As an Amazon Associate, Akermin earns from qualifying purchases. Product links on this page are affiliate links.
A thermal camera works by focusing infrared radiation, which is emitted by every object above absolute zero, onto a detector array that converts the incoming energy into an electrical signal. That signal is processed into a false-colour image where each pixel represents a temperature, calculated after the camera applies an emissivity value and compensates for ambient reflected temperature. The published figures that most affect whether you find a fault are the detector resolution, the thermal sensitivity (NETD), and the accuracy specification. A 320 by 240 sensor such as the one in the MiLESEEY TR20 gives 76,800 individual measurement points per frame, which is enough to resolve a hot connection in a distribution board. A lower-resolution sensor still works, but each pixel covers a larger area, and a small fault can be averaged away.
The infrared band, and why a camera can see what your eye cannot
Every object with a temperature above absolute zero emits electromagnetic radiation across a spectrum. The peak wavelength of that emission shifts shorter as the object gets hotter, which is why a piece of steel glows red at around 900 K and white at 1500 K. At the temperatures found in buildings, electrical panels and machinery, the peak sits in the mid-infrared, roughly 8 to 14 micrometres. That band is invisible to the human eye but carries almost all the thermal information a building or electrical inspector cares about.
A thermal camera is built around a detector that absorbs this mid-infrared radiation and converts it to a measurable signal. The most common type in handheld instruments is the microbolometer, an array of tiny resistive elements that change resistance as they warm. The camera reads that resistance change pixel by pixel and, after calibration against internal reference sources, converts it to a temperature value. The result is a grid of temperatures, typically displayed as a false-colour image where darker or bluer tones mean cooler and lighter or redder tones mean warmer. The colour palette is arbitrary; the numbers behind it are the measurement.
This is the first point where buyers get confused. A thermal camera is a radiometer, not a thermometer in the sense of a probe you touch to a surface. It measures the infrared power arriving at the lens and infers temperature from that power. The inference depends on how efficiently the surface emits, which is the next section.
Emissivity: the correction that decides whether your number is right
Emissivity is a number between 0 and 1 describing how efficiently a surface radiates compared with a perfect black body. A matt black surface at 60 degrees Celsius emits close to the theoretical maximum; a polished chrome fitting at the same temperature emits a small fraction of it. A thermal camera that assumes an emissivity of 0.95, which is typical for organic materials, plaster, paint and most dirty surfaces, will read a shiny metal surface as much cooler than it really is, because less radiation is reaching the detector.
The practical consequence is that you must set emissivity before you trust an absolute temperature. Many cameras let you choose a preset material or type a value. The AMPBANK M192 listing states that emissivity is adjustable within its app, which is the minimum you need for electrical work. If a camera does not publish adjustable emissivity, treat its absolute readings on shiny surfaces as indicative only. For comparative work, such as finding the hottest connection in a row of identical terminals, emissivity errors largely cancel because every terminal has the same surface finish. For absolute work, such as deciding whether a component is above its rated temperature, they do not.
There is a second correction that listings rarely mention: reflected temperature. A thermal camera receives not only radiation emitted by the target but also radiation from the surroundings reflected off the target. A shiny stainless steel vessel in a cold room will reflect the cold room and read colder than its true surface temperature. Instruments that allow you to set reflected ambient temperature, as the AMPBANK listing states it does, give you a way to compensate. Instruments that do not leave you with a systematic error you cannot remove.
Resolution, NETD and the arithmetic of finding a small fault
Two numbers dominate the specification sheet, and they do different jobs. Detector resolution, expressed as horizontal by vertical pixels, sets the smallest object the camera can resolve at a given distance. Thermal sensitivity, usually written as NETD and expressed in millikelvin, sets the smallest temperature difference the camera can distinguish from noise. A camera can have high resolution and poor sensitivity, or low resolution and good sensitivity, and the right balance depends on the fault you are hunting.
The resolution arithmetic is straightforward geometry. Each pixel subtends an angle determined by the lens focal length and the pixel pitch. A common approximation for the instantaneous field of view is the pixel pitch divided by the focal length, in radians. Without the focal length and pitch, which most consumer listings do not publish, you cannot calculate the exact spot size, but you can compare cameras: a 320 by 240 sensor has four times as many pixels as a 160 by 120 sensor, so at the same field of view each pixel covers one quarter of the area. That matters when you are looking for a single loose terminal in a panel of forty.
