Parameter extraction method for Cauer model considering dynamic thermal diffusion boundaries in IGBT module

Abstract

Considering the change of working conditions, the calculation of thermal resistance and thermal capacity with the fixed shape of heat conduction area will result in a large calculation error for junction temperature of the Cauer model. Based on the conventional Cauer model, a circular heat conduction area is chosen to calculate the thermal resistance and thermal capacity in this manuscript, which avoids the inaccurate calculation of heat conduction area under different working conditions by the fixed heat diffusion angle method. In addition, a few fiber grating sensors are arranged on the baseplate to track the selected isotherm in real time, and update the isotherm as the thermal diffusion boundary according to the simulation rule. Furthermore, accurate thermal resistance and thermal capacity values are calculated online according to the circular heat conduction area that its boundary is the selected isotherm. Finally, the correction of the Cauer model with real-time updated thermal resistance and thermal capacity can calculate a more accurate junction temperature.

Introduction

The Insulated Gate Bipolar Transistor (IGBT) is combination of Bipolar Junction Transistor (BJT) and Metal-Oxide-Semiconductor Field-Effect Transistor (MOSFET), which is widely used in the fields of photovoltaic power generation, wind power generation, electric vehicles and electric traction due to its high input impedance, low saturation voltage drop and low drive power [1]. With the increasing demand for power density, the reliable operation of IGBT modules in the more rigorous environments has become the focus of attention [2].

Studies have shown that about 60% of power device failures are due to increased thermal resistance (R) or additional power loss induced junction temperature rise [3]. However, due to the multi-layer structure of the IGBT modules, obtaining the junction temperature is a complex task. The existing IGBT junction temperature measurement methods mainly included optical method, temperature sensitive electric parameter (TSEP) method and thermal network model method [4]. Optical method is mainly measured by infrared cameras and optical fibers, which has the advantage of directly obtaining the temperature distribution of power devices, but require the removal of the package housing during the measurement [5]. The thermal electric parameter method relies on the relationship between TSEPs and junction temperature to obtain junction temperature. The common TSEPs include on-state voltage (V_ce-on) [6], gate voltage (V_ge) [7], on-time (t_on) [8] and off-time (t_off) [9], etc. However, the measurement circuit design of TSEPs is complex, the required device accuracy is high, and need to change the power converter hardware structure. It is difficult to measure and require high-cost experimental instruments [10]. The existing thermal network model can accurately obtain the junction temperature of IGBT module under certain conditions, and then judge the reliability of IGBT module. The thermal network model includes the Foster model and the Cauer model. The Foster model can be fitted by data, but when the reference temperature point changes, the junction temperature in the conventional Foster model changes synchronously with the reference temperature [11,12]. Cauer model can reflect the nature of physical heat conduction of IGBT module and clearly represent the heat conduction process from chip to heat sink [13]. Therefore, Cauer model based on device size and material characteristics is one of the preferred methods at present. To build the Cauer model, only the material and size information of the packaging structure is needed, which is simple and convenient. In Ref. [14], the real-time influence of chip junction temperature on loss is considered, but the dynamic response of each layer of the thermal network model is not considered, which is meaning that the parameter variation of the thermal capacity (C) is ignored. In Ref. [15], each power cycle is equivalent to the combination of full response and zero input response. The numerical relationship between junction temperature and thermal network parameters is analyzed, but only the changes of heatsink are considered, and the changes of IGBT module are ignored. In Ref. [16], the square heat conduction area is used in the calculation of R and C, but the finite element simulation results show that the heat conduction area is closer to the circle, and the circular heat conduction area conforms to the actual situation. Moreover, most of the existing research literatures employ constant thermal network parameters [[11], [12], [13], [14], [15], [16]], which is the main reason for the large error between the current calculation result and the actual junction temperature.

For the conventional Cauer model, P_loss generated by IGBT chip is regarded as heat source and the heat flow (fixed 45° angle conduction) through the seven-layer structure of the module is represent by the RC network [17]. In the actual working conditions, the collector current and the emitter saturation voltage drop do not remain constant, so the P_loss will also change. The changing heat loss makes the thermal diffusion boundary of each layer in IGBT module change in real time, which leads to a large error in the junction temperature calculation results. Based on this, in this manuscript, the selection rule of dynamic thermal boundary under different convection coefficients is analyzed through simulation，and the dynamic thermal diffusion boundary and its circular heat conduction area radius under different working conditions are obtained by placing few fiber grating sensors on the baseplate to capture the selected isotherms that meet the accuracy of R and C calculation. The R and C of the Cauer model is modified online to obtain accurate junction temperature to meet the requirements of actual variable operating conditions and junction temperature analysis during heating up.