Comparative Study of Feedback Architectures for Phased-Array Linearization under Load Modulation and Hardware Impairments

Abstract

As phased array transmitters become integral to 5G and emerging 6G systems, efficient digital predistortion (DPD) is essential to maintain linearity across large antenna arrays. However, implementing DPD at scale is challenged by the complexity of feedback acquisition, the presence of hardware impairments, as well as antenna coupling and resulting load modulation of the power amplifiers (PAs). This paper presents a comparative study of three DPD feedback architectures - anti-beamforming, shared feedback (switched per-PA), and over-the-air (OTA) - under both load-modulated and non-load-modulated conditions. Unlike prior works that focus on individual schemes or ideal conditions, this study evaluates each architecture in the presence of practical non-idealities such as PA characteristic mismatch, feedback gain imbalance, and feedback delay mismatch. Key performance metrics, including Adjacent Channel Leakage Ratio (ACLR) and Total Radiated Power ACLR (TRP-ACLR), are used to quantify linearization effectiveness. Among the considered non-idealities, it is found that delay mismatch deteriorates DPD performance the most. In terms of the feedback schemes, anti-beamforming feedback scheme gives the most robust performance under all non-idealities. These results provide critical insights into the trade-offs involved in selecting DPD feedback architectures for scalable, energy-efficient phased-array transmitters.