Graduate and Postdoctoral Studies
Electrical and Computer Engineering
Wireless Full-Duplex: From Practice to Theory
Tuesday, January 14, 2014
to 4:30 PM
1049 Duncan Hall
Full-duplex is the ability of a node to transmit and receive simultaneously in the same band. Ideal wireless full-duplex communication can double the spectral efficiency compared to the traditional half-duplex communication. In this dissertation, we study the challenges in realizing full-duplex communication. We tackle the challenges from two different perspectives: node and network.
Node perspective: Simultaneous transmission and reception results in a large self-interference due to the proximity of transmit and receive antennas at the full-duplex node. To establish the feasibility of wireless full-duplex, we develop a wideband real-time physical layer and evaluate its performance on the WARP testbed. Self-interference reduction in our proposed physical layer is achieved through passive suppression and active cancellation. Based on the constraints of the physical layer, we propose a MAC layer protocol which is designed specifically to discover and enhance opportunities to communicate in the full-duplex mode.
%Further, based on the physical layer, we propose a MAC layer protocol which is designed to discover full-duplex opportunities in a network.
Experimental evaluation of our physical layer design, as well as several other full-duplex designs proposed in literature, reveal that active cancellation does not push self-interference all the way upto the thermal noise floor. In this dissertation, we explore the bottlenecks limiting active cancellation in full-duplex systems. We show that the amount of active cancellation is limited by transmitter side noise, particularly by the phase-noise in the local oscillator at the transmitter of the full-duplex node. Thus, unlike conventional half-duplex systems where receiver thermal noise is a limiting factor, full-duplex systems are limited by transmitter side noise. As a key by-product of our analysis, we propose a signal model for a wideband MIMO full-duplex system. We use our proposed signal model to study the performance limits of a system where the start of transmission and the start of reception at a full-duplex node are not synchronized. Interestingly, we discover that the bit-error-rate of the communication mode where the start of transmission precedes the start of reception is better than the mode where start of transmission follows reception of a packet at the full-duplex node.
Network perspective: In order to extract gains in capacity from full-duplex operation in a multi-user network, we propose to use full-duplex capable nodes to simultaneously operate uplink and downlink in a network. Such operation results in a new type of interference in the network -- internode interference, i.e., the uplink transmission from each mobile user starts interfering with the downlink receptions at all the other mobile users. We show a physical layer coding strategy that aligns interference over time and extracts gains in degrees-of-freedom of the network.
Finally, we recognize that larger gains from full-duplex operation are possible by leveraging the fact that the strength of the internode interference channel is often different from the uplink/downlink channel. By analyzing the uplink/downlink capacity of a network composed of one base-station and two mobile users, we show that full-duplex not only out-performs half-duplex, but also recovers some of the degrees-of-freedom lost due to lack/delay of channel state information of the network.