Data-Clustering-Assisted Digital Predistortion for 5G Millimeter-Wave Beamforming Transmitters With Multiple Dynamic Configurations

Motivated by data science, in this article, a data-clustering-assisted digital predistortion (DPD) is proposed to linearize millimeter-wave (mmWave) beamforming transmitters with multiple dynamic configurations. Based on the data analysis, similar transmitter states with different configurations can be clustered, resulting in a significant reduction of linearization states. Model complexity within the cluster can be further reduced by utilizing a penalty factor method. To validate the proposed concept, experiments were carried out on a 16-channel mmWave beamforming transmitter with configurable beam angle, operating frequency, and input power. Total 216 transmitter states can be reduced to 8, 20, and 40 states with unequal optimal model parameters for each state, without losing much performance. The proposed method can be extended to the scenario where a large scale of dynamic states can occur in complex transmitter structures in future wireless systems.