First PRMI noise budget from the NB tool is attached.
Optical gain estimated from the openloop transfer function measurements are 5.5e3 W/m for MICH loop, 8.6e4 W/m for PRCL loop.
Power recycling gain estimated from this optical gains is ~ 3. This estimation was done by comparing the optical gain in simple Michelson and PRMI MICH.
Note that this estimation assumes perfect diagonalization of MICH and PRCL loop, but actually they are not diagonalized yet.

[Method]
1. Lock PRMI on sideband using REFLAIR_A_RF45_I_ERR and Q_ERR (see alog #10427).

2. Take OLTF of MICH loop and PRCL loop.

3. Keep it locked for a while to take feedback signal data (H1:LSC-MICH_OUT_DQ and H1:LSC-PRCL_OUT_DQ) for the noise budget. Data I used started from Feb 28 2014 18:25:00 UTC (local Friday morning).

4. Use the NB simulink model to plot OLTFs. Change optical gain for MICH loop and PRCL loop to match those with the measured OLTFs. Here I assumed that the error signal from BS motion only appear in REFLAIR45_Q, and PR2/PRM motion only appear in REFLAIR45_I. See attached Simulink diagram I used. Optickle block is not used since we want to make the measurement based NB model. The model lives in /ligo/svncommon/NbSVN/aligonoisebudget/trunk/PRMI/H1.

5. Use the same model to plot noise budgets for MICH loop and PRCL loop. Sensitivity curve is estimated from the feedback signal data.


[Result]
1. OLTF_MICH_1077647116.png: OLTF of MICH loop from the measurement and the model. UGF is ~6 Hz and phase margin is ~40 deg. For the model curve, optical gain of 5.5e3 W/m was used. The measurement and the model agrees pretty well.

2. OLTF_PRCL_1077647116.png: OLTF of PRCL loop from the measurement and the model. UGF is ~60 Hz and phase margin is ~30 deg. For the model curve, optical gain of -8.6e4 W/m was used. The measurement and the model agrees OK except for the dip at ~13 Hz, which comes from closs coupling with MICH loop. This is because we only use BS for MICH loop, but BS also changes PRCL.

3. NB_MICH_1077647116.png, NB_PRCL_1077647116.png: Noise budget for MICH and PRCL loop. Note that seismic noise is not real (copied from LLO model). Sensor noise is currently not contributing very much. MICH motion is larger than PRCL motion (because of BS motion?).


[Discussion on PRMI optical gain]
1. Measured optical gain (from BS motion * sqrt(2) to REFLAIR45_Q) for simple Michelson was 1.9 W/m, including the cable loss (see alog #10213). The things changed in MICH loop for simple Michelson and MICH in PRMI are;

                       MI        PRMI MICH
PD whitening gain      45 dB     0 dB
H1:LSC-MICH_GAIN       900       5    
output matrix for BS   0.05      1
measured optical gain  1.9 W/m   5500 W/m

This gives overall gain ratio of MI/(PRMI MICH) = 1.3. Since the UGFs for simple Michelson and PRMI MICH loop was 8 Hz (see alog #10127) and 6 Hz respectively, this ratio is consistent.

2. Optical gain ratio between simple Michelson without PRM and PRMI should be approximately equal to the power recycling gain (PRG). However, since we did optical gain measurement for simple Michelson with PRM misaligned, measured optical gain will be smaller by factor of T_PRM^2. Thus, the ratio between simple Michelson with PRM misaligned and PRMI will be approximately Gp/T_PRM^2. So, estimated power recycling gain is;

Gp ~ 5.5e3/1.9 * 0.03^2 = 2.6

Designed PRG for PRMI is 58 according to LIGO-T1300954. Considering the closs-coupling between PRCL loop and MICH loop, this estimation seems to be an upper limit to the actual PRG (see below).

3. According to Optickle simulation in LIGO-T1300328, sensing matrix for PRMI sideband is

            PRCL    MICH
REFL 45I    3.4e6   2.5e3
REFL 45Q    6.4e4   1.3e5  W/m

So, simulated ratio between diagonal elements is PRCL/MICH = 3.4e6/1.3e5 = 26. Our optical gain estimation gives 8.6e4/5.5e3 = 16.
Considering the fact that we are ignoring the off-diagonal elements in the optical gain estimation, I think this is reasonable. For example, BS to REFL 45Q could be 1.3e5+6.4e4 W/m and 3.4e6/(1.3e5+6.4e4) = 18.


