Title: 10/2 OWL Shift 7:00-15:00 UTC
State of H1: Low noise, finally
Shift Summary: Locking problems continue, packing it in
Activity log:
Ed and Jeff both had problems locking today, so I came in to poke around at a few things.
Sheila called and suggested I try fiddling with a couple things in Guardian, so I did, but I think other things (that I never touched!) are broken, still.
7:41 loaded new gain in SWITCH TO QPDS -41111 to -80000, sat at DRMI for 5 mins then -> ENGAGE ASC
8:14 Third Try
8:30 #4, also fails
9:02 increased Switch to QPDS Tramp, lost lock before I could try it
9:15 Trying longer TR CARM TRAMP
10:00 30 sec Tramp, gain -120000
10:20 Reduce gain to -20000, no good, AS port looked like a split mode
10:35 Packing up, revert all changes, leaving ISC_Lock in down
Filiberto, Richard, Gerardo
Per work pemit 5529.
The Mid-Y chassis was removed and repaired. The popped capacitor (C12) on the AA BNC interface board was replaced. All channels were tested and passed.
Then the gains on channels 1, 2, and 3, were updated by replacing R1 and R4 on the AA BNC interface board. These channels were tested and 10x gain was noted as expected.
Chassis was returned to the rack and powered up after installing a 5A fuse instead of a 3A.
TITLE: Oct 2 EVE Shift 23:00-07:00UTC (04:00-12:00 PDT), all times posted in UTC
STATE Of H1: Aligning/Locking
LOCK DURATION: N/A
SUPPORT: Sheila, Evan
INCOMING OPERATOR: Jim W
Activity log:
00:07 After much ALS alignment tweaking and the wind picking up to around 25mph, Re-locking has been inititated
02:09 Patrick called to let me know that he’ll be in the H2 building for a little while.
02:22 Patrick done in H2 building
04:57 Evan leaving (Patrick still here). There’s nothing more he can do. There seems to be an instability in ALS. Locking doesn’t make it past Switch to QPDs, if it even get’s that far. The majority of the time is spent relocking IMC when it unlocks trying to find IR. Winds are still in the upper 20s.
I placed a call to Sheila to get her thoughts on the situation and Evan suggested that maybe Jim stand down on the OWL shift until the ground motion/wind/ALS is sorted out.
05:10 Called Landry about present condition of IFO and waving off the OWL shift
05:20 Called Sigg about present condition of IFO and waving off the OWL shift. The consensus is Call Jim and let him decide to stand down or come in and confirm proper operation of the Seismic systems, make an aLog and leave.
05:30 Called Jim. He’s going to go ahead and come in NOW and check his stuff and probably NOT stay for the entire OWL shift. His call.
06:10 Jim on site
06:24 Called Landry to give him an update.
Shift Summary: The long and short: No locking for me tonight. Evan, Sheila seemed to be at an impasse and Daniel agreed that this trouble be addressed tomorrow during the day. Wind is still blowing ≥25mph. Ground motion is higher than typical. Not much noise above our raised seismic floor from any earthquakes reported as headed our way. Jim is coming in early to assess the seismic systems and probably won’t stay for the OWL shift.
Mid-Shift Summary: No locking. Sheila and Evan have been working on the problem. All the while the wind has been increasing at times to 40mph. there was a small bump in the EQ graph due to a small quake in Alaska and a larger one near Greenland. Microseism seems to be taking a trend downward in the last couple of hours.
We are having trouble locking this afternoon, there might be several factors contributing. The beamsplitter optical lever is definitely glitching, but this is only a problem while trying to lock DRMI, which has been working OK. The two attached screenshots show this.
We have been having lots of locklosses in the first steps of the CARM offset reduction. Looking at them it seems like the sevo that adjusts the DIFF offset based on AS45Q is running away.
