A Link Quality Model for Generalised Frequency Division Multiplexing

TL;DR

This paper applies a Mutual Information based link quality model to GFDM, demonstrating its accuracy and efficiency for system-level performance evaluation in 5G scenarios, significantly reducing computational complexity.

Contribution

It extends the PHY abstraction model from OFDM to GFDM, enabling fast and accurate system-level analysis of GFDM-based 5G systems.

Findings

Model achieves within 4% PER and throughput difference compared to bit-level simulation.

Simulation time is reduced by approximately 62,000 times.

Model is effective across various channel types.

Abstract

5G systems aim to achieve extremely high data rates, low end-to-end latency and ultra-low power consumption. Recently, there has been considerable interest in the design of 5G physical layer waveforms. One important candidate is Generalised Frequency Division Multiplexing (GFDM). In order to evaluate its performance and features, system-level studies should be undertaken in a range of scenarios. These studies, however, require highly complex computations if they are performed using bit-level simulators. In this paper, the Mutual Information (MI) based link quality model (PHY abstraction), which has been regularly used to implement system-level studies for Orthogonal Frequency Division Multiplexing (OFDM), is applied to GFDM. The performance of the GFDM waveform using this model and the bit-level simulation performance is measured using different channel types. Moreover, a system-level study for a GFDM based LTE-A system in a realistic scenario, using both a bit-level simulator and this abstraction model, has been studied and compared. The results reveal the accuracy of this model using realistic channel data. Based on these results, the PHY abstraction technique can be applied to evaluate the performance of GFDM based systems in an effective manner with low complexity. The maximum difference in the Packet Error Rate (PER) and throughput results in the abstraction case compared to bit-level simulation does not exceed 4% whilst offering a simulation time saving reduction of around 62,000 times.