Abstract: Full-Duplex (FD) transceivers are often too complex and power-hungry for widespread implementation in user terminals. This paper investigates the 3-node topology (3NT), a configuration where an FD-capable base station (BS) simultaneously serves two half-duplex (HD) users—one on the uplink and one on the downlink. Using a tractable mathematical framework based on stochastic geometry, the authors demonstrate that 3NT can achieve performance remarkably close to a full FD network (2-node topology). Results show that with efficient self-interference cancellation (SIC), 3NT suffers only a 5% rate loss compared to 2NT. Crucially, if UE-side SIC is imperfect, 3NT becomes the superior architecture. This suggests that FD gains can be harvested primarily through base station upgrades, regardless of the user terminal's duplexing scheme.
Introduction
The demand for higher data rates in cellular networks has pushed researchers toward Full-Duplex (FD) communication, which allows for simultaneous transmission and reception on the same frequency. While theoretically doubling capacity, FD faces significant hurdles: self-interference (SI) at the transceiver and increased network-wide interference. While base stations can accommodate the bulky and expensive hardware required for high-performance SI cancellation, doing so for every consumer smartphone is economically and technically challenging. This paper addresses whether we can achieve FD-level performance by upgrading only the infrastructure, keeping user equipment (UE) in traditional Half-Duplex (HD) mode.
The 3-Node Topology and α-Duplex Design
The core of this research is the 3-node topology (3NT). In this setup, an FD base station manages two separate HD users: one transmitting (UL) and one receiving (DL) in the same time-frequency resource. To analyze this, the authors developed a unified framework using Stochastic Geometry, modeling base station and user locations as a Poisson Point Process (PPP).
A key innovation is the α-duplex scheme, which introduces a parameter α to represent the degree of overlap between UL and DL channels. When α=0, the system behaves as a standard HD network; when α=1, it operates in full in-band FD. This flexibility allows for fine-tuning the balance between UL and DL rates. Furthermore, the model incorporates realistic pulse shaping (rectangular and triangular) and matched filtering, accounting for the spectral leakage that occurs in real-world hardware.
Performance Analysis and Key Insights
The study rigorously compares 3NT against the 2-node topology (2NT), where the base station communicates with a single FD-capable user. The results challenge the assumption that FD hardware is necessary at both ends of the link.
The analysis yields several critical insights:
- Diminishing Returns for FD UEs: The performance gap between 3NT and 2NT is surprisingly small. Even with perfect SI cancellation in FD phones, 3NT only loses about 5% of the total rate.
- The SIC Threshold: The graph above illustrates that 3NT is highly robust. If an FD phone's SI cancellation is not near-perfect (e.g., βu = -75 dBm), the 3NT approach with HD phones actually provides a higher downlink rate (ℛd) because HD phones are inherently immune to self-interference.
- Infrastructure-Centric Gains: The findings suggest that the duplexing capability of the base station is the dominant factor in harvesting network-wide rate gains.
Conclusion
This research provides a strong argument for prioritizing Full-Duplex upgrades at the base station level. By utilizing the 3-node topology and optimizing channel overlap via the α-duplex scheme, network operators can deliver near-FD speeds to existing half-duplex user terminals. This path avoids the cost and complexity of integrating FD hardware into mobile devices while still realizing the capacity benefits of next-generation wireless communication.
Reference: AlAmmouri, A., ElSawy, H., & Alouini, M. S. (2016). Harvesting Full-Duplex Rate Gains in Cellular Networks with Half-Duplex User Terminals. arXiv:1511.00207v2 [cs.IT].
Summary updated