EFFICIENT TESTING AND POST-MANUFACTURE TUNING OF BEAMFORMING MIMO WIRELESS COMMUNICATION SYSTEMS: ALGORITHMS AND INFRASTRUCTURE

波束形成MIMO无线系统的高效并行测试和调谐在过去的十年中，多输入多输出(MIMO)无线通信系统的使用发生了一场革命。预计到2020年，移动数据流量将比2010年增长高达1000倍。未来的5G无线系统(通信数据速率为50Gbps)将部署大规模的MIMO系统，具有大量的发射和接收天线，以及允许射频波束形成的新型射频收发架构。5G大规模MIMO系统的研究正在以令人兴奋的速度向前推进。它将有可能将电磁波束指向移动的目标，同时以极高的速度进行通信，并将对其他用户的干扰降至最低。下载高清晰度电影将在不到一秒的时间内成为可能。然而，随着电路复杂性水平的显著提高和运行速度的提高，底层电子产品将非常容易受到制造工艺变化、电气退化和缺陷的影响。在通信的高数据速率下，设备非理想性对RF系统性能的影响可能是巨大的。功率消耗将是一个主要问题，因为将涉及大量发射器和接收器链。如此庞大的MIMO系统将需要在销售前进行广泛的测试和质量调整。在极端情况下，这类系统将需要具有内置的自测试和自调整能力，以自动补偿因电、热和机械压力造成的现场损耗。首先要解决的问题是对一系列MIMO系统进行高效的低成本制造生产测试，该系统具有处理具有10-100个射频链的5G系统的能力。现有技术的测试方法要求发往各个射频链的测试信号具有频率分离，以便独立于其他链中的非线性来评估各个射频链的非线性。相反，给定一组可以生成的频率，只能并行测试特定最大数量的RF链。一个关键目标是设计“频率高效”测试和后端响应分析算法，这些算法不需要这样的频率分离，从而允许并行测试大量的射频链。第二个关键目标是使用智能测试和响应分析算法对尽可能多的射频链进行并行增益和相位调谐。加快调谐过程的一种方法是使用上述并行测试过程，以很好地执行MIMO系统的并行调谐。这种并行调谐可以由机器学习算法支持，机器学习算法基于从并行测试技术中提取的特定的时间和频域响应特征来预测每个链的最佳调谐旋钮配置。为了进行测试和调谐，需要捕获高速信号并对其进行信号保真度分析。为此，使用建议的非相干欠采样来获取测试响应信号，通过显著简化高速器件测试和表征所需的硬件，为降低测试成本提供了重要途径。总体而言，使用建议的技术将允许大规模MIMO系统在制造后和现场进行测试和调整，而不需要复杂的测试仪器，只需10毫秒的测试时间，显著降低测试成本，同时提高产品良率和现场可靠性。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

Efficient Parallel Testing and Tuning of Beamforming MIMO Wireless SystemsThere has been a revolution in the use of multiple-input multiple-output (MIMO) wireless communication systems over the last decade. It is expected that mobile data traffic will increase by up to 1000X by 2020 as compared to 2010. Future 5G wireless systems (communication data rates 50Gbps) will deploy massive MIMO systems with large numbers of transmit and receive antennas and novel RF transceiver architectures that admit RF beamforming. Research on 5G massive MIMO systems is moving forward at an electrifying pace. It will be possible to point electromagnetic beams towards moving targets while simultaneously communicating at extremely high speeds and minimizing interference with other users. Downloading a high definition film will be possible in less than a second. However, with the dramatically increasing levels of circuit complexity and higher operating speeds, the underlying electronics will be highly susceptible to manufacturing process variations, electrical degradation and defects. At high data rates of communication, the effects of device non-idealities on RF system performance can be dramatic. Power consumption will be a major issue since a large number of transmitter and receiver chains will be involved. Such massive MIMO systems will need to be tested extensively and tuned for quality prior to sale. In the extreme, such systems will need to possess built-in self-testing and self-tuning capability to automatically compensate for field wear and tear due to electrical, thermal and mechanical stress.The first problem to solve is efficient low-cost manufacturing production test of a range of MIMO systems with the capacity to handle 5G systems with 10-100 RF chains. State of the art test methods require that test signals to individual RF chains have frequency separation for the individual RF chain non-linearities to be assessed independent of nonlinearities in other chains. Conversely, given a set of frequencies that can be generated, only a certain maximum number of RF chains can be tested in parallel. A key goal is to design "frequency-efficient" tests and back-end response analysis algorithms that do not require such frequency separation allowing large numbers of RF chains to be tested in parallel. A second key goal is parallel gain and phase tuning of as many RF chains as possible using intelligent testing and response-analysis algorithms. One way to speed up the tuning procedure is to use the parallel testing procedure described above, to perform parallel tuning of the MIMO system well. Such parallel tuning can be supported by machine learning algorithms that predict the best tuning knob configurations for each chain based on specific time and frequency domain response features extracted from parallel testing techniques. To enable testing and tuning, high-speed signals need to be captured and analyzed for signal fidelity. To this end, the use of proposed incoherent undersampling for acquisition of test response signals provides a significant avenue for reducing testing costs by significantly simplifying the hardware required for high-speed device testing and characterization. Overall, the use of the proposed techniques will allow massive MIMO systems to be tested and tuned, post-manufacture and in the field, without the need for complex test instrumentation in 10s of ms test time, significantly reducing test cost while increasing product yield and field reliability.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.