Download Using 650-V High Speed 3 IGBTs in Power Modules for... ter Performance Improvement

PCIM Europe 2013, 14 – 16 May 2013, Nuremberg
Using 650-V High Speed 3 IGBTs in Power Modules for Solar Inverter Performance Improvement
Christian R. Müller, Infineon Technologies AG, Max-Planck-Straße 5, D-59581 Warstein,
Germany
Jens De Bock, Infineon Technologies AG, Max-Planck-Straße 5, D-59581 Warstein, Germany
Abstract
For applications which require an output filter or have a booster/buck choke, higher switching
frequencies can lead to advantages on system level. One application where both can be
found is a solar inverter. At the state-of-the-art efficiency and power density, high cost pressure can be observed for solar inverters. The High Speed IGBT is optimized for highfrequency hard-switching applications [1]. Therefore, this device is an ideal choice for power
modules which are used in solar applications.
In this paper, the difference between the 650-V IGBT3, 650-V IGBT4 and the 650-V High
Speed 3 (HS3) IGBT from Infineon Technologies used in power modules will be described. It
will be shown that due to the device design, the 650-V HS3 IGBT provides a superior performance and can be used as a high-efficiency switch. The requirements to achieve this high
efficiency by using the 650-V HS3 IGBT will be explained as well.
1.
Introduction
1.1.
Application-specific device design of IGBTs
The trench-field-stop technology is the most common concept for modern IGBTs with blocking voltages in the range of 600 V to 1200 V. This technology allows to realize devices with
low on-state voltages and high switching softness on the one hand and low switching losses
and a MOSFET-like switching performance for high-frequency applications on the other
hand. Within this technology, the device performance is mainly controlled by design parameters like cell geometry, chip thickness, and doping profile. These parameters determine the
carrier distribution in the device and, hence, influence the static and dynamic characteristics
of the IGBT significantly.
In principle, the performance of an IGBT is determined by a trade-off curve which describes
the correlation between the turn-off energy (EOFF) and the collector-emitter saturation voltage
(VCE sat). Devices with a high carrier density in the drift region provide a low VCE sat and allow
small static losses. During the turn-off, a high carrier density leads to a slower clear out of the
device and the dynamic losses are increased. Therefore, the performance of an IGBT can be
either optimized for high-frequency applications like solar inverters or boosters, which need
devices with low dynamic losses, or for low-frequency applications like drives, which benefit
from low static losses.
ISBN 978-3-8007-3505-1 © VDE VERLAG GMBH · Berlin · Offenbach
896
PCIM Europe 2013, 14 – 16 May 2013, Nuremberg
Fig. 1. Left-hand side: VCE sat-EOFF trade-off curve for 650-V IGBTs. On the right-hand side, typical
turn-off waveforms of the IGBT3 (top) and the HS3 IGBT (bottom) are shown with: VCE: black,
IC: red, and VGE: green.
In Fig. 1, the trade-off curve is shown for 650-V IGBTs. As mentioned above, EOFF decreases
with increasing VCE sat. The IGBT3 and the HS3 IGBT are highlighted on the trade-off curve.
As one can clearly see, the HS3 IGBT is optimized for high-frequency applications due to its
high collector-emitter saturation voltage VCE sat = 1.85 V and low turn-off energy EOFF =
0.6 mJ. The IGBT3 provides a more than a factor of 2 larger turn-off energy EOFF = 1.4 mJ,
whereas the VCE sat is reduced to 1.45 V. The right-hand side of Fig. 1 displays typical turn-off
waveforms of the IGBT3 (top) and the HS3 IGBT (bottom). For both measurements, the following parameters were used: DC-link voltage VDC = 400 V, collector current IC = 50 A, gate
resistor RG ȍDQGMXQFWLRQWHPSHUDWXUH7vj = 25 °C. The devices were operated with an
identical test setup with a stray inductance Lı = 25 nH. Compared to the IGBT3, IC decreases
much faster in the HS3 IGBT which results in a larger over-voltage peak on the collectoremitter voltage (VCE). In addition, the collector current snaps off. This leads to parasitic oscillations on the gate-emitter voltage (VGE) and on VCE. Hence, the switching softness of the
HS3 IGBT is lower.
