A reflective mounting pad can resemble a screw head, while an installed screw can disappear into glare. For automated PCB assembly checks, the useful image is therefore not necessarily the most evenly illuminated image. It is the image that keeps installed and missing screws distinguishable across normal production variation.
Start by comparing dome and coaxial illumination without changing wavelength. Then test wavelength separately, using the same inspection regions and acceptance criteria. This approach helps you select a camera, lens, light, and decision rule without mistaking a colour effect for a geometry improvement.
Define exactly what the station must distinguish
Write the inspection requirement around the mounting location, not around the appearance of the whole board. The primary classes are an installed screw and an exposed mounting pad. Include realistic confusing conditions, such as a washer without a screw, a loose screw beside the hole, and reflections from nearby components.
Presence is not the same as correct assembly. A visible screw head does not establish tightening torque, seating force, thread engagement, or electrical continuity. If those conditions matter, use additional process measurements or a separately validated inspection.
Build the initial sample set around variation that can change reflections:
- Board finish, solder-mask colour, and coating variation.
- Screw-head finish, drive orientation, and surface marks.
- Allowed board translation, rotation, height, and tilt.
- Nearby components that cast shadows or reflect light into the inspection region.
- Empty pads with normal handling marks or process residue.
Use location features outside the mounting pad to register the board before measuring presence. Otherwise, a small placement change can move a bright pad edge into the measurement region and imitate a missing screw.
For the broader selection process, the lighting selection guide by inspection task explains why each feature needs its own contrast objective.
Compare dome and coaxial geometry at one wavelength
A dome presents a broad range of illumination directions to the part. This can reduce isolated reflections from a curved screw head, but it can also make the screw and surrounding pad similarly bright.
Coaxial illumination directs light toward the board along the viewing axis through a beam splitter. A flat surface facing the camera may return strongly, while a curved or tilted surface sends some light elsewhere. This can create useful screw-to-pad separation, but sensitivity to tilt and component shadowing must be tested.
| Decision factor | Dome illumination | Coaxial illumination |
|---|---|---|
| Curved screw-head reflections | Broad angular coverage can reduce isolated highlights | Head shape and drive recess can create distinct intensity patterns |
| Flat exposed mounting pad | May look similar to other illuminated metal | May return strongly when aligned with the viewing axis |
| Board tilt variation | Broad illumination angles may help, but require testing | Reflection strength can change substantially with tilt |
| Nearby tall components | May obstruct portions of the illumination | May obstruct illumination or the camera view |
| Main selection question | Does diffusion preserve enough screw-to-pad difference? | Does the contrast survive normal alignment variation? |
Neither geometry is automatically correct. Compare the worst acceptable installed screw against the most confusing empty pad, rather than choosing from the cleanest demonstration image.
The dome lighting guide covers dome coverage and viewing-hole limitations. The discussion of diffusion and polarization on reflective metal is useful when screw-head highlights remain unstable.
Run a controlled geometry experiment
Use a monochrome camera for the initial intensity comparison unless colour measurement is itself required. Lock exposure, gain, gamma, and other image-processing settings. Disable automatic adjustments that could change the score between boards.
- Choose a common wavelength band. Compare dome and coaxial candidates with similar nominal wavelength and bandwidth. Comparing a red coaxial image against a white dome image changes two variables at once.
- Fix the camera and lens. Keep the viewing direction, aperture, focus, and inspection regions unchanged. Move only the light hardware needed for each geometry.
- Match a useful reference level. Adjust illumination output so a stable reference region has approximately the same nonsaturated brightness. Record the adjustment rather than assuming equal electrical settings mean equal illumination.
- Capture matched samples. Image the same installed and missing conditions under both geometries. Include the limits of normal board position and tilt.
- Compare decision margins. Evaluate class separation, clipped pixels, and sensitivity to position. Whole-board visual uniformity is secondary.
Place the reference region on material that stays visible in both assembly states. Do not normalize against the screw head, because the reference would disappear when the screw is missing.
