Beam-Dependent Active Array Linearization by Global Feature-Based Machine Learning
Mattia Mengozzi, Gian Piero Gibiino, Alberto Maria Angelotti, Corrado Florian, Alberto Santarelli
Abstract
An approach based on machine learning is proposed for the global linearization of microwave active beamforming arrays. The method allows for the low-complexity real-time update of the digital predistortion (DPD) coefficients by exploiting order-reduced model features, hence avoiding the need for repeated local DPD identification steps across the various operating conditions of the beamformer (e.g., different beam angles or RF power levels). The validation is performed by over-the-air (OTA) measurements of a <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$1$</tex-math> </inline-formula> <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\times$</tex-math> </inline-formula> <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$4$</tex-math> </inline-formula> array operating at 28 GHz across 100-MHz modulation bandwidth (BW).