Measurement of dynamic characteristics of 1200A/ 1700V IGBT-modules under worst case conditions

Measurement of dynamic characteristics of 1200A/ 1700V IGBT-modules under worst case conditions M. Helsper Christian-Albrechts-University of Kiel Faculty of Engineering Power Electronics and Electrical Drives Kaiserstr. 2 24143 Kiel Tel. +49 431 77572 355 Fax. +49 431 77572 353 mth@techfak.uni-kiel.de F. W. Fuchs Christian-Albrechts-University of Kiel Faculty of Engineering Power Electronics and Electrical Drives Kaiserstr. 2 24143 Kiel Tel. +49 431 77572 350 Fax. +49 431 77572 353 fwf@techfak.uni-kiel.de R. Jakob ALSTOM Power Conversion GmbH Culemeyerstr. 1 12277 Berlin Tel. +49 30 7496 2393 Fax. +49 30 7496 2303 roland.jakob@powerconv.alstom.com ABSTRACT A test stand for measuring dynamic characteristics of power semiconductor devices has been developed. The test stand is introduced and the measurement system is explained. Investigations by measurement of the dynamic behaviour of 1200 A/ 1700 V IGBT modules made by Mitsubishi Electric are presented and analysed, with special regard to the behaviour under worst case and worst case fault conditions. The measurements give information about the properties of the modules under high stress conditions. Keywords: IGBT, Diode, measurement, dynamic characteristics, short circuit. 1 INTRODUCTION Because of the continuous further development in the field of power semiconductor devices there is often a request for a characterisation of the properties of new types of power semiconductor devices. To perform a contribution to this information request, we have developed a test stand. This test stand is used to determine static parameters and especially dynamic parameters of power semiconductors such as IGBTs, diodes or MOSFETs. The aim of the characterisation is to test the semiconductors in terms of their qualification for applications in the area of electrical drives. Here exist high requirements for instance concerning switching, short circuit, parallel connection or high temperature. The information in the manufacturers data sheets is often insufficient for all possible working conditions, so additional measurements are needed. With the test stand we have investigated the dynamic behaviour of 1200A/ 1700V IGBT-modules of the type CM1200HA-34H, made by Mitsubishi Electric. The measurements are realized according to a measurement program that was developed corresponding to industrial applications. Aiming at high power applications, two IGBTs have been connected in parallel. Special focus was here to investigate the behaviour of the modules under worst case conditions. 2 TEST STAND The test stand principle schematic is shown in figure 1. The test stand consists of a supply circuit, a test circuit and a variable load. The supply circuit with a step up converter is used to provide the capacitance C ZK of the test circuit with the requested test voltage. If there are switching tests, the step up converter is switched off. One switching test takes only a very short time so the voltage V ZK keeps nearly constant. D1 D3 D5 S GR T LIM L HS D HS I HS T1 T2 L L 1 C ZK V ZK C GR V GR D LIM T HS L 2 L 3 D4 D6 D2 V GE3 V CE3 a) b) I C3 I C4 I L transformer rectifier step up converter test circuit with modules T1 - load supply circuit Fig. 1. Principle schematic of the test stand

The topology of the test circuit corresponds to that of a buck converter, whereas here the IGBT modules T1 and T2 respectively and are controlled together. The test circuit is built on an aluminium plate, which could be heated. So it is possible to control the temperature. The connections between the modules and the capacitance C ZK are realized by using a laminated bus bar, which consists of copper plates and a thin insulating material. This configuration leads to a very low stray inductance. The load is a self-wound coil or a shorting bar. Industrial drivers with over voltage and short circuit protection are used for the control of the IGBT-modules. Figure 2 shows the principle schematic of the driver. For the high current modules a higher driver power is necessary than one of the used standard pulse amplifiers can supply. That s why two coupled pulse amplifiers are used to generate the gate emitter voltage for the parallel IGBTs. current is measured by current probes. It is a big advantage that for this kind of current measurement no additional stray inductance has to be accepted in contrast to a measurement e.g. with a current transformer. But it is to note that no dc-current can be measured in this way. Figure 3 shows a comparison between a collector current measurement with a special coil and a current transformer (Pearson Current Monitor) [3]. Additional the load current is to be seen. This measurement is made in another test circuit with single modules at inductive load and a higher stray inductance through the current measurement loop. The measurement with