Glare is the reason most shiny-part images fail: the ring or bar light you chose reflects straight off the part and washes out the print, the seam or the defect you need to see. The reflex fix is a polarizing film on the light and an analyzer on the lens. It works, but it is the most expensive glare fix available, because every photon reaching the sensor has crossed two absorbing sheets. This guide gives you the numbers — how much light crossed polarizers pass, what that costs in f-stops, exposure and depth of field — and the decision rule for when moving the light beats filtering it. It quantifies what cross polarization for surface glare explains qualitatively.
The three transmission numbers
Put a linear polarizing film on the light and an analyzer on the lens, and three measured values describe the damage for typical plastic-film sheets used on machine-vision lights. A single film passes about 40% of unpolarized light: an ideal sheet would pass 50%, and real film loses a little more to absorption. The pair at its transmission maximum passes about 30%. Rotated to near extinction, the pair passes under 3%.
Converted to stops, where one stop is a factor of two in light:
- Single film, T = 0.40: log2(2.50) = 1.3 stops
- Pair at maximum, T = 0.30: log2(3.33) = 1.7 stops
- Pair near extinction, T = 0.03: log2(33.3) = 5.1 stops
Between maximum and extinction the pair follows the cosine-squared behaviour known as Malus's law, T(θ) ≈ 0.30 × cos²θ, with the real floor set by the film's extinction ratio rather than an ideal zero. Note the single-film line before any glare is removed: a lone polarizer already costs 1.3 stops.
Why the practical bill is 2 to 2.5 stops
The 1.7-stop figure is a best case: analyzer parked exactly at the transmission maximum with the glare already extinguished. A real part rarely allows both at once. The specular band sits at one polarization angle, so you open the analyzer past the maximum and toward extinction of that band, giving up the brightest return in the scene along with the glare. Matched-brightness tests on a flat specular lid and on a curved plastic bottle both land at 2 to 2.5 f-stops of aperture opened to bring the polarized image back to the unpolarized brightness. Budget 2 to 2.5 stops, not 1.7.
In f-numbers, k stops opened divides the f-number by 2^(k/2), and 2^(2.5/2) = 2.38:
| Starting aperture | Opened 2.5 stops |
|---|---|
| f/11 | f/4.6 |
| f/8 | f/3.4 |
| f/5.6 | f/2.4 |
| f/4 | f/1.7 |
The last row is the problem. f/1.7 is at or past the limit of many C-mount machine-vision lenses, and aberrations eat resolution wide open. If your unpolarized image already needs f/4 or smaller, crossed polarizers do not fit the optics — the fix is more light, not aperture.
Exposure or depth of field: you pay one of the two
You recover the lost factor of 5.7 in one of three ways: open the aperture, lengthen the exposure, or add LED light. The first two have hard limits.
Depth of field scales roughly with the f-number. Opening f/5.6 to f/2.4 keeps 2.4 / 5.6 = 43% of the depth of field, so a part with ±1 mm of focus margin now has about ±0.4 mm.
The depth-of-field calculator runs this for your own lens and distance.
Motion blur is the harder limit, and the worked example shows it.
Worked example: 150 mm/s conveyor, 1 pixel of blur allowed
- Object pixel pitch: 25 µm (a 2448-pixel, 3.45 µm sensor at 0.138× magnification; field of view 2448 × 25 µm = 61 mm).
- Exposure budget: t = 25 µm / 150 mm/s = 167 µs.
- Baseline without polarizers: well exposed at 150 µs and f/5.6. Blur = 150 µs × 150 mm/s = 22.5 µm = 0.9 pixels — fits.
- Add crossed polarizers and the scene loses a factor of 5.7. Exposure route: 150 µs × 5.7 = 850 µs, so blur = 850 µs × 150 mm/s = 127 µm = 5.1 pixels — fails.
- Aperture route: f/5.6 to f/2.4, keeping 43% of the depth of field — fails on parts with height variation.
- Light route: buy the factor of 5.7 back with a brighter head, a shorter working distance or a strobe pulse.
- Geometry route: move the light so the glare lobe never points at the lens. Loss factor 1.0, and the 150 µs, f/5.6 baseline stands.
On a moving line, the geometry route is usually the only one left, which is why it comes first.
Geometry first: the fixes that cost no light
A specular lobe leaves a surface at the mirror angle, and a coaxial ring — lens and light on one axis — aims that lobe straight back into the lens. Move the light and the lobe moves with it, at zero cost in stops.
- Offset or rotate the ring 10-30° from coaxial and the flat-surface glare leaves the frame. On a curved part such as a bottle, offset along the part's long axis so the lobe sweeps away over the whole silhouette.
- Drop to a low-angle ring: 0-15° emission close to the part turns the same idea into raking dark-field light. The lobe passes the lens, while edges, pits and texture still scatter light into it.
- Go transverse with a bar: mounted to the side, low, long axis across the direction of travel, a bar keeps its bright lobe off the lens from a longer working distance than a ring allows and leaves the space above the part free for handling.
Both fixes come from standard ring lights and bar lights, and you can prototype them in minutes by hand-holding the light.
Two low-angle rings cover most curved-part cases:
- Size
- 50-100 mm
- Type
- Standard
- Size
- 100-200 mm
- Type
- Standard
The LS-RNG-0256 is a 96 mm outer-diameter ring with low-angle (0-15°) emission on 24 V — the compact choice for tight stations. The LS-RNG-0730 is a 103 mm ring with a 48 mm inner aperture and emission angles from low-angle (0-15°) up to 85°, so you can tune the geometry by swapping heads instead of re-machining brackets.
For the transverse layout:
- Size
- 200-400 mm
- Type
- Standard
- Voltage
- 24 V
The LS-BAR-1047 is a 290 × 30 mm bar on 24 V, sized to graze across small and medium parts.
