Lighting solder joints on brass: dome plus coaxial fill

Four solder joints on a polished brass pad are among the hardest everyday lighting jobs in a vision cell. The pad is a mirror, its surface carries dents from the welding pressure, and every joint is a curved cap that images whatever light you point at it. One ring or spot light gives a glare band, black corners and joints you cannot separate from the dents. The geometry that works is a dome light, diffuse illumination from a surrounding hemisphere, with a coaxial light stacked above it to fill the dark circle the dome's camera hole casts at the center. The sections below size both lights, pick the lens and f-number, decide when blue light helps, and list the checks that catch overexposure and off-center imbalance before they reach production. The worked example uses a 40 mm pad; scale every number to your own part.

Why does a shiny, dented brass pad defeat a single light at a single angle?

Because mirror-bright brass obeys the law of reflection point by point: each small patch of surface returns light from exactly one direction into the lens, so one LED at one angle lights one stripe of pad and leaves the rest near black.

A specular surface is a mirror of its light source. Under a single ring light, a flat brass pad shows one bright annulus where the ring's reflection reaches the lens and darkness everywhere else; curvature and welding dents rotate each patch's normal, so that annulus breaks into blobs that move with every part you set down. The joints are convex caps: each one images the light source itself, so the blob you see has the lamp's shape, not the joint's.

Your task raises the bar. Locating four joints and judging whether each is centered on its pad position is a measurement, so the background must be uniform and the joint edges sharp, not merely visible. Single-angle lighting fails both requirements at once. The same law of reflection decides between bright field and dark field lighting on flat parts, and on flat circuit boards the reflections show up differently, as the solder pad inspection lighting guide shows. For what separates a good joint from a cold one electrically, read troubleshooting cold solder joints on industrial PCBs; this guide stays with the optics.

Which light shows solder joints evenly on a curved, reflective pad?

A dome light: its emitting surface surrounds the pad with diffuse light arriving from nearly every direction at once, so every surface slope on the brass and on each curved joint returns almost the same brightness to the camera.

The dome is the machine-vision version of an overcast sky. Each point of the pad sees a wide cone of emitting surface instead of one ray, so what it reflects is the dome itself, broad, even and independent of local slope. Dents stop being features: a dent only changes which part of the dome a patch images, and all parts of the dome are equally bright. The convex joints reflect the dome straight back into the lens and appear as smooth, bright caps with clean edges against the uniform field. That combination, a shadow-free background under evenly lit caps, is why dome lights lead reflective and curved-part applications; the guide to dome lights for shiny and curved parts works the same physics for wrinkled and blistered surfaces.

Two limits follow from the dome's own shape, and they decide whether the dome alone is enough: its size relative to the field of view, and the hole in its top.

Why is there a dark circle at the image center, and what fills it?

The camera hole: a dome needs an aperture for the lens, and the patch of pad directly beneath it receives almost no direct light, leaving a soft dark disc roughly the hole's own diameter across the middle of the image.

A joint that lands in that disc is a joint you cannot measure. The fix that keeps every dome advantage is a coaxial light mounted above the dome, sending light down through the hole along the camera axis. The fill lifts the shadow zone to the level of the surrounding diffuse field and adds a slight on-axis emphasis to whatever sits dead center, which usually sharpens a center joint's edges. Keep it weak; its job is to erase the disc, not to become a second light source. The guide to fixing the dark spot from a dome light's camera hole collects the alternatives, a larger dome, a moved part, an off-axis camera, for cells where a coaxial stack will not fit. Browse coaxial lights for the fill unit; a compact narrow-angle model keeps the beam on axis without crowding the camera:

How big must the dome be, and which lens goes with it?

Size the dome first: keep the emitting area at least twice the field of view, so a 56 mm field of view needs 112 mm of emitting diameter, which the common 116 mm emitting size meets.

The rule keeps the pad's edge points looking at a wide cone of emitting surface. A pad that fills more than about half the aperture shades its own rim, and brightness rolls off toward the image edges. Work the numbers on a concrete station:

• Brass pad 40 mm across, four joints about 5 mm each, 40 percent margin for locating the pad in the frame: field of view = 40 × 1.4 = 56 mm.
• Emitting area at 2 × 56 mm = 112 mm minimum.
• Camera: 1/2 inch format, 6.4 mm sensor width, 1440 pixels across, 4.4 µm pixels.
• Lens: 25 mm focal length.

The 25 mm lens puts the camera about 244 mm above the pad, enough clearance for the dome and the coaxial unit stacked over its hole. Each pixel then covers 0.039 mm of brass, a 5 mm joint spans about 128 pixels, and the ±0.5 mm centering tolerance you judge is ±13 pixels: comfortable margins for edge-based measurement.

The dome used in this example clears the rule with margin:

Stop the lens down for the dents. The pad's surface and the joints' caps span about ±3 mm of height. Total depth of field is approximately 2 × f-number × circle of confusion × (1 + m) / m², with the circle set to two pixels = 8.9 µm:

• At f/2.8: 2 × 2.8 × 0.0089 × 85.7 = 4.3 mm total, ±2.1 mm, which fails the ±3 mm requirement.
• At f/5.6: 2 × 5.6 × 0.0089 × 85.7 = 8.5 mm total, ±4.3 mm, which passes, and the Airy disc of 2.44 × 0.55 µm × 5.6 = 7.5 µm stays inside the 8.9 µm tolerance, so diffraction does not eat the resolution.

