Optimization of Cell-Aware ATPG Results by Manipulating Library Cells' Defect Detection Matrices

Abstract

Cell-aware test (CAT) explicitly targets defects inside library cells and therefore significantly reduces the number of test escapes compared to conventional automatic test pattern generation (ATPG) approaches that cover cell-internal defects only serendipitously. CAT consists of two steps, viz. (1) library characterization and (2) cell-aware ATPG. Defect detection matrices (DDMs) are used as the interface between both CAT steps; they record which cell-internal defects are detected by which cell-level test patterns. This paper proposes two algorithms that manipulate DDMs to optimize cell-aware ATPG results with respect to fault coverage, test pattern count, and compute time. Algorithm 1 identifies don't-care bits in cell patterns, such that the ATPG tool can exploit these during cell-to-chip expansion to increase fault coverage and reduce test-pattern count. Algorithm 2 selects, at cell level, a subset of preferential patterns that jointly provides maximal fault coverage at a minimized stimulus care-bit sum. To keep the ATPG compute time under control, we run cell-aware ATPG with the preferential patterns first, and a second ATPG run with the remaining patterns only if necessary. Selecting the preferential patterns maps onto a well-known N Phard problem, for which we derive an innovative heuristic that outperforms solutions in the literature. Experimental results on twelve circuits show average reductions of 43% of non-covered faults and 10% in chip-pattern count.