ELOG 40m

Mode Cleaner transfer function

IOO

I measured the transfer function of the input mode cleaner using a PDA255 and the ISS. First I put the PD in front of the ISS out-of-loop monitor diode and used an SR785 to measure the swept sine transfer function from the Analog IN port of the ISS to the intensity at the PD. Then I moved the PD to detect the light leaking out from behind MC2, using ND filters to get the same DC voltage, and measured the same transfer function. Dividing these two transfer functions should take out the response of the ISS and the PD, and leave just the transfer function of the MC. A plot of the data, along with a single-pole fit, are attached.

The fit is pretty good for a single pole at 3.79 kHz. There's a little wiggle around 9kHz due to ISS weirdness (as Tobin has not been giving it the attention it requires), but this shouldn't affect this result too much. Using the known MC length of 27.0955m, and assuming that MC1 and MC3 have a power transmissivity of 2000ppm and MC2 is perfectly reflecting, the total round trip loss should be about 300ppm. The fitted finesse is 1460.

Attachment 1: MCtf.pdf

Tue Nov 6 10:42:01 2007

Increased the power exiting the PSL by turning the half-wave plate after the MOPA, opened the PSL shutter, and aligned the mode cleaner to the input beam. It wasn't that hard to find the beam with the aperture open all the way on the MC2 camera. The transmitted power is now 2.9 arbitrary units, while the input power is 1.2 arbitrary units. Not sure yet if that's an increase or decrease in efficiency, since no one posted numbers before the vent. Also turned on the input-steering PZTs and saw a REFL beam on the camera.

Tue Nov 6 15:36:03 2007

Easily, after resetting the PSL Uniblitz shutters. There's no entry from David or Andrey about the recovery from last week's power outage, in which they could have indicated where the procedure was lacking/obscure. Tsk, tsk.

Tue Nov 6 15:37:34 2007

/cvs/cds/caltech/scripts/SUS/rampdown.pl is now in the crontab for op340m, running every half-hour at 15&45. It checks the suspension watchdog trip levels, and reduces them by 20 if they are above 150.

Wed Nov 7 13:54:44 2007

Mode Cleaner transfer function

IOO

I performed the same procedure described here, and re-measured the transfer function of the mode cleaner to see the effect of the drag-wiping. The results are attached in a pdf. We don't seem to have done any damage, but the improvements are barely measurable.

What	Then	Now
pole frequency	3.789kHz	3.765kHz
loss per optic	99ppm	91ppm
finesse	1460	1470
trans	86.7%	87.7%

Attachment 1: mctf.pdf

Fri Nov 9 17:33:33 2007

The 3.7 MHz is actually on the light. It's the beat between the 29.5 MHz sidebands and the 33.2 MHz sidebands. There are pads in the ISS PCB for a filter to notch this frequency--John is working on it.

I also found a 1.2 ND filter on the lens which focuses the beam on the ISS diodes. I replaced it with a 0.6 ND filter, which brought the ISS DC level (on the screen) up to ~4.2 (it saturates at 5). Once John finishes the filter we should be able to crank up the gain.

Fri Nov 9 21:36:14 2007

rob, rana

We looked at the FSS a bit today. The most we could get out of it with the gain sliders was a UGF of around 95kHz. After a bit of tweaking the waveplate after the AOM, this got up to ~115kHz. We should be able to get at least 500kHz. This system needs a fair amount of work.

Mon Nov 12 14:09:19 2007

The testpoint manager was dead on fb40m. I know I re-started it sometime after the power outage, so something must have killed it. If you get an error from DTT like
"diagnostic kernel does not support: testpoints", then log into fb40m as root, check for the tpman with a ps -ef | grep tpman. If it's not there, then run /usr/controls/tpman & and close the terminal window.

