ABSTRACT
Digital Signal Processing (DSP) is based on the representation of signals by sequences of numbers and the processing of these sequences. The basic purpose of DSP is to estimate characteristics of a signal or to transform a signal into a more desirable form. As recent developments have illustrated, more signal processing implementations are using DSP techniques than ever before.
A fault-tolerant system is one that is capable of correct operation despite the presence of certain hardware failures and software errors. By applying fault-tolerant techniques in the design of a new computer system, the system can be made significantly more reliable.
A number of improvements are needed in the field of DSP systems, including: increased reliability, performance, and growth potential; and improved simulation methods and techniques for test and evaluation. However, in order to begin to achieve these goals, a number of critical developments have been lacking until recently. These include high-performance IEEE-754 floating-point DSP microprocessors, multiple-device simulation software callable from a high-level language, and simulation software capable of single-cycle granularity and complete device pinout support.
Based on these developments, the potential now exists to design a new fault-tolerant DSP computing system for a broad class of DSP applications. This design emphasizes fault tolerance and parallel processing and is prototyped completely in software. This in turn leads to system test and evaluation for a fraction of the traditional cost, and makes more advanced studies possible such as fault injection and analysis.
