PCIM Europe digital days 2021, 3 – 7 May 2021 An Inverse Method to Evaluate the Chip Temperature Distribution within Press Pack IGBT Jie Chen1, Erping Deng1,2, Yongzhang Huang1,2 1 State Key Laboratory of Alternate Electrical Power System with Renewable Energy Sources (North China Electric Power University), Beijing, China 2 NCEPU (Yantai) Power Semiconductor Technology Research Institute Co., Ltd, Yantai, Shandong Province, China Corresponding author: Jie Chen, huadianchenjie@163.com Abstract Press Pack Insulated Gate Bipolar Transistor (PP IGBT) is quite famous for the high power applications but lacks in-depth reliability research, evaluation of chip temperature distribution will be more meaningful than classical virtual junction temperature for reliability assessment. In this paper, an inverse method combining the experimental measurement and numerical solution, called multi-current VCE (T) method is applied to evaluate the single chip temperature distribution within PPI in the power cycling test, which extends the capability in chip temperature measurement of the conventional VCE (T) method. Finally, the calculated results are verified by the designed experimental measurement. 1 Introduction Press Pack Insulated Gate Bipolar Transistors (PP IGBTs) are being increasingly attractive in many high voltage and high power density applications, such as High Voltage Direct Current (HVDC) transmission project [1]. Compared to typically wire bonded power modules, PP IGBTs have several superiorities due to different package structure (see Fig. 1), e.g., double side cooling, higher power density, and ease of connection in series [2]. Due to the complex operating conditions and high maintenance costs, how to improve its reliability has attracted more and more attention from academia and industry. Fig. 1: Exploded view of the PP IGBTs [2] Although the appearance of the PP IGBTs is not too late, there is not much research on reliability, ISBN 978-3-8007-5515-8 especially experimental investigation such as power cycling test (PCT) [3]. The PCT is considered as the most important test for the reliability evaluation of power devices with repetitive junction temperature swings which is typically generated by on-state losses resulting from the DC current. As early in the 1990s, the junction temperature swing ΔTj and the mean junction temperature Tjm were recognized as the two important factors affecting the PCT results [4]. Thus, a precise measurement of junction temperature is crucial for the PCT. Since a direct measurement is not possible for packaged PP IGBT, the method using the temperature sensitive electrical parameters (TSEPs) is an interesting and potential method [5]. Among these TSEPs method, the VCE(T) method, using the pn-junction voltage drop at a small current, has been established as the standard method for junction temperature measurement for IGBTs during power cycling test [6]. However, the method can only obtain an average junction temperature, called virtual junction temperature, rather than the temperature distribution. With the ever-growing power density in power devices, evaluation of temperature distribution will be more meaningful than the classical virtual junction temperature for reliability assessment. 1152 © VDE VERLAG GMBH · Berlin · Offenbach Authorized licensed use limited to: North China Electric Power University. Downloaded on March 02,2022 at 08:59:50 UTC from IEEE Xplore. Restrictions apply. PCIM Europe digital days 2021, 3 – 7 May 2021 Firstly, the temperature difference between the center and the edge of a chip has easily exceed 40K for single-sided cooling IGBTs by infrared (IR) measurement [7], but the temperature gradient of the chip within the PP IGBT is temporarily unknown. Although the finite-element simulation can be used to predict the junction temperature distribution [8], there is still no effective experimental verification method up to now for the simulation results. Secondly, as discussed in Ref. [9], the variables of the CIPS 2008 empirical lifetime model are not independent, such as the load pulse duration ton and load current IL. It is not possible to obtain the same ΔTj and Tjm with the same ton but different IL or the same IL but different ton, especially for the wide range. However, both the IL and ton are related to the lateral temperature gradient in the chip. As mentioned in Ref. [10], the temperature gradient may be an option for variable in empirical lifetime models if it can be determined without complex simulation. In [11], a sequential VCE(T) method with separated gate controller is proposed to measure the junction temperature distribution among paralleled chips within PP IGBT, this paper will focus on the single chip temperature evaluation using the multicurrent VCE(T) method, which essentially belongs to an inverse method with combination of experimental measurement and numerical solution. Moreover, the method is applied in the DC power cycling test and verified by the designed experiments. 