Mon Nov 12 15:05:49 2007

Spent a bit of time fiddling with the FSS again today. In a not-particularly-systematic manner, I raised the input-side of the 21.5MHz PC, adjusted the half-wave plate in front of it, touched up the RC alignment and the alignment onto the transmitted and reflected diodes. This got us a ~15% increase in
transmitted light, and I was able to push the UGF to 140kHz with the common gain slider at 30dB and the FAST gain slider at 22dB. The next options include adjusting the AOM setup, mode matching into the RC, and just increasing the pickoff fraction right from the getgo.

Mon Nov 12 15:18:34 2007

After John soldered a 3.7 MHz notch filter onto the ISS board, I took a quick TF and RIN measurement. The out-of-loop RIN is attached, including a dark noise trace, and with the gain slider at 10dB. The UGF is 35kHz with a phase margin of 30deg. John is currently doing a more thorough inspection, and will detail his findings in a subentry.

Attachment 1: ISS.png

121

Wed Nov 21 14:31:41 2007

I `tweaked' the FSS path today. Here's what I did:

1) Shut down the FSS autolocker

2) Turn off FSS servo

3) Assume the beam coming back from the AOM is double-first-order, and don't make any changes large enough to lose it.

4) Tweak the alignment of these components to maximize the incident power on the RC reflected diode:

a) PBS before AOM
b) AOM
c) curved mirror after the AOM

5) Translate the AOM such that the beam moves away from the PZT, then when it levels off (no more power gains with movement),
move it back just a little bit so there's a teensy drop in power. This should but the beam as close to the edge as possible,
but whether or not it's the best place is still to be determined.

6) Lock the FSS, and align the mirrors into the frequency reference cavity.

After all this, the RC transmitted power went from .57 to .73 -- probably not a big enough change to account for the missing loop
gain, but we'll know more once the loop gets measured (after Alberto stops hogging the Agilent network analyzer).

Other possible routes include a systematic check of the upstream path (e.g., the Pockels cell) and just increasing the pickoff fraction for the FSS.

124

Tue Nov 27 15:45:08 2007

It's unclear (to me, at least) what was the end result of the FSS path tweaking before Thanksgiving. Today I measured the open loop gain, and it was still around 100kHz, even with the gain sliders maxed out, but it looked really crappy with a sharp cutoff around the UGF. Then, on a lark, I pushed around the "Input Offset Adjust" slider, which sums an offset into the signal coming out of the mixer. By moving this slider to 7V, I got the UGF to 500kHz with 45 deg of phase. That would be fine, and we could go offset hunting, but the same thing happens if one puts in a large negative value! I don't really understand what's going on, but it seems like weirdness in the electronics. Unfortunately the web interface to the conlog is not running (presumably because the `new' linux1 doesn't have its apache server running) and my command line conlog efforts have been stymied. So, I don't know what the historical settings of this offset are, but zero is definitely not a good setting right now. Here's a snapshot:

FSS
UGF: 500kHz
CG : 24dB
FG : 19dB
input offset: 7V
Phase Adjust: 1.09V
Phase Button: 0
RF Amp Adjust: 7.38V

margins:
phase: 45 deg
gain: 8dB

Attachment 1: FSSsmall.jpg

125

Tue Nov 27 15:47:17 2007

After the FSS running pretty quick, I checked the MC loop. I used TPA 1&2.

MC loop
UGF: 70kHz
Input Gain: 29dB
Boost Level: 2
phase: 40 deg

Attachment 1: MCsmall.jpg

126

Tue Nov 27 16:18:58 2007

Reduced the common gain to 22dB in the mcup script, so that the WFS would not blow the lock. The above measure of the OLG was done without the mcWFS running, so may be a low estimate as compared to when the alignment is perfect.

134

Wed Nov 28 17:41:34 2007

I investigated the FSS a bit more today. I looked at the signals coming out of the FSS frequency reference, and saw that both the LO and PC drive were distorted, non-symmetric waveforms. In addition, the LO path had a 3dB attenuator, meaning the mixer was starved. I placed mini-circuits SLP-30 filters in both paths, and now both are nice sine waves. I also took out the 3dB att. With this work, and the CG slider maxed out at 30, the FSS open loop gain (for real this time) goes up to ~250kHz. Still needs more investigation.

