A Novel Electro-thermal Simulation Approach to Power IGBT Modules for Automotive Traction Applications

Transcription

1 Special Issue Recent R&D Activities f Pwer Devices fr Hybrid Electric Vehicles 27 Research Reprt A Nvel Electr-thermal Simulatin Apprach t Pwer IGBT Mdules fr Autmtive Tractin Applicatins Takashi Kjima, Yasushi Yamada, Maur Ciappa, Marc Chiavarini, Wlfgang Fichtner This paper describes a nvel electr-thermal cupling simulatin technique fr analyzing autmtive IGBT mdules. This technique uses a electric circuit simulatr and is based n a pwer semicnductr device mdel with temperaturedependent characteristics and a nvel cmpact thermal mdel suitable fr autmtive IGBT mdules. Fr the device mdel, a mdel parameter definitin methd was prpsed, and simulatin results f n-vltage characteristics using this mdel shwed gd agreement with Abstract measurement results. The cmpact thermal mdel can take int accunt lateral heat spreading within the mdules and thermal interference amng pwer devices. The thermal mdel was validated in a cmparisn f temperature transient respnses calculated using the prpsed mdel thse calculated by FEM, and thse which were measured. The usefulness f the electrthermal cupling simulatin technique was shwn in example simulatins which included tw parallel IGBTs with resistive lad. Keywrds Circuit simulatr, IGBT mdule, Cmpact thermal mdel, Mdel parameter, Hybrid electric vehicle, Pwer semicnductr device, FEM, Thermal interference

2 28 1. Intrductin In hybrid electric vehicles (HEVs), pwer IGBT (Insulated Gate Biplar Transistr) mdules, which cnvert the direct current f the battery int alternating current t rtate the mtr, generate a cnsiderable amunt f heat because f huge dissipatin f electric pwer. In additin, the electrical characteristics f the pwer semicnductr devices that are the main cmpnents f pwer IGBT mdules (IGBTs and Dides), depend strngly n their junctin temperature. An electr-thermal simulatin technique is therefre required t estimate the electrical and thermal behavir f the pwer IGBT mdules. Hwever, cnventinal appraches using simple thermal mdels based n circuit simulatrs have encuntered significant difficulties in accurately predicting the transient behavir f cmplex autmtive pwer mdules. 1) This paper describes a nvel electr-thermal cupling simulatin technique fr analyzing autmtive IGBT mdules. This technique is based n a pwer semicnductr device mdel fr determining temperature-dependent device parameters and a nvel cmpact thermal mdel suitable fr autmtive IGBT mdules. Since this technique uses a circuit simulatr, it has the ability t estimate the detailed electrical characteristics f the devices such as surge vltage and pwer lss. This technique is capable f slving prblems that cannt be slved by the cnventinal methd; fr example, imbalanced temperature rise in parallelcnnected pwer devices with different characteristics can be examined In this paper, first the utline f the electrthermal simulatin methdlgy is described. Secndly, ur prpsed parameter definitin methd fr the electrical IGBT mdel is presented. Onvltage characteristics simulatin results using an ptimized IGBT mdel and measurement results are cmpared. Thirdly, ur prpsed cmpact thermal mdel is validated. Cmparisn f thermal pulse respnses calculated using this thermal mdel with thse which were measured, and als with FEM (Finite Element Methd) analysis results is made. Finally, examples f electr-thermal simulatin are presented. 2. Electr-thermal cupling simulatin Figure 1 shws a diagram f ur electr-thermal simulatin technique fr pwer IGBT mdules. In this figure, the mdule mdel cnsists f an electrical mdel and a thermal mdel. The device mdel, where electrical characteristics f IGBTs r dides are defined, is cnnected t the thermal mdel. The instantaneus value f the device pwer lss is applied t the thermal mdel, in which the thermal characteristics f the mdule are defined. Then, the instantaneus device temperature is generated by the thermal mdel, and the temperature dependent device mdel parameters are determined using this instantaneus device temperature. These calculatins are perfrmed simultaneusly using a circuit simulatr. As described abve, the device mdel and the thermal mdel are essential cmpnents f an electr-thermal simulatin. 3. Simulatin mdels 3. 1 Electrical IGBT mdel T accurately predict the lss dissipated frm an IGBT, a mdel fr the IGBT needs t vary the device characteristics dynamically with variatins f instantaneus device temperature. Since, the widely used circuit simulatr knwn as "SPICE" cannt mdel temperature-dependent device characteristics, we chse the SIMPLORER circuit simulatr (designed by Ansft Crpratin) as the slver fr the electr-thermal cupling simulatin. 2) In the IGBT mdel, the dependence f the characteristics n the temperature can be expressed in the fllwing frm. Generally, the amunt f pwer lss dissipated frm an IGBT is determined by the cnductin lss, Fig. 1 Pwer Device mdel Lad Cntrller Inductance etc. Electrical mdel Pwer mdule mdel Lss Temp. Chip Substrate etc. Heatsink etc. Thermal mdel Diagram f electr-thermal simulatin fr pwer IGBT mdules in HEVs.

