Displaying reports 72941-72960 of 83266.Go to page Start 3644 3645 3646 3647 3648 3649 3650 3651 3652 End
Reports until 19:15, Sunday 09 March 2014
H1 SEI (ISC, SUS, SYS)
jeffrey.kissel@LIGO.ORG - posted 19:15, Sunday 09 March 2014 (10640)
Ground, HEPI Pier, and ST1 ISI Motion During 2014-03-06 High Winds
J. Kissel

I've grabbed spectra from all possible degrees of freedom of the ground, HEPI pier, and ISI ST1 for the four operational ground sensors and the four operational BSC-ISIs during a very windy and noisy afternoon this past Thursday, Mar 6 2014 at 22:15 UTC. This will hopefully assist modelling efforts to design blend filters that might be able to withstand such input noise. Unfortunately, because Sebastien was commissioning a few chambers, and we had to reboot ETMX to get his GND T240 stored in the right place, there wasn't any long stretch of time during that day in which all platforms were at comparable performance (hence the differences in the ISI ST1 performance). Also, I'm not sure I believe what the rotational HEPI L4Cs are saying at the microseism.

The spectra are attached, but they were created by the following DTT templates,
/ligo/svncommon/SeiSVN/seismic/BSC-ISI/H1/Common/Data/
2014-03-06_2215UTC_GroundMotion_ASDs.xml
2014-03-06_2215UTC_HEPIPierMotion_ASDs.xml
2014-03-06_2215UTC_ISIST1Motion_ASDs.xml
and they were also exported to the following files
/ligo/svncommon/SeiSVN/seismic/BSC-ISI/H1/Common/Data/
2014-03-06_2215UTC_Windy_GroundMotion_ASDs_ITMY-XYZ_HAM2-XYZ_HAM5-XYZ_ETMX-XYZ.txt
2014-03-06_2215UTC_Windy_ISIST1Motion_ASDs_BS-XYZRXRYRZ.txt
2014-03-06_2215UTC_Windy_ISIST1Motion_ASDs_ETMX-XYZRXRYRZ.txt
2014-03-06_2215UTC_Windy_ISIST1Motion_ASDs_ITMX-XYZRXRYRZ.txt
2014-03-06_2215UTC_Windy_ISIST1Motion_ASDs_ITMY-XYZRXRYRZ.txt
2014-03-06_2215UTC_Windy_PierMotion_ASDs_BS-XYZRXRYRZ.txt
2014-03-06_2215UTC_Windy_PierMotion_ASDs_ETMX-XYZRXRYRZ.txt
2014-03-06_2215UTC_Windy_PierMotion_ASDs_ITMX-XYZRXRYRZ.txt
2014-03-06_2215UTC_Windy_PierMotion_ASDs_ITMY-XYZRXRYRZ.txt
where the last part of the file name indicates the order in which the channels have been exported.
Non-image files attached to this report
H1 ISC
sheila.dwyer@LIGO.ORG - posted 18:16, Sunday 09 March 2014 (10639)
Refl DC bias path

Alexa, Sheila

We tried engaging the refl DC bias path with the 1.6Hz pole :40Hz zero engaged.  We were able to measure a tranfer function using 0dB on the CM in1 gain, and -80 in REfl DC bias, which has coherence to about 2Hz. (the UGF is around 3.5Hz).  We tried to push the gain up and engage the boost to get a better measurement, but now we seem to have an earthquake from Mexico tripping our ISIs, so we are done for the day.

Images attached to this report
H1 SEI (SEI)
jeffrey.kissel@LIGO.ORG - posted 18:14, Sunday 09 March 2014 (10638)
H1 ISIs Trip on Earthquake
All ISIs except for the BS tripped Mar 10th 2014 ~00:58 UTC, most likely from 6.3 Mag Mexican Earthquake (see USGS Page), which happened at Mar 10th 00:38:20 UTC. *sigh* Well, THAT's annoying. Delightfully however, once untripped, the guardian brought the HAM2 and HAM3 ISIs back up to full isolation. 
H1 ISC
sheila.dwyer@LIGO.ORG - posted 15:58, Sunday 09 March 2014 - last comment - 07:20, Monday 10 March 2014(10637)
PRMI+X arm alingment

Yuta, Sheila, Alexa

Today we found an alignment that should be acceptable to both PRMI and the Xarm, in the hopes that the green team and red team will not each have to change PR2 significantly each day. 

First Yuta aligned PRX, then we aligned the Xarm, locked the IR to the X arm, and moved PR2 and IM4 to point the beam down the xarm.  After this the alignment of PRX was way off and unlockable.  We checked that the beam was centered on POPAir B, then restored PR2 to the intial PRX alignment.  Then Yuta aligned PRM by hand, and walked PR2 back to the arm alignment. Finally Yuta realinged PRY.  A screenshot of the final positions is attached.

