Methods of self-interference cancellation in full duplex telecommunication systems

Abstract: With the wireless technology evolving quickly, so does the demand of speed and efficiency. This makes the companies look for new and better ways to improve the current systems. One way of improving the present systems would be to employ Full Duplex Technology. In recent years the standard has been Half duplex technology with either Time Division Duplexing (TDD) or Frequency Division Duplexing (FDD). The drawback of TDD is that the signals is put in different time slots, meaning if many signals is to be transmitted at the same time there will be a delay. For FDD the signals are sent at different frequencies. This takes up a lot of space in the spectral domain. Full Duplex Technology has the potential to double the spectral efficiency with it'spower to transmit and receive signals simultaneously at the same frequency. The main challenge with Full Duplex (FD) is the leaking Self-Interference (SI) from the transmitter to the receiver. Different methods can be used to suppress the SI in both the digital and the analog domain. Typically the Self-Interference Cancellation (SIC) is split into three parts. The passive Radio Frequency (RF) SIC, which suppresses the signal using for example. antenna separation, antenna polarization or a circulator. The active analog RF cancellation which could for example use a multi-tap analog least mean square adaptive and finally the Digital passband SIC, that is addressed in this thesis. The cancellation in the RF domain needs to suppress enough for the Low Noise Amplifier (LNA) and the Analog to Digital Converter (ADC) to not saturate. The Digital SIC should optimally suppress the signal to the noise floor to be able to demodulate the received signal. In this thesis modelling and reconstruction of the SI signal has been done. An attempt to model the non-linearities from the Power Amplifier (PA), the imbalance from the IQ-mixer and the effects of the circulator has been done to as correctly as possible comply to the real signal distortions. Simulations using experimental data provided by Syntronic SRD was used to evaluate the cancellation for Recursive Least Square (RLS), Ordinary Least Square (OLS) and Normalized Least Mean Square (NLMS) algorithms. The simulations shows that a cancellation of over 45dB within the bandwidth can be achieved using digital cancellation in the baseband. The result shows the importance of having a weakly nonlinear transmit signal to achieve a better performance using FD. It was showed that the linear model shows significantly worse results incomparison to the Generalized Memory Polynomial (GMP) and Memory Polynomial (MP). The MP and GMP model achieve similar results in the weakly non-linear cases but the GMP outperforms the MP model in strongly non-linear cases.