40m QIL Cryo_Lab CTN SUS_Lab TCS_Lab OMC_Lab CRIME_Lab FEA ENG_Labs OptContFac Mariner WBEEShop
  40m Log, Page 83 of 344  Not logged in ELOG logo
ID Date Author Typeup Category Subject
  811   Thu Aug 7 17:32:23 2008 JenneUpdateSUSAfternoon PRM activities
Rana, Jenne, Yoichi, Dmass

After Yoichi confirmed this morning that the wire was in both grooves, Rana attempted to lift the PRM a tiny bit, and twist it around (very gently of course) to see if we could make the wire slip back to its nominal position underneath the optic. On the first attempt, the wire ended up slipping the wrong way, causing slightly more tilt. On another attempt, the wire came out of the groove nearest the chamber door by about 0.5mm. We got the wire back in the groove by slightly lifting the optic, and pushing the wire back in. Then, on further attempts at making the wire slip back to its nominal position, the wire came out of the groove farthest from the chamber door. It is very difficult to get at that side of the PRM, because the table is crowded, and it is on the far side of the optical table from the chamber door. We decided to pull the PRM out of the chamber. Rana clamped the mirror into its cage using the earthquake stops and removed the OSEMS, and then we pulled the mirror out. We put it on a cart that was covered with foil and had a little foil house for the optic cage. We rolled the mirror+cage over to the flow bench at the end of the y-arm.

We saw that the wire is no longer even on the standoff (~3mm away from the groove) on the side that was farthest from the chamber door.

Since we have not confirmed that we have spare wire and spare magnets (and due to the time of day), we have decided to cover the cage with some foil, while it is sitting on the flow bench, and we'll fix the wire in the morning.
  812   Fri Aug 8 09:54:10 2008 ranaUpdateCamerasNew code + gstreamer allows for easy saving and compression of images

Modified the CamSnap code to output the image data stream to standard out. This can then be piped into a gstreamer plugin and then be used

Didn't work; Prosilica has only 1 "l". Even so, sshing from op440m to mafalda, I got this:
mafalda:SnapCode>CamSnap -F 'Mono8' -c 44058 -E 5000 -X 0 -Y 0 -H 480 -W 752 -l 0 -m 300 | gst-launch-0.10 fdsrc fd=0 blocksize=360960 ! video/x-raw
-gray, height=480, width=752, bpp=8,depth=8,framerate=30/1 ! ffmpegcolorspace ! theoraenc ! oggmux ! filesink location="./testVideo.ogm"
Setting pipeline to PAUSED ...
Pipeline is PREROLLING ...

** (gst-launch-0.10:27121): WARNING **: Size 60 is not a multiple of unit size 360960
Caught SIGSEGV accessing address 0x487c
ERROR: from element /pipeline0/ffmpegcsp0: subclass did not specify output size
Additional debug info:
gstbasetransform.c(1495): gst_base_transform_handle_buffer (): /pipeline0/ffmpegcsp0:
subclass did not specify output size
ERROR: pipeline doesn't want to preroll.
Setting pipeline to NULL ...
#0  0xffffe410 in __kernel_vsyscall ()
#1  0xb7deddae in __lll_mutex_lock_wait ()
#2  0xb7de9aac in _L_mutex_lock_51 () from /lib/tls/i686/cmov/libpthread.so.0
#3  0xb7de949d in pthread_mutex_lock ()
#4  0xb7e452e0 in g_static_rec_mutex_lock () from /usr/lib/libglib-2.0.so.0
#5  0xb7f1fa08 in ?? () from /usr/lib/libgstreamer-0.10.so.0
#6  0x080c1220 in ?? ()
#7  0x00000001 in ?? ()
#8  0x0809586c in ?? ()
#9  0x00000001 in ?? ()
#10 0x08095868 in ?? ()
#11 0xb7f7a2a8 in ?? () from /usr/lib/libgstreamer-0.10.so.0
#12 0xb7e8da80 in ?? () from /usr/lib/libglib-2.0.so.0
#13 0xb7f7a2a8 in ?? () from /usr/lib/libgstreamer-0.10.so.0
#14 0xb7f7a2a8 in ?? () from /usr/lib/libgstreamer-0.10.so.0
#15 0x00000000 in ?? ()
Spinning.  Please run 'gdb gst-launch 27121' to continue debugging, Ctrl-C to quit, or Ctrl-\ to dump core.
Caught interrupt -- 
  814   Fri Aug 8 11:04:34 2008 SharonUpdate MCL Wiener filter
I took some old data from Rana and converted the units of the Weiner filter to m/m so to make the effect of the seismometer and accelerometers more obvious.

The data is in counts, and so to convert to m this is what I did:

%%% MC_L calibration
v_per_counts = 5/32768;
v_per_v = 3;
amp_per_N = 1/0.064;

%%% Accelerometers calibration
v_per_counts_acc = 61.045e-6;
g_per_v = 9.8/100;

%%% Seismometer calibration
v_per_counts_seis = 61.045e-6;
m_per_s_per_s_per_volt = 9.8/100;
m_per_v_per_s = 1/345;

for jj=1:6
hfmatm(:,jj)=hfmat(:,jj).*(v_per_counts.*v_per_v.*amp_per_N.*f)./(v_per_counts_acc*g_per_v); %%% accelerometers' data
hfmatm(:,7)=hfmat(:,7).*(v_per_counts.*v_per_v.*amp_per_N)./(v_per_counts_seis*m_per_v_per_s); %%% Seismometer data
  816   Fri Aug 8 13:29:54 2008 YoichiUpdateSUSNo groove in the stand-off ... wait, it is not even a stand-off !
Yoichi, Steve, Seiji

We took magnified pictures of the stand-offs of the PRM.
Attm1: North side stand-off.
Attm2: South side stand-off.
Attm3: Zipped file of the full pictures.

