re WPs 4012 & 3884, I brought down & restarted the HEPI Pump Station Pressure maintenance servo yesterday.  I am working on an operations manual but for now know that the FWD button inside the Pump Station Control Panel has been hardwired and the Pump will respond as the Medm directs as soon as the Green start button is pressed on the PS Control panel.  This is only needed in the case of a Power Outage or Reservoir Level trip.

Anyway, attached are the old (boring) and the new (ever so exciting) control medms.  I snapped the EndX screen but EndY looks the same.  In general, the operator needs to do nothing with this system.  If anything takes the pressure down, the bottom line is to make sure the output of the medm or the Setpoint is zero'd before pushing the green start button on the Pump Station Controller.  The output or setpoint is then slowly ramped up(a few psi at a time) by the operator (could/should be automated) to minimize any great shakes of the HEPI/ISI/Optics in chamber.

[Daniel, Stefan, Kiwamu]

 For some reason we frequently failed handing the MC control off to the ALS CARM signal this time around. We checked several things to make sure we are back to the configuration we were in the past Friday. Also we aligned the in-vac picomotors to center the infrared beams on their trans QPDs on TMSY.

Also we took a new noise spectrum without the RGA pump running nearby HAM3, the noise spectrum didn't show a big improvement at around 60-70 Hz in which we thought it would improve.

Besides we did the following things :

• Stefan cleaned up the LSC_CARM filters and rearranged the order of the filters
  • The new filters were then double-checked by a filter archive and switch readbacks.
  • -40 dB gain in LSC_CARM was moved to FM2 of REFL_SERVO_SLOW
• We checked whether if the LSC-REFL_SERVO_COMFILTER was on or off.
  • Confirmed to be off
• We checked the UGFs of the ALS CARM loop both by analog and digital measurements without engaging the AO path in a couple of (slightly) different configuration.
  • The analog and digital measurements agreed with each other in every measurement. This is good.
  • The UGFs was measured in a couple of different settings as we proceed in the handing off script. The UGF could go to 300 Hz with a phase margin of 10 degrees.
  • We confirmed that once the AO path was introduced the phase margin got significantly improved so that it can be more than 30 degrees.
• We suspected that some sort of saturation made the loop instable
  • We monitored the feedback signals applied on the bottom two stages of MC2, but we didn't find an evidence of the saturation
  • When we didn't close the ALS CARM loop the individual loop seems running fine.
• There was a time band at around 21:30 PDT where we had periodical lock losses with a period of about 1-2 minutes.
  • According to the seismic striptool displayed in the control room, there was big disturbance in end Y and end X in a frequency band from 30 mHz to 100 Hz. This might be an earthquake.
  • After a couple of 10 minutes we became able to lock the ALS CARM
• We centered the beams on the infrared trans QPDs
  • There was an apparent improvements in the noise spectrum at around 1 Hz and all the mechanical structure we had had were gone.
  • However several hours later we found that these mechanical structure came back in the spectrum. Since the arm cavity alignment was good, this might be a usual angle to transmitted light coupling.
• Refcav became unable to capture a fringe for some reason
  • We tried tweaking some temperature scanning parameters, but this didn't help
  • After a couple of 10 minutes the refcav came back to normal operation by itself
• We found a sharp peak in the ALS_COMM VCo monitor analog channel at 171 kHz.
  • This is something Daniel unexpectedly found when he was checking out his electronics around the vertex electronics racks
  • The peak moves as we changes the frequency of the ALS_COMM. It is unclear where this signal comes from.
  • This sharp signal shows up only when the ALS beatnote signal presents.
• In the early evening the BS oplevs were found to be off by a lot (~ order of 10 urad)
  • This was found to be due to the BS HEPI. The BS HEPI was off.
  • We tried to turn the HEPI on by an automated command, but this didn't work --- this tends to trips the watchdogs for some reason.
  • We ended up manually turning each servo on --- we tuend on ISO_X, ISO_Y, ISO_RZ. ISO_HP, ISO_Z
  • This brought the IPS signals approximately to where they were in the last week.
  • A fine tuning of the green transmitted light path was done by tweaking the BS top stage bias.

(Dick G, Daniel S)

We added a second frequency difference divider to the rack. For now we just used the output of the DIFF VCO as its input. The division ratio is 10. The attached plot shows the stability of its output as measured by the timing frequency counter. The standard deviation of the distribution is around 2.0 Hz. This is about an order of magnitude better than the single stage measured here. Of course the range has been reduced accordingly and is now roughly +/-100 kHz.

