The new QPD installation is turning out to be much more hard than it originally seemed. After finsing the cable, QPD and interface board, when I tried to use the cable, it seems like it is not powered or connected to the interface board at all. I tried both QPD ports on the QPD interface board (D990692) both none worked. I measured the output pins of IDC style connector on the interface board and they seem to have the correct voltages at the correct pins. But when I connect this to our cable and go to the other side of the cable which is a DB25, use a breakout board and see for the voltages, I see nothing. The even pins which are supposed to be connected to each other and to GND are also not connected to each other. I pulled out teh DB25 end of the cable and brought it close to the IDC end to do a direct conitnuity test and this test failed too.

I even foudn another IDC end of a spare QPD cable hanging near 1Y2, but could not find the other end of this cable either.

So moving forward, we have following options:

• Assume the cable is bad and try to find another cable.
  • It is very hard to find these cables in the lab. Koji and I have already done one sweep.
• Source 26 pin 2 row IDC female connector and make a ribbon cable ourselves.
  • We probably will need to buy this connector for this to work.
  • Downs has apparantely thrown away all IDC connectors.
• Use clean room QPD that does not use this interface.
  • The QPD used in clean room tests for suspension hanging used a different board.
  • This board is just lying on the floor, mounted on one slot of a big 6U chassis.
• Use AS WFS
  • If used in current position, it would not be useful for BHD port or tuning LO1, LO2, and AS4.
  • If taken to ITMY oplev table, we will need to source LO and opther connections right at the PD head as that is design for these PDs.
• Use GigE camera
  • We can replace the analog camera with a GigE camera on the BHD output.
  • We will need to revide GigE camera code and medm screens for this, and run an ethernet cable to ITMY oplev table.
• Someone verify that the cable is indeed not working as I am seeing above. If I am wrong, I would be a happier person.

[Paco, Yuta]

We removed splitter to route BHD DC PD signals to c1hpc and c1lsc. This was necessary to circumvent IPC error, but this is no longer necessary. Now BHD DC PD signals are ADC-ed with c1hpc, and sent to c1lsc via IPC.
We also found that BHD DC PD signals have whitening filters as described in LIGO-T2000500 (Readout board is LIGO-D1400384).
We added unwhitening filter zpk([151.9;3388],[13.81],1,"n") to C1:HPC_BHDC_A and B, based on measured whitening stage gain (see Sec 3.1 of characterization reoprt in LIGO-T2000500).
This solved the signals leaking to minus (40m/17068).

Next:
 - Modify c1hpc model to send BHD DCPD signals to c1lsc after unwhitening. (Note added on Nov 15: The same unwhitening filter is also added to C1:LSC-DCPD_A and B for now. See attached.)
 - Redo visibility measurements,

[Yehonathan, Yuta, Paco]

We would like to estimate:

• LO phase sensitivty (for RF55 + audio dither scheme), as a function of RF demod angle (both I and Q); not to be confused with audio dither angle.
• LO phase sensitivity (for all schemes like in Attachment #2 of this previous post) but with some nonzero MICH offset.
• LO phase sensitivity (for RF55 + audio dither scheme) but with the uBHDBS (44:56) values from this post.

[Paco, Yuta]

MICH was locked with BHD DCPD A-B signal with LO phase controlled.
Locking procedure and configuration was as follows (see Attachment #1).

1. Lock MICH with AS55_Q, with C1:LSC-MICH_GAIN=-3, FM4, FM5, FM8, FM10 (boost filters are turned off to have more phase margin).

2. Lock LO PHASE with BH55_Q, with C1:HPC-LO_PHASE_GAIN=6, FM5, FM8, feeding back to AS1.
  - C1:LSC-BH55_PHASE_R=136.136 deg was tuned to minimize I when AS-LO is fringing with MICH locked with an offset of 50 (we first thought 136.136 deg - 90 deg is better from 40m/17216, but today, 136.136 deg seems to work better; Reason needs to be investigated).
  - We are supposed to use C1:HPC-BH55_Q_AS1_DEMOD_I_OUT to control the LO phase to give maximum MICH signal on BHD_DIFF (40m/17170), but somehow BH55_Q without audio dither was OK to get MICH signal. Line injection at 211.1 Hz on BS was seen in BHDC_DIFF (and AS55_Q), even if we use BH55_Q to lock LO PHASE (see Attachment #2; MICH_B is BHDC_DIFF and MICH_A is AS55_Q) or BH55_Q_AS1_DEMOD_I to lock LO PHASE (with both signs). Reason needs to be investigated.
  - Audio dither was done using AS1 with excitation of 15000 counts at 281.79 Hz. C1:HPC-BH55_Q_AS1_DEMOD_PHASE=60 deg was tuned to minimize Q with injection of line at 13 Hz using LO1.

3. Handed over MICH lock from AS55_Q to 0.66 * C1:LSC-DCPD_A - 1 * C1:LSC-DCPD_B. This was done by using C1:LSC-MICH_A and MICH_B gains. C1:LSC-MICH_A_GAIN=1 was handed over to C1:LSC-MICH_B_GAIN=-1.
  - 0.66 * A - B was tuned so that BHDC_DIFF will be zero (as it supposed to be with MICH offset of zero).
  - AS55_Q and BHDC_DIFF had roughly the same optical gain at 211.1 Hz (actually, BHDC_DIFF had higher optical gain; see Attachment #2), so we used MICH_A_GAIN=1 and C1:LSC-MICH_B_GAIN=-1
  - After handing over of BHDC_DIFF, OLTF was measured. UGF was ~70 Hz (Attachment #3).

Next:
  - Investigate how to get optimal LO phase. With BH55_Q or BH55_Q + audio dithering? How to optimize demod phases?
  - How do we balance DCPD A and B? What is the effect of BHD BS being 44:56 not 50:50?
  - Measure amount of MICH signal in BHDC_DIFF with different LO phases.
  - Improve SNR in BH55.
  - It will be much simpler if we send BHDC_SUM and BHDC_DIFF to c1lsc from c1hpc, instead of sending un-unwhitened BHDC_A and B.

BHD fringe visibility was measured again with unwhitening filters on on BHDC_A and B, which removed signal leakage to zero (40m/17265).
The result didn't change much from previous measurement (40m/17067) thanks to using the 'mode' of signal to calculate visibility.
Measured constrast of 74% indicate mode-matching AS beam to LO beam of 56%.

ITMX-LO fringe (10% percentile)
Contrast measured by C1:HPC-BHDC_A_OUT is 74.46 +/- 0.07 %
Contrast measured by C1:HPC-BHDC_B_OUT is 74.25 +/- 0.07 %
Contrast measured by all is 74.35 +/- 0.07 %

ITMY-LO fringe (10% percentile)
Contrast measured by C1:HPC-BHDC_A_OUT is 74.01 +/- 0.10 %
Contrast measured by C1:HPC-BHDC_B_OUT is 73.85 +/- 0.09 %
Contrast measured by all is 73.93 +/- 0.08 %

Errors are from standard deviation of 3 measurements.
The notebook lives in /opt/rtcds/caltech/c1/Git/40m/scripts/CAL/BHD/measureContrast.ipynb

Optical gains of AS55 and BH55 are calibrated for BHD MICH.

LO-ITM single bounce:
 With LO-ITM signle bounce fringe, optical gain of BH55_Q is measured using a method similar to MICH calibration in AS55 (40m/16929).
 Demodulation phase for BH55 is tuned to minimize I when LO-ITM is freeswinging (using getPhaseAngle.py).
 (Notebook: /opt/rtcds/caltech/c1/Git/40m/scripts/CAL/BHD/BHDOpticalGainCalibration.ipynb)
 Results are the following:

 LO-ITMY fringe: 7.84e9 counts/m (demod phase 147.1 +/- 0.3 deg) See Attachment #1
 LO-ITMX fringe: 8.44e9 counts/m (demod phase 149.6 +/- 0.4 deg) See Attachment #1

 Difference in the optimal demodulation phase 2.5 +/- 0.5 deg agrees with half of Schnupp asymmetry, as expected (40m/17007).
 Difference in the optical gain for LO-ITMY and LO-ITMX is probably from statistical fluctuation.

BHD MICH:
 Sensing matrix was measured by injecting a line at BS (300 counts @ 211.1 Hz), LO1 (5000 counts @ 287.1 Hz) and AS1 (5000 counts @ 281.79 Hz), when MICH is locked with AS55_Q and LO PHASE is locked with BH55_Q (both with no offset).
 Using the sensing matrix, demodulation phase was tuned to minimize I phase for MICH signal in AS55 and LO1 signal in BH55.
 After the demodulation phase tuning. sensing matrix was measured to be the following.
 See, also Attachment #3 for injected peaks. I phase signal is successfully suppressed by at least an order of magnitude.
 (Notebook: /opt/rtcds/caltech/c1/Git/40m/scripts/CAL/SensingMatrix/MeasureSensMatBHD.ipynb)

Sensing Matrix with the following demodulation phases (counts/counts)
{'AS55': -160.15695076011946, 'BH55': 154.13916838400047}
      Sensors           MICH @211.1 Hz          LO1 @287.1 Hz            AS1 @281.79 Hz           
C1:LSC-AS55_I_ERR_DQ    1.22e-05 (120.53 deg)   7.24e-07 (85.64 deg)     1.26e-06 (40.42 deg)    
C1:LSC-AS55_Q_ERR_DQ    2.95e-03 (-101.62 deg)  1.24e-06 (-80.43 deg)    1.69e-06 (152.31 deg)    
C1:LSC-BH55_I_ERR_DQ    1.28e-03 (80.95 deg)    3.44e-06 (109.31 deg)    2.22e-06 (154.40 deg)    
C1:LSC-BH55_Q_ERR_DQ    7.44e-03 (77.38 deg)    2.56e-04 (-59.85 deg)    2.42e-04 (6.40 deg)    
C1:HPC-BHDC_DIFF_OUT    2.21e-03 (82.45 deg)    4.37e-05 (121.87 deg)    3.61e-05 (-169.09 deg)

 Using BS actuation efficiency of 26.08e-9 /f^2 m/counts (40m/16929), optical gain for AS55_Q and BHDC_DIFF for MICH is

2.95e-03 / (26.08e-9/(211.1**2)) = 5.04e9 counts/m (AS55_Q for MICH)
2.21e-03 / (26.08e-9/(211.1**2)) = 3.78e9 counts/m (BHDC_DIFF for MICH)

 For AS55_Q, this is a factor of 4~5 higher than the previous measurement from free swing (40m/16929). Why?
 Free swing measurement was done again, and this gave 1.24e9 counts/m, which is consistent with the previous measurement (see Attachment #3).

 Using LO1 and AS1 actuation efficiencies of 3.14e-8 /f^2 m/counts (40m/17206), optical gains for BH55_Q for LO1 and AS1 are

2.56e-04 / (3.14e-8/(287.1**2)) = 6.72e8 counts/m (BH55_Q for LO1)
2.42e-04 / (3.14e-8/(281.79**2)) = 6.12e8 counts/m (BH55_Q for AS1)

Next:
 - Compare them with expected values
 - Measure them with different locking points (different LO phases, MICH offsets)
 - Investigate why MICH optical gain in AS55 is 4~5 times higher than free swing measurement (use different modulation frequency?)