the current transformer is used as a reference according to the high accuracy of this system [4]. +15V V CE-Control E-Signal 0V Pulse amplifier A1 C1 D 1 ZD 1 V1/C G1 E1 D 2 R G ZD 2 R E ZD 3 V1/G V1/E Pulse amplifier A2 C2 D 3 ZD 4 V2/C G2 D 4 V2/G R G ZD 5 E2 R E ZD 6 V2/E Fig. 2. Principle schematic of the driver circuit for two parallel IGBTs The short circuit protection is realized through a monitoring of the saturation voltage V CEsat of the modules. For this purpose the collector emitter voltage of every IGBT is detected with high voltage diodes (D 1 and D 2 ) and inside the amplifier compared with a reference voltage. If the reference voltage is exceeded a short circuit is triggered and all active IGBTs are switched off. An active clamping circuit is used for the overvoltage protection at switch off of the IGBTs, represented in figure 2 through a zener diode between the collector and the gate. If the collector emitter voltage exceeds the breakthrough voltage of the zener element the gate voltage is increased. The IGBT is working in the active area of his output characteristic and converts the energy of the stray inductivity into heat. So, this mode leads to high power losses in the IGBT and is usually used only for worst case or failure conditions [1]. 3. MEASUREMENT SYSTEM The slowly changing current through the inductive load is measured with a High Current Probe. The voltages are measured by using voltage probes. A digital scope is used for the display of the measured values. The measurement of the collector or diode currents during the switching interval is realized by using a special measurement system. Special coils which look like a pencil are fixed directly in a trench at the surface of the case of the IGBT-modules between the main connections of the collector and the emitter. The variable magnetic field caused through the fast changes of the semiconductor currents at switching induces a voltage in the coils. This causes a current flow in the closed coils [2]. This Figure 3. Switch on of an IGBT-modul Measurement of I C with special coils (Ch 4) and a current transformer (Ch. 3) V CE 100 V/DIV Ch. 1 I C 100 A/DIV Ch. 3 and 4 I Load 100 A/DIV Ch 2 Figure 3 shows that the special coil gives a correct image of the switching behaviour in terms of the switching quality. The measurements differ mainly only in the quantity of the measured signals. The measured current with the special coil is up to 9 % over the measured current with the current transformer. This is caused through a not fully correct scale for this current, which is determined mainly by the numbers of windings of the coil. That s why for all investigations also the load current is measured. It is used to verify the quantity of the measured collector current a short time before and after switching. Additional the measurement with the special coil shows a time delay in comparison to the measurement with the current transformer of approx. 20 ns. This is mainly caused by the amplifier of the current probe, which is used for the measurement of the current in the special coil. For the purpose of a determination of mainly characteristics of IGBT modules under certain application conditions the measurement with the special coils in connection with a measurement of the load current gives a sufficient image of the real collector currents.

4 TEST CONDITIONS High current modules have usually main and auxiliary connections. The auxiliary connections are used to connect the module with the driver on a very low inductive path. Figure 3 shows the test circuit and the used power semiconductors with relevant parasitic elements in the test mode. V ZK = Fig. 4. C3' G3 E3' L ZK1 L ZK2 R Gi E1 C3 L Gi C1 E3 L C2 L C1 L D1 D1 L E1 L E2 C4' G4 E4' R Gi E2 C4 L Gi C2 E4 L D2 D2 L C2 L C1 L E1 L E2 Principle schematic of the test circuit with parasitic elements at inductive load The voltage measurement at the main connections of the module includes the voltage drops at all parasitic inductivities of the module in the collector emitter path while a voltage measurement at the auxiliary connections gives a more accurate image of the real collector emitter voltage of the IGBT. The parasitic inductances between the main and the auxiliary connections are the most important for the collector emitter path of the module [5]. Figure 5 shows both collector emitter voltages at auxiliary and main connections at switch off. I load Also at switch on the influence of the module inductivity has to be respected. The safe operating area diagrams from the data sheets [6] of the measured modules are given only for the main connections. That s why the main attention is spent to measurements on these connections. Parallel measurements on the auxiliary connections are also realized. The investigations of hard switch on and off of IGBTs and diodes were performed at inductive load in the double pulse mode. Because of very short test pules (t puls < 60 µs) and currents up to the rated current there is no important self heating in the modules and the temperature of the cooling plate could be used as the reference temperature. For the short circuit tests the single pulse method was used. During these tests the chip temperature of the power semiconductor increases strongly caused by the high over current. If the IGBT works as here in his active area also at very short pulses (<10µs) the influence of the increased junction temperature has to be respected [7]. The temperature of the cooling plate is only the start value of the junction temperature at this test. To include the temperature depending behaviour of the IGBTs the measurements were performed at 25 C and at 120 C. The test voltage V ZK was controlled up to 1200V. For the industrial applications this is typically the worst case for the dclink voltage of the inverter. The driver output level was 15V for the on-state and 5V for the off-state. The gate resistance for each IGBT was after several tests fixed in terms of the existing requirements of the application to 2.45 Ω. 5 DYNAMIC BEHAVIOUR UNDER WORST CASE CONDITIONS In this part measurement results of the investigation of the switch on and off of the IGBTs and diodes of the modules are presented. The tests are performed at the worst case conditions for the dc-link voltage V ZK and nearly the rated current for every module. Figure 6 shows a switch on and figure 7 a switch off of the parallel IGBTs and measured on the main connections. Every module carries half of the whole load current in the conduction phase. Also in the switching phase there is a good symmetry between the currents of the parallel modules. Figure 5. Switch off of an IGBT-module I L = 1600A, V ZK =900V, T A =20 C Measurement of V CE at the main- and auxiliary connections V CE 200 V/DIV Ch. 1, Ch. 2 I C 250 A/DIV Ch. 3 and 4 It is to be seen that the collector-emitter voltage peak measured at the main connections is lower than at the auxiliary connections. caused through the module inductivity. The measurement shows that there is to note a difference between the behaviour of a whole IGBT module and the behaviour of the power semiconductor alone. At switch on the over current caused by the reverse recovery of the freewheeling diode is very soft and the peak current is quite low. The drop of the recovery current after its peak value is rather slow and a significant tail current can be observed. The switch off at worst case conditions shown in figure 7 is influenced by the over voltage protection. This is characterized by the second peak and the delayed drop of the over voltage. The active clamping of the collector emitter voltage limits the over voltage to values under 1450 V. The IGBT stays safely inside the RBSOA (reverse biased safe operating area) during the total switch off. The switch off time is increased opposite to the unclamped conditions. This causes higher switch off losses and so this mode should not be used periodically.

Because of the very short switch on time the diode has not reached its stationary state at the beginning of the switch off. This can support the tendency to an unstable reverse recovery of the diode. At the test conditions the investigated diodes show an uncomplicated behaviour. But in comparison with figure 6 a some stronger reverse recovery can be observed here. Fig. 6. Switch on of the bottom IGBTs I L = 2200A, V ZK =1170V, T A =20 C Measurement of V CE on the main connections V CE 400 V/DIV Ch. 1 I C 500 A/DIV Ch. 3 and 4 V GE 10 V/DIV Ch. 2 Fig. 8. Short switch on of the reverse diodes of T1 and T2 I L =2200A, V ZK =1170V, T A =120 C Measurement of V D1 on the main connections and of the diode currents I D1 and I D2 V D 500 V/DIV Ch. M3 I D 250 A/DIV Ch. 3 and 4 6 DYNAMIC BEHAVIOUR UNDER WORST CASE FAULT CONDITIONS Two typical kinds of failures are investigated. At first a short circuit of the load (case 1) and after this a short circuit of the branch between the poles of the dc-link (case 2). Reasons for the first case could be an isolation failure or wrong connections. A failure of the case 2 results for instance from a defect module or a failure in the control. Fig. 7. Switch off of the bottom IGBTs I L = 2400A, V ZK =1170V, T A =20 C Measurement of V CE on the main connections V CE 200 V/DIV Ch. 1 I C 250 A/DIV Ch. 3 and 4 The reverse recovery characteristics of a diode depend on internal parameters such as minority carrier lifetime, doping profiles, etc. The diode behaviour is also influenced by external parameters, which control the circuit operation. These are the forward current, the stray inductance of the circuit, the reverse voltage, the commutating di/dt and junction temperature [8]. Low inductive circuits, fast switching IGBTs, partly very short switch on times and the worst case conditions bring high requirements on the diodes for the investigated application. That s why a special