A saturated patch of specular glare hides exactly the feature you are inspecting — the print on shrink film, the laser mark on bare metal, the scratch on a glossy panel. The standard fix is a pair of polarizers: a polarizing film laid flat over the light and a polarizing filter on the lens, set 90° apart. The pair absorbs the mirror-like reflection and passes the diffuse signal underneath. This page gives you the two formulas behind that, the light-loss arithmetic that decides your exposure and aperture, a bench alignment procedure, and the straight list of surfaces where the crossed pair beats any diffuse light you can buy.
Two filters, one on the light and one on the lens
The setup has exactly two parts. A linear polarizing film sits flat in front of the illumination window — over the diffuser of a bar light, or as the last layer of a backlight. A linear polarizing filter screws into the lens front thread, specified like any vision filter: the LS-LXR-0225 band-pass filter, for example, prints its M30.5 x 0.5 thread right in the description, and you order a polarizing filter to match your own lens thread the same way. Screw-in filters live under lens accessories, with thread and size filters in the accessories finder; the light-side hardware the film mounts onto, such as bar light diffuser plates, lives under lighting accessories.
Unpolarized LED light oscillates in every plane. The film passes one plane and absorbs the rest. The specular bounce off a shiny surface keeps that plane; the diffuse return — light scattered inside the material or off micro-texture — comes back randomized, roughly half of it aligned with the lens filter's axis. Cross the two axes and the specular component hits an opaque filter while the diffuse component still passes. Print and texture survive; the white hotspot dies.
Set the two filters parallel instead and the pair works as a variable neutral-density filter: transmission follows the cosine-squared law, so you can dim an over-bright light continuously without touching the controller. The same knob, turned to 90°, is the glare switch.
Why the reflection dies: two formulas
The first is Malus's law: at 90° the pair passes nothing, at 0° it passes everything the first filter let through. The second locates the only angle where a dielectric surface polarizes a reflection completely. At Brewster incidence the reflected ray is entirely s-polarized — oscillating parallel to the surface — while the p-component refracts into the material. Glass with n ≈ 1.5 gives θB = arctan(1.5) ≈ 56°; water with n ≈ 1.33 gives 53°; acrylic and polycarbonate sit at 56-58°.
| Surface | Polarized on reflection? | What you fit |
|---|---|---|
| Glass, water, glossy plastic viewed near 53-58° | Yes, fully (s-plane) | Lens filter alone can catch it |
| Film, varnish, plastic under diffuse light | Partly | Film on light + crossed lens filter |
| Bare metal | Poorly — no Brewster zero | Film on light + crossed lens filter, or dome geometry |
| Stressed transparent plastic in transmitted light | Birefringent instead | Film on backlight + crossed lens filter |
Malus's law also sets the alignment tolerance. At 80° between the axes, cos²(80°) ≈ 0.03, so three percent of the bright specular leaks past the crossed filter; at 70°, cos²(70°) ≈ 0.12. Hold the pair within about ten degrees of a true cross and the leak stays invisible.
The light budget
The crossed pair taxes photons, and the arithmetic is short. An ideal polarizer passes half of unpolarized light; practical films and filters pass around 40%, the rest absorbed. The diffuse path through the crossed setup then carries 0.40 (film) × 0.5 (depolarized return, half aligned to the lens axis) × 0.40 (lens filter) = 0.08 — eight percent of the original light, which is log2(1/0.08) ≈ 3.6 stops.
The optimistic floor, with ideal films and a strongly depolarizing surface, is 0.5 × 0.5 × 0.5 = 0.125: exactly three stops, a factor of 8. Semi-gloss parts that depolarize only partly return even less. Plan exposure, aperture and drive current for a factor of 8 to 13, then confirm on the bench.
For comparison, a single filter on the lens costs about 1.3 stops (0.40 through) and buys almost nothing under your own lighting; the stop-for-stop accounting is in the polarizer light-loss article. When the budget cannot be met at your line speed, buy the missing stops as photons rather than as gain:
- Size
- 200-400 mm
- Type
- Standard
- Voltage
- 24 V
At 312 x 30 mm, this high luminance bar light is the bright engine behind a crossed pair. Pulsing harder than continuous duty allows is the other lever; the strobe duty-cycle article shows how much overdrive a duty cycle supports.
