MODELING OF HIGH-SPEED AND HIGH-POWER SEMICONDUCTOR-DEVICES

Authors
Citation
M. Andersson, MODELING OF HIGH-SPEED AND HIGH-POWER SEMICONDUCTOR-DEVICES, Acta polytechnica Scandinavica. El, Electrical engineering series, (79), 1994, pp. 1-45
Citations number
NO
Categorie Soggetti
Engineering, Eletrical & Electronic
ISSN journal
00016845
Issue
79
Year of publication
1994
Pages
1 - 45
Database
ISI
SICI code
0001-6845(1994):79<1:MOHAHS>2.0.ZU;2-W
Abstract
The existing SPICE semiconductor models of today do not model the devi ces accurately enough at high speeds or high power levels. To meet the challenges of rapidly evolving semiconductor technologies, electronic systems with higher performance and the increased use of computer-aid ed design methods, new improved models are needed. In the work describ ed in this thesis, two different approaches for solving this modeling problem for high-speed and high-power devices have been utilized: phys ical modeling and macromodeling. Physical models are strongly based on the device physics and typically quite complex. Therefore, they must be implemented into a specific circuit simulator by adding the model e quations to the simulator code. Macromodels are empirical models which can be constructed using elements already available in a circuit simu lator, and thus this type of model is more general and can be made to work for different circuit simulators without programming. The advanta ges and drawbacks of both modeling techniques are compared. Physical m odeling and macromodeling techniques yielding good results are present ed for silicon power devices and for silicon, gallium arsenide and sil icon-germanium high-speed devices. The work has been performed using t he APLAC circuit simulation environment, and as a result, a number of improvements in device modeling have been added to APLAC. An extensive device model library including APLAC programs for model parameter ext raction is also described. The use of this model library is based on t he methods for device characterization and parameter extraction develo ped simultaneously with the device models. Finally, some ideas for fut ure improvements of semiconductor models are presented.