Humidity Calibration: RH Sensors, References and Uncertainty

Calibrating a relative humidity sensor requires comparison under defined conditions, particularly temperature. Different readings taken in thermally different locations do not, by themselves, demonstrate a probe error. Identify the chain, assign a reference with uncertainty and demonstrate that the compared values describe compatible conditions.

1. Relative humidity and temperature belong together

Relative humidity, denoted here by RH, expresses the ratio between the partial pressure of water vapour and its saturation pressure under the specified conditions. Temperature changes the saturation reference: at constant water vapour partial pressure, increasing temperature reduces RH. A value of “50 %” alone therefore does not fully describe a calibration point.

Dew point, frost point and vapour concentration are different quantities. Conversion to RH requires gas temperature, an appropriate model and relevant pressure corrections; at low temperatures, the reference to water or ice also matters. Do not transfer a calculated value to another location without assessing local conditions.

2. Define the chain and reference

Identify the probe, protective filter, indicator, channel, cable, output and settings. Clarify whether the result concerns the display, analogue signal or value recorded by the system. An electronic input check does not characterise the sensing element or its exposure to air. The temperature sensor used in conversion also needs suitable evidence.

A pressure/temperature generator establishes conditions through measured quantities and a model. A chilled-mirror hygrometer measures dew point; obtaining RH also requires gas temperature. An RH reference sensor offers direct comparison but requires knowledge of corrections, uncertainty, drift and hysteresis. The chamber setpoint does not replace this evidence.

Salt solutions generate equilibrium conditions that depend on temperature: the tabulated value is not exact and does not automatically describe the actual vessel. Composition, solution condition, sealing, contamination and equilibrium matter. Within its method, DKD-R 5-8 does not accept them as a route to traceability, while allowing uses for intermediate checks and generation with a separate reference. This methodological choice is not a universal GMP prohibition.

3. Checklist: can the comparison be interpreted?

The following original checklist connects key checks to a practical decision. Completing it before interpreting a difference helps avoid adjustments that compensate for the test setup instead of the sensor.

An empty chamber and one containing warm instruments may create different conditions. Even a stable reading can differ consistently between locations. Avoid contact with liquids or salt aerosols and follow the configuration’s safety and compatibility conditions.

4. Points, response and hysteresis

Agree on RH points and temperatures relevant to use, the method, sequence, replicates and equilibrium criteria. No set of points suits every sensor. Calibration at room temperature does not automatically demonstrate performance at substantially different operating temperatures.

Record the approach to each point: response may depend on previous humidity. A purely ascending sequence does not necessarily quantify hysteresis, the response difference at the same value approached from opposite directions. Distinguish its effect from repeatability and response time. Do not indiscriminately average different paths; state how they are reported or included in uncertainty.

The stability criterion must consider the reference and instrument, not just the controller. More consecutive readings do not always constitute independent observations: faster acquisition does not eliminate drift or correlation. Retain times, conditions and reasons for exclusions allowed by the method.

5. Units and uncertainty: a simulated case

Define error as e = indication − reference and correction as c = −e. When RH is expressed as a percentage, the difference is expressed in RH percentage points. Relative error instead requires division by the reference value. Writing only “1 % error” leaves a critical ambiguity.

In the simulated case, the reference reads 51.0 % RH in a zone at 25 °C; the probe reads 48.0 % RH in a zone at 26 °C. The apparent difference is −3.0 percentage points. We cannot attribute all of it to the sensor or subtract an arbitrary temperature correction. Local measurements, volume characterisation and controlled comparison are needed, while preserving initial data.

After documenting comparable conditions at 25 °C, the reference value is 50.0 % RH and the probe indicates 50.6 % RH: e = +0.6 percentage points, c = −0.6 percentage points. Relative error is 0.6 / 50.0 × 100 = +1.2 %. The simulated expanded uncertainty is U = 0.8 percentage points, with k = 2: it is not 0.8 % of the reading.

Assume a rule agreed for this point: |e| + U ≤ 2.0 percentage points. The result is 1.4 ≤ 2.0. It meets the simulated criterion only under the conditions considered; it does not demonstrate the entire range or authorise changing the rule retrospectively.

The budget must assess relevant reference, temperature and gradient, stability, resolution, repeatability, drift and hysteresis contributions. Convert contributions to the same unit through the model; consider correlations and components already included. Certificate uncertainty does not automatically cover subsequent use, contamination or different mounting.

6. From the report to use

The report must make the item, configuration, method, RH/temperature points, results, uncertainties, coverage factor and treatment of hysteresis reconstructable. If adjustment takes place, separate results before and after intervention. Specify where corrections are applied: a documented correction that is not implemented in the system does not change the recorded value.

Authorised functions define permitted uses and checks consistent with risk and history. Assess filter changes, exposure to condensation or solvents, maintenance and drift. A useful intermediate check has a known reference and conditions; comparing two similarly contaminated probes may not reveal a common problem.

Explore temperature measurement chain calibration and the uncertainty budget. Chamber qualification and mapping remain a separate assessment. Other pathways are in the Calibration & Laboratory Metrology hub.

7. Sources and scope

Checked: 1 October 2026. Technical guides apply within their own scope; the checklist and case are original GuideGxP material, not a universal procedure.

1. PTB/DKD-R 5-8, revision 0, English edition October 2019 — RH hygrometer calibration, references, conditions and sequences; edition linked from the current official catalogue.
2. L. Greenspan, NBS, Humidity Fixed Points of Binary Saturated Aqueous Solutions, 1977, 81A(1), 89–96 — historical scientific study of equilibrium, temperature and saturated-solution uncertainties.