Data-Driven Model-Based Design Approach for Class EFJ Power Amplifier with Analog Predistortion Circuit

This study shows a data-driven model-based design method for a 2.4 GHz Class EFJ power amplifier (PA) with an analog pre-distortion (APD) circuit. The primary focus is on enhancing the amplifier’s linearity without compromising gain. The APD circuit is incorporated to improve linearity characteristics, resulting in notable enhancements in linearity, a high adjacent channel power ratio (ACPR), and a power added efficiency (PAE) of 36%, albeit with a decrease in power gain. The measured output power of the amplifier is 35 dB. A Conventional (feedforward) Artificial Neural Network (CANN) model is employed to analyze linearity patterns further and make future predictions. The C-ANN model is trained using a comprehensive dataset containing input features and corresponding linearity measurements. Utilizing the trained C-ANN model makes it possible to anticipate the linearity of forthcoming PAs based on their input characteristics.