Elsevier

Microelectronics Reliability

Volume 47, Issues 9–11, September–November 2007, Pages 1696-1700
Microelectronics Reliability

New developments of THERMOS3, a tool for 3D electro-thermal simulation of smart power MOSFETs

https://doi.org/10.1016/j.microrel.2007.07.012Get rights and content

Abstract

In this work, we present a novel 3D electro-thermal simulation tool capable of taking into account also particular driving strategies of the electron device, as it may be the case of smart power MOSFETs where a control logic interacts with the power section and controls its dissipated power and temperature. As an example, a thermal shutdown circuit, capable of reading the temperature on chip and switching the device off if the latter reaches dangerous values, usually embedded within smart power devices used in automotive applications to drive direction light or small motors/actuators, is simulated to validate our approach.

Introduction

The simulation of power electron devices is a complex numerical problem and it has been faced in many different ways in the recent past as the issue of power dissipation together with the knowledge of electro-thermal interactions in MOSFETs has become relevant. Being coupled, but on a different timescale with respect to the thermal problem, the electrical problem is usually reduced to few DC equation that are sufficient to model the temperature dependent static behaviour of the single cell. The time-dependent heat equation is solved, numerically or analytically, in a way such as to take into account layout geometries, boundary conditions and heatsink influence.

The THERMOS3 simulator we propose, entirely written in MATLAB, is based on a forward iterative finite difference time domain scheme [1] where the electrical equivalent of the thermal problem and the electrical quantities (drain current, source voltages) are solved self-consistently at each discrete time step. Non linearity in the thermal conductivity together with voltage depolarization due to finite resistance of the source metallization is explicitly taken into account. The ID = f(VGS, VDS,T) behaviour of the unit cell is modelled considering temperature dependence of the threshold voltage, mobility and MOSFET internal resistances. Convection on the surface and sides and isothermal boundary condition at the heatsink have been used.

As an exhaustive example of the capability of this new tool, we present the simulation of a smart power device produced by STMicroelectronics (Fig. 1). This device is protected during permanent short circuit by a current feedback loop which keeps the dissipated power inside the SOA together with a thermal shutdown circuit which shuts the device of if temperature on chip exceeds a reference value. Dynamic temperature behaviour of this device has been fully characterized by fast transient infrared imaging [2], therefore a quantitative comparison with the simulation is available for validation. One main issue is the optimization of time stepping since the overall behaviour of the device has to be observed in the 100 ms (or longer) timescale while sensitive information such as fast temperature transients can be as short as 100 μs. In our approach, an adaptive time stepping strategy together with switching event prediction and detection has been implemented.

Section snippets

The simulator

A multicellular power VDMOS can be treated as a number of equivalent single cells all of them sharing the same voltage on the drain and gate terminal and with the sources connected through a metal layer with a finite resistivity. Therefore, both VDS and VGS are dependent not only on temperature (the resistivity of the metal layer is itself a function of temperature) but on the layout of the device and position of the source bond wires. The single cell is modelled by the following set of

Experimental results

To compare and validate the outcome of THERMOS3, we have experimental temperature map at our disposal. The detection of the temperature distribution across the area of the device in transient conditions has been made possible by the use of a custom developed radiometric 2D measurement system that allows acquiring the transient temperature maps of the device surface [2]. By means of this equipment, we have obtained the thermal transient analyses used to calibrate the output of the simulator.

Results, comparisons and comments

The device has been discretized in Nx × Ny ×  Nz unit cells each with its own thermal parameters and operated in short circuit to the usual battery voltage Vbat = 12 V. Current feedback fixes the total current at Ilim = 10 A. In Fig. 4, we report a comparison between experimentally detected temperature distribution and the result of the simulation. As we can note, a good quantitative agreement is obtained both in shape and in peak temperature value.

To further investigate the features of the simulator, we

Conclusions and future developments

In our study on these kind of devices and in all the experience acquired in the electro-thermal simulation of power devices, we have understood that not only the localization of the bond wires is important in simulations but also a proper electro-thermal description of these components must to be taken into account. Into the next new version of the THERMOS3 simulation tool will be inserted also the possibility to well simulate and describe how the power bond wires, described in terms of

Cited by (27)

  • Automatic TCAD model calibration for multi-cellular Trench-IGBTs

    2014, Solid-State Electronics
    Citation Excerpt :

    The calibration procedure is specific for a multi-cellular T-IGBT because of two aspects: the trench-gate structure has a compact model description which is different from the others available (lateral, planar and so on); the multi-cellular structure of the device allows to treat the lateral boundary of the elementary cell with reflective (or ideal Neumann) boundary conditions and, from the thermal point of view, they are treated with adiabatic conditions. If the structure does not show a multi-cellular design, the latter boundary condition gives an overestimation of the temperature in the device and a non uniform current distribution can occur [16–21]. To achieve a good thermal calibration of the elementary cell it is mandatory to accurately specify the thermal boundaries of the elementary cell at the extremities [22,23].

  • Distributed electro-thermal model of IGBT chip - Application to top-metal ageing effects in short circuit conditions

    2013, Microelectronics Reliability
    Citation Excerpt :

    Nevertheless, the electrical representation used for such models must remain sufficiently light for the trade-off between computing time and complexity. Our modeling approach is a variation from [5] applied on IGBT chips in order to evaluate the ageing effects of the device through some electrical and thermal ageing material properties during severe operating conditions. The aim of this model is to allow the accurate analysis of effects in the electrical and thermal coupling over the chip.

  • FEM simulation approach to investigate electro-thermal behavior of power transistors in 3-D

    2013, Microelectronics Reliability
    Citation Excerpt :

    In this scheme the power transistor consists of N = nx·ny basic cells. The temperature of particular transistors can be calculated e.g. by a thermal network or reduced model as shown in [3,4]. This increases the accuracy of the circuit simulation.

  • 3D electro-thermal simulations of wide area power devices operating in avalanche condition

    2012, Microelectronics Reliability
    Citation Excerpt :

    At this point the I–V curves, as function of the operative temperature, are stored in a look-up-table used by the electrical solver. The simulator uses the relaxation method between the thermal and electrical problems [8]. The advantages of this method are the relative simplicity of the implementation and the reduction of the simulation time.

  • ANSYS based 3D electro-thermal simulations for the evaluation of power MOSFETs robustness

    2011, Microelectronics Reliability
    Citation Excerpt :

    The drawback is represented by a high computational effort due to the data transfers between the two simulators during the iterations. Simulators based on the direct method are faster but usually less accurate and more difficult to program because both the electrical and the thermal problems are solved using a matrices based circuit simulator approach [4,5]. An innovative approach exploits the reduction of the model order by means of a dedicated tool which leads to an equivalent thermal network.

View all citing articles on Scopus
View full text