With a small break in commissioning, I thought it would be a good time to try to optimize the ETMX optical lever since we're really close to hitting the noise floor at low frequency (f<2Hz). Previously, we had two levels of 1:10 whitening in analog land and it was good enough for a while until we fully commissioned the ISI.  With TBetter blend settings, the suspension resonances seen in the low frequency regime are getting closer and closer to not being visible at all, which will be a problem if we want to use the optical lever for active damping.

This is the premise for changing the gain settings in order to amplify the suspension mode signals at low frequency using the whitening chassis gains we also wanted to add more laser power to match the other optical levers on site and at LLO:

Old:
2 levels of 1:10 whitening
9,000 counts on the QPD

New:
1 level of 1:10 whitening
1 level of 1.414 gain
1 level of 2.0 gain
15,000 counts on the QPD


We've also had their respective anti-whitening and anti-gain stages in software in:
H1:SUS-ETMX_L3_OPLEV_SEG1_Name2
H1:SUS-ETMX_L3_OPLEV_SEG2_Name2
H1:SUS-ETMX_L3_OPLEV_SEG3_Name2
H1:SUS-ETMX_L3_OPLEV_SEG4_Name2

I've attached the pitch, yaw, and the counts on each individual segments to show if the gain switching is working.  The REFs are the old configuration.  You can see that it really doesn't help the situation at all when it comes to the low frequency stuff which means we might have to be more clever on how to make the optical levers more useful in the future.  In fact, one could argue that Yaw got worse, but it's hard to compare since the data sets differ by about 12 hours in time.  Jeff Kissel suggested looking at the laser intensity noise as well as the electronics chain to see if that could be a limiting factor in our spectra, this means that the adding extra laser power could be the cause of the increase in noise floor.

I have changed the X11 driver on opsws4 and 7 to use the proprietary AMD version; this matches the configuration of the other workstations.  These were originally kept on the open source driver due to problems using RDP with the other driver, at the expense of MEDM running slow - switching to rdesktop instead of remmina for RDP seems to have fixed this.

model restarts logged for Wed 16/Apr/2014
2014_04_16 03:22 h1fw1
2014_04_16 05:30 h1fw1
2014_04_16 06:08 h1fw1
2014_04_16 06:26 h1fw1
2014_04_16 09:16 h1fw1
2014_04_16 11:22 h1fw1
2014_04_16 18:23 h1fw1
2014_04_16 21:18 h1fw1
2014_04_16 23:13 h1fw1
2014_04_16 23:25 h1fw1

all restarts of h1fw1 are unexpected. We might try a reboot of the solaris QFS server h1ldasgateway1 today if this persists.

h1nds1 (the default NDS) reads frames written by h1fw1. This will show data gaps when h1fw1 restarts. If this poses a problem, the user has the option to switch to h1nds0 which has continguous data.

Green beam was almost clipping on PZT1 mirror and 1" fixed steering mirror downstream. This is because the path was set up before the arm was open. Now that we know a good angle of TMS, I changed the table path such that the beam was safely inside the mirror.

Restarted WFS path work but one problem is that the waist of the beam rejected by FI was much smaller than Bram's design (measured [91, 77]um P and Y, should be about 200-250um for the design to work).

The beam size injected into the chamber was [P,Y]=[2.09, 2.72]mm radius,  measured at around the splitter for Hartman (should be about 2.2 or so). The return beam size was matched to injection only in PIT ([P, Y]=[2.0, 1.7]). 

I adjusted the lens position (between PZT1 and PZT2) to split the error. Injeciton [P,Y]=[2.10, 2.66], return =[2.4, 2.1].

This made the waist of the FI-rejected beam larger (about 100um) but not enough. I'll see if I have a space to move things around to make it work.

JeffK, SheilaD, HugoP

We currently have 2 sets of optimized BSC-ISI blend filters that we are trying to choose from: TCrappy and TBetter. 

We produced the plots for these filters as installed on ETMX* in order to asses their strengths and weaknesses. Plots are attached:

• ETMX-TCrappy
• ETMX-TBetter

TBetter filters feature more inertial sensor signal in the blend, at low frequency, on the rotational degrees of freedom.

TBetter blends allows getting more performance ~0.5Hz, and the trade off is more low frequency motion amplification (see upcoming plots from Sheila).