Yuta has changed the dither guardian script for PRX so that the PR2 dither is now fed back to PRM instead of PR2. 

Hopefully this means that the PRMI and arm alignments will stay compatable and both the red team and green team don't have to spend as much time aligning each day. 

Images attached to this report
Comments related to this report
kiwamu.izumi@LIGO.ORG - 07:20, Monday 10 March 2014 (10644)

Just an observation from this morning:

I wanted to see how much power build up the PRC can achieve without touching the alignment. I reverted PRM, BS and ITMY to the values that Sheila had in her screen shot because their alignment wasn't saved and they had come back to some past values by the guardian. The highest build up I saw was about 12 uW in POPAIR_B_RF18 which is about 6-7 times smaller than the highest in the past. Not bad.

Note that ITMX, ETMX and TMXS had different alignment biases than that in the screen shot. I assumed that there had been a fine touch by the green team and I didn't revert them to the screen shot values. The green light mostly stays on a 01 mode and occasionally on a 00 mode.

H1 SEI (PEM)
jeffrey.kissel@LIGO.ORG - posted 15:11, Sunday 09 March 2014 - last comment - 15:23, Sunday 09 March 2014(10635)
Calibration for ETMX GND Channels Corrected for Temporary T240
J. Kissel

While gathering data of the ground motion from Thursday's windy afternoon, I found that the calibration filter used to turn the ground inertial sensor into (asymptoting to) 1 [nm/s] (at high-frequency) for the new, temporary, T240 at End X (see LHO aLOGs 10594, 9758, and D1400077) was incorrect. The "Cal" filters in FM1 of the H1:ISI-ETMX_ST1_GNDSTSINF_[A,B,C]_[X,Y,Z] banks were ON and still calibrating for an STS2 read-out chain, i.e.
1 / (1500 [V / (m/s)] * 40 [V/V] * 2^16/40 [ct/V] * 1e-9 [(m/s) / (nm/s)]) = 10.1725 [(m/s) / ct]
These filters were most likely installed and turned on by some script that was preparing for the permanent solution; not surprising.

I've installed a new filter in FM2 of all nine banks, called "T240Cal" with the following gain:
1 / (1200 [V / (m/s)] * 2 [V/V] * 2^16/40 [ct/V] * 1e-9 [(m/s) / (nm/s)]) = 254.3132 [(m/s) / ct]
where I've assumed that the Trillium Interface Chassis is configured in its low-gain mode, with a gain of 2.0. 

I attach a spectra of data in the past, comparing all four ground inertial sensors, with the ETMX GND sensor corrected for the factor of 
(254.3132 [T240] / 10.1725 [STS]) = 25.0 [T240/STS]
difference in gain between the STS and T240 calibration. Where we expect the ground motion to be coherent, namely between 0.1 and 0.5 [Hz], the ETMX sensor matches all three corner station's sensors exquisitely, and the micro seismic peak is roughly 1e-6 (m/s)/rtHz as expected.

As of now, until the T240 is replaced with an STS2 as designed, the ETMX_ST1_GNDSTSINF_[A,B,C]_[X,Y,Z] filter banks should have FM2 ON and FM1 OFF. Once the T240 is replaced by an STS2, we can switch back to using FM1, and "T240Cal" should be removed from FM2 to avoid confusion.

All calibration values were obtained from D1001575.
Non-image files attached to this report
Comments related to this report
jeffrey.kissel@LIGO.ORG - 15:23, Sunday 09 March 2014 (10636)
As of this entry (roughly 2014 March 9th 22:15 UTC) the channels
H1:ISI-GND_STS_ETMX_X_DQ
H1:ISI-GND_STS_ETMX_Y_DQ
H1:ISI-GND_STS_ETMX_Z_DQ
are therefore correctly calibrated version of the end stations ground inertial sensor, i.e. 1 [(nm/s) / ct] (above ~[4 mHz]). As with any STS / T240, the signal is trustworthy up to ~30 [Hz], where the instrument's response is rolled off internally.

Also, I've committed the current filter file with the correct gains into the userapps repo, 
/opt/rtcds/userapps/release/isi/h1/filterfiles/H1ISIETMX.txt, 
removed the local copy in the chans directory, and made it a soft link to the repository version,
controls@opsws2:chans 0$ ls -l H1ISIETMX.txt 
lrwxrwxrwx 1 controls controls 60 Mar  9 15:16 H1ISIETMX.txt -> /opt/rtcds/userapps/release/isi/h1/filterfiles/H1ISIETMX.txt
controls@opsws2:chans 0$

(So far this is the *only* ISI to have it's filter file soft-linked.)
LHO General
corey.gray@LIGO.ORG - posted 15:33, Saturday 08 March 2014 (10634)
Saturday Tour

Just finished up the March Public Tour.  I took the group up on the roof around 2:50-3:00pm.