We found no groove in the south side stand-off.
After some discussion, we concluded that it is actually a guide rod. You can see it from the size difference (the magnification is the same for the two pictures).
The stand off on the south side is missing (fell off, ran away, evaporated or whatever ...).
Also we noticed that the guide rod on the north side is missing.

We have to find a stand-off and place it on the south side.
Seiji suggested that it is better to put a guide rod next to the north side stand-off, otherwise the stand-off itself is too weak to hold the load.
He also said that the PRM was installed after he left, so it was not his fault.
  817   Fri Aug 8 15:10:35 2008 YoichiUpdateSUSNo groove in the stand-off ... wait, it is not even a stand-off !
I tried to find the missing stand-off and the guide rod in the BS chamber, but I couldn't.
  818   Fri Aug 8 17:54:52 2008 JenneUpdateSUSStandoffs and Guide Rods
After closer inspection of other small optics, it is clear that the guide rods should be above the standoffs on our small optics. Yoichi took a picture of the SRM that shows this clearly. This makes sense since the tension of the wire will make the standoff 'want' to go up during pre-epoxy adjustment, so the guide rod prevents the standoff from popping up and out.

Looking at the side of the PRM without the groove, it looks like there is lots of space between the guide rod and the alignment etch in the glass, so we can just place a standoff directly under the guide rod that is present.

A spare standoff is being shipped tomorrow morning, so we should have it by Monday for installation on the PRM.
  820   Mon Aug 11 00:58:31 2008 ranaUpdatePEM2 years of temperature trend
The PSL RMTEMP alarmed again because it says the room temperature is 19.5 C. Steve said in
an earlier log entry that this is a false alarm but he didn't say why he thought so...

I say that either the calibration of the RMTEMP channel has drifted, the setpoint of the HVAC
has shifted, or there's a drift in the RMTEMP channel. I don't know what electronics exactly
are used for this channel so not sure if its susceptible to so much drift.

However, since the Dust Monitor (count_temp) shows a similar temperature decline in the
last two years I am inclined to blame the HVAC system.

The attached plot shows 2 years of hour-mean trend.
  821   Mon Aug 11 09:39:29 2008 ranaUpdatePEM2 years of temperature trend
Steve and I went around and inspected and then adjusted the thermostats and humidostats.

All the thermostats were set at 70F in 2005 by Steve. We adjusted the ones on the arms up to 72F
and set the one on the wall west of the framebuilder up to 74F (this area was consistently colder
than all the others and so we're over-correcting intentionally).
  827   Tue Aug 12 12:05:36 2008 YoichiUpdateComputersHP color printer is back
I restarted the HP printer server (a little box connected to the HP color laser) so that we can use the HP LaserJet 2550.
After this treatment, the printer spat out a bunch of pages from suspended jobs, many of these were black and white.
I think people should use the black-and-white printer for these kind of jobs, because the color printer is slow and troublesome.
  829   Tue Aug 12 19:48:24 2008 JenneUpdateIOOPRM standoff is in....mostly
Yoichi, Jenne

The missing PRM standoff is now partially installed. The standoff is in, and the wire is in the groove, but we have not finished adjusting its position to make the PRM stand up straight. It turns out to be pretty tricky to get the position of the standoff just right.

We have set up a HeNe laser as an oplev on the flow bench (which we checked was level) in the clean assembly room, and are using it to check the pitch of the mirror. We set a QPD at the height of the laser, and are looking at the single-reflected light. When the single-reflected light is at the same height as the center of the QPD, then the mirror is correctly aligned in pitch. (Actually, right now we're just trying to get the single-reflected light to hit the diode at all...one step at a time here).

We'll continue trying to align the PRM's standoff in the morning.
  830   Tue Aug 12 21:38:19 2008 JohnUpdateLSCAccidental higher order mode resonances in the arms
Recently we had been having some trouble locking the full IFO in the spring configuration (SRC on +166).
It was thought that an accidental higher order mode resonance in the arms may have been causing problems.

I previously calculated the locations of the resonances using rough arm cavity parameters(Elog #690). Thanks to Koji
and Alberto I have been able to update this work with measured arm length and g factors for the y arm (Elog #801,#802).
I have also included the splitting of the modes caused by the astigmatic ETM. Code is attached.

I don't see any evidence of +166MHz resonances in the y arm.

In the attached plot different colours denote different frequencies +33, -33, +166, -166 & CR.
The numbers above each line are the mn of TEMmn.
Solid black line is the carrier resonance.
  832   Wed Aug 13 20:13:35 2008 YoichiUpdateSUSPRM stand-off is glued
Steve, Janne, Rob, Bob, Koji, Yoichi

We finally managed to balance the PRM and the stand-off is now glued.

The whole procedure was something like this:

(1) Measure the levelness of the optical table. It was done by a bubble level claiming that
the sensitivity is 60 arcsec (roughly 0.3 mrad).
There was no noticeable tilt along the longitudinal direction of the table.

(2) We put a He-Ne laser on one end of the table. Mounted a QPD on a X-Y-Z stage. Put the QPD very
close to the laser and centered it by moving the QPD.
Then we moved the QPD far away from the laser and centered the beam spot in vertical direction
by changing the tilt of the laser mount.
We then moved the QPD close to the laser again and adjusted the height to center it. By repeating
the centering at two locations (near and far) several times, we aligned the laser beam parallel to
the table.

(3) The PRM suspension tower was put on the other end of the optical table, i.e. far from the laser.
The QPD was moved next to the laser to form an optical lever. The height of the QPD is preserved from
the previous step.

