Why the detection range on the box is not the range you will get
Updated 2026-09-20
The box figure is real, but it answers a question you are not asking. It is the distance at which a 1.8 m tall person occupies about two pixels top to bottom — a warm smudge you could not tell from a fence post. Johnson's criteria, the standard behind every military thermal spec, use the target's narrow dimension and ask three separate questions: is something there (about 2 pixels across), what is it (about 8), and which one is it (about 13). Computed that way, a typical 640 scope with a 50 mm lens detects a person at around a thousand metres, recognizes one at a couple of hundred, and identifies one inside that. Those are the numbers on every page here, next to the box figure, so you can see both.
Where the box number comes from
Manufacturers publish a detection range for a standard target — usually a 1.8 m by 0.5 m human, sometimes a deer — and nearly all of them compute it against the 1.8 m height. At the range where that height spans two pixels the target is 'detected'. Some go further and use one and a half pixels. It is not a lie; it is a legitimate way of stating the furthest distance at which a warm vertical object registers at all. It is also of very little use to a hunter, because you cannot shoot a smudge.
What Johnson actually specified
In 1958 John Johnson at the US Army Night Vision Laboratory put observers in front of imagers and measured how many resolvable line pairs across a target they needed to perform three tasks: detect that something was there, recognize what class of thing it was, and identify which specific one. The results, at 50 % probability, were roughly one line pair for detection, four for recognition and six and a half for identification. A line pair is two pixels, so that is 2, 8 and 13 pixels across the target's critical dimension — the narrower one, because it is the width of an upright figure that distinguishes it from the ground. This is the basis of STANAG 4347 and of every serious thermal range specification since.
How the estimates on this site are computed
Each unit's page takes two published numbers — pixel pitch and lens focal length — and divides them to get the angle one pixel sees (the IFOV, in milliradians). Range for N pixels across a 0.5 m target is then 0.5 divided by N times IFOV, times a thousand. That is it. There is no fudge factor and no atmosphere model: the figures assume clear air, 50 % probability, and an average observer. The same function produces the numbers on cards, in compare tables, on the pillar page and in each unit's description, so they cannot disagree with each other.
Why our number is lower — and why it is still generous
Our detection estimate uses 0.5 m instead of 1.8 m, so it is about 3.6 times shorter than a box figure computed the usual way — that ratio is printed on every unit's page. But even our figure is an upper bound. Humidity, rain, fog and dust all absorb long-wave infrared and shorten every range. A deer in October against warm ground has far less thermal contrast than the laboratory target. And 50 % probability means that at the stated range you miss half the time. Treat the recognition figure as the working range of the instrument, and the identification figure as the distance at which you should be confident about what you are shooting.
What to compare instead
The recognition estimate. Detection tells you the sensor can see heat at that distance; recognition tells you that you can tell a coyote from a fox, or a hog from a calf. Every guide on this site that ranks by range says which of the three it is sorting on, and none of them sorts on the box figure.