C. Cahillane Between O1 and ER8 there have been some improvements to the O1 Calibration Model. This has significantly reduced our calibration's systematic errors between model and measurement. I have recalculated the uncertainty budget including these fixes. Although I have included a book of plots, I urge you to focus on Plots 1, 2, and 3. Plot 1 is the residuals of the systematic corrections model over the nominal model Recall the response function R = (1 + G) / C. The red line in Plot 1 represents {(1 + G_cor)/C_cor} /{(1 + G_nom)/C_nom} where G_cor and C_cor are the systematic corrections model and G_nom and C_nom are the O1 nominal model. Plot 2 and 3 are the money plots: Magnitude and Phase Uncertainty when we include our systematic corrections. It is currently dominated by uncertainty in the kappas, which remains at a flat 3% and 3 degrees for all frequencies. I believe that after outputting LLO uncertainty, getting accurate kappa uncertainty is our next most pressing issue. Plots 4 and 5 are the squared magnitude and phase uncertainty components. Here you can see that the kappas dominate our systematic corrections model uncertainty. Plots 6, 7, 8, and 9 are the same as 2, 3, 4, and 5, only including the systematic errors as uncertainty in the nominal model. Plots 10, 11, and 12 are the L1, L2 and L3 actuation systematic fits to the weighted mean residuals. Note that the delay in L3 is now 0. Plot 13 is all of the sensing measurements and their associated models placed on a single plot. This plot is very busy but needed to be included because this is where I get my weighted mean residuals for Plot 14. Plot 14 is the sensing systematic fit to the weighted mean residuals. Note that the advance is now 0. What remains to be done: 1) LLO Uncertainty 2) Kappa Uncertainty 3) Proper A_pu Uncertainty (Should take no time whenever I'm feeling organized. I don't expect this to change much) Far Future: 1) Correlated Uncertainty I believe that once we understand our kappa systematics we will have very, very low uncertainties, which is exciting for the GDS pipeline output which corrects for our systematics. In theory our calibration should be limited by detector noise. We are pushing to the limit.
Activity Log: All Times in UTC (PT) 15:00 (08:00) Take over from Jim 15:09 (08:09) Set Intent Bit to Observing 15:58 (08:58) Set Intent Bit to Commissioning – Sheila running measurements 16:00 (09:00) Power cycled Video1 and Video2 – Both were frozen 16:44 (09:44) Lock-Loss – Commissioning 17:51 (10:51) Gerardo – Going to Mid-Y to power off turbo pump and install new AA chassis 17:51 (10:51) Jodi – Going to Mid-Y (with Gerardo) to get some 3IFO parts 18:08 (11:08) Robert – Going into CER to check voltages 18:14 (11:14) Robert – Out of CER 18:35 (11:35) Richard – Going to Mid-Y to work on PEM problems 19:11 (12:11) Jim & Carlos – Going into the Mechanical building to check wireless access points 19:16 (12:16) Jim – Restarting Mid-Y PEM models after hardware repair 19:17 (12:17) Jodi – Back from Mid-Y 19:20 (12:20) Richard & Gerardo – Back from Mid-Y 19:31 (12:31) IFO locked at NOMINAL_LOW_NOISE 22.9W, 74Mpc 19:36 (12:36) Set Intent Bit to Observing 19:50 (12:50) Restart GraceDB 19:59 (12:59) Lockloss – Unknown 22:10 (15:10) Kyle – Going to Mid-Y and near End-Y to get Helium bottles 22:25 (15:25) Kyle – Back from Y-Arm 22:30 (15:30) Start initial alignment 23:00 (16:00) Turn over to Ed End of Shift Summary: Title: 10/02/2015, Day Shift 15:00 – 23:00 (08:00 – 16:00) All times in UTC (PT) Support: Sheila, Evan, Daniel Incoming Operator: Ed Shift Summary: - 15:00 (08:00) IFO locked. Intent Bit set to Commissioning. Wind is moderate, some seismic activity. - 15:09 (08:09) Set Intent Bit to Observing -15:58 (08:58) Set Intent Bit to Commissioning - Sheila running measurements -16:44 (09:44) Lockloss – Commissioning activities -22:30 (15:30) Start initial alignment Rest of shift spent trying to recover IFO. Daniel, Sheila, and Jeff B. went to End-Y. Seeing 8 to 10hz glitches in Y-Arm green, but found no obvious source, (aLOG 22184). Back to CS and will continue trying to relock.
