How do you calibrate a thermometer?

Place it next to a calibrated reference thermometer in a stirred bath or a dry-block calibrator, deep enough that heat flowing along the stem does not matter, and compare the readings once both are stable. Do this at two or more temperatures across the range you use. Record the error at each point and compare it with the thermometer's tolerance.

How do you check a thermometer at the ice point?

Pack shaved ice made from distilled water into an insulated flask, add distilled water, siphon off the excess and let it settle for 15 to 30 minutes. Insert the thermometer to its proper immersion with ice packed around it and read it once stable. It should read 0 °C within its tolerance. A tap-water ice bath can sit up to 0.025 °C below 0 °C.

What is the tolerance of a class A Pt100?

Under IEC 60751:2022, a class A platinum resistance thermometer may deviate by ±(0.15 + 0.002 |t|) °C, where |t| is the temperature in °C without its sign. That is ±0.15 °C at 0 °C and ±0.35 °C at 100 °C. Class A applies from −100 °C to +450 °C for wire-wound sensors and from −30 °C to +300 °C for film sensors.

How often should thermometers be calibrated?

No standard sets one interval. Base it on how stable the sensor is, how hard it is used and how tight its limit is, then adjust it from its as-found history. Ice point checks between calibrations catch a shift early, and NIST recommends checking even lightly used glass thermometers at the ice point at least once a year.

Should I calibrate the probe and the indicator together?

Calibrate them together when they are always used together, because the result is then the error of the reading you use. If probes move between indicators, calibrate each probe on its own and each indicator electrically, so each part carries its own result.

Can a thermocouple be recalibrated?

It can, but the result may not match how it behaves in use. A thermocouple's voltage comes from the temperature gradient along its wires, so wire that changed in use only shows up where it sits in a gradient. EURAMET recommends replacing base-metal thermocouples used at high temperatures rather than recalibrating them, or checking them in place.

Compliance and audits

Thermometer calibration: temperature probes, RTDs and thermocouples

13 min read Published November 8, 2026 Updated November 9, 2026

Thermometer calibration compares the thermometer with a calibrated reference thermometer at two or more temperatures, with both sensors held at the same temperature in a stirred liquid bath or a dry-block calibrator, and records the error at each point against the sensor's tolerance class or the maker's specification. Between calibrations, a reading at the ice point, 0 °C, is a quick check that shows whether the sensor has shifted.

On this page

What a thermometer calibration checks

A contact thermometer can only report the temperature of its own sensor. Calibration finds the relation between the temperature of the sensor and what the thermometer shows, so most of the method is about making sure the sensor really is at the temperature you think it is (DKD-R 5-1, clause 2). A non-contact infrared thermometer reads thermal radiation instead and is calibrated against a blackbody source.

There are two methods. In the comparison method, the thermometer and a calibrated reference thermometer sit side by side in a bath, a dry block or a furnace, and their readings are compared. In the fixed-point method, the thermometer goes into a cell that realizes a defining temperature of the International Temperature Scale of 1990 (ITS-90). For working thermometers the comparison method is the practical one, because it covers any temperature the bath or block can reach.

DKD-R 5-1 calibrates at two temperatures at least (clause 5), and the points should span the range the thermometer is used over. A calibration result is only valid between the lowest and the highest temperature calibrated.

Infrared thermometer calibration

Sensor, indicator or the whole system

Many thermometers have two parts: a probe, such as a Pt100 resistance thermometer (RTD) or a thermocouple, and an indicator or transmitter that turns its signal into a temperature. You can calibrate them together as one system, or each part on its own.

Calibrating the system in a bath gives the error of the reading you actually use, but the result belongs to that probe on that indicator. Swap either part and the calibration no longer applies. Calibrating the probe on its own gives its resistance or voltage at each temperature, and the indicator is then calibrated electrically over its range. EURAMET's thermocouple guide asks the certificate to identify each item that forms part of the measured system, including extension cables and digital indicators (cg-8, 14.2).

