LAUSR

Abstract Subregion constellation processing as a demodulation method improves noise tolerance and system error performance when the optical channel presents non-Gaussian distortions, common in optical-wireless communications systems. The proposed method consists of: first, grouping in clusters data symbols closely located -in the sense of Euclidian distances- on the 2D constellation diagram, for then tracing boundaries between clusters following the rules of Voronoi diagrams. The Voronoi contours define subregions on the constellation diagram, and each one of these subregions contains a cluster of data symbols. The proposed method allows both the definition of demapping rules based on identified Voronoi contours in the process of demodulation, as well as the characterization of transmission impairments on each cluster of data symbols by analyzing the 2D mean value of the cluster and its geometrical shaping. Subregion constellation processing is demonstrated and experimentally validated in a single-channel Radio-over-Fiber (RoF) system with a direct-detection (DD) receiver, with information propagated over an optical fiber link with length 78.8 km at a bit rate of 1 Gb/s. The fact of applying Voronoi contours and subregion constellation processing as a rule for demapping improves the noise tolerance in 3.1 dB for 16-QAM, and 1.4 dB for 4 + 12 PSK when compared to the standard grid for the studied modulation formats, for a bit error rate (BER) threshold of 1 × 10−2. A non-Gaussian behavior was observed on the recovered constellation diagram, creating asymmetric data-symbols distribution (irregular ellipsoidal shape on external symbols) on the IQ plane.