A 20 mm hot spot observed from 10 m needs roughly 0.67 mrad per pixel to span three detector elements [S3].
Resolution alone is insufficient: detector pixels, lens field of view, thermal sensitivity, and measurement distance determine whether a hot spot remains visible or is averaged into a cooler background [S1][S2][S3].
Resolution must be matched to hot spot size and distance
S5 gives 120×90 and 640×480 as example detector arrays [S5]. The common 640×480 class contains 307,200 pixels, while 320×240 contains 76,800 [S5]. A thermal imager with more pixels can resolve smaller features, but only when the optics deliver a sufficiently small instantaneous field of view, or IFOV.
A simple sizing test starts with the required target diameter divided by the required number of pixels. For a 20 mm spot measured from 1.50 m, a 3.05 mrad IFOV produces a 4.58 mm pixel footprint and about 4.4 pixels across the target. At 10 m, the same IFOV produces a 305 mm footprint, so the hot spot falls well below one pixel and its apparent temperature falls toward the surrounding 35 °C panel [S3].
For three pixels across a 20 mm target at 10 m, the required IFOV is approximately 0.667 mrad. That figure illustrates why detector pixel count alone is not a valid specification: lens focal length and horizontal field of view set the actual spatial sampling interval [S3].
The three-pixel rule sets the measurement threshold
For dependable temperature numbers, the target area should cover a cluster of at least 3×3 pixels rather than touch a solitary pixel [S5]. This requirement separates seeing a warm area from measuring its temperature. A single blended pixel can show an anomaly while substantially understating its magnitude.
At 1.50 m, a 20 mm hot spot on a 35 °C panel can return 90.0 °C in the hottest pixel when it occupies about 4.4 pixels across [S3]. Moving to 10 m with the same optical sampling makes the reading drift toward 35 °C.
Native detector values should be used for temperature measurements; super-resolution modes can increase displayed image detail by combining shifted frames, but they do not create additional independent detector samples [S3].
NETD determines whether weak anomalies are detectable
NETD, the noise-equivalent temperature difference, expresses the smallest temperature contrast the instrument can distinguish, and lower values indicate greater thermal sensitivity [S2][S4]. NETD measures thermal sensitivity, and lower values are better [S2].
Resolution and NETD solve different problems. A 1024×768 array supplies more thermal samples and finer contours, while a low NETD preserves contrast between a weak hot spot and its surroundings [S2]. Pixel pitch determines how finely textures and edges are captured [S2].
Cooled and uncooled detectors make a related trade-off. For a small anomaly at long range, a sensitive detector may detect the target while spatial sampling limits its temperature measurement.
Resolution classes suit different inspection duties
The practical comparison begins with the target and distance. A 640×480 detector provides more detail than a 320×240 detector [S1], and more pixels divide the scene into smaller, more precise points [S4].
This hierarchy is a starting point, not a distance guarantee. The controlling criterion remains the target's angular size relative to IFOV. A 640×480 sensor with a broad lens can perform less effectively on a distant hot spot than a lower-resolution instrument fitted with a narrower lens [S1][S3].
Use case selection and exclusion criteria
A laser profiler can establish the exact dimensions of a suspect component, while the thermal instrument establishes its temperature contrast. The required detector geometry then follows from that measured size, inspection distance, and acceptable spatial error rather than from a generic resolution label.
Thermal and spatial bandwidth must match the event
Radiometric thermal frames support point, area, and temperature analysis, but a transient event still requires a frame rate compatible with its duration [S8]. For a repeatable defect, spatial sampling and NETD should be specified before frame rate. For a short-lived event, the trade-off moves toward faster acquisition and adequate thermal sensitivity, with the thermal camera architecture determining detector responsiveness and operating constraints.
The comparison is straightforward: 320×240 favors acquisition economy and broad inspection coverage, 640×480 adds spatial sampling for small defects, and 1024×768 adds further detail at greater system cost [S1][S5]. NETD below 25 mK favors faint-contrast detection, while the three-pixel rule determines whether the hottest feature can be measured without averaging [S2][S3].
Electrical inspection practices are also affected by NFPA 70B and NFPA 70E, while ASTM C1060 and ASTM E1186 address building-envelope inspection activities [S5]. These references govern work practices, but they do not make a particular detector resolution adequate for every target. The engineering calculation still has to include hot-spot diameter, standoff, lens field of view, and measurement uncertainty.
Specify from target geometry backward
Record the hot-spot diameter, target temperature contrast, inspection distance, and required measurement accuracy before selecting a detector. Calculate the angular target size, choose an IFOV that places the target across at least three pixels, then verify that the selected lens supports that IFOV. Use native detector resolution for radiometric decisions and pair it with NETD below 25 mK when the expected contrast is weak [S2][S3].
Two signals will expose an undersized installation: a hot spot that becomes visible only after digital zoom, and a peak temperature that falls as the operator increases standoff. In both cases, IFOV or optical working distance is the first constraint to revisit, rather than image-processing settings.