Today we have spent most of the day trying to lock. We have locked, but only after some diifictulties.
Chris, Alexa
The first two attachments show the OLTFs of the various length DOFs. In these plots, the model is the blue trace and the red trace is measured data. The DARM TF agrees pretty well. The PRCL and SRCL loops appear to have a gain missing. The CARM and MICH model loops are not correct.
The fourth attachment depicts the noise budget produced by the model for the 8W lock on March 04, 2015 15:30 UTC. This roughly agrees with the conclusions Evan has made with his NB (LHO#17101). At high frequencies we are dominated by shot noise. At around 300 Hz we suspect intensity noise (or maybe beam jitter), as Gabriele noticed. From 30-100 Hz we are limited by DAC noise. And below 10 Hz we see some angular fluctuations.
The modeled MICH loop gain is looking better now. There were some aggressive BSFM plant inversion filters that weren't being linearized accurately in Simulink. This was fixed by using numerical TFs for them. The shape is still off above 60 Hz, but that's probably due to the filter change from alog 17148.
I put in sign flips as needed to ensure an even number of them are present in each OLTF. There remains some 10 dB of mystery gain to be hunted for in the DRMI loops.
SudarshanK, RickS
Today we had a quick look at our calibration data from Yend, taken after installing a polarizing beamsplitter downstream of the AOM (see aLOG 17145).
We will post a more detailed log after we have fully digested the data, but the preliminary result is that the new calibration of the Tx PD signal is within 6% of the previously reported number (see aLOG 16718).
This is about what we expected from our recent measurements of the signal drifts resulting from depolarization in the AOM.
Added 689 channels. Removed 123 channels.
The X direction at end X and the Y direction at the Y end are blended at 90mHz. Sensor correction is in the normal conifugration (no BRS). This was done at around 22:58 because we have had several lock losses due to large motions in the arms that ALS cannot handle.
This was at about -30 minutes on this screenshot, as you can see the arm control signals are reduced with the higher blends.
There is nothing that look like high ground motion on the seismic FOMs, although the winds are a little elevated (gusty 10-25 mph)
Here's a screen shot of the wind trend. Sheila switched the blends roughly 1 hour in the past on this trend.
Attached is a screenshot of the PEM's seismic FOM. So now we want to look at ISI T240s - 5 hours ago (~18:45 UTC -- after the Earthquake at -6 on the SEI FOM), - 3 hours ago (~20:45 UTC -- when the wind picks up, before the blends were switched) - and now (~22:45 UTC -- when the wind is still up, and when the blends have switched) Take a look at the time series of these few IFO channels: H1:ALS-(X/Y)_REFL_CTRL_OUT(MON or _DQ, whichever) this is what Sheila typically uses to assess badness. (and they're calibrated into [um]'s )
Corner station to X-1-5 double doors of the beam tube has been cleaned. Lights were relocated, support tubes vacuumed and moving north towards mid station from X-1-5 double doors. Looking into purchasing a suitable generator for this operation, the previous one not necessarily designed for this type of duty. We are still able to use site power for this section of cleaning. Beam tube pressures continually monitored by control room operator during cleaning operations.
Times in PST
07:00 Karen and Cris to LVEA
08:49 Richard to EY electronics bay
09:11 Peter King to Mid X
09:17 Richard back, Fil to EX electronics bay
09:30 Peter back
09:37 Fil back
09:39 Peter to Mid X
09:48 Corey to squeezer bay
10:10 Peter and Richard back from MX
10:26 Kyle to LVEA checking pumps
11:56 Corey out
13:38 Corey to squeezer bay for 20 min.
15:33 Dave/Ryan Fisher restarting OCD Master model
I'm posting some plots of the performance of the BSC-ISI's. The attached pdf has 18 plots, the first 12 show the contributions of each control path to the performance of the ISI, shown against aligo requirements, sensor noises and ground motion. The order is St1 X (p. 1), St2 X(p. 2), St1 Y(p. 3),... for x, y, z, rx, ry, rz. The configurations measured were offline, damped ISI, St1 ISI isolated / St2 damped, isolated (both stages), isolated with tilt decoupling, isolated with tilt decoupling and sensor correction, isolated with tilt, sensor correction and feedforward from HEPI. HEPI was off for the offline measurment as well. The blend filters used were the normal blend filters (the rdr compliment of filters with the T750 blend on St1 RZ). Please note that on the rotational plots, I used "paralell" ground direction (Y for RX, X for RY, Z for RZ). No, it doesn't make perfect sense, but X can be injected to RY or vice versa, and the BSC's have a coupling from Z to RZ, that we need to be aware of.
The last 6 plots show the final performance (i.e. our current configuration with damping, isolation, tilt decoupling, sensor correction and feedforward) for each dof for both stages.
