Introduction
Automated optical inspection (AOI) is the primary quality gate on modern SMT production lines, checking solder joints, component placement, and polarity after reflow. Standard AOI algorithms are trained on ceramic capacitors, plastic-packaged ICs, and standard tin-lead solder surfaces. Copper and brass components, however, present unique inspection challenges because of their high reflectivity, irregular geometry, and different surface coloration. Without proper algorithm tuning and classification rules, AOI systems generate excessive false alarms on copper parts or miss real defects that would be obvious on standard components.
Why Copper and Brass Challenge AOI Systems
Reflectivity is the primary obstacle. Bare copper reflects approximately 60 to 90 percent of visible light depending on surface finish, compared with 10 to 30 percent for matte tin solder and 5 to 15 percent for dark ceramic bodies. This high reflectivity causes two problems: glare that saturates the camera sensor at certain angles, and specular reflections that shift apparent edges depending on the lighting direction. Brass, with its golden-yellow color, adds a hue variation that standard color thresholds may not account for.
Geometry is the second challenge. Copper and brass SMT components often have non-standard shapes—shielding cans with tall walls, co
ector bodies with multiple protruding pins, and bus bar segments with wide flat terminals. These shapes create shadows and occlusions that confuse algorithms trained on rectangular IC packages or cylindrical chip components. The component height variation within a single board can also exceed the AOI focus range if the system is calibrated for a narrow height envelope.
Defect Classification Taxonomy for Copper/Brass Parts
Standard AOI defect categories include missing component, shifted component, reversed polarity, insufficient solder, excessive solder, and solder bridge. For copper and brass components, the classification taxonomy must be expanded to include defect types that are either unique to these materials or require different detection criteria.
Extended Defect Categories for Copper/Brass
| Defect Category | Standard Detection | Copper/Brass Adaptation |
|---|---|---|
| Missing Component | Pad bare detection | Check copper pad shine vs solder pad |
| Shifted Placement | Edge offset measurement | Use shadow-based offset for reflective parts |
| Insufficient Solder | Fillet height/area ratio | Account for copper terminal thickness in ratio |
| Solder Bridge | Co
ected pad regions | Wider search area for tall brass walls |
| Oxidation Stain | Not typically checked | Color deviation from fresh copper hue |
| Warped Shielding Can | Not typically checked | Height profile deviation across can top |
Lighting and Camera Optimization
AOI systems for copper and brass components need lighting configurations that minimize glare while preserving edge contrast. Multi-angle ring lights that illuminate from several directions simultaneously average out specular reflections and reduce the chance of sensor saturation. Diffused dome lights provide uniform illumination from all angles, which is effective for flat copper surfaces but less useful for tall shielding cans that cast shadows.
Coaxial lighting, where the light path shares the camera optical axis, is particularly effective for reflective copper surfaces because it eliminates directional glare while highlighting surface features perpendicular to the camera view. For brass components with varying surface angles, combining coaxial lighting with side lighting at a low angle creates shadow edges that reveal geometry without relying on color or reflectivity alone.
Camera exposure settings must also be adjusted. The auto-exposure algorithms that work for mixed-component boards tend to over-expose copper areas because they average the brightness across all visible surfaces. Setting exposure manually for the copper component zones, or using multi-exposure capture with separate analysis of bright and dark regions, prevents saturation while maintaining visibility of darker solder fillets adjacent to the copper surface.
Algorithm Tuning for Heavy Metal Parts
AOI algorithm tuning for copper and brass components involves three adjustments: component model definition, solder joint model adaptation, and threshold calibration. Component models must include the actual surface reflectivity range so that edge detection algorithms use the correct brightness gradient rather than the default gradient trained on matte components. For copper shielding cans, the model should also include height information to avoid false shifts caused by shadow-based edge detection.
Solder joint models need modification because the fillet on a copper terminal has a different visual profile than the fillet on a tin-lead terminal. The copper terminal’s bright surface creates a stronger contrast boundary with the solder, which standard algorithms may interpret as a defect edge. Training the algorithm on known-good copper joints—creating a golden board reference specific to these parts—teaches the system what an acceptable copper fillet looks like.
Threshold calibration requires ru
ing the AOI on a set of known-defective boards with copper and brass components to measure the false accept and false reject rates. If false rejects exceed 5 percent, the thresholds are too tight. If false accepts exceed 1 percent, the thresholds are too loose. Iterative adjustment with production-representative test boards is the most reliable way to reach an acceptable balance for these non-standard materials.
Inspection Workflow Integration
In a mixed-technology board with both standard and copper/brass components, the AOI inspection workflow must handle both populations without one compromising the other. The most practical approach is to create separate inspection recipes for the standard component zones and the copper component zones, then merge the results into a single defect report. Each recipe uses the lighting and threshold settings optimized for its component type.
For high-mix production lines where different boards have different copper component populations, maintaining a library of tuned recipes by board type ensures consistent inspection quality. Recipe management should be integrated with the board barcode or MES system so that the AOI automatically loads the correct settings when a new board type enters the line. This eliminates the risk of ru
ing a standard recipe on a board with un-tuned copper components.
Conclusion
AOI defect classification for copper and brass SMT components requires deliberate adaptation of lighting, camera settings, and detection algorithms beyond what standard inspection configurations provide. By expanding the defect taxonomy to include oxidation and warpage, optimizing multi-angle lighting to control glare, and tuning algorithms with copper-specific golden board references, manufacturers can achieve the same inspection reliability on heavy metal parts that they expect from standard SMT components. The key is treating copper and brass as a distinct inspection population rather than forcing them into algorithms designed for matte, rectangular parts.