Used double coincident lockloss time this morning to installl new Pcal inverse actuation filter at X-end. I took a transfer function between RxPD and EXC with Inverse actuation filter engaged. The result is attached below. The plot shows a good agreement at lower frequencies (upto few hundred Hz) and is about 2 percent off in magnitude and 7 degress off in phase at a KHz. This is expected because the foton implemetation of filter is little diffrent from its intended response.
We will test it with CBC waveform and detchar safety injection at the next opportunity.
The script used to make this plot is committed to SVN:
/ligo/svncommon/CalSVN/aligocalibration/trunk/Runs/O1/H1/Scripts/INVACT_PCAL/analyze_invactpcal.m
Procedure to perfrom Hardware injection through Pcal:
1. Turn of the pcal excitation.
2. There are two filters in H1:CAL-PCALX_SWEPT_SINE filter bank. Turn on both F1 and F2 if your injection is in units of Strain. Turn on F1 only if it is in units of meter.
3. Perform an injection.
4. Turn off the filters after you are done
5. Turn back the pcal lines on and make sure the SDF monitor for CALEX is all clear.
Looking at the checked in analyze_invactpcal.m, I noticed that you correct not only for the anti-aliasing analog and digital filters in the readback path (which you do need to get to get meters as read by the PCAL), but you also correct the plot for the anti-imaging analog and digital filters. This implies that the actual induced motion of the mass is off by the anti-imaging analog and digital filters as you had to apply this correction in post-processing. So the plot you attached tells us how good of a match we have if we pre-warp the injections by the known anti-imaging filters or take them into account in post-processing when looking for them in search pipelines. In the case where that is not done, then I believe the attached RxPD_over_EXC_no_IOP_compensation.jpg is closer to how the magnitude and phase will be. All I have done to the analyze_invactpcal.m script is removed the IOP upsampling filter , "par.A.antiimaging.digital.response.ssd" in the calculation of RX_over_EXC_corrected variable. The plot is still correcting for the analog anti-imaging (as its mostly delay like) and the zero order hold (again delay), which generally just creates a shift in the start time of the injection.
TITLE: Oct 2 EVE Shift 23:00-07:00UTC (04:00-12:00 PDT), all times posted in UTC
STATE Of H1: Locking
OUTGOING OPERATOR: Jeff B
QUICK SUMMARY:. Trouble locking. ETMY is very glitchy. Wind speeds are ≤15mph. Eq seismic Z axis is between .2-.1µm/s, so that looks ok. X and Y are a little elevated probably because of the wind. Microseism has been on an upward trend for the last 16 hours to about .4µm/s. GraceDB query is in failure at the moment.
Working on relocking the IFO after this mornings lockloss
Chris B., Adam M. Summary: We tested scheduling CBC waveforms with tinj. We had two successful coherent H1L1 hardware injections and one failure. I've attached omega scans of the hardware injections, except for the L1 omegscan for the injection scheduled at 1127789367 (will post this one later, there was an error generating it). Waveforms: All waveforms used are in the hardware injection SVN at: https://daqsvn.ligo-la.caltech.edu/svn/injection/hwinj/Details/Inspiral/${IFO}/coherentbbh*_*_${IFO}.out And the parameter files are: https://daqsvn.ligo-la.caltech.edu/svn/injection/hwinj/Details/Inspiral/coherentbbh*_*.xml Test 1: Originally we scheduled one hardware injection. It was successful: 1127783813 1 1.0 coherentbbh0_1126259455_ The GraceDB events for this injection are L1 and H1. Test 2: We then scheduled another injection. It was successful at L1 and unsuccessful at H1: 1127784860 1 1.0 coherentbbh1_1126259455_ The GraceDB events for this injection are L1 and H1. At the moment GraceDB is set up to treat hardware injections like EM alerts. In order to override the EM alert pausing of tinj and be able to schedule injections closer together on h1hwinj1/l1hwinj1 we did: ezcawrite H1:CAL-INJ_EXTTRIG_ALERT_TIME 112775939 The command worked. And the injection was successful at L1. At H1 the injection did not go in. Checking the MEDM screen for hardware injections we saw that tinj was paused. We were not able to unpause it. We thought that maybe there was an outside process (GraceDB?) that was keeping tinj paused. Therefore we removed the following scheduled injections from the schedule file: 1127785760 1 1.0 coherentbbh2_1126259455_ 1127786667 1 1.0 coherentbbh3_1126259455_ 1127787567 1 1.0 coherentbbh4_1126259455_ 1127788467 1 1.0 coherentbbh5_1126259455_ 1127790274 1 1.0 coherentbbh7_1126259455_ Since none of them would have been successful at H1. Test 3: We scheduled one last injection. It was successful: 1127789367 1 1.0 coherentbbh6_1126259455_ The GraceDB events for this injection are L1 and H1. We checked that tinj was unpaused before this hardware injection and it was.