Similar plots for the HAM's are next.
The Z and RZ ST2 isolation loops were mistakenly using a high-frequency, 750 [mHz] blend in these performance plots. See LHO aLOG 17222 for further discussion, but we think we can easily do much better in these DOFs between 0.5 and 1 [Hz]. We will remeasure and repost similar, improved "performance progression" plots in the future.
The data concentrator h1dc0 died from a kernel panic just before 9:00 PDT this morning and required the reboot button to be pressed for recovery. The startup was delayed by the need for a file system check on the boot disk (343 days since last check), so data collection didn't restart until 9:34 PDT. Also of note, monit did not cleanly restart the daqd process. The mx_stream needed to be manually restarted on h1sus2b, h1susauxh34, h1susex, h1susauxex, and h1susauxey.
Evan, Alexa
This morning we found the Y-ARM IR fiber polarization to be in the wrong polarization at 45%. Evan and I adjusted this back down to 2%. It's been a while since we have to do this.
Aidan, Nutsinee, Jim, Dave:
an apache web server was installed on h1hwsmsr to provide web access to the HWS camera image files. This is an internal web service with no offsite access running HTTP protocol. Its URL is http://h1hwsmsr
model restarts logged for Sun 08/Mar/2015
2015_03_08 00:26 h1fw1
one unexpected restart. Sunday reporting was given Saturday's date, possibly caused by time change to PDT. All times subsequent are now local PDT.
model restarts logged for Mon 09/Mar/2015
no restarts reported
model restarts logged for Tue 10/Mar/2015
2015_03_10 09:49 h1calcs
2015_03_10 09:54 h1calcs
2015_03_10 10:10 h1odcmaster
2015_03_10 11:30 h1pemex
2015_03_10 11:36 h1alsex
2015_03_10 11:36 h1calex
2015_03_10 11:37 h1calex
2015_03_10 11:38 h1iscex
2015_03_10 11:41 h1pemex
2015_03_10 11:45 h1iscex
2015_03_10 11:50 h1iopiscex
2015_03_10 11:50 h1iscex
2015_03_10 11:52 h1alsex
2015_03_10 11:52 h1pemex
2015_03_10 11:56 h1calex
2015_03_10 11:56 h1pemex
2015_03_10 11:58 h1alsex
2015_03_10 11:58 h1calex
2015_03_10 11:58 h1iscex
2015_03_10 11:59 h1iscex
2015_03_10 12:07 h1asc
2015_03_10 12:16 h1iscex
2015_03_10 12:21 h1iscex
2015_03_10 12:26 h1iscex
2015_03_10 12:47 h1omc
2015_03_10 13:41 h1omc
2015_03_10 13:46 h1lsc
2015_03_10 14:41 h1odcmaster
2015_03_10 15:01 h1alsey
2015_03_10 15:01 h1iopiscey
2015_03_10 15:01 h1pemey
2015_03_10 15:03 h1caley
2015_03_10 15:05 h1iscey
2015_03_10 15:09 h1broadcast0
2015_03_10 15:09 h1dc0
2015_03_10 15:09 h1fw0
2015_03_10 15:09 h1fw1
2015_03_10 15:09 h1nds0
2015_03_10 15:09 h1nds1
2015_03_10 15:09 h1odcmaster
2015_03_10 15:11 h1calcs
2015_03_10 15:13 h1omc
2015_03_10 16:26 h1pemex
2015_03_10 16:34 h1pemey
2015_03_10 16:36 h1broadcast0
2015_03_10 16:36 h1dc0
2015_03_10 16:36 h1fw0
2015_03_10 16:36 h1fw1
2015_03_10 16:36 h1nds0
2015_03_10 16:36 h1nds1
2015_03_10 16:56 h1pemex
2015_03_10 16:56 h1pemey
2015_03_10 17:05 h1dc0
2015_03_10 17:06 h1dc0
2015_03_10 17:07 h1broadcast0
2015_03_10 17:07 h1fw0
2015_03_10 17:07 h1fw1
2015_03_10 17:07 h1nds0
2015_03_10 17:07 h1nds1
no unexpected restarts. Please refer to slow controls alog for Beckhoff restarts. FE and DAQ details in CDS maintenance alog.
CDS maintenance summary, Tuesday 10th Mach 2015.
ODC work, WP5095
TJ, Ryan, Stefan, Daniel, Dave:
All three ODC models were slowed from 32kHz to 16kHz. Additional shared memory ODC IPC transmitter channels were added to the end station PEM, CAL and ALS models. Further ODC changes were made, please see the ODC team's alogs for details.
