Characterization of Microwave Transistors Using DC Bias Conditions

Transistor models are crucial for circuit simulation. Reliable design of high-performance circuits requires that transistor characteristics are adequately represented, which makes accurate and fast models indispensable. Scattering (or S-)parameters are perhaps the most widely used RF characteristics, employed in the design and analysis of linear devices and circuits for calculation of the input and output impedance, isolation, gain, as well as stability, all being important performance figures for small-signal or low-noise amplifiers. However, transistor S-parameters are highly-dependent on frequency, bias conditions (both voltage and current), and temperature. Consequently, transistor models that are to exhibit design utility must properly account for these relationships, while being flexible, cheap to evaluate, and easy to implement in computer-aided design environments.  In this dataset, the measured scattering parameter characteristics of three transistors (BFP193W, BFP720ESD, and VMMK-1218) are provided and split into the training and testing sets for the purpose of constructing regression models of the scattering parameters as functions of the DC bias conditions. The allocation of the training data is chosen to facilitate a rendition of globally accurate models while using a possibly small number of samples.  For more detailed information refer to [1]. [1] N çalık, F. Güneş, S. Koziel, M. A. Belen, and P. Mahouti, “Deep-Learning-Based Precise Characterization of Microwave Transistors Using Fully-Automated Regression Surrogates,” to be published, IEEE TMTT.