Massive multiple-input multiple-output (MIMO) is one of the key enabling physical layer technologies to address the massive capacity requirement demanded by 5G systems. Massive MIMO exploits the use of large antenna arrays at the base station (gNB) to simultaneously serve multiple users through spatial multiplexing over a channel. Massive MIMO relies on uplink pilots to obtain channel state information (CSI), exploiting channel reciprocity and time division duplexing (TDD) operation. In reality, however, the communication channel does not only consist of the physical channel in the air, but also the radio-frequency (RF) front-ends in transceivers which are not reciprocal. Therefore the system needs to be calibrated before channel reciprocity can be exploited. Distributed massive MIMO with spatially separated antennas gives a higher spectral efficiency and enhanced coverage area, compared to collocated massive MIMO. Nevertheless, coordinating a large number of remote radio units (RRUs), forming the gNB, is a big challenge. Hence, TDD reciprocity calibration and RRU synchronization are the two key factors to enable distributed massive MIMO.
In this thesis, we focus on deploying a distributed massive MIMO system on the OpenAirInterface (OAI) 5G testbed and applying real-time channel calibration algorithms in order to evaluate their performance. The main contributions can be summarized as follows. First, we implement the pre-coder function and the multi-thread parallelization for the optimal performance of the functional splits in our Cloud-RAN (C-RAN) system while increasing the number of active RRUs. Second, we present the low-cost solutions for the hardware issues resulting from our RRUs forming the distributed antenna system (DAS). Also, we analyze the methods used for time/frequency/phase synchronization and calibration in our testbed. Third, we carried out real-time measurements on our C-RAN testbed in order to prove the stable and precise synchronization between several RRUs and confirm the efficiency of the proposed group-based reciprocity calibration scheme. Fourth, we provide a ground truth for the evaluation of the group-based over-the-air (OTA) calibration framework through channel measurements on a simulated DAS. Last but not least, enabled by TDD reciprocity calibration, we built up a multiple-input single-output (MISO) testbed based on the OAI platform, in order to facilitate the evaluation of relative calibration and simultaneously access the performance of the MIMO antenna prototypes designed by the team in Orange labs.