The reason the second scheduled injection didn't happen at LHO was because the version of ext_alert.py running at LHO is evidently an older version which is still setting H1:CAL-INJ_TINJ_PAUSE (alog 22171). So to allow injections closer together than an hour, it was necessary to manually change the value of H1:CAL-INJ_TINJ_PAUSE as well as H1:CAL-INJ_EXTTRIG_ALERT_TIME. Sorry, I didn't anticipate that when we were making plans for the injections earlier today.
Screenshot attached. We have fast glitches in the ALS Y arm transmission and control signals, which make lock acquisition difficult. In the past they have disapeared after a few hours. see alog 51242 and others referenceing it.
Daniel,Jeff B and I went to the Y end to look into this, and found nothing conclusive. We watched several signals on the scope. We see no signs of any problems on the fiber servo signals, or the PDH error signal, but it is clear that something is wrong looking at the PDH control signal. It is not saturating.
Jeff B droped out of observing mode around 15:38 UTC so that I could make some injections on the ETMX ISI to look for scattered light at end X (WP 5528).
I broke the lock by disabling the CSOFT P input. I have done this many times without breaking the lock before, but this time it clearly broke the lock. Relocking now.
Title: 10/02/2015, Day Shift 15:00 – 23:00 (08:00 – 16:00) All times in UTC (PT) State of H1: At 15:00 (08:00) Locked at NOMINAL_LOW_NOISE, 22.9W, 74Mpc Outgoing Operator: Jim Quick Summary: No wind and no seismic activity. At the shift change, Jim had just finished recovering the IFO from an earlier lock-loss. Intent Bit set to Observing.
Title: 10/1 OWL Shift 7:00-15:00 UTC
State of H1: Low noise, finally
Shift Summary: Lost lock 10:00 UTC, next 5 hours spent trying to reacquire
Activity log:
Locked for first 3 hours
10:00 Lost lock
10:45 Figured out ALS-X wouldn't lock because of VCO issue, reset from VCO screens on ALS overview
12:30 Started initial alignment after several round of DRMI failure
13:30 Started initial alignment over again, after I realized I missed a step
15:00 Finally back to low noise, good luck Jeff
Way back on September 5th, I collected OMC mode scan data before and after the power-up step from 2.2W to 22.5W. The idea was to measure the time-evolution of the sideband and higher-order-mode content at the AS port as the IFO thermalizes and the alignment adjusts to the hi-power state. During the mode scans, I followed Koji’s beacon demodulation technique, and used a DARM excitation to tag the carrier light resonant in the arms. This lets us disentangle the junk carrier light (resonant in the corner) from the good carrier light (resonant in the arms).
There’s quite a bit of information in these mode scans, but the major results are:
- After the power-up step, the amount of 9MHz light at the AS port more than doubles, with a time constant of about 6 minutes. I’m not sure how this informs the studies by Elli & Stefan and Paul regarding the AS36 WFS sensing. Does this time constant agree with thermal effects in the SRC? Or is it from slow alignment loops responding to something like wire heating?
- The contrast defect (ratio of carrier junk light to total available carrier light) is very small, less than 70ppm.
- The mode-matching of the carrier light resonant in the arms into the OMC is excellent, better than 99%.
- Unfortunately, these data don’t completely solve the mysteries of the HAM6 power budget. The 45MHz sidebands saturate the DCPDs at 22W with the preamps in the Hi-Z state, and this makes it impossible to measure the 45MHz sideband power at the AS port using mode scans. But, we can accurately measure the DCPD photocurrent from the carrier and 9MHz sidebands. Carrier = 33.6 +/- 0.4 mA, 9MHz SB = 34.9 +/- 0.3 mA.