 

H1 ISC
keita.kawabe@LIGO.ORG - posted 01:06, Saturday 08 March 2014 (10632)
Green WFS today

Engaged DOF1P, DOF1Y and DOF2Y, dithered ITM and ETM, and measured the sensing matrix based on OL (assuming that OL calibration is now correct).

  WFSA WFSB
ITMP 82.8 cts/urad 901.8
ETMP -430.8 -1247.8
ITMY -587 671
ETMY -504 292

The sign follows the SUS convention that positive PIT = mirror points down, positive YAW = counter clockwise viewed from the top.

As was observed before, WFSA is a good YAW hard mode sensor, WFSB is a good PIT hard mode sensor.

Since neither of the WFSs is good soft mode sensor, I made these:

DOF2P = -4.19 WFSAP + WFSBP

DOF2Y = WFSAY + 1.1 WFSBY

After making these, all four DOF loops were closed and it worked, with one caveat: It seems like DOF2P wants a large offset in order for the green transmission to be maximized. I tried +2500 cts offset and it was good (transmission touches 850 cts). Without this, the transmission stays between 650 and 750.

Unfortunately I cannot give DOF2P an offset before the master switch, and I put this offset in the feedback filter itself. When IFO unlocks the integrator will keep integrating, destroying the IFO alignment.

So I disabled DOF2P input AND offset, and left other three DOFs, for tonight.

It's still slow.

H1 ISC
sheila.dwyer@LIGO.ORG - posted 23:54, Friday 07 March 2014 - last comment - 09:50, Saturday 08 March 2014(10629)
Green arm tonight

Daniel, Alexa, Keita, Arnaud, Sheila

Tonight we measured the same noise spectrum of the normalized arm PDH spectrum that we measured tuesday night, with the WFS on at low gain.  The noise was not improved, it even seemed worse. 

We then tried feeding the normalized PDH back to the CM board though input 2.  With a gain of 18 in -500 in REFL_DC_BIAS and 0 in the CM board Input 2 and positive polarity.

We saw that the error signal was supressed.  The transfer function we measured for this loop didn't make sense, we measured a ugf of 500Hz and 180 degrees phase.  We weren't able to calibrate the nosie spectrum we then measured, not trusting the measurement of the loop gain. 

The transmitted power fluctations were reduced by adding the feedback from the IR, but still fluctuated from 1 to 0.5, at around half a hertz. 

The IR was locked to the cavity starting at around 6:30 UTC, most of the time until the ITMX ISI tripped at around 7:30  UTC.  I reset the target offsets before bringing it back up, after which Alexa realinged the cavity by moving ITMX 2.5urad in pitch.  Trying to bring it up I tripped it twice more trying to reengage the Tcrappys (even though everything was green, it seems we need to wait until the number are around a few hundred.) Arnaud brought it back by switching to Tcrappy one DOF at a time on stage 2, and sensor correction is now left off. 

 

 

Comments related to this report
daniel.sigg@LIGO.ORG - 09:50, Saturday 08 March 2014 (10633)

Turns out we forgot to take into account that we turn the boost gain on in the path from the PLL. This would have made the two paths flat in respect to each other from 1.6Hz to 80Hz, rather than the desired 1/f.

H1 ISC
alexan.staley@LIGO.ORG - posted 21:08, Friday 07 March 2014 (10628)
COMM PLL Noise

I took another amplitude spectrum of the noise out of the PFD IMON in the COMM PLL to VCO loop. The PLL was locked; however we had not handed off. The two compensation filters were on and the input gain was 27dB. I have attached the data. The flatness which we were puzzled by last time has disappeared...

I did also see the traveling noise that seems to come from electronics cross-talk. This moving peak was about 3kHz wide (approzimately 12 degrees).

Non-image files attached to this report
H1 SUS (ISC)
jeffrey.kissel@LIGO.ORG - posted 17:22, Friday 07 March 2014 (10625)
H1 SUS ITMX UIM/L1, H1 SUS ITMY PUM/L2, and H1 SUS ITMY UIM/L1 Coils Balanced
J. Kissel

Following the same procedure outlined in LHO aLOGs 9453 and 9079, I balanced the coils on SUS ITMY UIM/L1, ITMX PUM/L2, and ITMX UIM/L1. The final balanced gains are

             ITMX UIM     ITMY UIM    ITMY PUM
UL            -0.965      -0.997       +1.028
LL            +1.025      +0.976       -0.911
UR            +0.970      +1.019       -1.091
LR            -1.030      -0.998       +0.968

The precision on the ITMX UIM and ITMY PUM numbers is within the usual +/- 0.5%, but for some reason I was able to get a crazy amount of SNR on ITMY UIM, so the numbers are good to with 0.05%. 

