research article online junction temperature cycle...

Research ArticleOnline Junction Temperature Cycle Recording ofan IGBT Power Module in a Hybrid Car

Marco Denk and Mark-M. Bakran

Department of Mechatronics, University of Bayreuth, Universitatsstraße 30, 95447 Bayreuth, Germany

Correspondence should be addressed to Marco Denk; [email protected]

Received 14 July 2014; Accepted 16 January 2015

Academic Editor: Pavol Bauer

Copyright © 2015 M. Denk and M.-M. Bakran.This is an open access article distributed under the Creative Commons AttributionLicense, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properlycited.

The accuracy of the lifetime calculation approach of IGBT power modules used in hybrid-electric powertrains suffers greatly fromthe inaccurate knowledge of application typical load-profiles. To verify the theoretical load-profiles with data from the field thispaper presents a concept to record all junction temperature cycles of an IGBTpowermodule during its operation in a test vehicle. Forthis purpose the IGBT junction temperature ismeasuredwith amodified gate driver that determines the temperature sensitive IGBTinternal gate resistor by superimposing the negative gate voltage with a high-frequency identification signal. An integrated controlunit manages the 𝑇𝐽 measurement during the regular switching operation, the exchange of data with the system controller, andthe automatic calibration of the sensor system. To calculate and store temperature cycles on a microcontroller an online Rainflowcounting algorithmwas developed.The special feature of this algorithm is a very accurate extraction of lifetime relevant informationwith a significantly reduced calculation and storage effort. Until now the recording concept could be realized and tested within alaboratory voltage source inverter. Currently the IGBT driver with integrated junction temperature measurement and the onlinecycle recording algorithm is integrated in the voltage source inverter of first test vehicles. Such research will provide representativeload-profiles to verify and optimize the theoretical load-profiles used in today’s lifetime calculation.

1. Introduction

The combination of an internal combustion engine and anelectric machine enables the improvement of the efficiencyand the performance of the drivetrain of personal cars,busses, and utility vehicles [1]. In view of the reliability andthe lifetime of voltage source inverters used in hybrid-electricpowertrains the IGBT power module can be considered asthe most lifetime critical component. This is especially trueif power modules with conventional linking and packagingtechnology are used. Those modules are characterized bya bond-wire connection, a direct copper bonded Al2O3-substrate where the chip is soldered on, and a copper base-plate. This results in a complex structure whose materialshave different coefficients of thermal expansion CTE. In caseof temperature cycles this CTE mismatch causes thermo-mechanical stresses in the modules interconnections andleads to the lift-off or the heel-cracking of bond wires or thedegradation of the die-attach or the substrate solder joint [2].

To estimate the lifetime of an IGBT power module in ahybrid car a simple lifetime calculation approach has becomedominant in recent years [3]. This calculation approach isderived from the lifetime estimation of mechanical parts anddemands the linkage of an application typical load-profilewith a lifetime model of the IGBT power module using acycle counting algorithm and a linear damage accumulationrule. For mechanical parts like shafts or gearwheels intransmissions this lifetime calculation approach could beverified over the years and today it is possible to design theirlifetime with a high accuracy. On the contrary the lifetimecalculation of IGBT power modules in hybrid cars is in avery early stage and currently it is not possible to quantifythe accuracy of the lifetime calculation approach. What isknown, however, is that the calculation approach suffersfrom different factors of uncertainty like the interaction ofdifferent failuremechanisms [4] and the information loss dueto the cycle counting. However, the most critical point inlifetime calculation is the limited representation accuracy of

Hindawi Publishing CorporationAdvances in Power ElectronicsVolume 2015, Article ID 652389, 14 pageshttp://dx.doi.org/10.1155/2015/652389

2 Advances in Power Electronics

today’s load-profiles.These theoretical profiles originate fromsimulation, but it is hard to consider different types of driver,different areas of operation, different hybrid strategies, andvarying ambient conditions in a load-profile that is rathershort in relation to the vehicle lifetime. Because of theseuncertainties, there is a need to verify the theoretical load-profiles and the lifetime calculation approach with data fromfield studies. For this reason this paper presents a temperaturecycle recorder that can be implemented in hybrid cars torecord the exposure of the IGBT power module during itsreal operation. The recorded load history of test vehicles orfirst field returns can be used to create an experience baseand to optimize the load-profiles and the lifetime calculationapproach of IGBT power modules. This paper shows newresults of the IGBT driver and the recording algorithm andcombines it with the results of the following publications [5–7]. In the following the state-of-the-art approach to calculatethe lifetime of an IGBT power module is briefly summarized.

2. State of the Art

The lifetime calculation of an IGBT power module in ahybrid car requires basically the linkage of an applicationtypical load-profile with an empirical lifetime model of thepower module using a cycle counting algorithm. Today thisload-profile originates from an application typical velocityprofile like the New European Driving Cycle (NEDC) andthe simulation of the entire hybrid-electric powertrain. Asubsequent electrothermal model of the powermodule deliv-ers the transient IGBT junction temperature over a certaintimespan.This transient temperature curve is called the load-profile of the power module. It is analyzed with a cyclecounting algorithm and valued with an empirical lifetimemodel. In recent years the following lifetime models andcounting algorithms have been presented.

2.1. Empirical Lifetime Models. Empirical lifetime modelsoriginate from the accelerated ageing of powermodules.Theyspecify the number of temperature cycles a powermodule canbear until a failure criterion is reached. In recent years variouslifetime models were publicized that differ primarily in thenumber of parameters used to describe a temperature cycle.The elementary lifetime model is a simple Coffin-Mansonlaw [8] that states that the number of temperature cycles tofailure𝑁𝑓 depends solely on the size of the amplitude Δ𝑇𝐽 ofa temperature cycle:

𝑁𝑓 ∼ Δ𝑁𝛼

𝐽. (1)

Today there are extended lifetime models [9–11] which con-sider additional parameters to describe a temperature cycle.In 1997 the LESIT [9] project investigated the temperaturecycle stability of powermodules with conventional packagingtechnology from European and Japanese suppliers. It wasfound that themedium cycle temperature𝑇𝐽,med has a notableinfluence on the sustainable number of cycles. For this reasonthe Coffin-Manson law was extended by an Arrhenius term.Equation (2) shows the LESIT model, where the number ofcycles to failure 𝑁𝑓 is a function of the cycle amplitude Δ𝑇𝐽

and the medium cycle temperature 𝑇𝐽,med. The parameters𝐴 = 640 and 𝛼 = −5 were derived from accelerated ageingand 𝑅 = 8.314 J/mol⋅K is the gas constant and 𝑄 = 7.8 ⋅

104 J⋅mol−1 is the activation energy:

𝑁𝑓 = 𝐴 ⋅ Δ𝑇𝛼

𝐽⋅ exp( 𝑄

𝑅 ⋅ 𝑇𝐽,med) . (2)

Since the technologies of conventional IGBT power moduleshave been improved, in 2008 the number of sustainabletemperature cycles to failure was reinvestigated by powercycling of several Infineon IGBTmodules. It became apparentthat many additional parameters have an impact on themodule lifetime. The developed CIPS08 [10] lifetime modeldescribes the number of cycles to failure 𝑁𝑓 as a functionof the amplitude Δ𝑇𝐽, the minimum temperature 𝑇𝐽,min =

𝑇𝐽,med − (1/2) ⋅ Δ𝑇𝐽, and the heating time 𝑡on of a temperaturecycle:

𝑁𝑓 = 𝐴 ⋅ Δ𝑇𝛽1

𝐽⋅ exp(

𝛽2

𝑇𝐽,min) ⋅ 𝑡𝛽3

on ⋅ 𝐼𝛽4

𝐵⋅ 𝑉𝛽5

𝐶⋅ 𝐷𝛽6 . (3)

Moreover the current per bond wire 𝐼𝐵, the nominal voltage𝑉𝐶, and the bond wire diameter 𝐷 were taken into account.The parameters 𝐴 and 𝛽1 to 𝛽6 and their validity rangesare given in [12]. For instance the heating time 𝑡on of atemperature cycle must be set to 𝑡on = 15 s for 𝑡on > 15 s.Equation (3) shows the CIPS08 lifetime model. For lifetimecalculation the current per bond wire can be set to 𝐼𝐵 =

10A. The diameter of the bond wire 𝐷 and the voltage classof the power module 𝑉𝐶 are constants, so that the lifetimeof the power module depends solely on the temperaturecycles the power module is exposed to during its operation.This comparison of different lifetime models shows that theaccuracy of the empiricalmodels used for lifetime calculationcould be improved due to the more accurate specificationof a temperature cycle. The present state of the art is theparameterization of a temperature cycle with its amplitudeΔ𝑇𝐽, itsminimum temperature𝑇𝐽,min, and its heating time 𝑡on.

