Thermal camera working distance and spot size for safe scans

A thermal camera does not measure a component; it measures an average over whatever fills its pixels. Whether the hotspot on a lug, fuse body or bus-bar joint shows its true temperature is decided by one product: working distance × IFOV, the instantaneous field of view of a single detector pixel. Get the ratio right and a routine cabinet scan taken from behind the arc-flash line yields trustworthy deltas. Get it wrong and a 5 mm hotspot smeared over two pixels reads far cooler than the joint really is, and a failing connection passes inspection. This page gives the IFOV formula, the 3 × 3 pixel rule, a worked example for a 320 × 240 camera at the stand-off distances electrical practice allows, and the levers that shrink the spot: focal length, detector resolution, IR windows and fixed monitoring.

What IFOV is, and how the spot-size number falls out of it

Every thermal image is a grid of detectors, and each detector sees a cone of the scene. The angular width of that cone is the instantaneous field of view, IFOV. Two data-sheet numbers fix it: the field of view of the lens and the pixel count across the detector.

The second line is the small-angle approximation, safe at these angles: 1 mrad is almost exactly 1 mm per metre, so the shop-floor arithmetic is easy. A camera with a 1.31 mrad IFOV covers 1.3 mm of target per metre of distance; at 5 m its pixels are 6.6 mm wide.

Data sheets state FOV outright. The LS-IRC-0044 lists FOV 88.5° × 73.2° on its 640 × 512 detector.

LWIR detectors sense the 8-14 μm band, and this module does so at 50 Hz from a 9-24 V DC supply. Divide 88.5° by 640 pixels and the pixel pitch is 0.138°, which is 2.4 mrad: 2.4 mm of panel per metre of stand-off, or 7.2 mm per pixel at 3 m. The wide view frames a whole board at once, and the cost is exactly this — a fat pixel. At a fixed distance, spot size is FOV divided by pixels and nothing else.

The 3 × 3 pixel rule

Geometrically, a target one pixel wide is at the limit, and a real image never reaches the geometric limit: the lens spreads a point across neighbouring pixels, the detector blurs further, and each pixel's cone gathers background around the target. Field thermography has settled on three tiers.

• 1 × 1 pixel: geometry only. Quote no temperatures from a single pixel.
• 3 × 3 pixels: the minimum for a reading you can act on, and the target should overfill the spot or box you draw.
• 5 × 5 pixels or more: what a ΔT comparison or a trend you will chart deserves, because edge pixels mix in cooler material.

The margin matters more than it looks. A hotspot that fills 2.5 pixels does not read 80 percent of its rise; it can read half the real rise or less. Count pixels on the saved image before the report goes out, not after.

Worked example: a 320 × 240 camera with a 24° lens at the 0.91 m stand-off

Inputs, with their sources:

• Detector: 320 × 240 pixels, a QVGA microbolometer — the classic format of the 0.3 MP LWIR class.
• Example camera in this class: the MV-CI003-GL-N25, a 0.3 MP LWIR camera with an EFL 25 mm lens.
• Lens: 24° horizontal field of view, the standard optic on this camera class.
• Stand-offs: 0.46 m (18 in), the distance an arc-flash study commonly assumes for exposed gear; 0.91 m (36 in), a stand-off facilities put on a special IR-scan label; 3.0 m, a scan taken from well back.
• Rule: 3 × 3 pixels minimum, 5 × 5 preferred.

The rule in words: a target is readable when it fills at least 3 × 3 pixels on the detector, so the smallest readable target is three pixel footprints across, and a confident comparison takes five. Swap in your own data-sheet FOV and pixel count; the method is identical.

The arithmetic:

IFOV = 24° ÷ 320 px = 0.075° per pixel = 0.075 × 17.45 mrad = 1.31 mrad

One pixel then covers:

• 0.46 m × 1.31 mrad = 0.60 mm
• 0.91 m × 1.31 mrad = 1.19 mm ≈ 1.2 mm
• 3.00 m × 1.31 mrad = 3.93 mm ≈ 3.9 mm

Decision: at the 0.91 m label stand-off this camera reads any target of 6 mm or more cleanly (5 pixels) and anything down to 3.6 mm with minimum margin. A 10 mm lug fills 10 ÷ 1.19 ≈ 8 pixels — a solid measurement. Push back to 3 m and the same lug fills 10 ÷ 3.93 ≈ 2.5 pixels, below even the minimum: the reading mixes with cooler material around it and under-reads the joint. At 3 m, quote trends only — or change the optics.

What sets the working distance: the arc-flash side

Spot-size math gives the distance you need; the electrical safety study gives the distance you are allowed.

In a facility with an arc-flash study, opening a panel happens at the working distance the study assumes — 460 mm (18 in) is the common assumption — and the PPE on the label belongs to that task. US practice under NFPA 70E treats infrared scanning as its own task: the person scanning can work in reduced PPE at a stated stand-off, while the person who removes covers carries the PPE for that job. The practical expression is a special IR-scan label, and 915 mm (36 in) is a stand-off seen on such labels.

