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 onto a microbolometer array, a grid of tiny detectors whose electrical resistance changes as they absorb heat. Each detector corresponds to one pixel and is read out electronically, then converted to a temperature using calibration data and the emissivity setting. The result is an image where brightness or colour represents apparent temperature, not visible light. The camera never touches the surface and never measures temperature directly; it measures radiance and infers temperature. That inference is only as good as the emissivity value you enter, the distance you are from the target, and the field of view that lands on each detector.
The chain from a warm surface to a pixel value
Every thermal camera follows the same sequence. Infrared radiation leaves a surface, passes through the atmosphere, enters the lens, and lands on a detector array. The lens is made of germanium or a similar material because ordinary glass absorbs the long-wave infrared band that room-temperature objects emit. The detector array, usually a microbolometer, contains thousands of tiny suspended plates. When infrared energy strikes a plate, its temperature rises a fraction of a degree, and its electrical resistance changes. The camera reads that resistance, subtracts a reference, and produces a number for each pixel.
That number is not a temperature yet. The camera applies a calibration curve, usually stored at manufacture, that relates detector response to radiance. It then applies the emissivity setting, the distance, and the ambient temperature to estimate the surface temperature. If any of those inputs is wrong, the displayed temperature is wrong even though the detector is working perfectly. The image you see is a map of apparent temperature, not a photograph of heat.
The principle is the same whether the camera is a handheld unit such as the MiLESEEY TR20 thermal camera with a 320 by 240 detector and a stated sensitivity of under 40 millikelvin, or a smartphone attachment such as the AMPBANK M192 USB-C thermal camera that relies on the phone for display and processing. The difference is in the detector, the optics, and the software, not in the physics.
What the detector resolution actually limits
Detector resolution is the number of independent temperature measurements the camera makes at once. A 320 by 240 detector makes 76,800 measurements per frame. A 192 by 192 detector makes 36,864. That matters because each detector sees a cone of infrared energy defined by the lens. If a small component, such as a fuse or a wire terminal, is smaller than that cone, its energy is averaged with the background and the camera cannot resolve it as a separate temperature.
This is why a high-resolution detector with a narrow lens can find a hot terminal that a lower-resolution detector with a wide lens will miss. It is also why super-resolution processing, which several listings mention, is not the same as a larger detector. Super-resolution uses movement between frames or interpolation to make the image look sharper, but it does not create new independent measurements. The listing for the DCTOPT thermal camera states 320 by 240 native resolution and 240 by 180 super IR resolution, which is a processing claim, not a detector count. The listing for the AMPBANK M192 states 192 by 192 native and 384 by 384 enhanced. Treat the enhanced figure as an image-processing feature and the native figure as the measurement limit.
The practical test is the instantaneous field of view, usually written IFOV. It is the angle subtended by one detector. If the listing does not publish it, you can estimate it from the field of view and the detector count, but the listing does not publish the field of view for every model. Where a figure is missing, write it down as not stated and ask the supplier before you rely on the camera for small-target work. For a fuller treatment of this arithmetic, see how to compare thermal camera resolution.
Sensitivity, frame rate, and the numbers in the listing
Thermal sensitivity, usually written NETD for noise equivalent temperature difference, is the smallest temperature difference the camera can show above its own noise. A lower number is better. The MiLESEEY TR20 listing states under 40 millikelvin. The DCTOPT listing states under 50 millikelvin. The AMPBANK listing states 40 millikelvin. These are typical figures for uncooled microbolometer cameras and they tell you the camera can resolve a fraction of a degree, which is enough for building envelope work and electrical inspection where you are looking for a few degrees of difference.
Frame rate is how many times per second the detector is read. A 25 Hz camera updates the image 25 times per second, which is smooth enough for handheld scanning. Lower frame rates produce lag, which makes it harder to track a moving target or to sweep a wall at walking pace. The listings for the MiLESEEY TR20, the DCTOPT, and the AMPBANK M192 all state 25 Hz. The Bosch PRO GTC600C listing does not publish a frame rate, so it is not stated in the listing bullets.
