As an Amazon Associate, Akermin earns from qualifying purchases. Product links on this page are affiliate links.
Buy on the detector array first, the noise floor second, and the lens angle third. A 320 by 240 sensor with a noise floor around 40 mK, such as the MiLESEEY TR20, resolves small temperature differences across a wide scene and is the sensible starting point for electrical and building work. Phone attachments like the FLIR One trade sensor size for convenience and cost; they are useful for quick checks but their native detector is small, and the higher figure in the listing is an upscaled image, not a bigger sensor. Accuracy figures of plus or minus 2 percent or 2 degrees Celsius are typical and are not the limiting factor in most inspections; what limits you is whether the camera can see the difference you are looking for.
The sensor is the camera; the screen is not
A thermal camera listing usually leads with two resolution figures, and they mean different things. The detector resolution is the number of individual infrared sensing elements in the array. The display resolution is the number of pixels on the screen you look at. A camera can have a large, sharp screen and a small detector, and the image will look smooth while carrying very little real detail.
The MiLESEEY TR20 publishes a 320 by 240 detector and a 2.8 inch screen at 480 by 640. The FOXWELL RT280 publishes a 2.8 inch screen at 320 by 240 and a native detector of 120 by 90, with a processing step the listing calls TISR that raises the effective resolution to 240 by 180. The FLIR One publishes a native detector of 80 by 60 and an upscaled image of 240 by 180. The FLIR One Pro publishes a native detector of 160 by 120 and an upscaled image of 480 by 360.
Upscaling interpolates between real sensor elements. It makes the picture less blocky and easier to interpret, and it is genuinely useful, but it does not create temperature data that was not sampled. If two objects a centimetre apart fall inside one detector element, no amount of interpolation separates them. That is why the native array is the first number to compare.
What that means in practice: a 320 by 240 sensor has 76,800 elements. An 80 by 60 sensor has 4,800. The ratio is 16 to 1. At a given distance and lens, the larger array puts sixteen times as many independent temperature samples across the same scene. For finding a warm connection in a crowded panel, or a stud behind plasterboard, that is the difference between a usable image and a blur.
NETD, or how faint a difference the camera can see
NETD stands for noise equivalent temperature difference. It is the temperature change that produces a signal equal to the camera’s own noise, expressed in millikelvin. A lower figure is better. A camera with a NETD of 40 mK can distinguish a surface that is four hundredths of a degree warmer than its surroundings from the noise; a camera with a NETD of 100 mK needs a tenth of a degree before the difference rises above the noise.
Why this matters more than accuracy: a thermal camera measuring a wall at 19 degrees Celsius with an accuracy of plus or minus 2 degrees is telling you the wall is somewhere between 17 and 21. That is fine if you are checking whether a radiator is hot. It is useless if you are looking for the cool patch where damp is evaporating, because the patch may be half a degree cooler than the dry wall around it. The accuracy figure sets the absolute number; the NETD sets whether the contrast exists at all.
The MiLESEEY TR20 publishes a NETD of better than 40 mK. That is a genuinely sensitive figure for a handheld at this level and is the specification that makes it useful for building fabric work, not just for electrical hot spots. The other listings here do not publish a NETD figure. Not stated in the listing bullets is not the same as bad, but it does mean you cannot compare them on sensitivity from the listing alone. For more on this, see what NETD means on a thermal camera.
| Model | Native detector | Image after processing | NETD | Field of view | Frame rate |
|---|---|---|---|---|---|
| MiLESEEY TR20 | 320 x 240 | 320 x 240 | Better than 40 mK | 50 degrees | 25 Hz |
| FOXWELL RT280 | 120 x 90 | 240 x 180 (TISR) | Not stated in the listing bullets | 50 x 38 degrees | 25 Hz |
| FLIR One | 80 x 60 | 240 x 180 (Vivid IR) | Not stated in the listing bullets | Not stated in the listing bullets | Not stated in the listing bullets |
| FLIR One Pro | 160 x 120 | 480 x 360 (Vivid IR) | Not stated in the listing bullets | Not stated in the listing bullets | Not stated in the listing bullets |
Field of view decides how much work you can do from one spot
The lens angle, or field of view, sets how much of the scene lands on the detector. A narrow lens gives more detail per object but forces you to stand further back to fit a whole panel or wall into the frame. A wide lens covers more area from one position but spreads the same number of detector elements over a larger scene.
The MiLESEEY TR20 publishes a 50 degree wide-angle lens. The FOXWELL RT280 publishes a 50 by 38 degree field of view. Those are both wide, which suits building inspection and electrical work where you want a whole wall or a whole panel in one image. The phone attachments do not publish a field of view in the listing bullets, so you cannot compare them on this point from the data given.
The arithmetic that matters: the instantaneous field of view, or IFOV, is roughly the field of view divided by the number of detector elements across. A 50 degree lens on a 320 element array gives about 0.16 degrees per element. At three metres, one element covers about 8 millimetres. That is the smallest object the camera can resolve at that distance. A 50 degree lens on an 80 element array gives about 0.63 degrees per element, or roughly 33 millimetres at three metres. Both cameras will show you a warm pipe; only the first will show you a warm terminal block inside a panel from across the room.
