Intelligent Modeling for GaAs pHEMT Device
Case Studies
Accurate compact models are essential for reliable RF circuit design, but conventional GaAs pHEMT modeling workflows are often time-consuming, expertise-dependent, and difficult to maintain across wide operating conditions. Analog Devices, Inc. (ADI) collaborated with Keysight to address these challenges using an intelligent modeling solution that combines Machine Learning (ML)-driven automated parameter extraction with a hybrid Artificial Neural Network (ANN)-enhanced ASM-HEMT compact model.
The derivative-free ML optimizer in IC-CAP automates parameter extraction by simultaneously fitting multiple device characteristics, reducing manual effort and improving robustness. A hybrid ANN framework further enhances model accuracy by introducing ANN-based capacitance and resistance elements into the ASM-HEMT model, enabling improved bias-dependent small-signal and large-signal performance.
The solution reduced extraction time from several days to a few hours while improving agreement with measured DC, S-parameter, output power, PAE, and IMD3 across wide bias conditions. By reducing reliance on expert knowledge and delivering more accurate simulations, ADI accelerated model development and improved confidence in RF product design.