Thermocouple types and tolerance classes (IEC 60584)

A thermocouple type is a promise about one thing: the voltage its junction produces at a given temperature, against a reference junction at 0 °C. IEC 60584-1 defines that curve for each letter type and then states how far a given piece of wire may deviate from it. Choosing a type is choosing a range, an output per degree, and a tolerance band — in that order of importance.

The five common base-metal types

Output below is the Seebeck coefficient at 20 °C, computed from the NIST ITS-90 reference functions, with what one microvolt of instrument error is worth in millikelvin at that point. That last column is the one that decides whether your readout is good enough.

Types K and T are almost indistinguishable in output at 20 °C — 40.33 against 40.27 µV/°C — and then diverge quickly: by 100 °C type T is producing 46.8 µV/°C against type K's 41.4. It is a reminder that a type's sensitivity is a curve, not a constant, and that a rule of thumb picked up at room temperature does not travel.

Tolerance classes

IEC 60584-1 states each class as the greater of a fixed value and a proportion of the temperature, over a stated range. For types K and N: class 1 is ±1.5 °C or ±0.004|t|; class 2 is ±2.5 °C or ±0.0075|t|; class 3 is ±2.5 °C or ±0.015|t| and covers sub-zero use only. The fixed term wins until the proportional term overtakes it — at 375 °C for class 1 and 333 °C for class 2.

Two things fall out of that table. Type T's class 1 band is three times tighter than everyone else's across the cold-chain range, which is why it is the default for freezers, cold rooms and transport qualification. And type K's class 1 tolerance at 1000 °C is ±4 °C — a number worth quoting to anyone who believes a thermocouple is an accurate device.

The noble-metal types

• Types R and S (Pt–PtRh) reach 1600 °C and are the laboratory reference types above the range of base-metal wire. Output is low — around 6 µV/°C at 100 °C — so they demand a good readout.
• Type B (PtRh–PtRh) reaches 1700 °C and produces almost nothing below 50 °C, which conveniently makes it insensitive to the temperature of its own reference junction.
• All three are expensive, and all three are poisoned by contact with metal vapour or reducing atmospheres. They are not a general-purpose upgrade; they are a high-temperature necessity.

What the tolerance class does not tell you

The classes apply to new wire, as delivered. A thermocouple used in a steep temperature gradient develops inhomogeneity — the composition of the wire changes along its length, and the emf it produces starts to depend on where the gradient sits rather than only on the junction temperature. Move the same thermocouple to a different immersion depth and it reads differently. No calibration corrects this, because the calibration was performed with the gradient in one place and your process puts it somewhere else.

That is the fundamental reason a platinum resistance thermometer beats a thermocouple for precise work in the range where both are available. If you need better than about a degree below 400 °C, the question is not which thermocouple type — it is whether a thermocouple is the right instrument at all.

Tolerance class is not measurement uncertainty. A class 1 type K is not a ±1.5 °C measurement until you have accounted for the reference junction, the extension leads, the readout, and the state of the wire. Calibrate the thermocouple with its own instrument and use the certificate instead.