2.2. Cycle Counting Algorithm. Counting algorithms enablethe evaluation of an application typical load-profile thatconsists of several different temperature cycles with anempirical lifetime model. For this purpose they extract andparameterize all temperature cycles within the load-profileand store them in a data vector. Widely accepted countingmethods are the half-cycle counting, the maximum-edgecounting, and the Rainflow counting [13, 14]. Figure 1 showsthe application of these counting algorithms on an exemplarytemperature profile.

In the half-cycle counting all rising and falling edgeswere counted as half temperature cycles. Their amplitudesare calculated as differences of two consecutive extremevalues. The minimum temperature of each cycle equals thesmallest cycle temperature. This applies to all mentionedalgorithms. Half-cycles with rising edges consist of a heatingtime, which is the time span between their occurrences. Thecooling time of half-cycleswith falling edges is not consideredin the reviewed lifetime models, so that these cycles do

Advances in Power Electronics 3

1

2

3

4

56

T(∘ C)

t (s)

(a)

1

2

3

T(∘ C)

t (s)

(b)

1

23

T(∘ C)

t (s)

(c)

Figure 1: Established cycle counting algorithms: (a) half-cycle, (b) maximum-edge, and (c) Rainflow counting applied to an exemplarytemperature profile.

9

9

25193114226

6

25

19

31

14

22

Vector

Win

dow

TJ

(∘ C)

t (s)

ΔT1

ΔT2

ΔT3

E1

E2

E3

Figure 2: Working principle of a conventional range-countingRainflow algorithm that scans a one-dimensional extreme valuevector.

not have time stamps. Maximum-edge counting means theinterpretation of the major edge of three consecutive extremevalues as full temperature cycle. The heating time equals thetime span between the occurrences of the first and the secondextreme value.

The unique feature of the Rainflow method is thecounting of closed temperature cycles, which appear asclosed hysteresis curves in the stress-strain diagram [15].This physical background distinguishes the Rainflowmethodfrom the previous described algorithms. For this reason theRainflow algorithm became the state of the art in the lifetimecalculation of power modules. The principle of the Rainflowmethod is standardized by ASTM E-1049 [16].

To calculate closed temperature cycles with a Rainflowalgorithm the extreme values of the entire load-profile mustbe known. Today there are different versions of Rainflowalgorithms, like fast Rainflow, range-counting, or graphicalalgorithms [17]. In the following the working principle of asimple range-counting Rainflow should be outlined. Figure 2shows an exemplary temperature profile, whose extremevalues are stored in a one-dimensional vector that is scannedby a moving window. A closed temperature cycle exists whenthe absolute difference of the extreme values 𝐸2 and 𝐸3 isgreater or equals the absolute difference of the extreme values𝐸1 and 𝐸2. Otherwise, the window has to be shifted. At thestarting position the extreme values in the window do notmeet the cycle condition so that the window is shifted by one

extreme value. The new values in the window fulfill the cyclecondition and the temperature cycle Δ𝑇1 = 𝐸1 − 𝐸2 = 6∘C iscalculated. Finally the extreme values𝐸1 and 𝐸2were deletedand the scanning window is shifted back to the beginning ofthe vector.

Figure 3 shows the complete algorithm to convert atransient temperature profile into temperature cycles that areparameterized with their amplitudes Δ𝑇𝐽, their minimumtemperatures 𝑇𝐽,min, and their heating times 𝑡on. Firstly asimple three-point algorithm determines all extreme valuesof the load-profile and stores them with a timestamp in atwo-column vector. Once all extreme values of the load-profile were determined, the vector is scanned by the range-counting Rainflow algorithm in search of full temperaturecycles. Whenever a closed temperature cycle is found it isparameterized and stored in a three-column data vector.

On the basis of the stored temperature cycles the lifetime𝐿 of the power module can be estimated according to (4),where 𝑁𝑓𝑖 is the number of sustainable temperature cycleswith a certain amplitude Δ𝑇𝐽, minimum temperature 𝑇𝐽,min,and heating time 𝑡on and 𝑛𝑍𝑖 is the number of cycles withsimilar parameters that were found in the load-profile. Inthe case of a vector storage each temperature cycle has to beconsidered with 𝑛𝑍𝑖 = 1. The lifetime 𝐿 of the power moduleresults from the linear accumulation [17] of the damage of alltemperature cycles 𝑁𝑍 and the scaling with the duration 𝑡LPof the application typical load-profile. Consider

𝐿 = (

𝑁𝑍

∑

𝑖=1

𝑛𝑍𝑖

𝑁𝑓𝑖 (Δ𝑇𝐽, 𝑇𝐽,min, 𝑡on))

−1

⋅ 𝑡LP. (4)

An exemplary load-profile of a hybrid car with a durationof 𝑡LP = 45 minutes incorporates 𝑁𝐸 = 5487 extremevalues that were further processed in a Rainflow algorithmto 𝑁𝑍 = ∑𝑛𝑍𝑖 = 2743 closed temperature cycles. Theevaluation of this temperature cycles with theCIPS08 lifetimemodel results in a lifetime of the power module of 𝐿 =

8016 operating hours, that is, about one-year continuousoperation. For this reason test vehicles enable the verificationof a great part of the load-profile within a short period oftime. It would even be possible to verify the entire lifetimecalculation approach with field studies.


Name Size

Extreme value detection algorithm algorithm

Temperature cycle storage vectors

ASTM-Rainflow

TJ

(t)

TJ

(t) ΔTJ

TJ,min

ton

1 × 2743

1 × 2743

1 × 2743

Figure 3: Conventional algorithm used on desktop computers to determine and store the temperature cycles that are incorporated into theload-profile.

System controller

measurement

Online junction

Online cycle recording algorithm

IGBT driver

temperature TJ

TJ (t) np

nz ton

ΔTJ TJ,min

Modified

Figure 4:The temperature cycle recorder consists of an online cyclecounting algorithm that is implemented on the system controllerand a modified IGBT driver with integrated online junction tem-perature measurement.

3. Temperature Cycle Recorder

The objective of this work is the verification of the state of theart to calculate the lifetime of an IGBT power module usedin hybrid-electric powertrains. Therefore the temperaturecycles of the power module should be determined during theoperation in the field and stored on the system controllerof the voltage source inverter. Basically this requires themeasurement of the IGBT junction temperature during theregular inverter operation and the online temperature cyclecalculation and storage on the system controller or the IGBTdriver. Figure 4 shows the block diagram of the recordingconcept. The junction temperature is measured with a mod-ified IGBT driver and transmitted to the system controllerwith an optical fiber. The online cycle recording algorithmcalculates closed temperature cycles and stores them in acompact data matrix. In the following the IGBT driver withintegrated junction temperaturemeasurement and the onlinecycle recording algorithm are presented in detail.