With a thermocouple, the indicator also does the cold-junction compensation, which adds an error of its own. Thermocouple reference tables assume the reference junction is at 0 °C. When it is at another temperature, the voltage that matches that temperature is added to the measured voltage, not the temperature itself (cg-8, 4.1 and 4.6).

Tolerance classes: IEC 60751, IEC 60584 and ASTM E2877

IEC 60751:2022 is the standard for industrial platinum resistance thermometers and sensors. It defines the resistance curve, with a temperature coefficient of 0.003851 per °C, and tolerance classes that grow with the temperature t. A Pt100 has a resistance of 100 Ω at 0 °C. Each class applies only over a range of validity, which is narrower for film sensors than for wire-wound ones.

|t| is the temperature in °C without its sign. A class B Pt100 may be off by ±0.3 °C at 0 °C and by ±0.8 °C at 100 °C.
OIML R 84:2003 gives the same formulas for classes AA, A, B and C, with the ranges of validity of that time, and adds a class D. It is free from the OIML.
IEC 60584-1:2013 sets the reference curves of the letter-designated thermocouple types (R, S, B, J, T, E, K, N, C and A) and their tolerances. The tolerances apply to new thermocouples made from wire normally 0.13 mm to 3.2 mm in diameter, as delivered, and do not allow for drift in use.
ASTM E2877-12(2019) is a guide for digital contact thermometers. It introduces nine accuracy classes for the sensor and the instrument together, over −200 °C to 500 °C or the part of that range a thermometer covers.

Tolerance classes of platinum resistance thermometers, IEC 60751:2022 Table 2
Class	Tolerance (°C)	Valid range, wire-wound	Valid range, film	Tolerance at 100 °C
AA	±(0.1 + 0.0017 \|t\|)	−50 to +250 °C	0 to +150 °C	±0.27 °C
A	±(0.15 + 0.002 \|t\|)	−100 to +450 °C	−30 to +300 °C	±0.35 °C
B	±(0.3 + 0.005 \|t\|)	−196 to +600 °C	−50 to +500 °C	±0.8 °C
C	±(0.6 + 0.01 \|t\|)	−196 to +600 °C	−50 to +600 °C	±1.6 °C

Tip: A tolerance class says how far a new sensor may sit from the standard curve. A calibration measures where this sensor actually sits, with an uncertainty. Once you use the calibrated values, the calibration and its uncertainty matter more than the class.

Baths, dry blocks and the reference thermometer

The reference thermometer needs a current calibration certificate, traceable to national standards, with an uncertainty small enough for the tolerance you check. DKD-R 5-1 recommends recalibrating industrial platinum reference thermometers at least every 1 to 2 years, depending on their history (clause 9, Table 1).

In a stirred liquid bath, the sensors sit in the liquid. Its temperature profile across the working volume should be checked with two or more thermometers before it is used for calibration (EURAMET cg-8, 7.2). The thermocouple guide spaces sensors about 1 cm apart and keeps them away from the bottom and the walls (7.4).

A dry-block calibrator heats a metal block with borings for the probes. It is portable and clean, but heat flows along the probes, and EURAMET calls that heat conduction usually one of the largest sources of uncertainty (cg-13, 2.2). The block's display shows the temperature of its control sensor. EURAMET's notes for users point out that the block's calibration relates to the block, and a thermometer in it can differ (cg-13, Annex C). A calibrated reference probe in the next boring measures the temperature where it matters.

Use a boring or an insert no more than 0.5 mm wider than the probe between −100 °C and +660 °C, and no more than 1.0 mm wider above that (cg-13, 2.2).

Immersion depth

Too little immersion can cause large errors, because heat flows along the stem and the sensor settles between the bath temperature and room temperature (DKD-R 5-1, 8.3). DKD-R 5-1 gives a first estimate: immerse at least 10 times the probe diameter for an error below 1% of the difference between the bath and room temperature, and each further 10 diameters cuts the error by a factor of 10. Count the depth from the top of the sensing element, not from the tip of the sheath. A 6 mm probe with a 30 mm element therefore needs at least 90 mm in the bath.

