Robust Small-to-Large Signal Transition Extrapolation for ANN-based Transistor Models

This work proposes a pre-processing technique to improve the small-to-large signal extrapolation in an ANN-based behavioral model for packaged transistors. By using the proposed technique, it is possible to use the large-signal load-pull measurements of the device together with its S-parameters and guarantee a robust transition to the small-signal region, which is useful in the simulation of complex PA architectures. The proposed model was tested for a very high-power class-C biased GaN HEMT, in a commercial harmonic balance simulator, both in single-ended and sequential load-modulated balanced amplifier configurations.