We repeatedly had CHARD run away when switching to LOWNOISE_ASC. 

The investigation was not helped lockloss tool, which started crashing suddenly. Guaridan also started having issues (connection errors, white epics channels.)

In an attempt to bring some clarity into the beam jitter discussion I looked at the Gouy phase evolution of the input beam. I collected distance and focal length information from a variety of sources, notably T1200470 and E1200616.

The biggest uncertainty I had was the PSL persicope to HAM1 viewport distance - if someone knows that, let me know.

The MATLAB script with all the numbers is in ~controls/sballmer/20160930/inputBeamCalc.m. It creates a structure of the following form, fully describing the beam and optical mode:

IMC =
    lambda: 1.0640e-06               % The wavelength in m
         q: 0.2325 +13.3832i          % The input Gaussian beam parameter q=z+izR in m
         N: 4                                 % number of optics
      dist: [16.2406 16.2406 0.2325 0.2325 0] % Distances between the optics (one more than N)
      ifoc: [0.0731 0 0 0]               % inverse focal length for all lenses and mirrors (f=R/2 for mirror)
     label: {'MC2'  'MC3'  'Waist'  'MC1'} % optic names
 

The attached .mat file contains the following structures of that form::

IMP:    Input beam: From PSL periscope to MC3
IMC:    Input Mode Cleaner
IMCp:  Input Mode Cleaner from MC1 to MC3 only (output path inside IMC)
IM:      Input Mirros: from IMC to PRM
PRC:  Power Recycling Cavity forward path
PRCr: Power Recycling Cavity return path
PRCrt:Power Recycling Cavity round trip

Below are plots and data for the different beam segments.

• LASER tripped off due to a Diode chiller error but I was able to reset and restart.
• TCS Y chiller water topped off
• some pesky earthquake action broke a lock which only resulted in a short down time. ISI blends were switched to EQ_v2
• µseism holdong steady below .02µm/s
• wind speeds barely exceeding 10mph
• Some locklosses occurring at LOWNOISE_ASC/COIL_DRIVERS

BS horizontal wedge doesn't have any couterpart Y path. Any horizontal beam shift on the BS will cause MICH path difference because of this wedge.

(CPY has a horizontal wedge but it's a mirror image of CPX about BS HR surface. ITMY has a vertical wedge but again it's a mirror image of ITMX.)

This effect doesn't look small, and a rough calculation shows that the HPO jitter peaks from DBB (but not broad humps seen in DBB) are consistent with what we see in DARM.

----

In the attached cartoon, when the beam on the BS is shifted in Y direction by y (blue line), the optical single-trip distance from the beam spot on the BS to ITMX gets shorter than to ITMY by:

EQN1: MICH = n*yW*sqrt(2)/cos(theta) - yW*(1+tan(theta)) ~ 0.7896yW = 1.0E-3*y

where W=1.3E-3 rad is the wedge, n=1.4496 is the refractive index and theta=29.2deg=0.5096rad is the angle of refraction.

DARM will only see about (1-rI)/(1+rI) of MICH where rI is the amplitude reflectivity of the ITMs, which is sqrt(0.986):

EQN2: DARM=MICH*(1-rI)/(1+rI) = 3.5E-3*MICH ~ 3.5E-6 * y.

Because of this, DARM should have a flat coupling to YAW displacement on BS.

Instead of the displacement y, we'll use the displacement A which is just the displacement normalized by the beam radius on the BS (53.4mm):

EQN3: DARM=3.5E-6*53.4[mm]*A = 1.9E-7 [m] * A.

----

Now, let B be a normalized misalignment parameter out of HPO (i.e. HPO jitter), e.g. B=disp/waistRadius+ i* angle/divAngle. Note that abs(B) is conserved unless 01/00 mode ratio is altered.

PMC gives us 1.6% HOM01 suppression in amplitude (T0900616).

For IMC there's some difference for PIT and YAW due to additional sign flip on top of Gouy shift per round trip, but PIT suppression is about 0.5% and YAW is 0.4% in amplitude assuming that the common round trip Gouy shift is 102 deg.

For PRC, assuming carrier recycling gain of 30, PRM reflectivity of 97% and the round trip Gouy shift of 32.5 deg, HOM01 suppression is 5.7% in amplitude.

PMC, IMC and PRC combined, HOM01 suppression is 3.7e-6 for YAW, 4.6e-6 for PIT.

Though A cannot be known without knowing the Gouy shift from HPO to BS, since |B| is conserved except through cavities, we can set the upper bound on |B| using the HOM suppression through PMC, IMC and PRC as

EQN4: |A| < 3.7e-6 * |B|

therefore upper bound on DARM

EQN5: DARM < 7E-13 [m] * |B|.

