Iterative Successive Nonlinear Self-Interference Cancellation for In-Band Full-Duplex Communications

Abstract

In-band full-duplex (IBFD) communications have recently been considered for wireless backhaul in the ultra-high frequency (UHF) band terrestrial broadcast systems since they can double the spectral efficiency compared to conventional half-duplex communications. The inherent challenge of IBFD communication is self-interference (SI), the power leakage from the co-located transmitter to the receiver. For wireless backhaul among transmitter towers, it's desirable to employ high-order modulation, e.g., 1024QAM or 4096QAM, for higher spectral efficiency. Furthermore, the transmitter emission power is much higher in low radio frequency (RF) bands. Due to the nonlinearity of the transmitter high-power amplifier (HPA), the nonlinear distorted SI becomes a performance-limiting impairment and must be effectively mitigated. Moreover, the SI channel exhibits a large multipath delay spread in these applications, especially in UHF-band terrestrial broadcasting systems where the transmission frequency is lower and the antenna directivity is limited. In this paper, a novel iterative successive nonlinear SI cancellation scheme, based on the previously proposed frequency-domain RF self-interference cancellation (RF-SIC) technique, is presented. The baseline RF-SIC is capable of cancelling SI with a large multipath delay spread. However, the RF-SIC performance is limited by the presence of the remote signal of interest (SOI) in the received signal. This SOI presence, referred to as "intrinsic noise" in the SI channel estimation process, occurs due to foregoing the training phase requirement. The proposed approach in this paper iteratively cancels the "intrinsic noise" and can suppress the nonlinear SI to the receiver noise floor. The proposed technique can be applied to 3rd Generation Partnership Project (3GPP) Integrated Access and Backhaul (IAB) technology to make it more feasible for deployment at the lower frequency band.