Trade-off and Design Considerations of the Erbium-doped Fibre Amplifiek

The trade-off between gain efficiency and noise figure is evaluated for an optimised EDFA pumped at 980nm. It is shown that operating at rrirxfmum gain efficiency elwavs results in an increased noise figure well above the 3dB quantum limit and that any attempt to reduce the amplifier noise figure results in a decrease in gain efficiency. Alternatively, for a fixed amplifier gain and pump power, increasing the fibre NA and the dopant confinement in the core reduces the amplifier NF. Finally, the input pump power required to achieve a target NF is calculated as a function of the amplifier gain, fibre NA and dopant confinement. INTRODUCTION. Erbium-doped fibre amplifiers (EDFAs) 11 1 will play an important role in the implementation of high quality, reliable, fibre-optic communication systems as line-, powerand pre-amplifiers. Gain efficiency and noise figure (NF) are among the most important figures of merit in describing and comparing the various EDFA configurations. It is well known that a combination of erbium-doped germano-silicate-based fibre and a 980nm pump-wavelength results in amplifiers with the highest gainefficiency. In addition, the co-propagating pump and signal configuration gives a reduced NF [2,31. The Performance of the EDFA is also influenced by a number of additional factors including the fibre NA, cut-off wavelength and dopant confinement. An optimisation of the fibre design is, therefore, required to maximise gain efficiency and minimise simultaneously NF. THEORETICAL MODEL. A 3-level theoretical model was employed in which the overlap-integral and equivalent ASEbandwidth approximations were used [41. A co-propagating pump (980nm) and signal (1 536nm) configuration was considered. The forwardand backward-propagating ASE were considered as quasimonochromatic waves of equivalent bandwidth Av=6OOGHt (=4.5nm), centered at the signal wavelength. The pump absorption, signal absorption and emission cross-sections were determined experimentally by power saturation measurements to be 2 . 5 5 ~ 1 02'mZ, 7 . 9 ~ 1 O-26m2 and 6 . 7 ~ 1 OZ6m2, respectively [51. The metastable lifetime was determined to be 12.1 ms. The signal input power was always -45dBm. EDFA OPTIMISATION. For each fibre NA and confinement factor (dopantlcore-radius ratio), the fibre length, input pump power and cut-off wavelength were optimised to give maximum gain efficiency. In Figures 1 (a) and (bl. the optimum gain efficiency and the accompanying NF are plotted against the fibre NA for confinement factors of 1, 0.7 and 0.5. The optimum cut-off wavelength depends only on the dopant confinement and it was found to be 833nm, 850nm and 880nm for confinement factors of 1, 0.7 and 0.5, respectively. The optimum gain efficiency is shown to increase quasi-quadratically with the fibre NA. For ultra-high gain efficiencies, confinement of the dopant inside the fibre core is needed. In Figure l(a), some of the best reported gain efficiencies 161. as well as, results obtained in our laboratories are plotted showing a very good agreement with the theoretical predictions. From Figure 1 (bl, it is deduced that, under fully optimised conditions, the amplifier NF increases with the fibre NA. It should also be noted that the amplifier NF is always well above the 3dB quantum limit. The amplifier NF deviates from the quantum limit since under fully-optimised conditions the backward-travelling ASE attains high levels and severely depopulates the metastable level, particularly close to the input end, thus, deteriorating the population inversion. This occurs even for the moderate gains ( 25dB) obtained at the maximum-gain-efficiency. From Figures 1, it is concluded that optimum (maximum) gain efficiency and near-quantum-