Fibre and components induced limitations in high capacity optical networks

The design of future all-optical networks relies on the knowledge of the physical layer transport properties. In this thesis, we focus on two types of system impairments: those induced by the non-ideal transfer functions of optical filters to be found in network elements such as optical add-drop multiplexers (OADM) and optical cross-connects (OXC), as well as those due to the interaction of group-velocity dispersion, optical fibre non-linearities and accumulation of amplifier noise in the transmission path.

The dispersion of fibre optics components is shown to limit their cascadability. Dispersion measurement techniques are first reviewed, and the limitations of the commonly used phase-shift technique is discussed. Additionally, an alternative method which enables the direct determination of small dispersion values in the pass-band of optical filters is proposed.

Available optical filter technologies are compared with respect to their dispersive properties. The cascadability of fibre gratings is investigated numerically and experimentally. The conventional Gaussian apodisation profile is shown to result in unwanted dispersion in the pass-band, which will limit its cascadability to less than five devices when a channel spacing of 50 GHz is used at 10 Gbit/s.

The use of narrow bandwidth modulation formats such as optical duobinary is suggested in order to improve the detuning tolerance of Gaussian apodised gratings. Alternatively, novel asymmetric apodisation profiles with multiple phase-shifts can be designed to provide reduced dispersion in the pass-band.

Large detuning tolerances are demonstrated experimentally for a variety of modulation formats. A numerical optimisation of pass-band flattened phased array (PHASAR) multiplexers is performed for use in high spectral efficiency metropolitan area networks at 40 Gbit/s. Even if conventional PHASARs are theoretically dispersion-less devices, the pass-band flattening process is shown to induce unwanted dispersion, which will ultimately limit the device cascadability.

A PHASAR based on a parabolic horn input coupler is found to be the most promising design in order to maximise the spectral efficiency in a four add-drop node ring network. The concept of "normalised transmission sections" is introduced in order to ease the dimensioning of transparent domains in future all-optical networks.

Normalised sections based on standard single mode fibre (SMF) and dispersion compensating fibre (DCF) are optimised numerically with respect to the positioning of the DCF, the degree of compensation and the input powers to the two fibre types. Experimental validations are performed for 10 Gbit/s non return-to-zero (NRZ) and chirped return to-zero (CRZ) modulation over 80 km pre-compensated spans.

Passive pre distortion at the transmitter is shown to significantly improve the reach of the systems. Based on the experimental results, transparent domains with a diameter of the order of 1000 km can be realised, thus demonstrating the applicability of the optimisation method to the design of large area networks.

Wavelength division multiplexing (WDM) systems not only require compensation of the dispersion of the transmission fibre, but also of its dispersion slope. The effectiveness of early slope compensating DCFs for broadband compensation of SMF is demonstrated experimentally for 10 Gbit/s NRZ modulation.

In particular, transmission in the L-band is achieved over more than 1000 km using a dispersion map optimised for the C-band, removing the need for separate band compensation. Novel DCFs enabling for the cabled compensation of the dispersion and dispersion slope of SMF (the so-called inverse dispersion fibres, IDF£n, where n is the SMF to DCF length ratio), are compared numerically.

For NRZ modulation at 10 Gbit/s, IDF£1 is found to maximise the transmission distance over 50 km spans for single channel, while being prone to cross-phase modulation in WDM systems where IDF£2 or 3 should be preferred. The benefit of using short return-to-zero (RZ) pulses over conventional NRZ modulation in a SMF+IDF£1 link is highlighted.

Short pulses disperse faster in the transmission fibre, which is in turn beneficial in terms of optical signal-to-noise ratio, resulting in a twofold increase in transmission distance over NRZ for a 3 dB power penalty criterion.