Machine vision English glossary: datasheet terms explained

Machine-vision datasheets compress a lot of engineering into a compact technical English: a dozen terms per line, each with a unit and a tolerance. If photography vocabulary leaks into industrial documents, or English is your working second language, the words fail before the optics do. This glossary defines the terms a lens, camera or lighting datasheet actually prints, one line each, with the unit and the number that matters, and a worked example that turns the vocabulary into a 25 mm lens choice at 300 mm working distance.

Lens datasheet terms

Focal length (f, mm) — a property of the lens. With the sensor width it sets image scale: halve f at the same working distance and the field of view doubles. The full selection procedure is in how to choose a machine vision lens focal length.

Field of view (FOV, mm) — the scene width you capture at the working distance. The shortcut FOV ≈ sensor width × WD / f is within about 10% while magnification stays below 0.1; the blister inspection working distance and field of view formula carries the exact version.

Working distance (WD, mm) — part to lens. Datasheets print it per magnification, and every WD moves field of view and depth of field with it; a liquid lens autofocus is one answer when WD never stops changing.

Magnification (m, ×) — image size on the sensor divided by object size. A 0.5x lens puts a 20 mm part across 10 mm of sensor.

Image circle (mm) — the diameter a lens covers. It must exceed the sensor diagonal: a 2/3" sensor measures 11 mm across the diagonal, so its lens must cover at least 11 mm.

Aperture, F-number (F) — focal length divided by entrance pupil diameter. Each stop, F2.8 → F4 → F5.6 → F8, halves the light; double the exposure time to compensate.

Depth of field (DOF, mm) — the distance range that stays acceptably sharp. It follows DOF ≈ 2·N·c·(1 + m)/m², where c is the circle of confusion you accept, in machine vision usually one to two pixels (the circle of confusion page gives the photography background). At a fixed F-number, DOF falls roughly with 1/m²: doubling WD about quadruples it. The depth of field calculator runs this arithmetic.

Distortion (%) — how far straight lines bend toward the edge: barrel bows out, pincushion bows in (distortion (optics)). Measurement accuracy inherits it; how to measure barrel distortion correction accuracy shows the test.

Telecentric — magnification constant over a depth range, so the part's size reading does not change when it moves closer or further. When that matters: entocentric vs telecentric lenses: when to use each.

Worked example: 90 mm part at 300 mm

Take a 2/3" camera with a 2448 × 2048 sensor of 3.45 µm pixels (8.45 mm wide), a 90 mm part and WD 300 mm.

• Magnification from FOV: 8.45 / 90 = 0.094.
• Focal length: f = WD × m / (1 + m) = 300 × 0.094 / 1.094 = 25.8 mm, so order 25 mm.
• Check at 25 mm: m = 25 / (300 − 25) = 0.091 and FOV = 8.45 / 0.091 = 93 mm — the part fits with 1.5 mm of margin per side.
• Object-space pixel: 3.45 µm / 0.091 = 38 µm, so a 0.1 mm feature spans 2.6 pixels.
• Depth of field at F2.8 with c = 6.9 µm (2 px × 3.45 µm): 2 × 2.8 × 0.0069 mm × 1.091 / 0.091² = 5.1 mm. At F8: 5.1 × 8 / 2.8 = 14.6 mm.

Decision: a 25 mm lens covers the part at 300 mm with margin; run it near F4-F5.6 for 7-10 mm of focus latitude, and stop down to F8 only if the mechanics cannot hold about ±3 mm.

Fixed focal length machine vision lenses print most of these terms on one line, and the lens finder filters by each. The LS-FFL-0679 is a 35 mm, 6 MP C-mount lens for 2/3" sensors.

The LS-FFL-0567 prints aperture and distortion together: F2.8-22 and <0.3% distortion, at 16 mm focal length for a 1.1" sensor.

The LS-FFL-1023 prints magnification and working distance as one rating: 0.5x at WD 150 mm, F13.4.

Camera datasheet terms

Resolution (MP) — pixel count. Detail on the part follows the object-space pixel, not the MP number: 20 MP behind the wrong lens resolves less than 5 MP behind the right one.

Pixel size (µm) — the pitch of one sensor pixel. Object-space pixel size = pixel size / m; calculate object space pixel size from magnification works it through.

Sensor format (") — legacy inch names standing for a diagonal: 2/3" ≈ 11 mm, 1" ≈ 16 mm. The lens image circle must cover it.

Global vs rolling shutter — global exposes every pixel at the same instant; rolling reads row by row, so a fast part shears or wobbles. For moving parts, global.

Mono vs color — mono gives every photon to brightness; color splits them across a Bayer pattern and interpolates. Sort by color, measure in mono.

Dynamic range (dB) — saturation capacity divided by temporal dark noise, measured to the EMVA 1288 method. As a feel for the scale: 10,000 electron saturation and 3 electron read noise give 20·log10(10000/3) ≈ 70 dB, and machine-vision sensors typically land between 60 and 75 dB.

Gain (dB) — electronic amplification applied after capture. It raises signal and noise together, so it never improves the signal-to-noise ratio; reach for exposure time and light first. Automatic iris, exposure and gain control in machine vision explains the loop.

Bandwidth — resolution × frame rate × bytes per pixel, through the interface: GigE about 120 MB/s (1 Gbit/s), USB3 about 400 MB/s (5 Gbit/s), CoaXPress CXP-12 about 1,250 MB/s per link.

The camera finder filters the area scan cameras category by every term above. The MV-CS200-10UM is a 20 MP, 1" mono USB3 camera; the MV-CU020-90GC is a 2.5 MP, 1/1.7" color global-shutter GigE camera.

Lighting terms: where contrast comes from

Back light — a light behind the part; edges and holes read as a clean silhouette.

Bright field vs dark field — bright field sends light in along the camera axis, so a flat shiny surface reflects straight into the lens; dark field lights at a low angle, so only scatter from scratches and chips reaches the sensor.

Ambient light — uncontrolled room light added to every exposure. Reject it with strobing, a narrow wavelength band or shielding; the ambient light rejection specification quantifies the acceptance test.

Diffuse and dome lights — even illumination from many directions for shiny, curved or wrinkled surfaces.

Strobe duration (µs) — a microsecond-scale flash freezes motion and can replace a long exposure on a moving line.

Flare — stray reflections inside the lens that lay a veil over the image and cost contrast.

Contrast, MTF (%) — the fraction of object contrast preserved at a given spatial frequency in line pairs per millimetre; lens datasheets print the curves.

Bring the terms together as a bench test: machine vision lens and lighting selection: a test plan.

Terms that get confused

From vocabulary to an order

The definitions are stable; what varies is which rows a manufacturer prints. Before ordering, name five numbers and the terms do the rest:

1. The part and its smallest feature, in mm — everything else is derived from these two.
2. The working distance your mechanics allow, in mm.
3. The focal length that covers the field of view at that WD (the worked example: 25 mm for 90 mm at 300 mm).
4. The F-number the depth of field formula demands, not the one habit suggests.
5. The bandwidth: pixels × fps × bytes per pixel against the interface limit.

Feed the five into the configurator and it returns a shortlist; the machine vision glossary for lens choice extends this vocabulary specifically for lens rows, and the systems topic page collects the worked selections. If your vendor documents arrive in German, the German-to-English industrial automation glossary covers the PLC-side words, and a PLC programming vocabulary reference corrects the 88 terms most often mixed up once the camera reports to a controller. When the numbers argue with each other, the contact page is the next stop.