Measurement Procedure
Here’s an outline of how the mode scan data were collected:
With the IFO locked on RF-DARM at 2.2W, unlock the OMC, turn off the OMC-LSC_SERVO output, turn off the OMC LSC dither. Turn off all stages of the the DCPD whitening (important).
Check that OMC ASC is on and using the QPDs. Zero the OMC PZT2 offset. Make sure the DARM boost (FM1) is on (important).
Set the DARM offset to 1.2e-5 counts in the LSC-DARM filter bank (this should be about 16pm).
Use AWG to set up an excitation on OMC-PZT2, I used a 70V ramp, 70 second period. Use AWG to set up an excitation on DARM for the beacon scan, I used 1e-8 counts at 201.7 Hz.
I collected ten minutes of data at low power, then engaged the power-up step in the Guardian. After power-up I collected about an hour of data. The GPS times of the data are:
Lo-power start: 1 125 478 482
Lo-power stop: 1 125 478 992
Hi-power start: 1 125 479 058
Hi-power stop: 1 125 482 221
Mode Fitting
For each span of data, the analysis code looks at PZT2_MON_DC and finds times when the PZT drive was slowly increasing. During these periods it grabs the DCPD_SUM data and fits the modes, using the measured transverse mode spacing of the OMC and the known sideband frequencies. I use the measured FSR and f_HOM from Koji’s lab measurements of the H1 OMC:
FSR = 261.72 MHz
f_HOM = 0.21946*FSR
The peaks are fit using the usual Lorentzian function of the PZT voltage. It would be better to do this as a function of optical frequency, but the PZT nonlinearity is small enough that I’ve ignored it. Anyways there's a chicken-and-egg problem, you have to fit the PZT voltage before you can convert voltage to optical frequency.
Problem: 45MHz sideband saturation
At 2.2W, the 45MHz sideband peaks generate about 16mA of photocurrent in DCPD_SUM. In the Hi-Z state, the DCPDs saturate at 20mA (the precise value varies slightly depending on the preamp electronics, these values have been recorded in, for example, the DCPD filter banks). At 22W we expect 160mA in each 45MHz peak, so these saturate the DCPDs.
Weirdly, the 45MHz peaks saturate at a slightly lower value than expected. During the mode scans each of the DCPDs would always flat-top around 27500 counts out of the ADC for each of the 45MHz peaks. See Figure 3. To get around this in the mode fitting, I fit the data before and after the flat-top from the saturation. Unfortunately this doesn’t return the correct peak height: the total power doesn’t agree with what we expect, and it doesn’t agree with the power measured by AS_C. So we still don’t have a complete picture of the HAM6 noise budget.
We could try mode scans with the DCPD preamps in the Lo-Z state, but this only gains us a factor of four in headroom, and the 45MHz peaks would still be on the edge of saturation.
Results: Contrast Defect, Mode Matching, and the Time Evolution of Sideband Power
Using the beacon dither demodulation, we can tag the fraction of the carrier modes which are resonant in the arm. For each PZT sweep, the DCPD data was demodulated at the DARM excitation frequency. A multiplicative factor was applied to match the carrier 00 mode signal in the demodulated signal to the raw DCPD data. From there, we calculate the fraction of each carrier higher-order-mode that is resonant in the arms. The procedure is the same as described by Koji. After some testing I settled on a 10Hz lowpass after the demodulation.
The junk light in the carrier higher order modes is used to calculate the contrast defect: 66.2 +/- 4.5 ppm. The uncertainty is a combination of the statistical uncertainty from mode heights and the variation from sweep to sweep, and systematic uncertainties described in section 5.6 of P1500136.
The fraction of good light in the CR2 (bullseye) mode is used to calculate the mode-matching of the resonant light from the arms into the OMC. Mode-matching: 0.997 +/- 0.001. The alignment into the OMC was not so good during these measurements (a large fraction of the CR1 mode was from the arms), but this was expected since we were using the QPD servo.
