a robust electro-thermal igbt spice model:...

Microelectronics Reliability xxx (2015) xxx–xxx

MR-11637; No of Pages 5

Contents lists available at ScienceDirect

Microelectronics Reliability

j ourna l homepage: www.e lsev ie r .com/ locate /mr

A robust electro-thermal IGBT SPICE model: Application to short-circuit protectioncircuit design

D. Cavaiuolo a, M. Riccio a,⁎, L. Maresca a, A. Irace a, G. Breglio a, D. Daprà b, C. Sanfilippo b, L. Merlin b

a Dept. of Electrical Engineering and Information Technologies, University Federico II, Via Claudio 21, 80125 Naples, Italyb Vishay Intertechnology, Diodes Division, Via Liguria 49, 10071 Borgaro Torinese, Turin, Italy

⁎ Corresponding author.E-mail address: [email protected] (M. Riccio).

http://dx.doi.org/10.1016/j.microrel.2015.06.0860026-2714/© 2015 Elsevier Ltd. All rights reserved.

Please cite this article as: D. Cavaiuolo, et al.Microelectronics Reliability (2015), http://dx

a b s t r a c t
a r t i c l e i n f o
Article history:Received 25 May 2015Accepted 18 June 2015Available online xxxx

Keywords:IGBTSPICE modelElectro-thermalShort-circuit

An optimized electro-thermal IGBT SPICE model based on the Kraus model was developed to allow reliablesimulation at application level. A particular emphasis to the temperature dependence of physical parameterswas given for both the on-state and breakdown conditions. The model was experimentally validated in steadystate and transient operation on a Field-Stop trench-gate 30 A–600 V commercial IGBT device. The effectivenessof the convergence–accuracy aspects was proved for the proposed model. The application to the short-circuitprotection circuit design for a hard switched fault in an inverter motor-drive application was also analysed.The ability of the proposed IGBT model was experimentally verified with a simplified-equivalent circuit wherea short condition was forced on the load, enabling the desaturation protection by the IGBT driver. The resultsshow an accurate prediction of the device electro-thermal behaviour under short-circuit conditions, with apossible optimal design of the desaturation circuit parameters when the device experiences hard-switched-fault or fault-under-load events.

© 2015 Elsevier Ltd. All rights reserved.

1. Introduction

Nowadays a careful choice and design of IGBT short-circuit (SC) pro-tection circuits have become crucial [1,2] also because in some applica-tion device may experience many non-destructive SC events [3]. In thisscenario, the abrupt increase of junction temperature may significantlyreduce the device lifetime expectancy according to the theorizedmodels [4]. Unfortunately, design of protection circuit parameters is ac-tually performed according to datasheet and application notes, whichprovide a general approach and rough design rules, neglecting devicetechnological specificity and self-heating effect: therefore, the optimaldesign cannot be pursued. Consequently, accurate electro-thermal(ET) simulations of IGBT under SC condition [5] would be very effectivefor protection circuit design. Besides numerical TCAD simulations,where detailed information on the device structure are needed, a validsolution can be the use of SPICE simulations with coupled electricaland thermal networks [6]. Unfortunately, despite a very high numberof compact IGBT models proposed in literature, only two models areimplemented within SPICE simulators as built-in library: the NIGBTmodel, based on the Hefner model [7], included in PSpice OrCad® andthe HiSIM model [8] implemented in ELDO SPICE®. Furthermore, IGBTmanufacturers provide itself SPICE sub-circuits to simulate his products,where a drastic trade-off between convergence and accuracy is present[9]. Hence, in this paper, we propose an accuracy–convergence

, A robust electro-thermal IGB.doi.org/10.1016/j.microrel.2

optimized IGBT SPICE model, that accounts for self-heating effect, withthe aim of evaluating the impact of a well-aimed protection circuit de-sign on device reliability.

2. IGBT SPICE model

The model we propose is an enhanced version of the Kraus model[10], optimized for PSpice implementation. A remarkable improvementconcern the addition of a linear-region transconductance factor KF

adopted in Hefner model for reproducing the PiN–BJT combined effectof trench-gate IGBTs [7]. The equation for theMOSFET current becomesthe following:

IMOS ¼KfKp Vgs−Vth

� �Vds−

KfV2ds

2

" #

2 1þ ϑ Vgs−Vth� �� ¼

Kp Vgs−Vth� �

V�ds−

V�2ds

2

" #

2 1þ ϑ Vgs−Vth� ��

KfVds ¼ V�ds≤ Vgs−Vth

� �:

ð1Þ

The current expression includes the effects of the high transverseelectric field for high gate voltage is also been included through the pa-rameter θ. Hence, it's possible to implement the Hefner approach formodelling the PiN effect in Kraus PSpice model by only means of a sim-ple modification: the internal IGBT VDS voltage is multiplied by Kf andthen imposed as drain-source voltage across the Level 1 MOSFETmodel through the a controlled voltage source.