This balancing has reduced the L3 P and Y caused by pringle excitation at 4 [Hz] by
Optic    Stage      DOF       Reduction Factor @ 4.0 [Hz]
ITMX      UIM        P               > 24.2 (peak below noise, totally incoherent)            
                     Y               > 1.39 (already balanced well, totally incoherent after balance)

ITMY      PUM        P               25.6 (peak still 95% coherent*)
                     Y               37.4 (peak still 80% coherent*)

ITMY      UIM        P               > 124.2 (peak still 68% coherent, but buried)
                     Y               19. 0 (peak still 90% coherent)

* For ITMY PUM, I could not reduce the Q phases to as small as I normally get, implying there is some other noise contributing to the imbalance unrelated to the knob I'm tuning. Unclear what that could be, but I guess this is why I couldn't reduce the imbalance low enough that the excitation became buried in the noise as I could with the other stages.

I attach the figures of merit for reducing the imbalance. All SUS have had these values captured in there safe.snaps, which have been committed to the repository... except for ITMY PUM which I found not stored while writing this entry. Will store them ASAP.
Non-image files attached to this report
H1 INS (SEI)
jim.warner@LIGO.ORG - posted 17:20, Friday 07 March 2014 (10626)
More tf's running on ETMY ISI

Running from OPSWS0. Already had a peek with DTT, and no change after todays fiddling with TMS cables. But, maybe more data will tell us...more. TMS is plugged in and damping now, as well as the quad.

LHO VE
kyle.ryan@LIGO.ORG - posted 17:17, Friday 07 March 2014 (10624)
~1330 hrs local -> Pumped down flex-line connecting RGA to 10" gate valve on BT near GV6
PT124B increased(ing) as a result -> The removal of atmospheric pressure from the O-ring valve's O-ring may have released some dissolved gas -> Flex line connecting RGA to 1.5" O-ring valve doesn't look to be applying any torque to 1.5" valve -> I will be monitoring PT124B from home
H1 SEI
hugh.radkins@LIGO.ORG - posted 16:51, Friday 07 March 2014 (10623)
WHAM4 SEI HEPI Progress/Status

Got the final Actuators attached to HAM4 today.  I still need to install the position sensor that was experiencing an interference.  And then access the total position before final closeup.

H1 SEI (AOS, INS)
hugh.radkins@LIGO.ORG - posted 16:47, Friday 07 March 2014 (10622)
WBSC10 ETMY SEI HEPI Springs adjusted for ACB Load

Tweeked the DSCW Springs to pull the system back up from the additional ACB weight.  We adjusted things attempting to bring the Dial Indicators back to the post final alignment numbers Jim recorded on 5 March.  There is a 10mil (1/4mm) west shift but we are still well within the +-3mm tolerance.  Otherwise we have less than 0.1mm vertical shift and much less than that N/S.

So, ready for HEPI Actuators.

H1 ISC
jaclyn.sanders@LIGO.ORG - posted 16:41, Friday 07 March 2014 (10621)
EY RF power adjustment (and asst repairs)

(Jax, Daniel)

Today we measured the RF levels in the ISC field and remote racks at EY and set nominal RF values in the MEDM interface. 

Installed attenuators by cable number:

2dB: 

18-1B2 (RF to phase modulator), 18-1, 13B (RF Preamp to Phase/Freq Discriminator), 14-2B (RF to VCO), 19-1 (71 MHz RF Dist. Amp to Oscillator), 14-1 (24.4 MHz RF Dist. Amp to Oscillator)

3dB: 

16-1 (Freq Divider to Phase/Freq Discriminator)

8dB:

17-1 (VCO to Freq Divider)

15 dB*:

18-5B (RF to VCO)

*Here we would have preferred 12 dB to be consistent with EX as per alog 9466, but didn't have one on hand in the big box o' attenuators.

---

During this process, we discovered the delay line had no output. After cracking it open, we found that one of the ICs (U22) had fried in spectacular fashion - likely a victim of an upside-down power cable at some point. I grabbed a spare (S1103442) from MY, gave it a quick once-over to make sure it works, then installed it at EY.