3.1. IGBT Driver with 𝑇𝐽 Measurement. Due to experiencingcurve effects voltage source inverters used within hybridcars contain standard IGBT power modules with conven-tional linking and packaging technology. The measurementof the IGBT junction temperature of those conventionalpower modules during their operation in a way that issuitable for series production is a very challenging task.Today’s sensor concepts are based on either the installation

of temperature sensors on the chip surfaces [18] or theutilization of temperature sensitive electrical parameters ofthe IGBT. In the laboratory the most popular temperaturesensitive parameter is the saturation voltage 𝑈𝐶𝐸 = 𝑓(𝐼𝐶, 𝑇𝐽)

that is measured at a constant collector current of 𝐼𝐶 =

100mA. In this case the saturation voltage has a temperaturesensitivity of approximately 2.3mV/K [19]. To realize the𝑈𝐶𝐸-method during inverter operation a switching sequencewith attached measurement phase could be used [20, 21], butthis also results in an interruption of the motor current anddeteriorates the systemproperties. In [22, 23] the temperaturesensitive quasi-threshold voltage is used for junction tem-perature measurement. Thereby the induced voltage acrossthe parasitic inductance between the power emitter andan auxiliary emitter is used to trigger the measurementof the gate emitter voltage. However, the method requiresan increased measurement and calibration effort and canbe considered to be very noise sensitive. The feasibility ofother temperature sensitive parameters for online junctiontemperature measurement, such as switching times [24], thecurrent slope [25], or the width of the miller plateau [26],is rather low. Reasons for this are their small temperaturesensitivity in the range of about 1 ns/K, their poor selectivity,the limited resolution of affordable sensors, and the need toadd subsequent compensating procedures.

A further temperature sensitive parameter that can befound in conventional power modules, where each IGBT isbuilt up of several paralleled IGBT single chips, is the internalgate resistor 𝑅Gi. Often this resistor is located directly underthe gate bond in the center of each chip and therefore closesto the junction of the semiconductors. To use this resistor astemperature sensor in [27] a modified IGBT single chip withdouble-sided bond connection of the internal gate resistor ispresented. This allows the impression of a small DC sensecurrent and the determination of the temperature sensitiveinternal gate resistance by measuring the voltage drop acrossthe 𝑅Gi. Although the internal gate resistor was found to bewell suited for junction temperature measurement the sensorsystem is inappropriate for paralleled chips and nonmodified,conventional IGBT power modules.

In summary, today there is no sensor concept suitable forseries production to measure the junction temperature of aconventional power module during the real inverter opera-tion. Existing solutions either are based on the modification


Terminals

LpG

E

C

UGACURG

RG

Cp

IGAC RGi

Figure 5: AC equivalent gate circuit of an IGBT powermodule con-sisting of the internal gate resistor 𝑅Gi and the parasitic parameters𝐿𝑝 and 𝐶𝑝.

UGoff

toff ton t (𝜇s)

UGon

U (V)

uGAC (t)

Figure 6: The negative gate voltage 𝑈Goff is superimposed with thehigh-frequency identification signal 𝑢GAC(𝑡) during the IGBT is inoff-state.

of the power module or require a large intervention in thepower electronic system. In the following an IGBT driverconcept with integrated junction temperature measurement[5] is presented that is easy to apply, does not affect the realinverter operation, and is applicable to conventional IGBTpower modules.

3.1.1.MeasurementMethod. To realize an easy to apply sensorsystem it is advantageous to integrate the junction tempera-ture measurement in the gate circuit or in the IGBT driver,respectively. It can be seen in Figure 5 that the temperaturesensitive internal gate resistor 𝑅Gi is in series with the para-sitic capacitance𝐶𝑝 of the IGBT, the inductance𝐿𝑝 of the gateconnection, and the external gate resistor 𝑅𝐺. To determinethe resistance of 𝑅Gi utilizing the available terminal elementsof a conventional power module the negative gate voltage issuperimposed with the sinusoidal identification signal𝑈GAC.During the feeding in period the voltage drop 𝑈RG acrossthe external gate resistor 𝑅𝐺 is measured. To ensure a highmeasurement accuracy an external gate resistor with a lowtemperature dependency has to be used.

Figure 6 sketches a short section of the gate voltage,where the negative voltage 𝑈Goff is superimposed with thesinusoidal, high-frequency identification signal 𝑢GAC(𝑡). Torealize the measurement during the switching operation ofthe IGBT with a switching frequency of 𝑓sw = 5 kHz thesensor response time has to be smaller than 𝑡off = 100 𝜇s.The frequency of the identification signal is in the range of𝑓𝑖 ≈ 5MHz.

𝜑

Z

RG

Im(Z)

Re(Z)

XC =1

j𝜔iCp

XL = j𝜔iLp

RGi

Figure 7: Vector diagram of the resistors 𝑅𝐺 and 𝑅Gi and thereactances𝑋𝐶 and𝑋𝐿 of the parasitic components 𝐿𝑝 and 𝐶𝑝 at thefrequency 𝜔𝑖.

The measurement of the amplitude ��RG and the phaseangle 𝜑 of the voltage drop 𝑈RG enables the calculation ofthe internal gate resistor according to (5). If the identificationfrequency 𝑓𝑖 equals the resonance frequency 𝑓𝑅 of the gatecircuit the phase angle becomes 𝜑 = 0∘. In this case thecos(𝜑) = 1 is constant and solely the amplitude ��RG of thevoltage drop across the external gate resistor 𝑅𝐺 containsthe temperature information and has to be measured. Themeasurement of the voltage amplitude ��RG can be done witha very simple sensor system. Consider

𝑅Gi =𝑈RGi𝐼GAC

= 𝑅𝐺 ⋅cos (𝜑) ⋅ ��GAC − ��RG

��RG. (5)

Within a feasibility study the measurability of the internalgate resistor 𝑅Gi in the series connection was investigated.Thereby it was assumed that the internal gate resistor canbe measured if it has a sufficient large share in the totalimpedance 𝑍 = |𝑍| of the series connection. Figure 7illustrates the external and the internal gate resistors 𝑅𝐺 and𝑅Gi as well as the reactances 𝑋𝐶 and 𝑋𝐿 of the parasiticcomponents of the gate circuit in a vector diagram.

If the frequency 𝑓𝑖 = 𝜔𝑖/(2𝜋) of the identification signalequals the resonance frequency 𝑓𝑅 of the gate circuit bothreactances 𝑋𝐶 = 𝑋𝐿 distinguish themselves and the ratio of𝑅Gi/𝑍 reaches its maximum value, where 𝑍 = 𝑅𝐺 + 𝑅Gi. Thepower module [28, 29] used in further investigations has an𝑅Gi/𝑍 ratio of 62% at a resonance frequency of about 𝑓𝑅 =25.1MHz. In the following a feeding-in method is presentedthat consists of a parallel path whose parasitic inductance𝐿𝑆 decreases the resonance frequency to 𝑓𝑅 = 4.5MHz.The driver circuit inductance 𝐿𝑝 and therefore the switchingbehavior of the IGBT remain uninfluenced.

3.1.2. Feeding-in of the Identification Signal. The superim-position of the negative gate voltage with a high-frequencyidentification signal is challenging. It is especially importantnot to alter the switching behavior of the IGBT, to considersafety aspects, and to ensure a continuous low-resistant con-nection of the gate to avoid an unwanted turn-on of the IGBT.


RG

IGBT driver concept

ConventionaldriverHSS

LSSENIN

HFADC

Con

trol u

nit

GND

AUX

Ls

TJ-measurement

UGon

UGoff

UGACURG

Figure 8: IGBT driver concept with an auxiliary MOSFET to feedin the identification voltage 𝑈GAC. The junction temperature of theIGBT is calculated on the basis of the measured voltage drop 𝑈RG.

EN

IN

HF

t (𝜇s)

t (𝜇s)

t (𝜇s)

t (𝜇s)

t0 t1 t2

UGon

UGoff

uGAC (t)

Figure 9: Signals to control the IGBT driver and to superimpose thenegative gate voltage 𝑈Goff with the sinusoidal identification signal𝑢GAC(𝑡).

