Writing
Notes from close to the hardware
Short write-ups of problems I have run into in embedded software, control and measurement: what went wrong, how it was found, what fixed it.
-
The brown-out I found in a simulator instead of on the bench
A battery climate meter that would have reset itself on half-empty cells, a 27-microsecond dip, and why the fix ended up in hardware.
-
STmin has to hold across block boundaries too
A small ISO-TP timing bug that no unit test would catch, found because a bus simulator made the transfer faster than it was allowed to be.
-
Where a Q31 filter quietly loses 60 dB
Low-cutoff IIR filters in fixed point look fine in a frequency plot and are wrong in practice. Measurements, and the one-line fix that recovers most of it.