A polished right-angle prism behaves like three mirrors bolted together. Whatever lamp you fit, only the rays the law of reflection steers into the lens come back, which is why the sloped 45° face and the vertical legs go black under an ordinary top light. The fix is a placement rule, not a brighter lamp: put the light at the mirror image of the camera across each face. This guide turns that rule into numbers — which face takes a coaxial light, how much tilt a dome covers, why a 45° face needs a tilted or folded camera — and works the arithmetic on a 20 × 20 × 50 mm prism you can repeat with your own part.
Where must the light sit on a mirror-polished prism face?
Place the light at the mirror image of the camera across that face: a polished face returns each ray at the angle it arrived, so a face tilted α from square to the lens sends light back only when it arrives at 2α from the camera axis. Work every face through that sum before you pick any hardware.
| Face tilt α | Rays must arrive at | What can deliver them |
|---|---|---|
| 0° | 0° | A coaxial light on the camera axis |
| 15° | 30° | A ring or dome with enough radius at the working distance |
| 30° | 60° | A wide dome only |
| 45° | 90° | Level with the bench: nothing above the part qualifies |
| 90° (vertical face) | 180° | No solution from above at all |
The same reflection law that separates bright field from dark field lighting decides the placement here; only the bookkeeping changes, because you now chase the single angle that returns to the sensor instead of avoiding it.
Which light suits the face that stares straight into the lens?
Use a coaxial light: its beamsplitter folds an LED panel onto the optical axis, so a face square to the lens sees a uniformly bright patch and returns it straight into the sensor, with no fixture shadow, as long as the emitting area exceeds the field of view.
That is also the one geometry that keeps camera and light on a single axis, which is what lets a folded or tilted camera reuse it — remember that for the 45° and vertical faces below. Two details decide whether the image is clean. First, the emitting area must exceed the field of view; the sizing section below works the margin. Second, check that the emitter is diffused before the beamsplitter, because a bare LED array acts like a point source and prints a hotspot on specular faces — the failure dissected in coaxial light hotspot on a specular surface. The coaxial light range runs from emitting faces under 50 mm to 400 mm and above, so sizing rarely forces a custom build.
What covers faces tilted slightly away from square?
A dome: it lights from a wide annulus of angles at once, so with emitting radius r sitting h above the part it covers tilts up to α_max ≈ atan(r/h)/2 — about ±28° for a Ø116 mm emitting area at a 40 mm working distance.
Run the sum on a real dome. The LS-DOM-0085 is a Ø166 mm dome with a Ø116 mm emitting area. Half of that is a 58 mm emitting radius, so at a 40 mm working distance the steepest usable ray arrives at atan(58/40) = 55°, giving 55/2 ≈ ±28° of tilt coverage. Lower the dome to a 25 mm working distance and the same geometry delivers atan(58/25) = 67°, or ±33°.
- Size
- 100-200 mm
- Voltage
- 24 V
That annulus behaviour is the same property that makes dome lights work on shiny, curved and wrinkled parts, so chamfers, curved bevels and engraved marks on a prism sit comfortably inside a dome's range. For a bigger part, the dome lights range climbs from under 50 mm to 400 mm and above, so a larger prism gets a larger annulus rather than a custom build. One geometry note: the camera hole at the top blocks the steepest near-vertical rays, and if the face images that hole you are seeing the dark spot from the dome's camera hole, not a lighting failure.
Why does the 45° face stay black under a top camera?
Because the mirror law asks for rays arriving at 2 × 45° = 90°, level with the bench: no ring, bar or dome above the prism can produce them, so the face returns nothing and images black apart from scattered scratches and dust.
The geometry is unforgiving — the required radius WD × tan(2α) runs away as α approaches 45° — so extra emitter area never fixes it. Three setups do:
- Tilt the camera. Mount camera and coaxial light together on a bracket square to the hypotenuse. The face becomes the face-on case and images uniformly.
- Fold the path. A front-surface mirror at 45° turns the vertical axis toward the face. Folding preserves angles, so the coaxial light above still works unchanged.
- Shoot it dark field. If the job is finding chips and scratches rather than uniform imaging, keep the camera vertical and fire low-angle light in level with the bench. The polished face stays black; rough defects scatter and glow.
How do you light the vertical side faces?
From above, never: the mirror image of a top camera across a vertical face sits at the same height beside the part, so the bright-field answers are a second camera square to the face or a 45° folding mirror that turns a horizontal view up into the top axis.
The folded path keeps everything on one axis. Light leaves the coaxial above, steps down through the beamsplitter, bounces off the fold mirror, crosses to the vertical face and returns along the same route, so the face behaves exactly like a face-on part. Two mirrors fold two different faces into one image — the pattern behind imaging two cube faces with one camera and mirror optics.
A dedicated side camera with its own small coaxial light is simpler to align if the cell has room. Choose by counting views: when one camera stops covering the visible faces, mirrors save cameras and cell height, while extra cameras save alignment time.
How much emitter area does the prism actually need?
Size every light from the field of view, not from the part: the emitter must fill the field of view with margin for fixture misalignment, so plan on about 1.5× the field-of-view diagonal — 90 mm for the 52 × 30 mm view below.
The part is a right-angle prism with 20 × 20 mm legs, 50 mm long, and the face under inspection is the hypotenuse: 50 mm long × 28.3 mm tall, since √(20² + 20²) = 28.3 mm. Camera and coaxial light ride a bracket square to that face.
