As an Amazon Associate, Akermin earns from qualifying purchases. Product links on this page are affiliate links.
Testing a thermal camera means checking three things in order: that it focuses and resolves detail, that it responds to small temperature differences, and that its absolute readings are plausible. You can do the first two with a cup of water and a piece of tape. The third needs a known reference or a calibration certificate. No consumer camera in this list publishes a calibration certificate, so treat absolute temperatures as indicative unless you have verified them against a reference.
What the sensor actually reports, and why that is not temperature
A thermal camera does not measure temperature. It measures infrared radiance in a narrow band, typically 8 to 14 micrometres, and converts that radiance to a temperature using an assumed emissivity. Every reading you see is the result of that conversion. Change the emissivity setting and the number changes without the object changing.
This matters because the most common mistake with a new camera is to point it at something shiny and believe the number. Bare metal, polished pipework and aluminium foil have low emissivity, often below 0.2. They reflect the thermal radiation of whatever is around them rather than emitting their own. A stainless steel flue can read colder than the room because it is reflecting a cold window. The camera is not broken. It is reporting radiance correctly and converting it with the wrong assumption. The emissivity problem explains the arithmetic in more detail.
The practical fix is to change the surface, not the camera. A strip of matte electrical tape, a dab of black paint or a piece of cardboard has an emissivity around 0.95 and stabilises within a minute. Measure the tape, not the metal. If you need to measure the metal itself, you need a contact probe or a camera with a calibrated emissivity table and a known surface finish. The listing for the MiLESEEY TR20 states a temperature range of -4 to 1022 Fahrenheit and an accuracy of plus or minus 2 percent. That accuracy figure applies to the camera’s radiance measurement, not to your ability to guess emissivity correctly. A 2 percent error on a 200 degree reading is 4 degrees. A wrong emissivity assumption can produce an error of 50 degrees or more.
Focus, resolution and the difference between a blob and a component
The second thing to test is whether the camera can resolve the thing you care about. Thermal resolution is not the same as the display resolution. A camera with a 320 by 240 sensor has 76,800 detector elements. Each element covers a patch of the scene. If the patch is larger than the component you are inspecting, the camera averages the component with its surroundings and you get a blob.
The arithmetic is simple. Take the horizontal field of view in degrees, divide by the horizontal sensor count, and you get the angular size of one detector. Multiply by the distance and you get the spot size. The TR20 has a 50 degree lens and a 320 pixel horizontal sensor. That is 0.156 degrees per pixel. At one metre, one pixel covers about 2.7 millimetres. A 2.7 millimetre spot is fine for a wall or a duct. It is not fine for a surface mount resistor or a small connector pin.
This is where super-resolution claims need care. The AMPBANK M192 states a native 192 by 192 sensor and a super-resolution output of 384 by 384. The DCTOPT states a 240 by 180 super IR resolution. Super-resolution is a processing technique that combines multiple frames or interpolates between detector elements. It can make an image look sharper and help you see a pattern. It does not create new detector elements. The native sensor count sets the physical limit on what can be resolved. Treat the super-resolution figure as an image quality claim, not a resolution claim. The resolution explainer covers the spot size arithmetic in more detail.
Focus matters as much as pixel count. A fixed-focus lens is set to a hyperfocal distance and is sharp from that distance to infinity. Close work needs a manual focus ring or a close-up lens. The listings for the MiLESEEY TR20, the AMPBANK M192 and the Dctopt Thermal Camera do not state whether the lens is fixed or focusable. Not stated in the listing bullets. If you plan to inspect circuit boards, ask before you buy.
Sensitivity, frame rate and the targets that reveal them
NETD stands for noise equivalent temperature difference. It is the smallest temperature difference the camera can show above its own noise. A lower number is better. The MiLESEEY TR20 states NETD less than 40 millikelvin. The AMPBANK M192 states 40 millikelvin. The Dctopt Thermal Camera states less than 50 millikelvin. The Bosch GTC600C does not publish a NETD figure. Not stated in the listing bullets.
You can test sensitivity without a lab. Fill a cup with water at room temperature and let it stand. Put a piece of matte tape on a wall and press the cup against it for a minute. Remove the cup. The tape will be slightly warmer than the wall. A camera with good sensitivity will show a clean gradient. A noisy camera will show a mottled patch where the gradient should be. The test is qualitative, but it tells you whether the camera can see a small difference or whether it is drowning in its own noise.
