毫米波大规模MIMO中支持低精度量化的空时频同步技术

在节能降耗需求的驱动下，使用低精度模数转换器（ADC）是解决毫米波大规模多入多出（MIMO）系统面临的高能耗和高成本的途径之一。但是，低精度ADC会影响系统同步精度和频偏估计性能，而信道相关性会导致波束间的功率泄漏。如何解决好系统空时频同步（即波束对准、时间同步和频偏估计）是低精度ADC面向工程应用的重点和难点问题。对此，通过构建信道空间模型和基于概率的虚拟量化模型，项目从三个方面展开创新研究：1）以广义似然比和信号-量化-噪声比为准则提出鲁棒的空时交替同步方案；2）考虑空间相关性，使用波束分层策略降低同步开销，基于带外辅助思路设计能够抑制波束间功率泄漏的波束码本。针对时间同步，设计对量化误差具有鲁棒性的同步序列；3）利用机器学习完成频偏参数的概率建模，考虑量化误差的统计特性，提出基于消息传递的频偏和信道联合估计方案以提升性能并降低复杂度。项目一定程度上推进低精度ADC在实际系统中的应用。

Under the motivation of saving energy and cutting consumption, the use of low-resolution analog-to-digital converters (ADCs) is a promising solution to reduce high power consumption and hardware costs in millimeter-wave massive multiple-input and multiple-output (MIMO) systems. However, we have found that low-resolution ADCs severely degrade the performance of synchronization and frequency offset (FO) estimation, and that channel correlation causes power leakage among different beams. In this sense, it is crucial to efficiently address space-time-frequency synchronization (i.e., beam alignment, time synchronization and FO estimation) for the application of low-resolution ADCs in the practical systems. To this end, by establishing spatially correlated channel and virtual quantization models, in this project, we perform investigations as follows: Firstly, we propose a robust space-time synchronization scheme based on generalized likelihood ratio and signal-quantization-noise ratio via alternative optimization. Secondly, taking spatial correlation into account, we divide beam vectors into different layers to reduce the overhead of beam alignment. Moreover, we design the codebook that can eliminate the power leakage among different beams based on an out-of-band method. In addition, for the time synchronization, we design the sequence that is also robust to quantization errors. Finally, we obtain the probabilistic model of FO parameters by means of machine learning and try to propose a low-complexity and performance-improved method based on message passing (MP) for joint FO and channel estimation by applying the statistical characteristics of quantization errors. To a certain extent, this program will promote the application of low-resolution ADCs in the practical systems.