For covered gear that leaves two honest designs: scan covers-on at the labeled stand-off and size targets to the pixel math above, or keep covers permanently on behind an IR window and scan through it at a few hundred millimetres.

Shrinking the spot: focal length, pixels and framing

IFOV = FOV ÷ pixels, so only two levers exist, and both are chosen when you buy.

Narrower field of view. Halve the FOV and every pixel footprint halves at every distance. For a given detector, FOV scales with focal length, so doubling focal length halves IFOV and doubles the distance at which a target stays readable. LWIR optics are a category of their own, because ordinary glass passes almost none of the 8-14 μm band; compare in the thermal, SWIR and UV lenses group. One point in that range: the LS-SWL-0006, a 17 mm, F1.2 thermal lens sized for a 1/1.2" detector.

More pixels. A 640 × 512 detector halves the IFOV of a 320 × 240 at the same FOV: an 88.5° field on 320 pixels is 4.8 mrad, on 640 pixels 2.4 mrad. Pixel count is often the cheaper lever, because a longer LWIR lens narrows the view at the cost of field of regard.

Frame before you chase IFOV. The 24° example lens at 0.91 m sees 2 × 0.91 × tan(12°) ≈ 0.39 m across — one phase row, which is normally what you want in front of the camera. Covering a 600 mm face at 0.91 m would need 2 × atan(300 ÷ 900) ≈ 37°, and on 320 pixels that pushes the 5-pixel target up to about 9 mm.

To turn a framing width and a stand-off into a focal length, use the calculator:

Two follow-ups matter at the bench: thermal lenses shift focus as camera and scene warm, which the infrared lens athermalization guide covers, and how sensor size and focal length set the field itself is in the field-of-view guide. Spot size is also not the only lens decision on an uncooled camera — f-number drives sensitivity, which the NETD and lens-choice guide takes up.

IR windows and fixed cameras: scanning without opening anything

An IR window is an infrared-transmissive optic mounted in the panel cover. The camera never looks past an open cover, so the hazardous task disappears from the job, and the optical working distance drops to the gap between window and target, typically 100-300 mm inside a cabinet. At 0.20 m the worked example's camera resolves 1.31 mrad × 0.20 m = 0.26 mm per pixel, so the 3-pixel minimum falls to 0.8 mm — about 4.6 times finer than at the 0.91 m stand-off. Fit a window that transmits the camera's band (8-14 μm for LWIR), keep it clean, and log its distance with the scan.

Continuous monitoring goes one step further: a fixed camera lives at the cabinet, watches the joints continuously and alarms on rate-of-rise, so nobody stands in front of energized gear at all. The LS-IRC-0016 is a 640 × 512 LWIR module on USB3 that suits fixed mounting; the MV-CI003-GL-N25 covers the same role in the 0.3 MP class with an EFL 25 mm lens.

Where the process itself is hot and distant rather than hidden behind a cover, the constraints change again — the steel casting monitoring guide walks through that case.

Long stand-offs: what a scan from 5-10 m can still tell you

At 5 m the example camera's pixels are 6.6 mm wide; at 10 m, 13 mm. The 5-pixel rule then demands targets of about 33 mm and 66 mm — cable runs, bus transitions, bushings, drive housings. A 10 mm lug is not measurable from those distances with this camera, whatever the PPE argument for standing back.

Treat far scans as comparison work: same target, same load, same distance, charted over time. The 8-14 μm band exists because air is fairly transparent there, but a long humid path still absorbs some signal, so absolute temperatures from 10 m carry an atmospheric error you do not control. Trend deltas, not degrees.

The decision, per case

For the worked-example camera (320 × 240, 24°, 1.31 mrad):

• Covers-on scan at the 0.91 m label stand-off: measure targets ≥ 6 mm with confidence, 3.6-6 mm at minimum margin, nothing smaller.
• Targets under 4 mm — fuses, set-screws, small lugs: fit an IR window and scan at 0.2 m, or double the focal length, which halves the pixel footprint and doubles the readable range.
• Whole-face framing demanded at 0.9 m: accept a 9 mm 5-pixel limit, or scan per phase row with the tighter view.
• No permission to open anything: fixed monitoring camera at 100-500 mm from the bus, alarming on rate-of-rise.
• Only far access: comparison and trending of targets ≥ 33 mm at 5 m and ≥ 66 mm at 10 m; no absolute degrees quoted.

Before trusting any reading: measure the true working distance to the target plane; count the pixels the target fills in the saved image and discard anything under 3; put a matte spot of tape or paint on the joint and set emissivity to match; re-check focus after the camera warms; and log distance, emissivity and load with every image so later comparisons compare like with like. Save raw data rather than palettized snapshots — the high-bit-depth export guide shows how the extra bits survive into analysis tools.

Pick the camera in the infrared cameras category, pair the field of view in the lens finder, check framing in the camera finder, or start at the camera step below and send us your stand-off and target list — the pixel math decides the rest.