Facing high thermal loads on Power Modules in ...

PCIM 2010, 4 – 6 May 2010, Nuremberg, Germany

Paper 72

Facing high thermal loads on Power Modules in Hybrid Electrical Vehicles André Christmann, Infineon Technologies AG, Max-Planck-Straße 5, D-59581 Warstein, Germany Krzysztof Mainka, Infineon Technologies AG, Max-Planck-Straße 5, D-59581 Warstein, Germany

Abstract In automotive industry the IEC standard Q100/101 is used as a guideline for stress testing and reliability requirements. The launch of Hybrid Electrical Vehicles with high power semiconductor modules integrated in the vehicles lead to the application of IEC standard also to power modules. Suddenly the traditional module supplier was forced to fulfill significantly higher thermal demands when building modules for automotive applications. Nowadays, most of the module suppliers solved this issue, but the demanding automotive industry has started to increase the thermal requirements for the next generation of modules. This paper deals with the questions if the thermal shock test mentioned in the IEC Q101 standard is appropriate and if an increasing demand for thermal load can be addressed by simply increasing the required number of cycles (e.g. to 2000). As an example, three different types of modules are compared and it is shown that reaching a simple number in a stress test is not representative of the application itself.

1.

Introduction

HEV typically combines a combustion engine with an electrical drive train. To realize an operation with variable speed, an HEV is equipped with power semiconductors typically packaged in a module. In automotive industry the IEC standard Q100/101 is often used as guideline for stress testing and reliability requirements of components which are mounted into a vehicle. Therefore power module manufactures for automotive applications are forced to fulfill the thermal requirements given in this standard (namely 1000 thermal shock tests compared to 50 cycles for industrial modules). The more appropriate approach to forecast the life time and to derive reliability requirements is the modeling of the response of an inverter system to a given mission profile. The life time of a component is defined by the reliability demands due to its operating environment and to the operating conditions. In power electronics the component temperature and temperature changes have major influences on the reliability. Therefore, a procedure based on power loss calculation and thermal modeling was developed [6], which computes the temperature over a whole driving cycle. With the aid of these calculations the active thermal stress on various joints such as solder or bond joints can be evaluated. By transforming the thermal stress into reliability test data a prediction of lifetime is possible.

While taking into account the response of a module to the application conditions, the result of a life time calculation is certain “number of cycles” (for each reliability test) which is only representative for the individual module. Only if the module is capable to reach this required “number of cycles” (proven in a reliability test), the module fits to the application. In the next chapters life time calculations for 3 types of power modules are described for the FTP 72 mission profile. It will be shown that the use of a base plate as a part of a power module significantly reduces the power cycling demands on the module. One further result is that the cooling capability of a module influences the number of equivalent test cycles (= lifetime). One consequence is that the suitability of a power semiconductor module for a given application can not be proven by comparing a simple number resulting from a reliability test without performing a life time calculation!

2.

A single number is not enough

Since the resulting equivalent required number of cycles of a life time calculation differs for the evaluated module types, it may happen that the better performing module fulfills the application requirements but the other module does not, even if both modules reach the same number in a reliability test.

ISBN 978-3-8007-3229-6 © VDE VERLAG GMBH ∙ Berlin ∙ Offenbach

432


2.1.

How module construction influences of the required cycle number

In a typical power module failures due to high thermal loads are often claimed to be to failures in the interconnection of the ceramic to the base plate. In the following sections two different approaches to overcome this weakness and the consequences for reliability testing are discussed.

Removal of the weak interconnection The most simple way to solve a problem with an interconnection layer is to remove the interconnection. The removal of the solder joint between ceramic and base plate leads therefore to modules without a base plate. Unfortunately together with the removal of the interconnection layer the benefits of the base plate are excluded. A common way to mount a module without a base plate is to screw it onto the cooler and to use thermal grease. The use of thermal grease leads to an increase of the thermal resistance Rth even if the grease applied is thin (here assumed grease thickness is 50µm). A second approach for reducing thermal stress is the use of an optimized thermal path with a minimized thermal resistance.

