I have been searching for the way we can subtract signal better since I could see the acoustic coupling signal remains in the target signal even though there are no coherence between them.
I changed the training time which is used to decide wiener filter.
I have total 10 minutes data, and the wiener filter was decided using the whole data before.
(Right: the performance with the data when the triangular sound was created. Left: the performance with the data when the gaussian sound was created.)
I found that the acoustic signal can be fully subtracted above 40 Hz when the training time is short. This means the transfer functions between the acoustic signals and MCF signal change.
However, if the wiener filter is decided with short-time training, the performances at lower frequencies get worse. This is because wiener filter do not have enough low-frequency information.
So, I would like to find the way to combine the short-time training merit and long-time training merit. It should be useful to subtract the broad-band coupling noise.
The attached plots display RMS noise from various accelerometers and seismometers over the past 90 days. One can see how after the reinstallation of the seismometers in November, RMS from the GUR1Z and GUR1X channels decreases by a factor of about 100 from data in August. Additionally, the RMS over the course of the last 90 days has notably decreased in all instruments. In many cases, the RMS is only the result of inherent electronics noise, rather than from a signal.
An error this evening on rossa: dataviewer not working due to some font errors:
controls@rossa:~ 0$ dataviewer
Warning: Not all children have same parent in XtManageChildren
Warning: Not all children have same parent in XtManageChildren
Warning: Not all children have same parent in XtManageChildren
Warning: Not all children have same parent in XtManageChildren
Warning: Not all children have same parent in XtManageChildren
Character '\52' not supported in font. Discarded.
Character '\56' not supported in font. Discarded.
Character '\170' not supported in font. Discarded.
The suspect line was in elog/fckeditor/editor/js/fckeditorcode_gecko.js. I hacked it up so that it stopped whatever crappy conditional user agent crap it was doing. It seems to be working now.
Edit by Koji: In order to make this change work, I needed to clear the cache of firefox from Tool/Clear Recent History menu.
I ve tested another bandpass filter today with similar set-up. This time I took the data with corrected reference level. To set this reference-level the filter was disconnected and the cable was connected "thru" according to the instructions provided in the manual of AG4395A at http://cp.literature.agilent.com/litweb/pdf/04395-90040.pdf, page 3-10. The transfer functions are as follows
Spare optics from the AP table were moved to glass cabinet in the east arm. I'm not sure this is the right place. We'll see what everybody thinks.
There were two UNMARKED optics. Shame on you! No pencil marks on the optics either. These optics were shipped to the FBI for finger tip analysis.
Jeff is still working on the filter banks.
It took 15 days to reach 1e-5 Torr
The RGA was turned on yesterday. The rga scan plotting was not working. Den wrote a Python script to log and plot because the Mathlab had trouble receiving data from our old c0raga.
It turned out that the Joe's log&plot system was working today. Thanks Den and Joe.
Here is the first rga scan.
Some obvious things to be fixed
- We need POP55 and POP CCD for diagnoses.
Done. The beam is also going vaguely in the direction of POP110, but I can't see the beam, so it's tricky.
Order of operations:
1. Find POP on the table, place iris so I wouldn't forget. Find beam by putting big IR card where I think beam should be, look at IR card with IR viewer.
2. Move and re-clamp 2" lens so beam is on center of lens.
3. Move and re-clamp 1st (2") mirror so that beam is on center of mirror.
4. Remove BS-33% so that all the beam goes to POP55, steer that 1st mirror so beam is on POP55's little mirror. Align little mirror so beam is centered on POP55 (as seen by looking at PD with viewer, finding "edges" of PD, going back to center).
5. Put BS-33% back in place. The reflected portion of this beam is not possible to see using card+viewer technique.
6. Remove BS-50% that reflects half of this beam to POP110. Find beam reflected from BS-33% by waving POP camera around. Steer BS-33 until beam goes back in the direction that the camera used to be mounted. Adjust camera mount and BS-33 so that beam is on camera.
7. Put BS-50% back in place. Steer it around with voltmeter on PD to see if beam ever hits PD. Unsuccessful. Give up, since we have POP55, and POP camera.
8. Make a youtube video: POP, AS, REFL, ITMXF (all on Quad3) - PRMI coarsely aligned, no IFO parts locked. MICH was locked earlier, but not during video time.
Last Thursday, I put the speaker and my laptop in the PSL table, and make triangular wave sound with the basic frequency of 40Hz, and Gaussian distributed sound.
(I create the sounds from my laptop using the software 'NHC Tone Generator' because I could not find the connector from BNC to speaker plug.)
