2019_programme: ROBUST DIGITAL SELF-INTERFERENCE CANCELLATION FOR FULL-DUPLEX UWA SYSTEMS: LAKE EXPERIMENTS

• Session: 05. Underwater Communications and Networking
  Organiser(s): Tsimenidis Charalampos, Mitchell Paul, Pelekanakis konstantinos
• Lecture: ROBUST DIGITAL SELF-INTERFERENCE CANCELLATION FOR FULL-DUPLEX UWA SYSTEMS: LAKE EXPERIMENTS [invited]
  Paper ID: 891
  Author(s): Shen Lu, Henson Benjamin, Zakharov Yuriy, Mitchell Paul
  Presenter: Shen Lu
  Presentation type: oral
  Abstract: Full-duplex (FD) underwater acoustic (UWA) communication has significant potential in increasing the capacity of acoustic links, but it suffers from severe self-interference (SI) caused by the near-end transmission. Existing SI cancellation techniques include analogue cancellation, digital cancellation and antenna beamforming. Among those techniques, the digital cancellation has the lowest complexity. However, the reported digital SI cancellation performance is limited. To improve performance, the non-linearity of the power amplifier (PA) should be taken into account. Here we use the digitized PA output as the reference signal for SI estimation to reduce the effect of the non-linear distortion. Such a system architecture allows us to use a low-complexity linear adaptive filter for SI cancellation. Specifically, we use the recursive least-squares (RLS) algorithm with dichotomous coordinate descent (DCD) iterations. This results in a low-complexity SI canceller. As observed from the experimental results with traditional single-branch digital canceller, the choice of the sampling time is crucial to the SI cancellation performance. In this paper, we present a robust digital SI cancellation scheme based on the RLS-DCD algorithm with the use of the PA output. To achieve robust SI cancellation performance, the PA output is oversampled to twice the symbol rate, de-multiplexed into two branches employing digital SI cancellation, and then combined based on the residual variance estimates in the two branches. The SI cancellation performance is investigated by conducting experiments in an indoor water tank and in a shallow lake. The tank experiments show that up to 66 dB of SI can be cancelled using the proposed scheme, which is high compared to existing FD designs. Meanwhile, up to 56 dB of SI is cancelled in the lake experiments. The difference in the SI cancellation performance is due to faster varying SI channels in the lake experiments. Both the water tank and lake experimental results show robust SI cancellation performance regardless of the choice of sampling time.
    Download the full paper
• Corresponding author: Ms Shen Lu
  Affiliation: University of York
  Country: United Kingdom
  e-mail: