Efficient EMI Analysis and Self-Shielding Design Method for Pin Map of System-in-Package

As the radiation leakage from solder balls of system-in-package becomes increasingly significant, this article presented a novel analysis and design method for pin map distribution scenarios by combining electromagnetic theory and machine learning technology. First, an empirical virtual boundary generalized T-matrix (VBGT) method is proposed in this article to analyze the pin map structure with finite dimensions and imperfect boundaries, enabling rapid computation of edge-field and far-field radiation. Subsequently, an enhanced deep Q-network that couples the VBGT method and deep reinforcement learning is developed to optimize the pin allocation. Finally, the results from various real-world pin map cases demonstrate the efficiency and reliability of the proposed VBGT-DRL algorithm, which holds promising application potential in the automatic pin assignment design of system-in-package for electromagnetic interference self-shielding.