🤖 AI Summary
To address the high deployment costs, regulatory constraints, limited coverage, and security risks associated with private 5G networks in licensed spectrum, this paper presents the first fully open-source, end-to-end private 5G system operating in globally unlicensed bands (e.g., 5 GHz). The system leverages commodity servers, Open5GS for the core network, srsRAN for the RAN, USRP software-defined radios, and a deeply optimized Linux protocol stack with custom PCIe drivers—achieving full-stack autonomy and controllability. It marks the first complete implementation of the 5G NR protocol stack in unlicensed spectrum, eliminating dependence on proprietary hardware and licensed spectrum allocation. Experimental evaluation demonstrates a downlink peak throughput exceeding 220 Mbps, end-to-end latency under 15 ms, and stable support for voice, video, and URLLC services. All source code, configuration files, and performance benchmarks are publicly released for full reproducibility.
📝 Abstract
The fifth generation of mobile networks (5G) represents the latest development in mobile communications. It has been designed to support several types of data traffic and to meet more challenging performance requirements than ever before. These characteristics make 5G attractive for current but also novel public and private services. However, because of reasons related to coverage, regulatory restrictions, business considerations, and cyber security, many of these novel applications can only be deployed as part of a private network. The cost of licensed frequencies makes such approach prohibitive for many stakeholders, and therefore unlicensed frequency bands represent a more affordable option. Even so, private 5G networks for use in globally unlicensed frequency bands do not yet exist. In this paper we present the first end-to-end private 5G network operating in a globally unlicensed frequency band, using general purpose computers, open-source software and software-defined radio. We evidence its working and show that the choice of the hardware can significantly affect the performance of the network.