Both IGBTs have an identical short-circuit robustness with a short-circuit-withstand time t SC =
6 µs at Tvj = 150 °C. This value is mainly limited by the thermal capacitance of the chip. For
applications which need an enhanced tSC, the IGBT4 was developed with an increased chip
thickness. This enlarged chip thickness raises tSC up to 10 µs at Tvj = 150 °C and improves
the switching softness. However, this also results in larger dynamic losses and a slightly increased VCE sat.
1.2.
Electrical setup
For the measurements, a 650-V IGBT3, a 650-V IGBT4, and a 650-V HS3 IGBT with a nominal collector current of 50 A were used. The electrical performance of the chips is determined by measuring the switching losses. Therefore, each chip was integrated in a power
module (EasyPACK 2B) with an identical electrical circuit and a stray inductance of 17 nH.
Due to the fact that the turn-on losses (EON) are mainly dominated by the used free-wheeling
diode, all chips were operated with a 650-V emitter-controlled diode with a nominal current IF
= 30 A.
Unless otherwise specified, all measurements were performed in the lab under the following
conditions:
A setup with an integrated current probe and a stray inductance Lı = 25 nH was used. The
DC-link voltage was set to VDC = 400 V, which is a typical voltage in the application, and the
ISBN 978-3-8007-3505-1 © VDE VERLAG GMBH · Berlin · Offenbach
897
PCIM Europe 2013, 14 – 16 May 2013, Nuremberg
chips were operated at nominal collector current, i.e., IC = 50 A. To drive the IGBT, a gateemitter voltage VGE = ±15 V was used. All measurements were performed at T vj = 25 °C.
2.
Switching behavior and chip comparison
The switching behavior of the chips is measured with the above mentioned setup. The corresponding energies and the characteristic switching parameters are extracted from the turnon and turn-off waveforms. Fig. 2 shows EOFF, EON, and the current slope di/dt during the
turn-on for HS3 IGBT, IGBT3, and IGBT4. For the measurement, RG ȍZDVXVHGLQWKH
gate-driver circuit.
Fig. 2. Comparison of the switching energies (EON and EOFF) and di/dt during turn-on for HS3 IGBT,
IGBT3, and IGBT4. Inset: Total switching losses Etotal of the IGBT.
It can be observed that the HS3 IGBT provides the lowest turn-off losses. When considering
the turn-on losses, both, IGBT3 and IGBT4, show a better performance than the HS3 IGBT.
The difference of the overall switching losses Etotal, which is defined as the sum of EON and
EOFF, is very small if the HS3 IGBT is compared with the IGBT3. It is remarkable that the di/dt
of the HS3 IGBT is a factor of 2 smaller than the di/dt of the IGBT3.
Fig. 3. dv/dt during turn-off (left-hand side) and di/dt during turn-on (right-hand side) versus the gate
resistor RG in the gate-driver circuit for HS3 IGBT, IGBT3, and IGBT4.
ISBN 978-3-8007-3505-1 © VDE VERLAG GMBH · Berlin · Offenbach
898
PCIM Europe 2013, 14 – 16 May 2013, Nuremberg
To gain an insight on dv/dt and di/dt during the turn-off and turn-on, respectively, Fig. 3
shows dv/dt and di/dt for different RG. For the HS3 IGBT two major differences can be found:
The dv/dt is approximately 50 % larger than for IGBT3 and IGBT4, whereas the di/dt during
the turn-on is a factor of 2 smaller compared to the IGBT3. Both observations lead to the
conclusion that a comparison of the switching behavior is only sensible if RG in the gatedriver circuit is modified. Target of this modification is to realize equal dv/dt and di/dt during
the turn-off and turn-on, respectively, for HS3 IGBT, IGBT3, and IGBT4. In case of the HS3
IGBT, RG has to be reduced significantly for the turn-on and increased for the turn-off.