Frame rate matters for moving targets. The MiLESEEY TR20, the AMPBANK M192 and the Dctopt Thermal Camera all state 25 hertz. That is 25 frames per second, which is smooth enough for handheld scanning. A slower camera, 9 hertz for example, smears a moving scene and makes it hard to hold a steady image. The listing for the Bosch GTC600C does not state a frame rate. Not stated in the listing bullets. If you are scanning a large wall while walking, 25 hertz is the difference between a usable image and a blur.
The Bosch GTC600C is a different kind of instrument. The Bosch GTC600C is described as a connected camera with a built-in microphone and speaker for voice notes, storing up to 500 images and 350 notes. It is built for documentation as much as for measurement. The listing does not publish sensor resolution, NETD or temperature range. Not stated in the listing bullets. That does not make it a bad camera. It makes it a camera whose published strengths are workflow and record keeping, not radiometric precision. If you need to prove what you saw, voice notes attached to an image are genuinely useful. If you need to compare two temperatures to within a degree, you need the missing figures.
A test sequence you can run on any camera
The table below sets out a sequence of checks, what each one establishes, and what it cannot tell you. Run them in order. A camera that fails an early check will not be rescued by a later one.
| Check | What you do | What it establishes | What it cannot tell you |
|---|---|---|---|
| Focus and resolution | Point at a printed text target at the distance you will work | Whether the camera can resolve detail at that distance | Whether the detail is at the right temperature |
| Sensitivity | Warm a tape patch with a cup, remove, watch the gradient | Whether small differences are visible above noise | The absolute size of the difference |
| Emissivity response | Compare tape on metal with tape on wall | Whether the camera is reading radiance or assuming a surface | The true temperature of the metal |
| Absolute accuracy | Compare with a calibrated contact thermometer on a matte surface | Whether the camera’s conversion is plausible | Accuracy across the full range |
| Stability | Leave the camera on for ten minutes and re-read a fixed target | Whether the reading drifts as the sensor warms | Long-term calibration drift |
The absolute accuracy check is the one most people skip. A camera can be perfectly linear and still be offset by several degrees. The only way to know is to compare it with a reference you trust. A calibrated contact thermometer is a reasonable reference for a matte surface in still air. It is not a reference for a shiny surface or a moving target. If the camera and the thermometer disagree by more than the stated accuracy, the camera needs attention. The MiLESEEY TR20 states plus or minus 2 percent. The AMPBANK M192 states plus or minus 2 degrees Celsius or 2 percent, whichever is greater. The Dctopt Thermal Camera does not state an accuracy figure. Not stated in the listing bullets.
Stability is the check nobody mentions. Uncooled microbolometers drift as the sensor temperature changes. Most cameras have an internal shutter or a flag that recalibrates periodically. You will hear a click and see the image freeze for a moment. That is normal. If the image drifts without a click, or if the click happens every few seconds, the sensor is struggling. Run the camera for ten minutes before you trust a reading. A cold camera just out of a vehicle in winter will read low until it stabilises.
When the camera is the wrong instrument, and what to use instead
A thermal camera shows surface temperature patterns. It does not see through walls, it does not measure air temperature, and it does not measure moisture. It shows the thermal effect of moisture, which is a different thing. A wet patch evaporates and cools, so it often appears colder. But a cold patch can also be a missing insulation, a draught or a pipe. The camera gives you a pattern. You still need a moisture meter to confirm water and a borescope to see inside the cavity. The comparison with a borescope explains where each one stops.
For electrical inspection, the camera is genuinely useful. A loose connection has higher resistance and gets hotter under load. The camera finds the hot spot. But it does not tell you the load, and a hot spot at no load may be harmless. You need to know the circuit is carrying current before you interpret the image. A camera with a low NETD is better here because the temperature difference between a good joint and a marginal one can be small. The MiLESEEY TR20 and the AMPBANK M192 both state 40 millikelvin. The Dctopt Thermal Camera states less than 50 millikelvin. The difference between 40 and 50 is small in practice, but the difference between either and an unstated figure is unknown.
For measuring a single component’s temperature, a contact thermometer or an infrared thermometer with a known spot size is often better. An infrared thermometer averages over its spot, which may be larger than the component. A thermal camera shows you the whole scene, which helps you find the component, but its absolute accuracy is usually worse than a good contact probe. The comparison with an infrared thermometer sets out the trade-off.
The phone-attached camera, the AMPBANK M192, is a different proposition. It has no battery and no display of its own. It uses the phone’s screen and processing. That makes it small and convenient for quick checks. It also means the camera is limited by the phone’s thermal management and the app’s implementation. The listing states it works with USB-C devices running iOS, Android or Windows, and notes it is not compatible with Mac. If you work across multiple devices, check the compatibility before you rely on it. The listing also states adjustable emissivity and distance settings within the app. That is a genuine advantage over a fixed-emissivity camera, provided you remember to set them.
