5.1 Introduction

In the previous chapter, we have seen that in lossless and constant-dispersion fiber, collisions between ordinary solitons of distinctly different frequencies are perfectly elastic. That is, after the collision is completed, there is no net exchange of energy or momentum of either soliton. This perfect transparency would seem to make ordinary solitons ideal for use in dense WDM. Unfortunately, however, the loss and/or varying dispersion of real fibers tend to destroy the symmetries necessary for such transparency, so that the emerging solitons suffer significant frequency shifts (which dispersion then converts into timing shifts) and loss of energy. In principle, these defects can be perfectly corrected through the use of fibers whose dispersion profile, D(z), tracks the loss/gain-induced intensity profile I(z) (for example, fibers with exponentially tapered D used between lumped amplifiers.) Indeed, as will be reported on later in this chapter, step-wise approximations to such exponentially dispersion-tapered spans have allowed for successful experimental demonstration of six-channel WDM over distances of 10,000 km or more. Further expansion of the channel count, however, is severely limited by the very narrow range of wavelengths over which single fiber types can provide the correct sub-picosecond/nn-km dispersion values necessary for ordinary solitons, and the cost of production of fibers with custom-shaped dispersion profiles is seen as prohibitively expensive. Thus, in the commercial world, dense WDM with solitons is now concentrated almost entirely on the much more versatile and practical dispersion-managed solitons, as will be described in the next chapter.

In...