Poster: SoftNull: All-digital Massive MIMO Full-duplex Wireless

Today’s wireless base stations are half-duplex, meaning that transmission and reception are relegated to separate time slots or separate frequency bands. Data rates would be multiplicatively increased if base stations were full-duplex, meaning they could both transmit and receive at the same time and in the same frequency band. The challenge to fullduplex operation is self-interference: the base station generates high-powered interference to its own receiver, swamping the receiver electronics and preventing the base station from receiving the much weaker uplink signal. Research over the last ten years [1, 2, 3, 4], has shown that full-duplex operation is feasible for small cells. The key enabler of full-duplex has been a combination analog cancellation and digital cancellation of the self-interference [3, 5]. Another promising wireless innovation is massive multipleinput multiple output (“Massive MIMO”), in which the base station uses very large antenna arrays (on the order of hundreds) to communicate with many users simultaneously. The benefit of Massive MIMO is that the beam to each user is very focused, which enables the base station to leverage simple signal processing and mitigates interference between cells [6]. The grand vision for next-generation wireless communication is to combine Massive MIMO and full-duplex in a single system. Full-duplex Massive MIMO brings both new challenges and new opportunities. Challenge: Analog cancellation has been considered necessary to prevent the self-interference from overwhelming the dynamic range of the receiver electronics. However, as the number of antennas grows, the complexity of the hardware required for analog cancellation grows superlinearly. Therefore solutions are needed to suppress self-interference prior to the receiver front end whose analog complexity does not scale with the number of antennas. Opportunity: Massive MIMO also presents a new opportunity for full-duplex. Many antennas provides transmit spatial degrees of freedom that can be leveraged for transmit beamforming to suppress self-interference. However, supPermission to make digital or hard copies of part or all of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. Copyrights for third-party components of this work must be honored. For all other uses, contact the Owner/Author(s). Copyright is held by the owner/author(s). S3’15, September 11, 2015, Paris, France. ACM ISBN 978-1-4503-3701-4/15/09. DOI: http://dx.doi.org/10.1145/2801694.2801716. pressing self-interference via transmit beamforming requires sacrificing transmit dimensions that could have been leveraged for reaffirming to the downlink users. In particular, for receive antenna that is nulled, a transmit dimension (i.e. a virtual transmit antenna) must be sacrificed. Proposed Solution: We consider removing the analog cancellation stage altogether, and relying on SoftNull, an all-digital-architecture for self-interference suppression, which uses transmit beamforming and digital cancellation to suppress self-interference. SoftNull can be implemented on existing base stations with existing radios; no special-purpose analog components would be needed. SoftNull leverages the observation that that the self-interference need not be zero-forced, but only suppressed to a level commensurate to the desired uplink signal, so that the selfinterference no longer overloads the receiver and can then be cancelled digitally. Given a required number of “virtual antennas” for the downlink, SoftNull choses the transmit beamweights which best suppress self-interference, which turn out to have a closed-form expression. Initial experiments on a 72-element array have shown that SoftNull can sufficiently suppress self-interference while maintaining strong links to the users, for moderate numbers of users (4-12), and moderate path loss (< 90 dB). Much more research needs to be performed in terms of both algorithm development and experimental evaluations.