Reducing optical power variation in amplified optical network

In an optical network where lightpaths may encounter a chain of Erbium doped amplifiers (EDFAs), power non-uniformity is a problem of concern. Practical EDFAs have non-flat gain profile. Thus, some wavelength channels experience higher gain than other channels. Automatic gain control (AGC) used in EDFA only ensures constant aggregate gain and does not address the non-flat gain shape. The AGC, combined with non-flat EDFA profile, cause the actual per channel gain to vary according to the input power profile. In fact, higher channels will often "steal" the available gain from lower power channels. This gain variation, compounded over a cascade chain of amplifiers, translates into received power variation. Together with limited receiver dynamic range, this severely impacts the lightpath performance and complicates lightpath engineering. Furthermore, the effect of gain variation can propagate from node to node and be exacerbated in a mesh network where lightpaths weaves in and out of each other. This paper analyzes the non-uniform gain problem and shows that even under the worst-case situation, the gain variation is bounded. As a result, simple engineering rules are adequate to ensure the performance of a lightpaths even in a dynamic environment. Furthermore, this paper suggests a novel pre-emphasis scheme. Appropriately de-emphasizing the strong wavelength channels and emphasizing the weak channels at the light path origins can reduce the power fluctuation at the receivers. This will allow the lightpath to traverse a larger number of EDFAs before regeneration is needed.