front ends crashing due to 208.5 day bug with 2.6.34 kernel

As has been reported in previous alogs, the 2.6.34 kernel has a timer counter overflow bug with makes the machines susceptible to freeze-up if they have been running in excess of 208.5 days. Last week h1build froze, and this weekend h1seiey, h1susauxb123 and h1susauxh34 did the same. The machines marked with an asterix in the list below have an uptime which exceeds 208.5 and could freeze at any time.  We should work on verfifying their SDF settings are up to date and then reboot them at our earliest convenience.

* h1psl0 up 211 days

* h1seih16 up 211 days

* h1seih23 up 211 days

* h1seih45 up 211 days

* h1seib1 up 211 days

* h1seib2 up 211 days

* h1seib3 up 211 days

  h1sush2a up 58 days

  h1sush2b up 201 days

* h1sush34 up 211 days

  h1sush56 up 192 days

* h1susb123 up 211 days

* h1susauxh2 up 211 days

  h1susauxh34 up  7:01

  h1susauxh56 up 191 days

  h1susauxb123 up 7:03

  h1oaf0 up 61 days

  h1lsc0 up 58 days

  h1asc0 up 202 days

* h1pemmx up 209 days

* h1susauxey up 211 days

  h1susey up 38 days

* h1iscey up 211 days

* h1susauxex up 211 days

* h1susex up 211 days

  h1seiex up 95 days

  h1iscex up 163 days

Comparison of HAM5 wall to table top motion for scattered light mitigation

As a quick check, the attached figure shows that the vacuum enclosure walls of LLO HAM5 move between about 10^2 and 10^3 times more than the table top in the 10-100 Hz band. Thus, even if baffles that were mounted on the table top were as good at back-scattering light as the HAM5 doors (unlikely), and they were only 10 cm away from an OFI scattering site, a linear estimate suggests they would cause less noise than reflections off of the more distant door. Of course, for the same geometric attenuation reason, the further the baffles are from the OFI, the better.

PRX and MICH aligned

T Vo, Sheila

Attached is a screenshot for which PRX and the Michelson are roughly aligned.  We still need to aling the RMs to get light on REFL.  

Our next steps towards locking DRMI are:

• to set darkoffsets on the AS WFS, and
• using the swinging  michelson phase the AS WFS so that the michelson signal is in the Q phases.  We can try this for both 9Mz and 45Mz to get an idea of how we should set the phases for the REFL and POP sensors as well. 
• Then we should be able to lock the michelson and fine tune the 45MHz phase
• Attempt to lock PRX to align PRM
• Align onto the LSC POP and REFL diodes
• Attempt to align and lock PRMI

Intermittent crashes while dragging channel names from medm

On Friday afternoon Terry McRae and I crashed several of the ZOTAC workstations in the control room by dragging and dropping channel names from the INMONs of the OMC QPDs into the terminal.  Dave Barker tried this on his workstation and it was fine for him. 

In this situation the workstation completely froze and the solution was to ssh in from another workstation and pkill medm.  

Now I am trying to redo the dark offsets for the AS WFS, and having a similar problem.  When I try to middle click on an medm screen and drag it into the terminal, sometimes it works but it seems like about 50% of the time medm crashes.  The workstation doesn't freeze, although my dataviewer did freeze.  I cannot reopen the sitemap.  Again this is a ZOTAC workstation.

four front end machines offline

I just noticed that four front end machines are offline: h1susauxb123, h1susauxh34, h1seiey, and h1pemmy.

All models on all these machines are dead, and the machiens themselves are inaccessible via ssh.

Entered PSL to set up IO GigE2

I got permission from Mike to continue working on my IO GigE camera 2, and did get light onto it, and it is working, however the beam is too big, so the camera is still blocked, work to continue next week.

Attempt to OMC Scan with IFO beam and SR3 Heater

I was hoping to do an OMC scan today and it's tantalizingly close with the angular loops closed but it seems like the both OMC_DC are saturating even though I've turned off the all the gains and whitening(de-whitening).   The attachment shows the scans going negative at 2 Watts into the IMC.

So I turned down the power into the IMC to about .1 watts and things got better but it was still very unclean scan data so I'm a little confused because Dan Hoak's OMC scan was able to handle about 4 miliamps on OMC_DCPD Sum.

I've misaligned SR2 so that the OMC isn't flashing and returned the DCPD whitening filters to their nominal observe state (1 whitening + 1 dewhitening).  Also I've turned off the OMC ASC and the AS WFS DC loop so that if the IMC loses lock the optics don't go crazy.

I've ran your scans through through my code. The plots are attached below.

