Algorithm for Fast Statistical Timing Analysis

Abstract

Problems of parameter variations are a main topic in current research and will gain importance in future technology generations due to the continuing scaling. Therefore, it requires appropriate timing analysis which is traditionally done with corner-case simulations. These are quite conservative and pessimistic approaches. In contrast, new statistical static timing analysis (SSTA) algorithms offer a more accurate prediction of the timing behavior of circuit designs. Further, correlations between various parameters and devices can be observed. Unfortunately, the SSTA algorithms mostly require high computational effort and accurate library characterization. This paper proposes an approach for a fast statistical static timing analysis (F-SSTA) with moderate requirements on computation time and library characterization. The approach considers the analysis of gates with multiple inputs. The simulation results show an average error of 5% compared to Monte-Carlo simulations but a significant speed improvement of around 20 times compared to a highly accurate SSTA algorithm.