Paper 72

module type is directly mounted on an open liquid cooler with direct contact of the pins and the cooling medium. Therefore the thermal grease with its poor thermal conductivity can be removed. Optimization of Zth: The presence of the base plate leads to an increase of the heat conductive area through thermal spreading and therefore to a lower thermal resistance Rth, ja between junction and ambient (coolant). Figure 2 shows a comparison of the transient thermal impedance junction to ambient curves of the 3 types of modules which are described in chapter 4.1. Since a significant number of the pulses in an inverter application are lying in a typical time frame above 0,1 sec the superior performance (factor 1,6 – 1,8) of the base plate module types is obvious.

Optimization of the cooling performance Two kinds of optimizations are implemented in a direct cooled module. Fig. 2. Comparison of the transient thermal impedance curves [junction to ambient (coolant)] for a direct cooled PIN FIN module with liquid cooled modules (with and w/o base plate) using thermal grease. The junction temperature swing for single power pulse is directly proportional to the Zth, ja(t). Therefore a given electrical power dissipation profile P(t) leads to different ∆T between coolant and junction. The junction temperature plot vs. time can be calculated as: t

∆T (t ) = ∫ P(τ )Z´′th , ja (t − τ )dτ

(1)

0

Fig. 1. Example of a PIN FIN structured basepTM late (HybridPACK 2) [14][15] in comparison to a flat baseplate Optimization of Rth: In order to get an ideal heat transfer from the module to the cooler a PIN FIN structured copper baseplate was used. This

Since the failure mechanism bond wire lift off is directly related to temperature swings the number of required cycles representing a certain lifetime in a power cycling test has to be adjusted to the application as a consequence. The result of this adjustment is shown in chapter 5.


433


3.

From drive cycle to reliability test

3.1.

Reliability tests

During the life time a module is exposed to passive temperature swings coming from the environment (climate) and active temperature cycles generated by operation. Temperature cycling and power cycling tests represent these conditions. Temperature Cycling: In temperature cycling tests the temperature of a power module is changed by variation of the ambient temperature (TST: Thermal Shock Test) or the temperature of the case (TC: Thermal Cycle Test) without electrical stressing. This test is applied mainly to evaluate the lifetime of the solder joints and to evaluate the resistance to the sudden changes in temperature the device can experience during storage, transportation or in use. Power Cycling: Power cycling (PC) tests are used to determine the resistance of a semiconductor device to thermal and mechanical stresses due to cycling the power dissipation of the internal semiconductor die and internal connectors. This happens when load currents are periodically applied and removed causing rapid changes of temperature. The power cycling test is representative to high temperature operating life [1]. The predominant failure mechanisms due to thermal stress are the degradation of solder layers (solder fatigue) and the bond wire lift-off.

3.2.

Life time modelling

Fig. 3 explains the procedure how the representing number of test cycles for a reliability test can be derived from the information of the inverter system (cooling conditions) and the driving strategy (mission profile, motor and drive control) for a given set of electrical parameters (electrical characteristics of the power module)[8].

Fig. 3. General approach for calculating the number of equivalent tests cycles. Only the parameters marked red were varied in investigation

Paper 72

In order to get independent from drive conditions, motor characteristics as well as chip characteristics a common set of input parameters was chosen.

Mission profile The drive cycle FTP 72 in combination with a standard electrical motor was used to derive a common load for each module type.

Electrical parameters The electrical parameters are based on the HyTM bridPACK 1 module FS400R07A1E3 (400A / 650V rating) [10].

Thermal model For calculating the power module reliability, it is necessary to know transient temperatures of semiconductor junctions, as well as the system solder temperature. The solder joint temperature is necessary for the calculation of the solder fatigue due the thermal cycling. For transient thermal impedance entries approximation, the Foster RC-networks was chosen [11]. The values of the R’s and C’s were derived by 3D transient finite element simulation using the system’s material properties and physical dimensions or they were extracted from measurement.

Temperature profile simulation With the aid of the thermal model the temperature of the IGBT, the diode and the system solder can be calculated for the load conditions of a given driving cycle. The temperature itself is not the major parameter which is responsible for solder and bond wire ageing but rather the temperature swings. Therefore an automatic algorithm is implemented in the simulation to extract the temperature cycles ∆T, maximum junction temperature and pulse duration time [ton].

