Comparative Analysis of Deep Neural Networks for Gallium Nitride High Electron Mobility Transistor Load Pull Data Reconstruction

Active load pull measurement is a dynamic field of research with promising applications in amplifier design, as it accelerates the search for an optimal transistor operating point. This study delves into the use of deep Artificial Neural Networks (ANNs) for load-pull data imputation. Our approach us...

Full description

Saved in:
Bibliographic Details
Published in2024 10th International Conference on Control, Decision and Information Technologies (CoDIT) pp. 1281 - 1286
Main Authors Alleman, Julien, Prigent, Michel, Courreges, Fabien
Format Conference Proceeding
LanguageEnglish
Published IEEE 01.07.2024
Subjects
Convergence
Data models
Gallium nitride
HEMTs
Load modeling
MODFETs
Periodic structures
Power amplifiers
Power generation
Transistors
Online AccessGet full text

Cover

More Information
Summary:Active load pull measurement is a dynamic field of research with promising applications in amplifier design, as it accelerates the search for an optimal transistor operating point. This study delves into the use of deep Artificial Neural Networks (ANNs) for load-pull data imputation. Our approach uses simulated data from a Gallium Nitride (GaN) High Electron Mobility Transistor (HEMT) model, ensuring relevance to actual measurement conditions. The efficiency of Autoencoder, U-Net, and Wasserstein Generative Adversarial Network (WGAN) models is evaluated with the dataset, utilizing masking rates from 90% to 99.9%. This study emphasizes the performance of U-Net and Autoencoder models, which provide a streamlined approach by directly generating results, in contrast to the iterative convergence loop needed for the WGAN approach. Furthermore, these models demonstrate remarkable performance in reconstructing load-pull measures when a substantial amount of data is available.
ISSN:2576-3555
DOI:10.1109/CoDIT62066.2024.10708344