Robust Fixture Design of CCGA Based on Clustering Algorithm with Outliers

Ceramic Column Grid Array (CCGA) Packages are commonly used for integrated circuits with high-density, which require high-performance and high-reliability. The alignment and placement of welding posts is a key step in the CCGA packaging process, which is mainly realized by matching pads using high-precision flux. Due to the variation in the pads plating process, the planting flux designed by the standard drawings cannot completely and accurately plant all the ceramic columns on the pads. In this paper, we apply two clustering algorithms, k-means and Self-Organizing Map (SOM), in planting fluxes design in order to cover as many integrated circuits as possible. The results show that SOM outperforms K-means in both coverage ratio and stability. The optimal results are reached by eliminating the outliers and using SOM with sigma=1.5, learning_rate=0.5. Considering the cost, this paper provides high-precision design of plating fluxes which efficiently improve CCGA processing.