T SPICE model: Application to short-circuit protection circuit design,015.06.086

http://dx.doi.org/10.1016/j.microrel.2015.06.086

mailto:[email protected]

Journal logo


http://www.sciencedirect.com/science/journal/

www.elsevier.com/locate/mr


Fig. 2.Model comparison on a) IGBTDCoutput characteristics and b) on inductive turn-offIGBT current and voltage waveforms at Vcc = 400 V, Ic = 30 A, Rg= 10Ω and L= 250 μHat Tj = Tamb.

2 D. Cavaiuolo et al. / Microelectronics Reliability xxx (2015) xxx–xxx

Furthermore, other important enhancements regard the introduc-tion of Miller temperature-dependent avalanche modelling [7]:

Mav ¼ 1

1− Vcb

BVce T jð Þ� �β ð2Þ

the use of explicitly evaluated solution for steady-state carriers chargeQb0 aimed tomodel optimization [11] and the definition of temperaturedependencies for physical parameters in case of trench-gate devices[12].

The newmodel was validated on a Field Stop (FS) trench-gate 30 A–600 V commercial IGBT and the algorithmproposed in [13] is used to ef-fectively evaluate the model parameters. In order to identify the mainproperties and the improvements brought by newmodel, it is comparedwith sub-circuit based model provided by manufacturer and NIGBTmodel present in the OrCAD Pspice library. In Fig. 1 an overview of themain performance characteristics for the three considered SPICEmodelsis given. Apart the comparative evaluation of the accuracy, the speedand the convergence of the model, other features were evaluated. Thephysics accounts for the number of formulas involved in the models;the parameters accounts for the number of physical parameters of themodels; the extensibility accounts for the number of parameters thatcan be customized; the simplicity accounts for the inverse of the com-putational complexity of the model. In particular results from DC char-acteristics, along with inductive load turn-off transient are normalizedand reported in Fig. 1. The other parameters are evaluated by the thor-ough evaluation of the model equations and implementation in SPICElike simulator. Simulation results are compared to the experimentalones.

The devicemeasured-simulatedDC characteristics and turn-off tran-sient voltages and currentwaveforms at Tj= 27 °C depicted in Fig. 2a–breveal the goodness of the improved model respect to NIGBT and man-ufacturer models.

Moreover, the effectiveness of new model is also confirmed bycomparison of simulated IGBT blocking characteristics (Fig. 3a) and in-ductive turn-off transient collector current waveforms (Fig. 3b) at in-creasing operative temperature with experimental ones.

3. Short-circuit test ET simulation

The first circuit application considered as a case-study for newPSpice ET model validation, is the standard short-circuit (SC) test, thatis usually performed by device users in order to verify its capability tosustain both high-voltage and high-current values, before destructivethermal runaway event occurs. The simplified schematic of experimen-tal short-circuit test setup used for ET simulations is depicted in Fig. 4. A

Fig. 1. Benchmark performance analysis on PSpice IGBT models: Manufacturer (Krausbased), NIGBT (Hefner based) and proposed model.

Please cite this article as: D. Cavaiuolo, et al., A robust electro-thermal IGBMicroelectronics Reliability (2015), http://dx.doi.org/10.1016/j.microrel.2

Foster equivalent thermal network provided by the device manufactur-er on the device datasheet, was used in PSpice simulation in order to in-vestigate the transient behaviour of junction temperature.

The test conditions are as follows: VGE = −5/15 V, RG = 10 Ω andTj0= 27 °C. Fig. 5 reports the comparison between experimental wave-forms of DUT short-circuit collector current and collector-to-emittervoltage during an SC pulse of τSC = 5 μs, since it's guaranteed on thedatasheet as themaximum short-circuit time interval before device fail-ure. The results prove that the newPSpicemodel is capable to accuratelypredict the ET behaviour for the DUT.

