Screen projection inspection for transparent part defects

A transparent part can have a clear outline while a shallow dent or flow line remains nearly invisible. Screen projection inspection addresses this problem by recording how the part redistributes directional light onto a separate screen. The camera measures the resulting intensity pattern rather than focusing directly on the material.

This approach suits transparent molded parts, films and glass when ordinary transmitted lighting reveals too little contrast. The practical decision is whether your defect changes ray direction enough to produce a repeatable pattern. You can test that decision by varying screen distance, illumination directionality and camera focus before committing to an optical assembly.

Choose screen projection when the defect redirects light

Start with the feature you need to detect, not the brightness of the light source. An opaque inclusion blocks light. A transparent surface depression may transmit almost as much light as its surroundings while changing the direction of the transmitted rays.

A diffuse backlight supplies rays from many directions. Their overlapping contributions can reduce the visibility of weak refractive features. More directional illumination can preserve the redistribution caused by those features, making a screen pattern easier to observe.

These methods can complement each other. Screen projection does not replace an edge measurement or establish defect depth automatically.

For the broader task of separating lighting problems from processing problems, start with glass defect detection before thresholding. For dimensional work, the comparison of directional and diffuse edge lighting addresses a different decision: stabilizing an edge rather than revealing weak internal patterns.

Understand how deflection becomes an intensity pattern

Consider a narrow bundle of rays crossing the part. A local surface slope or refractive-index variation changes the rays' exit directions. As those rays travel toward a screen, neighboring bundles can converge or spread apart.

Convergence raises the local irradiance; spreading lowers it. The screen converts that irradiance distribution into a visible pattern for the camera.

For a simplified one-dimensional model:

Here, x is the ray position at the part's exit plane, L is the distance from that plane to the screen, and theta is the small exit-angle deflection in radians. The intensity expression assumes uniform incident illumination, negligible absorption and no crossing of neighboring rays.

The distinction between displacement and contrast matters. A uniform angular deflection shifts a beam without necessarily creating a pattern across its interior. Spatial variation in deflection creates the local bright and dark regions.

For an illustrative deflection of 0.5 mrad:

• At 100 mm separation, displacement is 100 × 0.0005 = 0.05 mm.
• At 300 mm separation, displacement is 300 × 0.0005 = 0.15 mm.

Those figures describe ray displacement, not dent depth or guaranteed detection sensitivity. Both surfaces, part thickness and internal refractive structure can contribute to the same image.

Build the optical path around the screen

Arrange the trial as directional source, transparent part, diffusing screen, then camera. A translucent screen lets the camera view the transmitted pattern from the downstream side. Keep the screen approximately normal to the nominal illumination axis initially.

Focus the camera on the screen, not on the part. Place a temporary fine target at the screen plane, focus manually, then remove it. This separates poor screen-image focus from a weak optical signature.

Use a screen with uniform texture and sufficient spatial resolution for the expected pattern. A thick scattering sheet can spread light laterally and erase narrow features. A glossy sheet can produce directional reflections rather than a stable diffuse image.

Mount the part and screen independently so you can change their separation without disturbing camera focus. If you move the screen, refocus or move the camera to maintain its screen-plane relationship.

Start with fixed camera exposure, gain and image processing. Automatic adjustments can hide the contrast changes you are trying to compare. The lens and lighting test plan provides a useful framework for recording each configuration and its failure cases.

Sweep distance and directionality separately

Increasing screen distance can make weak deflections more visible, but it also increases sensitivity to illumination angular spread. Do not change distance and source geometry together during the first experiment.

An illustrative distance sweep is 25, 50, 100 and 200 mm, subject to available space. At each position, record the same acceptable part and the same known defect. Re-establish screen focus and record any change in camera field of view.

Finite illumination directionality introduces a useful estimate:

Here, delta_theta represents the effective angular width of illumination contributing to the pattern. At 200 mm separation, an angular width of 0.5 mrad produces an estimated spread of 0.1 mm. An angular width of 5 mrad produces about 1 mm.

These are geometric estimates, not specifications for any catalog light. Screen scattering, camera focus and optical aberrations can add further blur.

Compare a directional source with a diffuse source while holding the part, screen and camera positions constant. Match the reference screen brightness without clipping highlights. Otherwise, a brightness difference may be mistaken for a directionality improvement.

Also test a small change in part angle. Refraction depends on incidence geometry, so an indication that is strong at one orientation may weaken at another. Avoid choosing an angle that works only for an unusually well-positioned sample.

Work through coverage and camera sampling

Consider a hypothetical transparent coupon measuring 40 × 30 mm. Suppose the screen is 200 mm behind the part, and the expected deflections remain within approximately ±5 mrad.

