Reflected UV surface inspection: test contrast before buying

A scratch that disappears under visible illumination may become easier to detect in reflected ultraviolet light, but a brighter-looking defect is not enough evidence to buy a UV system. You need to identify which wavelengths reach the camera, compare contrast against normal surface variation, and respect the production exposure limit. This guide gives you a controlled test for scratches, texture changes and contamination, followed by the optical compatibility and safety checks needed before selecting parts.

Choose reflected UV only when the reflected signal helps

Reflected UV imaging records ultraviolet radiation returned from the part. Fluorescence imaging records longer-wavelength light emitted after the material absorbs excitation light. These mechanisms need different detection filters and may require different cameras and lenses.

A source can produce reflected UV and visible fluorescence simultaneously. Without suitable detection filtering, a convincing image may come from the wrong mechanism. Follow the fluorescence excitation and filter guide when the useful signal is visible emission rather than reflected UV.

Shorter wavelengths can change scattering and absorption, but they do not automatically reveal every small scratch. Surface material, defect orientation, roughness and collection angle all affect the result. Treat improved UV contrast as a hypothesis to test, not a universal property.

Contain UV before running a test

Plan optical safety before energizing the source. UV can injure eyes and skin, and visual brightness does not indicate exposure risk. Reflections from the test part also belong in the assessment.

Use a contained optical test arrangement with shielding whose attenuation is documented for the actual wavelength range. A cover that looks dark or transparent tells you little about its UV transmission. Prevent direct and reflected radiation from reaching accessible viewing positions.

The risk assessment should define access controls, any required safety interlocks, fault responses and maintenance procedures. Do not defeat protective functions for alignment. Establish an alignment method that keeps the UV source disabled until the protective arrangement is ready.

Have a qualified person verify accessible exposure under normal operation and credible faults against applicable requirements. Personal protective equipment, where required, supplements engineering controls rather than replacing them. Also check whether repeated illumination changes the inspected material; photochemical changes can affect both the part and later measurements.

Build a fair visible-versus-UV comparison

Start with representative acceptable parts and confirmed defects. Include different scratch directions, surface finishes and contamination examples that matter to the process. Record how each defect was independently classified.

Use the lens and lighting selection test plan to organize samples, settings and saved images. Establish two comparisons: a wavelength comparison under similar geometry, followed by a comparison of the best practical setup in each band.

1. Fix the view. Hold the part, camera angle, field of view and inspection region constant. If you change cameras or lenses, match object-space sampling and register the images.
2. Control image processing. Disable automatic exposure, automatic gain, gamma changes and adaptive enhancement. Save linear image data where available, with documented black-level handling.
3. Match geometry first. Compare illumination from similar directions before attributing an improvement to wavelength. Recheck focus after changing bands.
4. Record the exposure requirement. Adjust exposure or illumination within documented operating limits to avoid clipping. Record every change rather than hiding it through display normalization.
5. Then optimize each method. Allow each band its best usable lighting angle, but retain the same acceptance task and production constraints.

For scratches, test illumination from more than one azimuth. A shallow angle may direct the smooth surface reflection away from the camera while scattered light from a scratch enters the lens.

A visible-light dark-field arrangement may outperform a poorly positioned UV source. The LED color contrast guide helps establish a credible visible baseline before adding spectral complexity.

Calculate contrast and separation from variability

Measure image data, not the appearance of an automatically stretched preview. Use a defect region and a nearby background region with comparable underlying surface finish. Subtract a documented dark offset where needed.

For a nonzero background, signed Weber contrast describes whether the defect is brighter or darker. A separate variability metric helps reveal when apparently strong contrast remains unreliable.

Here, S is a practical separation metric, not a universal acceptance standard or a calibrated sensor signal-to-noise specification. Estimate variability from repeated captures and representative part regions. Document whether it includes temporal noise, surface texture, positioning changes or all three.

Consider these hypothetical linear grayscale measurements:

The visible contrast is (110 - 100) / 100 = 0.10. The UV contrast is (120 - 80) / 80 = 0.50.

Visible separation is 10 / sqrt(4² + 4²) = 1.8. UV separation is 40 / sqrt(6² + 4²) = 5.5.

That result supports further UV testing, but does not establish a production detection rate. Repeat it across the sample set, retaining unfavorable results. A low background can inflate contrast, so reject calculations dominated by the dark offset or clipping.

Check sampling and available exposure

Improved spectral contrast cannot recover a defect that the optical system does not resolve adequately. Specify the smallest defect, its critical dimension and the required object-space sampling before choosing a camera.

Suppose the inspection spans 40 mm and the smallest relevant scratch is 0.10 mm wide. As an initial test target, require five pixels across that width. This is a sampling assumption to validate, not a guaranteed detection threshold.

Check optical blur, focus variation and pixel response as well as the pixel count. The UV lens fundamentals guide explains why visible-light focus and transmission cannot simply be assumed at another wavelength.

