Study On The Key Technologies Of Dwdm Transmission System Beyond100Gbps

With widespread popularization of Internet applications and unceasing emergence of various broadband services, the broadband traffic and network capacity are growing rapidly. As the construction of national information infrastructure, optical transport network is faced with enormous pressure, and requires high bit rate, long-haul optical transmission technology urgently. The implementation of coherent optical transmission systems per wavelength channel beyond100Gbps will provide supporting technologies for high-speed, high-capacity and long-distance backbone transmission network in the future. Due to the growth of channel rate and spectrum efficiency, optical transmission systems beyond100Gbps are more sensitive to all kinds of impairments. There are several technical problems in the digital signal processing (DSP) algorithms that should be solved immediately including very high chip clock speed, very high computational complexity and the interactional impairments. In this paper, the research mainly focues on the key DSP algorithms for coherent optical transmission system beyond100Gbps, and achieves a number of innovative results. The research contents are joint scheme for symbol, sampling clock and carrier frequency synchronization in PM-CO-OFDM, parallel DSP structure and low complexity nonlinearity compensation algorithm for PDM-CO-OFDM, low complexity nonlinearity compensation and impairments compensation by aggressive optical filtering in Nyquist WDM system. The main work and innovative contributions are listed as follows.1. The CO-OFDM transmission system must keep the orthogonality between subcarriers, and the high accuracy of synchronization algorithms are indispensable to CO-OFDM. However, the existing algorithms are mostly designed for a specific impairment, and the interaction effects between the different impairments have not yet been considered. In practice, it is worth noting that the interaction effects will seriously deteriorate the system performance and should not be ignored. In order to solve this problem, a joint scheme for symbol, sampling clock and carrier frequency synchronization is proposed. The phase impairments caused by symbol synchronization error (SSE) and sample timing error (STE) are designed to be compensated in channel equalization. An algorithm is proposed to estimate sampling frequency offset (SFO) and residual carrier frequency offset (RCFO) simultaneously based on linear relationship between subcarrier, symbol position and SFO, RCFO. The estimation results of SFO and RCFO are fed back to time domain to correct the sampling clock frequency by using interpolation filter and compensate for the phase impairment caused by RCFO, respectively. Taking480Gb/s PDM-16QAM CO-OFDM system as an example, the proposed scheme can correct the interactional synchronization errors and achieve excellent BER performance, no OSNR penalty compared to the system with ideal condition.The simulation shows that the scheme can successfully track±3.5GHz carrier frequency offset (CFO)and±2000ppm SFO.2. We analyze the core DSP algorithms of CO-OFDM, including dispersion compensation, symbol synchronization, carrier frequency synchronization, sampling clock synchronization, channel estimation, phase recovery, and balances the contradiction between chip clock speed and resource. A parallel processing structure and modified DSP algorithm for real-time coherent optical OFDM system beyond100Gbps are proposed. To compensate dispersion, data need to be transformed to frequency domain by FFT which would convert the data from serial to parallel. Parallel channels are multiples of the subcarriers after dispersion compensation. The parallel processing mechanism reduces the request for the DSP-chip rate, and reach a balance between the number of parallel channels and the DSP-chip area. Double training symbols are used and double correlation peak decision mechanism is proposed to ensure that the starting point can be found even when one of the training symbols is destroyed by the parallel channels. Based on invariant error function of parallel channels and parallel location information known, the mechanism solves the problem that correlative control parameters of interpolation filter are not suitable for parallel processing. The simulation shows that the whole structure can achieve the same performance compared with serial systems. Besides, the request for DSP rate is reduced by three orders of magnitude.3. Back-propagation algorithm can be utilized to compensate fiber nonlinearity and dispersion in long-haul optical communication systems. It requires large step number to attain better results significantly, meaning that the relations between dispersion and nonlinearity could be ignored in each step. It is difficult to be implemented in real-time because of its high computational complexity. In this paper, we propose a low complexity compensation algorithm for nonlinearity in PDM-CO-OFDM system beyond100Gbps based on crosstalk weight. The chromatic dispersion (CD) will be estimated by channel estimation algorithm based on training sequence. The algorithm takes into account power crosstalk in neighboring samples induced by CD, and it introduces the crosstalk weight in nonlinear compensation. Then, Crosstalk weight is obtained by calculating the inter-symbol interference (ISI) and inter-carrier interference (ICI). The numerical simulations show that under the same performance the proposed algorithm can effectively reduce the computational complexity compared with the traditional back-propagation algorithm. The computational complexity is reduced by30%nearly in480Gbps PDM-16QAM CO-OFDM system.4. In order to avoid the contradiction between algorithm performance and computational complexity, we propose a low complexity maximum posterior probability (MAP) algorithm based on dot product for nonlinearity compensation and impairments compensation by aggressive optical filtering. The algorithm replaces the minimum euclidean distance by dot product, removes the calculation of training sequence power and received signal power. The performance is fully verified in1.2Tbps PM-QPSK Nyquist WDM simulations and1.024Tbps PM-QPSK Nyquist WDM off-line experiments. The proposed algorithm can effectively reduce by50%computational complexity compared with the traditional MAP algorithm. In addition, we finished and participated in1.2Tbps PM-QPSK Nquist WDM long haul transmission real-time experiment.