Digital self-interference cancellation in full-duplex wireless systems

Official URL: http://risc01.sabanciuniv.edu/record=b1640327 (Table of Contents)

Present half-duplex (HD) wireless technologies are currently striving to meet the growing demand of high speed wireless connectivity. Recent works have demonstrated the feasibility and potential of full-duplex (FD) wireless systems to double the spectral efficiency of HD systems, which makes FD communication an attractive solution to address the present wireless spectral congestion. Self-interference (SI) cancellation is the key to FD communication and the residual SI is the major factor determining the performance of an FD radio. At the receiver of an FD system, SI suppression is achieved in two stages, first in analog domain at RF level, and then in digital domain at baseband level. Digital SI cancellation, being the last stage, plays a crucial role, as it primarily quantifies the signal-to-noise ratio (SNR) of the desired signal. In this thesis, we present a novel frequency domain approach for the reconstruction of SI signal in digital domain. For the realization and performance evaluation of the proposed and the existing time domain reconstruction approaches with different SI channel estimation algorithms, we have considered the baseband model of FD implemented on an OFDM system under time dispersive fading channel. We have evaluated the performance of digital SI cancellation techniques for such an FD system via detailed simulations and extensive tests with WARP Software Defined Radio (SDR), also analyzing computational complexity. Through the simulation and test results, it is shown that, for the AWGN channel, the amount of digital cancellation increases with increasing SNR of the received SI signal, and a maximum cancellation of 36 dB is achieved. Under fading, the SI suppression capability of all digital techniques degrades, especially with increasing delay spread. However, since the frequency domain estimation is resilient to large delay spreads, better performance is observed as compared to the time domain estimation based techniques, which are more prone to frequency selectivity. Additionally, it is demonstrated that with least square frequency domain estimate, the cancellation obtained by the proposed frequency domain reconstruction, outperforms the existing time domain approach by 5 - 10 dB, while the computational complexity is reduced to one-fourth of that required by the time domain reconstruction. Furthermore, it is observed that the SI suppression capability of the digital cancellation techniques can be improved up to 1 dB, by increasing the number of training sequence symbols, which can be achieved by slight modifications in the preamble structure. Lastly, FD operation is demonstrated on the WARP SDR set up, by applying the frequency and time domain reconstruction approaches, showing simultaneous transmission and reception of a tone.