Search papers, labs, and topics across Lattice.
This study employs a deep learning-assisted approach combining convolutional neural networks and genetic algorithms to inverse synthesize compact, wideband inverted Doherty power amplifiers. The methodology successfully integrates multiple functionalities into a single pixelated structure, resulting in a prototype GaN HEMT Doherty PA that achieves a peak drain efficiency of 51%-63% and maintains a 6-dB back-off efficiency of 48%-54% across the 1.9-2.5 GHz frequency range. Additionally, the circuit demonstrates impressive performance with digital predistortion, achieving an adjacent channel leakage ratio better than -53.2 dBc, showcasing the potential for enhanced amplifier designs.
Achieving over 51% peak efficiency in a compact, wideband Doherty power amplifier through innovative deep learning techniques could revolutionize power amplifier design.
This paper presents a deep learning-assisted methodology for the inverse synthesis of a compact, wideband inverted Doherty power amplifier (PA). Convolutional neural networks (CNNs) and genetic algorithms (GAs) are jointly employed to generate pixelated Doherty combiner networks that integrate load modulation, impedance matching, power combining, and phase compensation into a single structure. As a proof of concept, we design and fabricate a GaN HEMT Doherty PA with a pixelated output combiner. The prototype achieves a measured peak drain efficiency of 51%-63% and a 6-dB back-off efficiency of 48%-54% over 1.9-2.5 GHz. Within the same frequency range, the measured output power is 44+/-0.3 dBm. Furthermore, with digital predistortion (DPD) applied, the prototype circuit demonstrates an adjacent channel leakage ratio (ACLR) better than -53.2 dBc.
