On the Dependability of Load Dependent X-Parameter Models for Varying Loads

Load dependent X-parameter models offer an accurate description of nonlinear devices, especially in mismatched environments. However, the model prediction is ambiguous in many cases, because the model contains X-parameter sets on a predefined load pull grid. Hence, the prediction of the model is strongly dependent on its implementation and the underlying interpolation of the X-parameters. As shown in this work, this has a significant impact on the overall modeling error. As the interpolation behavior is seldomly specified, even commercially available simulators tend to generate predictions that suffer from relatively large model errors. This work compares these results with other known interpolation approaches, namely with a nearest neighbor approach and with a weighted superposition using several sets of X-parameters. The modeling errors are determined and compared by using these model implementations and measured load dependent X-parameter models as well as verification measurements. All the measurements are gained from the same GaN HEMT produced by Cree.