What Is an RTD? Resistance Temperature Detector Guide

Those are the definitions. The error budget around them is what decides whether a measurement is any good, and a tolerance-class number alone does not capture it.

Published November 8, 2024. Updated October 27, 2026. Reviewed by the Focusens Engineering Team.

How an RTD measures temperature

The principle is simple: as temperature rises, the atoms in the platinum element vibrate more and block the flow of current more, so resistance rises too. A current is passed through the element, the voltage drop is measured, and that resistance is converted into a temperature reading with a known equation.

IEC 60751:2022, edition 3.0, published 27 January 2022, specifies that equation, referred to the International Temperature Scale of 1990. The metrological basis for that scale is documented in the BIPM's Guide to the Realization of the ITS-90, chapter 5, Platinum Resistance Thermometry. Above 0 °C:

R(t) = R₀ · [1 + A·t + B·t²]

Below 0 °C, a third term is added:

R(t) = R₀ · [1 + A·t + B·t² + C·(t − 100)·t³]

with A = 3.9083 × 10⁻³ °C⁻¹, B = −5.775 × 10⁻⁷ °C⁻², and C = −4.183 × 10⁻¹² °C⁻⁴. R₀ is the resistance at 0 °C: 100 Ω for a PT100, 1000 Ω for a PT1000. Every value in the table below comes from those equations and can be reproduced in a spreadsheet.

Values calculated from the IEC 60751 coefficients above.

This is where a common assumption breaks down. RTD response is close to linear, not perfectly linear. The slope column above falls by about a quarter between the ice point and 850 °C. A single-multiplier conversion (the familiar "0.385 Ω/°C" figure is just the 0 to 100 °C average) drifts noticeably above roughly 150 °C. For most process-control ranges that drift is small enough to ignore; for wide-span or high-temperature work it is not. Full lookup values across the whole range are in the PT100 resistance table.

PT100, PT1000, and construction

Within platinum RTDs, the two practical choices are nominal resistance and construction.

Both share the same curve shape and the same IEC tolerance classes: multiply any PT100 resistance by ten and it is the PT1000 value. What differs is how each behaves once wires and excitation current are attached, covered below. For a full side-by-side on choosing between them, see PT100 vs PT1000.

A thin-film element is a platinum track deposited on a ceramic substrate: small, fast, and cheap in volume. A wire-wound element is a platinum coil in a ceramic or glass body, which survives higher temperatures and, in the glass form, thermal shock better. IEC 60751 assigns each tolerance class a validity range that differs between wire-wound and thin-film construction, so a class is a promise over a stated window, not a property of the part everywhere. Check the current edition of the standard before writing a class into a specification.

Tolerance classes, and what they actually promise

IEC 60751 defines four classes. The tolerance is expressed in degrees and widens with temperature.

A tolerance class describes the element alone, at the moment it left the factory. It says nothing about the wires connected to it, the current pushed through it, or what the transmitter does with the signal afterward. Treating the class figure as the accuracy of the finished measurement is the single most common specification error in RTD work, because it ignores wiring and self-heating entirely, both of which are usually larger.

The error budget the class figure hides

A Class A PT100 at the ice point is specified to ±0.15 °C. One ohm of lead resistance in a 2-wire connection adds about 2.6 °C to the reading, because the meter cannot tell sensor resistance from cable resistance. That wiring error is roughly seventeen times the tolerance just paid for.

One ohm is not a hypothetical: ten metres of 0.5 mm² copper, out and back, is about 0.69 Ω at room temperature (around 1.8 °C of error on a PT100, 0.18 °C on a PT1000).

Lead resistance is the dominant term in a 2-wire PT100 installation, and it swamps the difference between Class A and Class B. Paying for Class A on a long 2-wire run buys almost nothing until the lead error is compensated. Self-heating runs the opposite way: at the same excitation current a PT1000 dissipates ten times the power of a PT100, which is why PT1000 front ends usually drop excitation to a few tenths of a milliamp. The two errors point in opposite directions, so no sensor is universally better: the right choice depends on the wiring.