3 29 dminated by the n-vltage cnductin lss, and the switching lss, dminated by the switching lss f the tail current in the transient current wavefrm. The IGBT mdel in SIMPLORER can represent the dependence f the lss f temperature n nly three parameters, which are determined by the simple parameter definitin methd described in this paper. Thse parameters are "saturatin current f BJT (Biplar Junctin Transistr)" and "base resistance f BJT" fr characterizing the n-vltage, and "time cnstant f tail current" fr the tail current. Each f these parameters can be expressed as a functin f temperature. Thse functins can be ptimally determined by minimizing the errr between the data calculated frm a trial functin and the data measured at different temperatures. Figure 2 shws the cmparisn between V CE -I C characteristics simulated frm the ptimum IGBT mdel and measured characteristics. Frm this figure it is clearly seen that the simulated results f the V CE -I C characteristics are in gd agreement with the measured results. The ptimized IGBT mdel is therefre useful fr evaluating the pwer lss generated by the IGBT Design f cmpact thermal mdel A cmpact thermal mdel, which is cmpsed f thermal resistance and thermal capacitance, is strngly required as a thermal mdel fr carrying ut the electr-thermal cupling simulatin, because the cmpact thermal mdel can be implemented using a circuit simulatr easily. The cnventinal cmpact thermal mdel such as the Elmre mdel cannt exactly represent a three-dimensinal structure I C (A) Gate MOSFET Cllectr Emitter V GE =15V BJT IGBT macr mdel (simplified) 15 C 1 C 5 C having temperature transient. 3) Specifically, the Elmre mdel cannt represent lateral thermal spreading and thermal interference n the real IGBT mdule munted n the water cler. We prpse a new cmpact thermal mdel t vercme these prblems. Figure 3 shws the "cell" fr each physical dmain, which cnsists f the thermal resistance and capacitance. In the cell, tw parallel thermal resistance and capacitance subcircuits are cnnected in series. The thermal impedance (Z th (t)) f the cell can be expressed by the fllwing equatin, Z th (t) = R 1 {1-exp(t/R 1 C 1 )} + R 2 {1-exp(t/R 2 C 2 )} (1) where the parameters R 1, R 2, C 1 and C 2 are determined t minimize errrs in cmparisn with results calculated by FEM. The cells crrespnding t physical layers in the IGBT mdule are cnnected in series t represent the ttal thermal mdel f the IGBT mdule. Figure 4 shws the time dependence f thermal impedances calculated using the ptimum cmpact thermal mdel and the transient heat FEM (the transient respnse t the heat unit step). Because the thermal impedance generated by the cmpact thermal mdel agrees with that by FEM, the prpsed cmpact thermal mdel is suited t represent the IGBT mdule. In additin, the prpsed thermal mdel has the capability t represent the interference f the heat flux frm different heat surces that are adjacent t each ther. Specifically, the interference f the heat flux can be represented by merging the individual thermal mdels at individual pints where the different heat fluxes interfere with each ther. Figure 5 shws the thermal mdel fr a mdule with tw IGBTs cnnected in parallel. In this mdel, the IGBTs which are the heat surces are represented as current surces. In Fig. 5, the interference area is indicated by the circle V CE (V) R 1 R 2 C 1 C 2 Z th (t) Fig. 2 V CE -I C characteristics f IGBT measured (dts), simulated (lines). Fig. 3 Cell f prpsed cmpact thermal mdel.