Summary of actuation calibration so far (counts from C1:LSC-xx_EXC or C1:SUS-xx_LSC_EXC):
BS   : 26.08e-9 /f^2 m/counts (see 40m/16929)
ITMX :  5.29e-9 /f^2 m/counts (see 40m/16929)
ITMY :  4.74e-9 /f^2 m/counts (see 40m/16929)
ETMX : 10.91e-9 /f^2 m/counts (see 40m/16977 and 40m/17014)
ETMY : 10.91e-9 /f^2 m/counts (see 40m/16977)
MC2 : -14.17e-9 /f^2 m/counts in arm length (see 40m/16978)
MC2 :   5.06e-9 /f^2 m/counts in IMC length (see 40m/16978)
LO1 : 3.14e-8 / f^2 m/counts (see 40m/17206)
LO2 : 2.52e-8 / f^2 m/counts (see 40m/17206)
AS1 : 3.14e-8 / f^2 m/counts (see 40m/17206)
AS4 : 2.38e-8 / f^2 m/counts (see 40m/17206)

[Paco, Yuta]

We found that MICH UGF was unexpectedly high, ~200 Hz, in the measurement yesterday, which makes the closed loop gain to be more than one at MICH line injection at 211.1 Hz.
We did optical gain calibrations for AS55, BH55 and BHDC_DIFF in BHD MICH again with UGF at around 10 Hz.
This solved the inconsistent result with free swing calibration.

What we did:
 Did the same measurement for BHD MICH as written in 40m/17274, but with MICH UGF of ~10 Hz and LO PHASE UGF of ~15 Hz (see OLTFs in Attachment #1, and filter configurations in Attachment #2).
 Updated sensing matrix is as follows

Sensing Matrix with the following demodulation phases (counts/counts)
{'AS55': -163.52789698340882, 'BH55': 152.7860744565449}
      Sensors        	MICH @211.1 Hz       	LO1 @287.1 Hz       	AS1 @281.79 Hz       	
C1:LSC-AS55_I_ERR_DQ	1.85e-05 (-118.82 deg)	3.31e-07 (-32.19 deg)	7.86e-07 (112.27 deg)	
C1:LSC-AS55_Q_ERR_DQ	7.32e-04 (59.57 deg)	1.19e-06 (158.17 deg)	9.07e-07 (-92.25 deg)	
C1:LSC-BH55_I_ERR_DQ	5.02e-04 (-123.21 deg)	1.79e-05 (-26.73 deg)	1.76e-05 (-120.23 deg)	
C1:LSC-BH55_Q_ERR_DQ	1.75e-03 (59.57 deg)	2.71e-04 (-22.64 deg)	2.56e-04 (-114.37 deg)	
C1:HPC-BHDC_DIFF_OUT	1.00e-03 (-115.93 deg)	3.09e-05 (-14.99 deg)	2.84e-05 (-110.23 deg)	

 Using BS actuation efficiency of 26.08e-9 /f^2 m/counts (40m/16929), optical gain for AS55_Q and BHDC_DIFF for MICH is

7.32e-03 / (26.08e-9/(211.1**2)) = 1.25e9 counts/m (AS55_Q for MICH) This is consistent with freeswing measurement (1.24e9 m/counts) 40m/17274
1.00e-03 / (26.08e-9/(211.1**2)) = 1.71e9 counts/m (BHDC_DIFF for MICH)

 Using LO1 and AS1 actuation efficiencies of 3.14e-8 /f^2 m/counts (40m/17206), optical gains for BH55_Q for LO1 and AS1 are

2.71e-04 / (3.14e-8/(287.1**2)) = 7.12e8 counts/m (BH55_Q for LO1)
2.56e-04 / (3.14e-8/(281.79**2)) = 6.47e8 counts/m (BH55_Q for AS1)

  (Notebook: /opt/rtcds/caltech/c1/Git/40m/scripts/CAL/SensingMatrix/MeasureSensMatBHD.ipynb)

Next:
 - Compare them with expected values
 - Measure them with different locking points (different LO phases, MICH offsets; LO phase can be calibrated using optical gain calibration of BH55_Q)

MICH optical gain was measured with different LO phases over ~90 degrees.
Zero crossing of BH55_Q_ERR seems to be roughly 55 degrees away from optimal LO phase.

What we did:
 - Locked MICH with AS55_Q with no offset, with UGF at ~10 Hz (same as configuration in 40m/17279).
 - Injected BS calibration line at amptilude of 300 counts at 211.1 Hz.
 - Locked LO Phase with BH55_Q with different offsets added at C1:HPC-LO_PHASE_OFFSET.
 - Measured sensing matrix at that frequency. Counts are calibrated into meters using actuator efficiencies as described in 40m/17279.
 - LO phase was obtained using a DC value of BH55_Q. This was calibrated into degrees from the following:

Amplitude of LO-AS fringe in BH55_Q was calculated to be

A = BH55optgain*lamb/(4*pi) = 60 counts

where BH55optgain is 7.12e8 counts/m, which is optical gain of BH55_Q for LO1 measured in 40m/17279.
(Actually, BH55_Q goes upto ~ +/-200 counts in time series data, but maybe 60 is the nominal fringe amplitude, considering alignment fluctuations and fluctuation in AS darkness? Note that, no offset in BH55_Q is assumed in this calculation, but AM etc can create an offset.) 
LO phase can be obtained by

LOphase = arcsin(BH55_Q/A)

where BH55_Q a DC value (10 sec average) of BH55_Q.

Result:
 Attachment #1 is uncalibrated plot C1:HPC-LO_PHASE_OFFSET of around +/- 50 was the maximum we could add, and more offset gave unstable lock.
 Attachment #2 is calibrated plot. AS55_Q does not depend on LO phase, as expected. BH55_Q and BHDC_DIFF depend on LO phase as expected. BH55_I and AS55_I stay at low level, as expected (this means that our RF demodulation phase is OK).
 Dotted gray line is an eyeball fit of expected curve (40m/17170) to fool your eyes.
 This tells you that we are roughly 55 deg away from LO phase which gives maximum MICH signal for BHDC_DIFF.
 Error bar in x-axis is from standard deviation of BH55_Q fluctuations. Error in y axis is probably ~20% at maximum.
 Notebook: /opt/rtcds/caltech/c1/Git/40m/scripts/CAL/SensingMatrix/MeasureSensMatBHD.ipynb

Next:
 - Repeat the measurement with
   - MICH locked with higher UGF, with notch at 211.1 Hz, for more robust AS dark fringe
   - DCPD A and B balanced at 211.1 Hz (null MICH signal for BHDC_SUM to balance?)
   - Measure optical gain also for BHDC_SUM and BH55_Q demodulated at audio dither
   - Lock LO phase at different sign so that we can sweep LO phase over ~180 deg
   - Sign-sensitive optical gain measurement (demodulation with BS motion necessary)
 - Compare with expected values from simulations
 - Why do we have 55 degrees offset? Expected offset is 90 degrees...
   - Check if there is any RAM in 55 MHz in the input beam by measuring AM with ITM single bounce

[Paco, Yuta]

We realized that bandstop filters ("violin" filters) we implemented in 40m/17206 had pass band gain of -1dB.
gain(1,"dB") was added to all the filters (see Attachment #1 for gain adjusted violin filters for AS1).
We also realized that audio dither frequency we chose to generate BH55+audio dither error signal and to measure sensing matrix at ~280 Hz was too close to violin filters.
These will affect calibrations by upto ~60%.
For example, actuation gains should be actually

• LO1 = 3.14e-8 / f^2 m / cts * 3dB = 4.44e-8 / f^2 m / cts (3 violin filters)
• LO2 = 2.52e-8 / f^2 m / cts * 0dB = 2.52e-8 / f^2 m / cts (no violin filters)
• AS1 = 3.14e-8 / f^2 m / cts * 3dB = 4.44e-8 / f^2 m / cts (3 violin filters)
• AS4 = 2.38e-8 / f^2 m / cts * 4dB = 3.36e-8 / f^2 m / cts (3 violin filters+ bandstop at 96.7 Hz)

Next:
 - Redo actuator calibrations for LO1, LO2, AS1, AS4
 - Redo sensing matrix measurements with different audio dither frequencies for LO1 and AS1

As there is some confusion in actuator calibration, we have done the measurement again from scratch.
Results are the following.
New values for LO1, LO2, AS1, AS4 are obtained from free swinging ITMY-LO, so it should be more robust.

BS   : 26.54e-9 /f^2 m/counts
ITMX :  4.93e-9 /f^2 m/counts
ITMY :  4.90e-9 /f^2 m/counts
LO1  : 26.34e-9 /f^2 m/counts
LO2  :  9.81e-9 /f^2 m/counts
AS1  : 23.35e-9 /f^2 m/counts
AS4  : 24.07e-9 /f^2 m/counts

BS, ITMX, and ITMY actuator calibration:
 Followed the procedure in 40m/16929.
 Calibrated AS55_Q using X-Y plot to be 9.72e8 counts/m (Attachment #1), locked MICH with UGF of 10 Hz, and measured the transfer function from C1:LSC-BS,ITMX,ITMY_EXC to C1:LSC-AS55_Q_ERR.
 The result is Attachment #2. They are consistent with 40m/16929.

LO1, LO2, AS1, and AS4 actuator calibration:
 Followed similar steps with ITMY-LO fringe.
 Calibrated BH55_Q using X-Y plot to be 7.40e9 counts/m (Attachment #3), locked ITMY-LO with UGF of ~15 Hz (Attachment #4), and measured the transfer function from C1:SUS-LO1,LO2,AS1,AS4_LSC_EXC to C1:LSC-BH55_Q_ERR.
 The result is Attachment #5. They are inconsistent with 40m/17284, but this one should be more robust (see discussions below).

LO1, LO2, AS1, and AS4 actuator calibration by taking the ratio between ITMY:
 We have also followed the steps in 40m/17206 to calibrate BHD actuators.
 This method does not depend on BH55_Q optical gain calibration, but depends on ITMY calibration.
 Measured OLTFs for ITMY-LO fringe locking is Attachment #6, and actuator ratio with respect to ITMY is Attachment #7. In this measurement, Bandstop filter at 96.7 Hz for AS4 was turned off, and gain was lowered by a factor of 2 to avoid AS4 oscillating.
 This gives

LO1  : 116.81e-9 /f^2 m/counts
LO2  :  51.69e-9 /f^2 m/counts
AS1  : 101.48e-9 /f^2 m/counts
AS4  : 117.84e-9 /f^2 m/counts

 These are not consistent with 40m/17284, and larger by a factor of ~2-3.
 These are also not consistent with the values from free swinging measurement, and are larger by a factor of ~4-5.
 I guess there are some gains missing when comparing ITMY loop in c1lsc and other loops in c1hpc.

MICH optical gain with a sign was measured with different LO phases over ~180 degrees, with updated calibration and higher MICH UGF.
Zero crossing of BH55_Q_ERR seems to be 68 degrees away from optimal LO phase.

Calibrated sensing matrix:
 - Locked MICH with AS55_Q at dark fringe, with UGF of ~200 Hz. Notch at 311.1 Hz was turned on.
 - Locked LO PHASE with BH55_Q, with UGF of ~10 Hz (C1:HPC-LO_PHASE_GAIN=-2, using LO1).
 - Measured the sensing matrix as written in 40m/17279, but with different dither frequencies to avoid violin mode frequencies and to match with already-installed notch filters.
 - Sensing matrix was calibrated into meters using actuator gains measured in 40m/17285
 - Sign was added by comparing the phase with C1:SUS-xx_LSC_OUT. If they are 90-270 deg apart, minus sign was added to the sensing matrix.
 - Resuts are as follows. At least important green ones are consistent with previous measurements (40m/17279).

