A Two-Tier Low-Complexity Linearization Architecture for Fully Digital mMIMO Transmitters

This paper presents a two-tier, low-complexity linearization framework for fully digital multiple-input multiple-output (MIMO) transmitters that jointly minimizes total radiated power–adjacent channel power ratio (TRP-ACPR) and user error vector magnitude (EVM). The architecture comprises: (i) a per-antenna single-input single-output (SISO) digital predistorter (DPD) trained using feedback processed by an in-band notch filter to isolate out-of-band distortion, and (ii) a low-order linear MIMO pre-equalizer inserted before the precoder to mitigate in-band distortion at the receivers. Joint training of both stages enables simultaneous TRP-ACPR and EVM reduction without increasing precoder complexity or requiring multi-input or high-order nonlinear DPD models. Experimental results using a 16-channel fully digital transmitter at 3.5 GHz in a two-user line-of-sight scenario demonstrate that the pre-equalizer reduces EVM to 0.7%, while a 13-tap SISO-DPD improves TRP-ACPR from −33 dB to −43 dB and two-user ACPR from −36.5 dB to −49.6 dB and from −37 dB to −51.2 dB, respectively.