Determination of ∆T occurrence Active cycles: Fig. 4 shows the number of occurrences for a certain temperature swing at the diode for a PIN FIN module and a module without a base plate. Temperature ranges below 4 K are neglected since they do not decrease lifetime significantly. Most of the swings lead to an increase of the junction temperature below 30K. Since a significant number of the load cycles lasted longer than 1 second, a shift from low ∆T to higher ∆T can be observed for the module without a base plate. The difference in the highest occurring ∆T (PIN FIN: 40°C – w/o base plate: 73°C) represents a several second lasting recuperation cycle given by the driving strategy for a FTP 72 cycle.


434


Fig. 4. Comparison of the number of active cycles for different ∆T between a PIN FIN module and a module without a base plate for one FTP 72 driving cycle Passive cycles: Overlaid to these active swings are always passive swings due to the operational environmental. The heating up of a cooling system during operation also leads to a temperature swing which has to be considered during life time calculation. Assuming a vehicle lifetime of 15 years and two cycles per day 10950 additionally cycles are applied to a power module. The environmental temperature was defined in table 1 reaching from 5 days with -25°C up to 35 days with 30°C outdoor temperature.

Fig. 5. Passive temperature swings due to heating up the cooling system from environmental temperature to operation temperature The temperature swing of the heating up sequence was always defined as the difference between the maximum temperature during the driving cycle and the minimum temperature at the start equal to the environmental temperature (see Fig. 5). A reliability test applying different temperature swings to a device is rather impractical. Therefore standardization to a common ∆T has to be done.

3.3.

Transformation from duty cycles to test cycles

Paper 72

to stress cycles or change in temperatures. A model of this type known as the (modified) Coffin-Manson model has been used successfully to model crack growth in solder and other metals due to repeated temperature cycling as equipment is turned on and off. The form of this frequently cited equation makes it clear that fatigue will result in much earlier failure when the joint experiences wider temperature excursions. The most useful derivative of this equation is probably the relationship between the number of cycles to failure with two different thermal ranges (∆Tduty_cycle and ∆Ttest) [12]. Although different exponents have been mentioned in the literature, the calculations that have been done use an exponent of 3,3. This model takes the form

nduty _ cycle ntest _ cycle

 ∆Ttest _ cycle   =  ∆T   duty _ cycle 

[2 - 4,5 ] (2)

The equivalent number of test cycles ntest_cycle for a specified test temperature range ∆Ttest can be calculated from the number nduty_cycle of duty cycles with a given ∆Tduty_cycle coming from the temperature profile.

Wire bond acceleration calculation The relation between an arbitrary parameter set under duty condition (current I, junction temperature Tj, operation time ton and temperature swing ∆T) and the number of equivalent cycles for a known reliability test setup is given in equation (3). Although different exponents have been mentioned in the literature, the calculations that have been done in this paper use an exponent of 4,3.

nduty _ cycle ntest _ cycle

 ∆T  =  test _ cycle   ∆T   duty _ cycle 

[2 , 5 − 5 ]

o cor

(3)

This formula also contains the ratio of the different temperature differences but it has been modified due to results of a large number of performed tests [13]. Based on equation (3) the number of equivalent test cycles (conditions: ∆Ttest=100K, Tj,max=150°C , ton, test = 2s and reference current of Itest = 400A) results from the summation of all p transformation for any duty cycle i.

4.

Variation of parameters

Solder joint acceleration calculation

4.1.

Module design

Models for mechanical failure, material fatigue or material deformation typically have terms relating

Three types of modules were compared:


435


1. Direct cooled PIN FIN module: The base plate is directly submerged into the coolant (see Fig. 1 and Fig. 6). 2. Module with flat base plate mounted on a cooler with the aid of thermal grease. 3. Module without base plate mounted on a cooler with the aid of thermal grease.