Calibrated sensing matrix with the following demodulation phases (counts/m)
{'AS55': -164.1726747789845, 'BH55': 169.57651332419115}
      Sensors            MICH @311.1 Hz           LO1 @147.1 Hz           AS1 @141.79 Hz           
C1:LSC-AS55_I_ERR_DQ    2.72e+06 (84.89 deg)    6.41e+05 (14.60 deg)    -1.98e+05 (206.79 deg)    
C1:LSC-AS55_Q_ERR_DQ    -1.20e+09 (-228.85 deg)    -1.43e+06 (-106.51 deg)    1.41e+06 (29.21 deg)    
C1:LSC-BH55_I_ERR_DQ    -2.45e+09 (-230.64 deg)    -6.57e+07 (167.84 deg)    7.28e+07 (-16.56 deg)    
C1:LSC-BH55_Q_ERR_DQ    7.81e+09 (-48.64 deg)    -7.34e+08 (159.70 deg)    8.06e+08 (-10.08 deg)    
C1:HPC-BHDC_DIFF_OUT    -9.91e+08 (-224.55 deg)    -1.13e+08 (164.14 deg)    1.26e+08 (-3.95 deg)    
C1:HPC-BHDC_SUM_OUT    -6.84e+06 (-104.69 deg)    1.50e+07 (-8.71 deg)    -1.79e+07 (173.80 deg)    
LO phase from C1:LSC-BH55_Q_ERR_avg 4.98e-03 +/- 1.65e+01 deg

Estimating LO phase:
 - Using 7.34e+08 counts/m, which is an optical gain of BH55_Q for LO1, LO phase can be estimated as follows.

A = BH55optgain*lamb/(4*pi) = 62 counts
LOphase = arcsin(BH55_Q/A)

 - When C1:HPC-LO_PHASE_GAIN is plus, LOphase was calculated with the following to take into account of the sign flip in the controls.

LOphase = 180 - arcsin(BH55_Q/A)

Balancing A-B:
 - BHDC_A and BHDC_B were balanced to give null MICH signal in BHDC_SUM at 311.1 Hz. This gave BHDC_DIFF = 0.919*A - B.
 - It seems like this balancing gain changes over time by ~30%.

Result:
 - Attachment #1 is uncalibrated MICH optical gain in different LO phases, and Attachment #2 is the calbirated one. Basically the same with 40m/17282, but with updated calibration and sign considerations.
 - In addition to the previous measurements, we can see that BHD_SUM is not dependent on LO phase (small dependence probably from not perfect A and B balancing).
 - 0 deg of LO phase means that it is a zero crossing of BH55_Q with a slope that LO PHASE loop can be closed with a minus C1:HPC-LO_PHASE_GAIN, feeding back to LO1.
 - Dotted and dashed gray lines are from scipy.optimize.curve_fit using the following fitting function (not an eyeball fit this time!).

def fitfunc(x, a,b,c):
    return a*np.sin(np.deg2rad(x-b))+c

 - Fitting results show that we are -22 deg away from our intuition that BH55_Q crosses zero when BHDC_DIFF give no MICH signal (68 degrees away from optimal LO phase).
 - Fitting results also show that BH55_Q sensitivity to MICH crosses zero when BHD_DIFF sensitivity to MICH maximizes. This suggests that BH55+MICH dither can be used to lock LO phase to optimal LO phase.

Notebook: /opt/rtcds/caltech/c1/Git/40m/scripts/CAL/SensingMatrix/MeasureSensMatBHD.ipynb

Next:
 - Compare with expected values from simulations
 - Why do we have -22 deg?
   - Check if there is any RAM in 55 MHz in the input beam by measuring AM with ITM single bounce (quick measurement shows it is small)
   - Unbalanced BHD BS?
   - Contribution from 55 MHz sidebands from LO beating with 55 MHz sidebands from AS?
       - Lock LO phase using audio dither only (demodulate BHDC_DIFF?).

[Paco, Yuta]

MICH displacement sensitivity was compared under AS55_Q locking and BHD_DIFF locking.
Sensitivity with BHD was better by more than an order of magnitude due to smaller sensing noise.
During the measurement, LO phase fluctuation was ~13 deg RMS.

Locking configurations:
 - MICH was first locked with AS55_Q, no offset, and then handed over to BHD_DIFF after LO phase locked. FM2, FM3, FM4, FM5, FM6, FM8, FM10 on, C1:LSC-MICH_GAIN=-3 gave UGF of around 80 Hz.
 - LO PHASE was locked with BH55_Q, no offset. FM5, FM8 on, C1:HPC-LO_PHASE_GAIN=-2 feeding back to LO1 gave UGF of around 40 Hz.
 - Attachment #1 shows the OLTFs.

Sensitivity estimate:
 - Sensitivity was estimated using measured actuator gains and optical gains. Following numbers are used.

C1:LSC-AS55_Q_ERR to MICH 1.08e-9 counts/m (measured at 311.1 Hz today)
C1:HPC-BHDC_DIFF to MICH 1.91e-9 counts/m (measured at 311.1 Hz today)
BS   : 26.54e-9 /f^2 m/counts (40m/17285)
LO1  : 26.34e-9 /f^2 m/counts (40m/17285)

 These numbers were also reflected to C1:CAL-MICH_CINV and C1:CAL-MICH_A.
 C1:CAL-MICH_A_GAIN = 0.5 was used to take into account of LSC output matrix of MICH to BS being C1:LSC-OUTPUT_MTRX_8_2=0.5.

 - Attachment #2 shows the displacement spectrum of MICH (top) and LO PHASE (bottom). Brown MICH curve is when locked with AS55_Q and black MICH curve is when locked with BHD_DIFF. RMS of original and in-loop LO PHASE was estimated to be

 Original LO phase noise: 393 nm RMS (266 deg RMS)
 In-loop LO phase noise: 19.4 nm RMS (13 deg RMS)

Next:
 - Improve LO phase loops to reduce LO phase noise
 - Estimate LO phase noise contribution to MICH sensitivity

[Anchal, Yuta]

To send BHD signals from c1hpc after unwhitening and taking sum/diff, c1hpc and c1lsc models are modified.
PDDC_DOF_MTRX medm screen was modified to reflect this change.
We don't need to unwhiten and take sum/diff again in c1lsc model anymore
 

c1hpc has option of dithering BS now (sending excitation to BS LSC port to c1sus over IPC). This is available for demodulating BHDC and BH55 signals. Also BS is a possible feedback point, however, we would stick to using LSC screen for any MICH locking.

c1sus underwent 2 changes. All suspension models were upgraded to the new suspension model (see 40m/16938 and 40m/17165). Now the channel data rates are set in simulink model and activateDQ script is not doing anything for any of the suspension models.

Changed the BHD BS transmissivity to 0.56.

Demodulation Phases

As was noted before. The LO phase sensitivity plot vs LO phase from the previous elog shows the optimal sensitivity at each LO phase. That means that the optimal demodulation phase might change as a function of LO phase. Attachment 1 shows the previous plot and a plot showing the optimal modulation phase for some of the methods. When double demodulation is involved I optimize one modulation and show the optimal demodulation angle of the second. As can be seen, optimal audio demodulation angles don't change as a function of LO phase.

Additionally, as expected maybe, for the single RF sideband methods that nominally should not have worked at nominal LO phase (angle in which BHD Diff is most sensitive to MICH), the optimal demodulation angle changes quite a bit around the nominal LO phase.

Fixed demodulation angle

Attachment 2 shows the LO phase sensitivity in the single 55MHZ sideband method when we fix the demodulation angle. -23.88 is the demod angle optimal for nominal LO phase. 66.12 is 90 degrees away from that. -75.21 is the is the demod angle optimal for LO phase at the amplitude quadrature and 14.78 is 90 degrees away from that. It can be seen that fixing the demod angle can be mostly harmless.

Effect of MICH offset

The simulations were run with 0 MICH offset. Attachment 3 shows the LO phase sensitivity of the different methods when MICH offset is introduced together with the optimal demod angle. As expected the single RF SB methods are sensitive to this offset while the double demod methods are not since they are not relying on DC fields.

[Yuta, Paco, Anchal]

We measured

(a) BHDC_DIFF sensitivity to BS dither for a set of MICH offsets.

Configurations

• MICH locked with AS55_Q
  • The MICH offset was varied below
• LO_PHASE locked with BH55_Q
  • Balanced DCPD_A and DCPD_B by applying a digital gain of 1.00 to DCPD_A
  • Changed the BH55 demod angle to 140.07 deg to minimize BH55_I
• BS dither at 311.1 Hz
  • Use newly added HPC_BS Lockins to readback the demodulated signals

Results & Discussion

The analysis was done with the '/cvs/cds/rtcds/caltech/c1/Git/40m/scripts/CAL/BHD/BHD_DIFFSensitivity.ipynb' notebook.

Attachment #1 shows the main result showing the sensitivity of various demodulated error signals at 311.1 Hz for a set of 21 MICH offsets. We noted that if we didn't randomize the MICH offset scan, we observed a nonzero "zero crossing" for the offset.
Note that, although LO_PHASE loop was always on to control the LO phase to have zero crossing of BH55_Q, actual LO phase is not constant over the measurement, as MICH offset changes BH55_Q zero crossing.
When MICH offset is zero, LO_PHASE loop will control the LO phase to 0 deg (90 deg away from optimal phase), and BHDC_DIFF will not be sensitive to MICH, but when MICH offset is added, BHDC_DIFF start to have MICH sensitivity (measurement is as expected).
For BHDC_SUM, MICH sensitivity is linear to MICH offset, as it should be the same as ASDC, and does not depend on LO phase (measurement is as expected).
For BH55_Q, MICH sensitivity is maximized at zero MICH offset, but reduces with MICH offset, probably because LO phase is also being changed.

Fields at the BHD BS. More on this later.

[Anchal, Paco, Yuta]

Attachment #1 is the same plot as 40m/17303 but with MICH sensitivity for ASDC and AS55 also included (in this measurement, BH55 demodulation phase was set to 140.07 deg to minimize I fringe).
Y-axis is now calibrated in to counts/m using BS actuation efficiency 26.54e-9 /f^2 m/counts (40m/17285) at 311.1 Hz.
2nd X-axis is calibrated into MICH offset using the measured AS55_Q value and it's MICH sensitivity, 8.81e8 counts/m (this is somehow ~10% less than our usual value 40m/17294).
ASDC have similar dependence with BHDC_SUM on MICH offset, as expected.
AS55_Q have little dependence with MICH offset on MICH offset, as expected.

This plot tells you that even a small MICH offset at nm level can create MICH sensitivity for BHDC_DIFF, even if we control LO phase to have BH55_Q to be zero, as MICH offset shifts zero crossing of BH55_Q for LO phase.

Notebook: /opt/rtcds/caltech/c1/Git/40m/scripts/CAL/BHD/BH_DIFFSens_pydemod.ipynb

Attachment #2 is the same plot, but BH55 demodulation phase was tuned to 227.569 deg to have no MICH signal in BH55_Q (a.k.a measurement (c)).
In this case, LO phase will be always controlled at 0 deg (90 deg away from optimal), even if we change the MICH offset, as BH55_Q will not be sensitive to MICH.
In this plot, BHD_DIFF have little sensitivity to MICH, irrelevant of MICH offset, as expected.
MICH sensitivity for BH55_I is also constant, which indicate that LO phase is constant over this measurement, as expected.

