Key attributes of an SRAM testing strategy required for effective process monitoring

Yield learning-a key process in assuring manufacturing efficiency-must be based on effective defect diagnostic procedures. One such procedure, using measurements of SRAMs and a computer-generated defect-fault dictionary, is presented in this paper. The discussion is focused on the 'resolution of diagnosis', which is the ability to resolve as large a variety of defects as possible. To assess the resolution limits, an experiment involving dictionary generation and SRAM testing was conducted. The test results were compared against simulated bitmaps, and the differences were analyzed using failure analysis. The results obtained confirmed that defect simulation-based diagnosis can be very effective. It can also be enhanced if appropriate SRAM designs, testing strategies and defect models are chosen.<>