Optimal Nonlinear Digital Signal Processing: A Dynamical Systems Approach

Wednesday, August 14, 2019 - 1:00pm to Thursday, August 15, 2019 - 1:55pm

Event Calendar Category

LIDS Thesis Defense

Speaker Name

Omer Tanovic

Affiliation

LIDS

Abstract

THESIS COMMITTEE:
Professor Alexandre Megretski (Thesis Supervisor)
Professor Pablo Parrilo
Professor George Verghese
 
ABSTRACT:

This thesis addresses optimal nonlinear signal processing problems aimed to improve power efficiency of modern wireless transmission systems. 
The first part of this thesis is motivated by peak-to-average power ratio reduction of communication signals. The problem is formulated as minimization of a frequency-weighted convex quadratic cost subject to a time-domain output amplitude constraints. A new method for converting optimality conditions into finite-latency stable systems generating optimal outputs with arbitrary precision is proposed.
 
The second part contains analysis of the nonlinear distortion introduced into the baseband (discrete-time) input-output dynamics of the communication systems by the (continuous-time) power amplifier nonlinearity. It is shown that when the nonlinearity is represented by a Volterra series model the resulting baseband equivalent model is a series interconnection of a discrete-time Volterra series model, of the same degree and equivalent memory depth, and a linear system. The result suggests a new, analytically motivated, structure of digital pre-distortion (DPD) of power amplifier nonlinearities.
 
The third part of the thesis focuses on analysis and design of digitally implemented pulse-width modulators (DPWM) used as quantizers for power amplifiers in switched-mode operation. A time-domain input-output model of DPWM which offers new insight into nonlinear behavior of this system is developed. A modified Lloyd-Max quantization based algorithm for linearization of the baseband of DPWM is proposed.