Following up on 47176, Craig did a series of MICH noise injections at different times during the lock, to check if the MICH to DARM transfer function is changing. He switched off the MICHFF for each measurement, so what we measure is the entire MICH to DARM coupling (not the residual after feed-forward).

The suspicion was that the coupling changes during thermalization. This is confirmed. The plot below shows the measured coupling at different times, reported as minutes elapsed after powering up to 30 W.

Although at first inspection the four measurements seem quite similar, if we compute the ratio of each measurement over the first one (used as reference), we see that there are changes of the order of 4% in amplitude and 2.5 degrees in phase. This is most likely a limitation on the performance of the MICH feed-forward cancelation.