Evaluation and Symbol Classification of Multi-Valued Signaling Using Two-Dimensional Symbol Mapping with Linear Mixture Model

This study presents an evaluation methodology using a linear mixture model (LMM) with 2-dimensional (2D) symbol mapping for multi-valued data transmission. A 2D map of the transmitted signal can be visualized on a 2D plane. The LMM enables numerical modeling of symbol distribution characteristics. Simulation results showed that the characteristics of the received end-symbol distribution can be quantified by adjusting and determining the LMM parameters from the measured symbols using a genetic algorithm, even in the absence of an eye aperture in the eye diagram evaluation. Additionally, by extracting features of the distribution of multilevel received symbols with the LMM, as in unsupervised learning, unknown received symbols can be clustered with the LMM. Therefore, symbol determination is possible even when the eye is completely closed owing to severe intersymbol interference (ISI). The results of the 4-level pulse amplitude modulation transmission simulation show that it is possible to completely cluster and determine the received symbols, even when the eye is completely closed.