As an Amazon Associate, Akermin earns from qualifying purchases. Product links on this page are affiliate links.
Point the camera at a surface, let it settle to the background temperature, and read the apparent temperature only after you have set emissivity, focus and distance correctly. The number is a surface temperature estimate, not a core temperature and not a moisture reading. A camera such as the MiLESEEY TR20 gives you 320 by 240 pixels and a stated NETD below 40 millikelvin, which is enough resolution and sensitivity for building and electrical work. The most common failure is not the camera; it is scanning a shiny or wet surface and believing the number.
What the camera is actually measuring, and why that changes how you point it
A thermal camera measures infrared radiance arriving at its detector. It then converts that radiance to a temperature using a model that assumes the surface emits like a greybody with the emissivity you entered, and that nothing between you and the surface is absorbing or reflecting. Every number on screen is the output of that model. Change the emissivity setting and the number changes while the physics does not.
This matters because most building materials are reasonably well behaved. Painted plasterboard, brick, timber and most matt plastics sit in a narrow emissivity band and read close to true surface temperature. Bare metal, polished pipework, foil-faced insulation and wet surfaces do not. A shiny copper pipe will read the temperature of whatever it is reflecting, which is often the room, not the pipe. If you scan a wall and see a cold patch, the honest first question is whether the wall is cold or whether it is reflecting something cold.
The camera also only sees the surface. It cannot tell you the temperature inside a wall cavity, the temperature of water in a pipe, or the temperature of a person’s core. It tells you the temperature of the outermost millimetre of material it can see. Everything you conclude about what is behind that surface is inference, and it is only as good as the thermal pattern you can justify.
Resolution sets the smallest object you can trust. A 320 by 240 detector has 76,800 pixels; a 192 by 192 detector has 36,864. The number matters because a small hot spot on a distant terminal block may fall inside a single pixel and be averaged with its cooler surroundings. That is why a smartphone thermal camera with 192 by 192 native resolution is fine for close work and less useful for scanning a distribution board from across a room. The arithmetic of pixel coverage is set out in how thermal camera resolution changes what you can see.
The setup sequence: focus, range, emissivity, then scan
Most bad thermal images are bad before the shutter is pressed. Work through the same four steps every time.
- Focus. A blurred thermal image smears temperature across pixels. Manual-focus cameras must be focused on the target surface, not on the display. Fixed-focus cameras have a minimum focus distance; inside it, the image is soft and the numbers are unreliable. The listing for the MiLESEEY TR20 states a 50 degree wide-angle lens but does not publish a minimum focus distance, so treat close-up work as something to check by eye.
- Range. Set the temperature span so the target fills the display. A camera auto-ranged to minus 20 to 450 degrees Celsius will show a 3 degree wall difference as almost uniform. Narrow the span until the pattern appears. The same camera’s stated range is minus 4 to 1022 degrees Fahrenheit, which covers most building and electrical work but tells you nothing about the span setting you should use.
- Emissivity. Set it for the material, not for the camera default. Matt paint and plasterboard are close to 0.9; bare metal is far lower and often needs a patch of tape or paint to read reliably. The full procedure is in how to set emissivity on a thermal camera.
- Scan. Sweep slowly, perpendicular to the expected gradient. A cold pipe behind plasterboard produces a linear cold streak; a missing insulation patch produces a broad area. Move at a speed that lets the detector refresh several times over each feature. A 25 Hz refresh rate, as stated for the TR20 and the AMPBANK M192, is adequate for handheld work but not for fast machinery.
Only after those four steps should you read a number. The number is a spot temperature at the cursor, and it is only as good as the emissivity and focus behind it.
Reading a pattern rather than a number: what different jobs look like
Thermal inspection is pattern recognition. The temperature value confirms what the pattern suggests; it rarely reveals it on its own.
| Job | What the pattern looks like | What it usually means | What it does not prove |
|---|---|---|---|
| Missing wall insulation | Broad rectangular cold area, edges aligned with studs or joists | Lower thermal resistance in that bay | That the cavity is empty rather than damp or draughty |
| Air leakage | Irregular cold streak at a joint, skirting or ceiling line | Cool air moving across the surface | Where the air enters, or the leak rate |
| Electrical fault | Local hot spot at a terminal or breaker, warmer than neighbours under load | Higher resistance or loose connection | That the circuit is safe; confirm with a clamp meter |
| Plumbing leak | Cold or warm streak following a pipe run, often with a damp halo | Water on or near the surface | That the pipe itself is leaking rather than condensing |
| HVAC duct loss | Temperature difference across duct insulation, warm or cold patches | Insulation or seal failure | Actual air flow or refrigerant charge |
The pattern is the evidence. The spot temperature is a supporting figure. If you find yourself reporting a number without being able to describe the shape that produced it, stop and re-scan.
For building work, the relevant guide is finding insulation gaps with a thermal camera. For plant and ductwork, the pattern is different and the guide is choosing a thermal camera for HVAC diagnostics.