NETD tells you whether a small temperature difference will be visible above the sensor noise. A NETD of 40 millikelvin, which the MiLESEEY TR20 listing states, means the camera can resolve a difference of about 0.04 degrees Celsius in laboratory conditions. A NETD of 50 millikelvin, stated for the DCTOPT thermal camera, is slightly less sensitive but still adequate for most building and electrical work, where the faults you care about are usually several degrees above ambient. The listing for the Bosch GTC600C does not publish a NETD figure at all, which is a real gap if you intend to compare it on sensitivity.
| Product | Stated detector resolution | Stated NETD | Stated accuracy | Stated temperature range |
|---|---|---|---|---|
| MiLESEEY TR20 | 320 x 240 | less than 40 mK | plus or minus 2 percent | -4 to 1022 F |
| Bosch GTC600C | Not stated in the listing bullets | Not stated in the listing bullets | Not stated in the listing bullets | Not stated in the listing bullets |
| AMPBANK M192 | 192 x 192, enhanced to 384 x 384 | 40 mK | plus or minus 2 C or 2 percent | -4 to 842 F |
| DCTOPT | 320 x 240, 240 x 180 super IR stated | less than 50 mK | Not stated in the listing bullets | Not stated in the listing bullets |
Note the inconsistency in the Dctopt Thermal Camera listing: it describes a 320 by 240 sensor and then a 240 by 180 super-resolution mode. Those are different arrays, and the listing does not reconcile them. Treat the resolution claim as stated but unresolved. The AMPBANK M192 listing is clearer in one respect and vaguer in another: it states a native 192 by 192 array and an enhanced 384 by 384 output, which is interpolation, not extra detector pixels. Interpolation sharpens the picture; it does not add measurement points.
Why the displayed image is not the measurement
Most thermal cameras display a colour image and, separately, a set of measurement tools: a centre spot, movable spots, a box or a line profile. The colour image is processed for human vision, often with a palette and a level-and-span adjustment that stretches the temperature range across the display. Two adjacent pixels that differ by half a degree might be shown in dramatically different colours if the span is narrow, or in almost identical colours if the span is wide. The number attached to a spot is the measurement; the colour is a guide to where to put the spot.
This is why the analysis features in a listing matter more than the palette count. The DCTOPT listing states that it can evaluate temperature data across points, lines and areas, which lets you find the maximum in a region rather than hoping your spot landed on the hottest pixel. The MiLESEEY TR20 listing states a customizable alarm temperature and a sound and light alarm, which is useful for scanning a panel quickly: set the threshold and let the camera tell you when a pixel exceeds it. Neither feature replaces a proper spot measurement, but both reduce the chance that you miss a fault because you were looking at the wrong part of the screen.
A related trap is the temperature range. A camera that spans -4 to 1022 F, as the TR20 listing states, covers most electrical and building work. A narrower range gives better resolution per degree, because the analogue-to-digital converter has fewer degrees to cover, but it may clip a hot fault. If the camera has autoranging, it will switch ranges as needed; if not, check that your expected fault temperature falls inside the stated range.
Where a thermal camera stops being the right instrument
A thermal camera measures surface temperature. It does not see through walls, it does not measure air temperature directly, and it does not identify a gas. If you are looking for a water leak behind plaster, the camera shows you the temperature pattern on the plaster, which may be cooler where water is evaporating or warmer where a hot pipe is running. That pattern is evidence, not proof. Confirmation usually requires a moisture meter, and the two instruments answer different questions. The same logic applies to air leaks: a thermal camera shows a cold draught as a streak on a surface, but a draught gauge or a smoke pencil confirms the flow.
For electrical work, a thermal camera is excellent at finding a high-resistance connection because the fault generates heat. It is poor at finding a fault that generates no heat, such as a broken conductor carrying no current or a relay that has failed open. It also cannot see inside a sealed enclosure unless the enclosure has an infrared-transparent window; most do not, and opening the enclosure changes the thermal environment. If you are choosing an instrument specifically for panel work, the guide to choosing a thermal camera for electrical work covers the trade-offs in more detail.
