First changed the mixing of B1s2 and B1_PD2 for the OMC1 error signal (the SDB1_B1x channels), as the ratio of power has now changed from 1/6 to 1/12.
This didn't have an impact.
Then noticed that the when OMC2 goes through the resonance the total power between B1s2 and B1 is not kept constant.
This is because SDB2_B1_DC has been for the past year the average of the two B1 PD powers, and with the current balance weight of 0.0 for B1_PD1 and 1.0 for B1_PD2 it actually corresponds to the total power on the B1 beam (so double the power in the situation with the 50/50 beam splitter).
Tried to divide the darm_b1dc_weight by a factor 2, it kept the total B1s2 and B1 power constant, but the ITF still unlocks.
To make it a bit cleaner (in case SDB2_B1_DC is used somewhere else), changed the value of the SDB2_ONE constant from 1.0 to 0.5, in this way SDB2_B1_DC = 0*B1_PD1 + 0.5*B1_PD2.
This didn't solve the problem either.
Figure 1 summarizes the situation when OMC2 goes through the resonance. The total B1s2 and B1 power is constant 12*B1s2 + B1_PD2 = 18mW.
But the B1p_56MHz_I signal is clearly moving away from zero as if the dark fringe offset is increasing.
An explanation could be that the beam is not well aligned on B1_PD1, and that only ~20% of the light reaches the photodiode.
