No abstract

No abstract

Multi-mode fiber (MMF) is the most cost-effective fiber for high-speed LANs. Modal dispersion leads to optical-energy spreading over several symbol periods, drastically limiting distance and data-rate. Compared with copper channels, equalization is challenging because the channel response varies enormously from fiber to fiber and also over time [1]. These aspects, paired with the practical difficulty of implementing TX pulse shaping, increase the equalization burden at the receiver. To date, electronic dispersion compensation (EDC) consisting of an FIR filter cascaded with a nonlinear equalizer, such as DFE, enables 10Gb/s up to 300m according to the 10GBASE-LRM standard. To satisfy the demand for greater network capacity, solutions to reach 25Gb/s on a single fiber, and up to 400Gb/s aggregated throughput with space-division multiplexing on 16 fibers are being investigated [2]. At this data-rate, robust DSP-based EDCs still need high power, indicating an analog approach to signal processing to reduce power. To have market impact and economic feasibility, the interface must be flexible, accommodating a variable data-rates for compatibility with legacy channels and different standards [2]. In addition, achieving high energy efficiency at each standard (i.e., data rate) is fundamental.

Recent work on the application of turbo decoding techniques to partial response channels has focused on additive white Gaussian 'noise channel models. Simulations using these ideal partial response channel models show gains exceeding 5 dB over uncoded systenb at bit error rates of lo''. Since the APP detectors of the turbo decoder assume uncorrelated Gaussian noise, the performance on more realistic channel models, using correlated and mediadependent noise, was unknown. In this work, we replace the white noise partial response channel model with more realistic channel models. First, the effects of colored noise are investigated with an equalized Lorentzian channel model. Then, media noise is added by incorporating the microtrack model into the system. Simulation results of the ,turbo decoding system with the various channel models will be presented. Additionally, since the use of an outer Reed-Solomon code is anticipated in an actual system, the burst statistics at the error floor are investigated.