Maximum and minimum sensitizable timing analysis using data dependent delays

Abstract

Modern digital designs require high performance and low cost. In this scenario, timing analysis is an essential step for each phase of the integrated circuit design cycle. To minimize the design turn-around time, the ability to correctly predict the timing behavior of the chip is extremely important. This has resulted in a demand for techniques to perform an accurate timing analysis. A number of existing timing analysis approaches are available. Most of these are pessimistic in nature due because of some inherent inaccuracies in the modeling of the timing behavior of logic gates. Although some techniques use accurate gate delay models, they have only been used to calculate the longest sensitizable delay or the shortest topological path delay for the circuit. In this work, a procedure to and the shortest destabilizing delay, as well as the longest sensitizable delay of a static CMOS circuit is developed. This procedure is also able to determine the exact circuit path as well as the input vector transition for which the shortest destabilizing (or longest sensitizable) delay can be achieved. Over a number of examples, on an average, the minimum destabilizing delay results in an improvement of 24% as compared to the minimum static timing analysis approach. The maximum sensitizable timing analysis results in an improvement of 7% over sensitizable timing analysis with pin-to-output delays. Therefore, the results show that the pessismism in timing analysis can be considerably decreased by using data dependent gate delays for maximum as well as minimum sensitizable timing analysis.