Investigating and Modeling Uplink Processing in 5G NR Multisector Scheduler Simulator

Sammanfattning

As the latest generation of wireless access technology called Fifth Generation (5G) New Radio (NR) evolves, computational efficiency is key to keep low cost and flexible deployments for vendors. Understanding how the Digital Signal Processing (DSP) load behaves in the digital units is one aspect to enable this efficiency. This thesis investigates how the DSP load in the base station behaves for a cell in high band, where frequencies are in the range 24.25 GHz - 52.6 GHz, when varying the number of allocated Resource Blocks (RBs) in both a single and multisector configuration. The computational behaviour is also characterized using regression models and implemented into an internal simulator at Ericsson. Additionally, the investigation is complemented with data from real User Equipments (UEs) and a more real-life scenario where error occurs in the transmission. The results show that the load scales with the number of RBs in such a way that it is possible to have more Connected Users (CUs) by limiting the allocated RBs for all users. Furthermore, the load generated by the uplink data should be considered on a Time Division Duplex (TDD) pattern basis. When doing so, it can be approximated as a linear or higher order function of the number of allocated resource blocks. When implemented into the simulator, the model is validated against data from real hardware. The validation showed that the model generates accurate normal distributed load values: it captures the essential behaviour but it is hard to predict correct mean values, due to the generated load being too stochastic in nature. These insights on the load behaviour provides input and considerations on how to design solutions that does not exceed the maximum computation utilization. For future work, the behaviour from downlink data and other traffic scenarios could be investigated to fully explore the limitations and requirements.