Partitioning and analysis of static digital CMOS circuits

Abstract

Performance optimization of automatic test pattern generation (ATPG) algorithms has received considerable attention. While the application of high-performance algorithms is often limited to simple gates such as AND's, OR's, and XOR, the cell libraries of silicon vendors usually contain more sophisticated structures. To deal with this problem, we present a library independent algorithm for the partitioning and analysis of static digital CMOS circuits described at the switch level. The algorithm recognizes inverters, NANDs, and NORs. It also checks whether a partition can set its output to a high impedance state, thus being capable of partitioning large bus structures with tristate gates. Our approach supports existing ATPG algorithms at the gate level. Moreover, it allows a mixed-level approach for ATPG using detailed fault models at the switch level for whatever partitions it is necessary. Our implementation processes in the order of 2000 transistors per second. This is for circuits containing combinational and sequential logic on a state-of-the-art workstation. We processed complete chips with up to 72000 transistors, which is clearly adequate for practical purposes.