Adaptive signal processing of constant modulus signals

Abstract

The constant modulus algorithm (CMA) compensates for the effect of multipath propagation and interference penetration in constant modulus signals by sensing the modulus variation of the received signal caused by multipath channel and interference. The CMA that uses the steepest descent weight-updating method resembles the well-known least mean square (LMS) algorithm, but differs significantly in that no explicit desired signal is required. But, if a modulus-corrected version of output signal is regarded as the desired signal of the CMA, the CMA may be thought of being operated as in the decision-directed mode of an adaptive equalizer. In this thesis work, the convergence behavior of the CMA and its frequency-domain implementation are studied by considering these facts. First, we investigate the optimum weight vector that minimizes the performance index of the CMA. We also analyze the convergence behavior of the squared output modulus and show that the convergence behavior of the CMA resembles that of the decreasing- LMS algorithm. With these analysis results, several convergence properties of the CMA are clearly explained. Next, a frequency-domain algorithm that reduces the computational complexity and has fast convergence speed is suggested, and its computational complexity is compared with time-domain algorithms. In addition, to reduce the complexity further, the constraints imposed in interference cancellation are realized in the frequency domain. Finally, we show the results of simulations on the effects of various parameters that affect the convergence behaviors of the CMA.