Widen Linearization Angle of Beamforming Arrays with Semi-Partitioned Digital Predistortion

In this paper, a semi-partitioned digital predistortion (DPD) structure is proposed to perform an affordable 2-target DPD to widen the linearization angle in beamforming phased arrays. Instead of using additional auxiliary phase-shifters to implement the second nonlinearity path for the 2-target DPD, in the proposed approach, the RF chains are grouped into two groups to form two DPD paths, without changing the original beamforming phase-shifting network. With this semi-partitioned predistortion, a multi-target iterative learning control structure is proposed to modify DPD extraction to enable online model update. This direct learning architecture is validated by simulations, and experiments with an indirect learning architecture are also carried out offline to validate the proposed DPD in a hybrid beamforming scenario.