Hugh R, Jess M
As of 00:53:00 UTC July 15, the ETMY chamber was in the following configuration:
- HEPI in robust_isolated
- Stage 1 in level 2 isolation: gain of 0.8 in Y, gain of 0.6 in X, TBetter on all DOFs except TCrappy on RZ
- Stage 2 in level 2 isolation: gain of 0.5 on all DOFs, Start blend on all DOFs
Please leave ETMY in level 2 isolation until 8am tomorrow (if it hasn't tripped by then)
Today Hugh and I spent the afternoon trying to get EMTY stably into level 2 isolation. For stage 1, Hugh brought up the DOFs one by one, and we discovered a giant peak in the actuator output signal (and the L4Cs) in all DOFs at 738Hz as the X DOF gain approached 1. (Giant = most of the actuator power spikes at that high frequency right before the ST1 actuators trips.) We're not sure what might cause this, but it's very reproducible and we plan to look into it later this week.
We noticed a small peak in the open loop transfer function at about the same frequency, and the level 2 isolation filter gain is much higher than levels 1 or 3 in this very high frequency range. (See attached.)
As for stage 2, we had no trouble using the control script to bring it up to level 2 with the Start blend, but sometimes it would trip on the actuators shortly after and we couldn't switch the blend for any DOF without tripping (also on the actuators).
I attach the .pdf of the loop design for reference. In doing so I noticed two things: - Why are we allowing gain peaking of 4+? I thought we'd agreed never to design a loop with gain peaking of more than 2ish? - The prediction of the minimum phase margin is busted. I was about to be saddened by a phase margin of 20.8 [deg] (again, I thought we'd agreed that no less than 35 to 40 [deg] phase margin is acceptable) BUT it turns out the phase margin is 41.9 [deg] at the upper unity gain frequency. So, at least the loop is still stable, but hopefully the designer didn't accept the phase margin of 21 [deg] as stable and move on... Regarding the 738 [Hz] resonance -- what's our high-frequency roll-off policy? Seems like still having high-frequency features at a loop gain of 0.2 is treacherous, especially when the feature is close enough to the Nyquist frequency that the *measurement* amplitude (perhaps not the real amplitude) falls off deceivingly because of the DAQ's down sampling filter. Can we not just toss in pole at ~300 [Hz] where it would cost us little in phase margin, but alleviate the worries of this high-frequency non-sense? I think perhaps, now that the era of "OMG we need to get every platform commissioned yesterday!" is over, we should consider dropping these automated scripts and going back to looking at the loop design in detail, understanding the compromises we've made there, and making sure it all makes sense.
