Search papers, labs, and topics across Lattice.
This paper introduces a deep learning-driven inverse design methodology for Doherty power amplifiers (PAs) using multi-port pixelated output combiner networks. A convolutional neural network (CNN) is trained as an electromagnetic (EM) surrogate model to predict S-parameters of pixelated passive networks, which is then integrated into a blackbox Doherty framework with a genetic algorithm (GA) for optimization. The resulting Doherty PA prototypes, incorporating three-port pixelated combiners and GaN HEMT transistors, achieve high drain efficiency (over 74% peak, 52% at 9dB back-off) and output power (over 44.1 dBm) at 2.75 GHz, while also demonstrating good linearity under 5G NR-like waveform testing after digital predistortion (DPD).
Deep learning slashes design time for high-efficiency Doherty power amplifiers, enabling complex pixelated combiners that extend the back-off efficiency range.
This article presents a deep learning-driven inverse design methodology for Doherty power amplifiers (PA) with multi-port pixelated output combiner networks. A deep convolutional neural network (CNN) is developed and trained as an electromagnetic (EM) surrogate model to accurately and rapidly predict the S-parameters of pixelated passive networks. By leveraging the CNN-based surrogate model within a blackbox Doherty framework and a genetic algorithm (GA)-based optimizer, we effectively synthesize complex Doherty combiners that enable an extended back-off efficiency range using fully symmetrical devices. As a proof of concept, we designed and fabricated two Doherty PA prototypes incorporating three-port pixelated combiners, implemented with GaN HEMT transistors. In measurements, both prototypes demonstrate a maximum drain efficiency exceeding 74% and deliver an output power surpassing 44.1 dBm at 2.75 GHz. Furthermore, a measured drain efficiency above 52% is maintained at the 9-dB back-off power level for both prototypes at the same frequency. To evaluate linearity and efficiency under realistic signal conditions, both prototypes are tested using a 20-MHz 5G new radio (NR)-like waveform exhibiting a peak-to-average power ratio (PAPR) of 9.0 dB. After applying digital predistortion (DPD), each design achieves an average power added efficiency (PAE) above 51%, while maintaining an adjacent channel leakage ratio (ACLR) better than -60.8 dBc.