Motion imposes a separate limit. At an assumed part speed of 200 mm/s and an allowed object-space blur of 0.020 mm, the maximum effective exposure is 0.020 / 200 = 0.00010 s, or 100 microseconds.

A stationary UV image requiring 5 ms does not meet that motion budget. You must demonstrate sufficient signal within the available exposure, change the imaging arrangement, or stop the part. Any pulsed operation must remain within the documented source and controller ratings.

Verify every component in the optical chain

A UV-capable source alone does not create a reflected UV system. Check the source spectrum, delivery optics, part response, detection filter, lens, camera window and detector response as one chain.

Camera and illumination

Use the area scan cameras category and camera finder to shortlist acquisition hardware. A monochrome designation does not establish UV sensitivity. Request spectral response for the assembled camera, including any protective window or internal blocking filter.

For a trial at 365 nm, confirm useful camera response at that wavelength rather than relying on a visible-band sensitivity figure. Check shutter behavior, triggering and whether the required exposure is available in the intended acquisition mode.

The special lights category and lighting finder are starting points for illumination selection. Before choosing a source, obtain its spectral distribution, usable irradiance at the test geometry, uniformity and permitted operating modes. The catalog examples here do not establish a validated UV source-and-camera pairing.

Lens and detection filter

Browse thermal, SWIR and UV lenses through the lens finder, but select by the actual inspection band. A lens intended for another infrared region is not interchangeable with a UV lens.

The LS-SWL-0291 is listed as a 25 mm UV lens with a 240-1000 nm range. Request transmission and imaging performance at the chosen test wavelength before treating it as suitable.

The LS-SWL-0280 is listed as a 1.4x UV lens with a 190-700 nm range. Confirm the intended conjugates, working distance and field coverage for your task.

These are lens candidates, not interchangeable replacements or a validated inspection bundle. A listed wavelength range does not promise flat transmission or equal resolution throughout that range.

For reflected UV, the detection filter must pass the intended UV band while sufficiently rejecting unwanted visible emission and ambient light. Check its transmission and blocking curves, including angle sensitivity. The optical filter selection guide explains why the filter must match the signal you intend to measure.

Challenge the result with production variation

Once one defect looks promising, stop tuning against that single sample. Separate development samples from a holdout set that has not influenced lighting or threshold selection.

Test acceptable surface variation as aggressively as defects. Include position shifts, part tilt, finish differences and the expected range of contamination. Repeat after removing and replacing the part. A useful method must survive normal handling rather than depend on a carefully staged reflection.

A practical screening matrix might include three illumination azimuths, two part tilts and ten representative parts. That gives 3 × 2 × 10 = 60 image conditions per wavelength. This is a starting experiment, not sufficient statistical proof for every production requirement.

Inspect the distribution of decision scores for acceptable and defective regions. Lock the decision rule before evaluating the holdout set. Record false rejects and missed defects separately, with the tested sample counts and defect classes.

The controlled scratch-detection image tests offer a useful pattern for separating real surface evidence from imaging artifacts. Save raw captures alongside the settings so later changes can be compared fairly.

Recognize when another method is better

Choose visible illumination if it delivers adequate separation within the production exposure and variation limits. UV adds compatibility checks and protective measures; it should earn those requirements through measurable inspection benefit.

Choose fluorescence when the material emits a useful signal and the task concerns its presence or coverage. Do not describe that result as reflected UV simply because a UV source initiated it.

Reflected UV may also be the wrong approach when:

• Background texture dominates. The same wavelength that emphasizes scratches may emphasize acceptable roughness.
• Contamination lacks spectral contrast. A film may resemble the substrate in reflection despite being physically present.
• Geometry drives the result. Small tilt changes may overwhelm the difference between good and bad surfaces.
• Exposure is insufficient. Good stationary images may not survive the available motion budget.
• The measurement needs depth. Intensity alone does not establish scratch depth or missing-material volume.

Optical appearance also does not establish cleanliness, sterility or biological identity. Those claims require an appropriate validated measurement method.

Check the evidence before ordering

Turn the successful experiment into a written acceptance specification. Keep application assumptions separate from product specifications so each requirement can be checked against documentation.

• Inspection target: define defect type, minimum size and acceptable surface variation. This prevents a visually impressive but irrelevant demonstration.
• Geometry: record field of view, working distance, camera angle and illumination direction. These determine whether the result can be reproduced.
• Spectral chain: document source output, detection filtering, lens transmission and assembled-camera response. Missing data anywhere can explain a weak signal.
• Timing: record exposure, trigger sequence and motion allowance. Bench performance must fit the production cycle.
• Image evidence: retain raw data, clipping checks, contrast measurements and holdout results. Display screenshots are not sufficient evidence.
• Safety acceptance: document protective measures, access behavior and exposure verification. Complete this before operational use.
• Change control: identify settings and replacement components that require revalidation. A different filter or window can change the recorded signal.

Start the lighting configuration process once these requirements are defined. If spectral compatibility remains uncertain, send the sample description, test wavelength, geometry and exposure budget through application support before selecting a complete system.