Performance evaluation of three adaptive equalization algorithms

The three simplified versions of the stochastic gradient algorithm are analyzed. Assuming gaussian probability distributions for the input signal and the output error signal, their stability is studied, their output mean square error performance is evaluated and their optimal step-size parameters are given. For digitally implemented equalizers we also indicate the minimum number of bits required to store the equalizer coefficients. Computer simulation results are reported which confirm the validity of the theoretical results.