Demultiplexing by independent component analysis in coherent optical transmission: The polarization channel alignment problem

Digital coherent receivers are essential to the attainment of high spectral efficiency in high capacity fiber-optic transmission. Coherent receivers detect both amplitude and phase of the optical field, which permits the use of advanced modulation formats by the transmitter. A real-time coherent optical receiver using Digital Signal Processing (DSP) enables the transmission system to perform dispersion compensation, carrier phase estimation, polarization de-multiplexing, polarization mode dispersion compensation and data recovery for an advanced modulated format. DSP methods that have been suggested to de-multiplex data from mixed polarizations fall into two classes. First class includes the Least Mean Square algorithm and the decision-directed algorithm require training sequences. The second class that includes the Constant Modulus Algorithms (CMA) and Independent Component Analysis (ICA) is preferable because they work blind and do not need training sequences. The blind source separation methods are preferable since they estimate the source signals directly from the observed signals and, in addition, do not limiting the spectral efficiency. Although, ICA solves the convergence-to-same-source problem of CMA while possessing similar polarization tracking capability, it inherently suffers from a drawback, which is called permutation ambiguity. A novel technique is proposed in this technique that enables ICA to accurately track polarization channel alignment. An improved ICA method is used to identify the input polarizations that carry two different channels but are mixed randomly while the light is propagating in the optical fiber. The addition of some steps to eliminate the permutation ambiguity through a novel technique involving the projection onto the nearest equivalent class technique results in an algorithm that is considerably more robust than the conventional ICA algorithm. The main contribution is tracking channel alignment when the changes in polarization are aggressive due to environmental perturbations (e.g. aerial fiber, multimode systems). Accurate polarization demultiplexing is verified by numerical simulation for the DQPSK modulation format.