M0 to M0 transfer function measurement in longitudinal (open loop with a gain of 3) on ETMY will start from 2:00 am

J. Kissel

Now that (we think) we trust our model to predict the open loop gain transfer function at all frequencies in the current H1 SUS ETMY damping loop configuration with an EPICs gain of 1.0 (see LHO aLOG 6969), I've made a model of how increasing the gain in L and P by 3.0 compares to a gain of 1.0. The gain of 3.0 configuration is what the ISC commissioners run to reduce the spot motion to what they desire (on a camera alone), and some how it is stable. (And yes, they still run with the ISIs only damped, because there is too much low-frequency motion generated by the isolation loops -- also a camera alone "measured" statement).

The first attachment are the loop figures of merit. Dashed lines are for a gain of 1.0, and solid lines are a gain of 3.0. We see that the Longitudinal loop should have a upper unity gain phase margin at 3.9 [Hz] of 9 [deg]. This causes a factor of 10 gain peaking at 4 [Hz]. In addition, the (originally) 0.43 [Hz] mode is no longer damped, but distorted to 0.35 [Hz]. In Pitch, although the loop is not as unstable as L, there is still a sharp gain peaking feature at 5.1 [Hz] because the gain margin is so small. Similarly, the lowest (originally) 0.56 [Hz] mode is distorted down to 0.43 [Hz].

We have yet to measure the open loop gain of this exact L/P Gain = 3 configuration, but we were able to get some passive measurements:
(1) A comparison of the in loop OSEM sensors, which show similar distortion, and 
(2) Optical lever spectra of the G = 3 configuration alone to quantify what a "good" level of motion for the cavity locking is. 
These are the second and third attachments. Confusingly, the L in-loop sensors seem to follow the model behavior, but we don't see similar modification on the middle frequency resonances in Pitch. Also, the optical lever spectra show none of the expected gain peaking.

J. Kissel, A. Pele

Given that we were suspicious of the one data point that appeared to show model inaccuracies at the most troublesome 0.43 [Hz] mode of H1 SUS ETMY (see LHO aLOG 6933), we took the time to gather a really high (0.001 [Hz]) resolution, open loop gain transfer function. Comparing this against the model reveals the model matches with measurement very very well. No such double peak is visible. 

The only remaining discrepancies are the "known" differences -- a little bit off in frequency for the 0.56 [Hz] first pitch mode, and the ~30 [Hz] zero that's from cross coupling of drive to sensor signal in the readout chain (i.e. not representative of the suspended dynamics).

For posterity, this was measured with the 2013-05-01 filters, with the extra 0.43 and 1 [Hz] boosts installed by Arnaud, and the overall EPICs gain was set to 1.0. We again see that the phase margin is only 19 [deg]. More to come on installation of a better loop design (i.e. installing the one we already have one in our pocket; see LHO aLOG 6760).

We found ETMY moving about a factor of 10 more (mostly pitch), as seen by the OPLEVs, as well as with beam spot motion. In addition the DAC noise level was increases. Similarly the IST GS13s showed at least two sick channels (higher DAC noise, ratty, 10x higher signal at lower frequencies.) After chasing our tail for about 4h, Dave suggested a remote power cycles.

Thus Jim remote power cycled all EY front ends, and we restored them to Friday's burt settings. That fixed the sick channels, and the arm locked right up with no extra motion.

8:46 Pablo D. and Colin S. went into the H2 PSL enclosure to work on PCAL.
9:05 Michael R. went into the H2 PSL enclosure to work on PCAL.
9:27 A truck arrived to pickup pallets.
Hugh R. restarted the end Y HEPI pump controller. (WP 4012)
Corey G. went to end Y and the end X TMS lab.
h1fw1 was taken down to attach a new SSD raid. (see entry 6963) (WP 4011)
10:51 Dick G. went into the LVEA to look at the racks by the H1 PSL and ISCT1 table.
11:12 Thomas V. went into the LVEA to take an inventory of baffles.
11:25 Jim B. went into the LVEA to restart the test stand frontend computers.
Dave B. rebooted the OAF and LSC frontend computers to try to fix IPC errors.
Dale I. brought a school tour through the control room.
Dave B. rebooted the ASC frontend computer to try to fix IPC errors. This unexpectedly created errors on most of the other frontends.
12:34 Dave B. powercycled the RFM switch at end Y to try to fix IPC errors.
1:05 Pablo D. and Colin S. went into the H2 PSL enclosure to work on PCAL.
2:32 Pablo D., Colin S. and Michael R. finished work for the day in the H2 PSL enclosure.
3:40 The commissioners rebooted seiey.
4:58 Jim B. powercycled all the frontend computers at end Y. (see entry 6966)

Jeff B. and Andres R. moved the TMS-X upper structure to end X and installed it on the Bosch frame platform. (WP 4013)
Richard M. worked on the computer for the access card reader.

