Displaying report 1-1 of 1.
Reports until 16:59, Wednesday 05 June 2013
H1 IOO
kiwamu.izumi@LIGO.ORG - posted 16:59, Wednesday 05 June 2013 - last comment - 12:53, Thursday 06 June 2013(6643)
IMC WFS loop closed part II

[Chris, Kiwamu]

This is a detailed version of the previous IMC WFS entry (see alog 6633).

 

Actuation issue wasn't really an issue:

 We kept driving the M3 stages (very bottom stage) of MC1 and MC3 to excite the angular motions (alog 6610) but it turned out that this was not the right way. The actuators on M3 are simply not strong enough to excite them with a high signal-to-noise ratio at the wave front sensors. Therefore we need o switch the driving point from the M3 to M1 (very upper stage) to have a big excitation. In fact this is the way the people in Livingston did. After all, exciting M1 gave us a prominent peak in the spectrum of the WFS signals and hence we were able to proceed with the measurement of the sensing matrix.

 

Balancing of the quadrant segment gain:

During adjusting the demod phase of each segment of the WFSs we noticed that they show different signal gain. In order to correct this gain discrepancy among the segments we changed the digital filter gain to make them identical. We excited the longitudinal motion of the IMC by injecting a 6 Hz sinusoidal signal at INPUT2 of the IMC common mode board servo with an amplitude of 80 mVp-p. This resulted in a peak in the spectrum of each quadrant. Then we adjusted IMC-WFS_A(B)_IX_GAIN and IMC-WFS_A(B)_IX_GAIN where X = 1,2,3,4. Here is the resultant gain coefficients. We are not sure why the gains vary by a factor of 5 at most among the segments. This variation can not be explained by the discrepancy in the transfer gain of the demodulation chain.

H1:IMC-WFS_A_I1_GAIN = 1
H1:IMC-WFS_A_Q1_GAIN = 1
H1:IMC-WFS_A_I2_GAIN = 5.33
H1:IMC-WFS_A_Q2_GAIN = 5.33
H1:IMC-WFS_A_I3_GAIN = 4.43
H1:IMC-WFS_A_Q3_GAIN = 4.43
H1:IMC-WFS_A_I4_GAIN = 1.33
H1:IMC-WFS_A_Q4_GAIN = 1.33
H1:IMC-WFS_B_I1_GAIN = 5.75
H1:IMC-WFS_B_Q1_GAIN = 5.75
H1:IMC-WFS_B_I2_GAIN = 1
H1:IMC-WFS_B_Q2_GAIN = 1
H1:IMC-WFS_B_I3_GAIN = 1.07
H1:IMC-WFS_B_Q3_GAIN = 1.07
H1:IMC-WFS_B_I4_GAIN = 5.47
H1:IMC-WFS_B_Q4_GAIN = 5.47

 

Compensation for the geometrical rotation:

As far as we remember there is a periscope in the IMC-REFL path which rotates the x-y relation of the beam, which rotates it by almost 90 degrees. Due to this effect the actual yaw motion happening in the IMC cavity shows up as almost pitch motion and vice versa at the RFFL port. We corrected this effect by manipulating the input matrix (IMC-WFS_A(B)_I_MTRIX) such that the yaw excitation shows up only in pitch basis after the input matrix. This adjustment was done by exciting the M2 stage of MC2 (which somehow we ended up with for some reason when exciting the angular motions) in pitch and yaw at 3 Hz. As a result the coupling between pitch and yaw were improved and separation ratio became approximately a factor of 50 at least which used to be 1 at worst. Here is the resultant matrix elements.


