Reduced-Precision Redundancy on FPGAs

Abstract

Reduced-precision redundancy (RPR) has been shown to be a viable alternative to triple modular redundancy (TMR) for digital circuits. This paper builds on previous research by offering a detailed analysis of the implementation of RPR on FPGAs to improve reliability in soft error environments. Example implementations and fault injection experiments demonstrate the cost and benefits of RPR, showing how RPR can be used to improve the failure rate by up to 200 times over an unmitigated system at costs less than half that of TMR. A novel method is also presented for improving the error-masking ability of RPR by up to 5 times at no additional hardware cost under certain conditions. This research shows RPR to be a very flexible soft error mitigation technique and offers insight into its application on FPGAs.

Approximation-based Fault Tolerance in Image Processing Applications

“…When working at the Register-Transfer Level (RTL), a commonly-used AC technique to trade accuracy and circuit area is the reduction of the data precision of each processed scalar value. This technique has been adopted in [10], [19], [20], to define approximate TMR schemes where the redundant replicas elaborate only on a selected subset of the most significant bits. Therefore, a new voting module is defined to check the numeric distance between the fullyaccurate nominal output value and the two lower-precision redundant ones; the effect is to confine errors in a subset of the least significant bits of the output values.…”

Mitigation of Radiation Effects in SRAM-Based FPGAs for Space Applications

“…The idea of applying RPR to FPGAs in space systems goes back to several works at the U.S. Naval Postgraduate School, Monterey [Snodgrass 2006;Sullivan 2008;Sullivan et al 2009;Gavros et al 2011] and was later followed up by Bratt et al [Pratt et al 2011;Pratt et al 2013]. Instead of using three redundant copies of a module, one module processes data with full precision while the two other modules process the data with reduced precision.…”

“…fixed-point numerical problems [Sullivan 2008]. A decision block determines if a failure has occurred as follows [Pratt et al 2011]:…”

“…The decision further depends on a threshold level T h : The full precision module is assumed to be correct if its output differs less than T h from the reduced precision module output RP 1 Out . For an FIR filter design, Bratt et al showed in [Pratt et al 2011] that the failure rate can be improved by ca. 200 times compared to an unmitigated design at the cost of ca.…”

Reduced Resolution Redundancy: A Novel Approximate Error Mitigation Technique

“…Several Approximate Triple Modular Redundancy (ATMR) techniques have been proposed for general logic circuits [4][5][6][7]. For Digital Signal Processing (DSP) designs, Reduced Precision Redundancy (RPR) [8][9][10][11] is another example of approximate error mitigation methods. RPR uses redundant modules with reduced precision that produce an approximate result.…”