(4) A stand-off was picked by a pair of tweezers. By gently lifting the mirror by the bottom earthquake stops,
the tension of the wire was relieved. Then the stand off was slid in below the guide rod.

(5) Using the microscope, it was confirmed that the wire is in the grooves on both sides.

(6) Without damping, it was too much pain to balance the mirror. So we put spare OSEMs in the suspension and
pulled a long cable from the suspension rack to the clean room with a satellite amp.

(7) It turned out that the pinout of the cable is flipped because of the vacuum feed through. So we asked Ben for help.
He made conversion cables. We also found UR OSEM was not responding. Ben opened the satellite box, and we found an op-amp was burnt.
Probably it was because we connected OSEMs wrongly at first and the LED current driver was shorted. We switched the satellite box
from the PRM one to the BS one. Ben will fix the PRM box.
Bob cleaned up some D-Sub converters for the interface with the clean OSEM pigtails.

(8) While waiting for Ben, we also tried to short the OSEM coils for inductive damping. We saw no noticeable change in the Q.

(9) After the OSEMs were connected to the digital control system, Rob tweaked the damping gains a bit to make it work efficiently.

(10) I pushed the stand-off back and forth to make the reflected beam spot centered on the QPD. I used the PZT buzzer to gently move the stand-off.
For fine tune, just touching it is enough. I found it useful to touch it without clamping the mirror, because if it is clamped, we can easily push
it too hard. When the mirror is freely hanging, once the tip of the buzzer touches the stand-off, the mirror escapes immediately. If the mirror
swings wildly by your touch, you pushed it too hard.

(11) After about an hour of struggle, I was able to level the mirror. We used about 1.5m optical lever arm. A rough calibration tells us that the
beam spot is within 0.6mm of the center of the QPD. So the reflected light is deflected by 0.4mrad. That means the mirror
is rotated by 0.2mrad. The OSEMs should have about 30mrad of actuation range. So this should be fine.

(12) We mixed the Vac Seal epoxy and put it under vacuum for 15min to remove bubbles. Actually 15min was not enough for removing bubbles completely. But
stopped there because we did not want the epoxy to be too stiff.
I dipped a thin copper wire into the epoxy and applied it on the top of the stand-off. I found the epoxy is already not fluid enough, so Steve made
another Vac Seal mixture. This time we put it under vacuum for only 3 min.
I also applied the epoxy to the sides of the stand-off.
While working on this, I accidentally touched the side of the PRM. Now there is a drop of epoxy sitting there (upper left of the attached picture).
We decided not to wipe it out because we did not want to screw up the levelness.

(13) We put an incandescent light about 1m away from the suspension to gently warm up the epoxy but not too much. We will leave it overnight to cure the
  834   Thu Aug 14 11:39:06 2008 steveUpdatePEMparticle counter is out of order
The 40m ifo paricle counter sensor failed last night.
  841   Fri Aug 15 16:45:43 2008 SharonUpdate Converting from FIR to IIR
I have been looking into different techniques to convert from FIR to IIR. This is so we can see how effective the adaptive FIR filter is in comparison to an IIR one with fewer taps.
For now I tried 2 matlab functions: Prony and stmcb (which works on LMS method).
I used the FIR wiener code, MC1_X, (c1wino) and applied the FIR to IIR algorithm.
Seems like the stmcb one works a bit better, and that there isn't much effect for having 1000 and not 400 taps.

Will keep updating on more results I have, and hopefully have the MC in time to actually check this live.
  842   Fri Aug 15 17:38:41 2008 YoichiUpdateSUSOSEM free swinging spectra before the pump down
I ran an overnight measurement of the free swinging OSEM spectra.
The attm1 shows the results. Everything look ok except for the ITMY UL OSEM.
The time series from that OSEM was very noisy and had many spikes.
We suspected the cable from the satellite box to the computer rack because we disconnected the cable
when we tested a spare cable which was used to connect the spare OSEMs to the PRM suspension in the clean room.
Janne remembered when she put the cable back, she trusted the latch on the connector and did not push it in too hard.
However, Rob suggested the latch does not work well. So she pushed the connector again. Then the signal from
the ITMY UL OSEM got back to normal.
The second attachment shows the ITMY spectra after the cable push.
We decided to pump down after confirming this.

There are still a lot of extra peaks especially in the suspensions in the BS chamber.
These may be inter modulations (by the non-linearities of the OSEMs) of the modes of the multiple
suspensions sitting on the same stack.
  846   Mon Aug 18 11:50:29 2008 YoichiUpdateSUSIn vacuum free swinging results
The first attachment is the results of the free swinging spectra measurement performed in vacuum this morning.
They are freely swinging, but the suspensions in the BS chamber got even more extra peaks.
Especially, the SRM spectrum looks like a forest.
If those extra peaks are inter-modulations of the primary suspension modes, the heights of them should be
enhanced (compared to the in-the-air case) by the increased quality factors of the primary modes (due to the less air friction).
This might explain the observed increase in the extra peaks.

While doing the free swinging, we had two big spikes in the OSEM signals of the ETMs and only in ETMs.
Those spikes screwed up the spectra of the ETMs. So the ETM spectra were calculated using the time series
after the spikes.
The second attachment shows one of those spikes. It looks like a computer glitch.
  848   Mon Aug 18 17:37:14 2008 robUpdateLockingrecovery progress

I removed the beam block after the PSL periscope and opened the PSL shutter.