LHO General
gerardo.moreno@LIGO.ORG - posted 16:38, Friday 07 March 2014 (10620)
Operation Summary

8:30 am, HFD department, fire hydrant flushing/testing.
9:00 am, Mitchell, Travis and Andres to Y-End, ACB install.
9:00 am, Aaron connect cables for new chassis, CM summing chassis by PSL area.
9:30 am, HFD department, second unit to check on RAFAR boxes ====> done by 11:15 am
10:08 am, Cyrus and Jim to Mid-Y, rack and stack work per WP#4466.===> break for lunch at 12:15 pm.
10:00 am, Thomas and Greg to LVEA North bay area, prep HEPA filter fan unit for craning.
10:55 am, Hugh to LVEA HAM4 area, HEPI installation.
11:21 am, Alexa, X-End, field rack measurements ====> done by 11:53 am.
11:30 am, Thomas and Greg to North bay area, crane HEPA filter on to a table.
11:37 am, Kyle to X-End, check up on access.  To Y-End, check up on purge air====> done by 12:10 pm.
11:40 am, Apollo crew moving ISI to high bay area ====> done by 11:59 am.
1:00 pm, Karen to Y-end, cleaning.
1:09 pm, Keita to Y-End, TMS cabling status.
1:24 pm, Andres to LVEA South and West bay area, hunting for SUS components ====> done by 2:28 pm.
1:30 pm, Alexa and Sheila to X-End, LVEA area.
1:48 pm, Jim and Cyrus, back to Y-Mid to continue with rack and stack ====> done by 3:14 pm.
1:50 pm, Hugh to LVEA HAM4 area, continue with work, then I saw him at Y-End, so who knows.
4:00 pm, Kyle done "making noise" West bay area of LVEA, on elevated BT slab near GV6.
 

H1 AOS (AOS, ISC, TCS)
thomas.vo@LIGO.ORG - posted 15:53, Friday 07 March 2014 (10617)
ITMX and ETMX OpLevs

THESE CALIBRATION FACTORS HAVE CHANGED

Per Keita's ALOG 10331 and 10454, I have adjusted the optical lever calibration gains for ETMX and ITMX to reflect a more accurate method of calibration that was performed by Keita with the baffle diodes.

ETM

  Old Delta New
Pitch 76.7 0.932 71.4844
Yaw 65.3 0.863 56.3539

 

ITM

  Old Delta New
Pitch 23.4 2.081 48.6954
Yaw 24.7 2.287 56.4889
H1 SUS (ISC)
jeffrey.kissel@LIGO.ORG - posted 19:57, Wednesday 22 January 2014 - last comment - 17:30, Friday 07 March 2014(9453)
H1 SUS PR2 Coil Balancing Complete
J. Kissel [with lots of help from K. Kawabe, K. Arai, S. Ballmer, and K. Izumi]

I've balanced the coils on the M2 and M3 stages of the H1 SUS PR2 using Keita's Technique (see LHO aLOG 9079 -- which, now that I understand -- I'll make sure to supplement this aLOG with kLOG comments about it). However, before I exercise my didactic tactics, the answer is:

H1 SUS PR2
Channel     Balanced COILOUTF Gain
M2 UL            -0.994 
M2 LL            +1.039 
M2 LR            +0.962
M2 UR            -1.005

M3 UL            -0.962
M3 LL            +1.043
M3 UR            +0.954
M3 LR            -1.034

I attach the results of the balancing as measured by the M3 OSEM sensors behind the optic, one for each stage of drive balancing. The left two panels of each attachment (Amplitude Spectral Density and Coherence) show the performance AFTER the balancing, and the right two panels show the performance BEFORE the balancing, where coefficients were just set to +/- 1.0. This balancing has reduced the coupling (at 4.1 [Hz]) as follows (using the M3 OSEMs as the figure of merit, which -- details in notes below -- are imperfect):

   DOF                  Reduction Factor @ 4.1 [Hz]
M2 Pringle to M3 P           > 178               (peak is in the noise, and only ~60% coherent)
M2 Pringle to M3 Y             35

M3 Pringle to M3 P             1.4
M3 Pringle to M3 Y             4.5

The next step is to take a full suite of M2 and M3 L/P/Y to P/Y, "off-diagonal" transfer functions with the newly balanced coils, as has been done with *unbalanced* coils on H1SUSBS and H1SUSPRM.

I have captured a new safe.snap that includes these new gains and committed it to the userapps repo.

Expert Notes for next time:
- Three different suspensions, three different people, and three different options for sensors resulted in three different details of how the the process was done, but the process is the same, in principle. With PR2, the only option for sensor demodulation were the sensor side of the OSEMs at the bottom stage. Because these sensors are not perfectly balanced, the precision to which I could improve the balancing was limited -- much more so on the M3 stage than the M2 Stage. I'll explain the details below.

- These coefficients were established to higher precision, but were rounded off to the nearest 1000th's digit because they will be easier to track, entirely visible on the MEDM screen, and the higher precision had little-to-no affect on the goodness of balancing. 

