Comparison of the electro-thermal constraints on SiC MOSFET and Si IGBT power modules in photovoltaic DC/AC inverters
Introduction
Silicon carbide (SiC) semiconductor components are being used increasingly in power electronic applications, mainly because of their high switching speeds which improve the overall efficiency and/or the compactness of the inverters [1] [2] [3].
In the case of photovoltaic installations, the inverter has the highest failure rate and the anticipation of its breakdowns is an important issue. Moreover, few studies have been conducted on the reliability of this type of converter using SiC MOSFETs [4].
In this context, the junction temperature TJ of the semiconductors and its variations over time ΔTJ contribute to accelerating the ageing of the DC/AC inverter [5] [6] [7]. Many studies have proposed methods to estimate this temperature especially for IGBT power modules using thermal models [8] [9].
The aim of our study is to compare the junction temperature swings in a SiC MOSFET and in a Si IGBT power module used in a 2 level photovoltaic inverter, having the same current and voltage ratings. A numeric tool is used to estimate the junction temperature from current mission profiles.
Knowing the temperature and its variations over time, coupled with the study of degradation modes and mechanisms as well as the mission profiles, will allow to estimate the lifetime of these semiconductors in photovoltaic applications.
In this paper, the method used to estimate the junction temperature of the devices is described. This model is composed of several sub-models, the main ones being the losses estimation model and the thermal model. These two models can be coupled in order to take into account the variation of the losses depending on the temperature of the components. Temperature, estimated as a function of time, is then injected into a cycle counting algorithm named “Rainflow” that allows to obtain, for a given temperature profile, the number of occurrences for each value of ∆ TJ [10]. This approach is shown in Fig. 1, where I is the current profile, P the losses profile for a given semiconductor component, TJ the junction temperature profile, ∆ TJ the variation of this temperature, TJM the mean temperature and TA the ambient one. All the calculations are made with analytical models and are implemented in Matlab software.
Section snippets
Selection of the components
The direct comparison of SiC and Si technologies is not obvious. In fact, the electro-thermal stresses on the packaging depend a lot on the chosen configuration. For example, using the same technology, these stresses are different when changing the current rating of the devices: the maximum temperature and the thermal management system will change. Furthermore these stresses depend a lot on the application (current level, switching frequency…). The cost would also be an important point. Thus,
Considerations for the calculation
The evolution of the losses in the semiconductor components as function of time is estimated using the mission profile presented above. Due to the large number of samples (one sample every 20 ms as explained in Section 2.2) throughout the year, the calculations are simplified using only the value of the average losses on the fundamental period Tbf. Therefore, the temperature variations within each fundamental period in our calculations [11] are not taken into account.
In the following paragraphs
Electro-thermal coupling
The electro-thermal coupling is used to take into account the dependence of some electrical parameters on the junction temperature. This coupling exists in the case of an IGBT, but its effect is much more important in the case of MOSFET, in particular for RDSon. Fig. 12 shows the temperature dependence of RDSon issued from the manufacturers' datasheets, plus its dependence on the current IT. The same work was done to estimate the variation of the parameters ET, RT, E0 and RD of the IGBT power
Modeling methodology
The thermal model has, as input, the losses profile. In this model, the thermal impedances ZthCH (Case-Heatsink), ZthHA (Heatsink-Ambient) and ZthJC (Junction-Case) are used. The latter is given by the following equation according to the Foster model [13]:where Rthi is one elementary thermal resistance of the model, τi one elementary time constant, at the point i, and N the number of RC cells in the model.
Usually, the use of Foster models is only possible if the
Integration of the models and Rainflow algorithm
“Rainflow” is a cycle counting algorithm (cycles contained in a mission profile) [15] [16].The TJM profiles of the various modules are the entries of this algorithm, which gives as output the number of occurrences N of each value of ∆ TJ at a given level of average TJ (TJM), represented in form of tables and histograms N = f (TJM, ΔTJ) [17] [18] [19].
By applying this algorithm on the temperature profiles obtained in Fig. 14, comparative histograms of TJM(Fig. 16) and of ∆ TJ (Fig. 17) in the case of
Conclusions
In this article, a model was used in order to estimate the junction temperature (losses estimation model + thermal model), to compare its evolution between two power module technologies using a current profile measured over a year in a photovoltaic plant. These modules use Si IGBTs and SiC MOSFETs with the same current and voltage ratings.
The obtained temperature profile is then injected into an algorithm named “Rainflow”. The results allows to observe that the average junction temperature and
Acknowledgments
This project has received support from the State Program Investment for the Future bearing the reference (ANR-10-ITE-0003).
References (20)
- et al.
Reliability enhancement by integrated liquid cooling in power IGBT modules for hybrid and electric vehicles
Microelectron. Reliab.
(2014) - et al.
Mission-profile-based stress analysis of bond-wires in SiC power modules
Microelectron. Reliab.
(2016) Determination of fragments of multiaxial service loading strongly influencing the fatigue of machine components
Mech. Syst. Signal Process.
(2009)- et al.
Application of Kriging and radial basis function in power electronic module wire bond structure reliability under various amplitude loading
Int. J. Fatigue
(Dec. 2012) - et al.
MOSFET technology advances DC-DC converter efficiency for processor power
SiC MOSFETs Enable High Frequency in High Power Conversion Systems
(Feb. 2016)SiC power devices and modules
- et al.
Failure modes and robustness of SiC JFET transistors under current limiting operations
- et al.
Impact of absolute junction temperature on power cycling lifetime
- et al.
Quantifying the thermal fatigue of CPV modules
Presented at the 6th Int. Conf. Concentrating Photovoltaic Syst
(2010)
Cited by (20)
A python implementation based lattice Boltzmann method for thermal behavior analysis in silicon carbide MOSFET
2024, Micro and NanostructuresCorrection strategy for wear-out prediction of PV inverters considering the mission profile resolution effects
2022, Microelectronics ReliabilityMission profiles and hygrothermal conditions: Its effects on the reliability of a soft switching converter
2020, Microelectronics ReliabilityCitation Excerpt :An example is the high gain DC-DC converter described in [5], which exhibits a maximum efficiency of 97%, and 96% at rated power. Another approach to improve efficiency is to employ switching devices with better characteristics, such as silicon carbide (SiC) diodes and transistors, which are commercially available in high voltage and current ratings [6]. For instance, a T inverter rated at 60 kW, built with SiC devices, achieved a peak efficiency of 99.2% [7].
Mission profile resolution impacts on the thermal stress and reliability of power devices in PV inverters
2018, Microelectronics ReliabilityCitation Excerpt :Thus, the reliability prediction results (e.g., estimated failure rate) may deviate considerably for different mission profile resolutions. In the previous studies, various time-resolutions of mission profiles have been applied in the reliability analysis of PV inverters, ranging from below 1 min to 1 h [3–15]. According to the literature survey, the most commonly used mission profile resolution is in the range of a few minutes (e.g., 1–10 min).
SPICE-Aided Nonlinear Electrothermal Modeling of an IGBT Module
2023, Electronics (Switzerland)Simplified selective harmonic elimination PWM for three-level five-phase T-NPC inverter
2023, International Journal of Circuit Theory and Applications