[Cheryl, Stefan, Sheila, Kiwamu, Joe, Paul]

The isolation ratio was calculated using both references for the rejected beam power (see Stefan's aLOG entry 7934), giving results different by 3dB.
Both are above the requirement of 30dB though.

The pick off beam splitter on the PSL table was characterized using power-meter measurements as follows:
Power in = 109.8mW
Power reflected = 38mW
Power transmitted = 67.8mW
R=0.3461
T=0.6175
L=0.0364

The first FI ratio calculation is made using the FI rejected beam power measurement on ISCT1 and the non-rejected beam power measurement on the PSL table.

In this case both measured beams pass forward through the IMC, through the FI, reflect off the PRM, and pass through the FI again before taking different paths (see attached drawing for measurement locations).

From this point, the rejected beam passes a R=90% BS and two 50% beam splitters before being measured on ISCT1. The power of the rejected beam just after the FI is therefore calculated as 6.94mW / 0.1 / 0.5 / 0.5 = 277.6mW. This is actually rather low considering that the power into the IMC should be around 555mW (1000mW*0.9*0.6175). Perhaps there was another 50% beamsplitter unaccounted for in the path somewhere? It's possible that some power is lost in the power control stage on the PSL just before the periscope, but I didn't think it shouldn't be as much as 50%. In this calculation, however, the isolation ratio inferred is unaffected by any loss at the power control stage because that stage is common to both the non-rejected and the rejected beams.

After returning through the FI, the non-rejected beam passes backwards through the IMC. The forward throughput efficiency of the IMC is taken as 94.25%, obtained from observing trends in the IMC REFL PD power (though this does not account for losses inside the IMC). The return efficiency is taken as the forward efficiency multiplied by the average x/y overlap of 94% between the measured mode in HAM1 (see entry 9750 and comments).
The beam then passes back through to the PSL and is reflected off the 34.6% reflective pick off BS. The beam does pass back through the power control stage, however it should have optimal polarization to pass this stage with high-efficiency, so this is not taken into account here.
The non-rejected beam power just after the FI is therefore calculated as 37uW / 0.9425 / 0.94 / 0.3461 = 123uW.

FI isolation ratio = -10*log10( non-rejected power / rejected power) = -10*log10( 123e-6 / 277.6e-3 ) = 33.5 dB

The second calculation is made using the other measurement of the input beam power on the PSL along with the non-rejected beam power measurement on the PSL table.

The non-rejected beam just after the FI is the same as before: 37uW / 0.9425 / 0.94 / 0.3461 = 123uW

From the PSL pick off beam splitter direct reflection point to just after the FI on the return path from the PRM, the rejected beam passes the pick-off BS (picking up a factor T/R = 1.7842 in power from the measurement point), the IMC in forward transmission (94.25% throughput), the FI (97.8%), reflects off the PRM (96.9%), and passes the FI again (97.8%).
The rejected beam power just after the FI is therefore calculated as 339mW / 0.3461 * 0.6175 * 0.9425 * 0.978 * 0.969 * 0.978 = 528mW

Calculating the isolation ratio with these numbers gives us:
FI isolation ratio = -10*log10( non-rejected power / rejected power) = -10*log10( 123e-6 / 528e-3 ) = 36.4 dB

This is quite a significant discrepancy between the two values for the isolation ratio. I am more confident in the 33.5 dB number though, because the measured beam paths have more in common in that case (especially the final power control stage). The 36.4 dB number is susceptible to any other losses in the IMC input beam path, such as the final power control stage before the periscope.

 

H1:PSL-FSS_MIXER_OFS set to zero.

H1:PSL-FSS_MIXER_OFS, which is a mixer offset adjust. was set to -1.634V. That seems high. I couldn't find any alog about setting it. And the ODC bit was set at 0.8V and was alarming. After discussing it with Rick, I set it to zero and verified that the FSS is still locking fine. We should verify (by blocking the light on the diode) that zero is a reasonable setting.

The default matlab version for the DTS linux-x86_64 is now matlab2012b.

J. Kissel, S. Ballmer, R. Mittleman

Executive summary: The seismic's HAM 2&3 front end was unable to drive output signals this weekend -- unclear whether it was related to the CDS maintenance. I have now restored its functionality with a soft-reboot of all front-end processes on the computer. The HAM-ISIs have been restored to level 3 isolation, HEPIs have been left with no drive requested but watchdogs are happy and master switch is on. Details below.


