In the surface mount technology (SMT) manufacturing process, automated optical inspection (AOI) is a crucial step in ensuring soldering quality and assembly consistency. To fully leverage the effectiveness of AOI in actual production, besides relying on equipment hardware performance, mastering a series of practical techniques is essential to improve inspection accuracy, reduce false alarm rates, and accelerate anomaly handling.
Firstly, appropriately selecting and combining light source modes is a fundamental technique for improving imaging quality. Different defects exhibit significantly different characteristics under different lighting conditions. For example, for solder joints with poor soldering or insufficient wetting, low-angle ring light can be used to enhance contour contrast; for the spherical characteristics and shadow interference of BGA solder balls, coaxial light or diffused light should be combined to reduce reflection; when inspecting characters and polarity markings, vertically incident light can be used to obtain clear boundaries. Skillful switching and combination of light sources can effectively highlight defect characteristics and avoid missed detections and false judgments.
Secondly, stencil creation and baseline calibration need to be precise down to the PCB version and panel differences. Experience shows that directly using a generic stencil can trigger false defect alarms due to differences in pad size, spacing, or surrounding silkscreen printing. Specific testing procedures should be established for different product models, and multi-point calibration should be performed using standard samples before implementation to ensure accurate matching of the coordinate system and magnification, thereby ensuring the comparability and repeatability of measurement data.
Thirdly, threshold settings should strike a balance between sensitivity and specificity. Blindly pursuing a high detection rate will lead to a large number of normal solder joints being mislabeled, increasing the burden of re-inspection. The key is to first collect a certain number of positive and negative sample images, analyze the differences in grayscale, shape, and texture between defects and good products, and then fine-tune the threshold parameters step by step, verifying the effect through small-batch trial runs, gradually approaching the optimal detection window.
Fourthly, making good use of multi-view and local magnification functions can improve the reliability of detection in complex areas. For difficult areas such as connector pins, fine-pitch QFPs, or densely packed RC arrays, a separate local detection area and higher resolution scan can be set to avoid missing detailed defects due to resolution limitations in global scanning.
Fifthly, establish habits for the classification, statistics, and trend analysis of defect data. By categorizing defects according to type, location, and time of occurrence, weak points in the process can be quickly identified. For example, increased bridging during a certain period may indicate abnormal squeegee pressure, and frequent misalignment in a specific area may be related to wear on the pick-and-place machine's nozzles. Linking data with sources such as SPI and the pick-and-place machine to form a closed-loop feedback loop significantly improves the targeted nature of process improvements.
Finally, strengthening operator training on the identification of typical defect images and establishing concise re-inspection and handling procedures can prevent production delays due to misread alarms. Combining these techniques, SMT automated optical inspection not only accurately intercepts defects but also transforms them into an effective source of information for process optimization, providing a solid guarantee for high-quality manufacturing.