Fig. 4. Comparison of the switching energies (EON, EOFF, and Etotal) with equal di/dt during turn-on and
dv/dt during turn-off for HS3 IGBT, IGBT3, and IGBT4. For the measurement, di/dt = 1.5 kA/µs
and dv/dt = 4.5 kV/µs were set by adjusting the RG for the turn-on and the turn-off. Inset: EON
and di/dt of the HS3 IGBT versus RG.
Fig. 4 displays the switching losses for HS3 IGBT, IGBT3, and IGBT4 for equal switching parameters. RG was modified in the way that di/dt = 1.5 kA/µs and dv/dt = 4.5 kV/µs were
achieved during turn-on and turn-off, respectively. Therefore, the HS3 IGBT has the lowest
switching losses (EON and EOFF) and, in sum, Etotal is more than a factor of 2 smaller compared to the IGBT3. The inset shows EON and di/dt of the HS3 IGBT versus RG. EON increases and, in turn, di/dt decreases with increasing RG. Especially for RGȍGLGW!N$—V
can be realized whereas larger RG lead to di/dt below 0.5 kA/µs.
The small turn-off losses indicate the superior switching performance of the HS3 IGBT and
are attributed to an optimized carrier distribution within the device. Therefore, the HS3 IGBT
is optimized for high-frequency applications and, with respect to the trade-off between E OFF
and VCE sat, provides small dynamic losses. The high turn-on losses of the HS3 IGBT are attributed to the very low di/dt compared to IGBT3 and IGBT4. To compensate this, it is necessary to decrease the turn-on gate resistor significantly. One possible way for realization is to
use a more sophisticated gate-driver design which allows to operate the HS3 IGBT as a very
efficient switch.
ISBN 978-3-8007-3505-1 © VDE VERLAG GMBH · Berlin · Offenbach
899
PCIM Europe 2013, 14 – 16 May 2013, Nuremberg
3.
HS3 IGBT under operating conditions
In the last chapter, it was shown that the HS3 IGBT outperforms IGBT3 and IGBT4 in highfrequency applications. The upcoming question is how the HS3 IGBT performs under operating conditions. Under typical operating conditions, the HS3 IGBT will mainly be operated with
collector currents smaller than the nominal chip current. In addition, also the DC-link voltage
may vary over a large voltage range. Therefore, the switching losses of the HS3 IGBT are
analyzed for DC-link voltages in the range of 150 V to 450 V and for collector currents up to
the nominal chip current.
Fig. 5. Switching energies (EON and EOFF) of the HS3 IGBT versus the DC-link voltage for IC = 10, 30,
and 50 A. Inset: Etotal of the HS3 IGBT versus the DC-link voltage for IC = 10, 30, and 50 A.
For the measurement, RG ȍZDVXVHGLQWKHJDWHGULYHUFLUFXLW)LJVKRZVWKHVZLWFK
ing losses of the HS3 IGBT versus the DC-link voltage. For small VDC, EOFF is low and increases linearly with increasing VDC, whereas larger collector currents result in higher turn-off
losses. In contrast to this, it is found that EON increases superlinearly with VDC and IC as well.
For IC = 10 A, the slope of EON versus VDC is almost constant. At IC = 30 and 50 A, a steeper
slope for VDC•9LVREVHUYHG,QWKHLQVHWWKLVVXSHUOLQHDULQFUHDVHLVDOVRLGHQWLILHGIRU
Etotal.
These measurements show that, compared to the turn-on losses, the turn-off losses of the
HS3 IGBT have minor impact on the device performance. Due to the superlinear increase of
the turn-on losses for VDC • 9 DW ,C • $ WKH KLJKHVW HIILFLHQF\ FDQ EH DFKLHYHG DW
small collector currents. The high turn-on losses for larger VDC and IC are attributed to a reduction of the di/dt. This effect is characteristic for the HS3 IGBT and is related to the carrier
distribution, which is determined by the device design.
