Experiments on Self-interference Cancellation in Full-Duplex Wireless Communication Systems

Abstract

Full-duplex is one of the promising technologies in the next generation of mobile access networks (5G) and its main drawback is self-interference resulting from power leakage between transmit and receive antennas. In this thesis we conduct the implementation of self-interference cancellation in full duplex wireless communica- tion system. Firstly, we present literature review on frequency division multiplexing (OFDM) technology, including basic OFDM principle, peak-to-average power ratio (PAPR) and sensitivity to frequency offset problem. The synchronization part of OFDM system is also emphasized, especially on the method of compensating frequency offset. We also present the history of full-duplex and its opportunities and challenges. In addition, methods in different domain in order to cancel the self-interference (SI) are studied in this thesis. Secondly, we use software defined radio (SDR), which is built with open-source software development toolkit and hardware platform USRP N210, to implement the full-duplex system. We introduce in details how to build a basic module with SDR and create the self-interference cancellation module based on a digital domain cancellation method. This adaptive gradient based algorithm uses cross-correlation between received and transmitted signals to estimate the channel characteristics in order to cancel SI. We encounter a delay problem when samples going through different buffers and the range varies from few to thousand samples. To avoid using more than thousand taps filter in cancellation, orthogonal preamble symbols are assigned to different transmitters and detection of the coming interference signal is implemented to reduce the number of filter taps. Thirdly, we measure the SI cancellation performance and bit error rate after cancellation. We report the SI cancellation performance with different amount of filter tap as well as gradient step size. The best SI cancellation rate observed in the experiment is around -31dB when filter tap equals to six.