Notebook: /opt/rtcds/caltech/c1/Git/40m/scripts/CAL/BHD/BH_DIFFSens_pydemod.ipynb

Attachment #3 is the same plot, but BH55 demodulation phase was tuned to 70 deg.
This demodulation phase was tuned within 5 deg to maximize MICH signal in BHD_DIFF with large MICH offset (20).
In this case, LO phase will be always controlled at 90 deg (optimal), even if we change the MICH offset, as BH55_Q will not be sensitve to LO carrier x AS sideband component of the LO phase signal.
In this plot, BHD_DIFF have high sensitivity to MICH, irrelevant of MICH offset (at around zero MICH offset it is hard to see because LO_PHASE lock cannot hold lock, as there will be little LO phase signal in BH55_Q, and measurement error is high for BHD_DIFF and BH55 signals).
MICH sensitivity for BH55_I and BH55_Q is roughly constant, which indicate that LO phase is constant over this measurement, as expected.

These plots indicate that BH55 demodulated at MICH dither frequency can be used to control LO phase robustly at 90 deg, under unknown or zero MICH offset.


Notebook: /opt/rtcds/caltech/c1/Git/40m/scripts/CAL/BHD/BH_DIFFSens_pydemod.ipynb

LO phase delay:
 From these measurements of demodulation phases, I guess we can say that phase delay for 55 MHz in LO path with respect to MICH path (length difference in PR2->LO->BHDBS and PR2->ITMs->AS->BHDBS) is

2*(227.569-70(5)-90)-90 = 45(10) deg

 This means that the length difference is (omegam=5*2*pi*11.066195 MHz)

c * np.deg2rad(45(10)+360) / omegam = 6.1(2) m   (360 deg is added to make it close to the design)

  Is this consistent with our design? (According to Yehonathan, it is 12.02 m - 5.23 m = 6.79 m)

  Attachment #4 illustrates signals in BH55.

Next:
 - Lock LO PHASE with BH55 demodulated at MICH dither frequency (RF+audio double demodulation), and repeat the same measurement
 - Finer measurement at small MICH offsets (~1nm) to see how much MICH offset we have
 - Repeat the same measurement with BH55_Q demodulation phase tuned everytime we change the MICH offset to maximize LO phase sensitivity in BH55_Q (a.k.a measurement (b)).
 - What is the best way to tune BH55 demodulation phase?

[Paco, Anchal]

I changed the script in /opt/rtcds/caltech/c1/Git/40m/scripts/SUS/outMatFilters/createF2Afilters.py to read the measured POS resonant frequencies stored in /opt/rtcds/caltech/c1/Git/40m/scripts/SUS/InMatCalc/resFreqs.yml instead of using the estimate sqrt(g/len). I then added Q = 3 F2A filters into FM1 output filter of LO1, LO2, AS1 and AS4 suspensions in anticipation of BHD locking scheme work.

I checked the LSC rack to evaluate what we might need to generate 44 MHz rf in the hypothetical case we go from BH55 to BH44 (a.k.a. double RF demod scheme). There is an 11 MHz LO port labeled +16 dBm (measured 9 Vpp ~ 23 dBm actually) on the left hand side. Furthermore, there is an unused 55 MHz port labeled "Spare 55 LO". I checked this output to be 1.67 Vpp ~ +8.4 dBm. Anyways the 55 MHz doesn't look very nice; after checking it on the spectrum analyzer it seems like lower frequency peaks are polluting it so it may be worth checking the BH55 LO (labeled REFL 55) signal to see if it's better. Anyways we seem to have the two minimum LOs needed to synthesize 44 MHz in case we move forward with BH44.

[Paco, Yuta]

We confirmed the noisy 55 MHz is shared between AS55, BH55 and any other 55 MHz LOs. Looking more closely at the spectrum we saw the most prominent peaks at 11.06 MHz and 29.5 MHz (IMC and PMC nominal PM freqs). This 55 MHz LO is coming all the way from the RF distribution box near the IOO rack. According to this diagram, this 55 MHz LO should have gone through a bandpass filter; interestingly, checking the RF generation box spare 55 MHz the output is *cleaner* and displays ~ 17 dBm level... ??? Will continue investigating when we actually need this RF.

Many changes have been done to c1hpc to support dual demodulation at audio frequencies. We moved away with ASS style of lockin setup as the number of connections and screens required would become very large. Instead now, the demodulation is done for a selected oscillator, on a selected signal. Similarly, the demodulated signal can be further demodulated for another selected oscillator. Please familarize yourself with new screen and test the new model. The previous version of the model is kept as backup alogn with all it's medm screens, so nothing is lost. Shown as an example in the screenshot, AS1 and BS oscillators can be turned on, and BHDC_DIFF signal can be demodulated first with BS and next with AS oscillator to get the signal.

[Anchal, Yuta]

We have modified c1cal model to support sensing matrix measurements for BHD PDs on Friday last week.
c1cal model now can inject dither to LO1, LO2, AS1, and AS4, and demodulate BH55_I, BH55_Q, BHDC_SUM and BHDC_DIFF signals.
Related models, c1lsc, c1hpc, and c1sus2 are also modifed accordingly.
MEDM screens are also edited accordingly.
Attachments highlight the modifications.

{Yuta, Yehonathan}

Today we lock LO phase using audio+RF method in two variants: AS1+RF and MICH(BS)+RF. We measure the TFs and find that AS1 variant has a UGF ~ 17Hz and MICH variant has a UGF ~ 32Hz.

Details

1. We lock MICH in the usual way using AS55. ITMs are aligned to make AS port dark. We use a single bounce and optimize mode-matching with LO beam by minimizing the BHDDC-A signal.

2. Using the new BHD Homodyne phase control MEDM screen we first try AS1. We put an elliptic filter with 80Hz corner frequency on the DEMOD1 filter bank. We find that the notch of that filter is at 281.768Hz and this is where we put the AS1 dither line.

AS1 is dithered with 20000 counts. We optimize the DEMOD1 demodulation angle by dithering LO1 at 27Hz and minimizing the Q quadrature in diaggui. We find that 45 degrees is the optimal demod angle. We lock the LO phase with a gain of ~ 45 and take the OLTF (attachment 1).

3. Next, we use MICH degree of freedom to lock LO phase. We dither BS with the same frequency as before with 4000 counts. Higher counts seem to put some offset on ASDC. As before we optimize the DEMOD1 demod angle and find it to be 115deg. We lock LO phase with a gain of 20 and take the OLTF (attachment 2).

[Yehonathan, Yuta]

We found that moving AS1 in yaw improves power on ASDC and AS55.
We compensated this move with AS4 and SR2 to keep the BHD fringe (ITM single bounce and LO beam fringes ~600 counts in amplitude at BH55).
We have also aligned BHD CCD camera to avoid clipping on a lens just before the camera (all the other optics on ITMY table remain untouched).
After the alignment, MICH BHD sensing matrix were measured with new C1CAL model (40m/17339) under the following conditions.
 - Locked MICH with AS55_Q at dark fringe. Notch at 311.1 Hz was turned on.
 - Locked LO PHASE with BH55_Q with C1:HPC-LO_PHASE_GAIN=-2, using LO1.

Sensing matrix with the following demodulation phases (counts/m)
{'AS55': -161.16488964312092, 'BH55': 162.57275834049358}
Sensors      BS @311.1 Hz           LO1 @147.1 Hz           AS1 @141.79 Hz           
AS55_I       (-0.19+/-1.45)e+07    (-0.26+/-2.43)e+06    (+0.35+/-2.39)e+06    
AS55_Q       (-1.74+/-0.02)e+09    (+1.61+/-8.31)e+06    (+1.08+/-8.59)e+06    
BH55_I       (+3.01+/-0.17)e+09    (+3.20+/-9.59)e+07    (-3.67+/-9.46)e+07    
BH55_Q       (-6.77+/-0.45)e+09    (+1.09+/-0.17)e+09    (-1.22+/-0.18)e+09    
BHDC_DIFF    (-8.41+/-4.81)e+08    (-1.26+/-0.94)e+08    (+1.38+/-1.03)e+08    
BHDC_SUM     (-2.75+/-9.14)e+07    (+1.18+/-1.13)e+07    (-0.97+/-1.02)e+07    

AS55_Q optical gain to MICH and BH55_Q optical gain to LO phase was improved by ~45%, compared with previous measurements (see 40m/17287).
The value for AS55_Q is consistent with the free swing measurement as attached.
SENSMAT part of c1cal seems to be working fine.

[Yehonathan, Yuta]

Sensing matrix measurements at different LO phases were performed under LO phase locked to both BH55_Q and BH55_Q+MICH dither.
We confirmed that BH55_Q+MICHdither can lock LO phase to around maximum MICH sensitivity for BHD_DIFF.

Locking configuratons
 - MICH was lockied using AS55_Q feeding back to BS, at dark fringe. Notch at 311.1 Hz was turned on. C1:LSC-MICH_GAIN=-6 (lowered to reduce BS DAC saturation).
 - LO PHASE was locked using BH55_Q, feeding back to LO1. FM2, FM5, FM8 on. C1:HPC-LO_PHASE_GAIN=+/-2.
 - LO PHASE was also locked using BH55_Q+MICHdither. BS was dithered with C1:HPC-BS_POS_OSC_CLKGAIN=4000 at 281.768 Hz (2nd notch of ELP80 used for demodulation). Feeding back to LO1. FM5, FM8 on (no LF boost). C1:HPC-LO_PHASE_GAIN=+/-20.
  -- Note that we could not increase the dither amplitude more as BS DAC starts to saturate (we are using BS for MICH loop, sensing matrix measurement, and audio dither; see 40m/17343).

Sensing martix measurements
 - Lines are injected to BS @ 311.1 Hz with amplitude of 1000, LO1 @ 147.1 Hz and AS1 @ 141.79 Hz with amplitude of 5000.

Estimating LO phase
 - Estimation of LO phase was done in the same way described in 40m/17287. We used measured sensitivity of BH55_Q for LO1 at BH55_Q zero crossing (-1.42e9 counts/m) to estimate LO phase offset from BH55_Q zero crossing.
 - In BH55_Q+MICHdither case, LO phase was flipped using the following equation when C1:HPC-LO_PHASE_GAIN is minus (to have consistend LO phase dependence with BH55_Q locking. NEEDS CHECK).

