 8, which was redundant with FM6 and FM7. The new filter file has been committed to the userapps repo.
J. Kissel We've discovered that the H1 SUS MC2 compensation filters are incorrect for the M2 stage coil driver, having never been changed after the analog coil driver was swapped out for a beefier drive (see LHO aLOG 9956). It turns out, by lucky coincidence, we've been operating the new coil driver in a state (State 1) where it's frequency response is flat, and the compensation filters -- though wrong -- cancel each other to also be flat, so the overall crossover loop gains remained unaffected. Just to be sure, however, I too a "pre" measurement of each crossover of the IMC_L loop gains. I attach the results, and summarize below. Crossover UGF [Hz] Phase Margin [deg] MC_L/MC_F 14.5 39 M2/M3 7 38 M1/M2 0.125 36 These numbers are comparable to LLO's numbers (see, e.g. LLO aLOG 4331). Templates can be found here: /ligo/svncommon/SusSVN/sus/trunk/HSTS/H1/MC2/Common/Data/ 2014-02-11_H1IMC_M1-M2_Crossover_OLGTF.xml 2014-02-11_H1IMC_M2-M3_Crossover_OLGTF.xml 2014-02-11_H1IMC_MCL-MCF_Crossover_OLGTF.xml
For posterity, the IMC loop had the following in its path:
Input Mode Cleaner, Common Mode Servo Board Settings
Common Path
- H1:IMC-REFL_SERVO_IN1GAIN = 9 [dB]
- p:z 40Hz:4kHz compensation filter ON
- p:z 1kHz:20kHz first stage boost ON (second and third stage boosts OFF)
- p:z 4kHz:17kHz filter ON
Fast Path
- H1:IMC-REFL_SERVO_FASTPOL = +
- H1:IMC-REFL_SERVO_FASTGAIN = -6 [dB]
- p:z 140kHz:70kHz highpass ON
Slow Path
- p:z [100 100]:[10 10] "generic filter"
LSC Model Settings
IMC_L Filter
- FM1; antiWhite = p:z [10 10]:[100 100] (compensates for "generic filter")
Gain = 1.0
LSC-MC Filter
- All filters OFF (when locked)
Gain = 1.0
SUS MC2 Model Settings
All coil driver states set to 1.
M3_ISCINF
- FM 6; notch_R3 = notch at highest roll mode of MC2
- FM 7; notch_V3 = notch at highest vertical mode of MC2
Gain = 1.0
M3_LOCK
- FM 3; z40:p140
- FM 9; CLP300 = Chebychev low pass at 300 [Hz]
Gain = -300
M2_LOCK
- FM3; z1:p0.1
- FM4; z1:p100
- FM10; ELP70 = elliptic low pass at 70 [Hz]
Gain = 0.1
M1_LOCK
- FM1; z0.01:p0
- FM2; z:p0.01,0.1
Gain = 1.0
I repeated the MCL-MCF crossover measurement that Jeff made since I could not get it to agree with my model. I made this measurement with the SR785 by injecting a source into excitation B through the slow path of the MC common mode board and looking at the ratio of test 1 and test 2. The results still differ from my model (likely something missing in the model..), but more suprisingly I found that this measurement showed a factor of 2 greater in magntiude in comparison to Jeff's open loop transfer function. See attached picture.
I caused a HEPI trip this evening, messing with arm cavity slow feedback to HEPI.
While trying to restet this, I once again became annoyed by how the isolate script sets the current setpoint to something other than the target setpoint for Ry and RZ. Forgetting this is part of why it took us several hours to recover from the hepi pump trip this afternoon, and it seems like by continuing to have the script do this we are shoting ourselves in the foot. I have heard that the script does this because of fear that HEPI can't handle these large offsets, but we have been using these large target positions for more than a month, so it seems HEPI can handle them. If that's true, is it possible to fix this script?
Having a closer look at the offsets, I noticed that the location was not coming back to the set point even after waiting a long time. I attached a plot of the target location over the last 60 days, it has really never been at the set point.
AS far as I understood, this after noon after our long struggle to get ETMX back after the hepi pump trip, Jim got it back one DOF at a time using 250mHz blends on stage 1, Tcrappy on stage 2, and using the 60 Hz notch in the stage 2 controllers (which is for level 3 but while we are still using level 2 controlers we need to engage FM5 by hand). (correct me if I'm wrong Jim)
Instructions for what worked for me now are:
HEPI:
untrip watchdog
Comands 2, Isolate level 1
Important! Check that DC bias current setpoints match the target setpoints.
ISI:
Set T240 watchdog thresholds to something large
Set stage 1 blends to 250, isolate level 3.
Wait a long time, get distracted by other work, come back in a half hour....
change stage 1 blends to Tcrappy
Isolate stage 2 level2 with Tcrappy
All set!