Temperature range and accuracy are separate from sensitivity. The MiLESEEY TR20 listing states a range of -4 to 1022 degrees Fahrenheit and an accuracy of plus or minus 2 percent. The AMPBANK M192 listing states -4 to 842 degrees Fahrenheit and an accuracy of plus or minus 2 degrees Celsius or 2 percent, whichever is greater. The DCTOPT listing does not publish a temperature range or an accuracy figure, so those are not stated in the listing bullets. Accuracy figures of this kind are typical for this class of instrument, but they apply only when emissivity, distance, and ambient temperature are set correctly.
| Feature | MiLESEEY TR20 | Bosch GTC600C | AMPBANK M192 | DCTOPT |
|---|---|---|---|---|
| Detector resolution | 320 x 240 | Not stated | 192 x 192 native, 384 x 384 enhanced | 320 x 240 native, 240 x 180 super IR |
| Thermal sensitivity | Under 40 mK | Not stated | 40 mK | Under 50 mK |
| Frame rate | 25 Hz | Not stated | 25 Hz | 25 Hz |
| Temperature range | -4 to 1022 F | Not stated | -4 to 842 F | Not stated |
| Accuracy | Plus or minus 2 percent | Not stated | Plus or minus 2 C or 2 percent | Not stated |
| Display | 2.8 inch LCD, 480 x 640 | Not stated | Phone screen | 2.8 inch LCD |
The table shows the pattern: the listings that publish the most numbers are not necessarily the most capable instruments, but they are the ones you can compare without guesswork. Where a cell says not stated, the listing is silent and you should not fill the gap with an assumption.
Emissivity, reflection, and the errors that look like findings
A thermal camera measures radiance, and radiance depends on both temperature and emissivity. Emissivity is a number between zero and one describing how efficiently a surface radiates compared with a perfect black body. Bare metal has low emissivity, often below 0.2, which means it radiates little of its own energy and reflects the surroundings instead. Point a thermal camera at a shiny copper pipe and you may see a reflection of the room, not the pipe temperature. The fix is to set the emissivity to the correct value for the material, or to apply a high-emissivity tape or paint to the surface and measure that.
Most cameras let you set emissivity as a number or choose from a material table. The AMPBANK M192 listing states adjustable emissivity within the app. The MiLESEEY TR20 listing does not mention emissivity adjustment, so it is not stated in the listing bullets. If a camera does not let you change emissivity, it is probably fixed at a value near 0.95, which is correct for most building materials and painted surfaces but wrong for bare metal. For a step-by-step explanation, see how to set emissivity on a thermal camera.
Reflections are the second common error. A cold window reflected in a shiny surface looks like a cold patch. A warm radiator reflected in a stainless steel panel looks like a hot spot. The way to tell a reflection from a real temperature is to move the camera slightly. A real feature stays in the same place on the object. A reflection moves with the camera angle. This is a simple test that costs nothing and prevents most false findings.
Where the method gives a number but not the answer
A thermal camera is an excellent tool for finding differences. It is a poor tool for measuring absolute temperature without care. The accuracy figures in the listings, plus or minus 2 percent or plus or minus 2 degrees Celsius, are typical, but they assume the emissivity is correct, the target fills the detector, and the camera has had time to stabilise. In a building inspection, you are usually looking for a difference of several degrees between a wet patch and a dry one, or between an insulated stud and an uninsulated gap. For that work, sensitivity and resolution matter more than absolute accuracy. For a guide to the building case, see how to choose a thermal camera for home inspection.
The method fails when the target is small relative to the detector, when the surface is reflective, when the temperature difference is less than the camera noise, or when the camera is out of focus. It also fails when you are trying to see through something. A thermal camera does not see through walls, water, or glass. It sees the surface temperature of the first opaque object in its path. If you want to find a pipe behind a wall, you are looking for the temperature pattern the pipe creates on the wall surface, which may be faint or absent.