What the specifications do not tell you
Every listing on this page publishes some figures and omits others. The omissions matter as much as the numbers. The MiLESEEY TR20 publishes resolution, frame rate, NETD, temperature range, accuracy and display size. It does not publish the spectral range, the detector type or the focus type. Not stated in the listing bullets. The Bosch GTC600C publishes storage capacity and voice note capacity. It does not publish sensor resolution, NETD, temperature range or accuracy. Not stated in the listing bullets. The AMPBANK M192 publishes native and super-resolution, frame rate, sensitivity, temperature range and accuracy. It does not publish the spectral range or the detector type. Not stated in the listing bullets. The Dctopt Thermal Camera publishes display size, super-resolution, sensitivity and frame rate. It does not publish the native sensor resolution, the temperature range or the accuracy. Not stated in the listing bullets.
This is not a criticism of the listings. It is a warning about comparison. A table of published figures compares what the manufacturers chose to publish, not what the cameras can do. Two cameras with the same NETD can have very different optical designs, and the optics often matter more than the sensor for close work. A camera with a wide lens and a high pixel count can resolve less detail at close range than a camera with a narrow lens and a lower pixel count, because the wide lens spreads the same scene over more pixels but each pixel covers more of the object. The spot size arithmetic in the resolution section is the only reliable way to compare.
The honest summary is this. You can test focus, sensitivity and stability yourself with simple targets. You cannot test absolute accuracy without a reference. You cannot test long-term drift without time. And you cannot compare two cameras on specifications alone if one of them does not publish the specification you need. If a listing is silent on a figure that matters to your work, ask the seller. If they cannot answer, that is your answer.
The test that matters most is the one you run before you buy
A thermal camera is a radiance meter with a display. It is useful when you understand what it is measuring, and misleading when you do not. The checks in this article take an hour and will tell you more about a camera than any specification sheet. Run them on the camera you have, and ask the same questions of the camera you are considering. If the seller cannot tell you the native resolution, the NETD and the focus type, the camera is not ready for the work you are planning.
Frequently Asked Questions
Can I calibrate a thermal camera myself?
You can check it against a reference, but you cannot calibrate it without a blackbody source and the manufacturer's service software. Checking means comparing the camera's reading with a calibrated contact thermometer on a matte surface in still air. If the difference is within the stated accuracy, the camera is plausible. If it is outside, the camera needs factory service. Do not adjust anything in a menu that claims to calibrate the camera unless the manufacturer documents it.
Why does my camera show a different temperature every time I press the trigger?
The camera is probably recalibrating its internal reference. Most uncooled cameras have a shutter or flag that moves across the sensor periodically. You hear a click and the image freezes for a moment. That is normal. If the reading changes every time you press the trigger without a click, the camera may be averaging a different number of frames or the target may be moving. Hold the camera still and let it settle before you read.
Is a higher resolution always better?
No. Resolution sets the smallest detail you can see at a given distance, but a high-resolution camera with a wide lens can resolve less at close range than a lower-resolution camera with a narrow lens. The number that matters is the spot size at your working distance. Calculate it from the field of view and the native sensor count. Super-resolution figures do not change the native count.
Can a thermal camera see through a wall?
No. It sees surface temperature. It can show you the thermal pattern caused by a stud, a pipe or a damp patch, but it cannot see the object itself. You still need a moisture meter to confirm water and a borescope to see inside the cavity. A thermal camera is a screening tool, not a confirmation tool.
What does NETD actually mean in practice?
NETD is the smallest temperature difference the camera can show above its own noise. A lower number means the camera can distinguish smaller differences. In practice, a camera with a NETD of 40 millikelvin can show a 0.04 degree difference in ideal conditions. Real conditions are never ideal, so treat the figure as a relative guide, not an absolute limit. The difference between 40 and 50 millikelvin is small. The difference between a published figure and an unpublished one is unknown.
Do I need a camera with a temperature alarm?
An alarm is useful when you are scanning a large area and cannot watch the screen continuously. The MiLESEEY TR20 and the AMPBANK M192 both state adjustable high and low temperature alarms. The alarm triggers when a pixel exceeds a threshold. That is useful for finding a hot spot in a panel. It is not a substitute for understanding the load on the circuit. A hot spot at no load may be harmless. The alarm tells you where to look, not what it means.