In this context the serial feeding-in of a high-frequency signalusing a transformer, whose secondary coil is integrated inseries to the driver circuit, is an unsatisfactory solution [30].On the contrary Figure 8 illustrates the parallel feeding-inof a negative gate voltage 𝑈Goff that is superimposed withthe sinusoidal identification signal 𝑈GAC. For this purposea parallel auxiliary MOSFET is used. During the feeding-in period this auxiliary MOSFET is turned on and theconventional IGBT driver becomes disabled so that its outputis forced to a high impedance state.

Due to the parallel feeding-in the original driver circuitand thus the IGBT switching behavior remain uninfluenced.To simplify the feeding-in process during the inverter oper-ation the driver concept consists of an integrated controlunit that receives the switching pattern and transmits themeasured junction temperature via an optical fiber. Figure 9

ADC

S1

S2

C D

IGBT driver concept

Uo

UGAC

RG

−−

×8

TJ-measurement

URG

Figure 10: Circuit tomeasure the temperature dependent amplitudeof the voltage drop 𝑈RG and to convert it into an ADC compatibledirect voltage.

sketches the control signals to feed in the identificationvoltage 𝑢GAC(𝑡) during the switching operation.

Initially the IGBT driver is enabled EN = 1 and its high-side-switch (HSS) is conductive (IN = 1), so that the positivevoltage𝑈Gon is applied at the gate and the IGBT is in on-state.At the time 𝑡0 the IGBT is switched off through the negativegate voltage𝑈Goff that is applied by the low-side-switch (LSS)by setting IN = 0. To apply the negative, superimposed gatevoltage 𝑢GAC at the time 𝑡1 the IGBT driver is disabled, settingEN = 0, so that its output is forced to a high impedancestate. At the same time the feeding-in of the identificationvoltage is activated by setting HF = 1. To sustain a low-resistant gate connection the signal voltage source 𝑈GACincorporates an impedance converter with a very low outputimpedance. During the feeding-in of the identification signalthe measurement setup is analyzing the voltage drop 𝑈RG togenerate an ADC compatible sensor voltage. At the time 𝑡2the driver concept received the instruction to reactivate theIGBT.Therefore the identification voltage is disabled (HF= 0)and the positive gate voltage is applied, setting EN = 1 and IN= 1. In the case of failure the current limited buffer becomesdeactivated, so that the gate signal of the IGBT driver isdominant.

3.1.3. Generation of the Sensor Output Voltage. As it wasoutlined earlier, in case of resonance the amplitude of thevoltage drop𝑈RG across the external gate resistor𝑅𝐺 dependslinearly on the junction temperature of the IGBT. In Figure 10ameasurement setup is presented that converts the amplitudechange into an ADC compatible direct voltage. Thereforethe voltage drop across 𝑅𝐺 is picked up differentially withthe subtractor S1, rectified with the Schottky diode 𝐷, andsmoothed by the capacitor 𝐶. The level of the resultingdirect voltage changes upon a certain offset linearly with thejunction temperature. To adjust the voltage change to theinput voltage range of anADC, as a first step, the offset voltage𝑈0 is subtracted in a way that the ADC input voltage equals𝑈ADC = 1V at an IGBT junction temperature of 𝑇𝐽 = 20∘C.Later, this offset voltage is used for the calibration of thejunction temperature measurement.

Secondly, an operational amplifier adjusts the ADC inputvoltage to be 𝑈ADC = 3V at 𝑇𝐽 = 120∘C. This results in


tR

tA

activeADC

Time base Trigger50𝜇s C1 DC

C1

50.0 𝜇s/div1.25 MS 2.5 GS/s

Stop 9.1VEdge Positive

FLT DC 1M FLT DC 1M FLT DC 1M10.0 V/div 10.0 V/div10.8500V

2.00 V/div−4.00000 V −29.8500V

0V1V2V3V

15V

4V

−9V

UA

DC

UG

TJ = 120 ∘CTJ = 70 ∘CTJ = 20 ∘C

C1 C2 C3

UGAC

Figure 11: Measurement result of the driver concept where theidentification voltage𝑈GAC leads to a temperature dependent sensorvoltage𝑈ADC that is sampled by an analog to digital converter duringthe time span 𝑡𝐴.

a temperature sensitivity of 20mV/∘C. When using a 10-bit analog to digital converter with a reference voltage of5V the junction temperature can theoretically be measuredwith a resolution of 0.24∘C/bit. The response time of themeasurement setup to generate a valid sensor voltage isprimarily defined by the capacitor 𝐶 and is set to 𝑡𝑅 < 70 𝜇s.

3.1.4. Experimental Results. To qualify the presentedmeasur-ing method a prototype of the IGBT driver with junctiontemperature measurement was developed and integrated inthe double pulse experiment with 𝑈𝐶𝐸 = 600V and 𝐼𝐶 =

250A. A tempering system enables the homogenous heatingof the power module to different junction temperatures 𝑇𝐽.Figure 11 shows the gate voltage 𝑈𝐺 at a switching frequencyof 𝑓sw = 5 kHz. After the IGBT is switched off the negativegate voltage 𝑈Goff is superimposed with the identificationsignal 𝑈GAC = 0.6V with a frequency of 𝑓𝑖 ≈ 4.5MHz.The sensor output voltage 𝑈ADC reaches, after a responsetime of 𝑡𝑅 ≈ 70 𝜇s, a constant voltage level that is sampledby the analog to digital converter of the control unit duringthe timespan 𝑡𝐴 ≈ 6 𝜇s. It can be seen that the sensoroutput voltage increases linearly with an increasing junctiontemperature.

To determine the calibration curve of the driver conceptthe sensor output voltagewasmeasured at four different junc-tion temperatures. At each temperature level the sensor out-put voltagewasmeasured several times to determine the stan-dard deviation and the remaining noise of the temperaturemeasurement. The measurement result in Figure 12 showsthat the sensor output voltage 𝑈ADC is a very linear functionof the junction temperature 𝑇𝐽. The standard deviation ofthe voltage measurement was found to be 𝜎UADC ≤ 24mVthat corresponds to a remaining noise of the temperaturemeasurement of less than ±1.0 K. To realize an easy to applyIGBT driver, this calibration curve is programmed on the

20 40 60 80 100 120

TJ (∘C)

0.5

1.0

1.5

2.0

2.5

3.0

3.5

UA

DC

(V)

deviation:Standard

𝜎UADC = 24mV𝜎TJ = 1.02 ∘C

Calibrationcurve

Figure 12: Calibration curve of the IGBT driver with junctiontemperature measurement.

control unit of the driver concept, so that themeasured sensoroutput voltage can be translated immediately in the junctiontemperature of the IGBT and transmitted to the systemcontroller of the voltage source inverter with an optical fiber.Because of process variations of the internal gate resistorthe IGBT driver has to be calibrated at a known junctiontemperature, for example, at room temperature. Thereby theoffset voltage 𝑈0 is set in a way that the sensor outputvoltage at a homogeneous calibration temperature complieswith the reference calibration curve on the control unit. Theautomatic calibration process is organized by the control unitand enables the pairing of the IGBT driver to a powermodulethat is still installed in the voltage source inverter.

To qualify the IGBT driver during the regular inverteroperation a test setup consisting of one inverter phase thatis connected to a virtual electric machine was developed.All IGBTs were controlled by a dSpace system. A MATLABSimulink model allows the emulation of different load-profiles of a hybrid-electric powertrain. During the regularswitching of the IGBTs the developed IGBT driver measuresthe junction temperature of one IGBT and reports it to thedSpace system, where it is displayed in real time. To verify thetemperature measurement with an IR camera a coated powermodule was used. Figure 13 shows the motor phase current𝐼AC and the junction temperature 𝑇𝐽 that is measured withthe developed IGBT driver 𝑇𝐽,DR and an IR camera 𝑇𝐽,IR. Thetimespan of the load-profile is one minute. It can be seen thatthe temperature variations measured with the IGBT driverare in a very good agreement with those obtained with the IRcamera. For this reason the developed IGBT driver is suitablefor the temperature cycle recording of a powermodule duringthe inverter operation in a hybrid car.