Check it. DKD-R 5-1 reduces the depth by 10% and accepts the result when the reading changes by no more than 10% of the target uncertainty. EURAMET lifts a probe by 20 mm in a dry block (cg-13, Annex C) and withdraws a thermocouple by up to 5 cm (cg-8, 8.2). A reading that changes shows heat conduction.

Where possible, EURAMET calibrates a thermocouple at the immersion it has in use, as long as that is deep enough to overcome heat losses or gains (cg-8, 8.1). For a resistance thermometer, DKD-R 5-1 aims for the deepest immersion the bath or block allows (8.3). Either way, record the depth on the certificate (DKD-R 5-1, 8.3; cg-8, 14.2).

Stability and uniformity

Two properties of the bath or block go straight into the uncertainty. Uniformity is how much the temperature differs from place to place in the working zone: along a boring, between borings and with the block loaded. Stability is how much it wanders over time once it has settled. EURAMET cg-13 has a block characterized for both, with stability observed over at least 30 minutes (3.5). It treats the largest difference found as the half-width of a rectangular distribution, so the standard uncertainty is that difference divided by √3 (4.2.2).

Let everything settle before you read. Then read the reference and the thermometer under test at the same time, or alternate them in a symmetric order so that drift averages out. EURAMET cg-8 gives orders such as S1, X1 ... Xn, S2, S2, Xn ... X1, S1, where S are the reference thermometers and X the thermometers under test (10.3).

RTDs and Pt100 probes: wiring, self-heating and hysteresis

Wiring. A 2-wire probe adds its lead resistance to the reading. In the DKD-R 5-1 example, a Pt100 with a 2.5 m copper lead of 410 mΩ at room temperature has 492 mΩ when the lead is at 70 °C, which reads 0.2 K higher (8.5.1). A 3-wire circuit compensates the leads if they match. The 4-wire circuit avoids the problem and is the preferred connection (8.5.2 and 8.5.3).
Self-heating. The measuring current warms the sensor. DKD-R 5-1 measures at two currents and extrapolates to zero current. Where self-heating is not measured, it counts 30 mK as a rectangular contribution in the uncertainty budget (8.8).
Hysteresis. An industrial platinum thermometer can read slightly differently at the same temperature after heating than after cooling, so DKD-R 5-1 asks for it to be investigated or estimated (3.1.2).
Stability. A knock or vibration can shift a platinum sensor. Measuring its resistance at the ice point before and after transport shows whether it moved (DKD-R 5-1, clauses 5 and 6).

Put these terms in a temperature uncertainty budget

Thermocouples: why they behave differently

A thermocouple does not measure at its tip. Its voltage comes from the temperature gradient along the wires, so wire that has changed in use only matters where it sits in a gradient. If the gradient in the calibration bath falls on a different part of the wire than in the installation, the calibration does not represent use (EURAMET cg-8, 1.3 and 9.1).

For that reason EURAMET recommends replacing base-metal thermocouples used at high temperatures rather than recalibrating them, or checking them in place (cg-8, 1.5 and 13.4).

Run the points in the right order. EURAMET calibrates from the highest temperature down, except for Type K, which changes when cycled to 500 °C or above. Type K is calibrated with rising temperatures and the first point is repeated at the end as a check (cg-8, 10.1).

Digital and liquid-in-glass thermometers

A digital thermometer is a sensor and an indicator in one, so it is calibrated as a system by comparison with a reference. DKD-R 5-1 applies by analogy to direct-reading electronic thermometers and data loggers with resistance sensors. If the thermometer is adjusted, the results before and after the adjustment both go on the certificate (clause 10).

A liquid-in-glass thermometer is made for a stated immersion. A partial-immersion thermometer goes in up to its immersion line. A total-immersion one goes in with the whole liquid column except about the last 1 cm. At the wrong immersion it can read far off, because the liquid in the stem is at the wrong temperature (NIST SP 1088, section 2).

