These four [4] Smath documents illustrate techniques to personalize
linearization of 4 current types of temperature sensors [CY7, T, J, Pt100].
CY7 [semiconductor] ... T, J [ThermoCouple] ... Pt100 [Platinum Resistance].
3 linearization methods are used:
1. Thiele Continued Fraction ... Smath loves
2. Normalized Rational fraction ... Not easy Smath
3. ChebyShev Polynomials ... Smath loves

If you wish so, examples of the 3 methods are available.

Pt100 [-259.3467 ... 961.78] the most accurate [secondary standard]
T [-268.935 ... 356.66] ... J [-210.002 ... 1084.5]
CY7 [0 °K ? ... 25 °K] only 3 eit90 reference points

To the contrary of vendors publication, NONE of the sensors will
be more accurate then 0.01 °C, many eit90 points are only ± 0.01 accurate.
To the contrary of many empirical opinions, Thermocouples [T/C 'T', 'J']
are very accurate if personalized, unfortunately supplied not personalized.

An industrial installation [sometimes X km long] affects the accuracy
compared to sensors embedded in chips. Estimate an industrial installation.
Pt100 is fragile for industrial use. T/C are rugged but may suffer from
heat cycling [Plant ON/OFF maintenance]. You may have to re-personalize.
Platinum that meet eit90 chemical composition need no more consideration.
T & J were extensively considered in NBS Monograph 125.
Their global tabulation is based on high order polynomials [not the best choice].
Equipment suppliers may offer specific range or global range [ $ cost].

For higher temperature, I'm not familiar with pyrometers.

JeanInst_Type Pt100 Copy.sm (58 КиБ) скачан 67 раз(а).
Inst_Type T Copy.sm (52 КиБ) скачан 64 раз(а).
Inst_Type J Copy.sm (111 КиБ) скачан 56 раз(а).
Inst_Type CY7 Copy.sm (342 КиБ) скачан 62 раз(а).