The maximum lateral displacement estimate is:

Allow another 5 mm on each side for positioning variation and setup margin. The required screen region is then:

A camera field covering 60 × 50 mm would include that region. With an illustrative image size of 1200 × 1000 pixels, sampling is:

A 0.30 mm screen pattern would therefore span six pixels before accounting for blur. That is a sampling calculation, not proof that the pattern has enough contrast for detection. The camera-resolution guide for dimensional inspection explains the broader distinction between pixel coverage and measurement capability.

Use the calculator to check screen-pattern sampling, entering the screen field rather than assuming it equals the part dimensions.

Illumination coverage is a separate check. A rectangular part measuring 40 × 30 mm has a diagonal of sqrt(40² + 30²) = 50 mm. A circular beam must cover that diagonal, plus positioning allowance, over its usable region. Screen size alone does not establish that the part receives uniform illumination.

Select trial hardware without assuming application performance

Browse collimated lights for directional-source candidates, then use the lighting finder to compare the available geometries. Beam coverage, mounting space and documented optical characteristics matter more than the category label alone.

The LS-CLM-0023 is a catalog candidate for a directional-illumination trial. Its listing does not establish screen-projection defect sensitivity or compatibility with a particular screen and camera.

Use a standard source from the backlights category as a baseline rather than assuming directional illumination must be better. Some defects produce adequate absorption or silhouette contrast without a projection screen.

The LS-BKL-1468 is a standard-backlight comparison candidate. Keep its baseline images separate from claims about the directional arrangement.

The supplied catalog information does not establish angular spread, screen compatibility or a complete projection assembly for these candidates. Ask for the relevant optical data and an application review before ordering around a sensitivity target.

Do not assume a telecentric imaging lens can be attached to a light source as a collimator. Such an arrangement requires documented optical and mechanical compatibility, including source placement and aperture matching.

Verify that indications belong to the part

Record a reference image without the part, then insert an acceptable sample. This reveals screen texture, illumination nonuniformity and the normal optical signature of the material.

Next, translate the part while leaving the screen and camera fixed. A material-related indication should respond consistently to the part movement, although projection geometry can change its appearance. A mark that remains fixed on the screen is more likely to belong to the screen or illumination system.

Use multiple acceptable samples, known defects, repeated insertions and the expected positioning range. Include samples whose defects approach the actual acceptance boundary. A dramatic demonstration defect is insufficient evidence for a production threshold.

Flat-field normalization can reduce fixed illumination variation, but it cannot restore clipped highlights or recover detail lost to blur. Register images before subtraction when positional differences would otherwise generate false indications. The guide to reducing contour artifacts in image subtraction explains that failure mode.

Moving parts need another check: exposure blur. If a screen pattern translates at 200 mm/s and the allowable blur is 0.025 mm, the exposure estimate is 0.025 / 200 = 0.000125 s, or 125 µs. Measure projected-pattern motion rather than assuming it always equals conveyor motion. See the exposure-time guide for moving parts for the timing trade-off.

Recognize where the usual answer is wrong

This arrangement is closer to shadowgraph imaging than knife-edge schlieren. A conventional knife-edge schlieren system converts angular deflection into intensity changes by partially blocking light at a suitable focal plane. A projection screen alone does not provide that angular cutoff.

Consequently, weak refractive-index gradients do not always create a useful screen pattern. The spatial variation of the deflection field matters. Strong convergence can also create caustics, where rays overlap and the simple intensity formula no longer applies.

Other limitations include overlapping signatures from front and rear surfaces, curved parts that redirect most light away from the useful screen region, and broad patterns that vary with normal material structure. Stronger contrast can mean greater ambiguity rather than better inspection.

If the real requirement is dimensional gauging, investigate telecentric lenses through the lens finder. Their role is different from focusing a camera on a projection screen. The guide to when telecentric gauging is appropriate explains that choice.

Check the complete trial before ordering

Use this checklist to turn an attractive image into a defensible inspection proposal:

• Define the reject condition. Describe the defect class and acceptance boundary, not simply a visible bright patch.
• Document coverage. Check the illuminated part envelope, screen region and camera field separately.
• Confirm screen focus. Verify it with a temporary target at the actual screen plane.
• Record distance and orientation. Both change the projection signature and must be repeatable.
• Preserve unsaturated images. Keep reference brightness, exposure and processing settings with the dataset.
• Challenge normal variation. Include acceptable material structure, positioning changes and repeated handling.
• Confirm optical compatibility. Obtain missing directionality, mounting and operating information before selecting an assembly.
• Set a verification criterion. Require separation between acceptable and reject populations, not a single successful image.

Start the lighting configurator with your part envelope, candidate screen distances and smallest observed screen pattern. For help reviewing the optical path, send those requirements and representative images through application support.