EDFA gain modeling using Pspice and neural networks

A new methodology in modeling the performance of an Erbium Doped Fiber Amplifier (EDFA) Gain is presented in this dissertation. It is based on the gain characteristics of an EDFA under different input signal power spectra and pump levels. The novelty of this dissertation is based on the training and use of a neural network applied to the performance characteristics of an EDFA.; The modeling is divided in two sections. The first model is an electrical equivalent circuit of an EDFA that can be simulated and analyzed by Pspice, a commercial electrical simulation program from OrCad. The second section is the development and implementation of a Neural Network to model the gain response of the EDFA for a given number of input power spectra.; In order to provide data and verify the performance of both models, a simulation program was developed. The simulation program's output performance was verified using an EDFA analysis program called Oasix.; The results from both models were accurate with respect to the theoretical and experimental data. The electrical equivalent model's output spectrum was proven to match the output of the simulation program. The neural network was trained successfully to identify the EDFA gain spectra to its correspondent input spectrum and pump power level with an accuracy of 98%.