In-Band Full-Duplex RF Canceller Tuning using Adaptive Learning Rate Functions

Enhancements in communication and spectral efficiencies can be derived from incorporating In-Band Full-Duplex (IBFD) technology into future wireless systems. These improvements are made possible by reducing each node's self-interference through a variety of techniques, including RF cancellation in the analog domain. While multi-tap RF cancellers have been proven to deliver high performance in realistic environments, their multivariate nature creates a problem when trying to tune them. This paper investigates different adaptive learning rate functions as well as starting locations for a multi-dimensional gradient-descent algorithm. Results are presented for this research using a two-tap canceller prototype that was measured over 20 MHz centered at 2.5 GHz and provided up to 48 dB of cancellation in a multipath interference scenario.