If boards are manually presented rather than held in a repeatable fixture, resolve positioning first. The handheld PCB camera setup guide addresses that different acquisition problem.
Test wavelength as a separate variable
After establishing a geometry baseline, compare available illumination colours while keeping the viewing geometry fixed. Start with representative installed and missing samples, then repeat promising combinations across the wider sample set.
Do not select wavelength solely from the visible colour of the board or screw. Surface coatings, metal finishes, camera sensitivity, and the light's spectral output all affect recorded intensity. A colour that darkens the solder mask may still leave the screw and exposed pad indistinguishable.
Record a small experiment matrix rather than relying on memory:
| Test variable | Keep fixed | Measure |
|---|---|---|
| Geometry | Wavelength band, camera, lens, sample set | Worst-case class separation and tilt sensitivity |
| Wavelength | Geometry, camera processing, sample set | Screw-to-pad separation and reference stability |
| Illumination output | Geometry and wavelength | Saturation, noise, and exposure feasibility |
The guide to choosing LED colour for machine-vision contrast explains the reflectance relationships behind these tests.
If another inspection genuinely requires a different colour, treat it as a separate acquisition requirement. The multi-wavelength sequencing guide covers that architecture; it is not a reason to add colour sequencing to a presence check that works with one exposure.
Size the image around the smallest useful feature
Determine which feature actually distinguishes the classes. It may be the screw-head boundary, the drive recess, or the exposed annulus around an empty hole. Sampling the entire screw head generously is not enough if the decisive detail occupies only a few pixels.
For an illustrative station, suppose the smallest useful detail is 0.6 mm wide. A starting target of 6 pixels across that detail gives an object-space sampling requirement of 0.1 mm per pixel. This is an engineering starting point, not a universal detection guarantee.
For a horizontal field of view of 120 mm:
This arithmetic establishes a sampling floor, not the final camera specification. Allow for board-position margins and confirm that the lens resolves the detail across the full inspection field. Check the vertical coverage separately.
Focus at the screw-head plane and inspect the pad plane as well. If both states must contribute sharp edges to the decision, both must remain sufficiently focused throughout the allowed board-height range.
Choose exposure from the available cycle time and motion conditions. If the board moves during exposure, assess blur using object speed multiplied by exposure time, then compare that travel with the object-space pixel size.
Measure separation with a worked decision example
A normalized intensity score can reduce sensitivity to common brightness changes when the reference region behaves consistently. It cannot correct changing reflections, saturation, or a reference region whose material varies between boards.
Let P be the median intensity in a registered candidate region and B the median intensity in a nearby reference region. One possible score is:
Reject or flag measurements when the denominator is too small for a stable calculation. Also check clipping separately; normalization cannot recover information from saturated pixels.
Consider illustrative 8-bit measurements, where intensity values range from 0 to 255. An empty pad has P = 180 and B = 60. An installed screw has P = 90 and B = 60.
Now suppose the full sample set produces installed-screw scores from 0.14 to 0.28 and empty-pad scores from 0.44 to 0.56. The observed gap is:
A provisional rule could classify scores below 0.32 as installed and above 0.40 as missing. Scores between those boundaries would require review or a controlled reacquisition. These boundaries illustrate a decision method; production limits must come from your own samples and acceptance requirements.
If intensity ranges overlap, try registered shape information or a different lighting arrangement. Avoid hiding poor optical separation behind increasingly complicated thresholds.
Choose candidate lights and complete the station
Browse dome lights when broad angular illumination appears promising. The LS-DOM-0085 and LS-DOM-0122 are catalog candidates for a mechanical review. Their inclusion here does not establish suitability for a particular PCB or wavelength configuration.
- Size
- 100-200 mm
- Voltage
- 24 V
- Size
- 100-200 mm
- Voltage
- 24 V
Check the illumination opening, camera viewing clearance, mounting arrangement, and space around the tallest component. Housing dimensions alone do not establish useful illuminated coverage.