diode test was realized. The switch off of the reverse diodes (modules T1 and T2) after such a test with a very short switch on time is shown in figure 8. This test is realized by using a double pulse with a very short break between the two pulses, which control the IGBTs and. The qualification for the modules here is that there is a good symmetry between the parallel diodes and no snappy recovery connected with undesirable oscillations at diode switch off. C ZK T1 T2 Case 2 Figure 9. Kinds of tested short circuits Case 1 The short circuit of the case 2 shown in figure 10 is a very strong test for the modules. At first IGBTs and are switched on. After the load current has risen up to the nominal value of 2300A, the short circuit starts with the additional switching on of T1 and T2. Only the low stray inductance s of the modules and the circuit limit the rise of the collector currents. Figure 10 shows that the modules limit the current by themselves at an amplitude in the range of 6 ka. The short circuit current decreases by time. This is caused by the self heating of the modules and the property of the current measurement system to transmit no dc-current. The modules T1 L l

and T2 lead the short circuit current, the modules and additional the load current. Figure 11 shows a switch off of short circuit case 2 of IGBT T2 in V CE - I C coordinates. At the begin of the switch off the collector emitter voltage increases up to a voltage peak of approx. 1490 V. This is caused through a something delayed reaction of the clamping circuit. After this peak the switch off overvoltage is clamped at nearly 1400 V. This representation gives the possibility of a precise analysis of the current and voltage values at switching and can be used also for the dimensioning of the driver and clamping circuits. Other measurements were performed in terms of the short circuit of case 1. Here a shorting bar (1µH, see figure 1) is used as load. Because of the higher inductance and a very fast short circuit switch off through the short circuit protection of the driver there is not such a stress as in the short circuit of case 2. 7 CONCLUSION Fig. 10. Short circuit of case 2 V ZK =1170V, T A =120 C, pre current I L =2300 A Measurement of V CE T2 on the main connections and the collector currents I C1 and I C2 V CE 500 V/DIV Ch. M1 I C 1,25 ka/div Ch 3 and 4 In the on state the collector emitter voltage on T1 and T2 is nearly the whole dc-link voltage. A small voltage drops about the stray inductance s when current changing. Another voltage part is carried from the IGBTs and, which are at the beginning of the short circuit in a saturated on state. After 6µs the not saturated IGBTs T1 and T2 switch off the short circuit at first, controlled by the active clamping circuit. This is a very critical operating point for these modules. After the upper IGBTs switched off the bottom IGBTs and lead only the inductive load current until they are switched off through the control. Then the reverse diodes of the switches T1 and T2 take over the freewheeling current. This is also to seen in figure 10 when the measured current falls below a plateau to the bottom range of the figure. Power semiconductors have been tested on a special test stand, which was developed for this purpose. The dynamic behaviour of 1200A / 1700 V IGBT modules has been measured. The measurements were performed according to worst case and worst case fault conditions in electrical drives for industrial applications. The measurements show a correct behaviour of the modules also at strong test conditions. The pass of this tests give a high safety for a cardinally function of the modules in the application. Moreover the properties of the modules must be tested in relation to the operation under real working conditions, for instance pulse width modulation mode. REFERENCES [1] SEMIKRON INTERNATIONAL, U. Nicolai, Applikationshandbuch IGBT- und MOSFET- Leistungsmodule, 1. edition, ISLE-Verlag, Ilmenau, 1998. [2] R. LAPPE, F. Fischer, Leistungselektronik Meßtechnik, Verlag Technik München, 1993. [3] C. WATERS, Current Transformer Provide Accurate Isolated Measurements, Power conversion & intelligent motion, no 12,.1986. Fig. 11. Switch off at a short circuit of case 2 in x-y coordinates V ZK =1170V, T A =120 C, pre current I L =2300 A X-axis: V CE T2 200V/DIV Y-axis: I C T2 1,25 ka/div [4] PEARSON ELECTRONICS, Inc., Data sheet model 2879 current monitor. [5] F. REDDIG, Untersuchungen zum Schaltverhalten von IGBT-Modulen und zur Ermittlung der Schaltverluste, Dissertation University of the Bundeswehr Munich, 1997. [6] MITSUBISHI ELECTRIC; Data sheets CM1200HA- 34H. [7] P. NANCE, M. MÄRZ; Thermal Modelling of Power Electronic Sytems; PCIM Europe Power Electronics, no 2 2000, pp. 20-27. [8] N. Y. A. SHAMMAS; Effects of external Operating Conditions on the reverse Recovery Behaviour of Fast Power Diodes; EPE-Journal Vol. 8 no 1-2, 1999, pp. 11-18.