LOphase = 180 - arcsin(BH55_Q/A)

Result
 - Attachment #1 shows the sensitivity of AS55, BH55, BHDC_DIFF/SUM to BS (upper panel), LO1 (middle) and AS1 (lower), under LO phase locked to BH55_Q. The upper plot is the same plot as 40m/17287. As we can see, "0 deg" in the x-axis is not the optimal phase for BHDC_DIFF to have maximum MICH sensitivity. "0 deg" is the optimal point in terms of BH55_Q sensitivity to LO1/AS1, as we tuned the demodulation phase to maximize it.
 - Attachment #2 shows the same plot, under LO phase locked to BH55_Q+MICH dither. Sensitivity of BH55_Q to MICH crosses zero at round these measurements, as we are zero-ing it with this locking scheme. Around these LO phases, sensitivity of BHDC_DIFF to MICH is maximized as expected. Also, sensitivity of BHDC_DIFF to LO1/AS1 is minimized, as expected (assuming residual MICH offset and contrast defect are small).
 - Attachment #3 is the combined data from #1 and #2. Data points from BH55_Q locking are marked with "o" and those from BH55_Q+MICH dither locking are marked with "x" (they have larger uncertainties in LO phase). Both measurements are somewhat inconsistent in some channels (BS to BHDC_DIFF and LO1/AS1 to BH55_Q). Needs further investigation.
 - Dashed lines are from scipy.optimize.curve_fit using the following fitting function.

def fitfunc(x, a,b,c):
    return a*np.sin(np.deg2rad(x-b))+c

Notebook: /opt/rtcds/caltech/c1/Git/40m/scripts/CAL/SensingMatrix/SensMatBHDvsLOPhase.ipynb

Next:
 - Lock MICH with BHDC_DIFF under LO phase locked to BH55_Q+MICHdither
 - Estimate LO phase noise contribution to MICH displacement sensitivity
 - Improve LO phase loop
 - Try audio+audio dither
 - Move on to FPMI
 - Move on to 44MHz
 - Estimate the amount of residual MICH offset and contrast defect from these plots

[Yehonathan, Yuta]

Here's a plot using same dataset from yesterday, but x-axis in raw BH55_Q data, not calibrated into degrees in LO phase.
This way you are free from calibration error in BH55_Q data to LO phase.
Elliptic fit is done using least squares.
dphi is calculated using the following equation where (ap, bp) are the semi-major and semi-minor axes, phi is the rotation of the semi-major axis from the x-axis.

beta=np.arctan(ap/bp/np.tan(phi))
dphi=-np.arctan(ap/bp*np.tan(phi))-beta

dphi gives you LO phase at zero-crossing.
For example, the top plot says that the sensitivity of BH55_Q to BS crosses zero at "-133.92 deg," which means BH55_Q+MICHdither can lock LO phase at -134 deg or 46 deg.
The top plot also says that the sensitivity of BHDC_DIFF to BS crosses zero at "127.45 deg," which means BHDC_DIFF sensitivity to MICH maximizes at 38 deg or 217 deg.
The middle plot says that the sensitivity of BH55_Q to LO1 crosses zero at "90.09 deg," which means BH55_Q+LO1dither can lock LO phase at 90 deg or -90 deg, and BH55_Q(no dither) can lock LO phase at 0deg or 180 deg (by definition).

Notebook: /opt/rtcds/caltech/c1/Git/40m/scripts/CAL/SensingMatrix/PlotSensMatBHDvsLOPhaseData.ipynb

Next
 - Use also BH55_Q+LO1/AS1dither to scan around 90 deg.

I've tuned one gold box RFPD to be resonant at 44.26 MHz and I left the notch to be near 66 MHz, however, it is only effective by 10 dB. Attached is the measured transimpedance using the test port. This measurement should be updated with PD testbed measurement.

This photodiode is ready to be installed for the dual RF lO phase locking scheme.

Thu Jan 19 15:06:43 2023 Updating the measurement with Moku:Pro calibration TF

[Anchal, Paco, JC, Yuta]

We have started hardware installations for BH44 RF PD. 44 MHz LO generation and signal chain from IQ demodulator was checked successfully, but found that IQ demodulator is not orthogonal.

RF PD Interface:
 - We have unplugged Ch2 of the RFPD Interface (labeled "Special") to re-use it for BH44. Ch2 was used for "UNIDENTIFIED" RFPD. DB15 cable was routed to ITMY table and connected to BH44 RF PD (40m/17398) now sitting on the cover of ITMY table. See Attachment #1.
 - Finding a DB15 RF PD interface cable was easy because of the organization work!

44 MHz LO generation:
 - LO for BH44 was generate following the scheme proposed in 40m/17319.
 - 11 MHz LO from RF distributor labeled "+16 dBm" (measured to be 16.5 dBm) and 55 MHz LO labeled "SPARE 55" (measured to be 2.26 dBm) was mixed with a mixer ZFM-1H-S+ (using 11 MHz as LO, and 55 MHz highpass filtered with SHP-50+ as RF). The mixer output was lowpass filtered with SLP-50+, and amplified with ZFL-500LN+, which gave 8.07 dBm at 44 MHz. The second heighest peak was -11.53 dBm at 22 MHz, which seems low enough. See Attachment #2 for the photo of the setup.

IQ demodulation:
 - We have used unused IQ demodulator labeled "AS165" to use it for BH44. See Attachment #3.
 - We have quickly checked if the IQ demodulator is working or not with LO from BH55, PD input of 55 MHz generated using Moku to see I and Q outputs. The outputs are sine waves at frequency consistent with the difference between LO frequency and "PD input" frequency, and the phase was off as expected. Q output was ~4 dBm higher than I output.

Measured diff of BH44:
 - After CDS modifications where done, BH44 IQ demodulator was tested by using 44 MHz LO generated in a method mentioned above, and injecting 11.066195 * 4 MHz signal from Moku as PD input. This gave ~75 Hz signal in C1:LSC-BH44_I and C1:LSC-BH44_Q.
 - With 0dB whitening gain and whitening/unwhitening filters off, gain imbalance was measured to be Q/I=137.04/62.49=2.19, and measured phase difference to be PHASE_D=27.21 deg (see Attachment #4; gpstime=1358051213).
 - With 0dB whitening gain and whitening/unwhitening filters on, gain imbalance was measured to be Q/I=138.44/63.21=2.19, and measured phase difference to be PHASE_D=26.95 deg (see Attachment #5; gpstime=1358051325138). This is consistent with whitening/unwhitening off, and noise is smaller, which mean whitening/unwhitening filters are probably working fine.
 - IQ demodulator board might be not working properly, as I and Q signals are not quite orthogonal.

Model changes:

• We modified c1lsc, c1hpc and c1cal model for BH44.
• BH44 ADC pins were identified and connected for RFPD phase rotator.
• The signals are sent to c1hpc through IPC where BH44 is now available for feedback loops in single and dual demodulation.
• The whitening filter controls and anti-aliasing filter enable buttons were created in c1iscaux slow machine db files.
• MEDM screens are updated accordingly (see Attachment #6).

Next:
  - Use different IQ demodulator board that has better IQ orthogonality.
  - Connect BH44 RF PD and use 44 MHz test input to check the signal chain.
  - Install BH44 RF PD optical path.
  - Try locking LO_PHASE with BH44.

I tested a spare IQ demod board labeled 33.3 MHz (closer to 44 MHz than the 165 MHz we had started using) and using the Moku adjusted the trim caps after the transformer T1 to adjust orthogonality (using an oscilloscope). The orthogonality seems quite good on this board and it seems to be working fine, so I decided to swap out the BH44 (previously AS165) IQ demod board for this one. To do this I first unpowered the amplifier, then disconnected the load (IQ demod board) then release from the Eurocrate, then add the new board.

Finally, using Marconi at 11.066195 * 4 to get close to the 44 MHz LO frequency, I measured a 63.9 Hz tone at the C1:LSC-BH44_I_IN1 and C1:LSC-BH44_Q_IN1 channels (whitening gain 0 dB). The measured angle between these two channels was 86.97 deg, so the orthogonality is much better now. The gain imbalance is also relatively better, Q/I ~ 0.57

By sending two frequencies offset from eachother to LO input and RF input, we measured the remaining phase difference between I and Q outputs of this demod board and correct that by setting C1:LSC-BH44_PHASE_D to 86.2 degrees and balancing the amplitudes by putting C1:LSC-BH44_Q_GAIN to 1.747. See attached for XYPlot after correction.

the problems with these circle plots:

1. you have to make the aspect ratio of the PDF correct or else it doesn't look like a circle
2. what we care about is the deviation from circle, so you should plot the difference in a way that let's us see the difference, not just that it sort of looks like a circle. This is similar to how we plot the residual when doing fits.

[JC, Paco, Anchal, Yuta]

- We installed the new BH44 beam path and RFPD.

- JC installed the new beam path for the LO/AS camera.

- We succeeded in locking LO Phase with BH44_Q_ERR, but didn't attempt FPMI BH44 because we noted large 60 Hz harmonics in most of our RF error signals.

BH44 RPFD/Camera installation:

- We picked off LO/AS beam path previously going to the camera, and installed a Y1 (45 deg, s-pol) mirror, a f=150mm lens and the RFPD (Attachment #1). We initially tested it using the incandescent light from a flashlight and then aligned the beam, we also made sure it's not saturating.

- Using the spurious transmission from the mirror mentioned above, we steered a new beam path for the camera and realigned it using another short focal length lens (f ~ 100 mm).

LO Phase control:

- We increased the whitening gain from 0 dB to 42 dB for both C1:LSC-BH44_I and C1:LSC-BH44_Q, and zeroed the offsets. Even before this step we could see a fringe from BH44, which is quite promising!

- After alignment was recovered on the LO/AS path, we succeeded in locking the single bounce (ITMX) LO phase using BH44_Q. Here the configuration was FM4, FM5 and a gain of ~ 5 * 0.5 = 2.5 (to match the typical BH55_Q error point).

- While BH44_Q was used to control the LO phase, we saw the BH55_Q was not zero but actually almost at max fringe value (see Attachment #2). This implies the BH44_Q is indeed orthogonal to BH55_Q with respect to the LO Phase!

FPMI lock:

- We locked electronic FPMI but noted a  large 60 Hz + harmonics component in the RF error signals including AS55, BH55, REFL55, and BH44 (see Attachment #3). We could hand off to FPMI and even locked the LO phase with BH44_Q, but we were not sure the BHD_DIFF error signal was fit for handoff to FPMI BHD. Therefore we stopped here.

60 Hz + harmonics:

- We did a quick investigation around the areas we have been working in the lab to see if we had introduced this noise in any obvious way. First we checked the new amplifier for the 44 MHz LO, we briefly removed its power but the 60 Hz noise remained. Then we checked the AP table, but nothing had really changed there. We also disconnected and removed the rolling cart with the marconi and other instruments from the LSC rack. Finally, we turned all the lights down. None of these quick fixes changed the amount of noise.

- We also tried looking at these error signals under different IFO alignment and feedback configurations. We always see the noise in the AS55 and REFL55 quadratures, but not in BH44, BH55 or BHD_DIFF unless MICH is locked.

Next steps:

- Investigate more into 60 Hz noise, why? where?

- Measure sensing matrix with LO Phase locked with BH44 and BH55 to make comparison.

- FPMI-BH44

Here's the beam path of BH44.

[Paco, Yehonathan, Yuta]

Since we have installed BH44, we are seeing side lobes of 60 Hz + harmonics in AS55, REFL55, BH55, BH44, preventing us from locking FPMI BHD (40m/17405).

BH55 RF amp removed:
  - We have noticed that the side lobes are there in BH55 (but not in BH44) when LO-ITMX single bounce is fringing (ETMs and ITMY mis-aligned).
  - Changing whitening gains and turning on/off whitening/unwhitening filters didn't help.
  - When LO-ITMX single bounce is locked with BH55, the side lobe in BH55 reduces.
  - Dithering LO1 at 11 Hz created 180 +/- 11 Hz signal, which confirms that this side lobes are from the up conversion of optic motion.
  - We thought it could be from RF saturation, so we have put a 55-67 MHz bandbass filter (SBP-60+) in between BH55 RFPD and RF amp (ZFL-1000LN+; 40m/17195). Didn't help.
  - We then removed the RF amp. This largely reduced the side lobes (but still some at 180 Hz). We could lock LO-ITMX single bounce without the RF amp, so we decided to remove it for now.

