Bipolar Transistor Parameter Extraction, Modeling and Test Structure Design
In this thesis, techniques are presented for determining the model parameter values required by compact bipolar transistor models in circuit simulators. Specifically, extraction of the Early voltages and the transistor series resistances are treated in detail. Extraction of other compact model parameters is also covered.
A general technique is suggested for extracting the Early voltages of different Early effect formulations found in compact models. This technique is shown to enable extraction of all Early effect parameters from dc measurements only. Thus, ac measurements required by previous techniques are not necessary. The Early effect is also characterized using a dedicated transistor test structure. This structure allows measurement of the bias dependence of the Early voltages, which are commonly assumed to be constant. It is shown that these parameters exhibit a geometry dependence not included in most compact models in use today.
The test structure used for characterizing the Early effect was originally designed for series resistance extraction. The use of the test structure considerably simplifies the extraction of these resistances using data from the transistor in normal operation. In contrast to previous methods, measurement of the bias dependent base resistance and the constant part of the collector resistance becomes a straightforward procedure. Furthermore, the valid bias range of the measured base resistance data is greatly extended by use of the test structure.
Besides relying on the use of the test structure, much of the base resistance extraction procedure is based on a new model for the bias dependent base resistance. This model, which is presented in the thesis, is intended for inclusion in a compact transistor model. Both emitter current-crowding and base conductivity-modulation are accounted for in our model. By using this model in combination with the test structure, we show how to extract the zero-bias intrinsic base sheet resistance using data from the transistor in normal operation.
General aspects of the design and use of test structures for parameter extraction are also discussed. It is shown that parameters extracted from test structures not necessarily translate directly into compact model parameters. In the case of the distributed base resistance, it is shown that parameters extracted using the test structure cannot directly be used in compact models. However, a simple conversion procedure for obtaining the desired compact model parameters is presented.