*: TBetter was loaded into TCrappy, and it was unclear which filters loaded in TBetter due to file shadowing in the folder tree. This was fixed on ETMX, and we now load the TBetter blend filters from the common part of the SEI svn:
/ligo/svncommon/SeiSVN/seismic/BSC-ISI/Common/Complementary_Filters_BSC_ISI/aLIGO/some_modified_bend_crappy_filters_040914.mat

This afternoon, after Jeff balanced the coils, I took measurements on the main chain of ETMY for filter design.
Results attached are showing Length/Pitch/Yaw drive to Pitch/Yaw Response for the top mass, UIM and PUM. Y axis is in urad/cts

Templates live in the usual sus path H1/ETMY/SAGL1/Data/2014-04-16_H1SUSETMY_P2PY_WhiteNoise.xml
 

But the xml file still needs updates.

I simplified the IOP software sus watchdog alarms, only alarming on the DACKILL status. I added all missing SUS and SEI systems (the current system was essentially the Y-arm one-arm-test version). I found some IOP safe.snap files had incorrect alarm levels, which I fixed and updated all safe.snap files as some were RCG2.7 versions.

I added guidance text to the alarms.

IOP Watchdog alarms should now be valid and acted on if raised.

08:50 David H. working around TCSY
08:56 Thomas V. working on ETMY optical lever
09:06 Hugh R. working on HAM5 HEPI
09:23 Jodi F. inventorying 3rd IFO baffle parts in LVEA west bay
09:23 Greg G. to crane dewers over X arm (WP 4570)
09:55 Mark B. to LVEA to take pictures of SRM and SR3
10:02 Travis S. to LVEA to take pictures of clean room HEPI interference in beer garden
10:24 Richard M. to end X to work on restoring weather station
12:15 David H. breaking for lunch
12:58 Karen to clean at mid Y
13:28 Jeff B. and Andres to LVEA test stand area to look for parts
13:45 Jeff B. and Andres done
14:01 Karen done
14:49 Thomas V. working on optical levers at end X
15:08 Arnaud P. starting measurements on ETMY
15:44 Thomas V. back from end X
16:27 David H. done work on TCSY

mid X weather station started
tour in the control room
dust monitor at end X resumed communication, possible side affect of work to restore weather station

Praxair drivers had noted this on previous occasions -> I installed a 3/8-1/4 tube adapter to the line and briefly applied 40psi of helium (only because it was "handy") and cleared the line out -> seems to be working fine now

DavidH/Greg

The calibration factors for the TCSX and TCSY paddlewheel flow sensors were adjusted so that their digital displays matched the measured flow rates for the in-line flow meters. Each chiller operated at ~60psi, while a flow rate of ~3.5GPM was displayed.

Initially a K-factor value of 335.000 was used, given the molded Tee fitting that houses the sensor. This resulted in the digital display to read high.
 
The flow sensor settings can be changed by using the menu function on the sensor itself. The following settings are now being used, and are the same both TCSX AND TCSY.

Model 2537 4 to 20mA Output
Flow Unit = Gallons per min (G/m)
K-Factor (PULSES per VOLUME UNIT) = 385.000
Average = 0 seconds(Default factory setting)
Sensitivity = 0 (Default factory setting)
4 Set (Set flow RATE to be represented by 4 mA) = 0.0000
20 Set (Set flow RATE to be represented by 20 mA) = 10.000
Contrast (Adjust visibility of liquid crystal display) = 3

Sample displays for the paddlewheel sensor/inline flow meter after this adjustment to the K-factor are attached.  

While switching blends back to TCrappy

It continues to trip as I am trying to bring it back, basically as soon as I reset the watchdog it trips again.  

this is sheila

JimW and I worked on getting the performance of BSC-EY comparable to the performance of BSC-EX, before handing off to SUS.

BSC-EY now has the following control tools installed:

• Blend filters: Start, T750mHz, TCrappy, TBetter
• Isolation loops: Lv1, Lv3
• Tilt Decoupling: GND-ST1
• Sensor correction: gains need to be tweaked.

We compared the performance of BSC-EY with the performance of BSC-EX, with both platforms under BSC-EX's most used configuration:

• Blend filters: TBetter
• Isolation loops: Lv3
• Tilt Decoupling: GND-ST1
• Sensor correction:  off
   

Attached Plots show comparable performance of the ISIs, at the projected suspension point, for similar ground motion at both end stations. (plots calibrated in nm and nrad).

One can notice a 10Hz peak, on BSC-EY spectra. We went ahead and took performance spectra at EY with the Isolation off. The peak is also there when the isolation and damping loops are off. Hence, the peak seen at 10Hz is neither caused, nor amplified, by the ISI active controls.

