Study On Signal Separation Based On Complex-valued ICA In MGDM System

With the the City Informatization progress, the demand for informationincreases continuously,the number of users of broadband access is increasing yearby year. In order to meet user demand for broadband services, The access networkrequirements to provide high-speed and unobstructed broadband channel、goodbusiness support capabilities as well as efficient service quality assurance，so itgives birth to the large-scale application of fiber to the home, and gradually becomethe best guarantee to promote the growing integration of the Information society. themulti-mode fiber has many suitable for short-distance communication,for theadvantages of easy installation, and low cost and so on. However,due to the inherentinter-modal dispersion, transmission capacity of multimode fiber is severelyaffected. In order to improve the transmission capacity of multimode fiber, theMode Group Diversity Multiplexing(MGDM) technology has been put out,Thebasic idea is that,due to the many modes in a multimode optical fiber,to use differentmode group as an independent transmission channel,Thus Multi-channel signals aretransmitted simultaneously in the same multimode fiber to improve the transmissioncapacity of the multimode optical fiber.But there is still a crosstalk between the many channels in the MGDM system,it makes the received signals are a weighted superposition of source signals.Therefore, correctly recovering the source signal from the received signal by thereception-side signal processing is one of the problems to be solved by thesystem.There was a precedent that Independent Component Analysis(ICA) isapplied to the the MGDM system to complete the baseband signalseparation.However, in the the MGDM system, when the transmission signal is asignal of advanced modulation formats, the signal is in the form of a complexvalue.Therefore, on the basis of our predecessors, the complex ICA is applied to thethe MGDM system to achieve complex-valued signal demultiplexing. The focus isto build a2×2the MGDM system and a Complex-valued ICA method is applied tothis system to transmit two QAM signals uccessfully. The specific work of the paperare as follows:Firstly, introduce the underlying technology and model of the MGDM system.Classify and summarize a variety of multiple-input multiple-output technology usedin the optical fiber,and do a brief introduction. Analyzed the mode groups used asIndependent channels in the MGDM system and their divided method, the focus is the reasrech of the principle of the mode selective excitation technique. By buildingthe simulation system.Show the different mode groups and light intensitydistribution law in the end face of the optical fiber,When the excitation of thedifferent position of the multi-mode fiber. On this basis, introduce the principle andthe model of the M N MGDM and analyze the relationship between the M andN.Secondly, introduce the methods of signal demultiplexing successfully used inthe MGDM the system currently. Roughly divided the method of signaldemultiplexing into three categories, such as direct detection of reception、channelnon-blind equalization and blind signal processing,at the same time simply analyzethe principle, advantages and disadvantages of various types of methods. Introduceindependent component analysis one method of the blind signal processing,including the basic model, constraints, uncertainties and the basis of independencejudgment.Then introduce Complex-valued ICA model and the basis ofindependence judgment of the Complex-valued ICA. Show the main steps of thefixed-point complex ICA algorithm based on negentropy maximization,and selectthree different non-linear function of T-CMN to process the mixed signalsrespectively,Analyze the result of the signal separation.Thirdly, research appling the Complex-valued ICA to MGDM system in theory,firstly analyze the constraints of the Complex-valued ICA and the MGDM systemmodel, then through the simulation, apply the real_valued ICA and Complex-valuedICA into the processing of Complex-valued signals. And analyzed the simulationresults.Finally, design and build a22MGDM system, to transmit two differentsignals by selective excitation of the different mode groups and Coupled them into amultimode fiber via a multimode coupler in accordance with a certain coefficient.At the receiving end, process the received signal using fixed point complex ICAalgorithm (T-CMN)based on negative entropy maximization through DSP module.And do the transmission simulation respectively by changing different parameters inthe system, such as coupling coefficient, OSNR, a partial core excitationeccentricity distance and so on. To obtain the results of the processed signals by theComplex-valued ICA under different parameters, to compare and analyze theprocessed signal constellation diagram and the bit error rate. To get the influencelaw of the various parameters on signal separation, Show the error rate comparisonchart with the change of coupling coefficient and the optical signal-to-noise ratiowhen the the different non-linear functions are selected.The Complex-valued ICA completethe demultiplexing of the complex-valuedsignals successfully in the case of without the pilot estimation to the channel and without the matrix inversion to reduce the complexity of the system, while alsoconducive to maintaining a low-cost design of the system, making the MGDM costadvantage is more obvious.