2 The multi-current VCE(T) method 2.1 Theoretical foundation Generally, the virtual junction temperature measured by the conventional VCE(T) method is seen as average temperature of the chip with the three-dimensional temperature distribution. Thus, as long as the temperature distribution remains the same, the virtual junction temperature is a fixed value and will not change. In [12], the physical meaning of the virtual junction temperature, especially the averaging mechanisms, have been studied in-depth combined with theoretical analysis and electro-thermal simulation. The results show that the virtual junction temperature is not a fixed value and is related to the measurement current. The smaller the measurement current, the larger the virtual junction temperature and the closer it is to the maximum junction temperature. Furthermore, the root cause is also given in Ref. [12] that the measurement current distribution has ISBN 978-3-8007-5515-8 a positive exponential correlation with the temperature distribution. The higher the temperature, the higher the ratio of the measurement current, and the positive exponential correlation is more obvious when the measurement current decreases, as shown in Fig.2. It is such the coupling effect between the measurement current and temperature distribution that provides the theoretical foundation for the multi-current VCE(T) method. Fig. 2: Current distribution with temperature different measurement currents [12]. at 2.2 Method principle Unlike the traditional VCE (T) method measuring the voltage at only one measurement current, the multi-current VCE (T) method requires that the voltages are measured at multiple different measurement currents with the same test conditions. The basic principle of the proposed multi-current VCE (T) method is shown in Fig. 3. The IGBT chip is subdivided into N parts, and all parts are electrically paralleled. Thus, the voltage drop across these parts is the same, and the total measurement current is the sum of all partial measurement currents. When power loss is generated in the active area, the chip temperature is not uniform and the temperature of elements at different locations are different. According to the results of Ref. [12], the measurement currents flowing through each part with different temperature are different. When VCE measurement is performed at M different measurement currents, there is always the above relationship established, and a nonlinear equation group can be expressed as: 1153 © VDE VERLAG GMBH · Berlin · Offenbach Authorized licensed use limited to: North China Electric Power University. Downloaded on March 02,2022 at 08:59:50 UTC from IEEE Xplore. Restrictions apply. PCIM Europe digital days 2021, 3 – 7 May 2021 1 N ­ I ° 1 N ¦ I i (Ti ,V1 ) i 1 ° ° ° 1 N ° I ® j ¦ I i (Ti ,V j ) N i1 ° ° ° 1 N °I ¦ I i (Ti ,VM ) M °̄ N i1 (1) distribution evaluation requires N to be large enough, which seems to be a contradiction. In order to greatly reduce the number of nonlinear equations and solve this problem, the chip temperature distribution is represented by the quadratic function as shown in Eq. (2) since the chip surface temperature distribution has strong regularity. T ( xˈ y ) The I(T,V) relationship can be obtained through temperature calibration at different measurement currents. So, If M ≥ N, it is theoretically possible to solve the value of Ti in the nonlinear equations, which represent the temperature of each part in the chip. More description of the method is given in [13]. However, the numerical solution of the nonlinear equations is not an easy task if N is large. On the contrary, the accurate temperature p1 p2 ˄ x 2 y 2˅ (2) After substituting Eq. (2) into Eq. (1), the nonlinear equations contain only 2 unknown parameters p1 and p2, where the p1 is the maximum temperature at the center of the chip and the p2 represents the temperature gradient. In theory, 2 different measurement currents are enough to obtain the temperature distribution. To improve the accuracy of solving the nonlinear equations, the number of measurement currents can be greater than 2 to constitute the redundant equations. Fig. 3: basic principle of the proposed multi-current VCE (T) method. 3 Experiment For this experiment, a 3300V/75A diode chip instead of IGBT chip is selected as device under test (DUT) because the black coating used for infrared will cause the gate PCB to be insulated. The active area of diode chip is square with 10.2 mm ×10.2 mm. 3.1 I-V-T calibration The accurate I-V-T relationship of the DUT is the basis of the proposed multi-current VCE (T) method, which are obtained and fitted by temperature calibration at different measurement currents. Generally, the calibration of power devices is generally performed in a thermostat [14]. Raising the temperature of the thermostat to a certain temperature and keep it constant, then the temperature of the DUT is the same and uniform, ISBN 978-3-8007-5515-8 so the recorded case temperature can be regarded as the junction temperature of the DUT. This method requires that the DUT is small in size, so it can be considered that the air temperature distribution inside the thermostat is relatively uniform and stable under a limited spatial range and long heating time. However, there is inevitable error