Inspect glue on a reflective metal surface: dome or dark field

A white glue bead on a polished metal lid is one of the harder routine inspections to light. The metal mirrors whatever you aim at it, so two bar lights land as two glare bands. The glue you must measure hides between them. More intensity, a faster lens or a filter alone will not rescue that image; the fix is geometry. This guide walks the decision in the order you meet it: why bars fail on a mirror, which geometry restores contrast — dark-field bars, a dome, coaxial light or crossed polarizers — then the camera and lens arithmetic for a typical lid, and a bench checklist built on the worst sample you will ever pass. By the end you can name the light for your field size and surface shape, with the numbers to defend it.

Why do two bar lights wash out a shiny metal surface?

Because polished metal is a mirror: each bar you mount prints into the image as a bright band, the band's position follows the lid's tilt, and the glue between bands never sits on a stable background, so a fixed threshold false-rejects.

The law of reflection in bright and dark field decides the whole problem: the camera sees a light source only when source, part and lens sit at equal angles. A bar is a long, narrow source, so its reflection is a long, narrow band. Tilt the lid one degree and the band walks twice the standoff times the tangent of the angle — about 10 mm at 300 mm, since 2 × 300 mm × tan 1° = 10.5 mm.

Two bars meant to fill each other's shadows instead deliver two wandering bands. Hand-matching their angles, distances and drive levels loses on any line with part-to-part tilt. The ring light vs bar light comparison still holds for matte parts; on a mirror, no directed ring or bar wins. Brushed or plated metal that scatters a little sits in between, and the copper surface lighting guide covers that middle ground. Specular surfaces defeat presence sensors the same way, and detecting a veneer leading edge on a shiny white belt works the identical geometry from the sensor side.

What light geometry shows white glue on reflective metal?

Dark field or a dome — both leave matte white glue bright and polished metal dark — and crossed polarizers do the same when neither fits; coaxial light inverts the contrast and only suits small, flat fields.

You win by controlling where the specular return goes, not by adding intensity. Four routes exist, and each handles the mirror path differently: exclude it with low-angle light, surround it inside a diffuse dome, cancel it with crossed polarizers, or use it deliberately with coaxial light. The white bead is matte, so it scatters toward the lens under every one of them. The light geometry selection guide maps the same choice across surfaces and features; the next three sections apply it to glue volume on metal.

How does dark-field lighting turn the metal black and the glue bright?

Skim the light in at 10 to 20 degrees from the lid's plane: the flat metal reflects it onward and away from the lens, while the raised, matte bead scatters light upward, so glue becomes the only bright feature in the frame.

Start with two bar lights on opposite sides of the bead line, both at the same standoff — 150 to 250 mm is a workable starting range — and both at the same grazing angle. A bead 1 to 3 mm high lights well at 10 to 20 degrees; flatter features need a steeper 30 to 45 degrees before they scatter enough light back. Because the lit faces of the bead grow with its height, brightness itself becomes a volume cue: a thin bead reads dim and narrow, a full bead reads bright and wide.

Two stock bars do the job: the LS-BAR-1047 at 290 x 30 mm and the LS-BAR-0400 at 312 x 30 mm.

The same grazing logic drives engraved text lighting on metal and low-angle dent and pit detection; for strobing on moving product, low-angle bars on moving boards shows the timing discipline.

When does a dome light beat dark field?

When the surface curves, wrinkles or arrives tilted: the dome surrounds the lid with light from every direction at once, so no single angle can black out or blow out the metal, and the bead still scatters its own bright outline.

A flat lid under a dome shows the camera the reflection of the dome's own camera hole — a dark patch — so flat polished metal reads dark while the white bead scatters the bright dome walls. A curved lid instead mirrors a smooth sweep of dome surface and reads as an even gray. Either way the hot bands are gone, which is exactly what the two-bar set-up could not deliver.

Size the dome, as a rule of thumb, at 1.5 to 2 times the field of view; dome lights for shiny, curved and wrinkled parts works the sizing in detail. The dome lights category keeps 34 of its 143 units in the 400 mm and larger band, so large and elongated lids stay in range.

The LS-DOM-0062 is a dome with a 194 mm outer diameter.

That diameter lights a flat lid up to about 130 mm across under the same rule of thumb. Elongated parts do better with a long format such as the LS-DOM-0001-L4, 237.2 mm with a 629.9 mm length. One known artefact: on parts that throw the hole's reflection onto the bead, the camera-hole dark spot fix covers the tilt and aperture moves.

Is coaxial light an option, and how big can it go?

Yes, on flat lids, with the contrast inverted — polished metal returns the axial beam and glows while matte glue scatters it away and reads dark — but plan on nothing above a 400 mm field.

