Application Of LDPC Codes In High-speed Coherent Optical Communication Systems

With the construction of information increasingly deepening,network traffic has maintained a 100-fold increase in 10 years in the 21st century.The past several years witness a rapid development of the optical fiber communication,where a plenty of results based on Digital Signal Processing(DSP)in the field of long distance coherent transmission have been achieved by researchers in and abroad.Forward Error Correction(FEC)code is one of the key technologies of DSP.The soft-decision(SD)FEC codes generally hold a higher coding gain over hard-decision FEC codes due to the full use of channel information the SD decoding process,therefore SD FEC codes with over 10 dB net coding gain is a hot research trend of FEC technologies for coherent optical communication systems.Turbo codes and low density parity check(LDPC)codes are the main research directions of SD FEC codes.The two kinds of codes have similar BER performance,but LDPC codes overweigh Turbo codes in the aspect of error floor performance and complexity of hardware implementation,and thus have been paid more attention.In this paper,several research works have been done around the application of LDPC codes in high-speed coherent optical communication systems.Firstly,some classical algorithms for construction,encoding and decoding were introduced and simulated.Then the 'model mismatch' penalty occurred in the application was analyzed.The paper focused on the cycle slip phenomenon in the digital coherent receiver,which is a main cause of the 'model mismatch' penalty.To solve the problem,we made efforts in two aspects:a pilot-aided carrier phase estimation algorithm with enhanced phase unwrapping process was proposed,and Turbo demodulation was adopted to mitigate the differential penalty in a DQPSK system.Main research works of this paper are as follow.(1)Construction,encoding and decoding algorithms of LDPC codesWith a brief review on the concepts and classification of LDPC codes,quasi-cyclic(QC-)LDPC codes were adopted in the paper due to the its low complexity of implementation.A class of QC-LDPC codes with girth-6 were constructed by an algebraic construction method based on Galois field.Classical encoding algorithms,such as Gaussian elimination and approximate lower triangulation,and an efficient encoding algorithm of QC-LDPC codes were then introduced and simulated.Decoding algorithms like Log-BP algorithm and Min-Sum algorithm are compared and simulated.Finally,the impact of code length,code rate and maximum iteration times of the decoding algorithm on the performance of LDPC codes is simulated and analyzed.(2)Turbo demodulation based on LDPC codes for DQPSK modulation Differential encoding is a robust method to mitigate the impact of cycle slips.However the large differential penalty makes it difficult to apply SD-FEC codes.We adopted Turbo demodulation in a DQPSK modulation system to reduce the differential penalty.BCJR algorithm was used in the MAP decoder in Turbo demodulation structure,and equations were deduced to adapt BCJR algorithm for the differential decoding in DQPSK modulation.The effectiveness of Turbo demodulation was verified by MATLAB simulation,and the number of inner and outer iterations were optimized by simulation.A 1.8 dB reduction of differential penalty was obtain by Turbo demodulation in our simulation.(3)A pilot-aided carrier phase estimation algorithm with enhanced phase unwrappingThe traditional blind carrier phase recovery algorithm generally induces severe cycle slips when enduring high phase noise,which limits the effectivity of LDPC codes and leads to a reduced post-FEC performance.A pilot-aided CPE with an enhanced phase unwrapping process was proposed by us to reduce the cycle slip rate.In this paper,the post-FEC performance of the proposed algorithm is evaluated by simulation.Results show that the proposed algorithm achieved a similar performance to Turbo demodulation with much lower complexity.