Will Your Temperature Cert Survive the Audit? How Often to Calibrate Temperature Sensors

A temperature calibration certificate survives an audit when it shows five things an assessor checks in order: a genuine accreditation reference for the work claimed, an unbroken traceability chain to national standards, a stated measurement uncertainty at each point, a calibrated range that actually covers how the sensor is used, and a calibration interval you can justify, and it fails when any one of those is missing, overstated or expired. The single question behind every temperature finding in a pharmaceutical, semiconductor, food-safety or TIC audit is blunt: show me the accredited certificate and its traceability. If your file answers that cleanly, the auditor moves on. If it does not, one sensor can unravel a whole qualification. This article is the checklist an auditor runs, and how to set calibration frequencies that hold up against it.

The auditor's five-point checklist

When an assessor picks up a temperature calibration certificate, they are not admiring the layout. They are running a mental checklist, and it is worth seeing your own certificates through their eyes.

• 1. Accreditation. Is the calibration accredited for the specific measurement claimed, and does the certificate carry the correct accreditation reference? A logo on the letterhead is not the same as the activity being on the accredited schedule.
• 2. Traceability. Does the certificate state an unbroken chain to national standards, so the reading is anchored to ITS-90 through calibrated references rather than asserted?
• 3. Stated uncertainty. Is a measurement uncertainty given for each calibration point? A result with no uncertainty cannot be judged against a tolerance.
• 4. In-scope range. Does the calibrated range cover the temperatures the sensor is actually used at? A probe calibrated 0 to 100 degrees Celsius but used at 300 is out of scope for that use.
• 5. Interval and validity. Is the calibration current, and is the interval justified rather than arbitrary?

Every point below expands one of these, because passing the audit is simply making all five true at once.

1. Accreditation: the logo is not the point, the scope is

The most common misunderstanding is treating an accreditation logo as a blanket seal. Accreditation is always for specific activities on a defined schedule. Unitest holds SAC-SINGLAS accreditation, number LA-2023-0845-C, and within it the temperature sensor work is scoped precisely: contact RTD and PRT probe calibration under UNI-T001 (-80 to 660 degrees Celsius in the laboratory, extending to -95 degrees Celsius on site), and non-contact infrared and radiation thermometer calibration under UNI-T008 (35 to 500 degrees Celsius). A thermistor, by contrast, is calibrated as a traceable calibration by comparison and is not on the accredited schedule, so its certificate should never carry an accreditation claim. An auditor who finds an "accredited" stamp on work that is not actually accredited has found a serious problem, so honesty here protects you. When you receive a certificate, check that the accreditation claim matches the specific measurement, not just the laboratory's general status.

2. Traceability: the chain an auditor can follow

Traceability is the documented chain of calibrations linking your sensor's reading back to national and international standards, each step carrying its own uncertainty. For a contact RTD or PRT, that chain runs through calibrated references and ITS-90 fixed points such as the triple point of water. For an infrared thermometer, it runs through traceable blackbody sources whose temperature is set with traceable contact sensors. A certificate that states this chain gives the auditor something to follow; one that simply asserts a number does not. If you want the mechanics of how that chain is built for resistance thermometers, our guide on how to calibrate an RTD or PRT probe walks through it, and the sensor types overview shows how each family is anchored.

3. Stated uncertainty: the number that lets you judge fitness

A calibration result without an uncertainty is only half a result. The uncertainty tells you how well the error is known, and it is what lets you decide whether the calibration is good enough for your tolerance. If your process needs to hold plus or minus 1 degree Celsius and your sensor's calibration carries an uncertainty comparable to that, the calibration is not fit for the tolerance, and an auditor will say so. Accredited certificates state uncertainty at each point precisely because it is the figure that makes a certificate usable. Unitest's accredited best measurement uncertainties run from about 0.01 to 0.39 degrees Celsius for contact RTD and PRT work, and around 1.1 to 3.7 degrees Celsius for non-contact infrared and radiation work, with the exact value shown for each calibration point. Read the uncertainty, then compare it with your tolerance; that comparison is the point of the whole exercise.

