Study of X-Parameters Modeling for Microwave Power Devices Based on ANNs

In order to achieve the large signal modeling for the microwave power devices, the double hidden layer back propagation neural network (BPNN) and general regression neural network (GRNN) are utilized for X-parameter modeling of a GaN transistor here. Then, the harmonic balance experiments have been carried out to verify the validity of both models. As a result, the three harmonic and modulus value of predicted data and expected data can be obtained. In detail, the three harmonic errors of the double hidden layer BPNN are 5.287 dBm, 3.320 dBm and 4.483 dBm, respectively, and the three harmonic errors of the GRNN are 0.130 dBm, 0.001dBm and 1.235 dBm, respectively. In addition, the three harmonic modulus errors of the double hidden layer BPNN are 0.003, 0.521e-4 and 0.683e-6, respectively, and the three harmonic modulus errors of the GRNN are 0.001, 0.235e-4 and 0.304e-6, respectively. In conclusion, the proposedGRNN model is effective to model the large signals for GaN HEMT transistors.