And I measured the acoustic coupling in MCF signal. The all the 6 microphones were set in PSL table around PMC and PSL output optics.
The performance of the offline noise cancellation with wiener filter is below.
(The target signal is MCF and the witness signals are 6 microphones.)
I can see some effects on MCF due to the sound on PSL table. Though I can subtract some acoustic signal and there are no coherence between MCF signal and mic signals, still some acoustic noise remains.
This is maybe because of some non-linearity effects or maybe because we have other effective places for acoustic coupling measurement. More investigations are needed.
Also, I compared the wiener filter and the transfer function from microphones signal to MCF signal. They should be the same ideally.
(Left: Wiener filter, Right: Transfer function estimated by the spectrum. They are measured when the Gaussian sound is on.)
These are different especially lower frequencies than 50 Hz. The wiener filter is bigger at lower frequencies. I guess this adds extra noise on the MCF signal. (see the 1st figure.)
The wiener filter can be improved by filterings. But if so, I want to know how can we determine the filters. It is interesting if we have some algorithms to determine the filters and taps and so on.
The more investigations are also needed.
I grabbed the a plot of the iLIGO PSL frequency noise spectrum from the Rana manifesto:
Rana's contention is that this spectrum (red trace) is roughly the same as for our NPRO.
From the jenne/mevans/pepper/rana paper Active noise cancellation in a suspended interferometer I pulled a plot of the calibrated MC_L noise spectrum:
The green line on this plot is a rough estimate of where the above laser frequency noise would fall on this plot. The conversion is:
L / f = 10 m / 2.8e14 Hz = 3.5e-14 m/Hz
which at 10 Hz is roughly 1.5e-11 m. This puts the crossover somewhere between 1 and 10 Hz.
After some system updates this evening, firefox can no longer handle the html input encoding for the elog. I'm not sure what happened. You can still use the "ELCode" or "plain" input encodings, but "HTML" won't work. The problem seems to be firefox 17. ottavia and rosalba were upgraded, while rossa and pianosa have not yet been.
I've installed chromium-browser (debranded chrome) on all the machines as a backup. Hopefully the problem will clear itself up with the next update. In the mean time I'll try to figure out what happened.
To use chromium: Appliations -> Internet -> Chromium
[Rana, Jamie, Ayaka]
We could not lock the arms with MC_L loop on. Therefore we measured the change in YARM error signal when the MC_L is turned on.
(data; POYerr_MCF.xml in zip file)
Green line; POY error signal when MCL loop was on and the YARM loop gain (0.5) was so high that the saturated control signal made funny peak around 250 Hz.
Blue line; POY error signal when MCL loop was off and the YARM loop gain was low (0.2).
Pink line; POY error signal when MCL loop was on (the gain was -300) and the YARM loop gain was low (0.2).
Red line; POY error signal when MCL loop was on, another low pass filter (2nd order, cut off frequency of 55Hz) was added to MCL loop and the YARM loop gain was low (0.2).
We also changed the filter trigger in order to lock YARM. The FM7 and 8 trigger was turned off. It means that spectrum above was took with FM2,3,4,5,6,9,10 on. Whitening filters were also on.
MCL control signal makes the arm spectrum bad because the MCL control signal moves MC2 mirror additionally and adds extra frequency noise.
Ideally, error signal should be the same at higher frequencies and go down at the lower frequencies when the MCL loop is on because MCL signal should suppress the seismic noise.
Before we added the LPF, MCF/MCL loop cross over (which was taken with the template /users/Templates/MC/MCL-MCF_xover-2012-8-23.xml) is below;
(MCL-MCF_xover.xml in zip file)
After the LPF is added, the cross over has been changed as below;
(MCL-MCF_xover2.xml in zip file)
For now, I will just turn off the MCL loop for the acoustic noise experiments.
We have checked the transfer function of a bandpass filter using AG4395A network analyzer and retrieved the data through GPIB. The RF out signal of AG4395A had been divided by splitter with two outputs of the splitter going to through R and the filter which was connected to the A channel of the network analyzer. The GPIB data came in complex data format, from which the absolute value and phase had to be retrieved.
The plot for the TF is as following
We've been talking for a while about how we want to store data. I'm not in love with keeping it on the elog, although I think we should always be able to reference and go back and forth between the elogs and the data.
I have made a new folder: /data EDIT: nevermind. I want it to be on the file system just like /users, but I don't know how to do that. Right now the folder is just on Ottavia. Jamie will help me tomorrow.
In this folder, we will save all of the data which goes into the elog.
