Machine Learning-Based DFT Recommendation System for ATPG QOR

摘要

A key metric of ATPG Quality of Results (QoR) is the number of test cycles required by a given design for a targeted test coverage. For full-scan compression designs, two key configuration parameters, the number of scan chain in/out ports (SIO) and the maximum scan chain length (MaxLen), usually influence the QoR. While the optimal configuration can be obtained by trial-and-error, this process is iterative, laborious and inefficient. In this work, we have developed a supervised machine learning (ML) recommendation system that predicts the optimal DFT configuration for scan-compression configuration, which produces the minimum number of test cycles at the targeted test coverage for the given design.