Figure 14 shows the temperature profile of the IGBT fora timespan of 45 minutes. It can be seen that the load-profile consists of a majority of operational temperaturecycles caused by the transient power flow over the powermodule. Those active temperature cycles depend on thedriver command and the superior hybrid strategy and lead


0 10 20 30 40 50 60

0 10 20 30 40 50 60

0

0

100

200

300

400

20

40

60

80

TJ

(∘ C)I A

C(A

)

IAC

TJ,DRTJ,IR

t (s)

t (s)

Figure 13: Junction temperature during inverter operation mea-sured with the IGBT driver and an IR camera.

0 5 10 15 20 25 30 35 40 45

t (min)

0

25

50

75

100

125

TJ

(∘ C)

Cool-down

Start

Operation

Figure 14: Typical load-profile of an IGBTpowermodule in a hybridcar that specifies the IGBT junction temperature during amission of45 minutes.

to the heating of the coolant and the cold plate temperatureduring the first 15 minutes of the driving. The cool-downin the middle of the load-profile represents a time periodwithout electric power flow. Consequently the coolant andthe power module temperature decline until the subsequentoperation cycles will cause the reheating of the system.After about 42 minutes the mission ends and the entiresystem cools down. In view of the homogeneous temperaturevariation of the power module during the start and the cool-down, these cycles should be labeled as passive ones.

Because of the limited storage capacity on the systemcontroller it is not possible to store the whole temperatureprofile over a representative long period of time. For thispurpose, an online recording algorithm is presented thatprocesses the measured junction temperature into a compactset of information that is relevant for the verification ofthe theoretical load-profiles and the lifetime calculationapproach. In view of the state of the art the verification can be

done on the basis of temperature cycles that were calculatedwith a Rainflow algorithm and stored with their amplitude,their minimum temperature, and their heating time.

3.2. Online Temperature Cycle Recording Algorithm. Theonline calculation and storage of temperature cycles on thesystem controller require a modification of the state-of-the-art algorithms used on desktop computers with regard totheir online ability, their computing power, and the requiredstorage capacity. Thereby the overall objective is to reach acompromise between a minimum calculation and storageeffort and a maximum accuracy regarding lifetime relevantinformation. Figure 15 shows the block diagram of the devel-oped online cycle recording algorithm that consists of anextreme value filter algorithm, an online Rainflow algorithm,and a classified storage of the determined temperature cyclesin a four-dimensional frequency distribution. In the follow-ing the working principles of the algorithms are presented.

3.2.1. Extreme Value Filter Algorithm. The calculation oftemperature cycles using a cycle counting algorithm requiresprevious identification of extreme values. Therefore a three-point algorithm [31] with a sliding average filter can beused. The filter averages the temperature profile and enablesthe filtering of noise. It is not suitable to filter out smalltemperature variations in the range of Δ𝑇𝐽 ≈ 5∘C, as theywere caused by the hybrid strategy.The three-point algorithmpicks out three subsequent temperature values. If the secondvalue is higher than the first and third value a maximum isidentified. Because of the limited filter level this algorithmfinds a lot of extreme values, which have to be furtherprocessed in the online Rainflow. To reduce the computingpower a logical filter algorithm was developed that enablesthe filtering of these small temperature variations but stilldelivers the exact temperature of valid extreme values. Theprinciple of the filtering is outlined in Figure 16. In thefilter routine two variables 𝑇𝑆 and 𝑡𝑠 were increased withthe temperature 𝑇𝐽(𝑡). Only after the temperature declines,a predefined filter temperature 𝑇𝐹, a valid extreme value isdetected. It occurred at the time 𝑡𝑠 and has the temperaturevalue 𝑇𝑆.

In Figure 17 the filter algorithm is applied to a cutoutof the presented load-profile of a hybrid car. The filtertemperature is set to 𝑇𝐹 = 5∘C. It can be seen that the load-profile consists of many small temperature variations, whichdoes not contribute notably to the damage of the powermodule. These temperature variations are smaller than thespecified filter temperature so that they are not identifiedas valid extreme values. In a sequence of small temperaturevariations the logical filter algorithm determines the lowestextreme value. In this view the developed extreme valuefiltermeets all requirements and combines an adjustable filterlevel and a high data reduction but still delivers the accuratetemperature of valid extreme value.

Thefilter temperature should be set in away that it enablesthe highest possible data reduction without the rejection oflifetime relevant information. To parameterize the filter for itsuse in a hybrid car Figure 18 shows the number of identified


Extreme value filter algorithm

Online cyclecounting algorithm distribution

4D-frequency

TJ

(t)

TJ

(t)

ΔTJ

TJ,min

ton

ΔΔTJJ

TTJT ,

t

Figure 15: Online cycle counting algorithm to calculate and store temperature cycles with less effort during the operation of the IGBT powermodule.

0 5 10

t (s)

TS

TF

ts

0

10

20

30

TJ

(∘ C)

Validextreme

value

TJ(t)

Figure 16: Working principle of the extreme value filter algorithmwith the auxiliary variables 𝑇𝑆 and 𝑡𝑠 and the adjustable filtertemperature 𝑇𝐹.

500 510 520 530 540

t (s)

50

60

70

80

90

100

TJ

(∘ C)

Validextreme

value

Filter

Filtertemperature:TF = 5 ∘C

Figure 17: Application of the extreme value filter algorithm with afilter temperature of 𝑇𝐹 = 5∘C to a cutout of the load-profile of ahybrid car.

extreme values 𝑁𝐸 in the load-profile as a function of thefilter temperature 𝑇𝐹. Moreover the error 𝐸 in the calculatedlifetime due to the filtering of small temperature variations isdiagrammed.

It appears that a filter temperature of 𝑇𝐹 = 5∘C leads toa significant reduction of the number of extreme values by

0.01 5 10 15

0.8

0.6

0.4

0.2

0

4000

3000

2000

1000

0

Optimum filter temperature:TF = 5 ∘CNE = 923E ≤ 0.002% E

(%)

E

TF (∘C)

NE

NE

Figure 18:The number𝑁𝐸 of extreme values that were found in thepresented load-profile of a hybrid car decreases with an increasingfilter temperature 𝑇𝐹.

more than 85 percent. The filtering of the small temperaturevariations that are caused by the hybrid strategy results in adeviation of the calculated lifetime of 𝐸 = +0.002 percent. Inview of the major reduction of extreme values and thereforethe required computing power, thisminor calculation error isacceptable. Consequently the optimum filter temperature toanalyze the presented load-profile of a hybrid car is set to 𝑇𝐹= 5∘C.

Finally the result of the filtering in the lifetime calcula-tion should be demonstrated. Therefore Figure 19 shows thenumber 𝑛𝑍 of temperature cycles with similar amplitudesΔ𝑇𝐽 that were determined with a filter level of 𝑇𝐹 = 0∘Cand 𝑇𝐹 = 5∘C. Additionally the lifetime consumption due tothe temperature cycles of each class was calculated with theCIPS08 lifetimemodel and normalized to one operating hourLC/h. It is obvious that cycles with amplitudes of Δ𝑇𝐽 ≤ 5∘Cmake up amajority of all cycles but do not contribute notablyto the lifetime consumption of the power module, so that theaccumulated lifetime consumption is nearly identical.