The self-heating figures above come from the excitation current and nominal resistance alone; the actual temperature rise also depends on the dissipation constant of the specific probe in the specific medium (a probe in moving water sheds heat far better than the same probe in still air), which is a per-part figure rather than a category one. Ask for it against the exact part code if the application is tight on budget. For the wiring configurations themselves and the arithmetic behind each, see 2-wire vs 3-wire vs 4-wire RTD.

How RTDs fail, and how to tell the failure modes apart

An RTD that has failed outright is easy to spot. An RTD that is merely wrong is the expensive one. Measure resistance at the terminals with the loop de-energized and compare against the table above.

That last row is easy to miss, because the sensor measures correctly at its own terminals. A PT100 wired into an input expecting PT1000 sees a resistance far below anything in its table and either pegs at the bottom of the scale or raises a sensor-break alarm, because a PT1000 never reads as low as 100 Ω anywhere in its range.

Installation and immersion matter as much as the sensor

An RTD reports the temperature of its own element, which is only the process temperature if it is installed so the two are the same thing. Installation error is routinely larger than element tolerance.

Immerse enough of the probe. Heat conducted along the sheath and out through the mounting pulls the element toward ambient: the shorter the immersion, the larger that stem-conduction error. Give the tip room to sit in flowing process fluid rather than a dead pocket near a wall.

Watch the thermowell. A well protects the probe and lets it be replaced without breaking the process, but it also adds thermal mass and an air gap; a loose fit inside a well slows the response further, because the air gap insulates the tip.

Keep sensor cable away from motor and drive cable, ground the shield at one end only, and terminate with ferrules rather than tinned bare strands. A degrading crimp shows up as a slow upward drift in a 2-wire loop.

Where RTDs are actually used

Because of that combination of accuracy and long-term stability, RTDs show up anywhere a process depends on knowing the real temperature, not just a close estimate:

• Industrial process control: monitoring liquids, gases and machinery, where a few degrees of drift can shift an entire batch.
• Power generation: turbines, generators and reactor cooling, where overheating has to be caught early.
• HVAC systems: regulating comfort and energy use in heating, ventilation and air conditioning.
• Food and beverage processing: cooking, pasteurization and cold storage, where the temperature itself is a safety record.
• Scientific and laboratory work: experiments where repeatability matters more than any single reading.
• Medical equipment: incubators and sterilizers, where the acceptable temperature window is narrow.

How to specify an RTD probe

A part number has to answer six questions before anyone can quote it.

Focusens RTD sensors are built with PT100, PT500 or PT1000 elements in three range bands: −50 to 200 °C, −50 to 350 °C and −50 to 600 °C (series MZP-1/4/5/6 and MZT). They come in DIN Class A or Class B, with SUS304 or SUS316 sheaths and PVC, Teflon, silicone or fibreglass leads. Insulation resistance is 100 MΩ and response time is around 10 seconds at 0.63τ. The ordering code carries the element type, tolerance class, simplex or duplex build, mounting style, lead material and lead length, so a complete probe can be specified without a drawing.

Threaded versions are available in M8×1, M12×1, M14×1, M16×1.5, G1/4 and G1/2. Specifications above are from the Focusens product catalogue. The full range sits in the RTD sensor category, and the PT100 platinum resistance thermometer probe is a typical starting point.

RTD or thermistor?

Both change resistance with temperature, but they suit different jobs: an RTD is the wide-range, standardized, replaceable-without-recalibration choice, while an NTC thermistor wins on sensitivity near room temperature and lower unit cost over a narrower span. The full comparison, with real accuracy and cost figures for each, is in RTD vs Thermistor: PT100 or NTC for Accuracy and Range.

FAQ

What does RTD stand for? Resistance Temperature Detector, a sensor that measures temperature from the predictable way a metal element's electrical resistance changes with heat.

What is the RTD formula? For a platinum RTD, R(t) = R₀·(1 + A·t + B·t²) above 0 °C, with an added cubic term below 0 °C. The full equation, coefficients and a worked resistance table are above.

Is RTD response linear? RTD response is close to linear, not exactly linear. The slope changes by about a quarter between −200 °C and 850 °C, which is why a single-multiplier conversion drifts above roughly 150 °C (see the table and explanation above).

If you're unsure which RTD is right for your application, get in touch and we can help you work through it.