For many glass thermometers used below 200 °C, long-term change comes mainly from the bulb volume, and a reading at the ice point shows it (SP 1088, 7.2). NIST recommends checking even a thermometer used intermittently near room temperature at the ice point at least once a year. It accepts a single-point recalibration at the ice point for mercury thermometers used at 200 °C or below (section 4). That recalibration only updates the corrections from an earlier full calibration. NIST still advises a full recalibration for thermometers with a contraction chamber and fine graduations. Inspect the thermometer regularly and before each calibration for a separated liquid column, which causes large errors (sections 5 and 6).

The ice point check

The ice point, 0 °C, is the simplest reference you can make, and ASTM E563-22 is the standard practice for it. NIST describes the method in SP 1088 (7.7 and 7.8). Shave clear ice made from distilled or deionized water to particles of 2 mm to 5 mm, pack it into a Dewar flask with distilled water, siphon off the excess melt water and let it settle for 15 to 30 minutes. Keep at least 3 cm of the mixture below the bulb, and make a fresh ice point each day you need one.

Following NIST's method, an ice point can be reproduced with an expanded uncertainty (k = 2) of about 0.002 °C (SP 1088, 7.2). A quick bath of tap water and ice is less exact: EURAMET notes that its ice point can lie between −0.025 °C and 0 °C (cg-8, 4.1).

An ice point check confirms one temperature. It shows a shift, but not a change in the slope of the curve, which needs at least two temperatures (DKD-R 5-1, clause 9). For electronic thermometers and probes, treat it as an intermediate check, not a calibration: it does not reset the due date. The exception is a mercury-in-glass thermometer used at 200 °C or below, where NIST SP 1088 accepts an ice point reading as a single-point recalibration (section 4). That recalibration only updates the corrections from an earlier full calibration. NIST still advises a full recalibration for thermometers with a contraction chamber and fine graduations.

As-found readings and a thermometer out of tolerance

Record the readings before anyone adjusts the thermometer or changes the indicator's settings. If an as-found error is outside its limit, the thermometer was out of tolerance for an unknown part of the time since its last good calibration or check. ISO 9001 clause 7.1.5.2 asks you to determine whether earlier results were adversely affected and to act as needed. For a thermometer, look at what its readings controlled, such as an oven or bath setting, a process temperature on a batch record or a test condition.

How often to calibrate a thermometer

No standard fixes the interval. NIST does not recommend a set recalibration interval, and EURAMET's thermocouple guide says there are no formally specified frequencies because thermocouples and their uses vary so widely (cg-8, 13.1). ILAC-G24:2022 / OIML D 10:2022 lists the factors to weigh, including the tendency to drift, the severity of use and the effect of the measured temperature itself, with high temperature on thermocouples as its example (5.1).

Some published starting points:

Industrial platinum reference thermometers in a calibration lab: at least every 1 to 2 years, depending on their history (DKD-R 5-1, Table 1).
Liquid-in-glass thermometers: an ice point check as often as monthly when new or used above 200 °C, extended to once or twice a year as the history shows a stable bulb (NIST SP 1088, 7.6).
Base-metal thermocouples used at high temperatures: plan to replace them rather than recalibrate (EURAMET cg-8, 1.5).

Tip: Shorten the interval after a drop, an overtemperature or a failed ice point check, and lengthen it only on a record of as-found results inside the limit. Record the reason for each change.

What the certificate and record should show

Keep enough detail for someone else to repeat the calibration. EURAMET cg-8 (14.2) and DKD-R 5-1 (clause 10) list these items on top of the general certificate content of ISO/IEC 17025:2017 clause 7.8.

Each item calibrated: the probe, any extension cable and the indicator or transmitter that forms the system.
The temperatures calibrated and the method, such as comparison in a bath or block.
The immersion depth and any other conditions of the calibration.
For an RTD, the measuring current and the self-heating. For a thermocouple, any heat treatment before calibration.
The reference reading and the thermometer reading at each point, the error and the measurement uncertainty.
The results before and after any adjustment, the date, who did the work and the next due date.

