Hacker News Reader

Comment by Retr0id | Hacker News Reader

There is presumably still some possibility of them failing in an "always reports 0" way

There are some parts of engineering in safety systems where you have a single thing that could go wrong that would have serious consequences, and the result of the FMEA is that "it has to not do that".

I don't think you'll find a wheel speed sensor without a few "outputs incorrect speed" failure modes.

Detectability is one dimension of an error, and "bad wheel speed" has decent detectability I reckon - either through redundancy, grey codes, index pulse checking, bound checking.

Also the issue experienced in the post wasn't an issue with a sensor per se.

Sure, it is possible theoretically.

However, most relevant regulation (IEC61508, ISO26262, DO-178X) requires that systems controlling machines in automotive, rail or aerospace have a possibility of dangerous faults lower than 10^-9 (over the expected lifespan).

Many critical control systems like this are formally verified and/or extremely well-tested and have redundancy in both software and hardware.

…Like when dismembered and spread across a bench.

It's not unthinkable that whatever transducer takes rotation and turns it into a signal which is processed by an MCU and translated into CAN messages could get stuck producing one signal and trigger such a scenario without the car's control system needing to be on a bench.

Reply on news.ycombinator.com