In summary the filter algorithm prevents the calculationof temperature cycles that are smaller than a filter tempera-ture of 𝑇𝐹 = 5∘C, so that the computing power of the onlineRainflow algorithm can be significantly reduced. In Figure 19all greyed temperature cycles with Δ𝑇𝐽 ≤ 𝑇𝐹, which make up


01 10 20 30 40 50

0

50

100

2

4

LC/h

Operationalcycles LC/h

(accumulated)Defined filtertemperatureTF = 5 ∘C

nZ

nZ

LC/h

·10−5

ΔTJ (∘C)

Figure 19: Number 𝑛𝑍 of temperature cycles Δ𝑇𝐽 determined withthe filter temperatures 𝑇𝐹 = 0∘C and 𝑇𝐹 = 5∘C and a subsequentRainflow algorithm.

a share of about 85 per cent, will not be calculated anymore.Despite the filtering the exact temperature of valid extremevalues is determined.

3.2.2. Online Cycle Counting Algorithm. The ability of count-ing algorithms to calculate cycles during the inverter opera-tion depends on their working principle. As it was illustratedin Figure 1 the half-cycle or the maximum-edge methodcalculates a temperature cycle on the basis of two or threesubsequent extreme values. For this reason the online imple-mentation of those simple countingmethods would be rathersimple. Online Rainflow algorithms are described in [32–34]. The feature of the Rainflow method is the counting ofclosed temperature cycles. For this reason an online Rainflowalways demands the intermediate storage of extreme values ina working storage, so that its online implementation on thesystem controller is complicated.

Substitutability of the Rainflow Method. In the followinginvestigations it should be analyzed whether the Rainflowmethod can be substituted with an easy to implement half-cycle or maximum-edge algorithm [6]. For this purpose thecapability of the three algorithms to convert the presentedload-profile into a distribution of defined temperature cyclesis investigated. Figure 20 shows the number of temperaturecycles𝑁with similar amplitudesΔ𝑇𝐽 that were extractedwitha Rainflow algorithm. Beside themajority of operational tem-perature cycles with amplitudes of Δ𝑇𝐽 < 70∘C two passivecycles with amplitudes of Δ𝑇𝐽 > 100∘Cwere calculated.Thesecycles are caused by the homogenous heating of the entirepower module during the start and the cool-down. All cycleswere valued with the CIPS08 lifetime model and normalizedto their lifetime consumption per operating hour LC/h. Itturns out that the lifetime consumption of the two passivecycles is greater than the accumulated lifetime consumptionof all operational cycles. In this view it is very important toextract and parameterize passive cycles with a high accuracy.Because of its physical background the results of the Rainflow

LC/h

·10−5

0

3

Startcycle

0

300

0 50 100

150N

N

Operationalcycles LC/h

(accumulated)

Cool-downcycle

LC/h

ΔTJ (∘C)

Figure 20:Number of temperature cycles𝑁with similar amplitudesΔ𝑇𝐽 that were determined with a Rainflow counting algorithm.The cycles were valued with the CIPS08 lifetime model; LC/h =

𝑓(Δ𝑇𝐽, 𝑇𝐽,min, 𝑡on).

counting should be taken as reference for the evaluation ofhalf-cycle and maximum-edge counting.

The half-cycle method extracts twice as many operationalcycles. Their valuing as half temperature cycles leads to areduced operational lifetime consumption compared to theRainflow algorithm [6]. This applies equally to the passivecycles. Additionally the passive cycle amplitudes were foundsignificantly smaller than in the Rainflow counting, so thattheir lifetime consumption is drastically undervalued. Themaximum-edge counting compensates the undervaluing ofthe passive cycles partially due to the interpretation of thehalf temperature cycles as full ones. Nevertheless the cycleamplitudes differ from the Rainflow counting and cause asmaller lifetime consumption.

In summary, simple counting methods calculate a signif-icant lower lifetime consumption than the Rainflow methodso that they predict a longer lifetime of the power module. Inthis view it is not possible to substitute the Rainflow methodwith a simple counting algorithm and there is a need toimplement an online Rainflow.

Online Rainflow Algorithm. For the calculation of temper-ature cycles with a state-of-the-art Rainflow algorithm allextreme values of the load-profile have to be stored in avector. In the case of longer test drives this also results ina large quantity of data that has to be stored temporarilyin a working storage. Contrarily to the presented range-counting Rainflow that scans the working storage with awindow in search of closed temperature cycles Figure 21shows an online version of this principle.Thereby thewindowis fixed at the beginning of a working storage with a variablesize. Once a new extreme value appears the working storageis shifted and the new extreme value is written to the 𝐸3position. The Rainflow cycle condition is checked and if itis true the temperature cycle is calculated and exported tothe storage. The accompanied extreme values were deletedand the working storage is shifted for two positions. Sinceone new extreme value can close more temperature cycles theRainflow cycle condition is checked again until it becomesfalse. To calculate the cycle heating times a corresponding


TrueRainflowExit Rainflow

functionFalse Export to

data storage

New extremum

Check again

∙ Shift working storage∙ Save at E3-position

cycle condition

E1 E2 E3

Working storage (TJ,Ext -values)

Working storage (ΔtExt -values)

NE

tE1 tE2 tE3

Function Rainflow (TJ,Ext , tExt )

Figure 21: Working principle of the online Rainflow algorithm that consists of a variable working storage for extreme values and their timeinformation and a routine to check the Rainflow cycle condition (|𝐸2 − 𝐸3| ≤ |𝐸2 − 𝐸1|).

working storage for the time stamps of the extreme values isavailable. To enable online cycle recording for an unlimiteddriving time with a storage capacity of a few bytes onlythe timespans between the extreme values were stored. Thisenables a reconstruction of heating times of up to 255 secondsthat is enough comparedwith themaximumspecifiedheatingtime of the CIPS08 model.

Working Storage Overflow. A critical parameter of an onlineRainflow algorithm that should be implemented on a micro-controller is the size of the required working storage. Inthe following the number of extreme values in the workingstorage during the analysis of the load-profile should beexamined. Therefore Table 1 outlines four exemplary tem-perature profiles 𝑇𝐽(𝑡), the number of extreme values 𝑁𝐸 inthe working storage, and the extracted temperature cycleswith the amplitudes Δ𝑇𝐽. It becomes clear that solely thetemperature profile 3 causes an increase of the number ofextreme values in the working storage above the overflowlevel of𝑁𝐸,max = 10. After the start temperature is reached allcycles were closed simultaneously and the working storage isemptied. All other temperature profiles enable the instanta-neous calculation of cycles and do not increase the numberof extreme values in the storage.

In a hybrid car temperature profile 3 typically arisesduring the voltage- and frequency-controlled starting pro-cedure of the electric machine [35]. Figure 22 sketches anexemplary junction temperature profile during the start-up ofthe electric machine in a hybrid car. It consists of many smalltemperature cycles, whose amplitude and frequency dependon the size and the fundamental frequency of the inverteroutput current.

The oscillations of the junction temperature during thestart-up of the electric machine cause an increase of thenumber of extreme values 𝑁𝐸 within the working storage.This leads to an overflow of the available working storagewith a maximum size of 17 extreme values and a loss ofinformation. To avoid an overflow the developed onlineRainflow algorithm was expanded by an overflow routine,which is scanning the working storage in search of thestarting point of profile 3 once the overflow level is reached.Subsequently all temperature cycles of the detected sequence

80

70

600 0.5 1 1.5 2

Compliance with profile 3

TJ

(∘ C)

Figure 22: The junction temperature profile during the start-upof the electric machine complies with the exemplary temperatureprofile number 3.

were calculated and deleted from the working storage. Forthis reason the overflow routine allows the use of a smallworking storage and improves the implementation of theRainflow counting on a microcontroller.

3.2.3. Temperature Cycle Storage. In view of the state of theart the lifetime calculation on the basis of the presentedload-profile demands the storage of each cycle with itsamplitude Δ𝑇𝐽, its minimum temperature 𝑇𝐽,min, and itsheating time 𝑡on [6]. To store the calculated temperaturecycles on the system controller with a minimum storagecapacity a four-dimensional frequency distribution shouldbe used [7]. Figure 23 outlines the storage of 𝑛𝑍 cycles withsimilar amplitudes Δ𝑇𝐽 and minimum temperatures 𝑇𝐽,min.To consider cycle specific heating times, each temperatureclass is linked with a histogram containing the number 𝑛𝑃 ofcycles with equal time stamps. In Figure 23 the class width𝑇CW to group cycles with similar amplitudes is set to 𝑇CW= 5∘C. To reach a compromise between the required storagecapacity and the achievable accuracy in lifetime calculation,the resolutions of the temperature and the time axis werevaried.

