【作者】 李新；

【导师】 顾畹仪；

【作者基本信息】 北京邮电大学 ， 通信与信息系统， 2012， 硕士

【摘要】 相干检测是一种全息检测技术,它可以允许信息被编码在光纤的所有自由度上,包括幅度,相位以及偏振。因此,它可以提高信号功率以及系统的频谱效率。近年来,由于高清电视、视频会议等新业务的驱动,网络容量迅速增长,之前的带宽已经不够使用,具有高频谱效率的相干检测技术开始受到人们的广泛关注。同时,高速率模数转换器(ADC)的实现以及大规模集成电路(VLSI)的发展使得对光纤传输损伤在GHz波特速率上进行数字补偿成为可能。因此,与数字信号处理(DSP)技术的结合将使相干检测技术更加具有吸引力。对于相干检测后的信号,DSP算法可以对光纤信道的线性效应做到无损补偿,例如色散效应和偏振模色散效应。此外,DSP算法也可以实现发射端激光器与本地激光器的载波同步。基于以上原因,对于相干检测技术中DSP算法的研究非常火热。本文首先介绍了基于DSP的相干光系统的基础知识,然后介绍了目前应用在相干光通信中的DSP算法,包括色散补偿、解偏振串扰、非线性补偿、频率估计以及相位估计算法五个方面。这些算法普遍存在着复杂度大的缺点,为了降低硬件功耗,复杂度必须要降低。作者在相位估计和频率估计这两个方面做出了改进。对于频率估计中的共轭M次方算法,它需要采用复数乘法器来去除调制相位信息,这会带来硬件功耗的增大。基于此,作者提出了一种低复杂度的M次方算法,这种算法采用查表器件来去除调制相位信息。与传统的共轭M次方算法的性能一样的同时,这个算法不需要采用复数乘法器,因而使复杂度大大降低。此外,作者还在相位估计算法方面做出了创新。盲相位搜索算法是目前相位估计中精度最高的算法之一,它首先假设出所有可能的相位噪声值,然后再借助欧式距离最短原则选出最接近于真实相位噪声的测试相位。为了保证盲相位搜索算法的高精度,在盲相位搜索算法中一般采用逐符号估计,同时测试相位个数也要求很多,这会带来很大的复杂度。在本文中,基于相位噪声慢变的特性,作者提出了粗估计和细估计相结合的两步法以及滑动块的思想,降低了测试相位的个数以及每个符号重复计算的次数。通过采用两步法以及滑动块的思想,原始的盲相位搜索算法的复杂度大大降低。仿真结果表明,在激光器线宽容忍度降低11%的同时,算法复杂度至少可降低一半。更多还原

【Abstract】 Coherent detection can detect all the information encoded in the optical signal, which allows information symbols to be encoded in all the degrees of freedom available in a fiber, leading to improved power and spectral efficiency. Because of the business requirements of high-definition television and video conferencing, the capacity of network grows rapidly recently, which leads to the fact that the existed bandwidth cannot meet the network’s need. As a result, coherent detection technology arouses strong interest of researchers because of its high spectral efficiency.At the same time, the realization of high-speed analog to digital converter (ADC) and very large scale integrated circuit (VLSI) enables the digital compensation of fiber impairments in the GHz speed. Thus, the digital-signal-processing (DSP) technology will bring coherent detection more advantages. After coherent detection, DSP algorithm can compensate linear effects of fiber lossless, such as chromatic dispersion and polarization mode dispersion. In addition, carrier synchronization can also be done by using DSP algorithms. For these reasons, many researchers begin the study of DSP algorithms after coherent detection.This paper first introduces the basics of the DSP-based coherent optical system, and then describes the DSP algorithms used in the current coherent optical communication systems. The DSP algorithms discussed in this paper mainly include dispersion compensation algorithm, polarization crosstalk solution algorithm, non-linear compensation algorithm, frequency estimation algorithm as well as phase estimation algorithm. However, most of these algorithms have the shortcoming of high complexity. Focusing on the problem of high complexity, the author has made improvements in frequency estimation algorithm and phase estimation algorithm. The conjugate M-th power algorithm uses complex multiplier to remove the phase modulation information in frequency estimation, which will bring high computational effort. Based on this shortcoming, the author proposed a low complexity M-th power algorithm, which uses look-up table device to remove the phase modulation information. Besides the same performance with the conjugate M-th power, the new algorithm does not require complex multipliers, thereby greatly reduces the computational effort. In addition, the author also made innovation in phase estimation. Blind phase search is one of the most accurate phase estimation algorithms, which first assumes all the possible phase noise values, and then chooses the nearest test phase based on the principle of the shortest Euclidean distance. In order to ensure its high accuracy, the blind phase search algorithm commonly uses symbol-by-symbol estimation and a large number of test phases, which will bring great complexity. Based on the slow-changing characteristic of phase noise, the author proposed a two-step method which combines coarse estimation and fine estimation, as well as the sliding block method. The modified algorithm reduces the number of test phases and the repeated computational times of each symbol, and therefore reduces the complexity of the original blind phase search algorithm. The simulation results show that the complexity can be reduced by 50% at the cost of a 11% reduction of laser linewidth tolerance.更多还原