DBB measurements of B are available in alog, e.g. alog 29754 (direct link to the HPO jitter plot is here).

If you look at 1kHz peak which does not appear in DBB RIN measurement, the peak height is somewhere between 5E-6 to 1E-5, and using EQN5 and these numbers,

EQN6: DARM < 3.5 to 7E-18 [m/sqrtHz] at 1kHz peak.

In DARM spectrum, 1kHz peak is visible at or somewhat above 1E-19 [m/sqrtHz], so it seems consistent with EQN5.

----

Caveats:

IF the broad noise in DBB jitter measurement is actually jitter,  assuming the same jitter coupling as the above, you should see the huge broad thing in DARM, which we do not.

This coupling only applies to YAW. As for PIT, there could be similar coupling mechanism if ITM wedges don't cancel with each other, but the imbalance is only 0.001 degree.

MICH is not the only thing that is modulated by this, SRCL and PRCL are simultaneously modulated. Full simulation might be useful.

I added ≈100mL to the TCSY chiller .  Attached are pics of BOTH levels BEFORE water was added. (TCS_1569.jpg is TCSY).

J. Kissel

Now that we have updated EPICs records that store the DARM loop model (LHO aLOG 29992), I wanted to confirm the Nutsinee's recent 15 Mpc BNS range gain (LHO aLOG 30085)was not because the calibration of the DARM response was changing. I've confirmed the gain is not because the calibration has changed. I also wanted to sanity check the new DARM model parameter time dependence channels (LHO aLOG 29855). I'm sure if I trust them yet; we still need a comparison against offline calculations of the same thing (aka "SLM Tool").

See all of the attachments.
Plot (2) -- Recreates her picture of the BNS range trend over 24 hours. Note that the two times I compare are shown with vertical lines on the dataviewer trace in 
   - red (1159272420, Sep 30 2016 12:06:43 UTC, Sep 30 2016 05:06:43 PDT) and in 
   - magneta (1159270000, Sep 30 2016 11:26:23 UTC, Sep 30 2016 04:26:23 PDT).

Plot (1) -- Recreates her ASD, at the above mentioned times. This shows the full spectrum, and the drastic difference in noise performance between the two times.

Plot (3) -- This shows the same ASD comparison, but zoomed in around the calibration line frequencies, all generated by PCALY, and read out in the calibrated reflected PD, or RXPD. This qualitatively shows that there's very little change in the line heights at any of the line frequencies.

Plot (4) -- This shows the magnitude and phase of the transfer function between PCAL and DELTA L EXTERNAL in (m/m) around the low frequency calibration lines; 7.93 and 36.7 [Hz]. At the calibration line frequencies, we can see a change of at most 5% and 3 [deg]. This is about the uncertainty in the overall calibration anyways (though this has not yet been officially quantified for the current state of the instrument). This is encouraging, with respect to the SRC detuning optical spring -- whatever detung we have appears to be unimpacted by these alignment changes. A *much* more rigorous study than two points of one lock stretch should be done, however.

Plot (5) -- Same as plot 5, but for the high frequency calibration lines; 331.9 and 1083.7 Hz. We see similarly little change in these magnitudes and phases.

Plots (6-9) -- These show trends of the optical gain, cavity pole, and the actuation strength of the QUAD compared against the BNS range over the 4 hour period in which Nutsinee was playing. While there is a clear correlation between the cavity pole and optical gain against the BNS range, I'm not sure I believe that the cavity pole has changed by 30 [Hz] given such little change in magnitude and phase of the 331.9 [Hz] line. Again, a *much* more rigorous, offline comparison should be done between these values which are calculated from the calibration line amplitudes and phases.

This is a plot of the jitter measured by the IMC WFS DC PIT/YAW sensors during last nights lock. The 280 Hz periscope peak reaches about 1x10^-4/√Hz in relative pointing noise, or about 3x10^-4 rms. The relative pointing noise out of the HPO is about 2x10^-5/√Hz at 300 Hz. After the attenuation through the PMC this would correspond to a level below 10^-6/√Hz. The jitter peaks show up in DARM, if they are high enough. This is clearly visible in the coherence spectra.