Side lobes only when one of the arms is locked:
  - When ETMs are mis-aligned, MICH fringing and BHD fringing, there are 60 Hz + harmonics, but the side lobes are not there.
  - But with Xarm is locked (ETMY, ITMY mis-aligned) or Yarm is locked (ETMX, ITMX mis-aligned), the side lobes appear in AS55, REFL55, BH55, BH44.
  - Changing whitening gains and turning on/off whitening/unwhitening filters didn't help.
  - As the error signals are normalized by TRX and TRY, we turned on/off the power normalization, but didn't help.
  - Switching 60 Hz comb in BS, ITMX, ITMY, ETMX, ETMY suspension damping didn't help.

POY11 Investigations:
  - When ETMs are mis-alined, POX11 had relatively large 60 Hz + harmonics, but almost none in POY11 (unlike other RFPDs; see Attchment #1).
  - However, when ETMY is aligned and Yarm is loked with POY11, the side lobe grows in POY11.
  - Changing the feedback point from ETMY to ITMY or MC2 didn't help.
  - We have unplugged the IQ demod board for BH44 from the eurorack (without removing the cables) and removed the fuse for the power supply of the RF amp for 44 MHz generation (40m/17401), but these also didn't help.
  - We have also tried locking Yarm with REFL55(= ~2 x POY11), BH55(= ~10 x POY11), ALSY(= ~2000 x POY11), but the side lobes were always there.
  
Next:
  - Disconnect cables in BH44 to open possible ground loops made during BH44 installation (especially 44 MHz generation part??).
  - Check if the noise was there before BH44 installation using past data.

Just to show how bad 60 Hz noise is, I compared FPMI displacement noise with pre-BH44 era (measured on Jan 13, 40m/17400).
Blue curve in Attachment #1 is the sensitivity with FPMI locked with RF in pre-BH44 era, and pink curves are that measured today (C1:CAL channels are currently unavailable due to 0x2000 appeared after running restatAllModels.sh).
60 Hz + harmonics pedestals are apparent today, but was not there on Jan 13. Today, DARM could be handed over to AS55_Q from POX11-POY11, but CARM could not be handed over to REFL55_I from POX11+POY11 (this was possible last night).
Attachment #2 shows FPMI error signals when electronic FPMI is locked. Too much 60 Hz, especially in REFL55_I_ERR and AS55_I_ERR (note that REFL55_Q is used for MICH lock, but AS55_Q is not in-loop yet when this screenshot was taken.)

Next:
  - Fix c1cal 0x2000 issue
  - Fix REFL55 loose RF output
  - Disconnect cables in BH44 to open possible ground loops made during BH44 installation (especially 44 MHz generation part??).

1) do a comparison with the whtiening before the ADC on/off. This will tell us if it is pickup before the whtiening filter or not.

2) If there are ground loops made by 44 MHz setup, we want to draw a simple diagram which includes which sides are grounded and which have transformers. How about make a drawing to bring to the group meeting tomorrow? IN the lab we have these coaxial BALUNs for making a 1:1 transformer coupling.

3) Another source of 60 Hz is the unintentional demodulation of spikes made by the Sorensen switching supplies: they produce spikes all the way up to 100 MHz, so if they land near 44 MHz, you may get some 60 Hz on the demodulation. You should be able to see this with a dipole antenna or a hoop antenna.

1) Turning the whitening filter before the ADC on/off didn't changed the relative height of 60 Hz peak compared with the noise floor. When the whitening filter was turned on, increase of the dark noise measured at C1:LSC-****_(I|Q)_IN1 was roughly consistent by eye with the whitening filter transfer function (gain of 1 at DC, ~15 Hz zero x2, ~150 Hz pole x2), which suggests the 60 Hz pickup is before the whitening filter.

2) Attached is the electronics diagram around BH44 and BH55.

Following was done to investigate 60 Hz noise issue, but no significant change in the FPMI noise observed.


REFL55 inspection:
  - Even before BH44 installation, we have been experiencing flaky REFL55 RF output. When some work was done at AP table or something, sometimes the amplitude of REFL55_I and REFL55_Q goes very low, and/or offset changes. This was usually fixed by touching the RF output of REFL55.
  - So, we took out REFL55 and opened the back lid to inspect. RF output seemed rigid and the SMA connector was properly grounded to the box; didn't find any issue (Attachment #1).
  - REFL55 was put back to its original position, and the cables were also put back.

BH44 Kapton tape:
  - I realized that other RFPDs have Kapton tape in between the RFPD golden box and the black mount, but not for BH44 we recently installed.
  - I have checked that the golden box of BH44 and the optical table is not grounded when RF output and the DB15 cable was disconnected, but is gounded when they are connected, just like BH55.
  - Anyway I removed BH44 and put a Kapton tape (Attachment #2), just in case, and BH44 was put back to its original position, and the cables were also put back.

FPMI noise spectra after the work:
  - Attachment #3,4,5 are noise spectra of FPMI BHD sensors when FPMI is RF locked with AS55_Q, REFL55_I, and REFL55_Q, and LO_PHASE is locked with BH55 with the following configurations.
  - Attachment #3: whitening/unwhitening filters for AS55, REFL55, POX11, POY11, BH55, BH44 turned on (nominal configuration after lock acquisition)
  - Attachment #4: whitening/unwhitening filters for AS55, REFL55, POX11, POY11, BH55, BH44 turned off. No significant change except for expected whitening filter transfer function.
  - Attachment #5: whitening/unwhitening filters for AS55, REFL55, POX11, POY11, BH55, BH44 turned on, 30 dB resonant gain at 60 Hz, Q=10 in CARM loop. Significant 60 Hz reduction everywhere. This was not observed when resonant gain at 60 Hz was put in DARM loop (only 60 Hz at AS55_Q was reduced). 60 Hz noise mainly coming from something in CARM loop?

Don't forget to:
  - Put a beam dump for BH44

Kapton tape is a good insulator, so its a good block for 60 Hz. But it is mostly useless for RF since the capacitance between the mount and the table is

C = epsilon * A / d

For Kapton the dielectric constant is 3.5, the PD mount area is 10 cm x 10 cm, and the film thickness is ~50 um. So C ~ 5 nF.

Z ~ 1 / (2 * pi * 44 MHz) / C

  ~ 0.5 Ohms

[Paco, Anchal, Yuta]

Isolating BH44 setup from the rest didn't help reducing the 60 Hz noise.
Frequency noise from IMC also seems unchanged before and after BH44 installation.

Isolating BH44:
  - To see if BH44 setup installed is causing the 60 Hz issue, we compared the spectra of FPMI sensors with BH44 setup and with BH44 setup disconnected.
  - In the latter configuration, BH44 setup was isolated from the rest by disconnecting the SMA cables and the RFPD power cable, as shown in Attachment #1.
  - There was no significant difference in the spectra with BH44 and with BH44 isolated.
We have even put the old AS156 IQ demodulator board we have pulled out to insert BH44 IQ demodulator board back, but didn't change.
  - We have also disconnected the 22 MHz generation setup around 40m Frequency Generation Unit at 1X2 for switchable IMC/AS WFS, but it also didn't help.
  - Attachment #2 is the orignal spectra with both arms locked with POX and POY, feeding back to respective ETMs (MICH is not locked), and Attachment #3 is those with BH44 setup isolated, AS156 IQ demod back, and 1X2 22MHz generation isolated. Both look basically the same.
  - BH44 setup was reverted after the comparison.

IMC frequency noise:
  - As adding a resonant gain at 60 Hz helped reducing the 60 Hz noise (40m/17419), the noise might be from frequency noise. It also explains why it is not present in MICH when ETMs are mis-aligned, and only present when one of the arms is involved (40m/17413).
  - To see if the frequeny noise at 60 Hz increased after BH44 installation, I compared the spectrum of C1:IOO-MC_F_DQ on January 11 (same Wednesday) with that measured today at almost the same time.
  - Attachment #4 is the result. 60 Hz noise and its harmonics seems almost the same in MC_F. It is rather noisy today in other frequencies, but not at 60 Hz.

Next:
  - Read the book.

So far different measurements are consistent with the hypothesis that the 60 Hz noise is from PSL frequency noise (40m/17423).
We have done several measurements in REFL55 and POX/POY to show this hypothesis, and all are consistent with the frequency noise hypothesis.
However, 60 Hz noise in the IMC error point seems too small to explain the 60 Hz noise in DARM.
 

[Koji, Paco, Yuta]

REFL55 attenuation experiment:
 - To check if the 60 Hz is in the light or not, we have compared the spectrum of REFL55_I_IN1 with different ND fiters in front of REFL55 RF PD.
 - Attachment #1 shows the result. Spectrum was taken when both arms are locked indivitually using POX and POY, feeding back to ETMs, MICH freely swinging. Red curve is nominal, blue is with OD0.5, and green is with OD1.
 - Attenuation in 60 Hz noise and side lobes are consistent with OD filter attenuation, which suggets that the noise is from the light.
 - Note that having side lobes is natural, as MICH is fringing (sorry for confusing plot). However, if the side lobes come from the RF saturation, we expect side lobes to decrease more than OD filter attenuation, but this was not the case.

Phase measurements between POX and POY:
 - When both arms are locked indivitually using POX and POY, feeding back to ETMs, transfer function from POX to POY had gain of ~1 and the phase of -10 deg (Attachment #2).
 - With PSL shutter is closed, transfer function from POX to POY had gain of ~0.1 and the phase of -100 deg, with lower coherence (Attachment #3).
 - These also support that 60 Hz noise in POX and POY when the arms are locked are from common origin, such as frequency noise.

[Michael, Paco, Yuta]

IMC error point measurement:
 - Attachment #4 shows the IMC Servo Board configuration we used for the all three measurements below.

For this measurement we took TP1A (from MC Servo board) and buffered it with a battery powered SR560 (DC coupled, low noise, gain x1) before connecting it to the single ended A1 channel on a SR785. The noise level was set to -42 dBVpk, and three different noise spectra were acquired:

• In Blue, the IMC is locked
• In Orange, IMC is unlocked by closing PSL Shutter (dark)
• In green, IMC is unlocked by closing PSL Shutter and the Servo board IN1 is disabled.

The estimated (in loop) line noise (60 Hz) levels are 70 uV/rtHz, which using the calibration 13 kHz/Vrms (from 40m/14691) amounts to 0.9 Hz/rtHz of (supressed) frequency noise at IMC Error point.

This number (0.9 Hz/rtHz) in terms of displacement corresponds to 1.28e-15 m/rtHz. The measured DARM noise (2e-10 m/rtHz @ 60 Hz from 40m/17414) is not accounted for by this amount.

