Design and analysis of short word length DSP systems for mobile communication

Memon, T 2012, Design and analysis of short word length DSP systems for mobile communication, Doctor of Philosophy (PhD), Electrical and Computer Engineering, RMIT University.

Document type: Thesis
Collection: Theses

Attached Files
Name Description MIMEType Size
Memon.pdf Thesis Click to show the corresponding preview/stream application/pdf;... 1.46MB
Title Design and analysis of short word length DSP systems for mobile communication
Author(s) Memon, T
Year 2012
Abstract Recently, many general purpose DSP applications such as Least Mean Squares-Like single-bit adaptive filter algorithms have been developed using the Short Word Length (SWL) technique and have been shown to achieve similar performance as multi-bit systems. A key function in SWL systems is sigma delta modulation (ΣΔM) that operates at an over sampling ratio (OSR), in contrast to the Nyquist rate sampling typically used in conventional multi-bit systems.

To date, the analysis of SWL (or single-bit) DSP systems has tended to be performed using high-level tools such as MATLAB, with little work reported relating to their hardware implementation, particularly in Field Programmable Gate Arrays (FPGAs).

This thesis explores the hardware implementation of single-bit systems in FPGA using the design and implementation in VHDL of a single-bit ternary FIR-like filter as an illustrative example. The impact of varying OSR and bit-width of the SWL filter has been determined, and a comparison undertaken between the area-performance-power characteristics of the SWL FIR filter compared to its equivalent multi-bit filter. In these experiments, it was found that single-bit FIR-like filter consistently outperforms the multi-bit technique in terms of its area, performance and power except at the highest filter orders analysed in this work. At higher orders, the ΣΔ approach retains its power and performance advantages but exhibits slightly higher chip area.

In the second stage of thesis, three encoding techniques called canonical signed digit (CSD), 2’s complement, and Redundant Binary Signed Digit (RBSD) were designed and investigated on the basis of area-performance in FPGA at varying OSR. Simulation results show that CSD encoding technique does not offer any significant improvement as compared to 2’s complement as in multi-bit domain. Whereas, RBSD occupies double the chip area than other two techniques and has poor performance.

The stability of the single-bit FIR-like filter mainly depends upon IIR remodulator due to its recursive nature. Thus, we have investigated the stability IIR remodulator and propose a new model using linear analysis and root locus approach that takes into account the widely accepted second order sigma-delta modulator state variable upper bounds. Using proposed model we have found new feedback parameters limits that is a key parameter in single-bit IIR remodulator stability analysis.

Further, an analysis of single-bit adaptive channel equalization in MATLAB has been performed, which is intended to support the design and development of efficient algorithm for single-bit channel equalization. A new mathematical model has been derived with all inputs, coefficients and outputs in single-bit domain. The model was simulated using narrowband signals in MATLAB and investigated on the basis of symbol error rate (SER), signal-to-noise ratio (SNR) and minimum mean squared error (MMSE). The results indicate that single-bit adaptive channel equalization is achievable with narrowband signals but that the harsh quantization noise has great impact in the convergence.
Degree Doctor of Philosophy (PhD)
Institution RMIT University
School, Department or Centre Electrical and Computer Engineering
Keyword(s) Short Word Length (SWL) DSP Systems
Sigma Delta Modulation
Ternary FIR-like filters
Version Filter Type
Access Statistics: 471 Abstract Views, 1937 File Downloads  -  Detailed Statistics
Created: Fri, 21 Sep 2012, 16:40:13 EST by Kelly Duong
© 2014 RMIT Research Repository • Powered by Fez SoftwareContact us