As an example of this variation Figure 24 diagrams thestorage capacity 𝑆 and the lifetime calculation error 𝐸 independency of an increasing class width 𝑇CW to group cycleswith similar amplitudes Δ𝑇𝐽. It turns out that a class widthof 𝑇CW = 5∘C enables a reduction of the storage capacity to


Table 1: Number of extreme values 𝑁𝐸 in the working storage during the analysis of four exemplary temperature profiles with an onlineRainflow.

Profile 1 Profile 2 Profile 3 Profile 4

𝑇𝐽(𝑡)

𝑁𝐸

NE,max

Δ𝑇𝐽

0

0.5

1

1.5

2

nZ

(a)×105

050

100150

200 150

10050

0

−50

1 5 10 15

(b) Time classes

np

ton (s)

TCWΔT

J ( ∘C)TJ,m

in(∘ C)

cyclescycles

PassiveActive

Temperatureclasses

Figure 23: Storage of temperature cycles with similar (a) amplitudesΔ𝑇𝐽 and minimum cycle temperatures 𝑇𝐽,min in a matrix. (b) Eachmatrix element contains a histogram that gives information aboutthe cycle heating times 𝑡on.

𝑆 = 51.6 kilobytes. A further increase of the class widthwouldlead to an increased error, which is caused by the temperaturecycles that were rounded up into the next higher temperatureclass and their nonlinear overvaluing in the CIPS08 lifetimemodel. In the case of a class width of𝑇CW = 5∘C the calculatedlifetime is affected by less than 𝐸 = −0.5 percent, so that it isa suitable compromise between a minimum storage capacityand a good storage accuracy.

3.2.4. Validation of the Online Cycle Recording Algorithm.Finally the properties of the developted online cycle record-ing algorithm should be evaluated. The major advantage ofa conventional algorithm, which was presented before, isthat each temperature cycle becomes valued with its exactindividual cycle parameters. For this reason a conventionalalgorithm enables the most accurate lifetime calculation, so

0 5 10 15

TCW (∘C)

0

30

60

90

120

1500

−1

−3

−5

E(%

)

S(k

B)

S = 51.6 kB

|E| ≤ 0.5%

−2

−4

Figure 24: Required storage capacity 𝑆 and the error 𝐸 in thecalculated module lifetime in dependency of an increasing classwidth 𝑇CW.

that it can be used as reference. In Figure 25 the requiredcomputing power to determine temperature cycles in theload-profile of a hybrid car and the required storage capacityto store them for lifetime calculation are diagrammed forthe conventional algorithm and the developed online cyclerecording algorithm.

It turns out that the required computing power to calcu-late the temperature cycles within the presented load-profilecould be reduced by 85 percent due to an extreme value filteralgorithm. Moreover the storage of the determined cycles ina four-dimensional frequency distribution and the variationof the class widths enable a reduction of the required storagecapacity by 99 percent compared to minimum class widths ofone degree and one second. Despite these major reductionsthe calculated lifetime on the basis of the temperature cyclesdetermined by the online algorithm deviates less than 0.8percent from the lifetime calculated with the conventionalextracted cycles. Because of its properties the developed


Computingpower

Storagecapacity

Calculatedlifetime

Conventional algorithmOnline algorithm

−99%−85%−0.8%(64h)

8kB

138

E/m

in

20E/min

8016

h

1.3

MB

Figure 25: Comparison of the conventional algorithm used ondesktop computers with the developed online cycle recordingalgorithm that should be implemented on the system controller ofthe voltage source inverter.

algorithm can be implemented on the system controller ofthe voltage source inverter to record the temperature cyclesof the IGBT power module during its operation [24].

4. Conclusions

The analysis of the state of the art to design the lifetime ofIGBT power modules used in hybrid cars showed that thereis a need to verify the theoretical load-profiles with datafrom the field. For this purpose a temperature cycle recorderwas presented that records the temperature cycles of anIGBT power module during its operation in a voltage sourceinverter. For the measuring of the junction temperatureduring inverter operation a modified IGBT gate driver waspresented. The driver concept determines the temperaturesensitive IGBT internal gate resistor by superimposing thenegative gate voltage with a high-frequency identificationsignal. For this purpose the conventional IGBT driver wasextended by an auxiliary MOSFET and a control unit thatmanages the feeding-in process during the regular switch-ing operation, the transmission of the measured junctiontemperature, and the automatic calibration of the sensorsystem. Finally, it is shown that the driver enables real-time junction temperature measurement during the regularinverter operation.

In order to reduce the quantity of data that has to be storedon the system controller an online cycle recording algorithmwas developed that processes the measured junction tem-perature into a frequency distribution of temperature cycles.A review of empirical lifetime models showed that eachtemperature cycle has to be parameterized with its ampli-tude, its minimum temperature, and its heating time. Thealgorithm consists of a filter that identifies all extreme valuesthat are relevant for the lifetime calculation. An online range-counting Rainflow algorithm processes them into closedtemperature cycles and stores them in a four-dimensionalfrequency distribution. In relation to the cycle countingalgorithm used on desktop computers the computing powercould be decreased by 85 percent and the required storage

capacity could be decreased to 8 kilobytes. Despite thesereductions the compliance of the online algorithm and theconventional desktop computer algorithm amounts to 99%.

Currently the IGBT driver with junction temperaturemeasurement and the online temperature cycle recordingalgorithm are integrated in the voltage source inverter of firsttest vehicles. On the basis of the recorded load-histories it willbe possible to verify and adjust the theoretical load-profilesused in today’s lifetime calculation approach. For this reasonthe presented temperature cycle recordermakes an importantcontribution to improve the accuracy of the lifetime calcu-lation of IGBT power modules and the reliability of futurevoltage source inverters.

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper.

Acknowledgments

This project was supported by the ZF Friedrichshafen AG.Special thanks go to the Department of Engineering ElectricMobility & Mechatronics, Auerbach i. d. Opf.

References

[1] W. Sch, A. Hensler, J. Lutz et al., “Hybrid drive as a variationof a gear box,” in Symposium Hybrid- und Elektrofahrzeuge,Braunschweig, Germany, February 2011.

[2] M. Ciappa, “Lifetime modeling and prediction of powerdevices,” in Proceedings of the 5th International Conference onIntegrated Power Systems (CIPS ’08), pp. 27–36, 2008.

[3] M. Thoben, K. Mainka, R. Bayerer, I. Graf, and M. Munzer,“From vehicle drive cycle to reliability testing of powermodulesfor hybrid vehicle inverters,” in Proceedings of the InternationalExhibition and Conference for Power Electronics, IntelligentMotion and Power Quality (PCIMEurope ’08), pp. 651–656,May2008.

[4] J. Lutz, H. Schlangenotto, U. Scheuermann, and R. D. Doncker,Semiconductor Power Devices, Springer, New York, NY, USA,2011.

[5] M. Denk and M.-M. Bakran, “An IGBT driver concept withintegrated real-time junction temperature measurement,” inProceedings of the International Exhibition and Conference forPower Electronics, Intelligent Motion, Renewable Energy andEnergy Management (PCIM ’14), pp. 1–8, May 2014.

[6] M. Denk and M.-M. Bakran, “Comparison of counting algo-rithms and empiric lifetime models to analyze the load-profileof an IGBT power module in a hybrid car,” in Proceedings of the3rd International Electric Drives Production Conference (EDPC’13), October 2013.

