Statistical Digital Predistortion of 5G Millimeter-Wave RF Beamforming Transmitter Under Random Amplitude Variations

摘要

The nonlinearity of the fifth-generation (5G) millimeter-wave (mmWave) phased array transmitter (TX) depends on the variations in the beamforming coefficients, e.g., due to unwanted amplitude errors of the phase shifters or purposely introduced amplitude variations to shape the beam, as well as other analog component variations. Therefore, also the digital predistortion (DPD) coefficients depend on the beamforming coefficients and require a continuous update for different beamsteering directions in a conventional DPD. We propose a robust DPD approach that uses the average array response for DPD training. The average array response is estimated using the known experimental histogram of power amplifier (PA) input power variation resulting from the amplitude variations due to beamforming. The DPD training based on the average array response makes the DPD coefficients insensitive to beamforming variations. The proposed DPD strategy requires a shared feedback path for training and only a single set of DPD coefficients to linearize the array response in all beamsteering directions. The performance of the proposed DPD strategy is validated by over-the-air (OTA) measurements of a 28-GHz phased array TX over different steering angles. Experimental results show that the proposed DPD method using one set of DPD coefficients provides similar linearization performance across the steering angles compared with the DPD trained to all steering angles through the OTA reference antenna.