Since SUS provides plenty of isolation at 10Hz, we decided to give them the go ahead to start their testing.

Meanwhile, JimW and I kept investigating. We noticed that HEPI L4Cs were seeing the 10Hz peak too, and that it is mostly seen on the vertical L4Cs (see Page 1 of last attachment). We were able to reduce this peak's amplitude by a factor of ~10 by simply turning HEPI position loops off (see Page 2 of last attachment). It looks like HEPI-EY position loops are somehow amplifying the 10Hz peak, but they are not its source, as the peak still appears clearly with HEPI off.

J. Kissel

Thanks to Thomas' incremental improvements to the H1 SUS ETMY optical lever (see LHO aLOGs 11356 & 11387), he got the noise low enough that we were comfortable pushing forward with the large suite of measurements needed to get the QUAD ready for ISC use. the first on this list is the coil balancing on the middle two stages, which has now been completed (following instructions in LHO aLOGs 9453, 9079 and parameters taken from LHO aLOG 10493) with the results below.

The final balanced gains are

H1 SUS ETMY
Channel     Balanced COILOUTF Gain
L1 UL            -0.957
L1 LL            +1.021
L1 UR            +0.976
L1 LR            -1.042

L2 UL            +0.964
L2 LL            -1.039
L2 UR            -0.959
L2 LR            +1.034

The SNR was particular good this time around (thanks to TBetter instead of TCrappy filters on the ISI? [Less wind/ better ground motion]), so I was able to determine these values to 0.25% (instead of the usual 0.5%). The first page of each attachment shows the ASDs of the optical lever signals during a pringle drive at each respective stage, before and after balancing. The second page compares the coherence between blanaced and unbalanced. In summary,

   DOF                  Reduction Factor @ 4.1 [Hz]
L1 Pringle to L3 P           > 9.4               (peak is in the noise, and only ~70% coherent)
L1 Pringle to L3 Y             38 

L2 Pringle to L3 P           > 1.8               (peak is in the noise, totally incoherent)
L2 Pringle to L3 Y             33

Now we begin measuring L P Y of each stage to P&Y of the test mass, to gather all data necessary for 
- iStage L to test mass P&Y for frequency dependent length to angle decoupling, 
- iStage P to test mass Y & iStage Y to test mass P for frequency dependent alignment decoupling, and 
- iStage P to test mass P & iStage Y to test mass Y for alignment plant compensation.

Scott & I got the last Horizontal L4C leveled and then a few more Actuators attached.  Should complete this initial install tomorrow morning.  IAS in the afternoon.  I say initial because commissioning may find some details that need addressing.

Since the Faraday swap on monday, we have more green power. 

On the table we measured 50mW going towards the chamber, where previously we had measured 47mW (6% increase).  On QPD A we had a 10% increase in the sum, on QPD B a 30% increase in the sum.

Alexa and I measured 19uW arriving on ISCT1, compared to 12 uW measured in alog 9191 ( 60% increase from January)

We also noticed that the beam quality was better than it had been.  We measured some profiles, it is a guassian beam, they will be attached to this post later.

When the arm is locked we see a 35% increase in the transmitted power measured by the COMM BBPD, compared to last week. 

We measured the phase noise of the low noise VCO. We used the COMM PLL which deploys a single stage frequency-difference divider (FDD). The PLL was locked to a fixed frequency oscillator (ifr locked to GPS running at 78.95 MHz). The output filters of the PLL were: boost engaged, generic disabled, VCO comp off and low pass off. We added a SR560 with 3 kHz low pass and a gain of 10 to clean up the high frequency part. We then looked at the error point of the common mode board (H1:LSC-REFL_SERVO_ERROR). The calibration is as follows:

25kHz/V (VCO sensitivity) / 25 (boost gain) * 5 (gain divider in VCO output path) / 10 (SR560 gain) / 100 (gain in ERROR readback) * 1V/3200cts = 0.0016 V/cts

For the SSB (single-sided sidebands) we divided by an other sqrt(2). This plot matches nicely to the high frequency SSB curve measured in T0900451. Above 100 Hz we see some excess phase noise which is most likely due to the ifr. The two curves together characterize the low noise VCO all the way from 1 mHz up to 100 kHz. To get the phase noise of the VCO without the FDD we can just multiply the attached curve by 10.

The rms frequency noise of the VCO was previously measured in alogs 6972 and 6865. The numbers were 2 Hz rms for a VCO with FDD and 16 Hz for a VCO without. This is in good agreement with the current measurement.