Why we moved to a 32‑bit floating‑point fuel model
Moving the M1 fuel model from fixed‑point to 32‑bit floating point removed a whole class of scaling bugs — for about 4 µs…
Archive
Author: fupsrl
Bench test: knock detection on the M1 prototype
Validating the wide‑band knock detection pipeline on the M1 prototype against a reference engine — latency, false positives, and one surprising noise…