Measurement-based Interference Analysis in Massive MIMO Systems

Many users worldwide experience faster connections with lower latency, which are crucial for accessing futuristic services such as: efficient emergency mobile communications, over the air monitoring with UAVs, streaming virtual and augmented reality, and more. This is thanks to the successful implementation of 5G and its vital technology, massive MIMO. In massive MIMO systems, a large number of antennas are deployed at the base station, enabling simultaneous data transmission to multiple users at the same time and with the same frequency resources, which leads to an increase in spectral efficiency and capacity. Massive MIMO can be deployed indoors and outdoors; in both cases, the allocation of the antennas and interference suppression are critical aspects to obtain a better system performance. This thesis begins with the measurement-data analysis of three different indoor antenna arrays: URA (Uniform Rectangular Array), ULA (Uniform Linear Array) and D-ULA (Distributed ULA). The results show three main advantages for the D-ULA scenario: First, based on an approximated method, we conclude that this scenario provides better favourable propagation conditions, representing close-to-zero orthogonality between the normalised users' channels. Second, the D-ULA scenario provides better coverage as the users will be closer to at least one antenna. Third, due to better favourable propagation conditions, the D-ULA scenario has a better power focus in contrast with URA. The D-ULA benefits are then quantified in terms of spectral efficiency, showing that this particular scenario has the highest sum peak spectral efficiency. Mainly because the favourable conditions of this scenario are translated into intrinsic better interference suppression. Hence, by implementing ZF received combining in this experiment, the D-ULA scenario could simultaneously serve 18 more users than the URA scenario, before reaching a saturation point. The second analysis of this thesis is interference suppression. A system achieves a higher performance when there are noise and full interference suppression, which means that all the signals from interfering users or to interfered users are suppressed. However, this procedure is not scalable. Therefore, interference suppression of only a subset of harmful users, also known as partial interference suppression is analysed as a scalable solution. Partial interference suppression is studied with the aid of three proposed combining vectors based on ZF and MMSE techniques. These combining vectors are E-ZF (Extended ZF), PMMMSE (Partial Multi-cell MMSE) and P-MMSE (Partial MMSE), the latest one is a versatile version that can be applied to any system given its user clustering strategy. While the proposed combining vectors are essentially known, the selection of the users for partial interference suppression is still an area of research. Our analysis provides a complete study of different user selection methods combined with known methods for partial interference suppression. The results of the experimental data analysis show that in an outdoor scenario, the users that cause the highest interference are those with the larger channel gain towards the analysed users. On the other hand, in an indoor scenario, the suppression of eigenvectors obtained from the wireless channel matrix of the interfered users, is more beneficial. The conclusions of the indoor experiment validate the results obtained by a simulated indoor scenario under the WINNER II channel model. Another option to overcome interference is cell cooperation. This thesis also studies the cooperation between cells in terms of spectrum and infrastructure sharing based on experimental data. When the spectrum is shared across all cells, this system has a higher capacity. Moreover, as it is expected, when all the antennas serve all the users by sharing infrastructure and spectrum, it reaches the highest capacity and spectrum efficiency but requires a massive computational cost in a centralised CPU. This thesis mainly relies on experimental data collected, with the aid of the KU Leuven Massive MIMO testbed, in indoor and outdoor experiments.

Publication year:2022

Accessibility:Open