Annual Verification: When a Thermal Payload Needs Recalibration

Thermal payloads drift slowly and silently. Nothing fails, no error appears, and the numbers gradually stop being true. If your findings carry numbers that anyone acts on, you need a verification schedule — and you need to know the difference between verifying and recalibrating.

Verification and calibration are different things

Verification measures a known reference and checks whether the payload reads it correctly. It produces a pass or fail against a tolerance and does not change the instrument.

Calibration adjusts the instrument so it reads the reference correctly. It changes the payload and is normally a factory or authorised-service operation.

The distinction matters commercially. Annual verification is cheap and fast; it tells you whether calibration is needed. Sending every payload for calibration annually regardless of condition is a common and unnecessary expense — verify first, calibrate on failure.

When to verify

Trigger	Action	Rationale
Annually, radiometric use	Full verification	Detector response ages slowly
Annually, non-radiometric use	Functional check only	No absolute numbers produced
After a hard landing / impact	Verification + boresight check	Mechanical shock shifts alignment
After water ingress	Full verification	Window and internal optics affected
After window replacement	Full verification	Transmission characteristics change
After firmware update touching imaging	Spot check	Processing changes can shift output
Readings inconsistent with expectation	Immediate verification	The instrument is telling you something
Before a contract requiring certification	Verification with certificate	Client or regulator requirement
Extended storage over 6 months	Spot check	Uncertain drift over long idle periods

A field check you can run monthly

You cannot calibrate in the field, but you can detect drift cheaply. Keep a stable reference — a vessel of water at a known temperature measured with a traceable contact thermometer works well, as does a commercially available emissive reference plate.

Warm the payload fully, set emissivity to match the reference surface, image it from a fixed short distance at a near-normal angle, and record the reading against the contact measurement. Log it every month.

The value is in the trend, not the single measurement. A payload that has read within 1 °C for eight months and then reads 3 °C low has told you something, and it did so before a client did. Anything beyond ±2 °C or ±2 % against a good reference is worth escalating to formal verification.

What a verification certificate should contain

• Payload model and serial number.
• Reference source used, its calibration status and traceability chain.
• Ambient conditions during verification.
• Measured versus reference values at several points across the operating range — not a single point.
• Pass or fail against a stated tolerance.
• Date performed and date next due.
• Whether any adjustment was made, and the as-found and as-left values if so.

Related reading

Technology: shutters, blackbodies and two-point correction and radiometric temperature measurement.

Operations: radiometric vs non-radiometric, report deliverables and maintenance.

Commercial: calibration service in the purchase contract.

What to do when verification fails

A failed verification is not automatically a broken payload, and jumping straight to a factory return is often the expensive wrong answer. Work through four possibilities first.

Procedure error. Wrong emissivity, insufficient warm-up, reflected temperature not entered, or too oblique an angle. Re-run the check with the procedure tightened before concluding anything about the instrument.

Reference error. Water baths stratify, contact thermometers drift, and emissive plates degrade. Verify the reference against a second source before trusting a failure it produced.

Window condition. A scratched, filmed or partially coated germanium window changes transmission and shifts readings low. It is a cheap replacement and a common cause.

Genuine drift. If the first three are excluded and the deviation is consistent across the range, the payload needs recalibration. Record as-found values before sending it — you will want them if a client questions past reports made with that instrument.

That last point deserves emphasis. A payload that has drifted was producing wrong numbers for some period before you caught it. Decide, and document, how far back you will review affected reports.

• Re-run with tightened procedure before suspecting the instrument.
• Verify the reference itself against a second source.
• Inspect and, if in doubt, replace the window.
• Record as-found values before any adjustment is made.
• Decide the look-back period for reports produced since the last passing verification.
• UAV thermal camera calibration: chambers, blackbodies and checks
• Choosing a thermal payload for ArduPilot or PX4 aircraft

FAQ

How often does a thermal payload need calibration?

Verify annually if you produce numeric findings; calibrate only when verification fails. Non-radiometric payloads used for search or security need a functional check but not radiometric calibration, because they produce no absolute numbers.

Can I verify a payload myself?

You can run a useful drift check against a stable reference measured with a traceable contact thermometer, and the monthly trend from that will catch drift early. Formal verification with a certificate requires a calibrated blackbody source and is normally a service-provider task.

What tolerance should I expect?

Most radiometric UAV payloads specify ±2 °C or ±2 % of reading, whichever is greater, under stated conditions. Persistent deviation beyond that against a good reference is grounds for formal verification.