Abstract
5G networks aim to support diverse services through Radio Access Network (RAN) slicing, but mixing different signal numerologies (subcarrier spacings and sampling rates) leads to Inter-Service/Slice-Band Interference (ISBI). This paper proposes a mathematical framework centered on a "Generalized Circular Convolution Property" (G-CCP) to restore orthogonality in mixed-numerology systems. By combining G-CCP with Filtered-OFDM (F-OFDM) and collaborative interference cancellation, the authors enable low-complexity receivers to operate efficiently in heterogeneous 5G environments.
Introduction: The Mixed-Numerology Challenge
To accommodate high-speed video (eMBB), massive IoT (mMTC), and ultra-reliable robotics (URLLC), 5G carves the spectrum into slices tailored to specific needs. However, these slices use different numerologies—meaning they have different subcarrier spacings, symbol durations, and sampling rates.
As illustrated in Figure 1, when these different "rhythms" coexist in the same frequency band, they lose their orthogonality. This results in ISBI, where signals bleed into adjacent lanes. Furthermore, standard low-complexity receivers break down because the mathematical assumptions they rely on, such as circular convolution, no longer apply when sampling rates between the transmitter and receiver are misaligned.
Methodology
The authors address this by systematically analyzing scenarios of "imparity" between slices. The primary technical contribution is the Generalized Circular Convolution Property (G-CCP). The authors prove that by mathematically accounting for the phase shifts in the Discrete Fourier Transform (DFT) matrix caused by sampling mismatches, the receiver can still utilize simple "one-tap" equalization even when the transmitter's sampling rate is a multiple of the receiver's.
As shown in the system architecture (Figure 3), the framework integrates several key components:
- F-OFDM Waveform: Applies subband filtering to reduce Out-of-Band Emissions.
- Collaborative Detection: An algorithm that treats interference as a known quantity to be estimated and subtracted rather than random noise.
- Power Pre-compensation: A precoding scheme to ensure balanced Signal-to-Noise Ratio (SNR) across all subcarriers, overcoming frequency selectivity introduced by filtering.
Key Results and Insights
The framework was evaluated using Monte Carlo simulations based on the LTE Extended Pedestrian A (EPA) channel model. Key findings include:
- Guardband Efficiency: The proposed F-OFDM and cancellation algorithms allow for a guardband of only 4 to 5 subcarriers, significantly improving spectrum efficiency compared to traditional methods.
- SINR and BER Performance: In complex scenarios where both baseband and RF settings differ, the algorithms successfully flattened the Signal-to-Interference-plus-Noise Ratio (SINR) across subcarriers. Bit Error Rate (BER) simulations showed that mixed-numerology systems could perform nearly as well as "clean," single-numerology systems.
- Hardware-Awareness: The design specifically accounts for low-end IoT sensors with limited RF bandwidth, ensuring they can communicate effectively alongside high-speed base stations.
Conclusion and Future Work
This research provides a robust mathematical foundation for enabling 5G RAN slicing without sacrificing hardware simplicity. While the current model assumes perfect synchronization, future work will explore the impact of timing and frequency offsets and the scaling of this framework to support a massive number of heterogeneous slices in congested cells.
Summary updated