先进数字预失真理论及应用

发布时间：2018-02-27 17:51

本文关键词： 功率放大器数字预失真单比特频谱外推归一化增益 I/Q非平衡　出处：《中国科学技术大学》2017年博士论文　论文类型：学位论文

【摘要】：功率放大器(Power Amplifier,PA)是无线通信系统中主要的非线性器件。数字预失真(Digital PreDistortion,DPD)以其成本低、编程灵活等优点成为目前射频前端的主流线性化技术。未来DPD的发展趋势主要集中在宽带应用场景下的低复杂度结构及算法研究。模数转换器(Analog-to-Digital Converter,ADC)是DPD系统反馈回路中功耗较大、成本较高的器件之一。本文围绕降低ADC量化精度的要求与ADC的采样速率展开研究,前者能降低系统功耗与成本,以应对基站"小型化",后者则能在降低功耗的同时降低反馈信号带宽,来应对"大带宽"的趋势。除此之外,本文还会解决DPD系统中存在的诸多关键问题。传统的DPD系统依赖于带宽不低于原始信号5倍且量化精度足够高的反馈信号,使得DPD在面向宽带应用的时候受到很大限制。本文考虑降低ADC的量化精度降低到极端情况下的1比特,并对此时的DPD算法进行理论推导,证明单比特DPD系统的可行性。同时本文对单比特DPD系统中存在的问题逐一提出了解决方案,包括基于单比特采样数据的频域时间对齐算法,迭代步长的估计,系统复杂度的分析,并通过实验测试验证单比特DPD系统的线性化性能。此外本文在此基础上提出一种改进的单比特DPD系统,即前向建模辅助的单比特DPD参数提取方法,也被证明能够有效对宽带功放进行线性化。在宽带场景下一方面DPD系统反馈带宽受限,另一方面传统模型精度的提升也很困难。为此本文阐述一种采用频谱外推思想的方法,在迭代的过程中不断地用降低采样速率的反馈数据来恢复出真实的功放输出,从而能够进一步得到精确的DPD参数。本文也对频谱外推算法的数值稳定性进行理论分析,并提出一种提高算法数值稳定性的方法。同时考虑到硬件实现成本,本文为降低算法的计算复杂度设计出一种分段快速傅里叶变换方法来实现整个算法,极大地提升了计算效率。本文通过理论和仿真详细分析在基于间接学习结构的DPD系统中,不同归一化增益对DPD性能的影响。随后以DPD模块不改变输入信号的平均功率这一基本原则,推导出具有明确的物理意义的一步计算最优归一化增益的算法。另外,针对频率相关I/Q非平衡对DPD线性化性能的影响,本文提出一种消除反馈回路I/Q非平衡的DPD参数提取算法,随后将这种算法扩展到同时存在正交调制解调器两种I/Q非平衡的场景,通过实验验证算法的有效性。
[Abstract]:Power Amplifier (PA) is the main nonlinear device in wireless communication system. The advantages of flexible programming have become the mainstream linearization technology of RF front-end at present. In the future, the development trend of DPD is mainly focused on the low complexity structure and algorithm research in wideband application scenarios. Analog-to-Digital conversion (ADC) is the feedback of DPD system. High power consumption in the loop, One of the most expensive devices. This paper focuses on the requirement of reducing the quantization accuracy of ADC and the sampling rate of ADC. The former can reduce the power consumption and cost of the system. To deal with the "miniaturization" of base stations, which can reduce power consumption while reducing the bandwidth of feedback signals to cope with the trend of "large bandwidth". In this paper, many key problems in DPD system will be solved. The traditional DPD system relies on feedback signals whose bandwidth is not less than 5 times of the original signal and the quantization accuracy is high enough. In this paper, the quantization accuracy of ADC is reduced to 1 bit in extreme cases, and the DPD algorithm is theoretically deduced. The feasibility of single bit DPD system is proved. At the same time, a solution to the problems in single bit DPD system is proposed one by one, including the frequency domain time alignment algorithm based on single bit sampling data and the estimation of iterative step size. The complexity of the system is analyzed, and the linearization performance of the single-bit DPD system is verified by experimental tests. On this basis, an improved single-bit DPD system is proposed, which is a single-bit DPD parameter extraction method assisted by forward modeling. It has also been proved that the wideband power amplifier can be linearized effectively. On the one hand, the feedback bandwidth of DPD system is limited, on the other hand, it is very difficult to improve the accuracy of the traditional model. In the process of iteration, feedback data with low sampling rate are used to restore the true output of power amplifier, thus the accurate DPD parameters can be obtained. The numerical stability of extrapolation method is also analyzed theoretically. A method to improve the numerical stability of the algorithm is proposed. Considering the cost of hardware implementation, a piecewise fast Fourier transform method is designed to reduce the computational complexity of the algorithm to realize the whole algorithm. In this paper, the DPD system based on indirect learning structure is analyzed in detail by theory and simulation. The effect of different normalized gain on the performance of DPD. Then, based on the basic principle that DPD module does not change the average power of input signal, an algorithm for calculating the optimal normalized gain in one step with clear physical meaning is derived. In view of the influence of frequency dependent I / Q nonequilibrium on the linearization performance of DPD, this paper proposes a DPD parameter extraction algorithm to eliminate the feedback loop I / Q nonequilibrium, and then extends the algorithm to the scenario where there are two kinds of I / Q nonequilibrium of orthogonal modems at the same time. The validity of the algorithm is verified by experiments.
【学位授予单位】：中国科学技术大学
【学位级别】：博士
【学位授予年份】：2017
【分类号】：TN722.75

【参考文献】