[Komori,Michimura]

We had discrepancy between the actuator efficiency measurement using 3km Michelson and the measurement using laser sensor (klog #1169).
During the investigation, we found that ETMX/Y coil signs were messed up.
We corrected the signs as

K1:VIS-EX_TM_COILOUTF_H1_GAIN =  1
K1:VIS-EX_TM_COILOUTF_H2_GAIN = -1 (was +)
K1:VIS-EX_TM_COILOUTF_H3_GAIN =  0.7
K1:VIS-EX_TM_COILOUTF_H4_GAIN = -0.95 (was +)
K1:VIS-EY_TM_COILOUTF_H1_GAIN =  0.78
K1:VIS-EY_TM_COILOUTF_H2_GAIN = -1 (was +)
K1:VIS-EY_TM_COILOUTF_H3_GAIN =  1
K1:VIS-EY_TM_COILOUTF_H4_GAIN = -0.87 (was +)

This solved the discrepancy, and the new measured values are;

     Michelson           laser sensor
ETMX  5.1 +/- 0.2 um/V    4.7 um/V
ETMY  3.3 +/- 0.1 um/V    3.1 um/V    (all at DC)

Here, DAC factor of 20/2^16 V/counts is assumed (this effects both measurements).

We flipped the signs of the gains for the yaw oplev damping for ETMX and ETMY accordingly.
We also made K1:LSC-MICH_GAIN = 1 (was 10) and updated the MICH_LOCK guardian accordingly.


Why this happened?
 Since ETMX and ETMY have + configuration of coils, we never checked the relative signs between "pitch" coils and "yaw" coils until we have the length sensor, Michelson.
 We should have checked the signs when laser sensor measurement was done.


Revisiting laser sensor measurement
 First of all, I fitted Ohishi-san's actuation efficiency measurement using laser sensor (KEYENCE LK-H022K), with a simple pendulum transfer function.
 The data lives in /users/VIS/TAMAtest. Using the calibration factor of the sensor (10 V / 5 mm), the fitting result gives (ETMXYlength.png);

      res. freq.  Q     DC gain
ETMX   0.92 Hz     5.1   4.7 um/V
ETMY   0.94 Hz     4.6   3.1 um/V

 Ohishi-san carefully did this measurement by flipping the signs of the coils to get the maximum efficiency.

 From the discussion below, I assume the resonant freqeuncy to be these numbers.


Actuator efficiency measurement using 3-km Michelson
 We followed the similar steps by Ando et al.(klog #1169; see also klog #1323).

 1. Aligned IMC to give the IMC transmission K1:IMC-MCL_TRANS_OUT = 0.43.

 2. Aligned Michelson to give the fringe K1:LSC-REFL_PDA1_DC_OUT to be 7200 at max, and 1400 at minimum. This gives the contrast of 67 % and the slope at the midfringe of the (3.4 +/- 0.1)e10 counts/m. Here, meters are in the Michelson one-way differential length (lx-ly).

 3. Locked MICH using only ETMX and measured the openloop transfer function (MICHOLTF_Xonly.png). Compared it with the NoiseBudget model (klog #1323), to get the ETMX actuation efficiency of 5.1 +/- 0.2 um/V at DC.

 4. Locked MICH using only ETMY and measured the openloop transfer function (MICHOLTF_Yonly.png). Compared it with the NoiseBoudget model, to get the ETMY actuation efficiency of 3.3 +/- 0.1 um/V at DC.

 5. Locked MICH using both ETMX and ETMY and confirmed that the measured openloop transfer function match the model (MICHOLTF_XYboth.png).


By the way
 The actuation efficiency when coil signs were messed up were
      Michelson
ETMX   0.84 um/V
ETMY   0.31 um/V  (all at DC)

 This means that the coils are balanced within roughly 10-15 %.


NoiseBudget update
 I have done minor updates on the Michelson NoiseBudget model during the investigation.

 1. Included 15.3 Hz notch filter in the LSC_MICH filter
 2. Updated to use demodulated calibration lines (K1:LSC-LKIN_CAL1_(A|B)_(REAL|IMAG)_OUT_DQ) to adjust the optical gain
 3. Measured PD dark noise (currently limited by ADC noise) and included in the model. Dark noise was measured with PR3 mis-aligned.
 4. Updated the actuator efficiency to use the values above.

 MICHOLTF_calibrated.png is the measured vs model openloop transfer function plot. Calibrated optical gain 3.3e10 counts/m is consistent with the measured optical gain.