I propose that we should have a common format for the names of the data files, so that we can easily find things.
My proposal is that one begins ones elog regarding the data to be saved, and submit it immediately after putting in the first ~sentence or so. One should then make a new folder inside the data folder with a title "elog#####_Anything_Else_You_Want" Then, data (which was originally saved in ones own users folder) should be copied into the /data/elog#####_AnythingElse/ folder. Also in that folder should be any Matlab scripts used to create the plots that you post in the elog. One should then edit the elog to continue making a regular, very thorough elog, including the path to the data. Elog should include all of the information about the measurement, state of the IFO (or whatever you were measuring), etc.
Riju will be alpha-testing this procedure tonight. EDIT: nevermind...see previous edit.
I got two seismometers and one microphone back from Tara.
They are now near the Gurlap under the MC.
I have finally plugged GUR1 back in....it is down at ETMY for now, since that's where the cable was. BLRMS are back up on the projector.
Air conditioning maintenance is scheduled for tomorrow morning till noon.
I uploaded a zip file that contains data files used for the calibration.
OLTF_x/y.txt: the open loop transfer function (measured by IN1/IN2 in arm servo filter bank).
coh_x/y.txt: coherence of OLTF. I used the data where coherence > 0.98.
ext_err_x/y.txt: the transfer function from ITM excitation signal to POX/POY error signal.
coh_x2/y2.txt: coherence of ext_err. I used the data where coherence > 0.98.
The LSC whitening filter was off because the xarm was unlocked when the POX Q whitening filter was turned on. (We have to study what was wrong.)
The SUS whitening filters were on.
The all digital filters except +6dB filter were on.
Today I found the IMC was misaligned significantly by WFS feedback.
Once the feedback was cleared, it locks with nice visibility.
But WFS misaligns it again as soon as the intergrators are engaged.
I checked the beam on the table, but found nothing really wrong.
The offsets of the error signals were nulled at the input filter modules of the WFS segments.
They did not fix the problem.
The instability started about 48hour ago, that means my work on the AP table did not
made immediate trouble. But it does not mean anything.
For now, the WFS outputs are off. More work is needed to find what's wrong.
On Wednesday (21st) night, I checked the AP table as I wanted to try PRMI locking.
It was difficult to work with the table as there were so many unnecessary components on it.
Also the beams went through complicated paths as they have funny angles.
So I decided to clean up of IMC REFl WFS, IFO REFL, and IFO AS paths.
I found that the AS beam was highly astigmatic as the beam went through a (too-much-) tilted lens.
I made several blocked optical paths for REFL and AS for future extension of the detection system.
The current status of the table was uploaded below.
The optical spectrum analyzers and the aux NPRO were left untouched but they should be moved
somewhere (either on the table or outside) which does not disturb the other optical paths.
After the cleaning, I started locking PRMI. I could lock PRMI stably. But I could not figure out how
the intra-cavity mode looked like as I did not have the POP camera. The power recycling
gain was not quantitatively evaluated as I did not have POP and I wasn't sure how the beam was aligned at POX/POY.
We need to know:
- Quantitative evaluation of the beam shape in the PRC
- Quantitative evaluation of the power recycling gain
- The POX whitening filters seem not switching. This issue should be checked at the circuit module itself and at the BIO.
- The POX beam is not well focused on the PD. This was particularly clear when PRMI was locked with carrier.
- The POP beam is going nowhere. We need POP55 and POP CCD for diagnoses.
I haven't checked ITMY table.
For the loop diagnosis, its best to use the method of "IN1/IN2", rather than manipulate the close loop gain. In this way, you can directly plot the swept sine measurement from DTT as the open loop gain.
Also, for reporting calibration, we should all try to record the current settings better. Anything that may change the loop gain should be recorded along with the Bode plot and the DATA must be posted to the elog - no more of just posting plots.
We need to know, e.g.
what is the power in the arms?
are the LSC whitening filters on?
are the SUS dewhitening filters on?
What normalization is being used in the LSC?
What digital filters are on in the X/YARM loop filter bank?
Resistance is feudal.
In order to estimate whether we can see acoustic coupling in arms or not, we have to calibrate signals to phase noise.
I used the same method as Yuta and Jenne did (6834).
I switched from ETM locking to ITM locking since only ITM actuators are calibrated (5583), and measured the open loop transfer function and the transfer function from ITM excitation to POX/POY error signal. Then I can estimate the calibration value H [counts/m] from POY/POX error signal to displacement.