4 3 Figure 6 shws the temperature transient respnses t the heat unit step calculated using the prpsed cmpact thermal mdel. In the case that thermal interference des nt ccur (peratin f ne IGBT), the lateral thermal spreading is accurately mdeled, because the result calculated using the prpsed mdel is in agreement bth with the FEM and the measured results. Als in the case that thermal interference ccurs, (peratin f tw IGBTs), the result calculated using the prpsed mdel is in agreement with that calculated by the FEM, shwing that the thermal interference is accurately mdeled. 4. Examples f electr-thermal cupling simulatin In this chapter, a representative example f the electr-thermal cupling simulatin is presented. The cmmercial pwer mdule prduct "2 in 1 mdule" was chsen as a test sample, which is Thermal impedance (K/W) Thermal impedance (K/W) Fig. 4 Silicn (IGBT) Slder Cpper (DBC) AlN (DBC) Cpper (DBC) Slder Cpper (Baseplate) Aluminum ally Bttm f Heatsink Silicn (IGBT) Slder Cpper (DBC) AlN (DBC) Cpper (DBC) Slder Cpper (Baseplate) Aluminum ally Bttm f Heatsink Time (s) (a) Calculated by FEM Time (s) (b) Calculated using the prpsed mdel Thermal impedance fr each layer. cmpsed f tw IGBTs cnnected in parallel. The simulatin circuit is shwn in Fig. 7. Figure 8 shws the detailed junctin temperature (T j ) transitin in an electr-thermal cupling simulatin in the case f tw successive switching cycles. The T j f the IGBT increases rapidly with bth turn-n and turn-ff f current, and it rises slwly during the "n" state and decreases slwly during the "ff" state. Fig. 5 Fig. 6 IGBT pwer lss Temperature f IGBT Silicn (IGBT) IGBT temperature ( C) Slder Cpper (DBC) AlN (DBC) Cpper (DBC) Slder Cpper (Baseplate) Aluminum ally (Heatsink) Bttm f heatsink Thermal interference part Cmpact thermal mdel f package/heatsink system including tw IGBTs Effect f thermal interference Operatin f ne IGBT Operatin f tw IGBTs Time (s) Measured RC mdel FEM Calculated (lines) and measured (symbls) pulse respnse system withut (lwer curve) and with thermal interference (upper curve). Pwer lss at the steady-state (befre turn-ff) is 93.7 W.

5 31 Tw examples f simulatin using the prpsed technique are shwn belw. A. In the case f different threshld vltages Figure 9(a) shws the transient behavir f T j and the cllectr current fr the tw paralleled IGBTs that have slightly different threshld vltages (IGBT 1 : 6.83 V, IGBT 2 : 6.33 V). Bth the inrush current and the steady-state current are larger in IGBT 2 because f its smaller threshld vltage. Because the inrush and steady-state current f IGBT 2 is larger than that f IGBT 1, the lcal pwer lss f IGBT 2 becmes higher than that f IGBT 1. The higher lcal pwer lss led t a faster T j increase in IGBT 2. B. In the case f different thermal resistances In this example, the slder layers under the tw IGBTs are assumed t have different heat resistances because f the presence f vids r cracks. In spite f the fact that the cllectr currents f the IGBTs were almst the same, the T j f IGBT 2 increased faster than that f IGBT 1 because f the higher thermal resistance f IGBT 2 (Fig. 9(b)). In the abve cases, cmputatinal time f the electr-thermal cupling simulatin was nly 1% lnger than that f the cnventinal electric circuit simulatin. Because the abve results explain expected phenmena qualitatively, the electr-thermal simulatin technique using the prpsed mdel is useful. Here we must address undesirable results caused by the small time increments in the circuit simulatin. Fr example, a simulatin using detailed pwer device mdels requires shrt time increments n the rder f nan-secnds. Hwever, an unexpected lss f a few kw is ften generated in the simulatin using such shrt time increments. If this large lss is applied t the cmpact thermal mdel, an unrealistic increase f T j ver 1, C will ccur mmentarily. This unrealistic simulatin result can be eliminated by applying average pwer lss t the current surce implemented int the cmpact thermal mdel. Figure 1 shws the transient behavir f T j in the simulatin where pwer lss averaged ver a time span f.1 ms r 1 ms Vltage (V), Current (A) 2 1 IGBT 1.6 Temperature Current 1 Ω IGBT 2 Vltage 2 V NOT included in simulatin circuit Fig. 7 Simulatin circuit Increase f IGBT temperature ( C) IGBT current (A) IGBT current (A) Temperature (a) The case f different threshld vltages (IGBT1: 6.83 V, IGBT2 : 6.33 V) Current 1 Current (almst the same) Current 2 Current 1 Temperature 2 Temperature Temperature (b) The case f different thermal resistances (IGBT1 :.49 K/W, IGBT2:.54 K/W). 5 5 Increase f IGBT temperature ( C) Increase f IGBT temperature ( C) Fig. 8 Simulated temperature f ne IGBT in switching peratin. Fig. 9 Simulated transient behavir f T j and f I C in paralleled IGBTs.