[Next]
- Measure PRMI sensing matrix, compare with the simulation, and use it in the NB model
- Update the simulink model so that it can handle off-diagonal elements
- Output matrix diagonalization for PRMI
- Include online seismic noise, frequency noise and intensity noise in the NB model

Today was another good day of arm locking. 

The comm handoff was still reliable this morning.  I had a look at the open loop gain, and decided to try using higher gain. I was able to push the loop gain up to 35kHz, where we have 40 degrees phase margin.  I turned the PLL gain down to 27dB to move the gain peaking to lower frequencies in order to make the higher ugf stable.  With a CM board gain of 18, I was able to engage the first boost in addition to the common compensation.  Data and plots of the new open loop gain are attached. (59 is a GIF, 58 is phase and 57 is magnitude)

Qualitatively, this makes the IR resonance in the arm more stable.  I've attached a stiptool showing the transmitted IR at the settings we have been using up to now, 9dB, and in the laster part with 18dB and the first boost on.  Maybe there is hope after all for ALS COMM. 

I also lowered the UGF of the IMC VCO frequency servo to 0.7 Hz, since it oscillates sometimes. 

We need to start thinking about how to align the IR to the arm cavity.  Today I tweaked PR2 and IM4 by hand.  The red team is already doing a dither alignment on PR2, it would be nice if we could add the IR transmitted signal to this so we could use it to align to the arm. 

I also had made a twinCAT library a while ago that was intended to automate finding the IR resonance in the arm cavity.  I added an medm screen to the ALS overview screen, but this still needs work.

The green team had some sucess today.

We saw that there is a large wandering peak (similar to what we have seen in the IMC alog 10289 alog 10327) in COMM PFD Imon.  This PFD chassis is directly under the diff RF doubler, by disconnecting the 79MHz input from this doubler we saw the peak become much smaller.  We are leaving this unplugged since we don't need it yet. (Alexa has some movies of the two wandering peaks)

The Xarm guardian is working, at least the states up to LOCKED_W_SLOW.  We didn't test the dither alingment states since the dither alignment isn't really working anyway.

The COMM PLL slow feedback to the IMC VCO had low bandwidth, which was limited by the randomization in the VCO servo library.  Daniel edited the low noise VCO library, the servo now has two modes, internal where the frequency comparator is used for an error signal, and external, meant for use with the COMM PLL.  The bandwidth is now limited to about 1 Hz, by the VCO response.  Right now we have the gain setting for the COMM PLL slow feedback set to 14000Hz/V, and the ugf of the IMC VCO servo set to 0.8Hz.  This seems stable.  We also turn this off after the COMM handoff now. 

We found a few gremlins in the COMM handoff-  with the change to LSC input matrix our script was no longer setting the matrix element.  For some reason the fast path was disabled in the REFL servo.  Both of these things could be solves with state control code or for the time being by improving the script/writting a guardian. Now we are reliably doing the COMM handoff, which is stable for ~20 minutes to a half an hour right now.  I've just committed the handoff and down scripts to the svn. 

We rewired the REFL_DC_BIAS, so now it is routed into the CM board input 2

Once we had this locked we started looking at the IR transmission.  When the alignment is good, we stay within about 160Hz (the transmitted power wanders over the peak back and forth, but doesn't move as far as we have seen some nights.  We have seen that the amount of noise we see depends on the alingment. 

We found the IR resonance, moved the ETM in YAW, and found the resonance again.  We made three measurements like this and got 400Hz/urad.  When the alingment was worse this measurement got harder.

It seems clear we need a good alingment to have a chance of locking COMM on resonance.

Right now COMM has been locked for half an hour.

After the MC2 trip, the MC2 guardian turned off the outputs from the lock fitlers for all three stages.  The IMC guardian did not reset these (except for M1 and M2 lock) , so the IMC could not lock.   I added a few lines in the acquire state of the IMC guardian that turn these back on. 

We had another ocurence of TwinCAT near the time of a trip of HAM2+3.  This is the first screen shot attached.  (MC2 and PR2 were also tripped)

Then as the guardian was bringing back HAM3 it tripped again.

