Maximum likelihood symbol synchronization in channels with data dependent noise

Summary form only given. We deal with symbol synchronization in channels with data dependent noise. Examples of such channels arise in optical communications using APD receivers, or direct detection of optically amplified signals where the noise power is higher when a logical one is received. Another situation of data dependent noise arises in the detection of signals in the presence of clutter. In these systems the disturbance exhibits cyclostationary statistics that are ignored by the conventional synchronizers designed for the additive Gaussian noise (AGN) channel, although this cyclostationarity contains timing information that can be explored to improve the tracking performance of the symbol synchronizer. We consider a channel model where the additive disturbance corrupting the received signal consists of the sum of an AWGN process and a cyclostationary component modeled by a Gaussian process with power proportional to the data symbol transmitted. This model represents a good approximation for many direct detection optical systems with APDs or in line optical amplifiers. The maximum-likelihood (ML) data aided (DA) symbol synchronizer for this class of channels is derived and its performance assessed. Comparing the new synchronizer against the well known MLDA synchronizer designed for the AWGN channel, shows that basically the new structure includes in addition to the operations performed by the AWGN-MLDA synchronizer, processing that explores the cyclostationary characteristics of the additive disturbance to enhance the accuracy of the time-delay estimation.