BSP 3: when the bus voltage falls

One real run: the board's real bsp_api.c compiled on a Mac, the bus falling from 48 V to 0 in 10 ms. The bus has two copies, both refreshed once a millisecond: raw for the PWM scaling, filtered for the controller and the under-voltage check. In the first millisecond the compare value stays 3187, worked out for 47.999 V, while the true bus falls to 43.44 V: U gets 0.818 V, 10 % short, a sawtooth every millisecond. The true bus passes 17 V at 6.5 ms, but protection fires at 9 ms: the next read, then the filter. Below 8 V the compare value freezes at 2733 instead of going to zero; in the 10 ms fall the freeze and the trip arrive together, at 30 ms protection wins. The SIL's bus is 48 V plus noise and shows none of it.