Finally, a thermal camera is not a gas detector, not a moisture meter, and not an alarm. It can show you a temperature pattern that suggests moisture or a gas leak, but it cannot confirm either. Use it to find the place to investigate, then use the appropriate instrument to confirm. The Bosch PRO GTC600C thermal camera listing mentions voice notes and image storage, which are useful for documenting a survey, but the listing does not publish detector resolution or temperature accuracy, so you should not assume it matches the more fully specified models on those points.
Matching the camera to the job
If you are inspecting electrical panels, you want enough resolution to see a terminal and enough sensitivity to see a few degrees of rise. A 320 by 240 detector with a 25 Hz frame rate and a sensitivity under 50 millikelvin is a reasonable starting point, which is the specification the MiLESEEY TR20 and the DCTOPT both state. If you are doing building envelope work, a wider field of view helps you cover a wall quickly, but it also makes each detector see a larger area, so you need to get closer to small features. The MiLESEEY TR20 listing states a 50 degree wide-angle lens, which covers more area per shot but reduces the effective resolution at a given distance. The trade-off is covered in how to choose a thermal camera for HVAC diagnostics.
If you already carry a phone and want a camera for occasional use, a USB-C attachment such as the AMPBANK M192 avoids a second screen and battery, but it depends on the phone and the app. The listing states it works with USB-C iOS, Android, and Windows devices and is not compatible with Mac. If you need a standalone instrument with a screen and a long battery life, a handheld unit is the better fit. For more on that decision, see how to choose a smartphone thermal camera.
None of these cameras is a substitute for a calibrated contact thermometer when you need a defensible absolute temperature. They are survey instruments. They tell you where to look, and they do it very well.
The short version
A thermal camera is a radiance meter with an image sensor. It converts infrared energy into a temperature map using calibration and emissivity, and its usefulness depends on resolution, sensitivity, and correct settings. Use it to find differences and to guide further investigation, not as a stand-alone thermometer.
Frequently Asked Questions
Can a thermal camera see through walls?
No. A thermal camera sees the surface temperature of the first opaque object in its path. It cannot see through walls, glass, or water. It can show the temperature pattern that a hidden object, such as a pipe or a stud, creates on the wall surface, but that pattern may be faint and is not a direct image of what is inside.
Why does a shiny metal surface read the wrong temperature?
Bare metal has low emissivity, so it radiates little of its own energy and reflects its surroundings. The camera reads the reflected energy as if it were the surface temperature. Set the emissivity to the correct value for the material, or apply high-emissivity tape or paint and measure that.
What does NETD mean and what is a good number?
NETD is noise equivalent temperature difference, the smallest temperature difference the camera can resolve above its own noise. Lower is better. Figures under 50 millikelvin are typical for uncooled microbolometer cameras and are adequate for building and electrical inspection. The MiLESEEY TR20 and AMPBANK M192 listings state 40 millikelvin, and the DCTOPT listing states under 50 millikelvin.
Is super-resolution the same as a higher-resolution detector?
No. Super-resolution is image processing that makes the picture look sharper, but it does not create new independent temperature measurements. The detector count is the measurement limit. The DCTOPT listing states 320 by 240 native and 240 by 180 super IR, and the AMPBANK M192 states 192 by 192 native and 384 by 384 enhanced. Compare native figures when comparing measurement capability.
Can I use a thermal camera to check for a gas leak?
A thermal camera can show a cold spot caused by gas expanding, but it cannot confirm a gas leak and it is not a gas detector. Use it to find a place to investigate, then use a calibrated gas detector or a leak-detection fluid to confirm. A thermal camera is not a safety device for combustible gas.
Do I need to let the camera stabilise before measuring?
Yes. Uncooled microbolometer cameras need time to reach thermal equilibrium, especially when moved between environments at different temperatures. The listings do not publish a warm-up time, so it is not stated in the listing bullets. Allow the camera to settle until the image is stable and the readings stop drifting before you record a temperature.