A comparison of cyclostationary blind equalisation techniques

Interest in blind equalisation has grown in recent years due to the fact that in some communication systems the transmission of a training sequence is not physically feasible. Most blind equalisation schemes to date have sampled the channel output at the symbol rate to produce a stationary output sequence. This paper presents a comparison of the performance of two algorithms for nonminimum phase blind channel equalisation using fractionally-spaced sampling in the receiver. The major problem associated with nonminimum phase channel equalisation is that the measured output statistics must preserve the phase characteristics of the channel. Cyclostationary statistics, unlike the conventional second-order statistics, have been shown to be efficient in this respect. Two families of algorithms are available which exploit the cyclostationary nature of the oversampled received signal in different ways. Both philosophies provide algorithms which are superior to existing symbol-spaced blind equalisation techniques in terms of reliability and speed of convergence. To date no comparison of these two approaches has been carried out and this paper examines these techniques by means of simulation using multipath channels