Millimeter Wave Antenna Array Calibration and Validation for 5G New Radio Access

Abstract: This Master thesis work was performed in Ericsson AB, Lund. It studies and validates two classes of algorithms to be used in mm-wave massive MIMO antenna arrays, for their use in the future 5G mobile communication systems. The first class of algorithms relates to the calibration of the transceiver (TRX) chains responses of the antenna array and makes use of the antenna array mutual coupling, which is considered as known by the system. It is used to compensate for inaccuracies in the TRX base-band complex responses, caused by environmental conditions such as temperature, humidity and aging, which are much more critical in mm-wave than conventional bands. After the calibration, the beamforming capabilities of the massive MIMO systems are increased. The second class of algorithms is related to the estimation of antenna array mutual coupling matrix, due to the fact that this information is used in the first method. The antenna array mutual coupling estimation algorithm was proposed by Ericsson's engineers and tested in this work. Since the systems considered will work at mm-wave frequencies, small construction errors can create big differences in antenna array coupling properties, so the coupling matrix must be estimated for each constructed system of the same kind. Several estimation and calibration algorithms were simulated, using Matlab® as a software for simulation, and analyzed. The estimation of TRX chains' complex responses is needed. Two estimation algorithms are used, referred as linear and non-linear least squares estimation. These estimation algorithms need to use the information regarding the antenna array mutual coupling matrix and over-the-air (OTA) self-measurements between pairs of elements in the antenna array. These measurements can be done considering all the possible pairs of elements in the antenna array (full measurements) or just a subset of the closest pairs (neighbour measurements). Firstly, simulations using a generic case were done, and later, simulations considering constraints in an Ericsson radio module proprietary system were done. Internal and external unwanted interference in the radio system were considered, to check for limitations in the estimation algorithms. So as to validate the proposed methods and algorithms, a testbed system using a radio module working at 28 GHz was built and measured. The signal levels and frequencies in the HW components of the testbed were calculated using the data-sheets of the components and later measured using a vector network analyzer and an spectrum analyzer. The last task was to write the code for controlling the radio module, then perform calibration using it and finally measure the performance of the algorithm using an an-echoic chamber. Due to a lack of time, the code needed to do the OTA measurements, the OTA measurements and the validation of the algorithms are left as future work. Observing the results of the simulations, several recommendations are made for future measurement validations. One common conclusion is that it is best to perform the minimum amount of self-measurements available, with those being the ones corresponding to the strongest coupling gains between elements in the antenna array. There were different preferred algorithms for calibration depending on the value of the signal-to-noise ratio (SNR) and the value of the signal-to-interference (SIR) ratio. Regarding the antenna array coupling matrix estimation, the conclusion is that the algorithm proposed by Ericsson's engineers works, but in order to achieve good results, the required post-processing SNR values of the related measurements may be too big. Therefore, it may take too long to perform these measurements. Possibilities to improve this algorithm are recommended and left as future work. Regarding the HW to be used in the testbed, it is recommended to add some extra components in order to improve the quality of the signals in the system, for future measurements.