Adaptive equalizer MSE performance in the presence of multipath fading, interference and noise

Analyzes the MSE performance of three fractionally spaced adaptive equalization schemes, viz., (1) an MMSE DFE computed with estimated channel parameters, and directly adapted; (2) RLS and (3) LMS DFEs, which all require training at the beginning of a TDMA burst to minimize intersymbol and interuser interference and noise. The authors derive for the computed MMSE DFE an upper bound expression for its excess MSE which can indicate that this scheme performs worse than the directly adapted RLS DFE. They also employ an MSE convergence analysis to show that the directly adapted LMS DFE exhibits prohibitively slow MSE convergence in the presence of interference due to its sensitivity to the eigenvalue spread of the equalizer input signal autocorrelation matrix and the existence of a significantly lower MMSE in interference than in an equivalent amount of stationary noise. Thus, of the three equalization approaches considered, the directly adapted RLS DFE yields superior performance in interference.