Automatic DRAM cell location in the SEM

Abstract

The highly repetitive nature of a DRAM cell array requires the failure analyst to perform the tedious task of manually counting and tracking cells in order to locate on the die the position of a failed cell that has been identified through electrical testing. This paper describes a vision-based system that automates the task of cell location without the need for a high-accuracy specimen stage. The algorithm makes use of cross-correlation to track a moving die, and mimics the action that would otherwise be carried out by the SEM operator in locating a cell position.