Identification of Soft-Error at Gate Level

Due to shrinking feature size and significant reduction in noise margins, as we are moving into very deep sub-micron technology, circuits have become more susceptible to manufacturing defects, noise-related transient faults and interference from radiation. Traditionally, soft errors have been a much greater concern in memories than in logic circuits. However, due to technology scaling, logic circuits have become equally susceptible to soft errors. Moreover, enhanced usage of commercial off the shelf (COTS) electronic components for avionics has also increased the importance of analyzing soft errors in hardware circuits. Conventionally, understanding soft error glitches requires circuit level modeling, which requires information available only at late stages in the design flow. Instead of this approach some researchers have produced modeling techniques using Reduced Order Binary Decision Diagrams (ROBDD) and Algebraic Decision Diagrams (ADD), which does allow analyzing soft error at an earlier stage in design flow. In this thesis, a new methodology for modeling soft errors glitch propagation path using Multiway Decision Graphs is introduced. This modeling technique is applicable on both combinational and asynchronous circuits. The proposed glitch propagation path modeling technique jointly takes care of logical and electrical masking. Our methodology involves new ways of injecting glitches including glitch injection in feedback paths of asynchronous circuits. This work presents a complete framework to exhaustively provide all the possible sequences of signals that lead to the possibility of glitch propagation to the primary output in combinational and asynchronous circuits. In addition, a new tool is developed based on the proposed methodology called Soft Error Glitch-Propagating Path Finder (SEGP-Finder) to automate the identification of these sequences of signals. This work helps designers identify the vulnerable circuit paths at the logic abstraction level. Also, this methodology allows designers to apply radiation tolerance techniques on reduced sets of possibilities. By applying our methodology on different combinational and asynchronous circuits an improvement in terms of possible-fault injection vectors is observed. As an example, approximately 8% of all the possible input vectors and sequences is required for obtaining exhaustive glitch propagation path identification in a representative implementation of a bundled data asynchronous circuit. To the best of our knowledge, this is the first time MDG based decision diagram based soft error identification approach is proposed for combinational and asynchronous circuits