There was no MC Refl beam on the camera, so I decided to trust the PSL launch
and aligned the MC to the PSL beam. Here are the old and new values for
the MC angle biases:
 __Epics_Channel_Name______   __OLD_____    ___New___
 C1:SUS-MC1_PIT_COMM          4.490900        3.246900 
 C1:SUS-MC1_YAW_COMM          0.105500	      -0.912500
 C1:SUS-MC2_PIT_COMM          3.809700	      3.658600 
 C1:SUS-MC2_YAW_COMM          -1.837100	      -1.217100
 C1:SUS-MC3_PIT_COMM          -0.614200	      -0.812200
 C1:SUS-MC3_YAW_COMM          -3.696800	      -3.303800

After this, the beam looks a *little low* going into the Faraday Isolator.
Nonetheless, after turning on the IFO input steering PZTs, I was able to
quickly steer the PRM get a beam on the REFL camera and into the REFL OSA.
The PRM optical lever beam is also striking the quad.

I then used the ETMX optical lever as a reference for realigning. After
steering around the input PZTs and ITMX, I saw some flashes in Xarm trans, then got
it locked and ran the alignment script ~5 times. The arm power went
up to 0.9, so I tweaked the MC1 to put the MC refl beam back on MCWFS.
The XARM power then went up to .96. Good enough for now.

Then I started to try and re-align the YARM. Since the oplevs for both ITMY
and the BS are untrustworthy, I first tried to get the beam bouncing off ITMX
and the BS back into the AS OSA, to try and recover some BS alignment. This
didn't work, as the AS OSA may not be a good reference anyways. After
wandering around in the dark for a little while, I decided to try an automated
scan of the alignment space. I used the trianglewave script to scan
the angle biases of BS, ITMY, & ETMY, then looked at the trend of the transmitted
power to find the gps time when there were flashes. I then used
time_machine_conlog to restore the biases to that time. This was close
enough to easily recover the alignment. After several rounds of aligning &
centering oplevs, things look good.

Also locked a PRM. Will work on the DRM tomorrow.

I'm leaving the optics in their "aligned" states over night, so they can
start their "training."

Note: The MC is not staying locked. Needs investigation.

For tomorrow:

lock up the DRM
fix the mode cleaner
re-align mode cleaner to optimize beam through Faraday
re-align all optics again (will be much easier than today)
re-align beam onto all PDs after good alignment of suspended optics is established.
  849   Mon Aug 18 22:47:12 2008 YoichiUpdateIOOMC unlock study
As rob noted, the MC keeps unlocking in a few minutes period.
I plotted time series of several signals before unlocks.
It looks like the MC alignment goes wrong a few hundred msec before the unlock (the attached plot is only one example, but all unlocks
I've looked so far show the same behavior).
I will look for the cause of this tomorrow.

The horizontal axis of the plot is sec. The data values are scaled and offset-removed appropriately so that all curves are shown
in a single plot. Therefore, the vertical axis is in arbitrary units.
  850   Tue Aug 19 10:36:34 2008 SharonUpdate Calibrating accelerometers
I took apart the accelerometers near MC1 and MC2.
The 2 sets of 3 accelerometers are now covered by a box on the floor. Please try not to move them... I will place it all back once I am done calibrating.
  851   Tue Aug 19 13:12:55 2008 JenneUpdateSUSDiagonalized PRM Input Matrix
NOTE: Use the values in elog #860 instead (20Aug2008)

Using the method described in LIGO-T040054-03-R (Shihori's "Diagonalization of the Input Matrix of the Suspension System"), I have diagonalized the input matrices for the PRM.

Notes about the method in the document:
  • Must define the peak-to-peak voltage (measured via DataViewer) to be NEGATIVE for PitLR, PitLL, YawUR, YawLR, and POSITIVE for all others
  • As Osamu noted in his 3 Aug 2005 elog entry, all of the negative signs in equations 4-9 should all be plus.

New PRM Input Matrices:

UR1.18771.0075 -1.0135
LR0.8439 -0.9425 -0.9653
LL0.9684 -1.05001.0216
  853   Tue Aug 19 14:25:38 2008 SharonUpdatePEMAccelerometer's calibration - update
Goal - Make sure the accelerometers are calibrated among themselves (have the same power spectrum when they are all together reading the same movements).

What I did - took the accelerometers off their usual X Y Z setting and set the 3 MC2's and 3 MC1's next to each other covered by a box.
Then I brought MC2 X to MC1 X and placed them in a box so I have a referance between the 2 groups.

Result - Seems MC1 accelerometers are much alike and have the same power spectrum when placed together for all frequencies. MC2 accelerometers seem to do the same until approximately 30 Hz. (decided not to correct for that since we don't really care about the accelerometers in such high frequencies).

When comparing the 2 X's, they also seemed to be almost perfectly correlated. I chose the gain by dividing the two and finding the mean of that in the range of 2 to 30 Hz. After correcting for all the accelerometers, I matched the gains of each group to its X accelerometer.

You can see the plots, taking into consideration that the groups were never together (pretty messy getting the cables all around).

Here are the numbers, when the MC2 and MC1 gains are calculated by comparing them to their X direction.

gain MC1 X_over_MC2 X=


gain_MC2_Y =


gain_MC2_Z =


gain_MC1_Y =


gain_MC1_Z =

  854   Tue Aug 19 17:00:19 2008 SharonUpdate Wiener TF calibration - update
This is an update for post 814

I added the calibration gains I got for the accelerometers (I realize I am just calibrating the accelerometers to themselves and this is not m/m exactly since we don't really know which accelerometer is doing exactly what we want it to do. However, since we are talking on relative small numbers, this shouldn't really change much).

