Automated Visual Inspection Algorithm for Solder Joint of Surface Mount Devices Based on Three Dimensional Shape Analysis

An automated visual inspection system has been developed for use with surface mount devices (SMDs) on a printed circuit board (PCB). The system is capable of inspecting minute solder joints of fine pitch components as small as 0.3mm pitch QFP leads. A solder joint is judged using a height image and an intensity image that are detected by a unique confocal height sensor. A solder joint is classified based on the extracted features of solder fillet shape and correlation between a lead and its pad position. Therefore, accurate recognition of these components from the images is extremely important for reliable inspection. However, the shape of a solder fillet varies greatly, depending on factors such as the soldering process, material of the lead and the solder itself. Furthermore, the height sensor sometimes detects erroneous height data, observed as spike noise caused by insufficient reflection light from the inclined specula surface of a solder fillet. The present authors developed a unique noise reduction algorithm that utilizes a model shape of a solder joint as well as a statistical edge detection algorithm based on highest probability to recognize a solder joint without influence of noise. Each component of a solder joint is recognized by a method customized for the type of it, in order to achieve reliable recognition of various shapes of solder joints. Evaluation of the system confirms that it realizes a 100% defect detection rate and a very low false alarm rate (0.16%). The present paper describes the employed inspection algorithm and briefly explains the developed automated inspection system.