Next:
  - Check the IMC error signal calibration
  - Measure the calibrated out-of-loop frequency noise using various signals (POX, POY, REFL55, AS55 with single arm, ALSBEAT, PMC_CTRL)

Timeline (as far as written in the elog):
 - Dec 20: FPMI BHD locked using BH55 (40m/17367).
 - Dec 21: FPMI RF locked, but not BHD, DARM noise 1e-11 m/rtHz @ 60 Hz (40m/17369).
 - Jan 10-11: AS WFS boards testing at 1X2 (40m/17391, 40m/17393).
 - Jan 11: FPMI BHD locked using BH55, DARM noise 2e-11 m/rtHz @ 60 Hz (40m/17392).
 - Jan 13 2pm: FPMI BHD locked using BH55, DARM noise 2e-11 m/rtHz @ 60 Hz (40m/17399).
  - After measuring the sensing matrix etc., LO_PHASE locking became unstable and FPMI BHD could not be recovered (I thought something similar to Dec 21 is happening).
 - Jan 13 6pm: FPMI BHD locked using BH55 recovered. DARM noise 2e-11 m/rtHz @ 60 Hz. Discovered that the 60 Hz noise is higher when LO_PHASE locking is unstable (40m/17400).
 - Jan 17: BH44 hardware/software installed. Found IQ demod board needs tuning (40m/17401).
 - Jan 18: Tuned IQ demod board for BH44 installed (40m/17402).
 - Jan 19: BH44 RF PD placed and connected to IQ demod board. FPMI RF locked, LO_PHASE locked with BH44, but found 60 Hz noise everywhere (40m/17405).
 - Jan 24: FPMI RF locked (with CARM locked with POX11+POY11 instead of REFL55_I). DARM noise 2e-10 m/rtHz @ 60 Hz (40m/17414).
 - Jan 24: AS WFS boards mounted in 1X2 (40m/17416).
 - Jan 25: Isolating BH44 setup didn't help (40m/17423).
 - Jan 26: Fixed tripping of -5V supply in 1X1 (40m/17425).
 - Jan 27: IMC error point measurements at 1X2 (this elog).

One of these V beam dumps was installed for BH44 RF PD.
The rest is now stored in the box in the shelf along Yarm, together with RF PD mounts.

[JC, Paco]

I tried restarting all models this afternoon to see if the DC error 0x2000 on c1cal model cleared (currently no frame data is being written from c1cal). By running ./restartAllModels.sh on the scripts/cds/ folder, and then BURT restoring, but unfortunately this didn't work. I also tried restarting DAQD without success. Finally, I guess since the error has been related to channel inconsistencies in the model's .INI file (what where when?) I tried rtcds build c1cal and rtcds install c1cal and then another full restart. No success.

The 0x2000 error happens when the rtcds model can not acquire the requested number of channels at their data rates. Basically, there is a maximum total data acquisition that one model can do (which is unknown to me). To fix this, I removed 2048 Hz acquiring of C1:CAL-SENSMAT_<DOF>_<RFPD>_<I/Q>_DEMOD_SIG_IN1. This would not allow us to do software demodulation of calibration lines ourselves. but C1:CAL-SENSMAT_<DOF>_<RFPD>_<I/Q>_DEMOD_<I/Q>_IN1 are still acquired at 2048 Hz to do our own low pass filtering in software and C1:CAL-SENSMAT_<DOF>_<RFPD>_<I/Q>_DEMOD_<I/Q>_OUT are acquired at 256 Hz.

This removal worked, and restarting DAQD worked and now c1cal does not have any DC errors. Current total data acquisition C1:DAQ-FEC_50_TOTAL is 2521 which is less than our other heavy models like c1lsc, c1sus etc. So c1cal can probably acquire more in future, but care is required while adding new channels. This issue happened because we added BH44 to the calibration model.

[Paco, Yuta]

We estimated the frequency noise of IMC output beam at 60 Hz using different methods to see if they are consistent.
They are not inconsistent, but seems hard to explain by an easy single dominating noise source (multiple noise sources at similar noise level?).

IMC suspension damping:
 - We checked that 60 Hz comb filters are all on for all OSEM sensors of MC1, MC2, MC3 (Attachment #1), and they all have comb(60,30,-40,3), which is 60 Hz comb filter of Q=30, -40 dB, 3 harmonics.

Revisiting IMC error point calibrations:

 - We revisited this calibration in 40m/17431. First, 0.9 Hz/rtHz corresponds to 1.3e-13 m/rtHz, as L/nu = 40 m / 282 THz = 1.4e-13 m/Hz.
 - Also, we need to add a loop correction. MC servo board settings when we took this data was as follows:
   - +4 dB in IN1
   - 40 Hz pole, 4000 Hz zero filter was on
   - 0 boost
  assuming 1/f around UGF of 200 kHz (40m/17009), and 1/f^2 between 40-4000 Hz, openloop gain at 60 Hz will be (4e3/60)**2*(200e3/4e3)=2e5. So, the estimated frequency noise at the output of IMC in terms of arm length is 1.3e-13 m/rtHz * (1+G) = 2.6e-8 m/rtHz (or 1.8e-8 m RMS considering 0.5 Hz bandwith).
  - Noise measured with the same condition but PSL shutter closed was 7 uV/rtHz at 60 Hz (40m/17431). This correspond to 1.3e-14 m/rtHz (or 9.2e-15 m RMS), which is an estimated dark noise.

Measuring frequency noise using arms:
 - We then proceeded to measure frequency noise using arms locked with POX11 and POY11. Attachment #2 and #3 is calibrated XARM and YARM noise using the error signals and feedback signals. For both, it is 1e-10 m/rtHz at 60 Hz (or 4.3e-11 m RMS considering 0.187493 Hz bandwidth). And this is more than x10 higher than what we have measured in August 2022 (dotted lines).
 - MC_F calibrated using 1.4e-13 m/Hz reads 7.1e-9 m/rtHz at 60 Hz (or 3.1e-9 m RMS considering 0.187493 Hz bandwidth).
 - Noise measured at DARM in FPMI locked with RF (but CARM with POX11+POY11, as 60 Hz was too much to switch to REFL55_I) was 3e-10 m/rtHz at 60 Hz (or 1.8e-10 m RMS considering 0.374994 Hz bandwidth) (Attachment #5), which is roughly the same as past measurements (40m/17414).
 - To check if MC_F is calibrated correctly, we injected a line at 57 Hz with 3000 counts in amplitude into MC2. Using MC2 actuation efficiency -14.17e-9 /f^2 m/counts in arm length (40m/16978), this should give

14.17e-9/(60**2)*3000 = 1.2e-8 m -> 0.93e-8 m RMS

 in XARM length noise. RMS value of YARM calibrated spectra reads 1.1e-8 m (Attachment #4), which is consistent within ~20%, so MC_F calibration is OK. Note that MC_F at 60 Hz are at the same level in August 2022 (green curves).

Summary of frequency noise measurements at 60 Hz:
 - 1.8e-8 m RMS as measured at IMC error point TP1A
    This gives you total of IMC length noise, error point noise, PSL free run noise, feedback noise.
    Estimated dark noise at error point TP1A is 9.2e-15 m RMS, and is small.
    Calibration might be wrong, as this rely on IMC loop gain estimate and error signal calibration of 13kHz/V a while ago in 2018 (from 40m/14691, which is from 40m/13696), which might not be true for TP1A at 60 Hz (note that there is a 40 Hz/4000 Hz filter).
 - 3.1e-9 m RMS as measured at MC_F
    This gives you total of IMC length noise, error point noise, PSL free run noise, but the noise injected at feedback point before MC_F is suppressed by ~2e5.
    As estimated dark noise is much less, it is IMC length noise, PSL free run noise or noise injected after MC_F.
    Note that typical NPRO free run noise at 60 Hz is 1e4/60 Hz/rtHz * 1.4e-13 m/Hz = 2.3e-11 m/rtHz, and is small, but we might be having large NPRO noise.
 - 4.3e-11 m RMS as measured using XARM and YARM
     This gives you total of IMC length noise, error point noise, but PSL free run noise and feedback noise are suppresed by ~2e5.
     But this also includes noise injected in XARM and YARM loops.
     If this is mainly from PSL free run noise or feedback noise, we expect 3.1e-9 m RMS/2e5 = 1.6e-14 m RMS, so it doesn't explain 4.3e-11 m RMS.
     If this is mainly from IMC length noise, this should be equal to frequency noise measured at MC_F, but MC_F is higher by nearly two orders of magnitude.
     Noise in POX11 or POY11 are smaller by a factor of more than 100 when dark (see 40m/17431), so contribution from dark noise of POX11 and POY11 at 60 Hz to XARM/YARM noise is negligible.
     These mean that the noise might be from combination of IMC and PSL. (For example, if noise injected at error point is 9.2e-15 m RMS, IMC length noise is 4.3e-11 m RMS, PSL free run noise is 3.1e-9 m RMS, and noise injected at feedback point is 1.8e-8 m RMS, it explains all the measurements so far.)
 - 1.8e-10 m RMS as measured using FPMI DARM
     Frequency noise in DARM should be suppressed by common mode rejection, but it is actually x3 higher than what we see in XARM and YARM.
     There might be extra noise from FPMI loops (note that CARM is controlled by POX11+POY11 in this measurement).

Next:
  - Check IMC error point calibration (is 13 kHz/V correct?) by driving MC2 at around 60 Hz (but not at 60 Hz) by known amount
  - Measure frequency noise at IN1 of MC servo board to avoid 40 Hz/4000 Hz filter
  - Check what exactly are we measuring at MC_F. Are there possibility of additional noise for MC_F, which is not fed back to laser frequency?
  - Drive MC2 at around 60 Hz (but not at 60 Hz) to see if MC_F and X/YARM spectra matches
  - Estimate IMC length noise from MC OSEMs
  - Touch electronics around 1X2 to see if 60 Hz at IMC error point changes (monitor the live spectrum!)

[Paco, Yuta]

We devised a plan to systematically hunt for the line noise source assuming it has something to do with BH44 and/or our recent changes in the LSC rack. Our noise estimate is the IMC_error point (TP1A) at the MC servo board. Our traces from Attachment #1 represent, in the following order:

PSL shutter open, IMC locked, C1:IOO-MC_SW1 = 1

PSL shutter closed, IMC unlocked, C1:IOO-MC_SW1 = 1

PSL shutter closed, IMC unlocked, C1:IOO-MC_SW1 = 0

PSL shutter open, IMC locked, C1:IOO-MC_SW1 = 1,  rewired the delay box on LSC rack.

Same as above, plus we connected a loose "ground" wire to the delay box supply.

Same as above, plus we removed the PD interface unit connection.

Same as above plus we disconnected the 44 MHz local oscillators and terminated their outputs at the RF distribution box.

Same as above plus we disconnected the RFPD (BH44) from the IQ demod board at the LSC rack.

Where all the measurements used +4 dB input gain, 40:4000 filter enabled, and Boost = 0 settings on the MC servo board. In between measurements 3 and 4 we had to replace the SR560 (buffer) because the starting one ran out of battery... We found a good one in the YEND, used to buffer the OLTF excitation for the YAUX loop TF measurements.

For the loop calculation, don't you have to consider the IMC cavity pole? What about the analog filter on the output of the HV driver for the laser PZT?

[Anchal, Yuta]

We have measured OLTF of IMC loop, and revisited IMC error point calibration again.
Also, we have tried to break the ground loop between MC servo board and TTFSS, but didn't help.

IMC OLTF measurement:
 - IMC OLTF was measured using SR785 at TP1A and TP1B. MC servo board settings are the following.
   - +4 dB in IN1
   - 40 Hz pole, 4000 Hz zero filter was on
   - 0 boost
 - Eye-ball fit of OLTF gives zeros at [30e3,30e3] Hz, poles at [40,3e3,3e3] Hz (Attachment #1). 40 Hz pole is from 40:4000 Hz fiter in MC servo board and 4kHz zero is compensated by IMC cavity pole (~ 3.79 kHz). We are not sure where two 3k:30k are from.
 - Anyway, eye-ball fit gives OLTF gain of 1.7e5 at 60 Hz, which is accidentally roughly the same as previous estimate (40m/17446).

