Research On Self-interference Suppression Of CO-time CO-frequency Full Duplex Wireless Communication Systems

The radio spectrum has become one of the scarcest resources with the explosive growth of wireless mobile devices, and new generation wireless communication technology has been put on the agenda. The co-time co-frequency full duplex(CCFD) transmission technology, which provides transmitting and receiving signal on a common frequency at the same time, has the potential to double the air-interface throughput compared with traditional time-division duplexing and frequency-division duplexing.The key point of CCFD is how to remove the strong self-transmitted signal(self-interference) from the received signal. Three concatenated cancellation technology are used, i.e., antenna cancellation, analog cancellation and digital cancellation. The principles of analog and digital cancellation are as follows: the self-interference is firstly estimated and rebuilt, and then substracted from the received signal. The rebuilt errors and amplifier nonlinearity have become the bottleneck of self-interference cancellation in full duplex transmission, and in this thesis we focus on the researching of the self-interference cancellation ability.In our thesis, we studied the impact of self-interference on the capacity of full duplex wireless communications, from which we have learned the importance of self-interference cancellation. We also discussed the impact of self-interference cancellation errors, i.e., amplitude error, phase error, delay error on bit error ratio(BER) and interference cancellation ratio(ICR), and we finally discussed the influence of amplifier nonlinearity. The details of our investigation are listed as follows:Firstly, we studied the weighted sum-rate(WSR) of full duplex with separated antennas in static and fading channels. The decision criterion of full duplex or half duplex was given, and the closed form of WSR upper and lower bound were derived.The full duplex outperforms the traditional half duplex in additive white Gaussian noise(AWGN) channel when ICR is greater than the threshold on conditions of the same transmission power, noise level and attenuation. For example, at least 112.5d B self-interference must be cancelled to choose CCFD, when the transmission power is 30 d Bm, the power attenuation by distance is 100 d B and the noise power is ?95d Bm.The full duplex transmission with antenna selection(AS) mechanism achieved better performance compared with fixed antenna transmission in fading channels on conditions of the same transmission power, noise level and attenuation. Particularly, full duplex with AS outperforms that without AS about 2.4bps/Hz, when the transmission power is 30 d Bm, the background noise power is ?95d Bm, the power attenuation by distance is 100 d B, and the ICR>140d B.Secondly, we analyzed the impact of self-interference cancellation errors, i.e., amplitude error, phase error, delay error on BER and ICR in AWGN channels, and derived the closed form BER and ICR expressions. We also discussed the influence of different shaping pulses, i.e, root raised cosine(RRC) pulse and rectangular pulse.For binary phase shift keying(BPSK) signal, 55 dB ICR can be achieved when amplitude relative error | h|<0.2% and phase error | Dy|<0.1?.When the signal to interference ratio(SIR) is ?45d B, the standard deviation of amplitude error h is 0.001 and the standard deviation of the phase error Dy is 0.1?, the full duplex transmission got 5d B(resp. 1.6d B) performance loss at BER 10-4 for BPSK signal with rectangular shaping(resp. RRC shaping).Thirdly, we analyzed impact of self-cancellation errors, i.e., amplitude error, phase error and delay error on BER and ICR in flat and frequency selective fading self-interference channels.The scatter paths of self-interference became the bottleneck of cancellation when the line of sight(LOS) path was cancelled in frequency selective fading channel.Particularly, when Ricean factor K=30d B, the scattered paths should be considered if more than 30 d B ICR expected. When SIR is ?30d B, all paths have same errors(the standard deviation of amplitude relative error is 0.001 and that of phase error is 0.1?), the full duplex transmission got 4.3d B performance degradation at BER 10-2.Finally, we studied the influence of typical amplifier nonlinearity on ICR and BER. During self-interference cancellation stage, two type of reference signal source were considered, i.e., type A, from the input of the nonlinear amplifier and type B, from the output of the nonlinear amplifier. The analytical results have shown that the performance of type B outperforms that of type A at same cancellation errors level. For a given BER level, e.g., 10-4, there is an optimal point of power back-off for nonlinear amplifier to achieve minimum total performance degradation. In particular, for a 64-QAM full duplex transmission by using a given nonlinear travelling wave tube amplifier(TWTA), the optimal output back-off(OBO) is about 24.5d B(for type A) and 8d B(for type B) at BER 10-4 when SIR=-33 d B.The total work in our thesis can be used to link budget, hardware architecture design, which hopefully could provide a guideline to devising efficient self-interference cancellation algorithms that are robust to estimation errors in future works.