H1:IMC-WFS_A_I_MTRX_1_1 = 0.478812
H1:IMC-WFS_A_I_MTRX_1_2 = -1.33069
H1:IMC-WFS_A_I_MTRX_1_3 = -0.478812
H1:IMC-WFS_A_I_MTRX_1_4 = 1.33069
H1:IMC-WFS_A_I_MTRX_2_1 = 1.33069
H1:IMC-WFS_A_I_MTRX_2_2 = 0.478812
H1:IMC-WFS_A_I_MTRX_2_3 = -1.33069
H1:IMC-WFS_A_I_MTRX_2_4 = -0.478812
H1:IMC-WFS_B_I_MTRX_1_1 = 0.394297
H1:IMC-WFS_B_I_MTRX_1_2 = -1.35813
H1:IMC-WFS_B_I_MTRX_1_3 = -0.394297
H1:IMC-WFS_B_I_MTRX_1_4 = 1.35813
H1:IMC-WFS_B_I_MTRX_2_1 = 1.35813
H1:IMC-WFS_B_I_MTRX_2_2 = 0.394297
H1:IMC-WFS_B_I_MTRX_2_3 = -1.35813
H1:IMC-WFS_B_I_MTRX_2_4 = -0.394297
 

Measurement of the sensing matrix

The measurement of the sensing matrix went very smooth. We excited the M1 stage of the IMC suspensions. We considered two DOFs in pitch (yaw), namely common (differential) MC1-MC3 and MC2. This is the DOF basis that the Livingston people came up with in the IMC WFS commissioning. The excitation was injected to DOF_1(2)_Y(P)_EXC with an amplitude of 10-12 counts at about 3 Hz. To be sure that we are exciting the pure pitch or yaw we looked at the witness channels for the bottom stage and confirmed that we were exciting the right angular motion. Since we didn't have an auto-script to measure the sensing matrix we used diaggui to measure the transfer coefficient from the excitation to the WFSs. After measuring the 2x2 sensing matrix we inverted it and put it to the control input matrix IMC-INMATRIX_P(Y). In diaggui the frequency resolution was set to 0.01 Hz. Here is the resultant input matrix.

H1:IMC-INMATRIX_P_1_1 = 0.64725
H1:IMC-INMATRIX_P_1_2 = -0.35275
H1:IMC-INMATRIX_P_2_1 = -0.159639
H1:IMC-INMATRIX_P_2_2 = 0.840361

H1:IMC-INMATRIX_Y_1_1 = -0.163134
H1:IMC-INMATRIX_Y_1_2 = -0.836866
H1:IMC-INMATRIX_Y_2_1 = 0.529564
H1:IMC-INMATRIX_Y_2_2 = 0.470436


Closing the loops:

The trial of closing the loops were relatively easier compared with what we did to reach this point. We simply started from a small gain of 1e-4 and we fiddled with the sign. Since this was the first time for us to close the loops we enabled only fundamental control filers, namely FM6 (0.06, 0.1) and FM7(0:1) of IMC-DOF_1(2)_P(Y) and disabled the rest of the fancy filters. We aimed an UGF of  a few 10 mHz or so. The gains were adjusted by looking at the time scale of the control against a disturbance. The gains we ended with were :

H1:IMC-DOF_1_P_GAIN = -1e-05
H1:IMC-DOF_2_P_GAIN = -1e-05
H1:IMC-DOF_1_Y_GAIN = -1e-05
H1:IMC-DOF_2_Y_GAIN = -1e-05
 

What are missing ?:

Comments related to this report
christopher.wipf@LIGO.ORG - 12:53, Thursday 06 June 2013 (6653)
Link 

The DOF3 (MC2 TRANS centering) loops have been turned on, currently with very low UGF (step response ~minutes).  They worked "out of the box" using the LLO filters, and a bit of tweaking of the DOF3 gains.

Under $USERAPPS_DIR/ioo/h1/scripts/imc there's a new script, mcwfsrelieve, which offloads the DOF1/2/3 outputs onto the downstream filter modules (H1:IMC-MC1_PIT et al).

There's also a rudimentary_autolocker script for the IMC (temporary stopgap until the Guardian is up and running here).

Though we have plenty of optimization left to do, the IMC now has a reproducible baseline configuration, which is its main requirement during HIFO-Y.
Displaying report 1-1 of 1.