One way to compensate this effect is the downsizing of RG, which, in turn, leads to a decrease of the softness. To avoid current snap off in the free-wheeling diode, one possible solution is to use diodes with a larger nominal current. As a result, the switching losses increase due to the enhanced diode-recovery current. Another and more promising solution is
the implementation of fast-frequency-optimized diodes. SiC Schottky diodes offer the possibility to overcome the current snap off in the free-wheeling path. In addition, the turn-on
losses are lowered due to the missing diode-recovery current. By using SiC diodes in the
free-wheeling path, EON can be reduced from 2.5 mJ to 1.2 mJ for VDC = 400 V, Tvj = 25 °C,
and IC = 50 A.
When fast-switching devices are used, an upcoming requirement for the application is the
reduction of the stray inductances in the setup [2], [3]. Hence, both, module and setup must
provide a low inductance to avoid parasitic effects. Two of the most common effects, closely
related to stray inductances, are the over-voltage peak (VPeak) on the collector-emitter voltage
ISBN 978-3-8007-3505-1 © VDE VERLAG GMBH · Berlin · Offenbach
900
PCIM Europe 2013, 14 – 16 May 2013, Nuremberg
and the reduced switching losses due to the collector-emitter voltage drop during turn-off and
turn-on, respectively.
Fig. 6. Switching energies (EON and EOFF) and VPeak of the HS3 IGBT versus the stray inductance for
VDC = 300 and 400 V. Inset: Etotal of the HS3 IGBT versus the stray inductance for VDC = 300
and 400 V.
Fig. 6 displays the switching losses and the over-voltage peak of the HS3 IGBT versus the
stray inductance of the setup for equal switching parameters. The equal switching parameters were realized by modifying RG in the way that di/dt = 1.5 kA/µs and dv/dt = 7.2 kV/µs for
VDC = 400 V and di/dt = 1.6 kA/µs and dv/dt = 6.0 kV/µs for VDC = 300 V were set during turnon and turn-off, respectively. With increasing Lı, the turn-off energy increases slightly whereas the turn-on energy decreases significantly. As a result, the total switching energy is reduced for larger Lı. This general trend is independent of the applied DC-link voltage but the
reduction of EON is more pronounced for VDC = 400 V than for VDC = 300 V. On the other
hand, large Lı leads to a raise of VPeak. As a consequence of this, the DC-link voltage in the
application becomes limited since VCES • 9DC + VPeak with the maximum blocking voltage
VCES. A countermeasure for this is slowing down the device by increasing RG, which, in turn,
results in higher switching losses.
Increasing the stray inductance of the setup reduces Etotal of the IGBT due to the fact that the
reduction of EON is more pronounced than the increase of EOFF. However, parasitic effects
like oscillations caused by resonance frequencies of the setup or the diode snap-off will lead
to electromagnetic interference, which has to be considered in the application [4]. In addition,
these parasitic effects can also interact with the gate-driver circuit and may have a negative
impact on the device.
4.
Simulated device performance with IPOSIM
To analyze the device performance for different switching frequencies, the inverter performance was simulated with IPOSIM [5]. To ensure comparability, the dynamic losses of HS3
IGBT and IGBT3 at equal switching parameters, which are shown in Fig. 4, are used. In the
simulation, the output current of a single-phase H-bridge with an output power of 4 kVA was
calculated and the following operating conditions were considered:
The output current IOUT was set to 17.4 ARMS and a power factor of 1.0 was used. The modulation index was 0.8 and the DC-link voltage was 400 V. For both devices, an equal thermal
management with a fixed heat-sink temperature of 80 °C for the module (EasyPACK 2B) was
used.