139

Thu Nov 29 11:10:54 2007

Without the RGA controller responding, the RGAlogger script just hangs. Rather than fix it, I just put it to sleep by commenting out the line in op440m crontab file. Once we get it running again, we'll move the cronjob to op340m.

141

Thu Nov 29 15:17:53 2007

I put some ISS beam on the diode on the PSL table. In the previous layout, this was the monitor diode (and it's labeled monitor) but I plugged it into the sensor jack anyways so we can run with the loop closed for now; we can just switch the cables later. The reason the beam was unclear is because someone popped up a flipper mirror which redirects the beam from the ISS into an OSA.

With the ISS gain slider at 15 dB the UGF is around 40kHz.

Why do we have such short cables for the ISS diodes?

145

Fri Nov 30 11:44:57 2007

In the interests of getting the Xarm alignment script working again, I reset the local damping gains for the test masses to their previous known working values (1), then I noticed that the ETMX oplev was dead. Since the scripts use the oplev motion as a readback for the optic motion, this means the script was basically blindly swinging the optics around. Some monkeying around with swapping HeNe power supplies eventually led to the conclusion that the power strip is funky, since the laser works when plugged into another power strip. Even weirder, the HeNe and the power supply indicator light have some sort of XOR relationship going on. When one works, the other doesn't. Steve will sort out this confusion later; we're good for now.

146

Fri Nov 30 13:46:50 2007

Electronics

ETMX oplev dead again

Quote:

Ech. The HeNe quit again. Let's replace it and see what happens.

175

Thu Dec 6 18:11:15 2007

HowTo

Making DMF monitors

I was able to use the matlab compiler to compile a version of the linetracker written by Rana, and run the compiled version on mafalda.

I believe I made the necessary edits to our mDV config file so that it should be easy for others to follow these steps:

1) Write the DMF routine you want, as a matlab function (not a script).

2) If it runs correctly in matlab, then from the matlab command line compile
it using the -m flag (i.e., mcc -m myfun.m). You should run the
compiler from the directory where you want the executable to end up (don't use the mDV/extra
directory so it doesn't get all cluttered).

3) prior to running the resulting executable (which should be called simply myfun),
prepend the directories
/cvs/cds/caltech/apps/linux/matlab/bin/glnx86
/cvs/cds/caltech/apps/linux/matlab/sys/os/glnx86/
to the LD_LIBRARY_PATH enviroment variable. These directories must be prepended as the
versions that already exist in /usr/lib don't work; I'm loathe to do this in the cshrc.40m
for fear of later conflicts, so it will need to be done in some sort of shell script which
calls the matlab executable.

180

Fri Dec 7 14:14:48 2007

Metaphysics

tdsread problems on Solaris

tdsread has developed a strange new illness, whereby it cannot read EPICS values from two subsystems at once (e.g., getting an LSC and SUS value simultaneously). I thought this might have something to with the fact that both losepics and iscepics are running on the same box,
but the same thing happens with IOO EPICS records, so that's not the culprit.

This is new behaviour, and it's only happening on the solaris machines. I suspect some ENV/cshrc juju has caused it, as the tdsread executable is the same one from April, and I don't think our EPICS infrastructure has changed otherwise. In the near term we can either try running the scripts on linux, or modify the IFO scripts to not do these types of calls.

184

Mon Dec 10 13:54:26 2007

HowTo

Don't blame the Matlab compiler

For human error. We should be careful to only run the compiler outside the mDV directory (thus placing the _mcr outside of the range the addpath command in the mdv_config files). Or maybe there's a smarter solution...

224

Thu Jan 3 12:38:49 2008

Quote:

I did not feel anything wrong yesterday, but unfortunately I have a very much sore throat today. I need to drink warm milk with honey and rinse my throat often today. So far I do not have other illness symptomes (no fever), so I hope that this small disease will not last for a long time, but I feel that it is better for me to cure my sore throat today at home (and probably it is safer for others in 40-m).

