On blind fractionally-spaced equalization using analytical constant modulus

In this contribution we discuss the applicability of the so-called analytical constant modulus approach to the case of blind fractionally-spaced equalization of communication channels. We demonstrate that such an approach cannot be pursued when the equalizer is straightforwardly parameterized by the samples of its FIR impulse response. Then, we obtain a new parameterization of the equalizer, exploiting the modulation redundancy associated to the excess bandwidth usually available in fractionally sampled PAM and QAM signals. It is pointed out that the FIR equalizer belongs to properly defined linear subspace, related to the geometrical structure of a suitable matrix build up using time-varying correlations of the received signal. Exploiting this new parameterization, we finally show that the analytical constant modulus approach is successfully applicable in blind fractionally-spaced equalization.