The breakdown of DCPD photocurrent from the carrier is:
34.00+/-0.06 mA total carrier light
22.42+/-0.06 mA of light from the arms (note: this is not quite the standard DARM offset)
11.59+/-0.06 mA of junk that's not from the arms
Probably in typical low-noise operations, we have 20mA of carrier light from the arms (fixed by the DC readout loop), and 11.6mA of junk carrier from the corner.
The figures attached are the following:
Figure 1 is a GIF movie showing the evolution of the peak heights following the power up. Note the dramatic increase in lsb3, a higher-order mode of the 9MHz lower sideband.
Figure 2 is a GIF demonstrating the peak fitting procedure.
Figure 3 illustrates the saturation of the DCPDs by the 45MHz sideband peaks. The fit to the peaks (which is necessary for the subtraction of the peak shoulders from the surrounding data) is performed using the data on either side of the flat-top from the saturation. To the eye this looks pretty good, but the peak heights from the fit are way less than what we expect, so there's something bogus going on here.
Figure 4 shows the result of the mode fitting (the same data as Fig. 2).
Figure 5 overlays all of the hi-power mode scans and labels the peaks. Not all of the peaks that are labeled are fit in the analysis.
Figure 6 shows the fit of the peak locations (in PZT voltage) to the expected optical frequency, using a 4th-order polynomial fit of voltage to frequency. This is a sanity check that we correctly labeled the peaks. The error bars are the standard deviation of each peak location, across the few dozen mode scans. This is a crude measure of the statistical variation in the peak fitting.
Figure 7 shows the results of the beacon dither demodulation for one sweep. Black is the raw DCPD data, blue is the demodulated data at the frequency of the DARM excitation, and green is the background demodulation. This is a replica of Koji’s plot from April. The blue trace has been multiplied by a constant so it matches the black trace (raw data) at the CR0 peaks.
Figure 8 shows the fraction of each carrier mode that is tagged by the DARM excitation. The fraction of the 00-mode from the arms is unity, by definition. Except for the 01,10 mode (due to misalignment from the QPD servo), most of the carrier HOMs are due to junk light, i.e. the fraction of each mode from light resonant in the arms is small.
Figure 9 plots various interesting results as a function of time since power-up. This plot is probably the most interesting collection of results. The contrast defect is fairly stable (upper left). Notice how the carrier mode-matching into the OMC improves over time (middle left), and how the 9MHz power increases (lower right). The total photocurrent in the 45MHz sidebands (lower left) is bogus due to the saturated peaks. The time evolution of various measured quantities were fit with exponential curves, the time constants returned by the fits are:
Total photocurrent in 9MHz modes: 370 seconds
AS_C SUM: 400 seconds
Carrier mode-matching (using beacon scan): 830 seconds (note, data are noisy)
Total photocurrent in carrier modes: 320 seconds (note, data are noisy)
Figure 10 demonstrates the change in power in the carrier, 45MHz, and 9MHz modes around the power-up. Except for the 45MHz data (which is wrong because of the saturated peaks), this is a nice before-and-after picture of the power at the AS port. In this plot, I have normalized the total DCPD photocurrent in [carrier, 9MHz, 45MHz] modes by the input power (measured by IMC-PWR_IN).
Finally, Figures 11, 12, and 13 show the change in the individual mode heights over time. There is a large increase in the amount of 9MHz HOMs after the power up. (Since the 9MHz light is not well-matched to the OMC, it couples to higher order modes of the cavity.) The 45MHz LSB5 mode increases, but this is a small peak in a fairly noisy part of the mode scan, and might be sensitive to a nearby 9MHz mode. The 6th-order carrier mode loses a lot of power, this is responsible for most of the reduction in carrier power in Fig. 10.
Analysis Code
I have pushed a version of the mode-fitting code to git.ligo.org. This code can’t run on the control room workstations because of the crummy version of scipy that doesn’t have the peak-finding routines, but there is a script included that will download the data with cdsutils, and you can hack away at it on a laptop from there.
Since the beacon dithering required a high sample rate, across one hour of data, most of this analysis was performed on the LHO cluster. The code and results are saved in this directory.
