Development of an Automated Solder Inspection System With Neural Network Using Oblique Computed Tomography

The high density LSI packages such as BGA is being utilized in the car electronics and communications infrastructure products. These products require a high-speed and reliable inspection technique for their solder joints. In this paper, an automated X-ray inspection system for BGA mounted substrate based on oblique computed tomography are proposed. Automated inspection consisted of OCT capturing, position adjustment, bump extraction, character extraction and judgment. Five characteristic features related to the bump shape are introduced. And by combining the characteristic features using artificial neural network, the condition of solder bump was judged. In the experiments, these techniques were evaluated using actual BGA mounted substrate. As a result, the correct rate of judgment reached 99.7%, which shows the clear evidence that proposed techniques may be useful in the practice.