Low-IR-Drop Test Pattern Regeneration Using A Fast Predictor

Shi-Tang Liu1, Jia-Xian Chen1, Yu-Tsung Wu1, Chao-Ho Hsieh1, Ying-Shiun Li2, Wen-Tze Chuang2, Norman Chang2, Chien-Mo Li1
1National Taiwan University, 2Ansys


Abstract

IR-drop becomes an important issue for testing in advanced technology nodes. In this paper, we propose a low-IR-drop test pattern regeneration to produce IR-drop-safe patterns. To speed up IR-drop analysis, we apply an existing machine learning model to predict IR-drop of test patterns. Because we already know the IR-drop of test patterns, we learn from test patterns to determine low-IR-drop preferred values and extract important bit assignments. By applying our techniques, we regenerate test patterns without predicted IR-drop violations. Experimental results show that our test length overhead is only 2.37% on average, and there is no fault coverage loss. Finally, we perform accurate IR-drop simulation on 10 IR-drop-safe patterns and no IR-drop violations are found.