This report is a summary of some of the issues that came up during the recent ISI_HAMX guardian commissioning (alog 10394, alog 10300):

There is quite a bit of overshoot in the X/Y location when the platform are ramped to RX/RY setpoints with offsets.  The image below shows the CPS_<dof>_LOCATION readback while the setpoint ramps with 10 urad offsets in RX and RY:

Note that X and Y locations swing through more than 100,000 counts during and after the ramp.  Fabrice's suggestion is therefore that we break up the isolation procedure such that we engage the isolation loops and ramp the biases for the RX and RY degrees of freedom first, before engaging the isolation loops for the other degrees of freedom.  This will help prevent the isolation loops from saturating during the isolation process, even if there are large cart bias offsets.

During deisolation we should also be ramping down any biases before we begin the deisolation procedure.

I have written two cron jobs which run on script0 as user controls. They log any front end restarts and produce daily reports of such restarts.

The logger runs every minute, the daily logger runs at 5 minutes past midnight.

Details are available in the wiki page:

https://lhocds.ligo-wa.caltech.edu/wiki/FrontEndModelRestartLogging

the code should be transparent to the front end model user except for the reboot.log file being removed within a minute of the restart completing.

The goal is to have the daily logger make a robo-log entry in the ALOG.

While TMS and SUS were removing extra (transport Payload) from the system, Jim & I confirmed corner connections on the SEI Sensors and wired up the remaining parts of position sensors.  With the payload now very close to final, we'll do another level check maybe (Dial Indicators show ~0.2mm tilt change) and adjust that if needed.  We'll then unlock and balance followed by locker/CPS Target adjustments in this final configuration.  Maybe by days end Monday we'll be ready for SUS to unlock but plan on Tuesday earlist. Then TFs can ensue.

(Sheila, Daniel)

This afternoon the noise eater of the ALS laser in H1 EX was oscillating. After some TwinCAT work, we were able to toggle the noise eater off and on remotely. This worked as expected and cleared the noise eater oscillation.

PS. There are 2 relay contacts wired up. Only the first one seems to be needed. I disconnected the second.

With the RF doubler for the 79.2 MHz source moved on top of the ALS distribution chassis, we took another look at the modecleaner error signal. The intermodulation product with the PSL VCO is still visible but small. It is about an order of magnitude smaller than before. This seems good enough for now, but could potential effect the noise of the full interferometer.

- Craig C. and Rick S. going to H2 PSL Enclosure for PCAL work. 8:50 AM PT

- Jeff B to EY to install dust monitor. 9:25 AM PT

- Aaron S to EY to do cabling check-off. 9:30 AM PT

- Hugh to EY to work on SEI. 9:43 AM PT

- Instrument air alarm, Kyle fixed this.

- Corey at EY to take off payload. 9:45 AM PT

- Filaberto running cables from beer garden to X and Y manifolds.

- Jim B, Cyrus R, Dan M trouble shooting instability with frame writer 1, won't bother anyone else.

- /ligo server froze, crashing all of the work stations. ~11:35 AM PT

- Mitchell in LVEA 2:14 PM PT

- Gerardo in LVEA and EY throughout the day.

- Arnaud starting a pitch measurement on ITMY. 3:16 PM PT

It seem like the sensor correction on ETMX is good. Several times H1:ISI-ETMX_ST2_SENSCOR_X_MATCH_GAIN has been reset to 0, I'm not sure what does this but maybe it is in the safe.snap or in one of the scripts. 

My question for SEI people is:

is there a correct time to turn this on durring the ISI turn on process?  Can we leave it on all the time?  Can we have turning it on be automated?

Yuta, Stefan, Kiwamu

Our goal in this morning was to get the PRMI locked with the 3f signals and test the stability. However we didn't get there yet.

A good news is that the initial dither alignment worked well, so that the alignment became less painful. Once PRX and PRY were aligned by the dither system, we could get a decent alignment in PRMI.

What we did in this morning:

• We ran the dither system on PRX and PRY via the guardian
  • This worked well.
  • We still need to make the whole processes sequential. Also the PRM suspension model needs to be updated to get the newly separated dither excitation via IPC blocks as was done in the PR2 model.
• We adjusted REFLAIR_A_RF45 demod such that there is less cross coupling in the two loops.
  • We empirically ended up with a phase of 142.7 deg.
  • This gave a less funny dip in the PRCL open loop transfer function.
• We increased the PRCL gain from -0.4 to -1.3 to have a stable PRC loop.
• We found that the ITMY ring heater had been off
  • This probably explains why we had to gain up the PRCL loop.
  • The cavity build up in POPAIR_B_RF18 was measured to be about 18000 counts with the whitening gain set at the maximum of 45 dB.
  • This is certainly smaller than before. We used to see it reaching 30000 counts with 4 W of the power in the ITMY ring heater.
• We let the PRMI be locked for a long enough time so that we can make an noise budget attempt for the first time.
  • Yuta is working on making a NB plot
• We are leaving the ITMY ring heater with a bit higher power of 7 W (i.e. 3.5 W in upper and lower heaters). See the attached.

