Artificial neural network application in vertical interconnection modeling

This paper proposes a neural network-based method for modeling vertical interconnect balls. The pi equivalent circuit with lumped element was used to characterize the electrical performance of balls. The values of lumped elements were extracted from the full-wave simulate result, and the extracted data was used to train back-propagation (BP) neural networks to obtain the relationship between the lumped element values and the physical size of the balls. Then this model was used to estimate the electrical character of ball with other layout parameters.