Shuffle of end station ISC models WP5095
Daniel, Jim, Dave:
The PEM and CAL models, which were sharing a single core, were split back into their own cores. The ODC model, which had its own core, was combined with the ISC model and given an RFM IPC sender. The original CAL DCU-IDs were used, and the new ones assigned to ALS were returned to the pool. The CAL core number was changed as it was conflicting with ISC
In the following table, each model's dcuid and core number is shown. Models which share the same core are color coded.
model | was | is now |
h1pemex | 84,2 | 84,2 |
h1calex | 84,2 | 124,4 |
h1iscex | 126,5 | 86,5 |
h1odcx | 86,5 | 86,5 |
h1alsex | 85,3 | 85,3 |
h1pemey | 94,2 | 94,2 |
h1caley | 94,2 | 125,4 |
h1iscey | 127,5 | 96,5 |
h1odcy | 96,4 | 96,5 |
h1alsey | 95,3 | 95,3 |
The dcuids 126 and 127 are no longer used.
The data from the PEM models were slow in recovering, I eventually remembered that the ADC part in the model should be changed from ADC1 to ADC0 and all the internal bus selectors modified accordingly.
PCAL model changes
Rick, Sudarshan, Dave:
The PCAL common model was modified to give a more logical ordering of its inputs (in order of ADC channel number, PCAL signals first, Timing signals last). h1calex and h1caley were modified accordingly.
Attempt to pre-load the PSL ODC DAQ channel type change
Daniel, Stefan, Ryan, Dave:
It was suggested that the fix of the PSL ISS ODC DAQ data type problem (it is FLOAT, should be 32bit UINT) could be preloaded and applied whenever the PSL is next restarted. Well, in hindsight this is not possible since the new INI file was taken by the DAQ on restart. I backed out this change and verified that only the ODC channel was corrupted for a few hours when this was tried.
DAQ restarts were necessary to support the above work, a few more than originally intended due to PEM and PSL changes.
Stefan, Elli
After maintenance day today there seemed to be a lot of changes to the interferometer confirguration, so we decided to try to lock the full interferometer to check this still works. When requesting Check_IR in LSC_Lock guardian, the ALS_COMM was not finding the comm beat note. The IR buildup in the x-arm (LSC-TR_X_A_LF_OUTPUT) was 0.7 rather than 1, and guardian would not recognise the IR resonance. We suspected a bad alignment but we did the initial alignment twice (getting an x-arm buildup of 1 at the input_align step) and repeating the initial alignment didn't help. In order to push on to ALS_DIFF, Stefan was lowering the LSC power scale from 3 to 2W (IMC input power stayed at 3W) so that the guardian would rocognise the comm beatnote.
Once IR comm and diff beat notes were found, we brought the LSC power scale back 3W and let DRMI lock. DRMI would catch lock but soon after the ASC loops engaged they would break the lock. PRC1 broke the lock twice, so Stefan turned off the PRC1 loop, and could keep DRMI locked for a bittle longer, but then SRC2 took it down. These are the two pointing loops.
Nutsinee, Elli
This morning we visited end-x and end-y to get the HWS working there. We can remote login to the h1hwsey and h1hwsex computers and the relevant software is installed. Daniel connected camera and frame grabber enable switches in the Beckhoff system (alog 17176). We confirmed cameras and frame grabbers turn on when requested by epics. We can run serial_cmd and take. On Thursday we will check alignment of the green beam onto the cameras.
Alexa, Sheila, Evan, Chris, Stefan We implemented Peter's DARM 'SUScomp' from alog 16728. Since we can't lock with that filter directly (see alog 16840), deleted the old 'LLO' filter, but instead loaded a difference filter called 'acqLP' that makes the 'SUScomp' look like the old 'LLO' filter: zpk([80;500-800i;500+800i],[50;70;200],1,"n") Guardian was updated to turn off acqLP' in FM8 instead of turning on a lead. A note on the previous filter is also found in alog 16381.
I have attached a plot comparing our various configurations:
1. Red trace: RF DARM sus compenstation as designed by Peter to obtain more phase margin. The LSC DARM configuration is: FM1(suscomp), FM2(2:0), FM3(resG), FM4(4^2:1^2), FM5(2:0), Gain 800 (LHO#16728, 16840)
2. Blue trace: Our old RF DARM sus compenstation where we used the LLO control filer and a 200Hz lead filter. (LHO#16381)
3. Green trace: ALS DIFF sus compensation. This is the configuration we use to lock ALS DIFF. The LSC DARM configuration is: FM1(suscomp), FM2(2:0), FM3(resG), FM7(SB60), FM8(acqLP),FM10(RLP33), Gain 400. As Stefan mentioned, FM1+FM8 returns our old LLO control filter.
I have also attached the RF DARM OLTF model with the new (red) and old (blue) configuration as described above, along with the respective measured data. The RF DARM UGF is now 55 Hz with a phase margin of ~45deg.