Set f/5.6 and move on. For exposure: if the station stops the pad, 2 to 10 ms continuous works. If the pad indexes past at 100 mm/s and blur must stay under one pixel, the shutter has 0.039 mm ÷ 100 mm/s = 0.39 ms, so strobe both lights with a 0.3 ms pulse and the blur stays below one pixel with margin.

Three domes bracket the usual brass-pad stations; the emitting-area rule picks between them:

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

Does blue light add contrast on yellow brass?

Yes, and the gain is measurable: brass reflects red and green strongly but blue poorly, while a silvery solder cap reflects every band about equally, so blue light can drop the pad from around 200 toward 80 gray levels on an 8-bit camera while the joints hold near 200.

That is the complementary-color rule at work: light a surface toward the color it reflects least and you darken it against everything neutral nearby. Yellow's true complement is violet; industrial LED ranges stock the nearest practical band, blue, so order the dome and the fill in the same blue, because a blue dome with a white fill rebuilds the brightness hierarchy you just removed. Both the dome and the coaxial ranges list blue models.

Decide by the pad, not by the metal's name. Bare yellow brass takes blue light; a silver or nickel-plated pad with silvery joints gains almost nothing from blue, because both surfaces return it alike, so keep white and lean on the dome's geometry. Pair blue light with a monochrome camera so every pixel contributes. The LED color contrast guide tabulates this by surface, the article on monochrome cameras and light color quantifies the sampling penalty on color cameras, and copper surface inspection lighting applies the same trick to copper, the other yellow-red metal.

Why is one corner brighter, and how do I stop the overexposure?

Two faults, two fixes: an off-center pad brightens the corner nearest the dome wall because it sees unequal solid angles of emitting surface, and overexposure means the strobe pulse or drive is simply too strong, so recenter first, then trim the light.

Measure instead of eyeballing. Draw four 20 × 20 pixel regions on bare brass near the image corners and read their means. The diffuse field should sit at 150 to 180 on the 8-bit scale, the four corners should agree within ±10 gray levels, and joint highlights should stay below 250 so fillet edges never clip. If the corners disagree, the pad is off the dome axis: recenter it and confirm the dome sits level. If one whole side runs brighter than the other, the pad or dome face is tilted, so shim it square.

Trim exposure in this order: shorten the strobe pulse, then lower the drive current, and only then stop the lens down. The f-number is already spent on depth of field at f/5.6, so pulse width is the free control. Keep dome and fill on one trigger so their ratio holds shot to shot. The lighting geometry guide covers how camera, light and part positions trade brightness against contrast.

When is a flat coaxial light or a diffuse ring light the better choice?

Choose a flat coaxial light alone when the pad is flat and smooth and you also read engraved marks on it, because it stacks far lower than a dome with its fill, and a diffused ring light when headroom rules the dome out and you only judge presence.

A coaxial unit folds the light through a beamsplitter onto the lens axis: a flat specular pad returns it straight back, and any patch tilted more than the unit's few-degree acceptance angle goes black. That behavior is perfect for engraved characters in recessed metal parts and exactly wrong for welding dents, which would print as dark halos the dome would have erased. So the decision line is surface shape: flat pad with marks to read means coaxial alone; dented or curved pad with joints to measure means dome plus fill. Tall parts end the dome discussion too, because anything standing well above the pad blocks the dome's own light and shadows its base, and four-direction bar lighting is the better tool there.

A diffused ring light with a diffuser plate is the shallower, simpler substitute: partial multi-angle coverage, enough to find joints and report positions, but a less uniform field at the center and real sensitivity to part height, since lifting the pad 5 mm shifts the balance where a dome barely notices. Use it for presence checks, not for the ±0.5 mm centering judgment. Sizing lives in the ring light selection guide. Cross polarization, the other stock glare answer, deletes the specular return you are actually measuring on a joint cap and costs most of the light; the cross-polarization guide shows the flat-part problems it does solve. For presence-only checks, start with a diffused ring light that carries its own diffuser plate:

What do I check before ordering the lights?

Run this eight-point check on the worst pad you can pull from production, not the cleanest, and in this order, because each point closes one failure mode from the sections above.

• Field of view and dome size. Pad width × 1.4 for margin, then emitting area at 2 × field of view. Do this arithmetic before the dome order, not after.
• Color decision. Bare yellow brass takes a blue dome and blue fill; a silver pad takes white light and geometry.
• Camera-hole overlay. Draw the hole's dark disc on the joint pattern; any joint inside it puts the coaxial fill in the order, in the same color as the dome.
• Two channels, one trigger. Dome and fill on separate driver channels, strobed together, trimmed until the center disc matches the field within ±10 gray levels.
• Lens arithmetic. Magnification, working distance and f-number from the formulas above, f/5.6 at about 244 mm for the example, before the lens order.
• Corner balance. Four corner regions within ±10 gray levels, field at 150 to 180, joints peaking below 250.
• Motion budget. Strobe width below one pixel of travel, 0.3 ms at 100 mm/s in the example, or a station that stops the part.
• Worst-sample proof. Capture the most dented, most tarnished pad you have; if the joints segment cleanly on that image, the cell is ready.

Then build it: run the lighting step of the system configurator to pair the lights with a controller and trigger, or filter the machine-vision light finder by size and color. More geometry decisions like these sit on the lighting topic page. If your pad falls outside these numbers, bigger fields, taller joints, mixed plating, send the part details through the contact page and we will run the arithmetic with you.