Temperature calibration: sensor, measurement chain and uncertainty

Temperature calibration applies to the item and configuration described in the report. A sensor, an indicator and the complete chain are not interchangeable calibration objects. Before reading the error, identify where temperature enters, which components transform the signal and which value the laboratory actually uses.

1. Define the chain’s boundary

Record the probe, serial number, cable, connector, indicator or transmitter, channel, configuration and acquisition. For a logger, the exported data also matter: units, rounding, active corrections and the association between channel and probe must remain identifiable. A correct display reading does not automatically demonstrate that the record used for a decision is correct.

If the probe is calibrated in resistance or electromotive force, the report does not necessarily include the customer’s indicator. If the assembled chain is calibrated, the results concern that combination. Separately calibrated components can support a chain assessment if the model, interfaces and contributions are documented: simply adding certificate values is insufficient.

Define the operating range, critical points and measurement requirement before agreeing the service. Thermal calibration applies a temperature to the probe; electrical simulation replaces the sensor with a signal and assesses the electronic portion. EURAMET cg-11 addresses the latter activity. [1] Neither alone describes temperature distribution throughout a chamber.

2. Relate technology to the relevant contributions

For a resistance thermometer, consider two-, three- or four-wire connections, lead resistance, measuring current, self-heating, insulation and the resistance–temperature function. A four-wire connection reduces certain lead effects but does not automatically eliminate the other contributions. Changing current or thermal coupling can change self-heating. DKD-R 5-1 covers these aspects for resistance thermometers. [2]

For a thermocouple, document type, polarity, extension or compensating cables, connectors and reference-junction treatment. Voltage depends on the materials along the temperature gradient: conductor inhomogeneity can make a change in immersion significant. EURAMET cg-8 distinguishes these effects. [3] Do not attribute all chain performance to the tip alone.

The sensor’s nominal class or conversion table cannot replace results for the individual configuration. Do not transfer a correction between channels or instruments merely because they are set to the same probe type.

3. Make the thermal comparison meaningful

Select a reference and medium compatible with range, geometry, materials, available immersion depth and required uncertainty. The reference thermometer needs relevant results and traceability; the bath or block setpoint is not automatically the reference value. A performance ratio between reference and probe may help design the comparison, but it is not a universal constant.

In a bath or block, check sensing-element positions, effective immersion, contact and heat loss along the stem. Consider axial and radial gradients, stability and the influence of other inserted probes. EURAMET cg-13 concerns block characterisation: it does not make every bore or loading equivalent. [4]

Wait for equilibrium demonstrated by reading series, using criteria consistent with target uncertainty. A stable generator display does not guarantee that both probes have equilibrated. Accredia DT-09-DT describes immersion comparisons and explicitly excludes air-temperature thermometers. [5] Non-immersible probes require an appropriate method; do not force them into liquid.

Fluids, hot or cold surfaces and sheath and cable limitations require appropriate procedures and protection. Keep activities within authorised conditions; this article is not an operating instruction for a bath or furnace.

4. Points, sequence and interpretation

The plan represents the use and chosen model: relevant extremes, decision zones and points needed to evaluate response. The method must justify the number of points, repetitions and cycle direction. Where relevant, compare ascending and descending runs to distinguish hysteresis, drift and insufficient stabilisation; do not conceal the difference in an overall mean.

We use e = t_indicated − t_reference; additive correction is −e. The reference has already been corrected according to its certificate. Stating only “±0.1 °C” is ambiguous: it could mean a specification, maximum error or uncertainty. The report must identify result, units, expanded uncertainty U and factor k.

The budget considers reference and drift, comparison medium, repeatability, resolution and chain-specific influences. Address correlations and contributions already included, without counting the reference uncertainty and the same effects twice. Cofrac LAB GTA 08 discusses modelling and use conditions within its accreditation framework. [6] Calibration uncertainty does not automatically cover different immersion, dynamic response or a new environment; interpolation and corrections require a justified model.

5. Simulated case: same probe, new indicator

A chain with a resistance probe indicated 30.07 °C at a reference value of 30.00 °C: e = +0.07 °C and correction −0.07 °C. After replacing the indicator, a new thermal comparison of the chain reads 30.31 °C at the same reference value: e = +0.31 °C. Assume U = 0.06 °C, k = 2, for each comparison, already including the relevant contributions.

For this example only, the uncorrected indication requirement is ±0.20 °C and the agreed rule is |e| + U ≤ 0.20 °C. The previous result gives 0.13 °C and meets the criterion; the new result gives 0.37 °C and does not. Reusing the old correction would leave a central-value residual of +0.24 °C: it does not restore performance.

These data are invented for teaching. A single-point comparison neither demonstrates the entire range nor identifies the cause by itself. Check settings, wiring, conversion function and current; plan the necessary tests before release. A separate probe certificate retains its own scope and must be assessed, while the old chain certificate does not describe the new combination. Preserve both and the intervention history.

6. Original matrix and return to use

The report must allow reconstruction of the item, points, conditions, corrections and limits. Record results before and after any adjustments. Authorised functions define permitted use, checks and review based on risk and history, connecting them to the calibration programme.

For decisions near limits, use the uncertainty budget. For performance throughout a volume, see stability chamber qualification and mapping. Other pathways are in the Calibration & Laboratory Metrology hub.

7. Sources and scope

Checked: 1 October 2026. Public documents were consulted in the editions listed; these guides are not universal GMP procedures. The matrix and case are original GuideGxP material.

1. EURAMET cg-11, version 2.0, March 2011 — electrical indicators and simulators.
2. PTB/DKD-R 5-1, November 2023, revision 0 — resistance thermometers.
3. EURAMET cg-8, version 3.1, February 2020 — thermocouples.
4. EURAMET cg-13, version 4.0, September 2017 — calibration blocks.
5. Accredia DT-09-DT, revision 00, 2019 — immersion thermometers.
6. Cofrac LAB GTA 08, revision 04 — applicable from 15 October 2023.