exist for press-pack packaged power devices since it has a large volume and requires a complicated clamping force fixture, which causes the total volume to become larger. Therefore, based on the structure of the PP IGBT, a double-sided heating calibration method is proposed here, requiring the same heating power on both sides, as shown in Fig. 4. Two heating plates are placed on the collector side and the emitter side to heat the DUT, and the aluminum plate with a larger heat capacity needs to be added between the heating plate and the DUT as a 1154 © VDE VERLAG GMBH · Berlin · Offenbach Authorized licensed use limited to: North China Electric Power University. Downloaded on March 02,2022 at 08:59:50 UTC from IEEE Xplore. Restrictions apply. PCIM Europe digital days 2021, 3 – 7 May 2021 thermal buffer to ensure that the heat from the heating plate reaches the surface of DUT more uniformly. In addition, the outside of the heating plate is insulated with a thermal insulation plate to ensure that the heat only flows to the DUT to ensure the symmetry of the heat flow. Lock nut Force sensor Thermal insulation board (FR4) Heating plate Heating plate Aluminum plate Groove for thermocouple Aluminum plate PPI PPI Aluminum plate Aluminum plate Heating plate Heating plate Thermal insulation board (FR4) Disc spring Fig. 4: Thermal calibration fixture for PP IGBT. Where a = 5.52×10−3, b = 24.9559, c = 0.0604, d = 1.6703×10−4. 3.2 Test setup The multi-current VCE (T) method is applied and verified in power cycling test at Iload=100 A, ton=20 s to investigate the junction temperature distribution within single-chip submodule PP IGBT. In addition, an external clamping force of 1.2 kN is necessary to make the internal components contact well. As mentioned in the second part, the key to the proposed method is to apply different measurement currents under the same temperature distribution. Obviously, it is impossible to implement it in one cycle since the junction temperature will decrease during injecting the measurement current. Thus, only one measurement current is injected in each cycle, and then others measurement currents are injected in other cycles at the same state. During power cycling test, all the test conditions are the same to reach a stable junction temperature in each cycle, so, it can be considered that the chip surface temperature distribution of each cycle is the same, which is the theoretical foundation of injecting these measurement currents in different cycles. To reduce the accidental error of measurement, each measurement current lasts for 5 cycles and the measured voltage VCE takes an average of 5 times. (a) External diagram Fig. 5: I-V-T characteristics of the DUT. The I-V-T characteristics of the DUT are obtained as shown in Fig. 5, which can be firstly fitted with temperature T as a parameter, then the appropriate fitting function are selected to perform the fitting: A(T ) exp >V B (T ) @ (3) a exp T b , B (T ) c +d T (4) I (V , T ) A(T ) ISBN 978-3-8007-5515-8 (b) Internal structure Fig. 6: Customized single FRD chip submodule. Experimental verification is not an easy task for temperature distribution evaluation within PP IGBT, a combination of the thermocouple and IR is adopted. The former is used to measure the 1155 © VDE VERLAG GMBH · Berlin · Offenbach Authorized licensed use limited to: North China Electric Power University. Downloaded on March 02,2022 at 08:59:50 UTC from IEEE Xplore. Restrictions apply. PCIM Europe digital days 2021, 3 – 7 May 2021 maximum temperature at the center of the active area, and the latter is used to measure the minimum temperature at the edge of the active area. To this end, a single FRD chip submodule is specially designed and manufactured for experimental verification, as shown in Fig. 6. There is a channel on the surface where the molybdenum and the chip in contact for inserting a thermocouple. The cross-sectional area of the channel is 0.6 mm×0.6 mm, only slightly larger than the diameter of the thermocouple 0.5 mm, which has almost negligible effect on the heat flow. In addition, the components exposed to the IR camera is coated with black paint to achieve an identical emissivity. 41.6K of traditional bond wired IGBT at a power density of about 210 W/cm2, which indicates that the advantages of double-sided cooling in thermal management, not only can decrease the average junction temperature, but also can greatly reduce the non-uniformity of the chip surface temperature. 