Summary--Looking closer shows that the disturbance seen on the dofs are occurring before the loops are driving; what the hay?  GS13 switching very likely the thing.

Details: Hugo gave me some changes for the guardian and it does as expected but that did not correct the problem.  There was some suspicion that Guardian may have been doing things at the same time and I think these result bear this out but I ignored one thing I should have ruled out first.

The first attachment shows a manual isolation turn on from Tuesday.  This is a zoom in from alog 17042.  MICH is NOT on.  First I use the Reset CPS offset button on the medm which gets 12 samples over 3 seconds for an average (T=1).  It then loads a ramp time of 5 seconds, puts the average values determined into the SETPOINT_NOW, and, the model takes over from there loading the SETPOINT_NOW into the Current Setpoint (BIAS_RAMPMON) over the prescribed 5 seconds.  The next step is turning on the Z Isolation Loop.  It is only the vertical we are engaging and I do this with the Command script (with boost, T=2.)  Notice that nothing shows up on the RZ_LOCATIONMON until T=3 when I start the RZ loop.  This all looks reasonable.

The second attachment is a few minutes later using the guardian to isolate state2, MICH is still NOT in the picture.  T=1 shows that Guardian does not use a TRAMP and the front end puts it in with no delay.  Theoretically, this should not be any problem as again, the loops are not on yet.  However, at T=2, RZ and Z channels start to show motion, even though the loop switches are on until T=3, about 5 seconds later!  T=2 is hard to pinpoint but it is looks to be 2 or 3 seconds after T=1. 

The 3rd attachment is from this afternoon and now MICH is Dark Locked.  I've modified the Guardian to do a 4 second ramp of the Setpoint to RAMPMON.  You can see this as the RESIDUALMON goes to zero as well as the RAMPMON.  Then, 1 or 2 seconds after the RAMPMON is finished, the RZ starts moving, again, all before the Z-loops engage.  With Mich though, the lock breaks (see Watchdog upper right) and who knows after that.

What is the Guardian doing before this... aha!  The switching of the GS13 is being done by the guardian and although it would not impact the CPS_Z_LOCATIONMON, it may glitch the damping loop enough.  Yup, there it is.  The last plot I've added the GS13 SWSTAT and it corresponds to the time the RZ_LOCATION starts to drive off.

Until the Isolation loops are adjusted to work with the MICH LOCK, we should keep the GS13 gains in whitened low gain.  And, unless we make the gain switching even smoother, there needs to be more settling time after the gain switching when we do switch.