Conditions that make a thermal image lie
Thermal contrast depends on a temperature difference across the surface. Remove the difference and the camera has nothing to show. Three conditions account for most false negatives.
Steady-state conditions. A wall with no heat flowing through it looks uniform regardless of what is inside it. Insulation defects show best when inside and outside temperatures differ by a useful margin, typically ten degrees Celsius or more, held for several hours. Scanning on a mild afternoon after a cold night often shows nothing.
Solar gain and recent heating. A wall that has been in sun stores heat and releases it for hours. The pattern you see may be the sun’s path across the facade, not the construction. The same applies to underfloor heating and to pipes that have recently carried hot water.
Reflections and wet surfaces. Low-emissivity surfaces reflect. Wet surfaces evaporate and cool. Both produce patterns that look like defects. A useful check is to move your own position: a reflection moves with you, a real defect does not.
The camera’s sensitivity figure tells you how small a difference it can resolve. NETD below 40 millikelvin, as stated for the TR20, means the detector can distinguish a few hundredths of a degree under ideal conditions. That is not the same as measurement accuracy, which for the same camera is stated as plus or minus 2 percent. Sensitivity is what makes the pattern visible; accuracy is what makes the number trustworthy. They are different specifications and the marketing often blurs them.
Recording, reporting and the limits of a single image
A thermal image without context is nearly useless six months later. Record the emissivity setting, the distance, the ambient temperature, the time since the heating or cooling was applied, and what you were looking at. Many cameras embed some of this in the file; the listing for the Bosch PRO GTC600C states storage for up to 500 images and voice notes up to 350, which is a practical way to capture the context at the moment of the scan. The listing does not publish detector resolution or NETD, so it is not possible to compare its imaging performance with the others from the bullets alone.
Analysis software matters more than people expect. Spot temperatures at a cursor are easy; line profiles and area statistics are what let you compare one visit with the next. The listing for the DCTOPT camera states point, line and area analysis and a 4GB detachable card, which is the sort of feature set that supports a written report rather than a verbal one.
Where the category stops being the right answer: a thermal camera cannot measure moisture content, air flow, gas concentration or electrical current. It can point you toward a damp patch, but a moisture meter confirms it. It can show a warm duct, but a balometer measures the air. It can show a hot breaker, but a clamp meter measures the load that made it hot. Use the camera to find the question, then use the right instrument to answer it.
The habit that separates a useful scan from a colourful picture
Set emissivity before you trust a number, focus before you read a pattern, and describe the shape before you report the temperature. If you cannot explain why a surface should be at a different temperature from its neighbour, you have not finished the inspection. The camera is a guide to where to look with a contact instrument, not a replacement for one.
Frequently Asked Questions
Can a thermal camera see through walls?
No. It reads the surface temperature of the wall. What you infer about the cavity behind it comes from the pattern of warm and cold areas, and that inference depends on a temperature difference across the wall and on the surface being dry and matt. A camera cannot image pipes or studs directly.
Why does my thermal camera show a different temperature from my contact thermometer?
The camera is solving for temperature using an emissivity value and assuming no reflection. A contact probe measures its own tip once it has reached equilibrium. On bare metal, shiny surfaces or wet surfaces the two will disagree, often by several degrees. On matt plasterboard in still air they should be close. Check emissivity and focus first.
Do I need high resolution for home inspection?
For close work on a wall or a distribution board, a 192 by 192 detector is often enough. For scanning a whole elevation or reading a terminal block from a distance, more pixels help because each pixel covers less area. The listing for the MiLESEEY TR20 states 320 by 240; the listing for the AMPBANK M192 states 192 by 192 native with processing to 384 by 384. Processing improves the display, not the information captured.
What does NETD mean and does a lower number matter?
Noise equivalent temperature difference is the smallest temperature difference the detector can distinguish from its own noise. A lower figure means subtler patterns are visible. Below 40 millikelvin, as stated for the TR20, is a good working figure for building and electrical inspection. It is not the same as accuracy, which is stated separately as plus or minus 2 percent for the same camera.
Can I use a thermal camera to find a water leak?
It can show you where a surface is cooler or warmer than its surroundings, which may be where water is evaporating or where a hot pipe is losing heat. It cannot confirm water. A damp patch that has dried, or a leak under a layer of insulation, may show nothing. Use the camera to choose where to put a moisture meter.
Is a smartphone thermal camera good enough for professional work?
It depends on the job. The listing for the AMPBANK M192 states a 192 by 192 native detector, a 25 Hz refresh rate and plus or minus 2 degrees Celsius or 2 percent accuracy, which is workable for close-range diagnostics. The limits are ergonomics, the phone's battery, and the lack of a hand strap or stand. For repeated reporting on the same asset, a dedicated handheld with onboard analysis, such as the DCTOPT camera with point, line and area tools, is easier to use consistently. The choice is set out in choosing a smartphone thermal camera.