The Researh On Bias Control Algorithms Of Electro-optic Modualtor Based On Machine Learning

Driven by bandwidth-hungry such as cloud computing and the Internet of Things,Internet data traffic will increase dramatically in the next few years.As a main transmission form of backbone networks and metropolitan area networks,optical communications will also face unprecedented challenges in the future.As an electro-optical conversion device,electrooptic modulators are widely used in optical communication and microwave photonic due to their outstanding performances.The electro-optic modulator needs a DC bias voltage to ensure that it works in a specific state for the reason of application scenario requirements or machining error.However,the operating point of the modulator might drift in practical applications due to some external factors such as temperature,mechanical stress,or the uneven electro-optical characteristics of the modulator,affecting the performance of the link.Therefore,the electro-optic modulator usually requires a bias control circuit to ensure the long-term stable operation of the modulator.This paper analyzes the causes of the bias drift of the electro-optic modulator based on the basic principles of the electro-optic modulator,and also analyzes the basic principle of implementing the bias control.After analyzing the shortcomings of the current bias control algorithms,we propose a bias control algorithm based on an artificial neural network.This algorithm uses a two-layer artificial neural network to automatically learn the non-linear relationship of the average optical power and bias voltage of the modulator.Then it can quickly predict and correct the bias of the modulator based on the monitored signal.Finally,we validate the effectiveness of the algorithm on a Mach-Zehnder modulators operating at the quadrature point and the null point driven by a 56-Gb/s PAM4 signals and 28-Gb/s BPSK signals,respectively.It is proved that the algorithm can meet the requirement of locking the modulator at any point of its transmission curve.In addition,this paper also analyzes the bias control algorithms for IQ modulator,which are widely studied in optic commnication.Most bias control algorithms for IQ modulator can only address the situation where the bias point is set to the null point of the modulation curve,and cannot meet the application of arbitrary point locking.Also,most of these algorithms require integration of multiple cycles of signals or a high computational complexity fast Fourier transform algorithms for spectrum analysis.Therefore,this paper also proposes a bias control algorithm for the IQ modulator based on ANN.The ANN learns and analyzes the half-cycle of the time-domain signal output by the IQ modulator and extracts the harmonics for bias control.Finally,we verified the feasibility of the algorithm with matlab.