[Stefan, Jenne]
We have been looking at different PD signals to see if we can set up a guardian check that the fast shutter indeed closes upon lockloss. Right now we don't have a good way to code it in guardian yet, so Stefan has made a template for the lockloss script that we can run every lockloss.
Note that AS_C and ASAIR_B are both before the shutter, so should not see a change in light upon shutter close. AS_A, AS_B, and OMC_DCPD are all after the shutter, and should lose light if the shutter successfully closes.
For now, we have some DTT snapshots. NDS2 is under maintenence, so not all plots have the same traces, sorry.
- The first one (2016_08_01_0843utc.png) shows an example of the shutter not closing. Signals are scaled arbitrarily to make them all visible. In this particular case, the wire was probably already broken, because we send the trigger signal requesting the shutter to close, but it doesn't actually close (NB the shutter state is readback by Beckhoff, so the timing of that signal can only be trusted to within ~1 sec or so). All of the signals except for the DCPDs saturate, but it's clear (particularly in comparison to the next 2 plots) that the light doesn't get shuttered. Note that the OMC was already offline at this time, so the OMC DCPD signal isn't super useful here.
- The second plot (2016_03_08_0430utc.png) shows a time from before the April HAM6 vent, so a time when the shutter was definitely working. In particular, note that the AS_A and AS_B sums get cut off at 5.4sec. This is the shutter closing. However, there is some kind of bounce or something, where the shutter allows light through for another 20 msec or so before closing and staying closed.
- The third plot is the most clear version of the shutter working (2016_08_01_0020utc.png), showing again AS_A and AS_B sums getting cut off. You can also see ASAIR gets the true peak of light after the shutter does the 'bounce and re-close' thing.
- The fourth plot (2016_07_22_0745utc.png) shows a time when the shutter's trigger failed, and so the shutter didn't close.
Once NDS2 is back online, we'll post a few more plots of a few more locklosses, with both ASAIR and OMC DCPD included.
Also, we don't want to use signals such as the OMC input pointing QPDs that are whitened for this fast information.
NDS is back. Here are several plots, one where the shutter failed, and 5 where it worked. We see that this "bounce" is very repeatable.
The problem is that the big pulse of power can come before, during or after the bounce, so it is possible that we still send high power light to the AS WFS and the OMC unintentionally. As long as the PZT shutter has worked, the DCPDs will be protected, but the black glass dumps and AS WFS will not be. See in particular the 28July 20:19utc plot for an example when the big pulse comes just after the fast shutter originally closes.
Rich saw this bounce behavior when testing the fast shutters (see T1400335 starting at page 8), so this is a known issue, but we need to fix it.
Optical way of looking every acquisition whether the shutter is moving:
The fast shutter is kept closed through DRMI lock, and only opened when we're ready to start using the AS WFS in the DHARD_WFS state. So, we can look to see that there is some power at the ASAIR PDs, or also the AS_C diode. Initially, there should be no light on the AS WFS or going to the OMC. However, after the shutter is requested to be opened, we should suddenly see signal on those diodes.
Attached are 2 instances of this behavior, so it shouldn't be too hard to write a guardian check to ensure that the fast shutter is moving during every lock acquisition sequence. If instead we see light on the AS WFS as soon as DRMI is locked, and there isn't a sudden change when the shutter is requested open, then we should get warnings so that we know the shutter is perhaps stuck open.
Hopefully the new shutter installation will solve the problem that Jenne mentions where the shutter let light pass for about 20 msec after the original closing. We believe that the previous shutter was too low (9 mm between the beam and the top of the shutter mirror), so that the downward 'bounce' of the shutter was enough to let the beam by. The shutter is now 3 mm higher than it was, so hopefully it will continue blocking the beam even during the downward part of the shuttering cycle.