6 32 is implemented int the cmpact thermal mdel. It is clear that any unexpected rapid rise and fall f T j is eliminated by the averaging taken ver a time span f 1ms. Implementatin f average pwer lss int the cmpact thermal mdel will nt have a fatal influence n the estimatin f T j in the pwer mdule, because it takes frm a few tens f secnds t a few hundreds f secnds fr the mdule t reach thermal equilibrium. 5. Cnclusin We have described a nvel simulatin technique fr perfrming electr-thermal simulatin. Key cmpnents f this technique are an electrical IGBT mdel with temperature dependent characteristics and a nvel cmpact thermal mdel. The parameters f the IGBT mdel are defined as analytical functins f temperature, and dynamic variatins f the parameters are calculated in a simple manner. The prpsed cmpact thermal mdel f pwer IGBT mdules can take int accunt lateral thermal spreading and thermal interference. The usefulness f the electr-thermal cupling simulatin technique was shwn in example simulatins which include tw parallel IGBTs with resistive lad. Acknwledgments The authrs are grateful fr the verall supprt f Mr. Ishik and Dr. Tadan (Tyta Central R&D Labs.). We als wuld like t thank Mr. Kari Oila (ETH) fr his technical supprt. References 1) Hefner, A. R. and Blackbum, D. L. : "Thermal Cmpnent Mdels fr Electrthermal Netwrk Simulatin", IEEE Trans. Cm., Packag. and Manuf. Tech., 17(1994), ) "SIMPLORER 6. User Manual", Ansft Crp. (22) T ( C) span f.1 ms span f 1 ms ) Yun, C.-S., Ciappa, M., Malberti, P. and Fichtner, W. : "Thermal Cmpnent Mdel fr Electrthermal Analysis f IGBT Mdule Systems", IEEE Trans. n Adv. Packag., 24(21), (Reprt received n Oct. 14, 24) Takashi Kjima Research fields : Pwer device, Pwer mdule, Pwer electrnics simulatin Academic sciety : Inst. Electr. Eng. Jpn. Yasushi Yamada Research fields : Pwer device, Pwer mdule, Pwer electrnics simulatin Academic degree : Dr. Eng. Academic sciety : Am. Inst. Phys., Inst. Electr. Eng. Jpn., Sc. Autmt. Eng. Jpn., Electrchem. Sc. Maur Ciappa* Research fields : Micrelectrnics, reliability physics, Physical characterizatin f semicnductrs, Pwer device reliability Academic degree : Dr. Eng. Academic sciety : IEEE Award : IEEE Third Millenium Medal Marc Chiavarini* Wlfgang Fichtner* Research fields : Micrelectrnics, Device physics, Device mdelling and simulatin Academic sciety : IEEE, Swiss Natl. Acad. f Eng., Austrian Natl. Acad. Sci. Award : IEEE Andy Grve Award *Swiss Federal Inst. f Technl. (ETH) Fig. 1 Effect f pwer lss averaging n eliminating unrealistic temperature.