Abstract:
Dynamic Convergence Analysis of Blind Equalization Algorithms
Equalization techniques counter channel distortion in wireless digital communication systems. In
such techniques blind equalizers are quite important as they do not require channel response and a
training sequence to restart after a communications breakdown; thus reducing overhead and delay of
such systems. The algorithms used for these blind equalizers are based on minimizing a certain cost
function. The cost function incorporates the knowledge of the transmitted signal constellation, using a
stochastic gradient approach.
The convergence properties of some of these algorithms have been studied and documented.
But a lot of work is still to be done. This is due to the nonlinearity of these algorithms which makes it
very difficult to perform the convergence analysis of these algorithms. In this dissertation we consider
the Multi Modulus Algorithm (MMA) and perform its dynamic convergence analysis. We derive
analytical expressions of the mean square error at the output and the expected tap weights.
These analytical results are then compared with the simulated results to verify the validity of
these expressions. This dissertation attempts to give insight as to how parameters like the initial tap
weights and the step size parameters affect the convergence behavior of MMA.