ISBN 978-3-8007-3505-1 © VDE VERLAG GMBH · Berlin · Offenbach
901
PCIM Europe 2013, 14 – 16 May 2013, Nuremberg
Fig. 7. Simulated semiconductor power losses of HS3 IGBT and IGBT3 in an H-bridge inverter topology versus the switching frequency. The simulated power losses are the power losses of the
H-bridge inverter and not of a single chip.
Fig. 7 shows the simulated semiconductor power losses (PLosses) in an H-bridge inverter under the above mentioned operating conditions. The analysis of the H-bridge inverter shows
that the static losses of the IGBT3 are only about 70 % of the static losses of the HS3 IGBT.
With increasing switching frequency f, the dynamic losses become dominant. For f = 7.5 kHz,
the overall losses, i.e., the sum of dynamic and static losses, of the HS3 IGBT are equal to
the overall losses of the IGBT3 (highlighted in Fig. 7). A further increase of f enhances this
effect and it can be clearly seen that the advantage of the HS3 IGBT comes to the fore at
higher switching frequency.
Fig. 8. Maximum achievable output current versus the switching frequency for HS3 IGBT and IGBT3.
In Fig. 8, the maximum achievable output current is displayed. For the calculation, the above
mentioned operating conditions were used, whereas IOUT was not fixed but limited by the
maximum junction temperature of 150 °C of the devices.
With increasing f, IOUT decreases which follows directly from the findings in Fig. 7. At low
switching frequencies, the IGBT3 provides higher IOUT YDOXHV WKDQ WKH +6 ,*%7 )RU I •
7.5 kHz, the output current of the HS3 IGBT is higher than the output current of the IGBT3.
The difference between IOUT of HS3 IGBT and IGBT3 increases at higher switching frequencies. This is attributed to the lower switching losses of the HS3 IGBT compared to the switching losses of the IGBT3.
ISBN 978-3-8007-3505-1 © VDE VERLAG GMBH · Berlin · Offenbach
902
PCIM Europe 2013, 14 – 16 May 2013, Nuremberg
5.
Conclusion
In this paper, a comparison of HS3 IGBT, IGBT3 and IGBT4 is presented. It is shown that the
HS3 IGBT outperforms IGBT3 and IGBT4 in high-frequency applications by providing up to a
factor of 2 lower switching losses. To take advantage of the superior switching performance
of the HS3 IGBT, an application-optimized operation mode is needed. Therefore, the operating current and the gate resistor have to be considered carefully. For the latter, one possible
way is to use a more sophisticated gate-driver design. This allows to operate the HS3 IGBT
with a high di/dt during turn-on and to limit the influence of parasitics efficiently.
The HS3 IGBT is a cost optimized very efficient switch for high-frequency hard-switching applications like solar inverters or UPS. The results of the simulation support these findings and
show that the HS3 IGBT should be considered as state-of-the-art switch for applications
which are operated with switching frequencies larger than 7.5 kHz. At switching frequencies
above 16 kHz, the HS3 IGBT can provide up to 30 % higher output currents under the same
operating conditions compared to the IGBT3.
6.
[1]
[2]
[3]
[4]
[5]
Literature
D. Chiola, H. Hüsken, T. Kimmer: “High Speed IGBT with MOSFET-like switching behavior”,
PCIM 2012, Nuremberg, Germany
R. Bayerer, D. Domes: “Power Circuit Design for clear switching”, CIPS 2010, Nuremberg, Germany
K. Vogel, D. Domes: “IGBT inverter with increased power density by use of a high-temperaturecapable and low-inductance design”, PCIM 2012, Nuremberg, Germany
M. Bäßler, A. Ciliox, P. Kanschat: “On the loss-softness trade-off: Are different chip versions
needed for softness-improvement?”, PCIM 2009, Nuremberg, Germany
IPOSIM: IGBT Power Simulation available at www.infineon.com
ISBN 978-3-8007-3505-1 © VDE VERLAG GMBH · Berlin · Offenbach
903