While trying to get he ISI-BS up and running (to either Level 2 or Level 3) we saw instabillity whenever turning the isolation loops on, and has been so for most of the week.

Judging by the high DC content of the supersensor signals coming out of the Blend filters, we deduced that there was some offset that was (not?) accidentally applied.

The CPS typically have some level of bias voltage, we compensate for this with a digital offset, so the error point of isolation loops -- which are DC coupled -- is small. Those offsets used to be set in the local basis. We are preparing (see E1300548) to roll off a new update to set those offsets in the cartesian basis. The latter way allows to keep track of the platform's alignement over time, in the basis of interest.

Looking more into the ISI-BS screens, we found that offsets were set in the "current" block of the blend filters. Those offsets are the equivalent of the upcoming update's "cartesian biases", and were probably set there to get the benefits of the update earlier. However, some BURT restore had left the local basis offsets on as well.

Having both offsets on at the same time, created a huge DC offset at the input of the isolation loops, which the actuators had a very difficult time pushing to zero, and eventually causing the (or many) watchdog(s) to trip, usually very early in the ST1 turn on process.

Local CPS biases were turned off. It was then possible to turn on Level 3 isolation loops on ISI-BS. 

One can still determine the alignment of the ISI by looking at the input to the CPS input to the blend filters.