Variation-aware array-level digital predistortion for phased arrays: Methods and OTA analysis

Energy efficiency improvements and reducing the carbon footprint are among the main goals of global sustainability efforts, they also heavily affect the field of wireless communications. In radio frequency transmitters, power amplifiers (PAs) are responsible for a significant amount of overall power consumption, and thus enabling their efficient operation while maintaining good signal quality is crucial.

Digital predistortion (DPD) is an effective way to enable PA operation in its high-efficiency region while mitigating its nonlinear distortions. However, modern transmitter architectures relying on analog and hybrid beamforming cause specific challenges for DPD design, as many amplifiers must be linearized using a common digital input. In such scenarios, varying PA behavior over the array elements has a major impact on their common linearization approaches. In this thesis, the array DPD is approached in multiple ways, focusing on PA measurement methods in arrays, variation-aware array DPD solutions, and performance evaluation of array DPD solutions in the array's radiated near field and far field. For observation receiver strategies, the thesis presents a new method to de-embed individual PA responses with a specific shared observation receiver approach. The variation-aware approaches focus on the linearization strategies enabling robust DPD operation when the nonlinear responses of individual amplifiers vary for multiple reasons, including variations of the beamforming. The presented method trains the array DPD for a certain distribution of amplitude variations of the PA input, which results in solid DPD performance under random amplitude variations over the array. For example, the presented solutions can maintain the DPD performance over varying steering angles. The method is also demonstrated in practice with over-the-air measurements using a 28 GHz phased array transmitter and 5G waveforms.

The final part of the thesis focuses on observing the behavior of array-level nonlinear distortion in the radiated near field, together with the DPD linearization. It is shown that the radiated distortion in the near field is both distance and direction-dependent. This is due to the distance-dependent channel delay differences in the near-field, in conjunction with the varying PA behavior of the individual elements. In addition, the differences in distortion and delays induce additional nonlinear memory to the overall array response. The variation of the PAs over the array elements is also shown to dictate the spatial behavior of nonlinear distortion in the radiated near field. The variation, combined with DPD, is also shown to lead to spatially focused DPD performance in the radiated near field, while the far-field trained DPD is shown to have a different performance in the near field than in the far field.