Yarm; H = 9.51 x 1011 counts/m
Xarm; H = 6.68 x 1011 counts/m
Phase noise in arms:
blue; Xarm, green; Yarm
I will calibrate the acoustic signal and see if it is reasonable that we can see the acoustic coupling signal in the arms.
But I guess it is difficult. Actually I have not seen coherence between ETM feedback signals and acoustic sounds yet. (I measured acoustic noise near POX and in PSL table.)
If I find that it is hopeless, I will create some sounds and try to measure transfer function from acoustic sound to arm cavity signals.
I am interested in how the transfer function calculated by wiener filtering is different from the measured transfer function.
I found that we do not have enough phase margin. This is why the arm locking is not so stable.
As MCL is disturbing arm locking by injecting a lot of noise, I have modified 'mcup' to disable MCL
As MC WFS keeps failing to start up when it is locked, the lines in 'mcwfsoff' to clear WFS filter history were restored.
Since the transmission beam on ETMXT camera seemed to be clipped, we checked the optics on ETMX table.
We aligned the lens so that it is orthogonal to the beam, then the beam shape looks fine.
Also we removed some an-used optics which were used for fiber input.
Last night I found that the MC autolocker has not been updated since the chamber was closed.
i.e. The low power version had been used.
I logged into op340m and modified crontab via "crontab -e" so that the normal power version is spawned.
If anybody wants a fancy single-axis seismometer for a while (GS-13), then please let me know.
New Lumitek IR Sensor Cards are here. We got 2 pieces of Q-11-T (2" x 2"), 2 pieces of Q-11-T (0.75" x 0.75") and one Q-11 (4" x 5")
We aligned and locked x and y arms.
MCL loop makes arms lock unstable, adds a lot of noise at frequencies 60-100 Hz. We'll fix it.
At some point we were not able to lock because of ADC overflows of PO signals. They happened if whitening filters were enabled. So we reduced the gain of POX whitening filters down to 36 dB and POY - to 39 dB. Now cavities can be locked with whitening filters.
Also we changed the pedestal of the lens in the beam path to the POX because the beam was too high.
Adjusted focus on ETMYF camera so that the IR beam is in focus.
For convenience, I would include a steering mirror in front of the TRX PD. Also, don't we usually have lenses in the oplev paths? Also, also, do we need lenses in front of the green refl and TRX PDs? Do you have a place in mind for the shutter? Is there a way to compactify the layout a little bit, so that even if the lenses are different for each table, the general layout for both ETMX and ETMY is the same (with an empty space on ETMX where IPANG belongs on ETMY)? I'm sure it is, since you've talked to Steve about this, but just to check: is the green refl PD far enough away from the edge of the table to accommodate the fancy new box?
I have attached the possible layout of the optics on the new ETMX endtable. More optics have been added when compared to the early cartoon layout considering that we need additional steering mirrors for reasons like: the table height in and out-of vac are different and several mounts have restricted movement in certain degrees of freedom.
As you can see, there is enough room for filters and other last time additions that may arise.
I will proceed with calculations based on the distances from the CAD drawing and the spec of the optics if there are no comments or suggestions about the layout.
Charles Blakemore our new undergrad help received 40m specific safety training today.
Oplev values that were changed to zero:
PRM P=0.15, Y=-0.3
SRM P=-2.0, Y=2.0
BS P=0.2, Y=-0.2
ITMY P=2.1, Y=-2.0
ITMX P=1.0, Y=-0.5
ETMX P=-0.2, Y=-0.2
ETMY P=0.5, Y=0.6
Also, PRCL was changed in the LSC input matrix from REFL33I to AS55I, since there is no REFL beam out of the IFO :(
Ayaka and I restored all of the oplev gains to these values. The exception is ETMY, which has both gains negative. I am unsure if this is a transcription error on my part, or if something physical has changed. The layout of the ETMY oplev was modified (since Rana took out the offending lens) but that shouldn't affect the sign of the gains.
POY11 does not go out of the vacuum
It does but slighty low and does not get on mirrors. We need to change optic mounts to adjust the height. Red is flashing in yarm at 00 and 10 modes. TRY is ~0.4-0.5.
I've adjusted BS angle, camera and TRX PD at ETMX table so I can see red flashing at 03 mode while green is locked to 00 and its transmission is maximized. I thought that by adjusting BS angle, I will be able to align red to 00 not disturbing green, but this was not the case. Maximum TRX I could get was 0.1. I've adjusted POX to get into PD and I can see PDH signal though I can't lock as cavity is still misaligned for red.
We put the POY beam onto the POY PD. The Yarm is currently locked on IR with ~0.65 transmission.
