As an Amazon Associate, Akermin earns from qualifying purchases. Product links on this page are affiliate links.
For a building fabric survey, thermal sensitivity matters more than resolution. The defects you are hunting, damp, missing insulation, air leakage and thermal bridging, show up as temperature differences of a few tenths of a kelvin spread across large areas of wall, so the limit on finding them is the camera noise floor rather than the pixel count. Resolution takes over as the binding constraint in the opposite situation: a small target at a distance you cannot close, such as a single terminal in an energised panel or a flashing detail on a roof seen from the ground. Both numbers are worth checking, but if a listing publishes only one of them, the sensitivity figure tells you more about building work.
Sensitivity decides what you can see, resolution decides how small it can be
These two specifications are not competing measures of quality. They constrain different axes of the same image. Thermal sensitivity, published as NETD in millikelvin, is the temperature difference that produces a signal equal to the detector own noise. It is a contrast limit. Resolution, published as a pixel count, is how finely the detector divides the scene in front of the lens. It is a geometry limit.
Put a feature in front of the camera and both limits apply independently. A feature can be enormous and still invisible, because its temperature difference from the surroundings sits under the noise. A feature can be strongly warm and still invisible, because it is smaller than the area one pixel covers at that distance and its heat is averaged with cooler material around it. Neither number rescues the other. The NETD specification describes the first limit and nothing else; detector resolution describes the second.
The reason the priority flips between survey types is that building defects and electrical defects sit at opposite corners of that grid. A loose lug running warm is a small, hot object. A cold bridge at a wall to floor junction is a large, barely cool one. The same logic applies to every instrument on this site, which is why the relationship between measurement range and resolution is worth understanding separately from either camera specification.
Why wall and roof surveys sit in the low contrast part of the scale
A building envelope is a slow thermal system. The inside face of a solid masonry wall and the inside face of the same wall where the insulation has slumped will not differ by much, because both are connected to the same room air and the same structure. The temperature difference you are reading is the residue of a heat flow difference after the surface has had hours to even out. On a heated building in cool weather, the figures people work with in the field are commonly in the range of a few tenths of a kelvin up to two or three kelvin for a severe defect, and the smaller end of that range is where most real findings live.
Damp makes this harder rather than easier. An evaporating surface loses latent heat, which drives it below the dry fabric around it, but the depression depends on how much air is moving across it and how dry that air is. A sealed, still room can leave a genuinely wet patch almost indistinguishable from dry plaster. That is a physics constraint, not a camera one, and no pixel count addresses it.
So the useful question for fabric work is not how sharp the image looks but how small a difference the camera can resolve before the picture turns to grain. A worked treatment of that threshold is in the guide to NETD and thermal sensitivity. The short version is that a difference equal to the noise floor is invisible, a difference three to five times the noise floor is clearly visible on a single frame, and the gap between those two states is where a sensitivity specification earns its place.
What a pixel count buys you at four metres from a wall
Resolution becomes concrete once you convert it into a surface footprint. Divide the horizontal field of view by the horizontal pixel count and you get the angle one pixel subtends; multiply that angle in radians by the distance and you get the width of the patch of wall that pixel averages.
Only one of the four listings in this topic publishes both halves of that calculation. The FOTRIC TK5 states a 320 by 240 detector and a 49 by 37 degree field of view, which gives 49 divided by 320, or about 0.153 degrees per pixel, which is 2.67 milliradians. The table below runs that figure out to normal survey distances. The third column applies the common field guideline that a target should fill roughly three pixels across before you trust a spot temperature taken from it.
| Distance from the surface | Width of one pixel on the surface | Smallest patch filling three pixels across |
|---|---|---|
| 1 metre | About 2.7 mm | About 8 mm |
| 2 metres | About 5.3 mm | About 16 mm |
| 4 metres | About 10.7 mm | About 32 mm |
| 8 metres | About 21.4 mm | About 64 mm |
| 15 metres | About 40 mm | About 120 mm |
Read that against what you are actually looking for. At four metres, a camera of this geometry averages every ten millimetres of wall into one reading, and needs a feature about the width of two fingers to measure it properly. For a damp band, a cold floor junction or a missing batt of loft insulation, that is ample. For a hairline crack carrying air leakage, it is not, and no amount of sensitivity will recover a feature that has been averaged away.
True detector pixels against an upscaled picture
The complication in this category is that several listings advertise a resolution the detector does not have. Upscaling has real benefits on screen, and it is not dishonest when the listing says what it is doing, but it changes nothing about the footprint arithmetic above, because that arithmetic depends on how the detector samples the scene, not on how many pixels the image file ends up with.
The FOTRIC listing is unusually direct about this, calling out a genuine 320 by 240 infrared detector with 76,800 raw pixels and contrasting it with imagers that interpolate. The Thermal Master P3 is equally clear in the other direction: it states a native 256 by 192 detector upscaled to 512 by 384, and publishes a sensitivity of 0.035 degrees Celsius, which is the same interval as 35 mK. The REVASRI imager publishes 320 by 240 as a super resolution figure and does not state the native count. The DCTOPT camera is harder still to read, because its title gives a 240 by 180 super resolution while a bullet describes 240 by 240 super resolution technology, and 320 by 240 refers to the screen.
| Listing | Native detector count | Upscaled figure | Sensitivity | Field of view |
|---|---|---|---|---|
| FOTRIC TK5 | 320 by 240, described as a genuine detector of 76,800 raw pixels | None claimed; the listing rejects interpolation | 40 mK, with a 12 micrometre pixel pitch | 49 by 37 degrees |
| Thermal Master P3 | 256 by 192 | 512 by 384 | 0.035 degrees Celsius, equal to 35 mK | Not stated in the listing bullets |
| REVASRI | Not stated in the listing bullets | 320 by 240 | 40 mK | Not stated in the listing bullets |
| DCTOPT | Not stated in the listing bullets | 240 by 180 in the title, 240 by 240 in a bullet | Below 50 mK | Not stated in the listing bullets |
Two things follow. First, every one of these four publishes a sensitivity figure, while only one publishes a field of view and only two publish a native detector count. For this product class, the sensitivity number is simply the better documented one. Second, a listing that volunteers its native count and its optics is telling you it expects the arithmetic to be done, which is a reasonable thing to weight.
Putting the two numbers in order for a fabric survey
Work the decision in this sequence. Start with the sensitivity figure, because it sets whether the defect class you care about is visible at all; if a listing does not publish one, you have no way to judge that and should ask for the datasheet before anything else. Then check whether the published resolution is a detector count or a reconstruction, and treat the native number as the one that governs measurement. Then, if the field of view is published, convert it into a pixel footprint at the distance you actually work from and compare that with the size of the smallest feature you need to measure rather than merely notice.
- Interior fabric survey, working at one to three metres: sensitivity dominates, and almost any native detector in this class has footprint to spare.
- Roof or facade from ground level: resolution and field of view start to bind, because distance multiplies the footprint linearly while the defect stays the same size.
- Electrical and mechanical work: resolution dominates, and the temperature differences are usually large enough that the noise floor stops mattering.
- Reporting rather than screening: the three pixel guideline matters, because a spot reading taken from a feature narrower than that is an average of the feature and its surroundings.
If sensitivity is the number you have decided to buy on, the shortlist of cameras that publish a low NETD is where to start, because it is filtered on a figure the manufacturer has committed to in writing rather than on an image size that may have been reconstructed.
Frequently Asked Questions
Does a 640 by 480 detector find damp that a 320 by 240 detector misses?
Only when the damp patch is too small to fill enough pixels at your working distance. A rising damp band across the bottom of a wall is hundreds of millimetres across and will fill dozens of pixels on either detector. The higher pixel count helps with narrow features such as a single thermal bridge at a window reveal, or when you have to stand a long way back from a roof.
Is super resolution the same as a bigger detector?
No. Super resolution, sometimes labelled TISR or X3, takes the frames a smaller detector produces and reconstructs a larger image from them. It genuinely smooths edges and makes small features easier to see on screen, but it does not add independent thermal measurements. The detector still samples the scene at its native pitch, and the noise floor of each sample is unchanged.
What sensitivity figure is good enough for building work?
There is no standard threshold, and anyone quoting one is inventing it. What you can say from arithmetic is that a defect needs to produce a temperature difference several times larger than the noise floor to stand out on a single frame. A listing that publishes a figure below 50 mK gives you room to work with tenths of a kelvin; a listing that publishes nothing leaves you unable to make that judgement at all.
Does a lower noise floor help on an electrical survey too?
It helps less, because electrical faults are usually warm by tens of kelvin rather than tenths. On a panel survey the binding constraint is almost always how many pixels land on the terminal you care about, which is a function of detector count, lens field of view and how far back the safety boundary forces you to stand.
Why do so few listings publish a field of view?
Because it is a lens specification rather than a headline feature, and the number looks unimpressive next to a pixel count. The practical consequence is that you cannot calculate the surface footprint of one pixel for most consumer cameras. Where a listing does publish both the horizontal pixel count and the horizontal field of view, that calculation takes one line of arithmetic.