I also added another missing gain for the seismometer. Rana has previously installed a 4300 ohm resistor in the seismometer, which changed the gain to 4300/(4300+5000) = 0.46 (this is from the manual). Moreover, there is a gain of 100 on the SR560. This comes up to an extra gain of 46, meaning multiplying the seismometer's counts by 1/46.
  855   Tue Aug 19 17:15:34 2008 SharonUpdate MEDM
I plugged in the gains I got for the accelerometers in the accelerometers' filters in the PEM screen of the adaptive filter
  856   Tue Aug 19 18:55:41 2008 YoichiUpdateIOOMC unlock study update
In entry 849, I reported that the MC transmitted power drops before the sudden unlocks.
However, because C1:IOO-MC_TRANS_SUM is a slow channel, we were not sure if we can believe the timing.
So I wanted to use C1:IOO-MC_RFAMPDDC, which is a fast channel, to monitor the transmitted light power.
However, this channel was broken. So I fixed it. Details of the fixing work is reported in another entry.
The attached plot shows a recent unlock event. It is clear that in the fast channel (i.e. C1:IOO-MC_RFAMPDDC),
there is no delay between the drop of the MC power and the crazy behavior of control signals.
So it was concluded that the apparent precedence of the MC power drop in the slow channels (i.e. C1:IOO-MC_TRANS_SUM)
is just an artifact of timing inaccuracy/offset of the slow epics channels.

Sometime around 5PM, the MC started to be unwilling to even lock. It turned out that the PC drive of the FSS was going
crazy continuously. So I changed the normal values of the common gain and the fast gain, which the mcup script uses.
Now with this new setting, the MC locks happily, but still keeps unlocking.
  860   Wed Aug 20 12:04:47 2008 JenneUpdateSUSBetter diagonalization of PRM input matrix
The values here should replace those in entry #851 from yesterday.

After checking the results of the input matrix diagonalization, I have determined that Sonia's method (described in LIGO-T070168) is more effective at isolating the eigenmodes than Shihori's method (LIGO-T040054).

So, the actual new PRM input matrices are as follows:

UR1.0000.8025 -0.9993
LR0.7235 -1.1230 -1.0129
LL0.6648 -1.04521.0000

Attached are plots of the spectra of the eigenmodes, using both Shihori's and Sonia's methods. Note that there isn't a good way to get the side peak out of the eigenmodes.

I've put these into the SUS-PRM MEDM screen.
  862   Wed Aug 20 13:23:32 2008 robUpdateLockingDRMI locked

I was able to lock the DRMI this afternoon. All the optical levers have been centered.
  863   Wed Aug 20 17:02:01 2008 SharonUpdate More FIR to IIR
I tested another method for converting from FIR to IIR other than the 2 mentioned in post 841.
I got this one from Yoichi, called poles fitting, you can read about it more if you want at: http://www.rssd.esa.int/SP/LISAPATHFINDER/docs/Data_Analysis/DA_Six/Heinzel.pdf.

Seems it's not doing much good for us though.

I am attaching a PDF file with the plots, which have N=50,100,600,1000, respectivaly.
  864   Wed Aug 20 18:09:48 2008 YoichiUpdateIOOMC still unlocks
Being suspicious of FSS PC path as the culprit of the MC unlocks, I opened the FSS box and connected a probe to the TP7,
which is a test point in the PC path (before high voltage amplifier).
The signal is routed to an unused fast DAQ channel in the IOO rack. It is named C1:IOO-MC_TMP1 and recorded by the frame
builder. You can use this channel as a generic test DAQ channel later.

By looking at the attachment, the PC path (C1:IOO-MC_TMP1) goes crazy at the same time as other channels. So probably
it is not the trigger for the MC unlock.

Then I noticed the WFS signals drift away just before the unlock as shown in the attached plot. So now the WFS is the
main suspect.
Rob tweaked MC1 pitch to center the WFS QPDs while the MC is not locked. It improved the shape of the MC reflection.
However, the sudden MC unlock still happens. We then lowered the WFS gain from 0.5 to 0.3. Did not change the situation.
It looks like the MC length loop starts oscillating after the WFS signals drift away.
We will measure the WFS and MC OPLTF to see the stability of the loops tomorrow.

  868   Thu Aug 21 18:13:24 2008 ranaUpdateIOOMC WFS Control signals not responsible for lock losses
This is a 4 hour, second-trend of the MC WFS error and control signals.

There is no sign that the MC loses lock because of feedback signal saturations.
  869   Fri Aug 22 10:39:41 2008 JenneUpdateSUSTaking Free Swinging spectra of PRM, SRM, ITMX, BS
I'm taking free swinging spectra of PRM, SRM, ITMX and BS, so I've turned off their watchdogs for now. I should be done around 11:15am, so I'll turn them back on then.
  871   Fri Aug 22 16:06:29 2008 steveUpdatePEMparticle counter replaced, flowbenches & HEPAs checked
MetOne #2 counter was swapped in (on the top of IOC, facing SW direction, at ~75 deg upwards)
with channel one size 0.3 micron and channel two size 1.0 micron.
Sampling time was reduced from 60s to 6 sec at 0.1 cf/min at 25 min rate.

This means that displayed number needs to be multiplied by x100 to get particles/cf/min

HEPA filters and flow benches were checked:

PSL enclosure closed, HEPA speed at 60% 0-0 particles on optical table NW corner

AP covered optical table 1,000 particles of 0.3micron and 10 of 1.0 micron at NE corner

Flow bench at SE 0-0 particle (p)

on the top of SP cover at SE corner 60,000 p of 0.3 micron and 530 p of 1.0 micron

Mobile HEPAs 10cm from output screen in the center 800 p of 0.3 micron and 0 p of 1 micron
These filters will be replaced.

Clean assembly room:
both flow benches 0-0 p for 0.3-1.0 micron
east side bench 520 p of 0.3 micron and 210 p of 1.0 micron

Large hood in baking room with fan on 1.7 million p of 0.3 micron
and 16,000 p of 1.0 micron

Pasadena air just outside of main entrance:
3 million p of 0.3 micron and 30,000 p of 1.0 micron

My desk 743,000 p of 0.3micron, 63,000 p of 0.5 micron and 5,500 p of 1.0 micron cf/min

NOTE: existing COCHECKLIST.adl PEM displays needs to be corrected so it shows the 10 fold increase
and change particle size on this screen to 0.3 micron
  872   Fri Aug 22 17:03:41 2008 YoichiUpdateIOOMC open loop TF
I measured the open loop TF of the overall MC loop using the sum-amp A of the MC board.
I used the Agilent 4395A network analyzer and saved the data into a floppy disk. However, the data was corrupted when
I read it with my computer. I had the same problem before. The floppy is not reliable. Anyway, I have to re-measure the TF.
From what I remember, the UGF was around 25kHz and the phase margin was less than 15deg.
Above this frequency, the open loop gain was almost flat and had a small bump around 100kHz.
This bump has a gain margin of less than 4dB (the phase is more than 180deg delayed here).
So the MC is marginally stable and either decreasing or increasing the gain will make it unstable easily.
Probably, the broken FSS is responsible for this. We have to fix it.

During the measurement, I also found that the input connectors (IN1 and IN2) of the MC board are freaky.
These are TNC connectors directory mounted on the board. Gently touching the cables hooked up to those connectors
caused a large offset change in the output.
When Rana pulled the board out and pushed it in firmly, the strange behavior went away. Probably, the board was
not correctly inserted into the backplane.
This could have been the reason for the MC unlocks.
  873   Sat Aug 23 09:39:51 2008 rana, jenneUpdatePSLPMC Survey
Jenne, Rana

We scoped out the PMC situation yesterday.

Summary: Not broke. UGF ~ 500 Hz. Needs some electronics work (notches, boosts, LPFs)

Ever since we swapped out the PMC because of the broken PZT of the previous one, the UGF has been
limited to a low value. This is because the notches no longer match the mechanical resonant
frequencies of the body. The old one had a resonance at 31.3 kHz which we were notching using
the LC notch on the board as well as a dangling Pomona box in the HV line to the PZT. The one
has a resonance at ~14.5 kHz which we don't yet have a notch for. Jenne has all the real numbers and
will update this entry with them.

  • Implement the 4th order Grote low pass after the mixer.
  • Replace the AD797 with an OP27.
  • Change servo filter to have a boost (need DC gain)
  • Make a 14.5 kHz notch for the bode mode.
  • Put a 20 lb. gold-foil wrapped lead brick on the PMC.

Here's the link about the modified PMC board which we installed at LHO:
LHO PMC elog 2006
  874   Mon Aug 25 10:07:35 2008 JenneUpdatePSLNumbers for the PMC servo board (Re: entry # 873)
Jenne, Rana

These are the numbers that go along with Rana's entry #873:

The existing notch in the PMC servo is at 31.41kHz.

The power spectrum of the PMC has a peak at 14.683kHz when it is just sitting on the PSL table (no extra mass). When we put a pile of steel and aluminum (~20lbs) on top of the PMC, the body resonance moves to 14.622kHz, but is decreased by about 40 dB!

Rana has ordered a lead brick + foil that should arrive sometime this week. To complete the mechanical part of this installation, we need to extend the earthquake mounts around the PMC so that the lead brick can't fall off of the PMC onto the rest of the table.
  876   Mon Aug 25 10:51:06 2008 steveUpdatePSLpsl headtemp is coming down
The laser water cooler was overflowing this morning.
I removed 500 cc water from the chiller.

The 4 days plot shows clearly:
that the capacity of the chiller is depending on the water level.
Overflowing water is a heat load for the chiller, so laser head temp goes up.
  878   Mon Aug 25 12:13:49 2008 JenneUpdatePSLBroken PMC Servo Board
I broke the PMC servo board (on accident).

I was trying to measure the resistance of the extra resistor that someone put between the board and the HV OUT connector, since this is part of an RC filter (where C is the capacitance of the PZT on the PMC) that I need to know the values of as part of my mission to make a 14.6kHz notch for the PMC body mode. The resistance is 63.6k. I had to pull the board to get in to measure this resistance.

This resistor between the board and the center pin of the panel-mount HV OUT connector made a rigid connection between the board and the panel. When I was putting the board back in, I must have strained this connection enough that it broke. We don't have any of the same kind of resistor here at the 40m, so I'm waiting until after lunch to go to Wilson house and see if they've got any. The IFO is down until I get this sorted out.
  879   Mon Aug 25 14:18:36 2008 JenneUpdatePSLPMC servo board is fixed
The PMC servo board is back in place, all fixed up with a shiny new resistor. The PMC locks, and the MC locks (I'm not saying anything either way about how long the MC will stay locked, but it is locked for now). The resistor is connected to the connector using a short piece of wire, so this problem won't happen again, at least with this connector on this board.
  892   Wed Aug 27 13:55:43 2008 rana,jenneUpdatePSLPMC Servo Board
Board is back in. PMC is locked.

Nominal gain is now 15 dB with brick. We need to do more studies:

  • Find out why there is still 35 MHz signal at the error point. Order some low pass filters to cut off above 35 MHz.
  • Explore brick + no-brick loop shapes and error spectra.
  • Measure and set the OLG.

We've left the copper-wrapped lead brick installed to let it slowly conform to the glass better.
  895   Fri Aug 29 02:40:43 2008 rana,jenneUpdatePSLPMC Servo Board

Board is back in. PMC is locked.

This entry has details about the low pass filter after the PMC mixer. This filter has a few purposes:

1] Remove the beat signal (at 2*f_mod) between the PD RF signal at f_mod and the LO signal at f_mod.
2] Remove the beat signal (at f_mod) between the PD RF signal at 2*f_mod (which comes from the
beating of the upper and lower RF sidebands) and the LO signal at f_mod.
3] Remove other RF signals from non-ideal behavior of the LO drive signal and distortion in the RF PD pre-amp.

