Performance analysis of a FPGA based novel binary and DBNS multiplier

Abstract

Designing high performance Software Defined Radio (SDR) with low power and flexibility is a major challenge. While the high performance DSP processors are unable to meet the speed requirements of these SDRs, System on chips (SOCs) are also not suitable because of their limited flexibility. Recently dynamically reconfigurable FPGAs have emerged as high performance programmable hardware to execute highly parallel, computationally intensive signal processing functions efficiently. Since basic intention of an SDR is to implement different modulation / demodulation schemes and basic building blocks for such schemes are signal processing functions, FPGAs have become an important component for implementing these. However, the effectiveness of such an approach with respect to cost, performance and flexibility need to be examined. Double Base Number Systems (DBNS) have been gaining attention for compute intensive applications in signal processing because of their higher performance in arithmetic operations in general and particularly multiplication. Keeping these issues in view, this paper aims to present a new Software defined Radio. To Enhance the performance of the proposed architecture , analysis have been done employing both single index and multiple indices DBNS multipliers. Experiments and analysis on performance have also been done with its binary counterpart. Both DBNS and binary based architecture were implemented on Xilinx virtex iv FPGA using xilinx ISE 9.1 i.