---- Details ----

Stefan informed me that Rich was having trouble getting signals out of the h1seih23 front end (both HAM2 / HAM3 ISIs and HPIs), and after a few unsuccessful trouble shooting attempts gave up. The symptoms were strange -- all levels of watchdogs were untripped (including the IOP model) and master switches were on, but no signals were getting out to the DAC (as reported by, e.g. H1:FEC-53_DAC_OUTPUT_1_7 type channels). The only metric of the failure mode was on the CDS overview screen the "WD" bit in the CDS State Word (the eighth bit [the 128 bit] of H1:FEC-53_STATE_WORD) for the H1IOPSEIH23 was red (and only the IOP, none of the user models' WD bits were red), and the GDS_TP screen for the IOP model showed the timing (H1:FEC-53_TIME_ERR) was red and claimed NO SYNC. 

After trying a few iterations of soft-rebooting things without success, the gave up. They knew that there was a proper song-and-dance, correct order, to soft-booting but they admitted to not remembering what it was and not being able to find any documentation describing it.

I came in this morning, and did the proper procedure, and all is now functioning properly.

---- The proper front end soft-reboot procedure ----
I'm sure it's documented some where, but being able to find it... 

[in words]
(1) log into the front end,
(2) kill all the user models (order doesn't matter),
(3) restart the IOP model, and
(4) restart the user models (order doesn't matter).

[what the terminal looks like]
controls@opsws3:~ 0$ ssh h1seih23  # Step (1)
controls@h1seih23 ~ 0$ killh1isiham2   # Step (2)
controls@h1seih23 ~ 0$ killh1isiham3   #       |
controls@h1seih23 ~ 0$ killh1hpiham3   #       |
controls@h1seih23 ~ 0$ killh1hpiham2   #       v
controls@h1seih23 ~ 0$ starth1iopseih23   # Step (3)
h1iopseih23epics H1 IOC Server started
Burt restored /opt/rtcds/lho/h1/target/h1iopseih23/h1iopseih23epics/burt/safe.snap
Old : H1:FEC-53_BURT_RESTORE         1
New : H1:FEC-53_BURT_RESTORE         1
 * WARNING:  awgtpman_iop has already been started.
controls@h1seih23 ~ 0$ starth1hpiham2   # Start of Step (4)
h1hpiham2epics: no process found
h1hpiham2epics H1 IOC Server started
Burt restored /opt/rtcds/lho/h1/target/h1hpiham2/h1hpiham2epics/burt/safe.snap
Old : H1:FEC-54_BURT_RESTORE         1
New : H1:FEC-54_BURT_RESTORE         1
 * Starting h1hpiham2 awgtpman ...                                        [ !! ]
controls@h1seih23 ~ 0$ starth1hpiham3
h1hpiham3epics: no process found
h1hpiham3epics H1 IOC Server started
Burt restored /opt/rtcds/lho/h1/target/h1hpiham3/h1hpiham3epics/burt/safe.snap
Old : H1:FEC-55_BURT_RESTORE         1
New : H1:FEC-55_BURT_RESTORE         1
 * Starting h1hpiham3 awgtpman ...                                        [ !! ]
controls@h1seih23 ~ 0$ starth1isiham2
h1isiham2epics: no process found
h1isiham2epics H1 IOC Server started
Burt restored /opt/rtcds/lho/h1/target/h1isiham2/h1isiham2epics/burt/safe.snap
Old : H1:FEC-56_BURT_RESTORE         1
New : H1:FEC-56_BURT_RESTORE         1
 * Starting h1isiham2 awgtpman ...                                        [ !! ]
controls@h1seih23 ~ 0$ starth1isiham3
h1isiham3epics: no process found
h1isiham3epics H1 IOC Server started
Burt restored /opt/rtcds/lho/h1/target/h1isiham3/h1isiham3epics/burt/safe.snap
Old : H1:FEC-57_BURT_RESTORE         1
New : H1:FEC-57_BURT_RESTORE         1
 * Starting h1isiham3 awgtpman ...                                        [ !! ]
controls@h1seih23 ~ 0$    # End of Step (4)


After the turning on the front end processes, I
- was delighted to see the safe.snaps were reasonably up to date 
- Turned on the HAM2-ISI and HAM3-ISIs damping loops and restored to Level 3 isolation 
- left HAM2 and HAM3 HEPIs as untripped and master switches on, but no requested drive since I figured Rich was just going to immediately play with it once he got in. 