Can you make something like Figure 12 without normalization?
For one thing I'd like to see the ratio of 0 mode VS higher order modes, and for another it seems to me that the SB imbalance is not small for 9MHz at t=4 and becomes worse as time goes, while 45MHz is just fine.
Here are AS LF, 18 MHz, 90 MHz and 36 MHz length signals during the most recent lock stretch. One can clearly see that the 9 MHz is in trouble.
Due to small finesse, only 00 and 1st order mode for 9MHz are anti-resonant. Especially, LSB 4th order HOM as well as USB 6th are very close to resonance.
"Transmissivity" of SRC against LSB4 and USB6 coming out of BS (which is due to differential mismatch from the ITMs or BS lensing) are about a factor of 7 larger than 00 mode.
In this comment I'll try to answer some questions about the calculation details, and post more data on the mode heights.
Parameters for the Contrast Defect Calculation
The contrast defect is calculated as the ratio of junk carrier light at the AS port to the total available carrier light incident on the beam splitter.
Available carrier light on the beam splitter:
P_carrier = p_in * J9^2 * J45^2 * tIO * g_cr = 673 +/- 40 W
Losses between beamsplitter and DCPDs (including photocurrent --> power calibration):
P_loss = tSRM * tOFI * rOM1 * rOM3 * tOMC * PDresp = 0.241 +/- 0.008 A/W
The uncertainties on the parameters above are guesswork, not motivated by any direct measurements. The dominant source of uncertainty turns out to be the recycling gain.
The total photocurrent in carrier HOMs measured by the DCPDs is about 12mA. Of this, about 0.7mA is tagged as good light from the arm cavities. Most of this is due to the CR1 mode -- this is expected, since the OMC alignment is not optimal on the QPD servo. The CR1 mode is quite small, so nearly all of the carrier HOM content is tagged as 'junk light' not resonant in the arms. This is the measurement used to calculate the contrast defect:
P_junk = 11.3 +/- 0.03 mA
contrast defect = P_junk / (P_carrier * P_loss) = 69 +/- 5 ppm
**Note: in the initial calculation I used a recycling gain of 38+/-2. Now I use 36+/-2, this has changed the result from what was presented in the main entry.
Mode Matching Worst-Case
While the calculation of the contrast defect is somewhat immune to mistakes in the beacon scan measurement (since the amount of carrier HOM content is so small to begin with), the calculation of the carrier mode-matching is highly sensitive to systematics in the beacon scan results. As is shown in Fig 8 above, the fraction of the CR2 mode that is tagged as 'good light' starts around 20%, but decreases as the IFO thermalizes to around 2%. If this is incorrect, we have overestimated the mode-matching into the OMC.
To calculate a worst-case scenario, the photocurrent in the CR2 mode for the last 15 mode scans is 2.6 +/- 0.3 mA. The fraction tagged as good light is 0.025 +/- 0.014. The carrier 00-mode photocurrent is 21.7 +/- 0.4 mA. If all of the CR2 light is from the arms, the mode-matching is 88%.
From the mode scans at low power, we know that a substantial amount of CR2 light can be present at the AS port even when the DARM offset is zero, implying the small CR2 fraction from the arms could be real. (Note: I think the low-power mode scans were taken with different TCS settings, certainly different ETM ring-heater settings.)
Mode Height Plots
In the attached plots, I try to answer Keita's question from above. These plots show the mode heights of the carrier, 9MHz, and 45MHz peaks over time, starting at the end of the power-up step. Some things to note:
I also attach two text files. The first has the median measured mode height, in mA of photocurrent, for all the modes fit within a single FSR. The value and uncertainty for each mode are calculated as the median and std() of the mode heights across the last 15 mode scans in the dataset. The final column is the measured frequency of the mode location, based on the fit of PZT voltage to optical frequency. (Remember, we use upper case LSB and USB for the 45MHz sidebands, lower case lsb and usb for the 9MHz sidebands.)
The second text file lists the carrier modes (zero through eight) and the measured fraction of the mode due to the 'good light' resonant in the arms, calculated from the beacon scan. Again, the uncertainty is calculated from the std() of the final 15 mode scans.