Paper 72

Power Cycling: For the failure mechanism bond wire lift off the maximum temperature of the bond wire is set to the maximum chip temperature Tj_max. derived from the drive cycle calculation. The life cycle modeling allows the number of equivalent power cycles for passive and active cycles to be calculated. That implies that the highest occurring temperature swing is assumed in the passive load range from starting temperature (if required from minus degrees) to highest occurring chip temperature during one (active) drive cycle. Thermal Cycling: A similar procedure was applied while transforming passive and active cycles as described in chapter 3.3. Maximum temperatures for the solder were also derived from the drive cycle calculation.

5.1.

Fig. 6. Example of a PIN FIN structured baseplate in comparison to a module without a baseplate using thermal grease.

4.2.

Cooling conditions

One liquid cooler and one direct cooled system were compared. In all cases the coolant temperature was set to 70°C. For the liquid cooled system a thermal grease layer between the flat ceramic and the flat cooler was assumed (thickness 50µm). The heat transfer coefficient α applied to the rear of the heat sink was assumed as α = 20000 W/m²K (strong liquid cooler). The PIN FIN module type is directly mounted on an open liquid cooler with direct contact between the pins and the cooling medium. Due to this direct contact the value α is not defined. In this case the flow rate of the liquid (10l/min) was chosen to represent an equivalent cooling capability.

5.

Results

As shown in Fig. 3, the active swings generated during the drive cycle and the passive swings due to the operational environmental have to be considered. For the silicon, the worst case condition of IGBT and diode has to be considered. For the given mission profile FTP 72 and the assumed driving strategy the highest load was on the diode. Therefore the shown examples represent the diode.

Calculated cycle numbers

Fig. 7 and Fig. 8 show a comparison of equivalent test cycles for the different module types for a given FTP 72 drive cycle representing 15 years lifetime. Power cycling: In Fig. 7 the number of equivalent test cycles (conditions: ∆Ttest=100K, Tj, test=150°C , ton, test = 2s and reference current of Itest = 400A) for a power cycling test is given as the sum of cycles generated by active operation and the number of cycles representing the passive swings (climate load). .

Fig. 7. Number of equivalent power cycles for three module types for a given FTP 72 drive cycle representing 15 years lifetime. Thermal cycling: In Fig. 8 the number of equivalent test cycles (conditions: ∆T = 80K) for a thermal cycling test are given as the sum of cycles generated by active operation and the number of cycles representing the passive swings. Since a module without a base plate does not have an interconnection between base plate and DBC the values for thermal stressing of this interconnection are obvious not available. (Of course this does not mean that a module without a base plate always passes the relevant thermal stress test but it will definitely have a different failure mode.)


436


Fig. 8. Number of equivalent thermal cycles for different parameters for a given FTP 72 drive cycle representing 15 years lifetime. The influence of the active cycles can be neglected. This is justified because of the very low temperature swing of the solder during operation compared to the high ∆T of the passive swings.

5.2.

Statements

Even the two reliability tests can not be compared although the trend for both test are similar. This is because in both cases a higher ∆T leads to a higher number of equivalent test cycles. 1) A better cooling capability (PIN FIN) leads to lower reliability requirements. (Of course, such a trivial statement can be made by everyone – the aim of this paper is to make clear how much the cooling capability influences the reliability requirements.) For thermal cycling the number of required cycles is for the PIN FIN case is only 66% that of the case using a module with base plate and thermal grease (Fig. 8). 2) The use of a base plate reduce significant the demands for power cycling capability on a module. A module without a base plate needs to fulfil the requirement for a 600% higher number of power cycles compared to a PIN FIN module (Fig. 7). 3) Since the resulting equivalent required number of cycles differs for all types it may happen that the better performing module fulfils the application requirements but the other module does not, even if both modules reach the same number of cycles in a reliability test.

Paper 72

Fig. 9. Comparison of suitability of the given module types for assumed reliability capability compared to number of equivalent power/thermal cycles for a given drive cycle Since the equivalent calculated number of test cycles differs significant in this case only the PIN FIN module fulfils all requirements at the same time!

6.