The "mismatch" (i.e. ratio of the average of all second order peaks to the average of all zeroth order peaks) is 0.08 +/- 0.02, 0.09 +/- 0.02, 0.083 +/- 0.005 for 0W, 0.5W, and 1W respectively. The uncertainties came from the standard deviation of the height of the peaks. Most of the uncertainty came from fluctuations in the height of the zeroth order peak.

As it stands now I can't really draw any conclusion about the SR3 heater.
It could be that the pzt is scanning too quickly, or the beam/cavity could be fluctuating in time. I can't say for sure.

There's also this weird thing where the resonances on the upward pzt ramp appear in different locations to the ones on the downward ramp, even though they should correspond to the same length changes in the cavity. Could be some hysteresis in the pzt response. For the analysis I only used downward ramp on the pzt. Driving the pzt with a sawtooth waveform should get rid of hysteresis (since it would only ramp one way), but there might be some artifacts from the sudden voltage change.

AS WFS and OMC ASC loops closed

Armed with the knowledge about the OM3 sign flip, I was able to close the angular loops on the AS WFS DC centering as well as the OMC ASC loops at the same time with the IFO beam.  I had to go pretty far with the alignment sliders on OM1 and OM2 to get the IFO beam back on the QPDs but this seems to let the control loops converge and the alignment offsets are closer to their zero position on OM1 and OM2.  However, I was not able to turn on the integrators but this configuration might be good enough to do an OMC scan. 

The squeezer crew could try to walk the OPO squeezer beam from last night towards this new alignment with the ZMs and try scanning from here, maybe it'll be less noisy with the angular loops closed.

One thing that is a little odd is that there seems to be an oscillation in the power of AS_C_SUM and AS_A/B_SUM, however, none the optics' suspensions seem to be moving excessively and all but one ISI was in isolated state.  HAM5 was the only one in "ISI Damped HEPI Offline" but when I tried to go to "Isolated" the HEPI ACT limit watchdog tripped so I left it alone.  This oscillation occurs both when the AS WFS DC centering loops are open and closed so it might be coming from further upstream of HAM6.  Particularly, it seems like AS_A_DC_PIT is the noisiest of the WFS signals but I don't know where there source is coming from.

 

attempt at OMC mode scan with OPO seed beam

Terry, Thomas Vo, Sheila

We set the dark offsets on the OMC QPDs, which now give us reasonable signals with the sewed beam from the OPO on them.  We weren't able to close the OMC ASC loops around the QPDs, although we siwtched the sign on the OM3 actuation as described in 41436.  

I manually aligned OM3 and the OMC suspension with the AS centering loops on to roughly zero the error points of the OMC QPD loops. (screenshot attached).  Then we were able to scan the OMC.  The scanning time was from 1:03 to 1:06:25 UTC April 14th.

We also made a quick attempt to lock the OMC, but it is a very noisy lock.  It might not be possible to lock the OMC on the seed beam with the purge air all the way up like this.  We need to have the door cover partly open to lock the OPO, so we didn't try turning down the purge air.  

 

New EOM installed. IMC beam currently misaligned.

[Peter K, Koji]

Quick report. We went into the PSL table and worked on the EOM replacement. After some recording measurements, the EOM was replaced. This (of course) caused some misalignment of the beam after the EOM. We investiagted how to recover the beam alignment to the IMC, but the beam alignment has not been fully recovered. We saw a spot on the IMC REFL camera and some HOM fringes on the IMC Trans camera.

- Measured the previous modulation depth with an optical spectrum analyzer (OSA).
- Measured the delivered RF power to the each port of the EOM + new cable.
- Measured the previous EOM optical transmissivity.
- Installed an HP RF amp (gain of +16dB) to the new 24.1MHz RF path
- Combined 24.1MHz and 45.5MHz with a power combiner

- Removed the previous EOM from the aluminum mounting block (found two of the four previous screws were not tight at all and two others were not fully fastened!)
- Installed the new EOM on the aluminum mounting block.
- Maximized the EOM transmission with the knobs of the tilt-aligner. We saw good amount of transmission.

- Checked the alignment of the EOM transmission. Found the beam is not going through the first iris after the EOM.
- Touch the HR mirror right before the EOM to have the beam recovers the centering on the first iris -> Saw some beam on the REFL DC and some higher-order mode fringes at the IMC transmission camera.

- We proceeded to make progress on the EOM installation.
- Measured the new modulation depth and adjusted it to have the same modulation depth as before. For the 45MHz modulation, the modulation depth did not reach the previous level with the max power.

I (Koji) will add more quantitative notes later.

IO GigE Camera 1 on the PSL is now measuring the PMC Trans beam (through 2 steering mirrors)

IO GigE camera 1 is now reading the PMC Trans leakage beam through the first steering mirror (in the main beam path), and the second steering mirror (which directs most of the power in the leakage beam to a DC PD).

