Protecting FPGA-based automotive systems against soft errors through reduced precision redundancy

Due to their beneficial performance/power/cost ratio, hybrid systems of CPU and FPGA devices are gaining interest from automotive domain. However, FPGA devices suffer from their soft error susceptibility in safety-critical applications. Traditional protection mechanisms like Triple Module Redundancy are well known from space applications, but seem too costly for automotive. In this paper, we introduce first ideas on extending the well-known reduced precision redundancy methods from Shanbhag towards protecting FPGA devices against soft errors by adding CPU-based redundancy. The cost of protection is estimated for a fuel injection control unit with respect to chip area and CPU time overhead, as compared to conventional TMR-based protection.