Conclusion

Simple numbers gathered from reliability testing do not represent the requirements of the application. It was shown that the equivalent number of thermal cycles during a validation test is strongly dependent of the module design and of the cooling capability. Facing higher loads in the future therefore does not directly lead to an increase of cycle numbers if better cooling performance can compensate the stress. Especially for a TST test the increase from 1000 TST to e. g. 2000 cycles is a rather impractical approach (> 6 month physical test time). Since the technology of power semiconductor modules is already reaching the physical limits a further acceleration by increasing the temperature seems not to be feasible and a changed test method is needed [16]. Activities with the aim of adapting the Q101 procedure which was derived for discrete components to be applicable to a complex system like a power module driven by the ECPE [17] and the ZVEI Fig. 3[18] are highly appreciated. One way for supporting such activities from a module manufacture side can for example be to provide combination of different tests in one common test, like a test “Power Cycling of IGBTModules with superimposed thermal cycles”, which was presented last year on PCIM [19].

In table 2 an example is given in which all three module types have the same arbitrary assumed PC and TC capability here given as a simple number.


437


7.

Remark

Some remaining correlation of variables used in the model restricts the model to ranges of test conditions of selected data. Therefore, the author strongly recommends not applying the model without consulting experts at Infineon Technologies. Furthermore some of the showed tests results may differ to results of released products due to unmentioned changes in material or processes.

8.

Literature

[1] IEC 60749-34, Ed. 1.0, 2004-03; IEC607479, 1998 [2] Graf, I.: Reliable IGBT Power Semiconductor Modules for Hybrid Electrical Vehicles, Bodo´s Power, 08-07, p. 22-24, 2008. [3] Münzer, M.; Thoben, M.; Christmann, A.; Vetter, A.; Specht, B.: Halbleiter für das „grüne“ Auto, Elektronik Praxis, 2007 [4] Münzer, M.; Thoben, M.; Vetter, A.; Christmann, A.; Ferber, G.: Suitability of Power Semiconductor Modules for HEV Applications, Automotive Power Electronics, 2006 [5] www.infineon.com/hybridpack [6] Thoben, M. et al.: From vehicle drive cycle to reliability testing of Power Modules for hybrid vehicle inverter, PCIM Europe 2008, Nuremberg, 2008. [7] Kanschat, P.; Rüthing, H.; Umbach, F.; Hille, F.: 600V-IGBT³: A detailed Analysis of Outstanding Static and Dynamic Properties, PCIM Europe 2004, Nuremberg, 2004. [8] Christmann, A. et all: Reliability of Power Modules in Hybrid Vehicles, PCIM Europe 2009, Nuremberg, 2009 [9] Mainka, K.; Aurich, J.; Hornkamp, M.: Fast and reliable average IGBT simulation model with heat transfer emphasis, PCIM Europe 2006, Nuremberg, 2006. [10] Datasheet FS400R07A1E3, www.infineon.com [11] Lutz, J.: Halbleiter-Leistungsbauelemente, Springer-Verlag, 2006. [12] Karya, Y.; Otsuka, M.: Mechanical fatigue characteristics of SnAgX solder alloys, Journal of Electronic materials, volume 27-11, p. 1229-1235, 1998. [13] Bayerer, R. et al.: Model for Power Cycling lifetime of IGBT Modules – various factors influencing lifetime, CIPS 2008, Nuremberg, March 2008.

Paper 72

[14] Münzer, M.; Thoben, M.; Christmann, A.; Vetter, A.; Specht, B.: Application Focused Power Semiconductor Module Design for Hybrid Electric Vehicles, Power Electronics Europe, Issue 2, 2007 [15] Volke, A.; Bo, Z.; Christmann, A.; Schlörke, R.: Reliable IGBT Modules for (Hybrid) Electric Vehicles, PCIM, China, 2007 [16] VDE: KFZ-Anforderungen an ElektronikBauelemente, VDE Positionspapier, 2008 [17] European Center for Power Electronics: Power semiconductor Robustness, ECPE Workshop Munich, 2009 [18] ZVEI - Zentralverband der Elektrotechnikund Elektronikindustrie e.V.: Robustness Validation Handbook, 2008 [19] Marco Feller, J. Lutz, R. Bayerer: Power Cycling of IGBT- Modules with superimposed thermal cycles, PCIM Europe 2009, Nuremberg, 2009


438