For building diagnostics, the camera is a screening tool. It tells you where to look with a moisture meter or a borescope. The guide to thermal cameras for HVAC sets out the resolution and sensitivity you need for duct and plant work, and the PCB inspection guide covers the very different requirements of finding a hot component on a dense board, where a close focus lens matters more than a wide temperature range.
What the listing does not tell you, and how to read around it
Consumer thermal camera listings are inconsistent. Some publish detector resolution, NETD, accuracy and temperature range; others publish only a headline resolution and a screen size. The Bosch GTC600C listing, for example, describes image storage, the user interface and voice notes, but does not state detector resolution, NETD, accuracy or temperature range in the bullets. That does not mean the camera is incapable; it means you cannot compare it on those figures from the listing alone. If sensitivity and accuracy matter to your work, ask the manufacturer or choose a model whose listing publishes them.
Watch for resolution claims that mix native and enhanced figures. The AMPBANK listing is explicit that the native array is 192 by 192 and the enhanced output is 384 by 384, which is honest. The DCTOPT listing states a 320 by 240 sensor and a 240 by 180 super IR resolution, which is contradictory; one of those figures is probably a different mode or a different product variant. Where a listing is internally inconsistent, treat the lower figure as the safe assumption for measurement purposes.
Finally, remember that the camera is one part of a system. The lens determines the field of view and the minimum focus distance; the display determines whether you can see the image outdoors; the battery determines how long you can work. None of these appear in the resolution figure, and all of them affect whether the instrument is useful. The broader roundup of thermal imaging cameras compares models across these practical dimensions, and the guide to thermal cameras with WiFi covers the models that stream images to a phone or tablet, which is often more useful than a small built-in screen when you are working in a tight space.
The short version
A thermal camera converts infrared radiance into a temperature map, and the quality of that map depends on detector resolution, thermal sensitivity and the emissivity correction you apply. Buy on the figures that match your fault size and temperature difference, not on the screen size or the palette. Where a listing does not publish a figure, assume you cannot rely on it for that property, and ask before you buy.
Frequently Asked Questions
Can a thermal camera see through walls?
No. Mid-infrared radiation is absorbed by most building materials within a fraction of a millimetre. A thermal camera shows the surface temperature of the wall, which can reveal a stud, a pipe or a damp patch because those features change the surface temperature. It does not show the object behind the wall directly.
What NETD do I need for electrical inspection?
For most electrical work, a NETD of 50 millikelvin or better is adequate, because a loose connection typically runs several degrees above ambient. A lower NETD helps when you are looking for small temperature differences, such as a marginally warm breaker, but resolution and emissivity settings often matter more. If the listing does not state NETD, you cannot compare it on sensitivity.
Why does my thermal camera read a shiny metal surface as cold?
Because the camera assumes an emissivity of around 0.95, which is typical for non-metals. Shiny metal has a low emissivity, often below 0.2, so it emits less radiation than the camera expects and the calculated temperature is too low. Set the emissivity to the correct value for the material, or apply a matt tape of known emissivity to the surface and measure that.
Is a higher resolution always better?
Not always. A higher-resolution sensor gives you more measurement points and a smaller spot size at a given distance, which helps find small faults. But a high-resolution sensor with poor sensitivity may still miss a small temperature difference. The right choice depends on the size of the fault and the temperature contrast you expect. For building work, 160 by 120 is often enough; for dense electrical panels, 320 by 240 is a more comfortable starting point.
Can I use a thermal camera to find a gas leak?
Only some gases are visible to a thermal camera, and only if the camera is tuned to the right waveband and the gas has a different temperature from the background. Consumer thermal cameras are not gas detectors. A combustible gas sniffer or a dedicated gas leak detector is the correct instrument for safety-critical gas work.
What does the accuracy figure plus or minus 2 percent actually mean?
It means the camera's reading is expected to be within 2 percent of the true temperature, but the reference for that percentage matters. Some listings state plus or minus 2 percent of reading, others plus or minus 2 percent of full scale, and some add a fixed offset in degrees. The MiLESEEY TR20 and the AMPBANK M192 listings state plus or minus 2 percent, but neither states whether that is of reading or full scale. In practice, accuracy is also limited by emissivity and distance, so the figure is a best-case specification.