So its important to have a very good rejection at 35 MHz and higher. I used the Hartmut LC network design which is
installed on H1, H2, & L1. Since there is a high gain in the audio amps right after the mixer we have to get rid of
the RF or else we'll get slew rate limited or otherwise rectified downconversion of the RF signal into our audio band.

Of course, what everyone immediately realizes from the above 3 points, is that this filter can't protect the PMC
noise performance from homodyne mixing (e.g. 2*f_mod in the LO and 2*f_mod in the RF PD). To get around that, we're
ordering some filters from Mini-Circuits to remove the 2f from those signals by ~30 dB. As long as we install
the same filters on the RF and LO legs, there should be no significant phase shift in the demodulated signal.

The attached 2 page PDF shows the calculated before and after TFs of this filter. The 2 attached .m files
calculate the TF's and have ascii art which shows how the filter works.

Here's a comparison of the attenuation (in dB) of 2 candidate Mini-circuits filters:

31 0.5 0.4
35 1.3 0.4
38 6.1 0.4
40 10.8 0.42
61 46.3 14.8
71 60 29
91 76.9 48
10780 60

We don't have tabulated data at the same frequencies for both filters so I just made up some of the points by eye-balling the
plots from the catalog - but you get the idea: we can get away with using the SLP-30 at 35 MHz since it only attenuates the
signals by ~1.5 dB. So if someone can find 4 of these then Steve doesn't have to order any from Mini-Circuits.
  901   Fri Aug 29 15:01:45 2008 steveUpdatePSLMOPA_HTEMP in increasing
The laser chiller temp is 21.9C ( it should be 20.0C )
Control room temp 73F ok, no obvious block

Ops, there is a piece of paper blocking the intake of the chiller

This is a four day plot. The paper was blocking the air flow all day.
  905   Fri Aug 29 22:57:48 2008 YoichiUpdatePSLFSS loop transfer functions
I've been measuring a bunch of transfer functions of the FSS related stuffs.
There are a lot to be analyzed yet, but here I put one mystery I'm having now.
Maybe I'm missing something stupid, so your suggestions are welcome.

Here is a conceptual diagram of the FSS control board

                                                          TP3             TP4
                                                           ^               ^
                                                           |               |
RF PD -->--[Mixer]-----[Sum Amp]------>--[Common Gain]--->----[Fast Gain]----[Filter]--> NPRO PZT
              ^     |      ^        |                  |     
              |     V      |        V                  |
LO ---->-------    TP1     IN      TP2                 -->---[Filter]--[High Volt. Amp.] --> Phase Corrector

What I did was first to measure a "normal" openloop transfer function of the FSS servo.
The FSS was operated in the normal gain settings, and a signal was injected from "IN" port.
The open loop gain was measured by TP1/TP2.
Now, I disconnected the BNC cable going to the phase corrector to disable the PC path and locked the ref. cav. 
only using the PZT. This was done by reducing the "Common Gain" and "Fast Gain" by some 80dB.
Then I measured the open loop gain of this configuration. The UGF in this case was about 10kHz.
I also measured the gain difference between the "normal" and "PZT only" configurations by injecting 
a signal from "IN" and measuring TP3/TP2 and TP4/TP3 with both configurations (The signal from the Mixer was
disconnected in this measurement). 

The first attachment shows the normal open loop gain (purple) and the PZT only open loop gain scaled by the 
gain difference (about 80dB). The scaled PZT open loop gain should represent the open loop gain of the PZT
path in the normal configuration. So I expected that, at low frequencies, the scaled PZT loop TF overlaps the normal
open loop TF.
However, it is actually much larger than the normal open loop gain.
When I scale the PZT only TF by -30dB, it looks like the attachment #2.
The PZT loop gain and the total open loop gain match nicely between 20kHz and 70kHz.
Closer look will show you that small structures (e.g. around 30kHz and 200kHz) of the two
TFs also overlap very well. I repeated measurements many times and those small structures are always there (the phase is
also consistently the same). So these are not random noise.

I don't know where this 30dB discrepancy comes from. Is it the PC path eating the PZT gain ?

I have measured many other TFs. I'm analyzing these.
Here is the TO DO list:

* Cavity response plot from AOM excitation measurements.
* Cavity optical gain plot.
* Reconstruct the open loop gain from the electric gain measurements and the optical gain above.
* Using a mixer and SR560(s), make a separate feedback circuit for the PZT lock. Then use the PC path
  to measure the PC path response.
* See the response of the FSS board to large impulse/step inputs to find the cause of the PC path craziness.
etc ...
  907   Mon Sep 1 04:34:00 2008 ranaUpdatePSLFSS loop transfer functions
I started from 6th item in Yoichi's todo list.

1) Increased the setpoint of the thermostat next to the framebuilder from 73F to 79F. Its freezing over there
in the room with the drill press. Steve's illegal mercury thermometer is reading 19 C.

2) Looked the RFPD's output spectrum using the 20 dB coupled output from the coupler that's in-line.
The first attached PDF file (n.pdf) has several plots:
page 1: 0-500 MHz anomolous peaks at 138 & 181 MHz but nothing too crazy
page 2: 0-100 MHz 80 MHz peak is RF pickup from the VCO Driver - not on the light
page 3: 10-30 MHz totally nuts
page 4: 18-25 MHz that's just wrong

The RF spectrum should only have some action around 21.5 MHz and a little peak at 2x 21.5 MHz. All that extra
junk means that something is broken!

3) To see if I could rid of any of the 80 MHz signal or any of that other trash from 18-25 MHz, I wound the RF cable
around a large toroidal ferrite core. This should have given us many uH of inductance for any signals common to
both the center and shield of the cable with no effect on the differential RF signals. There was no effect.

