In the actual operation of the Jia Jinyuan SMT production line, AOI inspection must be paired with stable conveyor speed and precise light source angle calibration; otherwise, insufficient solder areas are easily misjudged as qualified due to reflection interference, while tiny protrusions formed by excess solder are often overlooked by planar imaging if the 3D height fusion algorithm is not enabled. We once experienced a sudden 17% drop in bridging recognition rate caused by aging LED backlight modules, which immediately recovered to over 99.2% after replacement, indicating that light source consistency is more critical than simply increasing resolution.

The judgment threshold for component offset cannot mechanically apply the ±25% terminal width specified in the IPC-A-610 standard; it must be dynamically adjusted based on the pad design of the specific component, stencil aperture shape, and actual wetting state after reflow. For example, after a customer adopted a stepped stencil for a 0.4mm pitch QFN, AOI relaxed the X/Y axis offset alarm threshold from 0.12mm to 0.15mm, reducing the false alarm rate by 43% without missing any genuine offset defects. Jia Jinyuan engineers found during on-site debugging that the same AOI device has 21% lower sensitivity in identifying offset of SOT-23 components fixed with red adhesive compared to when no red adhesive is used, because the cured red adhesive alters the reflectance distribution characteristics.

The difficulty in identifying insufficient solder defects lies in distinguishing genuine insufficient solder volume from visual illusions caused by Solder Paste printing collapse. During continuous tracking of 12 batches of the same model motherboard, we found that when SPI inspection shows solder paste volume deviation within ±8% but AOI repeatedly alarms for insufficient solder, it is most likely caused by insufficient wetting due to a peak temperature that is 15℃ lower than required in the reflow profile. In this case, adjusting the reflow zone temperature is more effective than recalibrating the AOI template. Excess solder problems are often accompanied by local blockages caused by incomplete stencil cleaning, appearing as burr-like highlighted areas extending from the pads on AOI images. This feature can be clearly resolved under a 5-megapixel lens, but is easily confused with Flux residue when using a 3-megapixel camera.

Bridging defects are extremely prone to missed judgment in components with pin pitch below 0.5mm. Jia Jinyuan's current AOI system introduces dual-angle side lighting, which increases the shadow contrast formed by solder bridging between adjacent pads by 3.2 times. Combined with an edge tracking algorithm based on grayscale gradients, it raises the BGA bottom bridging recognition rate from 76% to 94%. It is worth noting that when there is a slight oxide film on the PCB surface or OSP thickness fluctuation exceeds 0.3μm, AOI will mislabel normal solder joints as bridging. In such cases, the upstream OSP treatment parameters must be retrieved synchronously for cross-validation.

All AOI programs must be retested with standard defect samples every two weeks, with particular attention to the impact of ambient temperature and humidity changes on the image acquisition module. For example, after workshop humidity exceeds 65%RH in summer, lens fogging causes component offset coordinate drift of up to 0.08mm. In this case, simply cleaning the lens cannot solve the problem fundamentally; the air conditioning dehumidification linkage mechanism must be activated. Jia Jinyuan's internal regulations require that AOI alarm work orders must be accompanied by original image screenshots, corresponding SPI data packets, and the most recent reflow profile file. All three are indispensable; otherwise, the FA team will directly reject the work order without acceptance.