Sensitivity-Based Stochastic Analysis Method for Power Variations

Abstract:
With the increasing relevance of random process variations on device properties one faces an increasing necessity to model the resulting variations of circuit properties by appropriate statistical simulations to be able to optimize robustness and yield already in an early design stage. In principle, this kind of modeling may be reached by extensive Monte Carlo simulations which, however, may become extremely time consuming for larger circuits. The present paper discusses a new approach to model characteristic statistical properties of the power consumption of circuits due to variations in leakage current and switching energy. Where applicable, the underlying “sensitivity-based stochastic analysis” (SBSA) approach provides a means to derive the required statistical results within an accuracy comparable to extensive Monte-Carlo simulations, but with run-times that are orders of magnitudes smaller than that of the latter. The paper sketches briefly the underlying basic approach, discusses examples where this method has been applied for a statistical analysis of switching energy and leakage currents of standard cells.