Lighting for glue bead inspection on dark adhesive

Black adhesive on a metal-and-glass assembly looks like an easy inspection: a black bead across brighter materials. The lighting decides whether it stays that easy. Three different measurements share one station — bead width, coverage of the metal under the bead, and whether the bead stands proud of the glass — and each wants its own light. One light for all three gives you one good image and two mediocre ones. This guide gives the lighting decision per measurement, the numbers behind it, and a worked optics example you can rescale to your own bead width and run length.

One measurement, one light

The three checks fail for different reasons, so build the station around three lighting modes rather than one:

Black adhesive absorbs across the visible spectrum, so the bead reads dark under every color and contrast is never the problem. Geometry is: the bead is a low, curved, partly glossy ridge lying across surfaces with different finishes. Each row of the table is that physics in miniature — where the camera, the light and the part sit decides what contrasts — and the lighting geometry guide works it through step by step. For the dome-versus-dark-field argument on a similar glue-on-metal problem, the dome or dark field decision guide covers the case end to end.

Why red is the default color

The bead is black to every wavelength, so the color costs nothing on the glue and is free to spend on everything else:

• Sensor response. A silicon sensor's sensitivity peaks in the red and falls toward blue, so the same LED drive returns a stronger signal in red.
• Metal contrast. Bare metal mirrors red strongly. Under coaxial light an uncovered patch of metal glows inside the dark track, which is exactly the coverage signal.
• Cure safety. Most photoinitiators in UV-cured adhesives absorb below roughly 450 nm. A 620 nm inspection light sits outside that band and does not advance the cure.
• Ambient trimming. A red pass filter on the lens strips the non-red part of plant lighting; strobing removes what remains.

One exception earns a test: some adhesives fluoresce. Shine a 405 nm lamp on a cured bead — if the ridge glows, violet light turns the bead into a self-luminous target against a dark background, and violet 405 nm or UV 365 nm ring variants exist for exactly that job. The wider logic of picking a wavelength is in the LED color contrast guide, and the monochrome camera color contrast guide shows how the same bead shifts gray level between colors.

Worked example: optics for a 3 mm bead

Fix the image scale before choosing lights, because every number below assumes it. The worked case: a bead run of 24 × 20 mm per shot, a 3.0 mm target bead width, a ±0.05 mm width tolerance, a 5 MP camera with 2448 × 2048 pixels of 3.45 µm, and a 25 mm lens at 100 mm working distance. Substitute your print and redo the arithmetic.

Sensor size: 2448 px × 3.45 µm = 8.45 mm wide; 2048 px × 3.45 µm = 7.07 mm high.

Magnification: m = 25 / (100 − 25) = 0.333.

Field: 8.45 / 0.333 = 25.3 mm wide; 7.07 / 0.333 = 21.2 mm high — the 24 × 20 mm run fits with about 1 mm of margin per side.

Scale: 25.3 mm / 2448 px = 10.3 µm per pixel. The 3.0 mm bead spans 290 pixels; a 0.05 mm width step is 4.8 pixels, enough for a subpixel edge fit to hold repeatability in the hundredths of a millimeter. Longer run? Raise the working distance and redo the two lines.

Flat bead: coaxial light for width and coverage

Where the dispensed line crosses flat metal, a coaxial light solves width and coverage in one image. The splitter sends red light down the lens axis; flat glass and flat metal mirror it straight back, so the background saturates bright. The black bead scatters instead of mirroring and stays dark. Width is the extent of the dark band. Coverage runs on the same frame: flag any bright island inside the bead track. At the example's scale, a 0.2 mm bare spot is 19 pixels across — about 300 pixels of area, an easy blob threshold.

Size the emitting face to the field plus fixture drift. A 60 × 60 mm face clears the 25.3 × 21.2 mm field with about 17 mm of slack per side. The LS-CXL-0027 pairs a 60 × 60 mm emitting face with 4.4 W of red and 7.2 W of white at 24 V, so one light runs the red inspection and a white utility view.

Two things break the coaxial image: glare patches where the metal finish is uneven, and a bright stripe where the bead crown curves. The stripe is the next section. For uneven metal, add a polarizer on the light and a matching one on the lens — the cross-polarization guide shows the arrangement — or move the coverage check to dark field.

Curved bead: dome or diffuse ring

A dispensed bead holds a half-cylinder cross-section and keeps roughly that shape as it cools — the half-cylinder bead cooling reference explains why the profile keeps shifting with temperature. A cylinder crest mirrors on-axis light straight back to the camera exactly as flat metal does, so the coaxial width image grows a bright stripe down the middle of the bead, and width measurement dies there.

Two fixes, simplest first.

A diffuse red ring handles a gentle crown. Rays arriving at 60° light both flanks, and the diffused output spreads the crest glint into measurable noise rather than a stripe. The LS-RNG-0100 is a 120 mm dia. ring light at 60° with 620 nm red output at 24 V.

