Modélisation physique d'un transistor de puissance IGBT

The IGBT transistor, associating the conduction advantages of the bipolar transistor and the switching advantages of the MOSFET transistor, is widely used in medium and high power applications with an operating voltage of 1.2kV to 4.5kV. New topologies, reducing the losses in the semi-conductor devices, are often proposed and have to be validated through a simulation tool. The finite element method (FEM) simulators accurately analyse losses in components, but the associated simulation time is so long that they cannot be used for complex topologies in power electronics. An equivalent transistor macromodel, correctly and quickly predicting currents and voltages on the element is needed. This thesis proposes an IGBT transistor model based on semiconductor physics and validated through a comparison with finite element simulator results. The model yields very good results regardless of transistor load and operating point. Base conduction is modelled using a novel equivalent charge approach. This results in a relatively simple macromodel with a good representation of the static and dynamic phenomena. The solving method of the Poisson equation has been revisited to fit recent changes of the internal structure of IGBT transistors. All approximations have been validated through fi- nite element analysis using an appropriate simulation set. The approximation of the internal MOSFET, even though based on existing models, takes into account the specificities of the IGBT transistor structure. It shows the degenerated properties of the internal MOSFET compared to the properties of a standard MOSFET, regardless of the operating point. Using intermediate potentials within the semiconductor structure allows a separate analysis of the different voltage contributions. A complete IGBT transistor macromodel has been developed and tested in different power electronics structures. Good results were obtained in all the tested structures with the proposed IGBT transistor macromodel. Parameter extraction can easily be automated using a mathematical tool such as Matlab and only requires few measurements on the device. The obtained results can then be quickly adapted to an other IGBT device.