Using Fabian's calibration data in 17898, I have set calibration factors at K1:VIS-SR3_TM_OPLEV_{TILT, LEN}_{HOR, VER}_GAIN. These converts from, in Fabian's convention, displacement units to the beam spot displacements at the QPD planes (μm). See Fig. 1.

Using old data from 7781, I derived a sensing matrix from beam spot displacements to optics' longitudinal, pitch, and yaw displacements. See Fig. 2.

I measured all transfer functions and derived a decoupling matrix shown in Fig.3.

Decoupled transfer functions are shown in Fig. 4, 5, and 6.

Fig. 7 is a 10 second average of the IP LVDT and TM OPLEV readout when I offset the IP in longitudinal. This shows that pitch readout is decoupled from longitudinal displacement to a better extent, which is not observable in the transfer functions. 

One thing to note is that there's severe L2T actuation coupling at the IP stage.

All measurements files are at /kagra/Dropbox/Subsystems/VIS/vis_commissioning/sr3/sensor_diagonalization

Sensing matrix and SENSALIGN matrices are available at /kagra/Dropbox/Subsystems/VIS/vis_commissioning/sr3/notebook/oplev\ sensing\ matrix.ipynb