Large-Signal Modeling of Microwave Transistors
Doctoral thesis, 1999
The development of computer aided design tools for microwave circuit design has increased the interest for accurate transistor models. The circuit complexity has grown as the CAD tools have been improved and the need to predict how non-linear circuits behave has also been increased. Thus the CAD programs nowadays can simulate non-linear circuits accurately; the research challenge is more into the modeling field. This thesis deals mainly with non-linear models for microwave transistors, field effect transistors as MESFET's, HFET's and LDMOS's, and bipolar transistors as HBT's. Different model approaches are discussed and different measurement techniques as well.
In the appended papers, the empirical Chalmers MESFET and HFET model are developed to include thermal and dispersion effects. Moreover, the Chalmers model is also applied to an LDMOS device. A survey of widely used non-linear models and their capability to predict the DC and large-signal characteristics of an HFET device are presented. The standard Gummel-Poon model is first shown to be useful for large-signal simulations of an HBT by a power spectrum method and in a later paper is the model extended to include surface recombination effects in the base- emitter junction of HBT's. In the model development work for the silicon based LDMOS transistors the difficulties to de-embed the parasitic elements of the pad pattern were addressed. This was solved by a direct de-embedding technique, which took advantage of the algorithms developed for the calibration of the vector network analyzer. Long time measurements shown that the light sensitivity of a device can be important to consider, the daylight variation shown that for a probed HFET device the characteristics can be light sensitive and this is discussed as well. Finally, an 850 MHz class-E amplifier, which use a LDMOS as switching element is presented. Measured results shows state of the art performance for class-E operation in output power and efficiency in good agreement with simulations.
model parameter extraction