I took yesterday the book "Digital Signal Processing", so I have it for reading at home.

Hope to see you tomorrow.

I've added a new category--TMI--for entries along these lines.

240

Wed Jan 16 14:06:24 2008

rob, tobin, johnnie

We did some locking work monday night, with decent progress. Working in the PRFPMI style, we managed to get through the part of the script that hands off the offset-CARM DOF to the MCL, but were not successful in engaging the AO path.

We also confirmed the problem with tdsread which prevent it from reading from multiple TLS (Three Lettered Subsystems) at the same time. Tobin traced this to a problem with the ezca library which tds uses, but it's not clear how to fix it. For now we just split the tdsread calls so that there are no multiple TLS calls. Tobin will report further on this.

241

Wed Jan 16 14:09:45 2008

I got a little further with the locking (PRFPMI) last night, after discovering that the cable going from the CM board to the MC board was unplugged at the MC side. This explains why we weren't able to engage the AO path last night. Tonight, I got up to the point where DARM is handed off to OMC transmission, a step which repeatedly failed.
Eventually I realized that although all the lights are the green, the OMC Trans signal was not being updated in the LSC's memory. I suspect this is because the c1ass machine was powered down. Work continues.

244

Thu Jan 17 14:13:20 2008

Incremental progress on locking yet again. This time the handoff of DARM to the OMC worked, and progress halted at handing off control of the common mode to REFL166.

249

Fri Jan 18 15:31:47 2008

rob, johnnie, andrey

On Thursday night we got the intereferometer fully locked in a power-recycled FPMI state. The obstacles included the REFL166 phase being wrong by 180 deg (because that's the correct phase for DRMI locking) and getting confused (again) by the "manual" mode dewhite switching at the ETMs. After turning on the dewhites and the MICH correction, we took the noise spectrum below.

Attachment 1: DARMnoise080118.png

252

Tue Jan 22 02:33:45 2008

0) The ETMY oplev needs work/centering

1) recentered DRMI oplevs

2) Did some light DRMI locking. Looked at the loops and the DD signals. The PODD signals look flaky; the beam may not be on the diode. MICH and PRC handoffs to DD signals were spotty, but not a total disaster. Changed the PD9 phase by 115 degs. Work continues on the DD_handoff subscript.

3) John says "There are ants everywhere."

4) Andrey is now versed in the arts of decimation.

263

Fri Jan 25 08:55:26 2008

Changes to Dataviewer channels (XARM)

General

As a general rule,

Quote:

clicking random blue buttons chaotically

is not a good problem solving technique. It is thus now explicitly discouraged as an option in the LIGO 40m Lab.

265

Fri Jan 25 10:14:35 2008

Changes to Dataviewer channels (XARM)

SUS

Quote:

2) BAD NEWS. While "clicking the appropriate blue button" on the DAQ MEDM screen,
namely CODAQ_DETAIL,adl screen, I obviously clicked some blue button that I should not have clicked,
and as a result the signal in Dataviewer from the channel "C1:SUS-ITMX_POS" has disappeared (it is now a straight line).

Description of what has happened and of my wrong actions:
I had two channels opened in Dataviewer simultaneously (both "C1:SUS-ETMX_POS" and "C1:SUS-ITMX_POS"),
and after clicking some blue button on CODAQ_DETAIL,adl screen, the signal from "C1:SUS-ITMX_POS" became
a straight line, while signal from "C1:SUS_ETMX_POS" continued to be a random noise.

I was scared that I made worse for the channels and for Dataviewer, and I started clicking random blue buttons chaotically hoping that it will restore the signal from "C1:SUS-ITMX_POS". Random clicking on arbitrary blue buttons did not return the signal.

As the channel "C1:SUS-ETMX_POS" works normally, I will be measuring Q-factors of ETMX tonight,
but it is obvious that someone else (Rana, Robert,Steve?) needs to restore the correct settings for "C1:SUS-ITMX_POS".