4. Short-circuit protection circuit design

As a case study for the model validation, we consider an inverter formotor-drive application, where two main kinds of short-circuit condi-tions may occur: a shoot-through of the leg or a short of the load(Fig. 7a). Nevertheless, since experimental characterization of IGBT SCin application is very tricky to deal with, the effectiveness of themodel in this condition is proved by means of a simplified-equivalentexperimental circuit. In a step-down converter, a short is forced on theload and a SC event occurs on the conducting IGBT, enabling thedesaturation protection by the driver HCPL-316J, provided by AvagoTechnologies®.

In this case the well-known desaturation method is considered,which consists on the sense of device saturation collector-to-emitter



Fig. 3.Model comparison on a) IGBT blocking characteristic and b) simulated-measuredIGBT turn-off collector current inductive turn-off IGBT current waveform at Vcc = 400 V,Ic = 30 A, Rg = 10 Ω and L = 250 μH VS Tj.

Fig. 5.DUTmeasured and simulated a) current and b) voltagewaveforms during SC test atτSC = 5 μs, VCC = 360 V, VGE = −5 V, RG = 10Ω, Tj0 = 27 °C.

3D. Cavaiuolo et al. / Microelectronics Reliability xxx (2015) xxx–xxx

Vce,sat voltage. During normal condition, IGBT operates in linear region,so its collector-to-emitter voltage is small: if a short-circuit eventsoccurs, its collector current IC abruptly arises till saturation region,where the VCE voltage becomes larger. When the value of VCE exceeds

Fig. 4. PSpice OrCad equivalent schematics of short-circuit test experimental setup.


a defined threshold value, the digital control detects a “fault condition”and forces the “soft” turn-off of the IGBT. The sense circuit is depicted inFig. 6 and the logic circuits needed to achieve device protection are allintegrated within the IGBT gate-driver. However, it is required the de-sign of some external desaturation circuit parameters such as Vce,th,

Fig. 6. Simplified schematic of desaturation circuit Vce detection by means Vdesat voltage.



Fig. 8.Measured and simulated IGBT current–voltagewaveforms during short-circuitwithLwires = 1.9 μH and driver “soft” gate turn-off control.

4 D. Cavaiuolo et al. / Microelectronics Reliability xxx (2015) xxx–xxx

Cblank or Rpull-down according to IGBT characteristics in order to preventdevice from critical SC electro-thermal conditions which may affect itsreliability and lifetime. The typology of SC event is classifiable as a HSF(“hard switched fault” [14]) and the total path inductance is quitehigh (Lwires = 1.9 μH) due to the wires of the link to the motor. The cir-cuit (Fig. 7b) is simulated in PSpiceOrCad® environment and the equiv-alent Foster model of device-package thermal network is taken fromdevice manufacturer datasheet, assuming Tcase = Tamb. The simulatedIGBT current–voltage waveforms using the new model (Fig. 8) show abetter accordance with experimental data respect to the NIGBT model,while themanufacturer providedmodel is not even able to achieve con-vergence in this kind of electrical network.

The electro-thermal validation of the model was performed in sta-tionary conditions. More in detail, the step-down converter topology(see Fig. 7b) was analysed in non-fault conditions and the simulatedmean value for ΔTj was compared with the experimental measure-ments (a hole was made, through the plastic coverage, to achieve thedie surface in order to directly measure the Tj mean value).

In particular, the steady-state device junction temperature reachesabout Tj≈ 55 °C (bothmeasured and simulated), proving that the tem-perature dependence on IGBT saturation current value is properly tak-ing into account with new model. On the other hand, the differencesin gate voltage behaviour, especially after desaturation sensing, aredue to “soft” dynamic turn-off control of the gate by the driver, not suit-ably reproduced by the NIGBT model. Hence, the proposed IGBT modelallows to accurately predicting device behaviour under SC conditions,therefore it is possible to optimally designed desaturation circuit pa-rameters when the device experiences HSF or FUL SC events. In detail,the choice of Cblank capacitance value may be critical especially if thecase of short on the leg (SC1) is considered. The Cblank valuemust ensurethat the VCE(t) voltage fall-downunder a given threshold voltage duringIGBT turn-on and therefore the disabling of protection.

