The application of irregular data formats for improved performance in 16-bit digital signal processing systems

This thesis asserts that 16-bit programmable Digital Signal Processors (DSPs) suffer from inadequate dynamic range and noise performance due to the use of standard data formats, and that this problem is compounded when they are programmed using traditional higher level languages. A solution to this problem is developed that involves the use of highly irregular data formats.; First, the data formatting problem for programmable 16-bit devices is defined. Background information concerning the problem and its context is then provided. Existing approaches to address the problem are identified and analyzed. A set of format criteria important for 16-bit DSP applications is delineated. A new class of irregular floating point formats is developed. Specific formats are derived from this class that provide greater dynamic range and improved noise performance for 16-bit DSP applications. A broad simulation of the new formats is conducted for both strong and weak signal cases. The simulation results are analyzed, and the analysis indicates that the 16-bit DSP data formatting problem can be improved through the use of the new irregular data formats. Ramifications of implementing these formats in both hardware and software are then discussed. Potential applications and areas of future research are identified. Finally, the results of this study are summarized.