Date of Award

2011

Degree Type

Dissertation

Degree Name

Ph.D.

Department

Electrical and Computer Engineering

First Advisor

Ahmadi, Majid (Electrical and Computer Engineering)

Keywords

Electrical engineering.

Rights

CC BY-NC-ND 4.0

Abstract

The ever increasing demand for low power DSP applications has directed researchers to contemplate a variety of potential approaches in different contexts. Since DSP algorithms heavily rely on multiplication, there are growing demands for more efficient multiplication structures. In this regard, using some alternative number systems, which inherently are capable of reducing the hardware complexity, have been studied. The Multi-Dimensional Logarithmic Number System (MDLNS), a multi-digit and multi-base extension to the Logarithmic Number System (LNS), is considered as an alternative to the traditional binary representation for selected applications. The MDLNS provides a reduction in the size of the number representation with a non-linear mapping and promises a lower cost realization of arithmetic operations with a reduced hardware complexity. In addition, using the recursive multiplication technique, which refers to the published multiplication algorithm that uses smaller multipliers to implement a larger operation, reduces the size of operands and corresponding partial additions. As part of this research, 2DLNS-based multiplication architectures with two different levels of recursion are presented. These architectures combine some of the exibility of software with the high performance of hardware by implementing the recursive multiplication schemes on a 2DLNS processing structure. These implementations demonstrate the efciency of 2DLNS in DSP applications and show outvistanding results in terms of operation delay and dynamic power consumption. We also demonstrate the application of recursive 2DLNS multipliers to reconfigurable multiplication architectures. These architectures are able to perform single and double precision multiplication, as well as fault tolerant and dual throughput single precision operations. Modern DSP processors, such as those used in hand-held devices, may find considerable benefit from these high-performance, low-power, and high-speed recongurable architectures. In the final part of this research work, recursive 2DLNS multiplication architectures have been applied to a FIR lter structure. These implementations show considerable improvement to their binary counterparts in terms of VLSI area and power consumption.

Share

COinS