Expand description
The PPS link: one transaction per 25 ms slot, so a tick never skips a beat.
The module NACKs a transaction addressed right after another one ended, so
no two transactions ever run back to back. SCHEDULE gives each 25 ms
slot one purpose; per 100 ms tick:
| Slot | Purpose | Transaction |
|---|---|---|
| 0 | Slot::Command | write the setpoints (or ENABLE = 0, if a disable is due) |
| 1 | Slot::Read | read the bank 0x04..=0x1F |
| 2 | Slot::Enable | write ENABLE, only if the bank disagrees with the setpoint |
| 3 | Slot::Read | read the bank |
So a command reaches the module within one tick, a reading is never older
than 50 ms, and every write is verified by the next read. The spacing is
enforced, not assumed: a stalled executor skips slots (next_slot_start)
instead of running them back to back, and the board’s power read, which
shares the bus, takes a slot of its own once a second (Slot::Aux). The output is only
enabled once the bank shows the commanded setpoints: the module applies them
whether or not it is enabled, so enabling first would run the old ones.
The loop never stops. A failing link is reported as an episode (first error, then once a second, then recovery); a module that never answered is probed once a second rather than at the slot rate.
Re-exports§
pub use crate::driver::pps::PpsError;pub use crate::driver::pps::PpsRunningMode;pub use crate::io::pps_plan::PPS_SLOT_MS;pub use crate::io::pps_plan::PPS_TICK_MS;pub use crate::io::pps_plan::PpsSetpoint;pub use crate::io::pps_plan::SCHEDULE;pub use crate::io::pps_plan::Slot;
Structs§
- PpsReadings
- One bank read, as the consumer sees it.
- PpsResources