Fault Localization and Testability Approaches for FPGA Fabric Aware Canonic Signed Digit Recoding Implementations

Abstract

Canonic signed digit (CSD) recoding finds applications in real time VLSI signal processing. In this paper, we have proposed optimized FPGA implementations of CSD recoding techniques starting from a two’s complement input and a redundant signed digit (SD) input. The architectures exploit the fast, hardwired fabric resources of the FPGA logic elements to give rise to a circuit realization optimized for speed and area. The underutilized logic elements configuring the original design are further targeted to append suitable fault localization circuitry without any compromise in speed and area. This makes the designs attractive for implementation in an era where reliability issues of semiconductor chips are on the rise owing to extensive miniaturization of physical device dimensions. Primitive instantiation and constrained placement based design approach allow us to conveniently select the logic area for mapping or to detect and bypass any physical FPGA slice coordinates if deemed faulty.