4) Next went to look at the 21.5 MHz Crystal Oscillator Reference card (D980353...I bet you can't figure out how
this one works). These have the Mini-Circuits SMA 30 MHz low pass (SLP-30) filters on both the LO and EOM outputs.

FSSLO.PNG shows the waveform after 20 dB attenuation going into a scope terminated with 50 Ohms.
FSSLO-Spec.png shows the spectrum of this signal - its pretty distorted. Here's the levels
   f (MHz) |  before filter (dBm) | after filter (dBm)
     21.5  |       -12.8               -13.1       
     43            -24                 -46
     64.5          -50               < -80
     86            -64               < -80

This would be OK after the filter, but the level is very low. Only 7dBm (accounting for my 20 dB att) !!
The FSS uses a JMS-1H mixer which needs, as everyone knows, a +17 dBm LO signal. Que lastima.

There seems to be something wrong already, but wait...

5) PC25.PNG shows the output signal going to the EOM from 0 - 25 MHz. The step that's visible there at
around 10 MHz is just something inherent to the analyzer (??). But see all that crap there down below
5 MHz ? That is NOT supposed to be there.

pc.pdf shows on the first page the comparison in EOM drive with 2 different slider values on the
RF AM adjust screen for the FSS. But page 2 is the punchline of this long entry: There is a bunch of
excess junk on the drive signal going to the FSS's phase modulator.
The FSS is then trying to handle
this extra frequency noise and getting into trouble.

We have to fix this board. I have also ordered a few SBP-21.4 from mini-circuits (SMA bandpass around 21.4 MHz)
just in case. Another option is to just replace this thing with a Marconi and an RF amp.

  911   Tue Sep 2 10:09:03 2008 steveUpdatePSLhead temp is cooling down
The chiller was over flowing this morning.
800 cc of water was removed.
PSL-126MOPA_HTEMP peaked at 20.7 C (normal is 18.7 C)
  912   Tue Sep 2 14:28:41 2008 YoichiUpdatePSLFSS EOM driving signal spectra
Rich advised me to change the +10V input of the FSS crystal frequency reference board from whatever voltage supply we use now to a nice one.
This voltage is directory connected to the signal lines of both LO and RF output amps. Therefore, fluctuations in the voltage directly appear
in the outputs, though DC components are cut off by the AC coupling capacitors.

I changed the source of this voltage from the existing Sorensen one to a power supply sitting next to the rack.
The attached plots shows the difference of the RF output spectra between the two 10V sources.
The low frequency crap is almost gone in the new 10V spectrum.

I tried to increase the FSS gain with the new 10V, but still it goes crazy. I suspect it is because the LO power is too low.
  918   Thu Sep 4 00:38:14 2008 ranaUpdatePSLc1iovme power cycled
Entry 663 has a plot of this using the PSL/FSS/SLOWscan script. It shows that the SB's were ~8x smaller than the carrier.
P_carrier   J_0(Gamma)^2 
--------- = ------------
P_SB        J_1(Gamma)^2

Which I guess we have to solve numerically for large Gamma?
  919   Thu Sep 4 07:29:52 2008 YoichiUpdatePSLc1iovme power cycled

Entry 663 has a plot of this using the PSL/FSS/SLOWscan script. It shows that the SB's were ~8x smaller than the carrier.
P_carrier   J_0(Gamma)^2 
--------- = ------------
P_SB        J_1(Gamma)^2

Which I guess we have to solve numerically for large Gamma?

P_carrier/P_SB = 8 yields gamma=0.67.
  920   Thu Sep 4 07:46:10 2008 YoichiUpdateIOOMC is now happy
The MC has been locked for more than 12 hours continuously now !
Changes I made yesterday were:
(1) Removed the 20dB attenuator before the EOM.
(2) Reduced the Fast Gain from 21dB to 16dB, which made the PC to be a little bit more loaded (~0.6Vrms).

As Rana pointed out in the meeting, setting the Fast Gain a bit lower may have put the FSS in a stabler state.
  921   Thu Sep 4 10:13:48 2008 JenneUpdateIOOWe unlocked the MC temporarily
[Joe, Eric, Jenne]

While trying to diagnose some DAQ/PD problems (look for Joe and Eric's entry later), we unlocked the PMC, which caused (of course) the MC to unlock. So if you're looking back in the data, the unlock at ~10:08am is caused by us, not whatever problems may have been going on with the FSS. It is now locked again, and looking good.
  923   Thu Sep 4 13:48:50 2008 YoichiUpdatePSLFSS modulation depth
I scanned the reference cavity with the NPRO temperature (see the attached plot).
The power ratio between the carrier and the sideband resonances is about 26.8.
It corresponds to gamma=0.38.
The RF power fed into the EOM is now 14.75dBm (i.e. 1.7V amplitude). The NewFocus catalog says 0.1-0.3rad/V. So
gamma=0.38 is a reasonable number.

  924   Thu Sep 4 14:43:58 2008 JenneUpdatePSLPMC Open Loop Gain
I have measured the PMC's open loop gain. UGF is 629.7Hz, with a phase margin of 53 degrees.

I injected into FP2 on the front panel, and measured MixOut/Source from 100Hz to 100kHz using the SR785. I did this both when the loop was open, and when the loop was closed (open the loop by enabling FP1, which breaks the loop).

We have 2 transfer functions involved: The actual open loop gain of the PMC servo loop (G1), and the gain between FP2 and the MixerOut monitor point (G2). This gives us:

TF(closed loop) = G2*(1+G1)
TF(broken loop) = G2

G1 = TF(closed)/TF(broken) - 1

This G1 is the final open loop gain, and it is plotted below.
ELOG V3.1.3-