TL;DR
This paper introduces "SoftNull," a novel digital beamforming technique that enables full-duplex communication in massive MIMO wireless systems. By strategically reducing self-interference through digital signal processing rather than relying solely on analog cancellation, SoftNull allows for significant data rate gains over traditional half-duplex systems, especially in dense urban environments.
What Problem Does This Paper Address?
The dream of full-duplex wireless communication, where a device can transmit and receive simultaneously on the same frequency band, promises to double spectral efficiency. However, a major hurdle is self-interference: the powerful transmitted signal overwhelms the receiver. While analog cancellation has been the go-to solution, it becomes complex and costly with the increasing number of antennas in modern massive MIMO systems. This paper tackles the challenge of enabling full-duplex in these large-antenna systems using only digital signal processing, aiming to reduce the reliance on expensive analog components and explore the benefits of massive MIMO's spatial multiplexing capabilities for self-interference management.
Methods
The core innovation is "SoftNull," a two-stage approach. The first stage is standard MU-MIMO precoding. The second stage, the self-interference reduction stage, employs a "SoftNull precoder." Unlike traditional methods that aim for perfect self-interference nulling, SoftNull's precoder strategically reduces self-interference to a level that prevents swamping the receiver's dynamic range. The residual interference is then handled by digital cancellation. The SoftNull precoder is designed to minimize total self-interference power while preserving a specified number of "effective antennas" for downlink data transmission. This is achieved by projecting the transmit signal onto the singular vectors of the self-interference channel corresponding to the smallest singular values. The paper also investigates optimal antenna partitioning strategies (e.g., contiguous blocks of transmit/receive antennas) to maximize self-interference reduction.
Datasets
The paper utilizes channel measurements collected using a 72-element antenna array platform (Argos V2) in diverse environments: an anechoic chamber, an outdoor open field, and a highly scattered indoor environment. These measurements capture real self-interference channels and array-to-client channels. Over 12 million wideband channel measurements were collected, forming a dataset of more than 40 GB. These traces are publicly available for other researchers.
Key Insights and Ideas
- Digital is Sufficient: The paper demonstrates that digital beamforming alone can achieve significant self-interference reduction in massive MIMO systems, reducing the need for complex analog cancellation.
- Strategic Nulling: SoftNull doesn't aim for perfect nulling but rather "soft nulls" that are sufficient to protect the receiver's dynamic range. This trade-off allows for better preservation of effective antennas for data transmission.
- Antenna Partitioning Matters: The spatial arrangement of transmit and receive antennas significantly impacts self-interference reduction. Contiguous partitions (like East-West) are shown to be more effective than interleaved or random ones, likely due to reduced angular spread of the self-interference channel.
- Environment's Role: Scattering environments (indoors) tend to decorrelate the self-interference channel, making it harder to suppress compared to less scattered environments (outdoors). However, lower path loss in indoor environments can sometimes compensate for this.
- The Trade-off is Key: There's a clear trade-off between self-interference reduction and the number of effective antennas available for downlink data. The optimal number of effective antennas (DTx) depends on path loss, environment, and the desired rate.
Key Results
- Significant SI Reduction: SoftNull can achieve over 50 dB of self-interference reduction while preserving a substantial number of effective antennas (e.g., 16 effective antennas preserved for 50 dB reduction in an outdoor environment).
- Data Rate Gains: In small cells with a large number of antennas relative to users, SoftNull can provide significant rate gains over half-duplex. For instance, a 40% improvement was observed for 4 users at 85 dB path loss.
- Performance Varies: Performance is sensitive to path loss and environment. Higher path loss requires more self-interference suppression, leading to fewer effective antennas for data and potentially lower gains. Indoor environments, with more scattering, require more effective antennas for the same level of SI reduction compared to outdoor environments.
- Optimal DTx: The optimal number of effective antennas (DTx) for maximizing sum rate varies. For example, DTx=18 yielded peak performance in an outdoor experiment, achieving 23% better rates than half-duplex but still 15% less than ideal full-duplex.
Limitations and Future Work
- Limited Clients in Measurements: The channel measurements were limited to 4 clients, and simulations were used for scenarios with more clients. Future work could involve more extensive multi-client measurements.
- Scheduler Adaptation: The paper assumes a higher layer makes decisions on antenna partitioning. Investigating adaptive scheduling for optimal partitioning is left for future work.
- Joint Precoder Design: The current approach uses a two-stage design. Joint precoder design for MU-MIMO downlink and self-interference reduction could potentially yield better performance.
- Hardware Prototype: The paper mentions an extension of this work involving a software-defined radio prototype of SoftNull.
Summary updated