The final measurement obtained from the sideband sweep data taken last Thursday was of the IMC g-factor (or alternatively the MC2 radius of curvature).

For this measurement, we wanted to measure the resonant frequencies of higher-order modes in the IMC. To do this we need to observe beats between higher-order mode sidebands transmitted through the cavity with the HG00 mode carrier also transmitted through the cavity.

Due to the spatial orthogonality of higher-order modes, a single element photodiode would would usually be insensitive to this signal (hence using wavefront sensors to detect misalignment). In the absence of a broadband wavefront sensor in transmission of the IMC, we can instead break the spatial symmetry at the REFL AIR photodiode using any physical obstruction to recover the signal. In this case, we used an iris offset with respect to the beam centre to partially block the beam. In addition to this, the IMC input beam was misaligned by disengaging the WFS loops and applying offsets to the periscope PZT in order to better ring up higher-order modes.

The first attached plot shows a sweep of one whole FSR around the 45.5MHz FSR peak. Overlaid on the measured data is the output from a Finesse model of the IMC, with the IMC geometry as designed expect for the MC2 radius of curvature which was taken as 27.275m from the nebula optics page. The Finesse model allows us to quickly identify each peak in the measured transfer function.

The higher-order mode content in the model was hand-tuned to match the data, as the effects of the iris are effectively un-quantifiable.  Since for this measurement we are only interested in the higher-order mode resonance freuqencies, correct y-axis values scaling is not necessary.

Higher resolution scans were then made of 6 individual higher-order mode peaks. The second attached plot shows the results of fitting a Lorentzian function to these higher-order mode peaks. The cavity g-factors corresponding to the fitted higher-order mode difference frequency (using the mean FSR reported in 8087) were then calculated. From there, the MC2 Rc could be calculated.

The results are shown in the following table:

The "Rc_eff" column shows the effective MC2 radius of curvature experienced in either the tangential or sagittal planes. The IMC beam probes MC2 under a small angle (0.82^o - see T0900386), and as a result the Rc experienced in the tangential (xz) plane is shorter than the normal incidence Rc by a factor of cos(0.82^o), and the sagittal plane (yz) Rc is longer than the normal incidence Rc by a factor of cos(0.82^o). The difference frequency between the HG00 mode and the HG10 mode is determined by the effective Rc in the tangential plane, and the difference frequency between the HG00 mode and the HG01 mode is determined by the effective Rc in the sagittal plane. The Rc_eff, g-factor and HOM diff. freq. entries for the Nebula/design rows were calculated from the quoted Rc of 27275mm and the incident angle 0.82^o.

The Rc column shows the equivalent normal incidence Rc values. For the fitted peaks these values were calculated from the Rc_eff values and the incident angle 0.82^o. For the Nebula/design values these are just the quoted Rc from the vendor measurement.

The mean value of MC2 Rc from all 6 peak measurements is 27277mm, with standard deviation 11mm*. This can be compared with the polisher's reported value of 27275mm (from the nebula page), the manufacturing requirement of 27240±140mm (see E070079-A-D), and the manufacturing goal of 27240±30mm (see also E070079-A-D). The value of Rc from this measurement agreed with the vendor value within the error bars, and is well within the manufacturing requirements (although it is slightly outside the manufacturing goal).

* The fits for HG11 and HG20/02 are a little suspect due to the influence of other nearby resonances. In both cases this influence can be expected to lower the measured peak frequency. For the HG11 mode this has the effect of increasing the HOM diff freq. and therefore lowering the calculated Rc. For the HG20/02 mode this has the effect of lowering the HOM diff freq. and therefore increasing the calculated Rc. The mean value for the Rc when these two measurements are discounted is 27278mm, and the standard deviation is 10.5mm. It appears therefore that the nearby resonances have a roughly equal and opposite effect on the HG11 and HG20/02 Rc values, and thus don't strongly affect the final result.

I put together an ODC site overview screen that contains the bits of all ODC channels implemented so far.
It also contains links to per-chamber summary screens, which in turn have links to the ODC screens with bit labels.

For now I added a link to the sitemap ASC menu.

More stuff will be added as it becomes available.

The screens all are in sys/common/medm, and should work for both sites. They are in svn as version 5850.

The pictures below show the site overview, and example chamber overviews for end-x and ham2.

