Application Of Feed Forward Neural Network Equalizer Based On Millimeter Wave Radio Over Fiber System

With the continuous improvement of users' demand for audio,video and multimedia services,broadband has become an important development direction in the communications industry.At the same time,the entire world is entering an era of all-things interconnection,and the traditional fixed access method can hardly meet more and more flexible access requirements.Radio over fiber(RoF)communication technology combines the bandwidth advantage of optical fiber communication with the flexibility of wireless access,and has become a research hotspot in current communication technologies.However,with the increasing number of wireless services,although people have adopted the modulation format with higher frequency utilization,inevitably saturation occurs in the low frequency band below 10 GHz.As a result,millimeter-wave RoF technology working around 60 GHz has become an important solution.However,RoF communication system in the transmission process,due to the complexity of its channel,the signal will introduce a lot of distortion.In order to ensure the quality of communication,channel equalization need to be considered.The main research results of this paper include the following two aspects:First of all,completed the work of design and construction for millimeter-wave RoF communication system platform.Our work is divided into two parts.The first part is the design and construction of long-distance intensity modulation-direct detection(IM-DD)transmission system.The system transmits a OOK signal with an information rate of 2 Gbps and passes through a 75 km optical fiber transmission.The purpose is to preliminarily analyze the equalization ability of the designed equalization algorithm and serve as a reference for RoF communication system.The second part is to design and build a 60GHz millimeter-wave RoF communication platform.The platform uses phase-modulated filtering method to generate millimeter-wave signals,which effectively reduces the requirements on optical devices.By using the phase modulator to modulate the optical carrier and the microwave signal at 30 GHz,two first-order sidebands are generated.After the carrier is filtered by the optical comb filter,the high-frequency Mach-Zehnder intensity modulator is used to modulate the 5 Gbps OOK signal to two sidebands,after 10km optical fiber transmission through the photodiode beat frequency,60GHz millimeter wave signal is generated.Then through 1.2m wireless channel transmission,using the oscilloscope to receive,sample and so on.Millimeter-wave RoF communication with information rates up to 5Gbps is achieved.Then,the corresponding channel equalizer is designed according to the relevant characteristics of the constructed millimeter wave RoF communication platform.Including LMS equalizer,BP and its improved algorithm and RBF neural network equalizer.The activation function of the implicit layer of BP neural network equalizer is the sigmoid function.and the learning algorithm is the gradient descent method.At the same time,a variable step length improvement algorithm with iterative variation is designed.The activation function of the implicit layer node of the RBF neural network equalizer is the inverse multi-quadratic function,and the k-means algorithm is used to get the clustering center.The experimental results show that BP neural network equalizer can reduce the error rate of RoF system by 100 times below compared with the traditional LMS equalizer,and its performance has great advantages.At the same time,the performance and stability of the variable step BP algorithm are better than the basic algorithm.In addition,the application of RBF equalizer is not good,and some conjecture is given in this paper.