Research On Digital Signal Processing (DSP) In High Speed Coherent Optical Transmission Systems

In recent years, the development of high-speed analog to digital converter (ADC)technology, integrated optical devices and digital signal processing (DSP) resumed coherent detection technology in the field of optical communication. Laser phase noise is a major limited factor of the high-speed large-capacity digital coherent optical communication system performance. Usually carrier phase recovery algorithm is used to eliminate the influence of the phase noise of the laser, however the currently proposed techniques is either too complex or influenced by the phase jump disadvantage(called Cycle Slip). In this thesis the application of the digital signal processing algorithms is being studied, focusing on carrier phase estimation (CPE) and cycle-slip mitigation technique, as well as a combinative performance research of decoding by low density parity check (LDPC) code after CPE.The main innovations of this dissertation are being summarized as follows.In the field of optical transmission for QPSK, 16QAM modulation formats the frequently used blind CPE algorithm (such as BPS, QPSK-Partitioning) can not deal with the problem of phase slip or cycle slip (CS). A low complexity joint polarization CPE techniques and a novel phase unwrapping (PU) algorithm based on pilot symbols, referred to as PAPU, have been studied. PAPU combines the conventional phase unwrapping with itself, detect and correct the CS by using the phase reference from the pilot symbols. This technique successfully avoid the traditional differential encoding and decoding process and eliminate the pre and post FEC differential penalties. For the lOGbaud polarization multiplexed 16-QAM system by using 1.56% overhead, the 6MHz linewidth has been realized near the soft decision limit 2E-2 with 15dB SNR through simulation In addition for the 30 Gbaud QPSK system with 4 MHz linewidth and 1 dB SNR penalties limit slipless CPE is achieved by only 0.39% overhead through simulation For the single polarization 28Gbaud QPSK system by using 1.56% overhead 2MHz linewidth tolerance has been achieved near the BER 1E-3with 13dB OSNR margin in a back to back offline experiment. And 0.5dB differential penalty has been avoided successfully.By combining carrier phase estimation (CPE) and FEC , an enhanced pilot symbol aided CPE and CS mitigation algorithm is proposed. Comparison is performed for four scenarios (1) concatenated codes based soft differential decoding, (2) Turbo differential decoding,(3) traditional pilot symbols based carrier phase estimation (PA-CPE) and (4)enhanced PA-CPE . Through the simulation of more than 109 bits of data, it is found that there exists error floor for the traditional PA-CPE even if concatenated code is used. For Turbo differential decoding by using cascading code, error-floor is also eliminated,however there is still 0.2dB penalties compared with the enhanced PA-CPE scheme. For the enhanced PA-CPE scheme 2.2dB performance gain and error-free performance after FEC is achieved with only 0.78% pilot overhead, and the gap relative to the theoretical limit is just 0.4dB.Based on K-Means clustering algorithm a blind CS detection algorithm is proposed.The main idea of the algorithm is the classification of slipped phase to locate the transition when the cycle slip occurs. Through pilot symbols slipped phase can be compensated. With laser linewidth 200KHz the simulation is done under pure laser phase noise channel. 20%overhead 64,800 length DVB-S2 code is used to test the BER performance before and after LDPC decoding. Pre-FEC BER improvement or reduced CS probability is shown to the algorithm's effectiveness. By using a rectangular interleaver of 16-codeword depth, the error floor caused by the residual cycle slip has been successfully eliminated in the post-FEC BER curve, which exists only about 0.7dB gap relative to the BER curve over AWGN case.