Give the face a 52 × 30 mm field of view with fixture margin. On an 8.8 mm wide 2/3-inch sensor, m = 8.8 / 52 = 0.169, and at a 60 mm working distance the focal length is f = 60 × 0.169 / 1.169 = 8.7 mm. The view diagonal is √(52² + 30²) = 60 mm, so the 1.5× margin asks for a 90 mm emitter. Treat the 1.5× as a starting margin for alignment error and beamsplitter aperture; tighten it once the fixture holds position.
That points to the LS-CXL-0067. Its 102 × 100 mm emitting face clears the requirement with margin on both axes.
- Size
- 100-200 mm
- Voltage
- 24 V
A smaller prism scales the same arithmetic down. A 26 × 16 mm field of view has a diagonal of √(26² + 16²) = 30.5 mm, so the emitter target is 1.5 × 30.5 = 46 mm. That fits the LS-CXL-0027, a 60 × 60 mm coaxial light drawing 4.4 W in red or 7.2 W in white.
- Size
- 50-100 mm
- Voltage
- 24 V
Which light wins for each face of the prism?
Match the light to the tilt: coaxial at 0°, a dome to roughly ±28°, a tilted or folded camera with coaxial for the 45° face, fold mirrors or a side camera for vertical faces, and low-angle grazing when you are hunting defects rather than imaging uniformly.
| Face in a top-view cell | Tilt α | Rays needed (2α) | Decision |
|---|---|---|---|
| Leg face square to the lens | 0° | 0° | Coaxial light on axis |
| Chamfers, bevels, engraved marks | up to ±28° | up to 56° | Dome sized by α_max = atan(r/h)/2 |
| The 45° hypotenuse face | 45° | 90° | Tilted camera plus coaxial, or a 45° fold mirror |
| Vertical leg faces | 90° | 180° | Fold mirrors into the top axis, or a side camera with its own coaxial |
| Chips and scratches, any face | any | dark field | Low-angle grazing from the side: face dark, defects bright |
The tempting answer — put one big dome over the whole prism — fails exactly where prisms are hardest. Above about ±28° of tilt and always at 45°, the dome cannot deliver the return angle: at a 40 mm height the required radius grows from about 58 mm at 2α = 55° to 40 × tan(60°) = 69 mm at 2α = 60°, and the demand runs away entirely as 2α approaches 90°.
Where scatter must be separated from the specular return — micro-scratch contrast, engraved marks — cross polarizers on the lens and the light remove the reflected glare while rough features keep their brightness. For defect hunting on the sloped and vertical faces, low-angle shadowless lights are the tool: the LS-SDL-0067 is a 113 mm low-angle shadowless light that floods the part with near-horizontal diffuse light.
- Size
- 100-200 mm
- Voltage
- 24 V
The wider placement framework — camera, light and part positions as one contrast budget — is in lighting geometry for image contrast.
What do you check before ordering the lights?
Check four numbers before you order: emitter area against the 1.5× diagonal rule, coverage angle against each face's 2α, strobe rating for the shortest exposure you need, and mechanical clearance for the beamsplitter, dome or fold mirrors above the part.
- Emitter area. 1.5 × the field-of-view diagonal, per view: 90 mm for the 52 × 30 mm example, 46 mm for the small one.
- Coverage angle. Check WD × tan(2α) against the emitting radius for every tilted face; the 58 mm radius at a 40 mm height covers 2α up to about 55°.
- Wavelength. White light to start; move to 850 nm or 940 nm with a matched filter on the lens when plant lighting leaks into the image.
- Strobe discipline. One controller channel per view, fired from the camera's exposure output so light and exposure cannot drift — the pattern for stabilizing lighting changes with camera triggers.
- Mechanics. Beamsplitters, domes and fold mirrors all add height; confirm the cell clears them at the working distance you computed.
Send the per-face table — tilt, size, finish — through the lighting finder, or start the configurator at the light step to match the parts above to your own prism. The lighting topic page gathers the neighbouring geometry guides.
Questions
Can one camera inspect all three rectangular faces of a right-angle prism at once?
No single axis is square to two perpendicular faces, so one top camera images at most one face brightly. Fold both side views into the vertical axis with 45° mirrors and image them with the top camera and its coaxial light, or add cameras square to the remaining faces. Two to three views is the practical count for the three rectangular faces.
Why is the prism image bright in the center and black at the edges?
The lit patch on the face is smaller than the field of view, so the outer region of the face receives nothing to return. Size the emitter to at least 1.5× the field-of-view diagonal and confirm the beamsplitter's clear aperture covers the full sensor format, then re-check with the part at the edges of its fixture tolerance.
What wavelength works best on polished glass or metal prism faces?
Start with white: it shows scratches, chips and fingerprints on polished faces without wavelength-specific surprises. Switch to 850 nm or 940 nm infrared with a matched filter on the lens when plant lighting contaminates the image, because the filter blocks ambient visible light while the infrared illumination stays bright to a mono camera.
Do I need a strobe controller for a prism lighting cell?
Yes when faces are imaged sequentially: a multi-channel controller fires only the light for the face in view on each exposure, keeps duty cycle low and holds brightness stable across thousands of cycles. Take the strobe signal from the camera exposure output so light and exposure cannot drift apart, and set per-channel current to balance gray levels across views.
How close should the dome or coaxial light sit to the prism?
As close as focus allows, because coverage angle improves as the light nears the part: a Ø116 mm emitting area reaches tilts of about ±28° at a 40 mm working distance and about ±33° at 25 mm. Keep the lens's depth of field across the face depth when you shorten the height.