The scheduled mainenance for the LHO CDS fileserver is complete.  The workstations, web services, and remote login are once again available for use.  Workstations in the LVEA were remotely shut down on Friday; these can be powered back on at any time they are needed.  The control room workstations, opsws0-7 are powered off to conserve electricity but may also be powered on if needed over the weekend.

The RAID card has been replaced in cdsfs0.  The new controller is now in the process of verifying the blocks on disk, which is a lengthy process and will take in the neighborhood of ~20 hours.  I will be back on site tomorrow early afternoon to run some additional checks on the filesystem before booting the CDS workstations and starting the RAID cache battery test.  There should be no issues with any of the CDS web pages at this point, it looks like everything is running normally there.  The external login server will be turned back on at the same time the workstations are ready.

RichM, HughR
Check for valid HEPI signals at BSC3 showed problems.  Found Cable going from HEPI AA chassis going to incorrect card on I/O comp.  HEPI I/O adc had an ISI cable attached.  We made BSC3 look like BSC 1/2 to correct; this matched the drawings.  The cable running from HEPI AA chassis was unlabeled.  The labeled cable was unattached at both ends.  I suspect the EEguys know something about this--maybe the labeled cable was deemed bad.  It should be removed is so and the cabling being used should be labeled.

The main fileserver for CDS workstations will be going down for maintenance today at 5PM local time, as specified in WP4188. ALL CDS workstations and remote logins will be unavailable during this work, which is anticipated to finish at 5PM PDT Saturday. Some web services will have interruptions in service during this period as well, such as the MEDM screenshots.  The front end systems and DAQ will continue to operate normally.  A follow up logbook entry will be posted when the work is completed.

Please save your work and log off soon if you are logged in remotely, or are using a local CDS workstation.

LVEA Laser safe
Alarms: Dust in Vac Prep Lab, Diode Room, and CDS frontends

Apollo – Installing walking plates at BSC3
Doug – Shooting alignment of the Elliptical Baffle in BSC2
Richard – Pulling cables at End-Y

09:20 Gerardo working on BOSEMs and Cable Brackets in SUS Cleanroom in LVEA
09:30 Rick & Peter working in the H2-PSL enclosure
10:37 Alexa – To End-X working in VEA
13:10 Patrick working on Dust Monitor #15 calibration failures 
    
  Getting continual low level (100 to 150 counts) dust alarms during the day from the Diode Room. Rick S. and I looked in the chiller anti-room, the diode room, and the A/C system, but could not find an apparent source for the alarms. Further investigation is needed to find the source of these alarms. 

As for the HSTSs, I took a spectra of PR3 from Monday. The isolation and damping loops of the ISI were closed (isolation level 3), as well as the damping loops of the suspension. The chamber was still under vacuum. Status of HEPI needs to be double checked.

The attached plot compares the spectra of LHO and LLO PR3, in a similar configuration (ISI isolated, SUS damped), for the 3 levels (M1 M2 and M3) and every degree of freedom.

Looking at the plots, we can see that LLO get more isolation at each level, even though we are using the same damping filters. That said, Jeff discovered that the suspension medm gains are set to different values, (higher values at LLO) which would explain why the resonnances of the suspension are more damped there. Other than this, spectra is very close to sensor noise. The larger signal between 0.1Hz and 1Hz is highly due to ground motion (I quickly took a spectra of the streckeisen sitting next to HAM2 chamber,  which shows the same shape around 0.1Hz) More details to come next week.

We installed C P Stat on the X arm tube opening where the spool was removed yesterday. We also lowered the clean room over BSC 3 to prepare for cartdrige install early Monday. I did a recheck of the upper limit on the main crane and found that the block was traveling up just a little to close to the drum so I adjusted the anti two block limit and it is ready to go.

As I mentioned in my last aLOG entry [8087], when performing the IMC sideband sweep measurements, the broad peaks when the sideband frequency is around an integer multiple of the IMC FSR are only expected to appear in the presence of a length/frequency lock offset. As such, we can try to compensate any offset in the length error signal by applying a digital offset in the IMC-L feedback path until these peaks are minimized*.

I recorded sweeps across the 45.5MHz FSR peak while adjusting the offset in the IMC-L path. The results are shown in the attached plot. As the offset is increased towards 20k counts, the peak reduces in magnitude.