4. In-scope range: calibrate where you actually use the sensor

A calibration only covers the range it was performed over. A probe calibrated across 0 to 100 degrees Celsius says nothing about its accuracy at 250 degrees Celsius, and using it there leaves an unqualified gap an auditor will find. The fix is to specify calibration points that bracket your real operating temperatures, so the certificate covers how the sensor is used, not a convenient default range. This is especially important for thermocouples used across very wide spans and for infrared thermometers, where the calibrated emissivity and spectral band also form part of "in scope". Before you send a sensor for calibration, note the temperatures it actually works at and ask for points that cover them.

5. Interval and validity: an expired certificate is no certificate

The last check is the simplest to fail: a certificate that has expired, or an interval nobody can justify. A current calibration with a defensible interval closes the checklist. That raises the real question of this article: how often should each sensor type be calibrated?

How to set a defensible calibration interval

There is no single legal number, and that is deliberate. A calibration interval is a risk decision you must be able to justify, built from three inputs:

• Criticality: what does a wrong reading cost? A probe guarding a batch of biologics or a validated semiconductor process justifies a tighter interval than one on a comfort application.
• Conditions of use: a sensor cycled to high temperatures, subject to shock or vibration, or used in a dirty or corrosive process drifts faster than one in a steady, clean environment.
• Calibration history: the sensor's own past certificates are the strongest evidence. A probe that barely moves over successive calibrations can justify a longer interval; one that drifts should be pulled in.

Practical starting points, to be adjusted from those inputs rather than followed blindly, look like this:

• RTDs and PRTs: commonly 12 months in general use, tightened to 6 months for critical pharmaceutical, biologics or food-safety applications. PRTs used as working reference standards deserve tighter control.
• Thermocouples: often 12 months, but shortened where they run at high temperatures, because oxidation-driven drift accelerates with heat.
• Thermistors: commonly 12 months for critical monitoring, with a shorter interval or a check where the sensor has been cycled hard or handled roughly.
• Infrared and radiation thermometers: typically 12 months, with attention to any instrument that has been knocked about, since optics and drift affect them.

Whatever numbers you land on, write down why. An interval you can explain with criticality, conditions and history is defensible; a round number with no rationale is the kind of soft spot a thorough auditor probes.

The drift risk is different for each sensor type

One reason a single blanket interval is weak is that each sensor type drifts through a different mechanism, so the same 12 months carries different risk depending on what you are calibrating. RTDs and PRTs drift from thermal cycling and mechanical shock. Thermocouples drift from oxidation and compositional change at high temperature, which is why a furnace thermocouple is far higher risk than a fridge RTD on the same nominal interval. Thermistors drift from material ageing. Infrared thermometers are affected by optics, contamination and emissivity error rather than immersion. A calibration programme that recognises these differences, and shortens intervals where the drift mechanism is aggressive, is both safer and easier to defend than one that treats every sensor identically. The sensor types guide sets out each mechanism in detail.

Who gets audited on this in Singapore

The five-point checklist is applied, in one form or another, across the sectors that make up Singapore's regulated economy:

• Pharmaceutical and biologics QC laboratories facing HSA GMP and GDP expectations, where a temperature record traced to a drifted probe can mean an out-of-specification batch.
• Semiconductor and electronics manufacturers whose process temperature control feeds yield, and whose customer and quality audits scrutinise calibration.
• Food and beverage operators under SFA licensing and HACCP, where cold-chain and process temperatures must be evidenced.
• TIC (testing, inspection and certification) laboratories under ISO/IEC 17025, whose own accreditation depends on the traceability of every measurement they make.

For all of them, a temperature excursion investigation, a yield-loss enquiry or a surveillance visit eventually lands on the same page: the calibration certificate and its traceability.

Make every temperature certificate audit-ready

The way to stop worrying about the auditor's question is to be able to answer it before it is asked: accredited where it claims to be, traceable, uncertainty stated, in-scope for use, and current on a justified interval. Send us your sensor list and required ranges and we will confirm what is covered under our SAC-SINGLAS scope, calibrate to the right points, and return certificates built to survive the checklist, with a clear quote and no obligation. You can review our full SAC-SINGLAS accreditation and start on the contact page.