Observation
We noticed during beam commissioning that trims did not reach the FGC in time to be played the next cycle. This caused a ping-ponging behaviour in the trims where the trims were lagging behind by one cycle too many.
Outcome
We found through scripted approaches by trimming +/- 5E-5 T that we need to trim roughly 1.7s before cycle start for the function to be played in the MBs. The sps-app-hysteresis trims 1.7s +/- 0.2s before cycle start, where inference takes roughly 0.4s for SFTPRO before that, with other overhead likely provided by Python.
Action
Implementation of a custom forewarning based on the supercycle structure to be able to predict one full cycle ahead of time in Cycle forewarning