Browse coaxial lights when a flat pad's reflection may distinguish it from the screw head. The LS-CXL-0092 and LS-CXL-0067 provide alternative catalog candidates for that review.
- Size
- 50-100 mm
- Voltage
- 24 V
- Size
- 100-200 mm
- Voltage
- 24 V
Confirm the usable optical aperture and light-to-part spacing for the proposed camera field. For either geometry, confirm the exact colour configuration before comparing candidates. Category-level colour options do not establish the configuration of an individual part.
The remaining station elements should support the experiment:
- Camera: fixed acquisition settings and adequate sampling of the decisive feature.
- Lens: required coverage, resolved detail, and focus across the relevant height range.
- Light controller: compatible output, operating mode, and timing for the selected light.
- Fixture: repeatable board support without obstructing the mounting locations.
- Trigger: image capture after the board reaches the intended inspection position.
Use the lighting finder to narrow candidates, then carry the measured requirements into the light configuration workflow.
Validate the difficult cases and recognize the limits
Reserve some boards and screws for validation rather than using every sample to set thresholds. Capture repeated placements so the validation includes real registration and reflection changes, not merely repeated processing of the same image.
For example, an initial engineering challenge could use 10 board samples, 2 assembly states, and 5 independent placements per state. That produces 100 images for each lighting condition. This is a manageable comparison exercise, not statistical proof of a production escape rate.
Track false acceptance of missing screws separately from false rejection of installed screws. Also track indeterminate decisions. Combining these outcomes into one accuracy percentage can conceal the failure that matters most.
The usual dome-versus-coaxial comparison is insufficient when:
- A component physically hides the mounting location from the camera.
- Screw and pad appearances remain indistinguishable under the tested arrangements.
- The requirement is to detect a raised or incorrectly seated screw rather than simple presence.
- A washer or foreign object consistently imitates the accepted screw pattern.
Those cases may require another viewing direction, an additional measurement, or a revised assembly process. More illumination cannot reveal a feature hidden from the camera.
Check these items before ordering
- Decision definition: distinguish screw presence from seating, torque, and unrelated assembly checks.
- Sample coverage: include normal finishes, empty pads, screw orientations, and misleading objects.
- Controlled comparison: change geometry first and wavelength second, with camera processing locked.
- Image quality: verify sampling, focus, clipping, and registration at every mounting location.
- Mechanical fit: confirm optical clearance and component access using the proposed station layout.
- Electrical compatibility: confirm the exact light configuration and controller requirements.
- Acceptance evidence: document class ranges, decision boundaries, and treatment of uncertain images.
Choose the configuration with the strongest worst-case separation that also fits the station. Before committing to hardware, send representative installed and missing images, board dimensions, placement limits, and cycle-time requirements through applications support. That information is more useful than a request for a light based only on PCB colour.
Questions
Is dome or coaxial lighting better for PCB screw presence checks?
Neither geometry is universally better. A dome can stabilize reflections from curved heads, while coaxial illumination can emphasize differences between flat pads and screw surfaces. Compare them at a common wavelength, then choose using the smallest observed separation across acceptable parts and normal placement variation.
Should red illumination be the first choice for a green PCB?
Red is a useful test candidate, not a guaranteed answer. The recorded difference depends on the screw finish, exposed pad, coating, light spectrum, and camera response. Measure installed and missing conditions under several available colours while holding geometry and camera processing constant.
Can a screw presence image also verify correct tightening?
Not reliably from a simple overhead presence image. A head may be visible even when the screw is loose, cross-threaded, or insufficiently engaged. Define those defects separately and use an appropriate additional inspection or process measurement when they form part of the acceptance requirement.
What should happen when installed and missing screw scores overlap?
Do not force every image into an acceptance decision when the optical evidence is weak. Check saturation, positioning, reference stability, and sample variation first. Then consider different geometry, wavelength, or shape features. An indeterminate category can prevent uncertain measurements from being silently accepted.