Optical fibres : analysis, numerical modelling and optimisation

Abstract

For several decades silica-based optical fibres have been used for telecommunication and sensor purposes. The single-mode fibre is frequently employed in long-distance networks, whereas the multi-mode fibre is the preferred means of signal transport in campus and in-building networks. Because of the huge bandwidth of optical fibres in comparison to its electrical wireless and copper-based counterparts, the demand for optical fibres keeps increasing. In a competitive market, fibre manufacturers aim to produce ever better fibres that are as cheap and easy to employ as possible. As fibre research, development and manufacturing is a mature discipline, improvements in fibre design can only be achieved through the construction of robust, accurate and efficient numerical fibre models for the computation of those quantities that determine the behaviour of the fibre. We have developed a modular software code, based on Maxwell’s equations, to compute these quantities in a vectorial full-wave way for both single-mode and multi-mode optical fibres. Key is the refractive-index profile, or, more specifically, the dopant profile, as it defines the propagation, splicing and bending-loss characteristics of the fibre. For the single-mode fibre, the fibre quantities that we have concentrated on are dispersion, dispersion slope, mode-field diameter, effective area, bending loss, effective and theoretical cut-off wavelength and MAC-value. We highlight one fibre quantity in particular, viz. the computation of the bending loss in a single-mode fibre. Many approximate modelling techniques have been developed to estimate this loss in a fast way. Our numerical scheme, however, is the first rigorous one, as we have performed a vectorial full-wave analysis of the bent optical fibre. In this context, triple integrals involving products of Bessel functions with large, complex order and argument appear. Due to cancellations in the pertaining computation, a high relative accuracy is needed for the computation of each product. As a result, it takes weeks on a contemporary computer to compute the bending loss as a function of the radius of curvature. We have used the vectorial full-wave bending-loss results to determine the most appropriate approximate method. Subsequently, we have extended that approximate method to compute the bending losses of higher-order modes, since the required effective cut-off wavelength depends on the bending loss of the first higher-order mode. The selected approximate method has been used in the ensuing bending-loss calculations. Since the fibre properties are often conflicting, it is a challenging task to adapt the radial dopant profile to meet a set of predefined design goals. A design goal is a combination of desired values for (some of) the aforementioned fibre quantities, and can mathematically be translated into a cost function. The minimisation of this cost function provides us with the optimal dopant profile for that specific set. For the single-mode fibre, we have performed this minimisation for piecewise-linear profiles, by employing various global and gradient-based local optimisation strategies to speed up the design step considerably. Frequently,these optimisation strategies lead to counter-intuitive dopant profile designs that could not have been contrived otherwise. We have selected a deliberate mix of several optimisation routines and have compared their performances. Perhaps the most important conclusion is that there still appears to be room for improvement in the design of the radial dopant profile of commercially available fibres. For the multi-mode fibre, vectorial full-wave optimisation is not feasible yet because of the long computation times for the large number of propagating modes. Still, our numerical scheme allows for a manual fine-tuning of the popular power-law profile to minimise differential mode delay. Further, we have included mode coupling and differential mode attenuation in our model to obtain intensity patterns that match closely with measurements. We have also analysed the influence of profile variations, e.g. on-axis dips and kinks, on the intensity pattern. A selective excitation of different mode groups in a multi-mode fibre, offers the possibility to create several independent transmission channels, and thus a higher information capacity. Recently, the feasibility of this so-called mode group diversity multiplexing technique has been demonstrated. Simulations provide us with a means to better understand its operation and possibly increase its efficiency. The channel separation may be enhanced by employing a lens between the fibre and the detector, which is called mode-selective spatial filtering. Our numerical simulations of a mode group diversity multiplexing link, with and without mode-selective spatial filtering, are in agreement with the measurements. The above discussion makes clear that the developed software code has a wide range of applicability. Moreover, it is built in a modular way and thus extensions, like the inclusion of more fibre quantities or different profile dopants, are straightforward.