Variable Precision Filters

ABSTRACT: In this Phase I SBIR project, the Athena Group, Inc., will develop a mathematical framework and expert system design tool for robust and highly efficient filter design. The developed tool will be capable of designing filters that have challenging requirements, such as sharp transitions, narrow subbands, and/or complicated passband structures. The developed expert system will be able to accept both mathematical constraints and physical constraints that relate to the filter"s resource utilization, such as data precision and power, in order to produce a near-optimal result. The developed expert system is intended to be operated without the guidance of an expert human designer, yet produce results comparable to those produced by the expert designer. By eliminating the need for human intervention, the developed expert system may be implemented in embedded applications that require real-time filter design in order to adapt to a variety of operating conditions. BENEFIT: The proposed technology will provide filter designers the co-design tool needed to optimally realize complex digital filters while concurrently meeting stringent design constraints. Delivered as an integrated software system that can target ASICs, FPGAs, or DSP microprocessors, the technology can significantly impact the entire DSP community.