- The templates for measuring the performance can be found here:
/ligo/svncommon/SusSVN/sus/trunk/HSTS/H1/PR2/Common/Data/
2014-01-22_H1SUSPR2_M2_CoilBalancing.xml
2014-01-22_H1SUSPR2_M3_CoilBalancing.xml

- The script used to perturb the coil balancing based on the results of the LOCKIN tool (authored by Kiwamu, made slightly more generic by me),
/ligo/svncommon/SusSVN/sus/trunk/Common/PythonTools/perturbcoilbalance_fourosem.py
Note, the script isn't fancy enough to perturb the gains automatically to minimize the demodulated error signal, that's done by-eye using a few by-hand iterations of this script and watching the results in StripTool.

- I've used a compromise of filters than Kiwamu and Keita inside the LOCKIN, which I'll motivate in the comments below. For the oscillator clock frequency band pass (in the DEMOD_SIG banks) I used
butter("BandPass",2,3.5,4.5)
and called it "BP4.1Hz," and for the I and Q low-pass filter, I used
cheby1("LowPass",2,3,0.05)
and called it "CLP50mHz."
These seem to have worked out well (for my patience level), and can be copied as long as the clock frequency remains the same (which should be assessed anew for every SUS type.) 
Non-image files attached to this report
Comments related to this report
jeffrey.kissel@LIGO.ORG - 20:27, Wednesday 22 January 2014 (9454)ISC
On which sensor to use for your demodulated signal

In order to best balance the coils one wants a sensor that captures what affects the cavity alignment the best. This is typically some sensor measuring the optic motion. However, each suspension type tried thus far has had different options.

- H1 SUS BS, a BSFM (Balanced by Keita) has only an optical lever measuring the optic. So this was the obvious choice, and as such, the LOCKIN part is directly hooked up to it. Good.

- H1 SUS PRM, an HSTS (Balanced by Kiwamu) does not have an optical lever. However, it *does* have OSEMs on the bottom stage. These are hooked up to the LOCKIN as the optical levers are on the BS. HOWEVER, the OSEM sensors, which are in the same location as the actuators, so if the basis transformation from UL LL UR LR to Optic P and Y are imperfect, then that limits the precision to which you can balance the coils. As such, Kiwamu constructed a make-shift optical lever using the REFL WFS at DC, which act like a QPD and the light source is the transmitted light from the IMC: because PRM is a 3% transmission mirror, it reflects tons of light into HAM1 when the PRC is not locked. A rather expensive light source, but GENIUS! As Kiwamu mentioned however, this work-around wasn't hooked up to the demodulator, so he just used the ASD as his figure merit.

- H1 SUS PR2, which has also has no optical lever, and because of the transmissions of the mirrors in the PRC, one can't get enough light on any nearby WFS or QPD to pull the same Izumi trickery, so we're stuck with the imperfect OSEMs. 

Once we get cavities under stable lock, we can revisit these optics which have no lever (i.e. all HSTS), because then we'll have the full ASC system with light everywhere at our disposal. But in summary, PRM is the only mirror lucky enough to do this trickery. Thankfully, as shown above, the M3 OSEMs can get some of improvement by themselves (as reported by themselves; of course we should check the improvement with global control loops and interferometers).

*nudge*nudge* Integration Issue 461 *nudge*nudge*
jeffrey.kissel@LIGO.ORG - 22:37, Wednesday 22 January 2014 (9457)ISC
On the signal processing filters and OSC frequency for the LKIN part

In the lock-in amplifier process, we want the average amplitude of the product of the oscillator and the response signal (filtered at that same frequency). This average value provides a metric of the linear coupling to the drive at the oscillator frequency which one can minimize, with the benefit of directionality. In order to isolate this DC, average value of the product from the bilinear term, we low-pass the output of the demodulator. The design metrics of this low-pass are a function of the oscillator frequency chosen and, in our case, the patience of the user:
(1) One wants a significant amount of isolation at twice the oscillation frequency, such that the 2f "noise" does not interfere with the average DC value, but
(2) The response time of the low-pass filter defines how many cycles which the average includes, and therefore the response time of your metric to the knobs you have to change it.
As such, one wants a high oscillation frequency with respect to the corner frequency of the low pass. 

However, we have further design constraints: 
(3) Given that the SUS actuators are weak and the particular, off-diagonal, mechanical response to our excitation is so small at high-frequency, signal-to-noise and/or coherence limit how high we can push of oscillator frequency to roughly 5 [Hz]. 
(4) Ideally, the SUS response to pringle excitation should be frequency-independent, if we've done a proper job of digitally compensating for the analog frequency response of the actuation chain. But, if one chooses a low oscillator frequency -- say 0.1 [Hz] -- then the averaging low-pass filter will have to be much lower, and the response time becomes unbearably slow -- hundreds of seconds.
(5) In the middle, between ~0.5 [Hz] and ~5 [Hz] is a forest of SUS resonances, which, if excited, might result in all sorts of unexpected coupling to degrees of freedom not of interest, saturations of sensors, etc.