How to calibrate a thermometer, step by step

Identify the thermometer, its sensor type and class, the range it is used over and the limit that applies.
Choose at least two points spread over the range in use.
Check the reference thermometer's certificate and due date, and that its uncertainty suits the limit.
Inspect the probe, its cable and its connector, and check it at the ice point if you can.
Put the probe and the reference side by side in the bath or block, deep enough that the top of the sensing element sits at least 10 diameters below the surface. Record the depth.
Wait for stable readings at each point, then read the reference and the thermometer together or in a symmetric order.
Check the immersion by withdrawing the probe a little at the point furthest from room temperature.
Work out the error at each point and compare it with the limit. Record these as-found results before any adjustment.
If the thermometer fails, adjust or replace it, repeat the points as left and assess what it measured since its last good check. Then set the next due date.

Where Axiospec fits

Axiospec is calibration management software. It keeps each thermometer and probe with its calibration history, certificates and next due date, and each calibration names the reference standard used. One calibration record holds up to 100 test points, each with its nominal temperature, its own tolerance and the as-found and as-left readings, so a class whose limit grows with temperature fits point by point. The record also keeps the room temperature and humidity.

On an in-house calibration, a final reading outside its tolerance sets the result to Fail, and recording a pass anyway takes a written reason. A Fail puts the thermometer in quarantine, and the record asks what it measured since its last pass. Axiospec does not run the comparison or pick the points, the limit or the interval. Those stay with you or your lab.

The free plan covers up to 50 active instruments. A new workspace starts with a 45-day trial of Professional, with room for 1,500 active instruments and no credit card. We load your existing records for you at no charge on every plan.

Common questions

How do you calibrate a thermometer?: Place it next to a calibrated reference thermometer in a stirred bath or a dry-block calibrator, deep enough that heat flowing along the stem does not matter, and compare the readings once both are stable. Do this at two or more temperatures across the range you use. Record the error at each point and compare it with the thermometer's tolerance.
How do you check a thermometer at the ice point?: Pack shaved ice made from distilled water into an insulated flask, add distilled water, siphon off the excess and let it settle for 15 to 30 minutes. Insert the thermometer to its proper immersion with ice packed around it and read it once stable. It should read 0 °C within its tolerance. A tap-water ice bath can sit up to 0.025 °C below 0 °C.
What is the tolerance of a class A Pt100?: Under IEC 60751:2022, a class A platinum resistance thermometer may deviate by ±(0.15 + 0.002 |t|) °C, where |t| is the temperature in °C without its sign. That is ±0.15 °C at 0 °C and ±0.35 °C at 100 °C. Class A applies from −100 °C to +450 °C for wire-wound sensors and from −30 °C to +300 °C for film sensors.
How often should thermometers be calibrated?: No standard sets one interval. Base it on how stable the sensor is, how hard it is used and how tight its limit is, then adjust it from its as-found history. Ice point checks between calibrations catch a shift early, and NIST recommends checking even lightly used glass thermometers at the ice point at least once a year.
Should I calibrate the probe and the indicator together?: Calibrate them together when they are always used together, because the result is then the error of the reading you use. If probes move between indicators, calibrate each probe on its own and each indicator electrically, so each part carries its own result.
Can a thermocouple be recalibrated?: It can, but the result may not match how it behaves in use. A thermocouple's voltage comes from the temperature gradient along its wires, so wire that changed in use only shows up where it sits in a gradient. EURAMET recommends replacing base-metal thermocouples used at high temperatures rather than recalibrating them, or checking them in place.

Put it into practice

Got a gage list? Send it over and we load it for you, usually in a couple of business days. Free on every plan. Then log every calibration to a tamper-evident audit trail and produce records on demand.

Create your account

Axiospec is a documentation and workflow tool. It helps you keep clean, traceable, audit-ready records. Certification depends on your own processes, scope and assessor.