Hammerstein Self-Interference Cancellation for FDD in Adjacent Bands

This work considers wireless full-duplex operation using frequency-division duplex (FDD) on adjacent bands, where inevitable out-of-band emissions caused by nonlinear transmitters lead to self-interference (SI) on the reception band. In order to cancel this type of SI, the nonlinear characteristic must be estimated as well as the SI channel response between the receiver and the interfering transmitter, which together form a Hammerstein model. Conventional channel estimation methods cannot be applied here because the channel response of out-of-band emissions affecting the adjacent band is required. We describe three approaches to model and cancel the adjacent-band SI caused by nonlinearity that has propagated through a SI channel. A memory polynomial-based reference method is compared to two novel Hammerstein model-based approaches that perform SI channel estimation in frequency domain. Evaluations via simulation and hardware measurement of a WiFi device demonstrate the superiority of the latter methods with respect to cancellation performance.