[Jamie, Fabrice]
After much initial headaches, we now have guardian nodes running for HAM2 and HAM3 ISIs (node names ISI_HAM2
and ISI_HAM3
respectively):
The above control screens are accessible from the guardian overview screen (blue "GRD" button on the site map).
The guardians are holding the ISIs in the HIGH_ISOLATED state (corresponding to "isolation level 3" in the old parlance). The guardians are watching for watchdog trips, and will automatically bring the ISIs back to the requested state (HIGH_ISOLATED in this case) after the operator resets the watchdog. State graph of the ISI_HAMX module that the nodes are executing is attached.
The current configuration (high GS13 gain, lower blend filter cross over frequency) is not able to survive (de)isolation from/to very large bias offsets. Fabrice and I tweaked the isolation gain ramps such that turning on and off the isolation loops (and the corresponding biases) survives offsets that correspond to an overestimate of the expected extrema of the platform drift (more detailed report to follow). Given that we don't routinely hold large offsets on the HAMs we don't expect this to be too big of a problem. And too be clear, the previous configuration wouldn't survive large offsets either, so we're definitely better off than we were. After testing, we reset the CART_BIAS targets to the equilibrium position ("Reset CPS offsets" && "Store target offsets").
In any event, it's possible that over time the equilibrium position will drift far enough away from the current target that an attempt to move the isolation biases back will cause a WD trip. This would likely happen after a WD trip caused by something else. Resetting the target to the new equilibrium value should make the problem go away.
I'll be posting a more detailed report on the current HAM ISI configuration tomorrow.