Moreover, as I was clicking chaotically all the blue buttons on CODAQ_DETAIL,adl screen, someone else (Rana, Robert, Steve?) will need to check somehow that I did not destroy signals from some other channels.

I apologize for the negative consequences of my channel adding,

but Rana asked me in the very beginning in September to let others know if I spoil something, so that others would be aware of it and could fix the problem.

I eventually resolved the situation by restarting the c1susvme1 processor, which had somehow got confused by the clicking random blue buttons chaotically. The data acquisition should be working again.

280

Mon Jan 28 15:11:38 2008

running compiled matlab DMF tools

HowTo

DMF

I compiled Rana's seisBLRMS monitor, and it's now running on mafalda. To start your own DMF tools, here is a procedure:

1) build your tool in mDV, get it working the way you'd like.
2) Make a new directory /cvs/cds/caltech/apps/DMF/compiled_matlab/{your_new_directory} and copy the *.m file there.
3) Make the *.m in your new directory into a function with no args (just add a function line at the top)
4) compile it (from within a fully mDV-configured matlab) with mcc -m -R -nojvm {yourfile}.m at the matlab command line.
5) add a line corresponding to your new tool to the script /cvs/cds/caltech/apps/DMF/scripts/start_all
6) Run the start_all script referenced in part (5).

NB: Steps (4) and (6) must be carried out on mafalda.

284

Tue Jan 29 14:56:39 2008

The seisBLRMS 1.0 program crashed at ~7:20 pm last night, so we didn't get data from overnight. It crashed when framecaching failed. I added
a try-catch-end statement around the call to dttfft2 to let the program survive this, then compiled it and started it on mafalda. After ~45 minutes, the compiled version encountered the same error, and while it didn't crash per se, after 20 minutes it still wasn't able to read data. We may have to dig deeper into the guts of mDV to make this stuff run more robustly.

287

Wed Jan 30 20:39:31 2008

In order to reduce the probability of seisBLRMS crashing due to unavailability of data, I edited seisBLRMS.m so that it displays data from 6 minutes in the past, rather than 3. After compiling, this version ran for ~8 hours without crashing. I've killed the process now because it seems to interfere with alignment scripts that use ezcademod, causing "DATA RECEIVING ERROR 4608" messages. These don't cause ezcademod to crash, but so many of them are spit out that the scripts don't work very well. I guess running DMF constantly is just making the framebuilder work too hard with disk access. In the near term, we can maybe work around this by having DMF programs check the AutoDithering bit of the IFO state vector, and just not try to get data when we're running these sorts of scripts.

291

Fri Feb 1 12:37:39 2008

Here are DV trends of the output of seisBLRMS over the last ~36 hours (which is how long it's been running), and another of the last 2 hours (which show the construction crew taking what appears to be a lunch break).

Attachment 1: seis36hours.png

Attachment 2: seis2hours.png

307

Sun Feb 10 21:43:16 2008

I adjusted the alignment of the free hanging mode cleaner to best transmit the PSL beam.

309

Mon Feb 11 22:44:29 2008

Change in channel trending on C1SUS2

DAQ

I removed the MC2 optical lever related channels from trends, and the SRM POUT and YOUT (as these are redundant if we're also trending PERROR and YERROR). I did this because the c1susvme2 processor was having bursts of un-syncy lateness every ~15 seconds or so, and I suspected this might interfere with locking activities. This behaviour appears to have been happening for a month or so, and has been getting steadily worse. Rebooting did not fix the issue, but it appears that removing some trends has actually helped. Attached is a 50 day trend of the c1susvme2 sync-fe monitor.

Attachment 1: srm_sync.png

310

Tue Feb 12 13:53:27 2008

The new RGA head was installed a few days ago. I just ran the RGAlogger script to see if it works, which it does. I also edited the crontab file on op340m to run the RGAlogger script every night at 1:25 AM. It should run tonight.

Attachment 1: RGA-080212.png

311

Tue Feb 12 16:18:29 2008