This leaves a tiny region between the given suspensions' resonant forest and where coherence drops off to stick the oscillation frequency. Keita chose 2.9 [Hz] on H1 SUS BS, and Kiwamu chose 4.1 [Hz] on H1 SUS PRM after considering these constraints, and because H1 SUS PR2 was the same SUS type as H1 SUS PRM, I stuck with Kiwamu's frequency of 4.1 [Hz].

Once the oscillation frequency is chosen, and the response signal, band-pass filter can be designed. Here, you want to isolate the response signal at the given oscillator frequency, but you need to be mindful that the impulse response of the band-pass filter is shorter than of the demodulated signal's low-pass filter. Here's how we each designed our filters:

Keita's approach (for H1 SUS BS): Lots of isolation on both the response signal band-pass and demodulated low-pass, and forgo patience. With a sharp cut-off, elliptic, band-pass the response time was greater than 5 seconds. The a sharp cut-off, elliptic low-pass, the averaging time was more than 100 sec.

Kiwamu's approach (for H1 SUS PRM): Less isolation for the less patient. Reduce the sharpness of the cutoff of both the band-pass to second order Butterworth filters, and increase the corner frequency of the low=pass to 300 mHz to reduce the averaging time to ~20 seconds. Note that he had started with an oscillator frequency of 10 [Hz] (hence the center frequency of his band-pass), but found the could not get enough coherence, reduced the oscillator frequency, and got enough response signal at 4.1 [Hz] to leak through to make a viable demodulated signal, so he didn't bother to change the design of the band-pass.

Since I just learned all of this today, and had Kiwamu, Stefan, and Keita telling me different metrics for their design, I chose what made sense to me (for H1 SUS PR2): Compromise. Move the center frequency of Kiwamu's band-pass to the oscillator frequency, but leave the shallow isolation in order to preserve the short impulse response time. Take the happy medium regarding patience and move the corner frequency of the demodulated low-pass to 50 [mHz]. This gave me plenty of SNR, lots of averages, and a impulse response time of the demodulated signal metric similar to Kiwamu's; about 10-20 seconds.

I attach bode plots and impulse response times for each of the three filters, color coded as above, with H1 SUS BS, H1 SUS PRM, and H1 SUS PR2.
Non-image files attached to this comment
jeffrey.kissel@LIGO.ORG - 23:46, Wednesday 22 January 2014 (9459)
Measurement Technique Expanded

Though Keita did a fairly good job explaining the principles of the technique in LHO aLOG 9079, I expand on his instructions with a little more detail below for the non-Jedi, and such that we might one day automate the process.

(A) Install the signal band-pass and demodulated I & Q filters; the same BPs in both oscillators' SIG bank, and the same LPs in both oscillators' I and Q banks.
SIG band pass: BP4.1Hz = butter("BandPass",2,3.5,4.5)
DEMOD I & Q low-pass: CLP50mHz = cheby1("LowPass",2,3,0.05)

(B) Turn on both the Pitch and Yaw Oscillators. 
As Keita mentions,
   (i) Only the amplitude of the oscillator you send out the coils matters, but the other must at least be *on* in order for the demodulation to happen,
   (ii) For each oscillator, the amplitude of the sin and cos don't matter, as long as they're the same,
   (iii) You want to demodulate both oscillators at the same frequency, so the oscillator frequency should be the same for both.
For sanity's sake, I just made both oscillators have exactly the same parameters
OSC     Frequency [Hz]     Amplitude [ct]     Sin [ct]      Cos [ct]
 P         4.1              200000 (2e6)       1000           1000
 Y         4.1              200000 (2e6)       1000           1000

(C) Turn on the EXC_SW for the stage you wish to balance, and filling out the associated stage's LKIN2OSEM matrix such that only one oscillator drives in a pringle configuration (I, like Keita, chose to use the Pitch oscillator),
     P   Y
UL  +1   0
LL  -1   0
UR  -1   0
LR  +1   0

(D) Open up two StripTool charts, one for each oscillator. Put the output of the I & Q demodulator banks for each oscillator on each tool, (e.g. H1:SUS-PR2_LKIN_Y_DEMOD_I_OUTPUT and H1:SUS-PR2_LKIN_Y_DEMOD_Q_OUTPUT). Be sure to set the y-scale for each channel to be consistent (+/- 10 [ct] worked for me, and then I zoomed in and out as necessary).

(E) Tune drive amplitude. I was initially scared of the large drive amplitudes Kiwamu had chosen, so I crept up on 200000. However, this meant when I started out with 10000 [ct], I got very little response when I began to tune the oscillator phase. So, crank up the drive to where you get lots of response to changing the oscillator phase, while making sure not to saturate the DAC. 

