Advanced Polynomial Digital Pre-Distortion Using an Iterative Optimization Architecture

Abstract

The exponential growth of wireless communication systems and the increasing demand for high data rates have intensified the need for power amplifiers (PAs) with both high efficiency and high linearity. However, inherent nonlinearities in RF PAs introduce amplitude and phase distortions, leading to spectral regrowth, intermodulation products, and degradation of signal fidelity. Polynomial digital pre-distortion (DPD) techniques have been widely employed to mitigate these nonlinear effects, yet conventional static polynomial approaches exhibit limited adaptability to dynamic signal conditions, amplifier aging, and environmental variations. This paper proposes an advanced iterative optimization-based polynomial DPD framework, designed to dynamically adjust pre-distortion coefficients in real-time to accurately linearize PA behavior across a range of operating conditions. The framework integrates memory-polynomial modeling to address temporal distortion effects, while a normalized least-mean-square (LMS) iterative algorithm ensures rapid convergence and stability of coefficient adaptation.
Simulation studies and hypothetical implementation scenarios were conducted on a 2-W 2.4 GHz RF amplifier transmitting high-order modulation schemes such as 64-QAM over wideband OFDM signals. Key performance metrics evaluated include adjacent channel power ratio (ACPR), error vector magnitude (EVM), third-order intermodulation distortion (IMD3), and 1-dB gain compression point. Results indicate a reduction in ACPR from -30 dBc to -48 dBc, an improvement in EVM from 5.2% to 2.8%, and a significant suppression of IMD3 products by 12 dB relative to static polynomial DPD implementations. Convergence analysis demonstrates stable coefficient adaptation within 120–180 iterations, highlighting robustness across varying input powers and polynomial orders. Comparative evaluation confirms that iterative optimization outperforms conventional approaches in mitigating both amplitude and phase nonlinearities, while memory compensation effectively addresses temporal distortion effects (Saleh, 1981; Kennington, 2000).
The proposed methodology offers practical applicability for modern 4G/5G wireless systems, enabling high-efficiency amplifier operation without sacrificing linearity. By reducing spectral leakage and improving signal fidelity, this iterative polynomial DPD framework supports enhanced communication quality, extended amplifier operational lifetime, and compliance with regulatory spectral masks. Furthermore, the approach provides a foundation for future enhancements incorporating adaptive or machine learning-based coefficient tuning to further optimize linearization under rapidly varying operating  
conditions. Overall, this study establishes an advanced, scalable, and analytically rigorous solution for dynamic PA linearization, bridging the gap between theoretical modeling and practical high-performance RF system requirements.