Framework for Selective Flip-Flop Replacement for Soft Error Mitigation

Abstract

With increasing adoption of newer technologies and architectures targeted for automotive and aviation electronics with an objective to improve performance and/or reduce power/area, soft-error robustness is becoming an important issue to ensure reliable operation for an extended lifetime over a wide range of operating conditions. In this paper, we propose a modeling and optimization framework to systematically improve the FIT (failure-in-time) rate of a design with minimal impact on power, performance and area. We first propose a framework to model and evaluate the relative vulnerability to soft errors of the standard master-slave flip-flops and Dual Interlocked Storage Cells (DICE) in the cell library. Later, we formulate a linear optimization problem using this information to selectively replace the flip-flops so as to improve the FIT rate of the design with minimal impact on area and power. Employing the proposed technique on a popular industrial IP core shows a 32% relative improvement in the design robustness with just 2% increase in design area.