[7] M. Denk andM.-M. Bakran, “Efficient online-algorithm for thetemperature cycle recording of an IGBT power module in ahybrid car during inverter operation,” in Proceedings of the 8thInternational Conference on Integrated Power Systems (CIPS ’14),pp. 1–6, 2014.

[8] M. Ciappa, “Selected failure mechanisms of modern powermodules,” Microelectronics Reliability, vol. 42, no. 4-5, pp. 653–667, 2002.


[9] M.Held, P. Jacob, G. Nicoletti, P. Scacco, andM.-H. Poech, “Fastpower cycling test for IGBT modules in traction application,”in Proceedings of the 2nd International Conference on PowerElectronics and Drive Systems (PEDS ’97), pp. 425–430, May1997.

[10] R. Bayerer, T. Herrmann, T. Licht, J. Lutz, andM. Feller, “Modelfor power cycling lifetime of IGBT modules—various factorsinfluencing lifetime,” in Proceedings of the 5th InternationalConference on Integrated Power Systems (CIPS ’08), pp. 37–42,Nuremberg, Germany, March 2008.

[11] Y. Wang, S. Jones, D. Chamund, and G. Liu, Lifetime Modellingof IGBT Modules Subjected to Power Cycling Tests, PCIM, 2013.

[12] J. Lutz, “IGBT-modules: design for reliability,” in Proceedings ofthe 13th European Conference on Power Electronics and Appli-cations (EPE ’09), pp. 1–3, Chemnitz University of Technology,2009.

[13] K. Mainka, M. Thoben, and O. Schilling, “Lifetime calcula-tion for power modules, application and theory of modelsand counting methods,” in Proceedings of the 14th EuropeanConference on Power Electronics and Applications (EPE ’11), pp.1–8, IEEE, Birmingham, UK, September 2011.

[14] M. Ciappa, F. Carbognani, and W. Fichtner, “Lifetime pre-diction and design of reliability tests for high-power devicesin automotive applications,” IEEE Transactions on Device andMaterials Reliability, vol. 3, no. 4, pp. 191–196, 2003.

[15] F. Ellyin, Fatigue Damage, Crack Growth and Life Prediction,Chapman & Hall, New York, NY, USA, 1996.

[16] ASTM E-1049, “Standard Practices for Cycle Counting inFatigue Analysis,” Annual Book of ASTM Standards, 1999.

[17] M. Miner, “Cumulative damage in fatigue,” Jornal of AppliedMechanics, vol. 12, pp. 159–164, 1945.

[18] E. R.Motto and J. F. Donlon, “IGBTmodule with user accessibleon-chip current and temperature sensors,” in Proceedings of the27th Annual IEEE Applied Power Electronics Conference andExposition (APEC ’12), pp. 176–181, Orlando, Fla, USA, February2012.

[19] D. Wagenitz, A. Hambrecht, and S. Dieckerhoff, “Lifetimeevaluation of IGBT power modules applying a nonlinear sat-uration voltage observer,” in Proceedings of the 7th InternationalConference on Integrated Power Electronics Systems (CIPS ’12),pp. 256–260, March 2012.

[20] B. Ji, V. Pickert, and B. Zahawi, “In-situ bond wire and solderlayer health monitoring circuit for IGBT power modules,” inProceedings of the 7th International Conference on IntegratedPower Electronics Systems (CIPS ’12), March 2012.

[21] F. Richardeau, M. Morvan, and F. Mosser, On-Line TJ Monitor-ing Sensor Embedded onVSI Driver Board Application on a 5 kWHigh Speed PMSM for Aeronatic Blowing-Fan, PCIM, 2014.

[22] I. Bahun, V. Sunde, and Z. Jakopovic, “Estimation of insulated-gate bipolar transistor operating temperature: simulation andexperiment,” Journal of Power Electronics, vol. 13, no. 4, pp. 729–736, 2013.

[23] V.-K. Sundaramoorty, E. Bianda, R. Bloch, and F. Zurfluh,“Simultaneous online estimation of junction temperature andcurrent of IGBTs using emitter-auxiliary emitter parasitic indu-ctance,” in Proceedings of International Exhibition and Confer-ence for Power Electronics, Intelligent Motion, Renewable Energyand Energy Management (PCIM ’14), pp. 1–8, Nuremberg,Germany, 2014.

[24] H. Kuhn and A. Mertens, “On-line junction temperature mea-surement of IGBTs based on temperature sensitive electrical

parameters,” in Proceedings of the 13th European Conference onPower Electronics and Applications (EPE ’09), September 2009.

[25] D. Barlini, M. Ciappa, A. Castellazzi, M. Mermet-Guyennet,and W. Fichtner, “New technique for the measurement of thestatic and of the transient junction temperature in IGBT devicesunder operating conditions,” Microelectronics Reliability, vol.46, no. 9–11, pp. 1772–1777, 2006.

[26] V. Sundaramoorthy, E. Bianda, R. Bloch, I. Nistor, G. Knapp,andA. Heinemann, “Online estimation of IGBT junction temp-erature (Tj) using gate-emitter voltage (Vge) at turn-off,” inProceedings of the 15th EuropeanConference onPower Electronicsand Applications (EPE ’13), September 2013.

[27] W. Brekel, T. Dutemeyer, G. Puk, and O. Schilling, Time Reso-lved In Situ T𝑣𝑗 Measurements of 6.5 kV IGBTs during InverterOperation, PCIM, 2009.

[28] Infineon Technologies, “Infineon technical information,” Tech.Rep. FF225R12ME4, Infineon Technologies, 2013.

[29] Infineon Technologies, “Industrial IGBT modules, explanationof technical information,” Application Note AN 2011-05, 2013,http://www.infineon.com/.

[30] E. Hoene and T. Baumann, “Device for measuring a tempera-ture of a power semiconductor,” WO2013/000971, 2013.

[31] R. Schmidt and U. Scheuermann, “Using the chip as a tem-perature sensor—the influence of steep lateral temperaturegradients on the Vce(T)-measurement,” in Proceedings of the13th European Conference on Power Electronics and Applications(EPE ’09), September 2009.

[32] L. Gopireddy, L. M. Tolbert, and B. Ozpineci, “Lifetime predic-tion of IGBT in a STATCOMusingmodified-graphical rainflowcounting algorithm,” in Proceedings of the 38th Annual Con-ference on IEEE Industrial Electronics Society (IECON ’12), pp.3425–3430, October 2012.

[33] G. Glinka and J. C. P. Kam, “Rainflow counting algorithm forvery long stress histories,” International Journal of Fatigue, vol.9, no. 4, pp. 223–228, 1987.

[34] S. D. Downing and D. F. Socie, “Simple rainflow countingalgorithms,” International Journal of Fatigue, vol. 4, no. 1, pp. 31–40, 1982.

[35] R. Krummer, R. Teimann, G. Berger, J. Petzoldt, and L. Lorenz,“On-line calculation of the chip temperature of power modulesin voltage source converters using the microcontroller,” inProceedings of the 8th European Conference on Power Electronicsand Applications (EPE ’99), Lausanne, Switzerland, 1999.

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttp://www.hindawi.com Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014


Active and Passive Electronic Components

Control Scienceand Engineering

Journal of



RotatingMachinery


Hindawi Publishing Corporation http://www.hindawi.com

Journal ofEngineeringVolume 2014

Submit your manuscripts athttp://www.hindawi.com

VLSI Design



Shock and Vibration


Civil EngineeringAdvances in

Acoustics and VibrationAdvances in



Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation http://www.hindawi.com

Volume 2014

The Scientific World JournalHindawi Publishing Corporation http://www.hindawi.com Volume 2014

SensorsJournal of


Modelling & Simulation in EngineeringHindawi Publishing Corporation http://www.hindawi.com Volume 2014


Chemical EngineeringInternational Journal of Antennas and

Propagation




Navigation and Observation



DistributedSensor Networks


research article online junction temperature cycle...

Documents