(F) Tune the oscillator phase
    (i) Again, drive hard enough that the separation between the I & Q phase is much larger than the noise (noise = wiggles around the DC value)
    (ii) Spin through the DEMOD Phase until you get the Q phase near zero
    (iii) Use Kiwamu's script mentioned in the main entry, perturbcoilbalance_fourosem.py, to put a *large* coil imbalance into the COILOUTF bank. This should cause a step in both the I & Q signals. The arctangent of the ratio between the the step sizes gives you the remaining distance you are away from the ideal phase:
             dPhi = 180/pi * atan( (Q_DC^{before} - Q_DC^{after}) / (I_DC^{before} - I_DC^{after}) )
The sign with which you add this to the current oscillator phase is unclear, so try both. A well-tuned oscillator phase means that abs(Q) is close to zero, and it doesn't respond [it's DC value doesn't change] to coil imbalance changes. Note, any offset the Q phase has from zero is noise that you can't tune away with the COILOUTF gain knobs. As mentioned in the above comments, this can be a result of, for example, imperfect sensors. As long as Q doesn't respond to coil unbalancing, then it's OK -- it's just one of limits on how well you can balance the coils. You'll need to tune the phase of each oscillator independently. 

(G) Set COILOUTF gains back to +/-1. Open up a DTT session (like the templates shown in the main entry), and set up a rolling averaged transfer function between the oscillator and the response sensor input. For PR2, that's H1:SUS-PR2_LKIN_P_LO (as the A channel) and H1:SUS-PR2_M3_WIT_P_DQ, H1:SUS-PR2_M3_WIT_Y_DQ. Take a reference measurement of 10-15 averages to show how badly the pringle excitation causes pitch and yaw in your response sensor. 
Remember: (as Keita says) A Pitch imbalance shows up in the Yaw oscillator's I phase, and a Yaw imbalance shows up in the the Pitch oscillator's I phase.

(H) Begin tweaking the COILOUTF bank gains (using perturbcoilbalance_fourosem.py) until the abs(I) phase goes to zero. You should stop when the noise is larger than the distance between the DC value ad zero. For the M2 and M3 stages of PR2, a perturbation of +/- 0.0005 was the precision I was able to achieve. If your oscillator phase is tuned correctly, a PIT imbalance should not affect the Pitch I phase, and Yaw imbalance should not affect the Yaw I phase. Therefore, you can do these degrees of freedom in succession, and not have to worry about going back and fourth.

(I) Because perturbcoilbalance_fourosem.py was quickly written, the perturbation increments are not exactly what you request. So by then end of the tuning process for both DOFs you'll have overly precise gains (check by caget-ing the gains in a terminal). Round off these gains to the nearest 1000th for reasons mentioned in the main entry.

(J) Making sure to have captured the "before" measurement as a reference, take a new DTT spectra to prove how well you've done!
jeffrey.kissel@LIGO.ORG - 23:58, Wednesday 22 January 2014 (9460)
Demod Phases and Resulting I & Q Values

After balancing the coils, I took a 300 second tds average of each I and Q channel just so I'd have a quantifiable number of "how good" I tuned the balancing. Unfortunately, there's no command line standard deviation tool, so I don't have a number for how much noise was on each channel. It was roughly +/- 0.5 [ct].
PR2 M2
     Demod Phase [deg]         Balanced Value
P       165              I         -0.017
                         Q         -0.088
Y       90               I          0.14
                         Q          0.52

PR2 M3
     Demod Phase [deg]         Balanced Value
P       160              I         0.15
                         Q         1.91
Y       87               I         0.11
                         Q         3.14

As one can see, the residual offset in the Q phase was much larger on the M3 stage than on M2. Unclear why this is, but this most probably is the reason why the results for M3 are not nearly as good as those for M2. It would be nice to have an independent sensor to verify the results, but we'll have to wait for resonant cavities and the full ASC system.
jeffrey.kissel@LIGO.ORG - 17:30, Friday 07 March 2014 (10627)ISC
Here're a few screenshots of the StripTool session during the tuning process. They're attached in chronological order, showing
(1) Aligning the optical lever, to show the difference between junk signal and plenty of signal, then after hand-tuning the demod phase to get the Q phases close to zero
(2) Before and after the big perturnations to gather the tweak needed in the demod phase, as determined by math
(3) The process of balancing the coils once the demod phase is perfect, bringing the I phases to zero as well, with little perturbations to the balance
(4) What balanced coils look like (strikingly similar to junk signal, but just signal doesn't respond to perturbations), also what it looks like to have saturations (sudden increase in I and Q phase signal amplitude).

You should be able to get a good amount of SNR exciting with an oscillator amplitude between 115000 and 125000, depending on the strength of your driver, and how much you've imbalanced the coils.
Images attached to this comment
Displaying reports 72941-72960 of 83266.Go to page Start 3644 3645 3646 3647 3648 3649 3650 3651 3652 End