Blind identification of terrestrial microwave radio channels

In microwave radio links applications blind equalization is useful to limit communication outages caused by severe selective fading. Recently, a new approach was proposed that combines blind identification and equalization. The resulting algorithm converges globally, can resolve the nonminimum phase zeros of the channel transfer function, is robust with respect to timing recovery, and uses only second-order statistics. In this paper we assess, by combining analysis and simulation, the applicability of this new method to the equalization of digital radio links affected by selective fading. Our results show that this algorithm outperforms standard blind equalization in terms of convergence speed; however, its properties depend critically on certain assumptions stemming from our a priori knowledge of the channel. >