Thermal camera resolution and sensitivity measure different qualities

Thermal-camera comparisons often place pixel count and a sensitivity figure next to each other, inviting a quick ranking. The two numbers describe different qualities. Pixel count concerns the image's sampling structure; thermal sensitivity concerns the relationship between a temperature-related signal and noise. Neither number, considered alone, establishes that a camera will deliver the information a particular customer needs.

A product team evaluating an imaging module should therefore begin with the customer's measurement or inspection decision. An industrial user checking whether a component's temperature has changed has a different requirement from someone reviewing a broad image for visual context. The supplier needs evidence matched to that use, rather than a claim that one headline specification makes its product universally better.

Identify the kind of resolution being quoted

A data sheet may describe detector pixels, the dimensions of an output image or a processed image produced by the camera's software. Those descriptions should not be treated as interchangeable. Ask which specification refers to the physical detector and which refers to an output or processing mode.

The image also depends on the optical arrangement. A large pixel count does not explain which part of an inspected object appears within each pixel, or whether the relevant detail is in focus. The customer needs the lens and imaging conditions associated with the supplier's example.

Consider a hypothetical manufacturer assessing a camera for routine inspection of an electronic enclosure. Its engineers need to compare the thermal appearance of several nearby components. A broad image of the entire assembly and a close view of one component answer different questions, even if both are produced by the same detector.

The acceptance discussion should describe the intended observation. Can the user distinguish the relevant regions in the delivered image and retrieve the required information consistently? This is more informative than selecting a pixel count without specifying the physical scene or the output the customer's workflow will use.

Read sensitivity as a noise-related quantity

In its May 2025 explanation of infrared detectors, FLIR describes noise-equivalent temperature difference, or NETD, as the temperature difference corresponding to a signal equal to temporal noise. It explicitly distinguishes this from temperature accuracy. This is manufacturer documentation explaining a specification, rather than an independent comparison of competing products.

A lower quoted NETD can indicate better thermal sensitivity under the stated measurement conditions. The conditions remain part of the claim. A buyer should obtain the relevant configuration and method before comparing figures taken from different data sheets.

For the enclosure-inspection example, the team may need to see a modest change between two otherwise similar observations. Sensitivity is relevant to that task, but the image still needs enough spatial information to separate the components of interest. Improving one quality cannot automatically compensate for the absence of the other.

Avoid converting the number into a general prediction of field performance. A laboratory sensitivity value does not establish a complete inspection success rate, an operational detection distance or a reliable temperature reading for every material. Those are broader claims requiring their own evidence and clearly defined conditions.

Keep temperature accuracy separate

An image can display small differences clearly while the reported absolute temperature remains affected by other parts of the measurement process. If the customer intends to use numerical temperatures, the evaluation must address the validity of those values as well as the appearance of the image.

NIST's 2016 report on high-magnification thermal-camera calibration examines calibration and measurement procedures and highlights the difficulty of analysing uncertainty when important processing is hidden inside camera software. Its research includes spatial-response effects among the considerations. The report is a measurement-method study, not a current product ranking.

The commercial implication is to ask what the supplier can substantiate about the actual output being sold. Is it an image intended for qualitative review, calibrated temperature data or both? Does the evidence cover the delivered lens and processing configuration? A product can be useful in either role, but the customer should know which claim it is accepting.

A buyer using numerical results also needs a suitable calibration traceability package. That package connects the measured result to the relevant reference and uncertainty. It answers a different question from whether an image looks detailed on a demonstration screen.

Compare the same output under declared conditions

A useful comparison should identify the camera, lens, software version, output mode and the conditions of the example. It should also explain whether the displayed image has been enhanced or rescaled. These details allow the customer to understand what it could reproduce with the offered configuration.

Suppose one vendor supplies a carefully adjusted still image while another provides an ordinary stream from its standard interface. The apparent difference may reflect presentation and processing as well as detector capability. The evaluation should establish a comparable customer task and retain the relevant output from both products.

For the enclosure example, the customer could review the same benign assembly under agreed observation conditions. Its engineers would assess whether the required regions remain distinguishable, whether numerical values are available where needed and whether the output can be used in the inspection record. The exercise should follow the manufacturer's authorised product use and the customer's normal laboratory controls.

The purpose is not to design a universal camera benchmark. It is to reduce uncertainty about this purchase. A narrow, representative demonstration with clearly recorded conditions can provide better commercial evidence than an impressive collection of unrelated sample images.

Include the interface and processing commitment

The camera's delivered information may pass through another application before a user sees it. The product team should establish what the interface provides: image frames, temperature-related values, metadata or a combination. A compelling local display does not prove that the same information is accessible through the integration route the customer plans to use.

Ask how the software identifies configuration and changes in processing. If an update changes the presentation or interpretation of the output, the customer may need to understand the effect on comparisons with historical records. That matters particularly when a business is building a repeatable inspection process over several years.

The Black Widow thermal-payload supply-chain story provides a component-market example. A named module's inclusion in a larger product is evidence of a supplier relationship; it does not independently validate every performance claim for the integrated system. The buyer still needs the relevant configuration and acceptance evidence at its own product boundary.

For an imaging-module supplier, explaining this boundary can strengthen a proposal. It can describe which output and supporting information it provides, which integration work belongs to the customer and which additional assessment is needed when the module becomes part of a different assembly.

Buy the information the workflow needs

The cost comparison should include the offered lens, required software, integration effort, calibration services where relevant and the support needed to preserve the agreed output. A less expensive module may require more work before the customer's team can use its data consistently. A more expensive module may provide capabilities that the customer does not need.

Return to the enclosure-inspection decision. If the workflow requires a qualitative comparison saved with an inspection record, purchasing a feature solely because it improves an unrelated specification may add little value. If the workflow requires defensible numerical measurements, omitting the relevant calibration and uncertainty evidence may leave the purchase incomplete.

A supplier should present a small set of claims that correspond to these different uses. State the detector and output resolution, define the sensitivity figure and its conditions, and explain the scope of any temperature-measurement claim. Then show the output in the customer's representative workflow.

That gives the commercial team a defensible reason for the proposed configuration. It can explain how the product supplies the required information and where further integration or evidence is necessary. The resulting decision rests on a useful imaging capability, rather than on treating pixel count and sensitivity as competing entries in a single ranking.