Worked example: exposure after the filters go on
Scenario: a printed carton face, static under a diffused white bar, monochrome 8-bit camera. Baseline without polarizers: 1.5 ms exposure at f/4, ROI mean grey 170, glossy highlights saturating at 255.
Fit the crossed pair. The specular peak falls by the extinction ratio — better than 100:1 for a decent film pair — so 255 × 0.01 ≈ 3 grey, effectively gone, while the diffuse mean drops to 170 × 0.08 = 13.6, call it 14 grey.
Static part: spend the whole factor on exposure. 1.5 ms × 10.3 ≈ 15.5 ms. Exposure alone caps the frame rate near 65 fps, which a static station never approaches.
Belt at 100 mm/s: 15.5 ms smears the print by 100 mm/s × 0.0155 s = 1.55 mm, unusable. Open from f/4 to f/2: (4/2)² = 4, two stops. The remaining factor is 10.3 / 4 ≈ 2.6, so exposure becomes 1.5 ms × 2.6 ≈ 3.9 ms and blur falls to 100 mm/s × 0.0039 s ≈ 0.39 mm. If your pixel pitch is finer than 0.39 mm the smear still spans several pixels — recover the rest with strobe drive instead of opening further.
Fast line: hold exposure near 2 ms and buy the factor of 8 to 10 as strobe pulse energy or a brighter head. Raise gain last — it lifts noise with the signal.
Aligning the pair on the bench
- Fix the film flat over the light window, clamped or taped at the edges. A wrinkled film scatters and depolarizes its own light.
- Screw in the lens filter and set both axes parallel: brightness peaks near the parallel-pair level, 0.40 × 0.40 = 16% of bare light.
- Aim the light and set a working exposure that puts the diffuse mean near 170 grey.
- Rotate the lens filter slowly towards 90° while watching the histogram; the specular spike collapses long before the diffuse base does.
- Stop where the specular peak sits at single-digit grey and the diffuse mean still reaches 60 or more at an exposure your cycle time allows.
- Check extinction on a flat shiny reference: the mean should fall by a factor of 50 or more between parallel and crossed.
- Lock the filter ring and mark its position; vibration and lens changes rotate it quietly.
Judge the result by measurement, not by eye — the ROI and reference method puts two lighting states on one comparable scale.
Polarized backlight: stress, inclusions and hidden parts
Cross the pair in transmitted light and the empty field goes black; anything that rotates polarization lights itself up. Mechanical stress makes transparent plastic birefringent: injection-moulded parts, bottle walls and clear housings rotate the plane locally, and stressed zones appear as bright fringes — photoelasticity, usable as a pass or fail check on residual stress and as a comparison between moulding parameters. Inclusions and gels in transparent film show the same way, and an inserted display carries its own polarizing layers, which turns a presence check behind an opaque housing into a simple brightness decision.
Use a broad, even backlight under the film, or the dark field itself becomes the defect. For glare work only the relative 90° matters; for stress fringes the absolute orientation of the pair changes the pattern, so note it whenever you compare images across shifts or cells. The full film-inspection recipe is in the crossed-polarized backlight article.
Where the crossed pair wins and where it does not
It wins on flat, glossy, awkward-to-light surfaces:
- Printed and laminated packaging — shrink film, blister foil, carton varnish — where gloss buries the print. The material-by-material recipes are collected in the cross polarization overview, the companion piece to this one.
- Bare metal that must present a dark, even face to code reading or object search; the shiny metal pins article contrasts the crossed pair with dome and diffuse options.
- Screens and glossy panels in cosmetic inspection; the screen inspection setup and the glossy panel scratch guide cover both angles.
- Transmitted stress and presence checks, as in the section above.
Under a film you want a flat, even base to polarize, so diffused bars are the natural carriers. A diffused white bar suits small fields:
- Size
- 200-400 mm
- Type
- Diffused
- Voltage
- 24 V
Web and long-part cells want the same idea stretched out:
- Size
- 400 mm +
- Type
- Diffused
- Voltage
- 24 V
On both, the film mounts flat against the diffuser face, cut slightly oversize and clamped at the ends.