A coaxial unit folds the light through a beam splitter straight down the lens axis, so flat, square metal returns a bright, even field and the glue prints as a dark region — the steadiest area measurement of the four routes. The same mirror law caps it: every degree of tilt shifts the return by twice the path times the tangent, about 5 mm per degree on a 150 mm path (2 × 150 mm × tan 1° = 5.2 mm), so hold the lid square within two or three degrees. Curved lids need not apply.

The beam splitter must also be as large as your field, which is why supply thins with size: only two of 260 units in the coaxial lights category sit in the 400 mm and larger band, against 107 in the 50-100 mm band. Cross coaxial off above 400 mm and move to dark-field bars or a large dome. Two sizes bracket the choice:

Values as printed in each product's data sheet; open a part for its drawing and full specifications.

The in-line illumination guide covers the same axial route on other specular parts.

How do you size the camera, lens and exposure for the glue bead?

Start from the reject width, not the camera: for a 2.0 mm minimum bead on a 260 mm field, a 5 MP camera with a 16 mm lens at a 508 mm working distance puts the bead across 19 pixels, with depth of field to spare.

Take one pass with all the arithmetic shown. The part is a 250 x 180 mm lid with a nominal 3 mm bead and a reject rule of a bead under 2.0 mm anywhere. Add 5 mm of margin per side: field of view 260 x 190 mm. The camera is a 2/3-inch-class 5 MP unit, 2448 x 2048 pixels on a 3.45 µm pitch, 8.45 x 7.07 mm of active area.

Magnification first: m = 8.45 mm / 260 mm = 0.0325. Focal length at a 508 mm working distance: f = 508 mm × 0.0325 / 1.0325 = 16.5 / 1.0325 ≈ 16.0 mm, so a stock 16 mm lens. Check it back: m = 16 / (508 − 16) = 0.0325, and the field of view comes out at 8.45 / 0.0325 = 260 mm. Pixel footprint: 3.45 µm / 0.0325 = 106 µm, so the 2.0 mm reject bead spans 2.0 / 0.106 ≈ 19 pixels, and a 0.5 mm change still spans 4.7 pixels.

Depth of field at f/5.6, allowing a two-pixel circle of 6.9 µm: 2 × 5.6 × 6.9 µm × 1.0325 / 0.0325² ≈ 76 mm, or about ±38 mm. A bead 1 to 3 mm above the lid never leaves focus, so the aperture is free — stop to f/5.6 or f/8 for sharpness across the lid.

If the line runs at 200 mm/s, cap motion blur at half a pixel: 0.5 × 0.106 mm = 0.053 mm, so exposure stays at or under 0.053 / 200 s = 265 µs; run 250 µs, and blur = 200 mm/s × 250 µs = 0.05 mm. A large diffuse dome delivers fewer photons per microsecond than a directed bar, so short exposures usually need the light strobed overdriven from a controller on the camera trigger.

The focal length and field arithmetic above is exactly what the finder-backed calculator runs:

Which lighting geometry fits which glue job?

Set it by field size and surface shape: raised beads take dark-field bars at any width, curved or tilted shiny parts take a dome, flat flush smears under about 150 mm take coaxial light, and past 400 mm coaxial is out of the question.

Three cases flip the usual answer. A flush smear has no height, so dark field sees nothing — switch to dome or coaxial and measure area. A curved lid defeats coaxial entirely, because the return never re-enters the lens, while a dome absorbs the curvature. A tall bead at 10 degrees shadows its own base, so steepen toward 20 to 30 degrees or add a third bar.

Crossed polarizers deserve their row: one sheet on the light, one on the lens, and the metal's glare cancels while the bead's depolarized scatter passes — at the price of roughly a fifth of your light. The cross-polarizer set-up guide covers the alignment.

What should you check before you order the light?

Bench-test the worst lid you will ever pass, not a good one: the least glue, the most tilt and the dullest metal set your margins, and one dome plus one bar settles it in an afternoon.

• Write the reject rule in millimetres first — minimum bead width or minimum covered area — because that one number fixes pixels, geometry and light.
• Pull the worst samples: minimum glue, ±2 degrees of tilt, the dullest and the brightest metal batches.
• Match the bars: equal standoff and equal angle within a degree, confirmed on a flat matte-white target before any metal goes on.
• Fit the dome at 1.5 to 2 times the field of view, and check the fixture clears the dome mouth.
• Prove the exposure at full line speed: strobe width under the blur budget, controller on the camera trigger.
• Lock out ambient light: shroud the station, or run red bars behind a matching bandpass filter on the lens.
• Decide the metric: bead width at 10 to 20 stations, with alarms on the mean and on any gap.

The light finder filters by size and colour, and the configurator below walks the whole station from camera to stand:

If the sample is awkward, send the lid and the reject rule through the contact page and the applications team will shoot it before you commit.