The short answer
Pick by temperature first, then by accuracy. Up to a few hundred degrees, a platinum resistance thermometer (RTD, usually a PT100) is the more accurate, more stable and more interchangeable sensor. Above roughly 600 °C, where RTD tolerance classes end, a thermocouple is the only practical choice. A thermocouple also wins where the probe has to be very thin, react quickly or survive heavy vibration.
The table above shows why the line sits where it does. At 100 °C a class A PT100 may be off by ±0.35 °C, while a class 1 type K thermocouple may be off by ±1.5 °C, four times as much. By 600 °C the order has turned round: class A is no longer defined for wire-wound elements, class B allows ±3.30 °C, and a class 1 type K thermocouple allows ±2.40 °C.
How each sensor works
An RTD is a resistor whose value rises with temperature. A PT100 is made of platinum and has exactly 100 Ω at 0 °C and 138.51 Ω at 100 °C. IEC 60751 defines this curve with the Callendar–Van Dusen equation, so every PT100 follows the same characteristic within its tolerance class. The instrument passes a small current through the element, typically 1 mA or less, and measures the voltage. The sensitivity is about 0.39 Ω per kelvin near 0 °C.
A thermocouple is two different metals joined at the measuring point. It produces a small voltage that depends on the temperature difference between that junction and the terminals where the wires meet copper (the cold or reference junction). A type K thermocouple gives 4.096 mV at 100 °C with the reference junction at 0 °C, about 41 µV per kelvin. IEC 60584-1 publishes the reference functions for the eight standard types: K, J, T, E, N, R, S and B.
The difference in signal size explains most of what follows. An RTD is a resistance of around 100 Ω that changes by 0.39 Ω per kelvin and is easy to measure precisely. A thermocouple is a few millivolts that change by a few tens of microvolts per kelvin, and it only means something once the reference junction is known.
Accuracy at the same temperature
IEC 60751 defines four tolerance classes for platinum RTDs:
- AA: ±(0.1 + 0.0017·|t|) °C
- A: ±(0.15 + 0.002·|t|) °C
- B: ±(0.3 + 0.005·|t|) °C
- C: ±(0.6 + 0.01·|t|) °C
Each class applies only within its own temperature range. For wire-wound elements, class A runs from −100 to 450 °C and class B from −196 to 600 °C. Thin-film elements often have narrower ranges, so check the datasheet.
IEC 60584-1 defines classes 1, 2 and 3 for thermocouples. Each is a fixed value or a fraction of the temperature, whichever is larger. For type K, class 1 is ±1.5 °C or ±0.004·|t| from −40 to 1000 °C, and class 2 is ±2.5 °C or ±0.0075·|t| from −40 to 1200 °C.
These tolerances describe a new sensor as delivered. The measurement as a whole adds the wiring, the reference junction, the transmitter input and the effect of ambient temperature on the electronics. For a thermocouple, the reference junction error goes almost one to one into the reading near room temperature. A transmitter whose cold-junction compensation is good to ±0.5 °C therefore adds about ±0.5 °C to every reading. The temperature measurement error calculator combines all of these terms; the links above open it with a class A PT100 and with a class 1 type K thermocouple at 100 °C for comparison.
Temperature range
IEC 60751 covers platinum RTDs from −200 to 850 °C, but the tolerance classes stop well before the top: class B ends at 600 °C for wire-wound elements. Above that, the platinum and its insulation drift too much for an RTD to stay within a class.
Thermocouples cover far more. Type K works up to 1200 °C within class 2, and its reference function runs to 1372 °C. Types R and S reach about 1600 °C in continuous use, and type B reaches about 1700 °C. At the cold end, type T is the usual choice below zero, with a class 3 tolerance down to −200 °C. The thermocouple types pages list the ranges and limits of each.
Stability and drift
A platinum RTD in a good sheath drifts very little in service, which is why platinum thermometers are the reference for calibration laboratories.
Base-metal thermocouples age. Type K has two well-known mechanisms:
- Short-range ordering: a reversible change in the positive leg between roughly 250 and 600 °C that shifts the reading by a few degrees.
- Preferential oxidation of chromium (“green rot”): this happens in low-oxygen atmospheres at high temperature.
Type N was developed to reduce both and holds its calibration better between about 800 and 1200 °C.
Drift matters more than the initial tolerance in a process that runs for years without recalibration. A thermocouple that started within class 1 can be outside class 2 after long service at high temperature, and nothing in the signal shows it.