Before starting measuring Pitch to Pitch on ITMY, I realized the oplev signal was oscillating at high frequencies. I took a spectra and compared with other oplevs, and the noise floor from 1000Hz is ~2 orders of magnitude higher compared to ITMX. Thomas pointed out that this oplev doesn't have an analog whitening filter yet, so we should double check when it will be installed.
dtt template was saved under SusSVN/sus/trunk/QUAD/H1/ITMY/SAGL2/Data/2014-02-28_Oplev_Spectra.xml

Evan, Yuta

We redid the Michelson noise budget using the NB simulink tool.
The data we used was taken from the long strech MICH lock on Feb 17 (see alog #10127).
The noise budget agrees well with we have done by hand (see alog #10221, #10217).
NB model works pretty nicely.

[Result]
NB_MICH_1076736130.png: Michelson noise budget plot. "Measured" curve shows the sensitivity curve calibrated from H1:LSC-MICH_OUT. Seismic noise is copied from LLO measurement, so seismic noise curve is not real.

sensorNB_MICH_1076736130.png: Noise budget for sensor noises. Low frequency noise mainly comes from the demodulator board noise. Compared with the one we did by hand (alog #10221), there is a factor of 5 difference in ADC noise. This is because we forget to put the PD filter bank gain (H1:LSC-REFLAIR_A_RF45_(I|Q)_GAIN) in the previous one.

[Note on NbNoiseSink]
If you put NbNoiseSink block into your model, noise at the sink will be calibrated by taking out the model loop gain. If you specify a DAQ channel to that sink and if you run 'nbAcquireData' function, it gets the actual data from a NDS server (using Matlab "get_data" function). So, NB tool automatically calculates the sensitivity curve for given GPS time.

Also, note that LiveParts and data fetching fail if you specify GPS time more than ~7 days ago. I'm not sure why, but it is a limitation from "get_data" function.

H1fw1 and the LDAS gateway were rebooted at 14:42 PST to correct problems with keeping the frame writer process running.  On examination of the LDAS gateway for frame writer 0, it was determined that significant problems existed that required a reboot of that as well.  This interrupted the h1nds0 NDS server and the h1dmt data stream.  

Stefan, Kiwamu,

We could not close the oplev damping loops today. It turned out that the oplev servo gains had been wrong since the model reboot on the last Tuesday. In fact, the signs were wrong in safe.snap file. We updated the safe file. Now they are 0.03 and -0.03 for pitch and yaw loops respectively.

[Ed, Evan]

We are preparing to make a measurement of the length of the PRC using the phase-locked auxiliary laser technique described by Chris Mueller (T1400047). Previously, this has been used to measure the Livingston IMC length (LLO alog 9599).

We set down a 520 mW Lightwave NPRO on the IOT2R table, along with a Faraday isolator and steering mirrors. We will inject this beam into the PRM_refl side of the IOT2R periscope. The beam will hit the back of IM4, and a small fraction (2400 ppm) will be transmitted toward the PRM. This gives 1.2 mW of auxiliary power on the PRM, compared to 9 mW of 45MHz PSL single-sideband power on the PRM.

Most of the auxiliary power should reflect from the back of IM4 and return to the IOT2R table via the IO_forward side of the periscope. For mode-matching, we hope that we can simply send part of the IO_forward beam onto a New Focus 1611 and maximize the observed beat. Currently, there is 3 mW of power in the IO_forward beam.

Using this beat, or otherwise, we will phase-lock the auxiliary laser to the PSL carrier beam. Then with PRMI locked on the PSL sideband, we will sweep the offset to the auxiliary PLL and monitor the RF coming out of POPAIR_B. We should see the strength of the RF reach a maximum whenever the auxiliary beam is coresonant with the the PSL sideband. By tracing out the Lorentzian profile of the RF amplitude across successive resonances of the PRC, we can extract the FSR of the PRC. Given a design length of 57.6557 m, we expect an FSR of 2 599 850 Hz. If we can measure the FSR to within 100 Hz, we can get the PRC length to within 2 mm.