3.3 Experimental result During the power cycling test, the voltage drop of DUT at the end of each cycle is kept constant after the thermal equilibrium is reached, which presents that each cycle is equivalence. Finally, six different measurement currents were chosen to perform the measurement using the proposed method, the results are shown in Table 1. (a) Three-dimensional temperature distribution Tab. 1: Measurement results of the multi-current VCE (T) method Measurement current (mA) Voltage drop (mV) Virtual junction temperature (℃) 100 46.22 96.2 75 45.37 96.5 50 44.43 96.7 20 43.33 97.0 10 41.12 97.1 5 40.63 97.6 (b) Two-dimensional temperature distribution Fig. 7: calculated chip temperature distribution These 6 pairs of I-V values are substituted into the nonlinear equations Eq. (1) for iterative solution, the calculated temperature distribution is shown in the Fig. 7, and the temperature distribution function is: T ( xˈ y ) 102.3 0.33˄ x 2 y 2˅ (3) The calculated maximum temperature at the chip center is 102.3 ഒ, which is awfully close to the 101.5 ഒ measured by the thermocouple, and the calculated minimum temperature at the edge is 85.1 ഒ, which is also very close to the 86.3 ഒ measured by the IR, as shown in Fig. 8. The agreement between the calculation results and experimental results show the effectiveness and accuracy of the proposed method. The difference between the maximum temperature and the minimum temperature is only 17.2K at a power density of 283.7 W/cm2, much smaller than the ISBN 978-3-8007-5515-8 Fig. 8: Temperature measured by the IR camera 4 conclusion In this paper, the multi-current VCE (T) method is applied to investigate the single chip temperature distribution within PPI, and a combination of the 1156 © VDE VERLAG GMBH · Berlin · Offenbach Authorized licensed use limited to: North China Electric Power University. Downloaded on March 02,2022 at 08:59:50 UTC from IEEE Xplore. Restrictions apply. PCIM Europe digital days 2021, 3 – 7 May 2021 infrared camera and thermocouple is adopted for experimental verification. The experimental results are agreed with the calculated results. Some conclusions can be drawn as follows: 1) The proposed multi-current VCE (T) method is suitable for single chip surface temperature distribution evaluation of PP IGBT during power cycling test, which extends the capabilities of conventional VCE (T) method in junction temperature measurement. 2) Compared with traditional bond wired IGBT, the double-sided cooling of the PP IGBT has advantages in thermal management under the same power density, it not only can decrease the average junction temperature, but also can greatly reduce the non-uniformity of the chip surface temperature. 5 Acknowledgements Measurement and Management Symposium, 2004. [6] ECPE/AQG 324, Qualification of Power Modules for Use in Power Electronics Converter Units (PCUs) in Motor Vehicles, 2018. [7] U. Scheuermann, R. Schmidt: Investigations on the VCE(T)-Method to Determine the Junction Temperature by Using the Chip Itself as Sensor, PCIM Europe 2009, 802-807, 2009. [8] E. Deng, Z. Zhao, Z. Lin, R. Han, and Y. Huang: Influence of temperature on the pressure distribution within press pack IGBTs, IEEE Trans. Power Electronics, 33(7), 6048–6059, 2018. [9] R. Bayerer, T. Herrmann, T. Licht, J. Lutz, M. Feller: Model for Power Cycling lifetime of IGBT Modules - various factors influencing lifetime, 5th International Conference on Integrated Power Electronics Systems, 2008. 6 References: [10] M. Junghaenel, U. Scheuermann: Impact of load pulse duration on power cycling lifetime of chip interconnection solder joints, Microelectronics Reliability, 76-77, 480-484, 2017. [1] E. Deng, Z. Zhao, P. Zhang, J. Li and Y. Huang: Study on the Method to Measure the Junction-toCase Thermal Resistance of Press-Pack IGBTs, IEEE Transactions on Power Electronics, 33(5), 4352-4361, 2018. [11] Y. Zhang, E. Deng, Z. Zhao, J. Chen, Y. Zhao: Sequential Vce(T) Method for the Accurate Measurement of Junction Temperature Distribution Within Press-Pack IGBTs, IEEE Transactions on Power Electronics, 36(4), 3735-3743, 2021. [2] E. Deng, J. Zhang, Z. Zhao, J. Chen, J. Li and Y. Huang: Influence of the clamping force on the power cycling lifetime reliability of press pack IGBT sub-module, The Journal of Engineering, 2019(6), 2435-2439, 2019. [12] J. Chen, E. Deng, L. Xie, X. Ying, Y. Huang: Investigations on Averaging Mechanisms of Virtual Junction Temperature Determined by VCE (T) Method for IGBTs, IEEE Transactions on Electron Devices, 67(3), 1106-1112, 2020. [3] L. Tinschert, A. Rygg Ardal, T. Poller, M. Bohlländer, M. Hernes, et al.: Possible failure modes in Press-Pack IGBTs, Microelectronics Reliability,55(6), 903-911, 2015. [13] J. Chen, E. Deng, X. Ying, Y. Huang: Temperature Distribution Evaluation of Single Chip by Multiple Currents VCE (T) Method, IEEE Transactions on Components, Packaging and Manufacturing Technology, 2020. (Early Access) This work was supported by the National Natural Science Foundation of China (52007061). [4] M. Held, P. Jacob, G. Nicoletti, P. Scacco, M.-H. Poech: Fast power cycling test of IGBT modules in traction application, Proceedings of Second International Conference on Power Electronics and Drive Systems, 425-430, 1997. [5] D.L. Blackburn. Temperature measurements of semiconductor devices - a review, Twentieth Annual IEEE Semiconductor Thermal ISBN 978-3-8007-5515-8 [14] E. Deng, Z. Zhao, P. Zhang, J. Li, Y. Huang: Study on the Method to Measure the Junction-to-Case Thermal Resistance of Press-Pack IGBTs, IEEE Transactions on Power Electronics, 33(5), 43524361, 2017. 1157 © VDE VERLAG GMBH · Berlin · Offenbach Authorized licensed use limited to: North China Electric Power University. Downloaded on March 02,2022 at 08:59:50 UTC from IEEE Xplore. Restrictions apply.