Binary32 compatibility
Binary32 is the IEEE 754 single-precision floating-point format commonly
represented by C float on embedded systems. It has less precision than the
binary64 floating point normally used by Python.
rtd-sensor does not simply assume that a single-precision implementation will
be “close enough.” Its binary32_compatible conformance profile is based on
independent single-precision evaluation and explicit acceptance criteria for the
behavior covered by that profile.
Built-in binary32 profile available since: rtd-sensor 0.5.0. Characterized IEC 60751 reference-resistance coverage added in: 0.6.0.
Why this matters to an MCU
A constrained target may want only:
That can be a legitimate subset claim without implementing the complete Python package.
What the claim does not mean
binary32_compatible does not mean:
- every arbitrary user coefficient set is safe in binary32;
- every future model family is automatically covered;
- binary32 and binary64 are bit-identical; or
- the physical sensor uncertainty is determined by floating-point precision.
It means the implementation satisfies the published numerical acceptance for the specific claimed contract subset.