IMPACT OF NOISES AND NONLINEARITY ON ANALOG SELF-INTERFERENCE CANCELLATION IN IN-BAND FULL-DUPLEX COMMUNICATIONS
A wireless revolution has occurred resulting in the formation of a proverbial backbone of wireless devices that our everyday functionality, productivity, and general way of life have become dependent. Consequently, victimizing an already constrained and finite wireless spectrum with further demands for increased bandwidths, greater channel capacities, and an insatiable plea for faster access rates. In-band full-duplexing (IBFD) is an innovative and encouraging technology that aims to answer this tacit mitigation call by bolstering spectral efficiency through simultaneous same frequency band transmission and reception. Conventionally, transceiver-based systems have their respective transmission and reception dictated by occurring in either disparate time slots (half-duplex) or distinct frequencies (out-of-band full-duplex). By achieving simultaneous same band communication, a theoretical doubling in spectral efficiency is rendered feasible. However, transmitter to receiver leakage, or self-interference (SI), remains the most barring frustration to IBFD realization. Being locally generated, SI is considerably stronger (often 50-100dB) than the desired signal-of-interest (SOI). Left unresolved, this unwanted energy saturates the receiver’s amplifiers and desensitizes its analog-to-digital converters. Thus, rendering the SOI unintelligible. Therefore, a means of self-interference cancellation (SIC) is necessitated to suppress any polluting SI to levels that of or below the receiver’s noise floor.In this thesis an in-depth history of in-band full duplex technology is first presented, followed by a condensed examination of the SIC domains. Pertinent theory is presented pertaining to noise analysis and estimation relevant to a proposed IBFD transceiver architecture. Finally, a modelled simulation of this transceiver, developed in MATLAB, is presented. Subsequent results detailing an investigative study done on a fully adaptive tapped-branch analog self-interference canceller are shown. Said canceller’s variable phase and amplitude weights are set via real-time training using gradient descent algorithms. Evaluation of the results reveal marginal effect on the SIC efficacy due to transmission path nonlinearity and noise distortions alone. However, expansion of model consideration for conceivable cancellation hardware nonlinearities reveals an indirectly proportional degradation of SIC performance by up to 35dB as distortion levels vary from -80 dBm to -10 dBm. These results indicate consideration of such non-idealities should be an integral part of cancellation hardware design for the preclusion of any intrinsic cancellation impediments.