I measured the power on the camera to be 40uW, though this may be a few uW's more or less, since the PMC Trans power was jumping between 39W and 43W.

The camera gain is set at 50000, Auto Exposure is OFF, the sum is currently at 2.3e+8, the beam X location is 332, and the beam Y location is 235.

Attached are pictures, a 2 hour plot (from before my entrance to the PSL to after), and a 40 minute plot (showing the first good data from IOGigE1).

I didn't want to block the DC PD, so did not verify the beam position on it's steering mirror (mirror #2), but I plan to check this in the future, since the second picture suggests that the transmitted beam may be close to or going through the barrel of the optic.

Setup of IO GigE camera 2 will be at a later date, the fixed image would not launch in the PSL, but later launched in the CR.

OLD/NEW EOM configuration comparison

[Peter, Koji]

Follow up entry for ALOG LHO 41420

Update on Electric Field Meter (EFM) installation

Calum, Craig, Luis, Georgia, Peter, Travis, Betsy, Filiberto, Ed, Richard, Rich

We had a hard time tracking down the source of an open circuit in a single wire(Y-axis, negative leg of the differential signal sent from the electrometer out from the vacuum system) of the electrometer cabling.  We spent much time taking every section of the cabling apart (air and vacuum) only to find that we couldn't see any problem.  We are absolutely certain that this pin was not functional at the start of our hunt, but somewhere in the process it fixed itself.  Very disturbing.

We were able to verify the performance of all aspects of the design, but as we hoped, we now have a list of tweaks that we will apply to the electrometer back at CIT. A timeseries of the differential signals coming out of the EFM revealed the background noise in the chamber to be of order 50mV pk-pk.  We were able to easily see people moving around in the chamber, and were able to see the door cover being flapped around by Georgia (when we asked her to flap it).

1.  We feel the gain in the EFM should be increased by up to a factor of 100.  We will add a bit of gain inside the EFM, then add more gain to the in-air part of the system.  This is to ensure the background field noise is above the ADC input noise (whitening).
2.  We will improve the mechanical mounting of the calibration plates such that it is easier to shift them from one axis to another
3.  We will improve the robustness of the serial communications link used to adjust the common-mode rejection ratio (CMRR) as this proved problematic.
4.  We want to limit the high frequency bandwidth of the EFM.
5.  We want to investigate a 1.7MHz oscillation seen when the field meter was swung causing an electrical saturation.  It appears to be phase reversal on one of the amplifiers.  This went away upon power cycling.  

We verified that the CMRR can be adjusted in situ to better than -60dB.  We also concluded that this adjustment should be done in-chamber as the bench results taken in the optics lab were different than the results obtained in the installed environment.

We verified the 'dark' noise of the EFM to be approximately 200nV/rtHz at 100Hz, which is consistent with design

We measured the background differential noise in X and Y to be approximately 5uV/rtHz at 100Hz as measured at the differential output at the rack (plots to follow).  I believe the calibration of this device to be approximately 64mV per volts per meter at 100s of Hz or so, although this needs to be further investigated (I get this from 16mv/v/m times 2 for differential sensitivity, times 2 for the differential driver gain).  As Rai cautioned us, we need to be careful with the plate spacing distance due to irregular features between the plates (copper buttons), but a simplifying view is given by the capacitances between the calibration plate - sense plate - and ground of the EFM body.  The observed ambient spectrum seemed a bit devoid of features to me and was approximately a factor of 10 lower than the field observed in the optics lab.

Craig, Georgia

Here are some initial spectra and transfer function we took of the Electric Field Meter (EFM) in the X-end chamber.  More thorough analysis/noisebudgeting will be done by Craig and Georgia on Monday.  

Plot 1 is Georgia's plot of the optics lab ambient electric field spectra vs. our in chamber ambient electric field spectrum in cyan.  If this result is to be trusted, we see a significant reduction in electric field noise in chamber vs in the optics lab.
Plot 2 is the EFM positive Y calibration from driven volts on the EFM calibration plates to differential output volts as detected by the sensor plates.  This means that the calibration plate nearest to the positive Y electric field sensor plate was driven by the SR785 using a 1 V oscillation, and the EFM differential Y output response was measured.
Plot 3 is the ambient electric field in the X-end chamber according to the EFM Y direction.
Plot 4 is the Y common mode rejection.  To take this measurement, both calibration plates on the positive and negative Y ends of the EFM were driven with the same 1 V oscillation.  The EFM Y output was then minimized using a digital potentiometer to balance the response.  


More measurements we want on Monday: 
1) Negative Y Calibration TF
2) Grounded Y Spectrum
3) Noise Floor
4) ISI Driven Ambient Electric Field Spectrum 
5) Retake the above measurements