Beyond 20k counts, the IMC would not stay locked long enough to make the measurement. I suspect this is caused by one of the outputs railing due to the additional offset, but I wanted to move on to higher-order mode peak measurements so I didn't investigate further. Maybe there is a better location to add the offset, for example on the analog common mode servo board itself. Apart from the lock losses above 20k counts, I didn't notice any drop in the IMC transmitted power with added offsets. This eliminates the possibility that the offset was driving the cavity further from carrier resonance and thus causing a large drop in the carrier light power, in turn causing the peak to drop in magnitude.

On a side note, since a bigger locking offset makes the FSR peaks easier to observe, it might be worth deliberately making this offset worse (i.e. adding -20k counts) for future measurements of the IMC FSR/length. This could help improve the SNR around e.g. the 18.2MHz, 27.7MHz and 54.5MHz peaks to make those measurements useful.

*At low frequencies the IMC-L path dominates over the IMC-F path, so applying an offset in the IMC-L path should suffice and eliminate the need to directly cancel any offset at the analog servo board input.

  Ops has been getting continual low level (100 to 150 counts) dust alarms all day from the Diode Room. Rick S. and I looked in the chiller anti-room and the diode room, but could not find an apparent source of the particles. Further investigation is needed to understand the source of these alarms.  

[Kiwamu, Stefan, Joe, Paul]

Last Thursday, measurements of several IMC FSR resonances were taken using the sideband sweep method [see e.g. LLO alog entry 4849].

In short, this method involves phase modulating the light into the IMC with a swept sine signal from a network analyzer and observing the signal in transmission of the IMC with an RF photodiode to observe various optical resonances within the cavity.

In this case, the RF photodiode was the broadband REFL AIR photodiode on ISCT1. The network analyzer output was split with a power splitter, with one output going to the REF input, and the other fed to a mini-circuits RF amp with a gain of 18.8dB. The network analyzer source Voltage was tuned to give around 10Vpkpk output from the RF amplifier. The RF amplifier output was then passed to the 45.5MHz input of the EOM on the PSL table. The RF output of the REFL AIR photodiode was connected to the A input of the network analyzer.

When the sideband frequency gets close a multiple of the FSR, a broad peak in the the N.A -->REFL AIR PD transfer function is observed, as reported in [1] and subsequent errata [2]. According to these references as well as Araya et al. [3] this broad peak is only present if there is an offset in the length/frequency lock of the cavity (more on that in a forthcoming post). There is a narrow dip in this peak however, when the sideband frequency approaches the exact FSR multiple frequency. The frequency at which this dip hits a minimum is an integer multiple of the cavity FSR.

The first attached plot shows 5 FSR peaks measured in this way. Since we apply the signal to the 45.5MHz input of the EOM, the modulation depth, and thus the SNR of the measurement, is low at frequencies far from 45.5MHz (e.g. the 18.2MHz FSR, the 27.3MHz FSR and the 54.5MHz FSR). The 36.4MHz peak and the 45.5MHz peak measurements came out the best, so I fitted the central dip using a simple Lorentzian function to estimate the frequency of the minima (see second attached plot).

The results from these two fits are summarised and compared to the design values in the following table:

The calculated lengths from each peak fit agree to within 100um, and are within 0.5mm of the design length.

If higher precision on this measurement is required later on, we could try syncing the network analyzer with a rubidium standard, and also try implementing the measurement technique described in [3]. Also, separately using the 9.1MHz input of the EOM to measure a greater sample of FSRs could be helpful.

[1] K. Skeldon and K. Strain, Applied Optics Vol 36, Number 27 (1997): http://www.opticsinfobase.org/ao/abstract.cfm?uri=ao-36-27-6802

[2] K. Skeldon and K. Strain, Applied Optics Vol 37, Number 21 (1998): http://www.opticsinfobase.org/ao/abstract.cfm?uri=ao-37-21-4936

[3] A. Araya et al, Applied Optics Vol 38, Number 13 (1999): http://www.opticsinfobase.org/ao/abstract.cfm?uri=ao-38-13-2848

An NPRO was setup for aligning and testing the outer loop power stabilisation photodetector array in the H2 PSL Laser Area Enclosure.  The power incident on the array was measured to be 250-260 mW, the beam diameter close to the location of the 8 individual photodiodes was about 800 microns.

Attached are plots of dust counts requested from 4 PM October 9 to 4 PM October 10.

Jason and I set 2new monuments and a new elevation target in preparations for tomorrow baffle alignment. I will setup the total station in the AM and be ready in the first hr.