Spectral Efficiency Optimization in Flexi-Grid Long-Haul Optical Systems

Abstract

Abstract-Flexible grid optical networks allow a better exploitation of fiber capacity, by enabling a denser frequency allocation. A tighter channel spacing, however, requires narrower filters, which increase linear intersymbol interference (ISI), and may dramatically reduce system reach. Commercial coherent receivers are based on symbol by symbol detectors, which are quite sensitive to ISI. In this context, Nyquist spacing is considered as the ultimate limit to wavelength-division multiplexing (WDM) packing.In this paper, we show that by introducing a limited-complexity trellis processing at the receiver, either the reach of Nyquist WDM flexi-grid networks can be significantly extended, or a denser-than-Nyquist channel packing (i.e., a higher spectral efficiency (SE)) is possible at equal reach. By adopting wellknown information-theoretic techniques, we design a limitedcomplexity trellis processing and quantify its SE gain in flexigrid architectures where wavelength selective switches over a frequency grid of 12.5GHz are employed.

Signal Overlap for Elastic Optical Networks

“…In particular, I A =87.5 Gb/s, I B =95 Gb/s (I A +I B =182.5 Gb/s) in the maximized scheme, and I A =100 Gb/s, I B =76 Gb/s (I A +I B =176 Gb/s) otherwise. In this figure are also reported simulations with LDPC codes with different code rates (the same applied in [13]), which best approach the achievable bound, showing that although some penalty must be taken into account, a good matching is evident. Each dot on the graph indicates the code rate of S A and S B respectively, in addition to the corresponding C/I.…”

“…3, already introduced in sect. IIA, shows the main simulation results of the presented work, namely the net bit rate achievable bounds of signal S A and S B as a function of S A OSNR and C/I, which entails the S B OSNR; these bounds were theoretically computed by exploiting the principles in [23] (see also [13], [20] for details) to determine the average mutual information per symbol, successively multiplied by the nominal bit rate R.…”

“…The computed bound can be efficiently exploited to guide the estimation of parameters like required optical signal to noise ratio (OSNR) and power ratio between the overlapped channels. Then, biterror rate (BER) simulations of practical low-density paritycheck (LDPC) codes, as in [13], are used to confirm the behavior of the theoretical predictions. Thus, the overlap of two 112 Gb/s PM-QPSK is assessed in terms of OSNR performance, showing the achievable working conditions and the LDPC codes to be adopted.…”

“…On the contrary, we introduce a complete digital processing with the aim of allowing an effective detection of frequency overlapped signals. More specifically, the proposed technique is practically elaborated upon the work in [12], [13], and based on advanced coding technology, successive interference cancellation (SIC) strategies and proper configuration of optical transmission parameters.…”

Time–Frequency Packing Applied to Cost-Effective IM/DD Transmission Based on Directly Modulated VCSEL

“…Results are hence presented both as bit error rate (BER) computations by employing practical LDPC codes and as achievable bit rate lower bounds (ABR), that represents a useful method to evaluate the theoretical system performance without resorting to practical codes. 1 For insights on the usefulness of such bounds with respect to BER measurements, please refer to [27]. The computation of these bounds exploits indeed the principles of auxiliary channel and mismatched detection, and, in brief, it takes advantage of samples r k to yield a bound I LB ≤ I, which, despite being a lower bound, is achievable for the considered receiver, in the sense that it is theoretically possible to design a code that reaches the predicted performance.…”