Wiring and signal
An RTD is connected with two, three or four wires. In a two-wire circuit, the resistance of both leads adds to the reading. Ten metres of 0.5 mm² copper adds 0.69 Ω, which reads as about +1.8 °C on a PT100 (see the lead-wire error calculator). A three-wire circuit cancels the leads if they are equal. A four-wire circuit removes them entirely.
A thermocouple must be extended with cable of the same type, or with matching compensating cable, all the way to the instrument. Copper cable in between creates a new junction at the terminals and moves the reference point. The millivolt signal is also more exposed to electrical noise than an RTD’s. Over long distances, both sensors are usually connected to a head-mounted transmitter that sends 4–20 mA instead.
Response time, size and robustness
A thermocouple junction can be tiny. Mineral-insulated thermocouples are made with sheaths well under a millimetre in diameter, and a grounded junction touches the sheath directly. That gives fast response and lets the probe go into small spaces. An RTD element has a finite length and needs a supported resistor, so RTD probes are usually larger and slower.
The thermocouple junction is also mechanically simple, which makes it the safer choice under strong vibration or shock. Wire-wound RTD elements are more delicate. Thin-film elements are sturdier but have narrower temperature ranges.
Cost
A base-metal thermocouple is usually the cheapest sensor for a given probe size, and type K costs least of all. A PT100 element costs more, and three- or four-wire cable adds to that. Platinum thermocouples (types R, S and B) cost more than either because of the platinum and rhodium they contain. Count the whole loop when comparing: the transmitter or input card, the extension cable and any recalibration often cost more than the sensor itself.
When to choose an RTD
- The process temperature stays below about 450 to 600 °C.
- You need a tolerance of a few tenths of a degree, as in pharmaceuticals, food, HVAC energy metering or laboratory work.
- Long-term stability matters more than speed.
- Sensors must be interchangeable without individual calibration.
When to choose a thermocouple
- The temperature is above about 600 °C, as in furnaces, kilns, exhaust gas or molten metal.
- The probe must be very thin, or respond in a fraction of a second.
- The installation is exposed to heavy vibration or shock.
- Cost per point matters and a tolerance of a degree or two is acceptable.
Common mistakes
Comparing the sensor tolerance alone. A class A PT100 read through a two-wire connection over a long cable can be less accurate than a class 1 thermocouple with a good transmitter. Compare whole measurements.
Using copper cable to extend a thermocouple. The reading then refers to the temperature where the copper starts, not to the transmitter terminals.
Choosing class A and running it outside its range. For wire-wound elements, class A is defined only from −100 to 450 °C; above that, only class B applies.
Ignoring drift. A type K thermocouple that has spent months between 250 and 600 °C can read several degrees off without any visible fault.
Frequently asked questions
Is an RTD more accurate than a thermocouple?
Usually yes, within the RTD’s range. At 100 °C a class A PT100 allows ±0.35 °C and a class 1 type K thermocouple allows ±1.5 °C. At 600 °C a class 1 type K thermocouple (±2.4 °C) is tighter than a class B PT100 (±3.3 °C), and class A is no longer defined there for wire-wound elements.
What is the maximum temperature of a PT100?
IEC 60751 covers platinum RTDs up to 850 °C, but the tolerance classes stop lower: class B to 600 °C and class A to 450 °C for wire-wound elements. Above about 600 °C a thermocouple is the usual choice.
Which responds faster, a thermocouple or an RTD?
A thermocouple, in most cases. Its junction can be made much smaller than an RTD element, and a grounded junction in a thin mineral-insulated sheath reacts quickly. The response time of any probe depends mainly on sheath diameter, construction and the medium around it, so compare datasheet values for the same conditions.
Do I need a transmitter for a thermocouple or an RTD?
Not necessarily: many controllers and PLC input cards accept both directly. A head-mounted transmitter is useful over long distances, because it converts the small signal to 4–20 mA near the sensor and, for a thermocouple, removes the need for extension cable all the way to the control room.
Why does my thermocouple read wrong when the cable is long?
The extension cable must be the same thermocouple type, or a matching compensating cable, from the sensor to the instrument. Plain copper cable, or a cable of another type, moves the reference junction. Polarity swapped at one connection also causes a large error.
Can I replace a thermocouple with an RTD on the same input?
Only if the input can be configured for an RTD and the wiring allows it. A thermocouple input measures millivolts; an RTD needs an excitation current and two to four wires. The cable usually has to change as well.
What does “class A” mean for a PT100?
It is a tolerance class in IEC 60751: ±(0.15 + 0.002·|t|) °C, which is ±0.15 °C at 0 °C and ±0.35 °C at 100 °C, valid from −100 to 450 °C for wire-wound elements. Class AA is tighter and class B is about twice as wide as class A.