Design And GPU Implementation Of The DSP Algorithm In The Coherent Receiver Of Fiber Communication

Since the increasing demands of the bit rate and transmission distance in opticalfiber communication,the traditional methods of intensity modulation/direct detectionhas been unable to meet these requirement. Instead,there is the digital coherent opticalcommunication technology with high sensitivity, good wavelength selectivity and highfrequency band utilization ration. Moreover,the increasing rate of ADC is capable ofsampling the high-speed analog signal and convert it to digital signal,that makes itpossible to apply the digital signal processing technology to the compensation forfiber-optic transmission impairments with a low system costs. Therefore,the coherentoptical communication system based on DSP technology become hot research topicsgradually.To apply DSP on the high-rate optical fiber communication system,the needs ofa processor with powerful computing capabilities is unavoidable. While GPU is sucha processor which can provide powerful computing capabilities,the development of GPUgeneral purpose computation, make it possible to do very large scale calculations usingGPU. For this reason, the GPU is suggested to do the optical communication signalprocessing,which can greatly increase the processing speed to meet the demand forhigh-rate optical communications.The DSP algorithms in the receiver was researched in depth, and the whole DSPprocess structure was designed with the suitable parallel computing algorithms whichis combined with the features of GPU,was able to take advantage of the powerfulparallel computing capacity of GPU.Then the algorithms was implemented through GPU,and the receiver was simulated.The simulation result indicates that the designed DSP algorithms have a good performancein decoding, are able to compensate for the signal transmission impairments and recoverthe original transmitted symbol constellation. Compared with the same algorithmsimplemented by CPU, the result shows that the GPU can greatly improves the speedof DSP, shorten the processing time,and can delivers significant performance gains.