On the other hand, when the path inductance is small and thecurrent rise-time become very short, a delay on circuit protection re-sponse is present. A formula for Cblank rough evaluation is given ingate-driver datasheet: once a tblank time before fault detection is de-fined, the Cblank is estimated from the internal desaturation currentIch and the desaturation voltage threshold Vdesat according to the fol-lowing expression:

Cblank ¼ tblankIchVdesat

: ð3Þ

Fig. 7. a) Twomain typologies of short occurring in an inverter (short of the load or of a deviceproposed electro-thermal IGBT model, HCPL316J® driver model (with desaturation protection


Usually a practical value of tblank equal to 1.5 μs is recommended toavoid protection circuit false activation, and the corresponding Cblank

is 47 pF, if Ich = 250 μA and Vdesat = 7 V are considered. This designmethod do not consider the self-heating effect and the device phys-ics specifications. With this simplified design of the SC protection cir-cuit, the device may experiences large ΔTj during short-circuit timeinterval. This appears clear from simulation results of an HSF IGBTshort-circuit event, where the higher is the Cblank value for avoidingfalse activation of desaturation circuit, the higher is also the time in-terval before protection intervenes when SC occurs, with the conse-quence of abrupt increase of device junction temperature (Fig. 9).Another example of design optimization using electro-thermalSPICE IGBT model regards the wise choice of Rgoff when the drivercircuit does not provide an embedded “soft” control of the gate volt-age turn-off (Fig. 10).

The Rgoff value can be accurately selected in order to limit the over-voltage due to stray inductances when turning-off high short-circuitcurrent according to device speed characteristics (tail current and life-time). Using a fast and accurate IGBT ET model we can predict the

in a leg); b) PSpice simulated load-short (Lwires = 1.9 μH) in a step-down converter, using) and device Foster equivalent thermal network.



Fig. 9. Simulated desaturation tblank and IGBT junction temperature after SC versus Cblank.Fig. 10. Simulated IGBTVCE peak-voltagewith RGoff for different values of stray inductance.

5D. Cavaiuolo et al. / Microelectronics Reliability xxx (2015) xxx–xxx

behaviour of collect-emitter peak voltage versus Rgoff for differentvalues of stray inductance, when a shoot-through occurs to a singleleg of an inverter (Fig. 6). For instance, in the analysed case the devicecan experience a breakdown avalanche event for a Rgoff b 25 Ω under200 nH stray inductance.

5. Conclusions

An optimized ET IGBT SPICE model based on the Kraus model wasdeveloped and experimentally verified on a Field-Stop trench-gate30 A–600 V commercial IGBT device. The application to the short-circuit protection circuit design for a hard switched fault in an invertermotor-drive applicationwas also analysed. The results showed an accu-rate prediction of the device electro-thermal behaviour under short-circuit conditions,with a possible optimal design of thedesaturation cir-cuit parameters when the device experiences hard-switched-fault orfault-under-load events.

Acknowledgement

This work was partly supported by the ItalianMinistry of University,Education and Research (MIUR) under the Virtual Energy Managementproject (PON 01_02754).


References

[1] Rodrìguez-Blanco, et al., Industrial electronics, IEEE transactions on, 58.5 2011,pp. 1625–1633.

[2] W. Zhiqiang, et al., Applied Power Electronics Conference and Exposition, IEEE APEC2013, pp. 577–583.

[3] M. Arab, et al., Power Electronics Specialists Conference, 2008. PESC 2008. IEEE. IEEE2008, pp. 4355–4360.

[4] I.F. Kovacevic, et al., Power Electronics Conference (IPEC), 2010 International. IEEE2010, pp. 2106–2114.

[5] A. Ammous, et al., Power electronics, IEEE transactions on, 15.4 2000, pp. 778–790.[6] V. d'Alessandro, et al., In: Microelectronics reliability volume 53 (issues 9–11)

(2013) 1713–1718.[7] A.R. Hefner, et al., In: Power electronics, IEEE transactions on 9 (5) (1994) 532–542.[8] M. Miyake, et al., In: Electron devices, IEEE transactions on vol. 60 (no. 2) (2013)

571–579.[9] M. Cotorogea, In: Power electronics, IEEE transactions on vol. 24 (no. 12) (2009)

2821–2832.[10] R. Kraus, et al., In: Power Electronics Specialists Conference, 1998. PESC 98 Record.

29th Annual IEEE vol. 2 (1998) 1726–1731 vol. 2.[11] R. Azar, et al., Industry applications, IEEE transactions on, 40.3 2004, pp. 710–716.[12] E. Santi, et al., Industry applications, IEEE transactions on, 40.2 2004, pp. 472–482.[13] D. Cavaiuolo, In: Power electronics, electrical drives, automation and motion

(SPEEDAM) (2014) 133–138.[14] M. Oinonen, et al., Power Electronics and Applications (EPE'14—ECCE Europe), 2014

16th European Conference on. IEEE 2014, pp. 1–9.


http://refhub.elsevier.com/S0026-2714(15)00178-X/rf0005

























a robust electro-thermal igbt spice model:...

Documents