You have constraints for the IR beams (i.e. one PZT and one BS for 8 dofs), so now you need to align the arms for the input IR beams.
The PZT and BS should be aligned so that you have the beam spots as center as possible with the above restrictions.
Then realign end greens for the given arm alignment. You can replace the mounts if necessary to align the end green.
Even if you lose the coarse alignment of the green, realignment is not difficult as you know now
We aligned accurately 00 green in yarm, changed voltage on PZT2 to see red flashing at TRY at the normalized level 0.2-0.3. The plan was to lock yarm using POY11 and green from other side, maximize red TRY by adjusting PZT2. But POY11 does not go out of the vacuum, so we adjusted TRY by flashing. 2 DOFs of PZT2 is not enough to match 4 DOFs of red beam so we adjusted both PZT2 and cavity mirrors. TRY flashing is 0.5-0.6 and green is still locking to 00 though its transmission is not maximized. We'll fix it later by adjusting input green beam.
Next we wanted to get red beam on TRX PD. Beam steering was done by BS only. We misaligned BS in pitch and excited BS angle motion by 1000 counts. We could see red beam moving on the wall of ETMX chamber. We moved it to ETMX mirror frame, estimated position of the mirror center and moved BS to this position. The beam should be approximately in the middle. For now we can not see red beam on the camera at ETMX table, more work is needed.
We aligned and locked xarm for green.
That's really, really awesome!
People complained about the MC instability: If we aligned the MC, it locked nicely for a while.
Then suddenly you find that it got totally misaligned with the order of 0.2 with the alignment slider.
This misalignment usually happens for MC2, but it happend on MC3 once.
Surprisingly to me, this instability happened even without MCL and WFS, not only once but at least three times.
This suggests that the suspensions are the cause of the trouble.
I played with the MC2 suspension for a while in the afternoon. It seems that it has a hysteresis (or say, bistablity).
And the nominal alignment of MC2 seems close to the point where the transition happens. (Dunno why)
I further played with MC2 and found that a step of POS actuation by +/-10000 induces this transition go and back.
When the POS kick is in the negative direction (-10000), the MC2 seems to return to the preferrable
position. Therefore, I applied DC position force of -5000 to pull the mirror a bit from the nominal position.
I let the MC locked for ~4hours without MCL and WFS, it kept good alignment and the lock was stable
except for unlocks because of the activties by Den and Ayaka.
All of them has been done without previous monitor data as the tools were available.
The MC2 situation is not conclusive but we should check how we can prevent the bistable transition
by restricting MCL/WFS.
We aligned and locked Y arm for green:
I've switched error channel cable to output monitor. Whitening filter is need for scattering measurements.
I didn't make any concrete progress today. AS and REFL cameras are in place, and Manasa has put ND 0.5 filters on both. I used a
camera to look at the back of the Faraday, and aligned PRM such that it was retroreflecting, and then tried to align ITMY to have once
fringes with the PRM at that place. I failed in this, since the AS beam on the AS table was starting to dall off the first mirror on the table.
I then restored all the suspensions to where they were before I started touching them today.
I moved ETMY face camera so that it is looking at the front of the black glass, but it's hard to tell where the beam is. I was thinking
about setting up a temporary camera to look at the back of ITMY to help guide PZT steering, but I haven't done this yet.
Koji and I then talked about the several different options I have for references, and how many different knobs I can turn. I'm sleeping
on it for now, and hopefully I'll have more insight on what to do tomorrow.
I've applied LQR feedback technique to BS oplev in pitch. I think the most inconvenient thing in using LQR controller is the amount of additional states created during cost function shaping. It requires 1 filter bank for each state. To avoid this I wrote state estimation code so all states are calculated inside one function.
On the plots below cost function and oplev feedback controller performance are shown.
I tried to add a test point to C1MCS model and spent next two hours rebooting front-ends, restarting models and realigning MC.
dmesg told me that DAQ channels can not be allocated as they already exist. Last time we met this problem Jamie emailed Alex about it. Jamie, what is the output? Restarting iop model does not help this time.
We have discovered that the MCL loop squishes the length fluctuations that result from the MC spot measurement angular dither. This is good, in that the MCL is doing what it ought to do. However, we need to turn it off before doing a spot measurement.
This is totally non-sensical statement, of course.
We might also say that the DARM loop totally squishes the GW signal and so its better not to have any feedback in the interferometer.
Hmmm, indeed. To keep MC_L on, we should be looking at the control signal rather than the error signal. I think Den has MC_L running nicely such that it always comes on when the MC locks, so I can switch it.