An air conditioning unit failed in the MSR, and temperatures were a bit on the warm side so the doors to the MSR have been opened and a fan to circulate air has been started.  This is a precaution in case of failure of a second air conditioner which would cause the temperature to rise rapidly.

Power down h1susauxey, h1iscey, h1seiey, and h1susey, then power up computers to see if unexpected noise on ADC channels goes away.

Restarted DAQ at 17:17 PDT for channel changes. 

Added the vim-gtk package to opsws0-opsws7 (omitting opsws6) to provide "gvim" at the request of one of our scientists.  gvim is a nice implementation of vim which works nicely with an X11 connection.  So if you "ssh -X name@workstation", you can use gvim.

Cyrus, Dave, Jim

WP #4011

Attached a new SSD RAID to h1fw1.  This RAID is to be used for storage of raw minute data that was formerly written to the SATAboy hard drive array.  Reconfigured the h1fw1 daqd to write raw minute data to /trend/minute_raw instead of /frames/trend/minute_raw.  Reconfigured h1nds1 to read raw minute data from both /frames/trend/minute_raw and /trend/minute_raw.  The raw minute data in /frames/trend/minute_raw is now in two subdirectories of /frames/trend/minute_raw, and is considered archive data while still being available for viewing via h1nds1.

Note that frame files written to /frames/full, /frames/trend/minute, and /frames/trend/second are not affected by this change.

I created Mathematica versions of the quadopt_wire.m and quadopt_wireloop.m quad parameter sets for the quad main chain metal builds.

20121115TMproductionWire is the old-style metal build of the quad main chain. It corresponds to ^/trunk/QUAD/Common/MatlabTools/QuadModel_Production/quadopt_wire.m r3736 of 11/15/12. The Mathematica version is in the SUS SVN at ^/trunk/Common/MathematicaModels/QuadLite2Lateral/20121115TMproductionWire and the wiki at https://awiki.ligo-wa.caltech.edu/aLIGO/Suspensions/OpsManual/QUAD/Models/20121115TMproductionWire . 

20130523TMproductionWireLoop is the new-style metal build of the quad main chain with prisms on the PUM and the final wire loop under it. It corresponds to ^/trunk/QUAD/Common/MatlabTools/QuadModel_Production/quadopt_wireloop.m r4760 of 5/23/13. The Mathematica version is in the SUS SVN at ^/trunk/Common/MathematicaModels/QuadLite2Lateral/20130523TMproductionWireLoop and the wiki at https://awiki.ligo-wa.caltech.edu/aLIGO/Suspensions/OpsManual/QUAD/Models/20130523TMproductionWireLoop . 

To allow installation of the SSD RAID on h1fw1, the default nds server setting for applications has been changed to h1nds0.  This only affects new shells and processes, so if a process that was connected to h1nds1 fails to collect data as we work on h1nds1, you'll have to wait or restart the process from a new shell.

In-Vac ISC Table

• I bumped the green alignment laser which was unfortunate, because we had one final mirror to align, but Keita was able to get this aligned after a while
• Aligned the final ("topside") mirror on the table (this is the periscope mirror known as "M5")
• Helicoiled the QPD Cables (D1000231)

Telescope

• Continued inspection of 6" Mirrors we have been receiving for installation.  Yesterday we looked at a mirror we received from LLO.

Misc.

• Continue to post photos of progress to Resource Space.
• SUS crew aim to install the Upper Mass Suspension on the EX TMS Lab Bosch Frame this week.
• Working on making inventory/gathering cabling needed for TMS

Starting transfer functions measurements on ETMY from 2:00 am

Here are plots of the long term drift of the arm offset frequency (as measured my H1:ASC-Y_TR_A_NSUM_OUT), and the 3 relevant VCO frequencies. All curves are calibrated in Hz IR.

Between 1600sec and 2000sec we manually moved the ALS-C_COMM_VCO_FREQUENCY to the next fringe and back.
The alignment was tweaked up between 2600sec and 2800sec.

Traces, top:
 - Blue: IR transmitted light, calibrated in Hz using zpk([81.9545+i*81.9529;81.9545-i*81.9529],[850;850],0.656,"n")
 - Red: ALS-C_COMM_VCO
 - Bright green: Green light transmitted (arb. units)
Traces, bottom (note that the scale is 1000x bigger):
 - Green: ALS-Y_VCO
 - Gold: IMC-VCO

Conclusions
- For 3h the IR frequency offset was well predicted by the ALS-C_COMM_VCO - I don't what happened at about 11000sec.
- To epics channel precision the IMC-VCO cancels out ALS-Y_VCO exactly (actually, I'd expect the difference to be equal to ALS-C_COMM_VCO, but they are too noisy to tell)
- The alignment tweak at 1600sec to 2000sec changed the offset by about 50Hz.
- A typical RMS over 10min is about 12Hz, dominated by VCO motion.