As happens quite often, this morning the lock acquisition failed many times at the beginning of CARM NULL (fig 1). The typical behaviour is a large fluctuation of the dark fringe, sometime becoming too large. WI and WE optical levers show some unusual rotation in TX, but apparently too slow for being able to have an impact on DIFFp accuracy (fig 2). Moreove, this behavior of the optical levers occurs only in this particular moment of the lock acquisition; normally, when the lock is acquired and lasts hours, no SOFT loop malfunction or sensing glitch is visible.

I think the problem comes from DIFFp TX sensing. During the crisis, the accuracy of the in-loop signal looks good enough and cannot explain the fluctuation of the datk fringe, but in the quadrature (and also in the other B1p QD) a large noise is visible (fig 3).

Using the optical levers to reconstruct DIFFp TX, one can notice an oscillation well correlated to B1p fluctuation (fig 4). The amplitude is about 0.1 urad peak to peak, at the limit of OptLev sensitivity, but very large for DIFFp, and much larger that DIFFp_TX in loop. Sensing noise of this loop seems a good explanation of that behaviour.