This is the calculation most buyers never do, and it is the reason a high-resolution camera with a wide lens can still disappoint on small components. If your work is mostly circuit boards or small connectors, a narrow lens on a modest sensor can beat a wide lens on a large one.
Accuracy, emissivity and the numbers you should not trust blindly
The accuracy figure on a thermal camera, typically plus or minus 2 percent of reading or 2 degrees Celsius, applies to a surface whose emissivity you have set correctly. Emissivity is how efficiently a surface radiates compared with a perfect black body. Shiny metal has low emissivity and reflects the surroundings; matt paint, plasterboard and most plastics are close to 0.95. Point a camera at a bare copper busbar and the reading is largely a reflection of whatever is behind you, not the busbar’s temperature.
The MiLESEEY TR20 publishes a measurement range of -4 to 1022 degrees Fahrenheit and an accuracy of plus or minus 2 percent. That range covers most electrical and mechanical inspection. The other listings do not publish an accuracy figure or a measurement range in the bullets, so treat their temperature numbers as indicative unless the full manual says otherwise. For the underlying physics, see emissivity in thermal imaging.
The practical rule: use the camera to find contrast, then confirm with a contact thermometer or a clamp meter. A thermal camera is a screening tool. It tells you where to look. It does not replace a calibrated contact measurement when the number has to stand up in a report.
Software, storage and the report nobody budgets time for
The image on the screen is the start of the job. The end of the job is a report someone else can read. That is why storage and PC software belong in the buying decision, not in the accessories list.
The FOXWELL RT280 publishes 8GB of built-in storage, enough for more than 20,000 images according to the listing, and Windows software for point, line and area analysis, 2D and 3D views and automatic report generation. It is explicitly not compatible with Mac. If your workflow is a Windows laptop and you need to hand a client a structured report, that is a real advantage. The phone attachments rely on the app and the phone’s own storage; the listing describes step-by-step inspection guides inside the app, which is useful for a homeowner or a tradesperson new to thermal work.
What the listings do not publish is the file format, the radiometric data retention, or whether the exported images keep their temperature data. If you need to re-analyse an image later, check that before buying. A JPEG with a colour scale burned in is a picture; a radiometric file is data.
Frame rate is the last specification worth a mention. The MiLESEEY TR20 and the FOXWELL RT280 both publish 25 Hz. That is smooth enough for handheld scanning. A slow frame rate makes the image lag behind your hand and is the usual reason a cheap camera feels unpleasant to use.
Where a thermal camera is the wrong instrument
A thermal camera sees surface temperature. It does not see through walls, it does not measure air temperature, and it does not detect gas. If you are looking for a leak behind a sealed panel, a thermal camera will show you the cool patch where water is evaporating, but it will not show you the pipe. If you are checking for moisture, the camera is a screening tool and a moisture meter confirms it; see using thermal imaging for moisture problems.
For inspecting inside a wall cavity or a duct, a borescope rather than a thermal camera is the right tool. For confirming a suspect connection, a clamp meter or a multimeter gives the electrical number the thermal image only hints at. And for anything where the absolute temperature must be defensible, a contact probe or a calibrated reference is what you need.
The honest summary: buy a thermal camera to find where to look, not to tell you what the answer is. The best specification for that job is a large native detector, a low NETD and a lens angle matched to the size of the things you inspect. Everything else is convenience.
The short version
Compare native detector resolution before anything else, then NETD if it is published, then field of view against the size of your target. Accuracy figures are similar across this class and rarely the deciding factor. If you need a report, check the software and storage. If you need to see inside something, a thermal camera is not the tool. For a wider view of the market, see the best thermal imaging cameras and the Seek Thermal review.
Frequently Asked Questions
Is a higher megapixel number always better?
Only if it refers to the native detector. Many listings quote an upscaled image resolution, which is interpolated from a smaller sensor. The FLIR One quotes 240 by 180 but its native detector is 80 by 60. Compare native arrays, not processed images.
Do I need a low NETD for electrical work?
It helps, but a loose connection usually shows a large temperature rise, so a modest NETD is adequate. For building fabric, moisture and insulation faults, the differences are small and a low NETD such as the better than 40 mK published for the MiLESEEY TR20 matters much more.
Can a thermal camera see through walls?
No. It reads surface temperature. It can show the cool patch where damp is evaporating or the warm strip where a pipe runs close to the surface, but it cannot image what is inside the cavity. A borescope is the tool for that.
Why does my reading change when I move?
Reflection. Low-emissivity surfaces such as bare metal reflect the surroundings, so the apparent temperature changes with your position. Set emissivity correctly, or cover the surface with tape of known emissivity, and measure again.
Is plus or minus 2 degrees good enough?
For finding faults, yes. For reporting an absolute temperature that must be defensible, no. Use the camera to locate the anomaly and confirm the number with a contact thermometer or clamp meter.
Do phone attachments work as well as a handheld?
They share the same physics but usually have a smaller native detector. The FLIR One Pro has a 160 by 120 native array, which is useful for quick checks, but a handheld with a 320 by 240 array will resolve smaller targets at the same distance.