For the record, it's not that *HEPI* can't handle the large offsets, it's that the Stage 1 T240s on the BSC-ISI can't handle the large offset without saturating, and therefore tripping the ISI's watchdog. The seismic group has heard this complaint loud and clear and is considering if they even need to include the T240s in the watchdog system. We promise we're trying to make it better!
I forgot to say, the reason I was messing with the feedback from ISC to HEPI in the first place was that Brett and I noticed while he was making his measurement that the tidal relieve servo was unstable....
I will try to fix this in the morning.
Like Jeff said, we limit the transfer of the Target of HEPI in the Current Setpoint to 500 nrad to prevent saturating the ISI's T240s. One has then to copy the targets into the current setpoint manually if he wants to override this. We could consider removing this limit if we increased the T240 WD thresholds on the ISI at the beginning of the turn on process (starting with HEPI), and had the T240s removed from the blend when turning the ISI on.
Sheila mentioned another interesting behaviour that we noticed yesterday: even when the Target and Current Setpoint agree, the position on RY does not go all the way to the setpoint value along RY on ETMX. (See attached plot, and Cart Bias MEDM screen). It is not the case on ITMY and ITMX HEPI, but none of them request such high target values (see attached Cart Bias MEDM sceeens for HEPI ITMX and ITMY). One could argue that parts of HEPI may start contacting (each HEPI has a slightly different Range Of Motion which depends on how it was set up) and prevent the servo loops from pushing HEPI to the requested Current Setpoint. This would probably make the position loops go unstable though, and it is not the case. As a side-test, we increased the value of the current setpoint on RY to see if the location would follow and it does, which indicates that HEPI is not stuck contacting.
This oddity has been going on for 60 days, but has also remained consistent throughout, even though HEPI, and its pumps, got restarted more than once. HEPI alignment along RY, once controlled, has not changed for those 60 days.
As far as having trouble turning the ISI on, the shift between the floating, and requested controlled position are quite high. We tried making this smaller, which made the ISI turn on way easier, and request the needed alignment to SUS, which seems to have worked. Should we do it this way from now on, for the sake of making the ISI turn on process more repeatable on ETMX?
The HAM-ISI models, scripts and screens were successfully updated.
The update was tested on HAM2-ISI, before being ported to the other platforms. Test performed are:
Foton files had to be updated on HAM2 and HAM3 so the level 3 controllers would be in the correct filter banks for the updated command scripts to work. Filters were not modifed.
The content of the update is detailed in DCC T140012. See the attached Update_Log.txt For more details about the sequence of the update at LHO.
Work was performed under WP #4434, which is now closed
Restarted the DAQ to support Hugo's HAM ISI model changes and incorporate Daniel's latest X1PLC2 channel list.
Installed the following electronics at EX per D1001423-v18. D1002559 ISC Whitening Chassis for WFS_A D0902796 WFS I/Q Demodulator WFS_A D1100905 Demodulator Patch Panel D0902796 WFS I/Q Demodulator WFS_A D1002559 ISC Whitening Chassis for WFS_B D1100749 Legacy iLIGO WFS PD Interface Items still needed for tomorrow: 1. Power up electronics 2. Finish running long field cables to 384 Channel Acromag Binary and AA chassis 3. RF cabling from field rack to ISCTEX table Filiberto Clara
Modifications to the slow controls system were made to add WFS whitening channels and WFS demodulator readbacks.
I swapped in a new HTTS filter file with corrected dewhitening filters (what claimed to be 10:0.4 had actually been 40:1.6). I checked that RM1 and RM2 still damp stably.
I've started on a round of undamped TFs on RM1 and RM2 to check that I get the expected results without post-hoc corrections.
The TFs were taken successfully. The raw data was processed with meas.badFilt = '' (no post-hoc filter correction) and meas.E1201027 = true (coil driver resistors replaced):
^/trunk/HTTS/H1/RM1/SAGM1/Results/2014-02-12_1330_H1SUSRM1_M1.mat
^/trunk/HTTS/H1/RM2/SAGM1/Results/2014-02-12_1330_H1SUSRM2_M1.mat
These data sets are plotted in the attached PDF with the latest L1:RM1 and L1:RM2 results and the immediately previous H1:RM1 and H1:RM2 results of 12/20/13. As expected, the new results without post-hoc filter correction agree with the earlier results with correction, showing that the new H1SUSHTTS.txt file is good. (It has been committed.)
The agreement with the model is good for H1:RM1 so this new data (2014-02-12_1330) can count towards Phase3b testing. The RM2 data shows the same strong coupling of L into Y as the last few measurements, and this needs to be investigated at the next vent.
LLO should impement the same fix as soon as convenient. Specifically, in L1SUSHTTS.txt, all the ???_M1_OSEMINF_?? modules should have Section 1 set to have label '10:0.4' (probably that way already) and filter zpk([10],[0.4],1,"n") (will have been zpk([40],[1.6],1,"n")).