It loses on three fronts. Curved shiny parts throw specular in every direction and no flat film follows the curvature — a dome light wins by geometry instead, and the dome-light guide sizes it. Surfaces where the sheen is the signal, such as brushed grain or textured rubber, lose the very feature you inspect when the pair is crossed. And line speeds too fast for three or four extra stops, even with strobe, push you to geometry — dark field — or to wavelength as the contrast tool. Polarizing film also ages under heat and UV: mount it on the coolest outer face of the light and inspect it at maintenance intervals.
Checks before you order
- Filter thread: match the polarizing filter to each lens front thread, diameter and pitch, the same way any filter is specified.
- Film size and mounting: sheet larger than the light window, held flat, on the coolest face of the light.
- Exposure headroom: baseline exposure × 8-13 must stay under your blur or frame-rate ceiling; if it cannot, add strobe drive or a brighter head.
- A flat, shiny reference part for the extinction check at every changeover.
- Colour work needs no correction — linear polarizers are neutral in the visible band — so monochrome and colour cameras both qualify.
Order the light first and the filters second: the head has to survive the stop tax before anything else matters. The bar lights page and the light finder narrow the choice by length, colour and diffuser type, and the configurator walks camera, lens and light as one chain:
The configurator matches a camera, lens, light, light controller and cables to each other, step by step, and lists only parts that fit what you already picked.
If the geometry fights back — curved, dark or fast-moving parts — contact the applications team with a photo of the part and the line speed, and we will say whether the crossed pair or a dome belongs on it.
Questions
Can one polarizing filter on the lens remove glare by itself?
Only when the glare is already polarized at the surface. Dielectric reflections near the Brewster angle — glass, water and glossy plastic at roughly 53-58° incidence — arrive s-polarized, so a single lens filter can block them. Under your own diffuse light the reflections come back at every angle, so one filter costs roughly half the light and removes almost none of the glare. The crossed pair is the default in a controlled lighting cell.
Do polarizing filters shift the colours in the image?
No, not meaningfully. A linear polarizer absorbs by oscillation plane, not by wavelength, so the balance between red, green and blue survives. Colour cameras keep their white balance and monochrome cameras lose nothing. Practical films show a transmission loss and, in low-grade sheet stock, a faint neutral tint that a one-time white balance removes. Treat polarizers as intensity parts, not colour parts, when you plan exposure.
How much light does a crossed-polarizer setup cost?
Plan on roughly one eighth of the original light. A practical film or filter passes about 40% per pass, and the diffuse return from the part is only half aligned to the lens axis, so 0.4 × 0.5 × 0.4 ≈ 8%, or 3.6 stops. The best case with ideal films is 0.5 × 0.5 × 0.5 = 12.5% — exactly three stops — and partly polarizing semi-gloss surfaces return even less than 8%. Size the lighting for a factor of 8 to 13.
How accurately must the two polarizers be aligned?
The transmission axes should sit within about ten degrees of 90° apart. Transmission between the filters follows the cosine squared of the angle between them: at 80° you leak about 3% of the specular, at 70° about 12%, which is already visible next to a dim diffuse signal. Rotate the lens filter while watching the histogram rather than judging by eye, then lock the ring and mark its position.
Can crossed polarizers really show stress in transparent plastic?
Yes, and it needs no extra hardware. Put a polarizing film on a diffuse backlight and cross the lens filter: the empty field goes dark. Stress makes the material birefringent, it rotates the polarization plane, and stressed zones light up as bright fringes. Injection-moulded parts, PET bottle walls and transparent housings show their residual stress this way, and an inserted display with its own polarizer layers also stands out for presence checks.
Why do shiny metals stay slightly bright even with the pair crossed?
Metal never reaches a Brewster zero, so its reflection is only partially polarized and partly depolarized by texture and curvature; some of it always survives the lens filter. Improve extinction with a quality film pair and exact 90° alignment, then take the rest out with geometry: drop the light angle into a dark-field arrangement, or move curved metal under a dome light, where uniform diffuse illumination replaces the mirror-like return.