A dome handles a true round bead. Light arrives from every direction at once, so no single slope can mirror the whole beam into the lens, and the bead reads as an even dark band on a bright surround. Choose an emitting area at least twice the field width; that keeps the camera port shadow small and the arrival angles wide. The LS-DOM-0085 is a Ø166 mm dome light with a Ø116 mm emitting area. Its emitting face covers the 25 mm field more than four times over, leaving room for taller parts in the same station.

Domes trade away some center brightness: the camera hole blocks the middle of the emitting area at short working distances. The dome camera-hole fix guide lists the standard workarounds, and the dome selection guide covers sizing for curved parts in general. When a crest stripe still shows in the ring image, that bead belongs under the dome.

Height above the glass: grazing light, tilt or a profiler

No color and no diffuse panel reads height; height needs geometry. Three options, least hardware first.

Grazing dark field turns height into length. Mount a high-angle ring with its face level with the glass and the rays sweep just above the surface. Flat glass reads dark because the rays reflect onward instead of back — the law of reflection guide walks that argument — while the bead flanks scatter light and outline bright. The bead also blocks the sweep and casts a shadow across the glass behind it. Shadow length converts straight to height: s = h / tan a, with a the ray elevation above the glass.

The LS-RNG-0810 is a 120 mm dia. ring light with a 70° angle. Mounted face level with the glass, that geometry puts the rays 20° above the surface, so tan a = 0.364. At a 0.5 mm height limit the shadow runs 0.5 / 0.364 = 1.37 mm; every extra 0.1 mm of bead adds 0.27 mm of shadow. Draw the fail line 1.37 mm from the bead edge and the height check becomes one blob measurement — or invert it for an estimate: h = 0.364 × s.

The shadow trick needs a clean, flat strip of glass beside the bead. Dust, a mating boss or curved glass casts its own shadow, and the measurement inherits it.

Tilting the camera gives a parallax estimate. Turn the camera 30° off vertical and the bead's top edge shifts sideways by h × tan 30° = 0.577 × h relative to its base outline. At 10.3 µm per pixel, a 0.5 mm bead shifts 0.29 mm, about 28 pixels — readable with no extra hardware. Depth of field sets the limit:

With N the f-number, c one pixel (3.45 µm) and m = 0.333: at f/8 the estimate gives 2 × 8 × 0.00345 mm × 1.333 / 0.111 = 0.66 mm — enough for a 0.5 mm bead focused midway between glass and crest. At f/4 the same formula gives 0.33 mm, and one of the two planes blurs. Diffraction at f/8 spreads a 620 nm point to about 12 µm on the sensor, 0.036 mm on the part — still inside the 0.05 mm width budget.

A laser profiler measures the number. When the deliverable is a height profile, bead volume or process-trend data rather than pass or fail, scan a laser line across the bead and read the cross-section directly. On glossy black ridges, the shiny-metal 3D point-cloud guide applies to the same reflection problem.

Checks before you order

Cure one bead first; every choice downstream reads off it.

• Crown shape. Cut a cured bead and look at the cross-section under a desk lamp: flat crown → coaxial; rounded crown → ring, then dome if the stripe survives.
• Fluorescence. Shine a 405 nm lamp on the cured bead. A glow promotes violet; no glow keeps red.
• Run length vs field. Redo the two-line optics example with your run and camera; check the emitting face clears the field plus fixture drift.
• Glass strip. Confirm a flat, clean glass margin beside the bead, wide enough for the longest shadow you will accept.
• Exposure budget. On a moving line, divide the blur budget by line speed: one pixel at 10.3 µm per pixel and 100 mm/s means about 0.1 ms, which means strobing.
• Controller channels. Plan the red and grazing shots on separate outputs now; sequencing them later costs a rewiring visit.

When the usual answer is wrong

Four cases move the answer:

• Clear or translucent adhesive. Absorption disappears and the bead stops being the darkest thing in the image. Backlight it, or shoot coaxially through the glass from below; the transparent-substrate measurement guide covers that geometry.
• Mixed or brushed metal. Bright-field coverage thresholds break down on finish variation. Move coverage to dark field, where bare metal reads dark and the bead flanks glow.
• Tight enclosure. A dome needs clearance around the part; arch-shaped light sources give dome-style diffusion in a flatter form for cramped stations.
• Fast lines. Short exposures force strobe-type rings and sequenced shots; the exposure arithmetic above is the deciding number.

From bead to order

The station falls out of the bead: red coaxial for width and coverage on a flat crown, a red dome or diffuse ring once the crown curves, and grazing geometry for the height limit. The light finder filters coaxial, dome and ring lights by size, angle and wavelength; the light configurator steps through camera, lens and light together; and the lighting topic page collects every lighting guide on the site. If the crown shape or the height limit sits between two of the options here, send bead photos and the substrate drawing through the contact page.