It has been observed that the servo for the set point of DIFFp tx and ty is not active at low noise 3 (to minimize the B1p DC) and this has always had an impact on the structure on the detection bench caused by diffused ligth.

For this reason I' ve analyzed the last 20 days of LN3 locks and I' ve taken sets of 600 sec in which the correlation between the DIFFp TX and TY inputs and B1p DC are evaluated with a linear fit.

It is visible (for the tx dof) that the higher correlation (higher slope of the fit, which correspond to the set point far from the optimal one) corresponds to a larger amplitude of the peaks (in DARM).

This analysis is to stress the fact that we need to re-engage the servo @ LN3.