MIMO Signal Processing for Mode-Division Multiplexing: An overview of channel models and signal processing architectures

We present the fundamentals of multiple-input, multiple-output (MIMO) signal processing for mode-division multiplexing (MDM) in multimode fiber (MMF). As an introduction, we review current long-haul optical transmission systems and how continued traffic growth motivates study of new methods to increase transmission capacity per fiber. We describe the key characteristics of MIMO channels in MMF, contrasting these with wireless MIMO channels. We review MMF channel models, the statistics derived from them, and their implications for MDM system performance and complexity. We show that optimizing performance and complexity requires management of channel parameters-particularly group delay (GD) spread and mode-dependent loss and gain-by design of transmission fibers and optical amplifiers, and by control of mode coupling along the link. We describe a family of fibers optimized for low GD spread, which decreases with an increasing number of modes. We compare the performance and complexity of candidate MIMO signal processing architectures in a representative long-haul system design, and show that programmable frequency-domain equalization (FDE) of chromatic dispersion (CD) and adaptive FDE of modal dispersion (MD) is an attractive combination. We review two major algorithms for adaptive FDE of MD-least mean squares (LMS) and recursive least squares (RLS)-and analyze their complexity, throughput efficiency, and convergence time. We demonstrate that, with careful physical link design and judicious choice of signal processing architectures, it is possible to overcome MIMO signal processing challenges in MDM systems.