Experimental 5G Network Performance Analysis in a Semi-Anechoic and Semi-Reverberation Chamber

5G New Radio wireless networks based on the International Mobile Telecommunications specifications are fundamental for current mobile communications. This work focuses on the performance analysis of a private 5G network in an experimental and controlled environment. For this purpose, exhaustive experimentation considering 192 combinations of factors has been proposed and carried out in a controlled environment within a semi-anechoic and semi-reverberant chamber. This enables accurate observation of the impact of the studied factors on a series of Key Performance Indicators that are fundamental in 5G networks, such as latency, throughput, and reliability. This dataset has been analyzed using statistical algorithms such as Principal Component Analysis and SHapley Additive ExPlanations. The results show how network performance can be affected not only by the environment, but also by its configuration. Furthermore, the configurations contribute differently depending on the indicator, showing that these can be improved without an excessive impact on other indicators: (i) throughput improves by more than 26% by modifying parameters such as bandwidth, duplexing method, or number of antennas; (ii) latency improves over 7% by modifying the duplexing method or 12% by switching between radio access technologies; (iii) reliability is exceptionally high in controlled environments. Our study shows that reliability can achieve improvements of up to 196.5% in terms of lost and retransmitted packets. The insights found in this work, based on the proposed configurations and their respective improvements, can be extrapolated to proposed use cases for 5G networks.