It converts a Pt100 temperature sensor resistance into degrees Celsius using a lookup table taken from empirical data in the DIN 43760 / IEC 751 document. This library's conversion accuracy is authoritative such that other purely computational methods may be validated against it.
Because executing a Callendar-Van Dusen temperature calculation in software floating point is computationally expensive on 8-bit Arduinos.
Pt100 sensors can have uncalibrated accuracy which often exceeds that of the measurement hardware and firmware. Now that these sensors are affordably manufactured at "1/10 DIN" accuracy for the 0-100C range, the firmware should match them at least minimally.
It is intended for 8-bit MCUs lacking SRAM and speed, namely the one used in the Arduino UNO and similar.
Consuming ~3kB of Arduino program memory, this Pt100rtd library is larger than any collection of computational methods that might be used instead. For any ordinary temperature between -60C and 650C, the venerable Callendar -Van Dusen equation works well. Gas liquefaction enthusiasts, however, have different requirements.
If the hardware interface has insufficient resolution, an inaccurate reference resistance (0.05% is only a start) or too high an excitation current through the RTD, precise conversion can't correct inaccurate data acquisition.
A high excitation current causes RTD self-heating which militates against accurate measurement. Self-heating can be corrected for only if the measurement conditions are known beforehand . . . which they generally aren't.
Generally speaking, self-heating error is mitigated by liquid immersion thermometry but reading an accurate air or gas temperature is more problematic.
The Pt100 resistance lookup table uses unsigned 16-bit integers because:
DIN 43760 Pt100 resistances resolve at 0.01 ohms and are represented in the lookup table as unsigned integer values of (ohms * 100) with no loss of accuracy. Unsigned integers require 2 bytes vs. 4 bytes for a floating point object. Integer arithmetic is also computationally cheaper than software floating point operations, most significantly, numerical comparison.
Even so, at 2100 bytes, the table being too large a global variable for SRAM, it resides in flash program memory with all the special handling that implies, specifically, PROGMEM data type(s) and use of pgm_read_word_near() to fetch them.
Several computational conversion methods are included for comparison: Callendar-Van Dusen (aka 'quadratic'), cubic, polynomial, and rational polynomial. These functions are pedagogical and should be commented out eventually to save space.
If ported to a CPU with more SRAM and a floating point unit, (viz., ARM M0 or better) these defs will certainly help:
#define PROGMEM /**/
#define pgm_read_word_near((x)) ((uint16_t)(*(x)))It has been tested and and found suitable for the Adafruit Pt100 RTD Breakout w/MAX31865 although any Arduino hardware+software mix that produces a conformant Pt100 RTD output in ohms may use the library.
Written by drhaney for his own selfish purposes under BSD license. All text above must be included in any redistribution. 12/14/2017