The problem of analytical evaluation of the maximum rate at which information can be reliably transmitted on a nonlinear wavelength division multiplexing fiber-optic channel with a given modulation format and detection strategy is addressed. An approximate solution of the nonlinear Schrödinger equation is adopted to obtain an accurate analytical discrete-time channel model, valid for arbitrary link configurations and modulation formats. By exploiting the concept of mismatched decoding, considering a sub-optimum detection strategy that accounts for intra-channel nonlinearities, the proposed model is employed to derive closed-form expressions of the achievable information rate with various modulation formats. All the analytical results are verified through comparison with numerical simulations in different scenarios.

In flexi-grid optical networks, fragmentation of spectrum resources may significantly affect the overall network efficiency. Effective techniques for defragmentation (i.e., re-optimization) are then required to limit the wasting of spectrum resources. However, current defragmentation techniques can only be implemented thanks to the presence of additional resources, such as spare expensive transponders. In this study, we propose, discuss and evaluate a novel defragmentation technique called push-pull. The technique is based on dynamic lightpath frequency retuning upon proper reconfiguration of allocated spectrum resources. It does not require additional transponders and does not determine traffic disruption. All the relevant technological limitations that may affect the push-pull applicability are discussed in the context of both optically-amplified direct and coherent detection systems. The technique is then successfully demonstrated in two different flexi-grid network testbeds, reproducing the two aforementioned scenarios. In particular, the reoptimization of a 10 Gb/s OOK lightpath is safely completed in few seconds (mainly due just to node configuration latencies) without experiencing any traffic disruption. Similarly, the push-pull is successfully performed on a 100 Gb/s PM-QPSK lightpath, providing no traffic disruption.

A simple but accurate analytical model for the fiber-optic channel affected by cross-phase modulation (XPM) is proposed and validated through numerical simulations. The model describes XPM as a time-varying and frequency-selective complex fading, whose impact and statistical properties can be analytically investigated. As an example, the proposed model is applied to derive an analytic lower bound to the channel capacity and to investigate the coherence properties of the channel.

The concept and significance of the so called nonlinear Shannon limit are reviewed and their relation to the channel capacity is analyzed from an information theory point of view. It is shown that this is a limit (if at all) holding only for conventional detection strategies. Indeed, it should only be considered as a limit to the information rate that can be achieved with a given modulation/detection scheme. By virtue of some simple examples and theoretical results, it is also shown that, using the same approximated models commonly adopted for deriving the nonlinear Shannon limit, the information rate can be arbitrarily increased by increasing the input power. To this aim, the validity of some popular approximations to the output distribution is also examined to show that their application outside the scope for which they were devised can lead to pitfalls. To the best of our belief, the existence of a true nonlinear Shannon limit has still not been demonstrated, and the problem of determining the channel capacity of a fiber-optic system in the presence of Kerr nonlinearities is still an open issue.

One of the most severe impairments that affect coherent optical systems employing high-order modulation formats is phase noise due to transmit and receive lasers. This is especially detrimental in uncompensated links, where an ideal compensator for channel distortions and laser phase noise should first eliminate receive phase noise, then equalize channel distortions, and only later compensate for transmit phase noise. Unfortunately, the simultaneous presence of transmit and receive phase noise makes very difficult to discriminate between them, even in the presence of a pilot tone. Moreover, the picture is different for optical systems using single-carrier or orthogonal frequency division multiplexing, where transmit and receive phase noise components may have a different impact. All these aspects are analyzed and discussed in this paper. A novel digital coherence enhancement (DCE) technique, able to significantly reduce the phase noise of transmit or receive lasers by using an interferometric device plus a very simple electronic processing, is also described. The performance of this technique and the statistical properties of the residual phase noise are analytically derived and verified by simulations, showing a high increase of the maximum bit-rate-distance product. The practical implementation of DCE is finally discussed and some alternative implementation schemes are presented.

In wavelength-switched optical networks (WSONs), quality of transmission (QoT) has to be guaranteed during lightpath provisioning. In multibit-rate WSONs, this task is complicated by the coexistence of optical connections operating at different bit-rates and modulation formats. The major issue consists in accounting for the severe impairments due to cross-phase modulation (XPM) induced by 10 Gb/s lightpaths on neighbor 40 or 100 Gb/s lightpaths. In this paper, QoT modeling is first reviewed for 10, 40, and 100 Gb/s transmission according to the adopted modulation format and detection type. In addition, a Gaussian approximation to compute the bit error rate of differential quadrature phase-shift keying (DQPSK) and QPSK signals is proposed, as well as closed formulas to compute the nonlinear phase noise variance due to XPM. Also, discussions about the XPM cumulation over spans in a WSON and how XPM can be considered in a dynamic network are provided. Then, four lightpath provisioning schemes are proposed to effectively account for QoT and, in particular, for XPM. The schemes differently exploit: 1) augmented spectral separation among lightpaths at different bit rates; 2) XPM worst-case scenario; and 3) current and novel generalized multiprotocol label switching extensions. The performance of the proposed schemes is evaluated through simulations in several multibit-rate scenarios. Results show that the proposed schemes provide effective network resource utilization while guaranteeing the adequate QoT to lightpaths at any bit rate.