5G可重构高能效毫米波前端一体化设计理论及关键技术研究

With the common requirements of high-efficiency, high reliability and low latency of information transmission in intelligent upgrading of the industry manufacturing in Zhejiang province, and combining with the next generation mobile communication technology, this project aims to carry out innovative research in the integrated design theory and key technologies for 5G massive MIMO reconfigurable millimeter-wave front-ends. Design of hardware architectures based on the traditional front-end circuits suffers from large interconnection losses, which affects the overall performance. With novel low insertion-loss 2D electromagnetic tuning technologies, this project studies the fusion methods of multi-devices and broadband multi-element decoupling approaches to build high efficient reconfigurable mm-wave power amplifiers and massive MIMO integrated front-ends. The digital predistortion (DPD) linearization technology based on big-data and deep learning methodology for the broadband mm-wave transmitters in the massive of MIMO system will be studied to optimize the overall performance of the proposed front-end systems. Furthermore, a DPD linearization system will be constructed according to the data stream in the broadband massive of MIMO transmitters. Three key scientific issues, reconfigurability, integration and high linearity, will be fully investigated through the proposed research, and a scientific breakthrough of solving the critical bottleneck for energy efficiency in 5G mm-wave transmitter front-ends can be achieved. An integrated front-end prototype will be built to demonstrate the efficacy of the proposed research and its performance will be validated with a variety of experiments. This project will lay a solid theoretical foundation on promoting the intelligent level of the industrial manufacturing equipment, realizing intelligent manufacturing and providing the basic infrastructure for efficient and reliable information transmission in Zhejiang Province.

本项目面向浙江省工业制造智能化升级对高能效、高可靠、低延迟信息传输的共性需求，结合新一代移动通信技术，在5G毫米波可重构大规模MIMO前端一体化融合设计理论和关键技术上开展创新研究。在硬件电路架构方面，针对传统前端电路互联损耗大并影响整体性能的问题，研究前端多器件融合方法和宽带多天线去耦技术，结合新型低损耗电磁二维调谐技术，构建高效率可重构毫米波功放和大规模MIMO一体化前端；在前端系统整体优化方面，基于大数据和深度学习研究宽带大规模MIMO毫米波发射机数字预失真线性化理论，形成以数据流为中心的宽带大规模MIMO发射机智能数字预失真线性化系统；通过上述研究，解决可重构、一体化、高线性三大关键科学问题，突破5G毫米波发射前端高能效的瓶颈，并研制一体化前端原型样机进行实验验证。本项目研究将为提升浙江省工业制造设备的智能化水平、实现智能制造、提供高效可靠的信息传输基础设施，奠定坚实的理论基础。

针对浙江省工业制造智能化升级对高能效、低延迟、高可靠信息传输的共性需求，以5G毫米波MIMO传输系统为目标，开展高效前端一体化设计理论和关键技术研究。经过四年努力，取得了一系列的成果。提出功放与天线一体化集成、低噪放与天线一体化、双频滤波天线一体化、宽带双极化缝隙阵列天线、多通道带通滤波器、双极化滤波天线等一体化前端电路与技术；提出频率可重构功放“电路-天线”一体化、功能可重构滤波电路、可切换双工器、可切换单通道和多通道电路、毫米波双频滤波器、毫米波双工器、微带－缝隙双频耦合器、全极化可重构全金属相控阵等可重构前端电路与天线阵；提出可重构预失真线性化、可调谐去耦和匹配等新理论与方法；提出基于切比雪夫多项式的LSTM神经网络网络模型、基于GRU注意力机制的功放模型、基于独热编码与LSTM的多频段通用数字预失真模型等；提出数字预失真超参数优化方法和深度神经网络带宽增强方法；提出数模混合预失真线性化器、5G有记忆模拟预失真器、中频孪生非线性激励模拟预失真器等宽带模拟预失真电路；研制出基于SIW的10KHz -40GHz毫米波超宽带发射和接收前端原型；构建了基于RFSOC的4路800MHz大带宽5G毫米波MIMO功放预失真线性化测试验证原型、39GHz毫米波4X1 MISO预失真线性化测试验证原型，以及云边端协同功放预失真线性化原型。在IEEE Trans. MTT、IEEE Trans. IE、IEEE Trans. AP等国内外学术期刊正式发表论文54篇，其中SCI收录34篇（中科院SCI一区6篇、二区20篇），另有6篇预发表SCI二区论文，EI收录会议论文7篇，获授权发明专利9项，申报发明专利15项；培养博士后1人，博士毕业生7人，硕士毕业生29人；2022年荣获中国发明协会发明创业创新一等奖1项。本项目研究可以为提升移动毫米波的动态数据传输能力奠定基础。