The ISS second loop control signal is an indication of the intensity noise after the mode cleaner with only the first loop on. The flat noise level above 200 Hz is around 3x10^-6/√Hz in RIN, with peaks around 240 Hz, 430 Hz, 580 Hz and 700 Hz. Comparing this to the free-running noise in alog 29778 shows this RIN level at 10^^-5/√Hz. We can also compare this with the DBB measurements, such as in alog 29754: the intensity noise after the HPO shows a 1/f behaviour and no peaks. Looking at the numbers it explains the noise below 300 Hz. It looks like a flat noise at the 10^-5 level including the above peaks gets added to the free-running intensity noise after the PMC. The peaks in the controls signal of the second loop ISS line up with peaks visible in the pointing noise. But, neither the numbers nor the spectral shape matches. These peaks have coherence with DARM.

I have done a cross correlation analysis of the two OMC DCPDs in order to study the spectral shape of the broad band noise that has been limiting the sensitivity in 200-1000 Hz band. Here are some results.

The red curve is cross-correlated noise of the two DCPDs in terms of the DCPD current in mA (with DARM loop effect removed).The data I used is a lock stretch from Sep. 26th which lasted for 6 hours in undisturbed. The frequency resolution or bandwidth is set to 0.1 Hz. The data length is 3 hours starting at 9:00:00 UTC of the day. The DARM suppression is taken out from OMC DCPD SUM. I have used the latest PreER10 DARM model (29931) to obtain a reasonable DARM suppression function. As we all know, there is high noise in 200-1000 Hz whose level is as high as shot noise (or null stream shown in green) in this frequency band.

I also made another plot, which is an overlay of 11 cross correlated spectra each of which is from different time but from the same lock stretch. See below.

This time, the frequency resolution is set to 1 Hz, so that we can get several FFTs out of the same lock stretch. Each spectrum is produced by an 18 minutes integration. As shown in the plot, there was one spectrum which is polluted by some kind of glitch. Otherwise, the shape of noise stayed almost unchanged over 3 hours. Also fig files are available.

@ ≈23:45 UTC

Summary:  Morning was spent recovering from Power Glitch.  Afternoon H1 was handed over for Commissioning.  Here are some notable activities from shift:

• 6-11am Power Glitch Recovery (see specific alogs)
• 8-11am PCal Y Work (Travis & Rick)---->Turned OFF BRS while they were out there & turned back on when they were done.
• ~11 Started Initial Alignment (with some issues)
• 12:30 Started locking (with some issues)
• 1:30 H1 at INCREASE_POWER & staying here so Terra can monitor PIs (mainly because since the ESDs went down during Power Glitch, PIs can be finicky for the next 12hrs as the ESDs return to a nominal state).
• For the afternoon, commissioners took H1 at various "mid-states" (i.e. INCREASE_POWER, DC_READOUT, etc.)

Notes:

• CP4 Alarms today (Chandra)
• PCal-Y crew left an illuminator ON at EY.  They will turn it off on Tuesday (Travis)

Locking Notes:

• ALSy had trouble getting over 0.6.  Possibly Clearing History for ALSy WFS fixed this.
• In FIND_IR while trying to adjust the DIFF OFFSET, TR_Y's zero was not at zero...it was more like 0.25.  Did NOT run DARK_OFFSETS & instead addressed this by hand by going to LSC OVERVIEW / LSC_CUST_TRPD / LSC_CUST_TRPD_INPUT.  And on here we adjusted the offsets for -TR_X_QPD_B_SUM & -TR_Y_QPD_B_SUM.  (after getting H1 to lock, these offsets would be taken care of by Guardian later).

model restarts logged for Thu 29/Sep/2016
2016_09_29 03:35 h1nds0
2016_09_29 14:10 h1nds0
2016_09_29 14:27 h1nds0
2016_09_29 14:32 h1nds0

installed more memory in h1nds0 to make it the same as h1nds1, no further unexpected restarts of h1nds0 for the rest of the day (and none on Friday at time of writing).

Overfilled CP4 again:

1. Opened exhaust bypass valve
2. Doubled LLCV (66% open)
3. Took ~17 min. to fill reservoir from 92% to 100%.
4. Took another hour to see LN2 pour out the exhaust. 
5. TC 252B did not respond to temp. change (and I broke the other one yesterday)
6. Set LLCV to nominal at 32% open in manual mode after the overfill
7. Would like to see how long it takes CP to recover to 92% at nominal LLCV (but may not happen this time since it's the weekend)

I used the borescope to look at the TC junction on the exhaust to see if it's measuring the pipe or the fluid flow. I couldn't see much but I suspect it's measuring just the pipe since the time delay on CP3 is long and temps never drop below -12degC on CP3 (response is faster when heating pipe with heat gun). 

I would like to drill and tap TCs directly into the pipe and use the existing signal lines to measure a more direct temp. for CP3 automation.