Measurement of new DARM loop on dc readout is attached.
I've saved Evan's .xml to the calibration repository here: /ligo/svncommon/CalSVN/aligocalibration/trunk/Runs/PreER7/H1/Measurements/DARMOLGTFs/2015-03-09_DARM_OLGTF_LHOaLOG17153.xml and exported text files of the transfer function and coherence, /ligo/svncommon/CalSVN/aligocalibration/trunk/Runs/PreER7/H1/Measurements/DARMOLGTFs/ 2015-03-09_H1_DARM_OLGTF_LHOaLOG17153_coh.txt 2015-03-09_H1_DARM_OLGTF_LHOaLOG17153_tf.txt Transfer function contains the following columns (i.e. I exported IN1 / IN2 "as is"): Frequency [Hz] Real Part [ ] Imaginary Part [ ] We'll use later for calibration / noisebudget model verification!
Also including the CARM OLTF that we took at 9 W.
I checked the measured DARM open loop transfer function posted by Evan against my DARM open loop model. Even though I did not do a fitting or any fancy analysis yet, it seems that the optical gain was consistent -- the measurement matched the model with an optical gain of 1.1x106 cnts/m or 9.09x10-7 m/cnts which we have been using since Feb. 21st (alog 16843) for the CAL-CS front end model.
Here is a plot showing the model and measured one:
The matlab script to generate this plot is archived in calSVN:
/ligo/svncommon/CalSVN/aligocalibration/trunk/Runs/PreER7/H1/Scripts/DARM_OLTFGTF_LHOaLOG17153.m
Note for myself:
optical gain in the model = 1.1e6
ESD strength in the model = 2.8e-10 [N/V^2] (see alog 16843)
LSC_DARM_GAIN = 800 (instead of 400)
Only ETMX was actuated
In response to Evan's alog (alog 17065), I took a look at the DARM spectra. Here are conclusions at the moment:
(Noise spectra)
The data sets that I used are from:
Here is a comparison of all three curves with the GWINC theoretical curves above 400 Hz up to 7600 Hz.
As Evan reported, indeed the measured curves from last night are lower than the GWINC curve in 1 - 4 kHz band while the one from Feb-26 looks fine.
(Discrepancies between the curves)
Now, I want to answer how much the Mar-4 data differed from the one from Feb-26th by taking the ratio between them. I divided the Feb-26th by Mar-4th spectra in 400-7600 Hz band. Then I convert it into a histogram to see how they differ on average. Since there were many peaks whose amplitude varied as a function to time, I excluded them by limiting the histogram range from 0 to 2. The ratio is shown as red bars in the below plot.
Note that I could have done a fancy Gaussian fit for it, but for now I picked the highest bar in the histogram in order to coarsely estimate the ratio. As shown in the plot, the Feb-26 data had a higher noise level by a factor of 1.09 on average.
Then I did the same ratio analysis for the 2.8 W and 8 W data of Mar-4th. It is shown as blue bars in the same histogram plot. Picking the highest bar, I measured the ratio to be 0.58 which agrees with what we expected i.e. sqrt( 2.8W / 8W) = 0.59. So the power scaling from 2.8 W to 8 W seems to have been done correctly last night.
(Unexplainable dip at around 2.8 kHz in the 8 W data)
However, it is not the end of the story yet. The noise curve of the Mar-4 at 8 W had a funny feature at around 2.8 kHz where the noise go down below the GWINC curve even if i apply the 9 % correction.
The below plot shows "normalized" spectra of all the three data sets. In order to line up all the spectra at the same level, I "normalized" the Mar-4th-2.8W data by multiplying a factor of 1.09. In a similar manner, I "normalized" the Mar-4th-8W data by a factor of 1.09/0.59. In this way I checked the shape of all the spectra.
The Mar-4th-8W data was lower by the rest of the two curves by 10-ish %. I did not do a serious histogram analysis.
Finally , if I apply only the 1.09 correction factor to the data from last night, they look like this:
Apparently the Mar-4-8W data is lower at around 2.8 kHz than the GWINC curve.
At least part (maybe all) of the issue here is actually due to the GWINC curves that we're using right now. In particular, I had put in a value for the arm losses that was too high.
By putting in 50 ppm per optic (i.e., 100 ppm per arm, which is closer to reality than the 180 ppm I was using before), I get a GWINC curve that is below the measured calibrated strain curve. This is shown for the recent 8 W lock in the attached noise budget.
Out of curiosity, I've also shown a rough estimate for how much DAC-induced ESD noise we can expect if the proposed low-pass filtering is installed (pole at 1.6 Hz, zero at 53 Hz). Obviously this is subject to the same uncertainty as the current noise trace with regard to the magnitude of the actuation coefficient.
Notes: