Skip to content
Request a quote

Lens calibration

How we focus a body worn camera — and why it decides whether a face is usable

A body worn camera is not focused on whatever happens to be in front of it at the moment the lens is turned. It is focused at a single computed distance, chosen so that a subject at arm’s length and a subject at the far end of a street both land inside the depth of field. This page sets out how that distance is calculated, how the station is set up, and how you can check the result on any camera with a tape measure.

The short answer

A body worn camera is focused at its hyperfocal distance — for a typical wide-angle lens around 0.7 to 0.8 m. At that setting the depth of field runs from roughly half that distance to infinity, so 0.4 m and 30 m are both acceptably sharp. Focusing closer trades away far-field sharpness you actually need, in exchange for near-field sharpness you largely do not.

Getting that distance right is arithmetic, not art. Getting the field of view right is where most specifications go wrong, because a wide-angle lens does not follow the lens formula printed in textbooks.

The inputs

Three numbers decide the whole setup

Everything downstream — sharpness, field of view, how many pixels a face receives — comes out of focal length, aperture and the physical size of one photosite.

01

Focal length

For a fixed-focus body worn camera this is deliberately short — usually 2.8 to 3.2 mm. Short focal length buys the wide field of view the job requires, and as a side effect makes the depth of field very deep even at a wide aperture.

02

Aperture

f/2 to f/2.4 is typical: fast enough for indoor and dusk recording, and because the lens is wide, the loss of depth of field is small. A faster lens does not meaningfully help a camera that is already focused past its hyperfocal distance.

03

Pixel pitch

Sensor width divided by horizontal pixel count. A 1/2.9″ sensor 5.376 mm wide with 1920 photosites gives a 2.8 µm pitch. This single number sets the circle of confusion, and therefore the hyperfocal distance.

From pixel pitch to hyperfocal distance

We allow a blur circle of two photosites — the point at which defocus is no longer distinguishable from the sensor’s own sampling limit. From there the chain is short:

pitch = sensor width ÷ horizontal pixels   → 5.376 mm ÷ 1920 = 2.80 µm
c = k × pitch   → 2 × 2.80 = 5.60 µm
H = f² ÷ (N × c) + f   → 3.15² ÷ (2.3 × 0.0056) + 3.15 = 774 mm

So for this lens the hyperfocal distance is 0.77 m. Focused there, the near limit of acceptable sharpness sits at 0.39 m and the far limit at infinity.

Why two pixels and not one? A blur circle smaller than the pixel pitch cannot be resolved by the sensor, so demanding one pixel would mean chasing sharpness the sensor cannot record. Two pixels is the conventional working tolerance and matches what the eye accepts in the finished video.

Depth of field

Why we focus far away instead of on the subject

It feels wrong to point a camera at a chart two metres away and then focus it at 0.77 m. Here is what the two choices actually buy.

Focused at 1.0 m — near subject sharp, far subject soft sharp 0.72 – 1.6 m soft beyond 1.6 m 0 m 30 m Focused at the hyperfocal distance 0.77 m — everything from 0.39 m is usable sharp 0.39 m – infinity 30 m 0 m Same lens, same aperture. Only the focus position differs.
Focusing at the hyperfocal distance costs almost nothing at close range and buys the entire far field. This is why fixed-focus body worn cameras are focused this way, and why a camera that has drifted closer than its hyperfocal distance is the single most common reason for a “soft” far-field image.

There is a second reason, specific to this product category. Evidence footage is mostly examined for identification — a face, a plate, a reflection in a window. Those subjects are usually well away from the officer, not on the officer’s chest. Optimising for the near field optimises for the part of the frame nobody argues about.

Field of view

The wide-angle trap: why the textbook formula understates the frame by 25%

This is the part that separates a credible field-of-view figure from a marketing one. It also decides whether the test chart you buy is the right size.

The familiar relationship between image height and angle is r = f · tan θ. It describes a rectilinear projection, and it is exact — for a lens with no distortion. Point a rectilinear lens at a tall building and straight lines stay straight, but the corners stretch, because the projection has to map an ever-larger angle onto an ever-larger radius.

A body worn camera lens cannot do that with a 100° field. To keep the image roughly proportional to angle instead of to tangent, wide lenses follow one of three other mappings. All four are in use, and they disagree sharply at the edges of the frame:

ProjectionMappingHorizontal FoVVertical FoVFrame width at 1.2 m
Rectilinearr = f·tanθ81.0°51.3°2.05 m
Equidistantr = f·θ97.8°55.0°2.75 m
Equisolid angler = 2f·sin(θ/2)101.0°55.5°2.91 m
Orthographicr = f·sinθ117.2°57.4°3.93 m

Same 3.15 mm lens, same 1/2.9″ sensor, same 1920 × 1080 pixels. Only the projection model differs.

The 25% error. If you apply the rectilinear formula to a lens that actually follows an equidistant mapping, you compute a 81° horizontal field where the lens really delivers 98°. At 1.2 m that is 2.05 m of frame instead of 2.75 m — a quarter of the picture missing from the calculation. The same error runs through every downstream number: how big the test chart needs to be, how many pixels a face receives, whether a 4 mm band at the edge of the frame is inside the specification.

In practice, lenses with roughly 20% barrel distortion — which is what most body worn camera lenses carry — sit between the equidistant and equisolid mappings. Neither is “correct” in the abstract; the right one is whichever the lens actually is. That is not something a datasheet will tell you reliably, which is why we measure it. The last section of this page shows how you can measure it too.

Resolution

How far away can it see a face?

The honest answer is a pixel count, not a distance. Given the geometry, the count at each distance follows directly.

DistanceFace width, pixelsWhat that supports
0.5 m335 pxIdentification, with detail
1.0 m168 pxIdentification
2.0 m84 pxRecognition
3.0 m56 pxDescription only
5.0 m34 pxSilhouette

Assumes a 150 mm face width, a 3.15 mm lens, equidistant projection and a 1920-pixel horizontal count. Face turned away from the camera, low light, motion and video compression all reduce the effective figure.

These bands are conventions, not a standard. Different agencies draw the line in different places. What is not negotiable is the arithmetic: at 3 m a 150 mm face occupies about 56 pixels across, and no amount of sharpening adds information that was never sampled. If your requirement is identification at 3 m, the sensor or the lens has to change — not the firmware.

This is the calculation we would rather have a customer run before a project than after a pilot. It is also why we publish it: a supplier who quotes a recognition distance without stating the pixel count, the lens and the sensor size has not told you anything you can verify.

The station

How the calibration station is set up

Once the lens is characterised, the station is solved backwards from a target: how much of the frame should the test chart fill?

A chart fills too little of the frame and the operator is judging focus from too few pixels; fills too much and the chart runs into the extreme edges of a wide-angle image, where distortion is at its worst and straight edges are no longer straight. The working band we use is 35% to 60% of frame height, with anything above 78% treated as out of frame.

For a given chart, the distance that produces a target fill is found by inverting the projection, which has no closed form for all four models — so it is solved numerically. With a 420 mm chart height and the lens above, 60% fill lands at 0.71 m. But the hyperfocal distance requires 0.77 m. Taking the larger of the two gives 0.77 m, where the chart fills 55% of the frame — inside the band, and past the hyperfocal distance.

When the two requirements conflict. If the distance that fills the chart to 60% is shorter than the hyperfocal distance, the answer is a larger chart, not a shorter distance. Shrinking the distance to suit an undersized chart is how a camera ends up focused in front of its hyperfocal point, which is exactly the failure mode this whole page is about avoiding.

Calculator

Run the numbers for your own configuration

Change any input and the results update. The default is a typical 1080p wide-angle body worn camera; the page works without scripting too, and carries the default results as static values.

Inputs

mm
f/
× px
mm
m

Projection model

Not sure which one your lens is? Measure it — see the tape-measure method below.

Runs entirely in your browser. No libraries, no network requests, nothing sent anywhere. You can save this page and use it offline.

Recommended focus distance

0.77m

The hyperfocal distance for this lens. Focused here, 0.39 m to infinity stays within the depth of field.

Pixel pitch2.80 µm
Blur circle5.60 µm
Hyperfocal distance0.77 m
Near limit at H0.39 m
Horizontal FoV97.8°
Vertical FoV55.0°
Diagonal FoV112.2°
Frame at 1.2 m2.75 × 1.25 m
Chart fill at 1.2 m36.1 %
Chart height in pixels390 px

Chart fill against distance — green band is the 35–60% working range

Chart fill 35–60% working range Hyperfocal distance

Frame size and chart fill across distance

Narrow screen — swipe the table sideways to see every column.

