LAUSR

This paper addresses self-interference (SI) cancellation for millimeter-wave (mmWave) based full-duplex backhaul systems, which involves not only the near-end cross talk due to antenna leakage, but also the far-end cross talk due to reflections from obstacles around the transmission path. The SI channel is with a long response, particularly for mmWave systems where the bandwidth is large. In order to manage the SI comprehensively, we put forward an adaptive multi-segment frequency-domain cancellation (MSFDC) scheme based on the frequency-domain block least-mean-square algorithm, which can cover the long SI channel with low complexity. In addition, we introduce an output branch on the traditional adaptive training loop and perform tap selection, so that the cancellation error decreases while the convergence and tracking abilities are maintained. To demonstrate the effectiveness of the proposed tap-selection based MSFDC, comprehensive evaluations are carried out in an E-band full-duplex backhaul scenario. It is shown that the number of multiplication operations can be greatly reduced by dividing the SI channel into multiple segments. Meanwhile, in a wide range of cases, the proposed scheme can cancel the SI to the degree below the noise floor, and significantly outperforms the traditional method without tap selection.