A Fast and Accurate Method for Bond Wires Inductances Extraction Based on Machine Learning Strategy

Abstract

This paper presents a fast and precise method for bond wire inductances extraction based on machine learning strategy. The dataset for training is collected from electromagnetic (EA) simulator and equivalent circuit theory, with the trained model, the self inductances and mutual inductances of bond wires can be accurately extracted only in dependence on their geometric dimensions. Random testing samples are used to verify the proposed method and the results show a tiny discrepancy less than 1% between the real and the predicted values up to 200 GHz. The proposed method will save a lot of time in bond wires inductances analysis and the overall wire bond interconnection design.