Everything got back. Now, I feel much better and I could some IFO experiment.
The detail is as follows:
------------
I put excitation signal at MN_LOCK filter at 0.2 Hz with the amplitude of 10000 or more and tweaked the non-diagonal component in the drive align matrix so that the QPD don't have a signal at 0.2 Hz by looking the time series. I think the yaw could be diagonalized very well, but the pitch is not really. Also, I re-diagonalized yaw oplev signal by that excitation signal.