Optimization of Pseudo Functional Path Delay Test Through Embedded Memories
MetadataShow full item record
Traditional automatic test pattern generation achieves high coverage of logic faults in integrated circuits. Automatic test of embedded memory arrays uses built-in self-test. Testing the memories and logic separately does not fully test the critical timing paths that go into or out of memories. Prior research has developed algorithms and software to test the longest paths into and out of embedded memories. However, in this prior work, the test generation time increased superlinearly with memory size. This is contrary to the intuition that the time should rise approximately linearly with memory size. This behavior limits the algorithm to circuits with relatively small memories. The focus of this research is to analyze the time complexity of the algorithm and propose changes to reduce the time required to test circuits with large memories. We use our prior work on pseudo functional K longest path per gate test generation, and the benchmark circuits with embedded memories developed in the prior work. Since the cells within a memory array are not scan cells, a value that is captured in a memory cell must be moved to a scan cell using low-speed coda cycles. This approach will also support the test of any non-scan flip-flop or latch, in addition to embedded memory arrays. In addition to testing the critical timing paths, testing through memories eliminates the logic “shadows” around the memory where faults cannot be tested. In this research our complexity analysis has identified the reason for the superlinear increase in test generation time with larger memories and verified this analysis with experimental results. We have also developed and implemented several heuristics to increase performance, with experimental results. This research also identifies the major algorithm changes required to further increase performance.
Boga, Viswanath (2016). Optimization of Pseudo Functional Path Delay Test Through Embedded Memories. Master's thesis, Texas A & M University. Available electronically from