When polarization is still the right answer
Two tasks have no geometric substitute. Stress and molecular orientation in transparent plastics — birefringence — are invisible without crossed polarization: polarized backlight, analyzer on the lens, and strain fields appear where plain backlighting shows a uniform sheet. The other case is sheen that survives every geometry change, typical of a doubly curved part whose lobe points somewhere new at each pixel.
Then pay for the 2.5 stops deliberately. Buy light rather than gain, because a high-luminance head puts more photons through the absorbing sheets while gain only brightens noise:
- Size
- 200-400 mm
- Type
- Standard
- Voltage
- 24 V
The LS-BAR-0400 is a 312 × 30 mm high-luminance bar on 24 V, built for exactly this kind of light-starved geometry. A strobe pulse is the second lever: sized to the exposure, it buys back stops within a duty cycle the head can take — LED strobe duty cycle: matching lights and controllers works the arithmetic.
Watch the film itself. Plastic polarizer film mounted directly against lens-concentrating LEDs can burn spots in as little as 100 hours of operation, and heat plus short-wavelength light drive the breakdown. Run white or red through the film rather than blue or UV, stand the film off the concentrator, and re-check brightness over the first weeks of production; a lighting stability test before deployment catches a fading film before production does. Wavelength choice has a second footing too: which LED colour gives the best contrast in machine vision matches wavelength to the defect, and film longevity adds a vote for red. Where the duty cycle justifies it, glass-sandwich and wire-grid films survive the heat better; diffuser plates, brackets and cables for the head live in the lighting accessories category.
When the glare is the signal
Polarizers delete exactly what some inspections exist to see. Scratches, dents, pits and tool marks appear as interruptions in the specular reflection; extinguish the reflection and you have erased the defect. For those jobs the low-angle dark-field geometry above is the tool: it keeps the lobe off the lens but lets the defect scatter light into it. The can-end scratch inspection guide and the dent and pit detection guide both build the inspection on the glare lobe rather than against it. Before ordering a polarizer, decide whether the glare hides your defect or carries it.
The decision, and the checks before ordering
| Glare fix | Light lost | Flat specular part | Curved part | Weak spot |
|---|---|---|---|---|
| Offset or rotated ring | none | glare gone | mostly gone | coaxial-only stations |
| Low-angle (0-15°) ring | none | gone, edges lit | mostly gone | measurement and gauging tasks |
| Transverse bar, low angle | none | gone | mostly gone | stations without side space |
| Crossed polarizer pair | 2-2.5 stops | gone | partial | fast lines, tight apertures |
| Single polarizer | ~1.3 stops | no | no | brightness tool only |
The default order is geometry, then light, then polarization. Change the geometry first: offset, low angle, or transverse. If glare survives geometry, budget 2 to 2.5 stops and buy the light back with a high-luminance head or a strobe. Make polarization the first choice only when the measurand is polarization itself, such as stress in clear plastic, or when the station physically cannot move the light.
Before you order:
- Find the lobe by hand-holding the light 10-30° off axis before buying any filter.
- Compute the exposure budget: t = pixel pitch / part speed; multiply by 5.7 for polarizers; compare with the camera's shortest usable exposure.
- Check the aperture floor: your unpolarized f-number minus 2.5 stops must stay above the lens's limit. The same trade for small specular parts is walked through in lighting shiny metal pins without glare.
- Compare set-ups against a fixed target, using the machine vision reference target for comparing illumination, so before/after decisions are objective.
- If you polarize: white or red through the film, film off the concentrator, and a stability check in the schedule.
When geometry wins, pick the head in the light finder, sort out mounting and cabling in the accessories finder, or step through the lighting configurator. More lighting techniques sit on the lighting topic page.
Questions
Do crossed polarizers still work on curved parts?
Partly. A curved surface tilts the polarization plane locally, so no single analyzer angle extinguishes the whole reflection at once: expect residual bright bands near the extremes of the curvature. If those bands miss the region you inspect, crossed polarizers are adequate. If they cross it, move the light off-axis instead — geometry removes the lobe everywhere, not just at one angle.
Can I compensate for polarizer loss by raising camera gain?
Gain multiplies pixel values, not photons. Opening the aperture 2.5 stops collects 5.7× more light and improves the shot-noise-limited signal-to-noise ratio by √5.7 ≈ 2.4; the same gain on an unopened aperture just brightens a noisier image. Add light with the LED, aperture or exposure first, and treat gain as the last resort.
Why does polarizing film burn out in front of LEDs?
Absorption in the film turns into heat, and short wavelengths accelerate the damage: a 470 nm blue photon carries 1240/470 ≈ 2.6 eV against 1240/630 ≈ 2.0 eV for red, about a third more energy per photon. In front of lens-concentrating high-output LEDs, film can burn spots within as little as 100 hours. Use white or red light, stand the film off the concentrator, or specify glass or wire-grid film.
How much depth of field do I lose opening the lens 2.5 stops?
Roughly proportional to the f-number you give up. Going from f/5.6 to f/2.4 keeps about 43% of it: a setup with ±1.2 mm of usable depth at f/5.6 drops to roughly ±0.5 mm, so part-height variation that focused fine before now blurs. Check your own lens and distance with the depth-of-field calculator before committing to the polarizer route.
Is a single polarizer on the light useful without an analyzer on the lens?
As glare removal on metal, no — metal reflections stay largely unpolarized, so extinction needs the pair. On its own the film works as a neutral density filter of about 1.3 stops (40% transmission): a legitimate way to tame an over-bright light without touching the LED drive. A single analyzer on the lens can also trim sheen from glossy plastics, whose reflections are partially polarized, but metal glare needs the pair.