Revisiting IMC error point calibrations:
 - We realized that error signal calibration of 13kHz/V a while ago in 2018 (from 40m/14691, which is from 40m/13696) is a calibration for IN1.
 - So, 70 uV/rtHz at 60 Hz at TP1A corresponds to 70 uV/rtHz / 4dB / (4e3/60) * 13kHz/V = 0.009 Hz/rtHz, which corresponds to 1.2e-15 m/rtHz.
 - The estimated frequency noise at the output of IMC in terms of arm length is 1.2e-15 m/Hz * (1+G) = 2.0e-10 m/rtHz (or 1.4e-10 m RMS considering 0.5 Hz bandwith).
 - Noise measured with the same condition but PSL shutter closed was 7 uV/rtHz at 60 Hz (40m/17431). This correspond to 1.2e-16 m/rtHz (or 8.5e-17 m RMS), which is an estimated dark noise.

Summary of frequency noise measurements at 60 Hz:
 - 1.4e-10 m RMS (or 1.0e3 Hz RMS) as measured at TP1A (estimate of unsuppressed noise difference between IMC and PSL)
  - This being smaller than MC_F measurement is strange, as this should be an estimate of total unsuppressed noise (if 60 Hz noise is coherently cancelling each other, this can be explained).
 - 3.1e-9 m RMS (or 2.2e4 Hz RMS) as measured at MC_F
 - 4.3e-11 m RMS (or 3.1e2 Hz RMS)  as measured using XARM and YARM
 - 1.8e-10 m RMS as measured using FPMI DARM

Buffering MC servo board output to TTFSS:
 - We have inserted a battery-powered SR560 in between MC servo board output to TTFSS, trying to break the possible ground loop between 1X2 rack and PSL.
 - To do this, we had to lower IN1 gain to -6dB, to avoid saturation of SR560.
 - This didn't make any difference in MC_F or POY during YARM lock.

[Anchal, Yuta]

Yesterday, we have measured the frequency noise of PSL with IMC locked/unlocked using ALS BEATX/Y to narrow down where the 60 Hz is coming from.
All the measurements so far is consistent with a hypothesis that 60 Hz noise injected after MC_F is picked-off (it could be from MC_F DAQ readout or something in the IMC loop).

Method:
 - Measured YARM noise spectra when YARM is locked with POY11 to measure the frequency noise with respect to YARM, and compared with MC_F
 - Measured ALS BEATX and BEATY spectra when PSL is free running and when IMC is locked. Here, when "PSL is free running" is done with PSL shutter closed, but all the cables remained the same and FSS loop was in "down" state. Shutters at both ends were closed, and PZT inputs to AUX lasers were terminated to avoid noise injection from PDH locking with dark noise (this was necessary to reduce noise in BEATY).

Result:
 - Attachment #1 is YARM noise calibrated into Hz, and Attachment #2 is BEATX and BEATY spectra with PSL free running (solid lines) and IMC locked (dotted lines). Below are summary of noise level at 60 Hz (RMS is calculated using a bandwidth of 0.187493 Hz)

YARM (PSL locked vs Yarm): 6.5e2 Hz/rtHz (2.8e2 Hz RMS)
MC_F (sum of noises in IMC loop): 4.9e4 Hz/rtHz (2.2e4 Hz RMS)
BEATX free (PSL free vs Xend free): 3.3e3 Hz/rtHz (1.4e3 Hz RMS)
BEATX locked (PSL locked vs Xend free): 8.8e2 Hz/rtHz (3.8e2 Hz RMS)
BEATY free (PSL free vs Yend free): 1.6e4 Hz/rtHz (6.9e3 Hz RMS)
BEATY locked (PSL locked vs Yend free): 1.5e4 Hz/rtHz (6.5e3 Hz RMS)

Discussion:
 - "BEATX locked" measurement suggests that PSL locked to IMC (and Xend free) has noise less than 3.8e2 Hz RMS. This is roughly consistent with YARM measurement of frequency noise, and suggests that Yarm is stable enough to measure the PSL locked frequency noise.
 - "BEATX free" measurement suggests that PSL free run (with cables connected) has noise of 1.4e3 Hz RMS (note that Xend free is less than 3.8e2 Hz RMS).
 - MC_F measurement is the sum of noises in IMC loop, including IMC length noise + noise injected at error point (3.8e2 Hz RMS), PSL free run noise (1.4e3 Hz RMS), noise injected at feedback. Therefore, this suggests that 2.2e4 Hz RMS we see in MC_F is from noise injected after MC_F pickoff point (or in the MC_F DAQ readout).
 - BEATY having large 60 Hz noise probably comes from noise in the beat measurement.

Next:
 - Use BEATX to monitor 60 Hz noise.
 - Try terminating PZT input to see if 60 Hz noise reduces. Try different gains at different point of MC servo board and TTFSS when IMC is unlocked to see where exactly 60 Hz noise is coming from.

[Anchal, Yuta]

Measurements yesterday (40m/17458) suggested that 60 Hz noise is injected after MC_F is picked-off.
So, we terminated PSL PZT input at several points to see where 60 Hz noise is injected.
It seems like the 60 Hz frequncy noise we see in MC_F is from TTFSS box, but the 60 Hz noise we see in YARM is not limited by this.

The 60 Hz noise we see in YARM is probably limited by IMC length noise.

Method:
 - We terminated PZT input to the PSL laser at various points one by one and monitored 60 Hz frequency noise using BEATX. PSL shutter was closed and IMC was not locked.

Result:
 - Below is the result at 60 Hz (RMS is calculated using a bandwidth of 0.187493 Hz)

Reference from 40m/17458, YARM (PSL locked vs Yarm): 6.5e2 Hz/rtHz (2.8e2 Hz RMS)
Reference from 40m/17458, MC_F (sum of noises in IMC loop): 4.9e4 Hz/rtHz (2.2e4 Hz RMS)
MC_F when PSL shutter is closed but MC servo board configuration at IMC locked state: 2.7e2 Hz/rtHz (1.2e2 Hz RMS) -- this gives IMC loop gain enhanced sensing noise

BEATX free (PSL free vs Xend free): 3.5e3 Hz/rtHz (1.5e3 Hz RMS) -- consistent with previous measurements
With PZT input to NPRO terminated (Attachment #1): 8.1e2 Hz/rtHz (3.5e2 Hz RMS)
Connected a terminated small box (we see in Attachment #1) before NPRO PZT: 6.3e2 Hz/rtHz (2.7e2 Hz RMS)
Connected input terminated Thorlabs PZT driver (MDT694): 5.9e2 Hz/rtHz (2.6e2 Hz RMS)
Connected input terminated summing amp (Attachment#2): 4.4e2 Hz/rtHz (1.9e2 Hz RMS)
Connected input terminated TTFSS (C1:PSL-FSS_MGAIN=-10dB, C1:PSL-FSS_FASTGAIN=-10dB): 3.9e3 Hz/rtHz (1.7e3 Hz RMS) -- consistent with "BEATX free (PSL free vs Xend free)" measurement
Connected input terminated TTFSS (C1:PSL-FSS_MGAIN=-10dB, C1:PSL-FSS_FASTGAIN=+10dB): 7.6e3 Hz/rtHz (3.3e3 Hz RMS)
Connected input terminated TTFSS (C1:PSL-FSS_MGAIN=+4dB, C1:PSL-FSS_FASTGAIN=+19dB): 2.9e4 Hz/rtHz (1.3e4 Hz RMS) -- Nominal gains when IMC is locked; consistent with "MC_F" measurement 40m/17458
Connected input terminated TTFSS (C1:PSL-FSS_MGAIN=+19dB, C1:PSL-FSS_FASTGAIN=+4dB): 1.1e4 Hz/rtHz (4.8e3 Hz RMS)

Discussion:
 - Connecting TTFSS increased 60 Hz frequency noise, which suggests that TTFSS is creating this 60 Hz frequency noise.
 - Setting TTFSS gains to nominal gains to IMC locked, 60 Hz frequency noise matched with frequency noise measurement using MC_F. This quantitatively supports that TTFSS is creating this 60 Hz frequency noise.
 - Increasing C1:PSL-FSS_MGAIN and reducing C1:PSL-FSS_FASTGAIN reduced 60 Hz frequency nosie. This means that some portion of 60 Hz noise is from between these two gains.
 - Note that having 60 Hz noise in TTFSS does not necessarily mean that our YARM noise is limited by this, because IMC loop suppresses the TTFSS noise. Assuming all 1.3e4 Hz RMS is all from TTFSS noise, it is suppressed to less than 1.3e4 Hz RMS/2e5 = 6.5e-2 Hz RMS (where 2e5 is IMC loop gain without super boosts, but it is actually higher with them) as frequency noise we see in YARM. YARM noise is measured to be 6.5e2 Hz/rtHz (2.8e2 Hz RMS), so it is not limited by TTFSS noise.
 - Also dark noise measured at MC_F (1.2e2 Hz RMS) tells you that the dark noise is not limiting the frequency noise we see in YARM.

Touching various parts around TTFSS:
 - We moved on to touch various parts around TTFSS to see if 60 Hz noise reduces in MC_F. We removed unused cables around TTFSS interface, touched power cables into TTFSS (both at TTFSS interface in the rack and TTFSS box on PSL table), BNC cables into TTFSS, disconnected slow controls, tried to avoid grounding of cables going into EOM (there is a small box that sums FSS feedback signal and 33.5 MHz; Attachment #3), but 60 Hz noise we see in MC_F didn't change significantly.

Next:
 - Check grounding situation around TTFSS box.
 - Check IMC length noise and error point noise by monitoring BEATX.
 - Check coil drivers for MC1, MC2, MC3 by disconnecting drivers while IMC is locked.
 - Try feeding back IMC servo also to MC2 with 60 Hz resonant gain to cancel 60 Hz noise


Note added at 23:50 to clarify:
nIMC : IMC length noise in frequency
nPSL: PSL free run noise in frequency
ne: sensing noise in frequency
nf: feedback noise in frequency
G: IMC loop gain (estimated to be 2e5 at 60 Hz without boosts)

MC_F = G/(1+G) * (nIMC + nPSL + ne + nf) + [noises in MC_F DAQ]
 = 2.2e4 Hz RMS
 MC_F when dark, MC servo nominal gain = G * ne
 = 1.2e2 Hz RMS
PSL frequency noise after IMC lock = G/(1+G) * (nIMC + ne) + 1/(1+G) * (nPSL + nf)
YARM = [PSL frequency noise after IMC lock] + [noises from YARM loop]
 = 2.8e2 Hz RMS
BEATX when PSL is free run, TTFSS low gain connected = nPSL + [noises from Xend AUX and BEATX sensing]
 = 1.7e3 Hz RMS
BEATX when PSL is free run, TTFSS nominal gain connected = nPSL + nf + [noises from Xend AUX and BEATX sensing]
 = 2.9e4 Hz RMS
BEATX when IMC is locked = [PSL frequency noise after IMC lock] + [noises from Xend AUX and BEATX sensing]
 = 3.8e2 Hz RMS
 
So, our estimate is
ne ~ 1.2e2/G Hz RMS (small)
nPSL ~ 1.7e3 Hz RMS
nf ~ 2e4 Hz RMS (this dominates MC_F, but already suppressed enough in [PSL frequency noise after IMC lock])
[PSL frequency after IMC lock] ~ 3e2 Hz RMS (this dominates YARM and BEATX when IMC is locked)
nIMC ~ 3e2 Hz RMS (this dominates [PSL frequency noise after IMC lock])