DistanceFrame widthFrame heightChart fillChart pixelsStatus
0.5 m1.15 m0.52 m82.8 %895 pxOut of frame
0.8 m1.83 m0.83 m53.5 %578 pxUsable
1.0 m2.29 m1.04 m43.1 %466 pxUsable
1.2 m2.75 m1.25 m36.1 %390 pxUsable
1.5 m3.44 m1.56 m29.0 %313 pxToo small
2.0 m4.58 m2.08 m21.8 %235 pxToo small
3.0 m6.88 m3.12 m14.6 %157 pxToo small

Face width in pixels across distance

Narrow screen — swipe the table sideways to see every column.

DistanceFace widthWhat that supports
0.5 m335 pxIdentification, with detail
1.0 m168 pxIdentification
2.0 m84 pxRecognition
3.0 m56 pxDescription only
5.0 m34 pxSilhouette

Verification

How to check the field of view of a camera you already have

You do not have to take our word for any of the figures above, and you should not take another supplier’s word either. A tape measure settles it in about five minutes.

  1. Set the camera up at a known distance. Clamp it at the position it would occupy on a wearer, and measure the distance from the lens front to a blank wall. One metre is convenient; write down the exact figure.
  2. Mark the horizontal edges. Put a strip of tape on the wall at each side of the live image, exactly where the picture ends. Measure the span between the two strips.
  3. Repeat vertically. The top and bottom edges are harder to judge by eye — if a chart happens to fill the frame vertically, measure the chart instead. Otherwise use a straight edge held up to the image.
  4. Feed the two measurements back into the calculator above. Compare the resulting field of view with what the datasheet claims. A gap of more than a few percent means the datasheet is using a different projection, or a different definition of the frame.
An inconsistency worth watching for. If a specification quotes both horizontal and vertical field of view, their ratio is fixed by the aspect ratio alone — for a 16:9 sensor the two figures should always divide to about 1.78, whatever the focal length. A datasheet where that ratio varies between lenses has measured each one on a different basis, and cannot be used to compare them. Check this before you build a station around a number.

Questions

Common questions

What is the hyperfocal distance of a body worn camera?

For a typical wide-angle body worn camera — around 3 mm focal length at f/2.3, with a 2 MP sensor of 2.8 µm pixel pitch — it works out to roughly 0.77 m. Focused there, the depth of field runs from about 0.39 m to infinity, which is why a subject at 1 m and a subject at 30 m can both be acceptably sharp.

How far away can a body worn camera identify a face?

It is a geometry question, not a marketing one. With a 3.15 mm lens and a 1920-pixel-wide sensor, a 150 mm face occupies about 168 pixels at 1 m, 84 at 2 m, 56 at 3 m and 34 at 5 m. Identification is generally treated as reliable from about 120 pixels upward, recognition from 80, and description only below 60. Light, motion and compression all reduce the effective figure.

Why does a wide-angle lens need a different field-of-view formula?

The familiar formula assumes a rectilinear projection, which only holds for narrow lenses. A body worn camera lens with roughly 20% barrel distortion follows an equidistant or equisolid mapping instead. Applying the rectilinear formula to it understates the horizontal field by about 25% — 2.05 m instead of 2.75 m at 1.2 m in the example above.

Does focusing past the hyperfocal distance hurt close-up image quality?

Only slightly, and only below the near limit — 0.39 m for the example lens. In exchange the entire far field is sharp. For evidence work that trade is almost always the right one, which is why fixed-focus body worn cameras are set this way rather than focused on the subject in front of them.

Can I use this calculator for a different lens or sensor?

Yes — enter the focal length, aperture, sensor dimensions and pixel count directly, and pick the projection model you measured. The formulas are the standard ones and are not specific to our products. If you would rather we ran the numbers for a specific configuration, send us the details and we will.

Next step

Have a lens specification you want checked?

Send us the sensor, focal length and the field-of-view figures you have been quoted. We will tell you which projection those figures are consistent with — and whether they can support the recognition distance your project needs.

Sales and projects

sales@iotkam.com

Pricing, samples, lead time and shipping terms.

OEM and ODM

oem@iotkam.com

Custom optics, firmware adaptation, SDK and integration.