ANN based model of automatically gain controlled EDFA in WDM systems

摘要

Erbium Doped Fiber Amplifier (EDFA) has revolutionized the optical communication system as its ability to amplify the signals and enabling the transmission upto thousands of kilometres. With the advent of WDM technology, along with EDFA realized the effective utilization of bandwidth paving the way for several generations of advancement in the communication network. Several automatic gain control techniques are widely used to compensate the gain fluctuations in a WDM channel, arising due to the power fluctuations at the signal, as a flat gain spectrum across the whole usable bandwidth is preferred because of accumulated imbalance likely to happen in different ways. This launching power discrepancy between different channels give rise to imbalance in received power and signal to noise ratio (SNR) and directly affects the system performance. Firstly, the disparity in received power can be outside of the dynamic range of the receiver and then the SNR degradation would cause the BER to fall below the required minimum due to inadequate gain compensation. Therefore, an effective communication system requires optimized gain stabilization techniques along with all other requirements of quality signal reception. In this paper, we attempt to model a feed forward EDFA with automatic gain control (AGC) using artificial neural networks (ANN). Detailed study is carried out and the system is verified with the experimental results in C-Band, and, very promising results could be achieved. In the characteristics, we are mainly trying to quantify the gain and optical noise figure as a performance measure of the system. The flattened gain calculated as the ratio of maximum to minimum signal power at the receiver is 1.16 dB against the allowable range of 3 dB. The ANN model computes with an accuracy of mean square error (MSE) of 3.9717 x 10(-5), justifies an accurate forecast with a low computational time in milliseconds range.