Digital self-interference cancellation for in-band full-duplex communication

By allowing simultaneous transmit and receive operations in the same frequency band, In-Band Full-Duplex (IBFD) communication has been proven as a new technology with the potential of a two fold increase in spectral efficiency, as compared to the conventional half-duplex systems. However, as the transmitter creates a high level of self-interference (SI) at the receiver on the same radio, considerable amount of SI cancellation is required for achieving this gain and successfully decoding the signal-of-interest (SoI) arriving from the distant node. For this purpose, in IBFD radios, various SI cancellation techniques are proposed to suppress the SI signal by employing propagation domain antenna suppression, analog domain cancellation, and non-linear and linear digital SI cancellation in the literature. In this thesis, we study linear digital SI cancellation techniques and non-linearity problem induced by the hardware components on the IBFD radios, considering Orthogonal Frequency Division Multiplexing (OFDM) based wireless systems. OFDM is the pertinent waveform for current and next generation wireless systems, whose spectral efficiency can be potentially doubled by IBFD communication. However, in OFDM based IBFD systems, linear digital self-interference cancellation (DSIC) employed at baseband does not provide sufficient cancellation in the cyclic prefix (CP) region. When the propagation delay between two communicating radios is non-zero, the CP noise affects the data region of the desired signal. In the first part of this thesis, we propose CP noise reduction (CPNR) technique for OFDM based IBFD radios. In the CPNR solution, we enhance SI channel estimation and SI signal reconstruction for time and frequency DSIC techniques. We have evaluated CPNR with time and frequency domain DSIC via both simulations as well as MATLAB and FPGA implementations on our Software Defined Radio (SDR) based IBFD radio. In the laboratory tests, the total suppression of the IBFD radio is improved by 6 dB by employing CPNR in frequency-domain DSIC and EVM for bidirectional communication is improved by up to 5%, allowing realistic propagation delays. In addition to improving the total suppression and EVM performance, CPNR is also shown to enhance the multi-path resiliency of DSIC techniques. In the second part of the thesis, we consider the non-linearity problem in IBFD radios at high transmit power levels, and we propose a new architecture along with time and frequency domain non-linear estimation algorithms. In this architecture, non-linear and linear SI cancellation stages are isolated via an RF switch, so that non-linear estimation can be performed separately from and prior to linear SI cancellation. Following one of the alternative proposed estimation algorithms, the non-linear SI signal is obtained as a reference to linear SI cancellation. Our experimental results obtained on OFDM based IBFD SDR set-up demonstrate that the amount of total SI suppression is improved by up to 13 dB over state-of-the-art digital, integrated linear and non-linear SI cancellation schemes. Moreover, in our solution, since non-linear estimation is decoupled from SI channel estimation, the SI cancellation performance is immune to changes in the (multi-path) environment; unlike existing schemes, which require re-optimization of model parameters for each setting. Last but not least, estimation overhead for digital SI cancellation is eliminated, and computational complexity is lowered by four to six orders-of-magnitude with the proposed algorithms