A Low-Overhead Multiple-SEU Mitigation Approach for SRAM-based FPGAs with Increased Reliability

摘要

The mitigation of single-event upsets (SEUs) through modular or functional redundancy is a traditional approach for designing fault-tolerant systems; however, even in multiple redundant systems, SEUs can lead to a system failure if they occur simultaneously. Previous fault-tolerant approaches have proposed run-time reconfiguration to regain the lost functionality. We worked with a similar strategy to overcome failures caused by unidirectional SEUs occurring simultaneously in both frontline and redundant modules, but the approach we propose in this paper not only improves reliability but also requires low-overhead as compared to previous methodologies. The proposed architecture is an array of computation tiles containing computation cells and corresponding hot-spares. Each computation tile has a separate region for spare cells. The simultaneous faults are handled by an on-chip fault-tolerant core and external host software that partially reconfigure the spare-cells region of a computation tile. The proposed architecture is implemented on a Xilinx Virtex-5 FPGA device and verified with the aid of simple digital application. Compared to previous schemes, our approach requires up to 9.6x less area overhead while providing 57.6% more reliability to mask multiple unidirectional SEUs.