Академический Документы
Профессиональный Документы
Культура Документы
JUAN COLMENARES
Doctoral Thesis
Stockholm, Sweden 2016
KTH
Electric Power and Energy Systems
TRITA-EE 2016:145 School of Electrical Engineering
ISSN 1653-5146 SE-100 44 Stockholm
ISBN 978-91-7729-109-1 SWEDEN
Tryck: Universitetsservice US AB
iii
A mis padres...
v
Abstract
Sammanfattning
This thesis concludes the work that I have carried out at the department
of Electric Power and Energy Systems (former department of Electrical Energy
Conversion), KTH Royal Institute of Technology since June 2012. Even though the
Ph.D. journey could be lonely, I would never have managed to complete this work
without the valuable help of various persons.
First of all, I would like to express my deep and sincere gratitude to Professor
Hans-Peter Nee, my supervisor, who not only gave me the opportunity to work as
a master thesis student back in 2011, but also received me again in 2012 as a Ph.D.
candidate. I am really thankful for his guidance, enthusiasm, and useful comments
and suggestions during these years. His knowledge in power electronics is unique.
Without him this thesis would have never been completed.
Many thanks to guest Professor Jacek Rabkowski from Warsaw Institute of
Technology. I learned a lot from his guidance and lab experience. He was a great
mentor during his stay, which was my very beginning as a doctoral student.
I would like to acknowledge Dr. Mietek Bakowski from Acreo AB. Thanks for
the really useful discussions during these years. Appreciation also to Dr. Patrik
Hilber from the department of Electromagnetic Engineering at KTH, who helped
me and guided me in the world of reliability.
Special thanks also to my part-time supervisor at University of Illinois at
Urbana-Champaign, Assistant Professor Robert Pilawa-Podgurski who received me
as an exchange student at UIUC this year. Thanks for the opportunity, I have
learned a lot during my stay there. Many thanks to my colleagues at UIUC, Tomas
(Dr. Modeer), Chris and Thomas, you were great company and made me feel very
welcome.
I would like express my very great appreciation to the rest of the SiC group
members at KTH: Dr. Dimosthenis Peftitsis, now Associate Professor from Norwe-
gian University of Science and Technology, Dr. Georg Tolstoy, and soon-to-be-Dr.
Diane Sadik. Our scientific discussions and countless hours sitting in the lab have
always been very valuable. I have learned a lot in the lab and office with you, and
you are a big part of this thesis. You have been not only my colleagues, but also
very good friends.
My gratitude also goes to the amazing people at the department. Thank
you for the great working environment. I would also like to thank Eva Pettersson
ix
x
and Brigitt Högberg for being more than our economists. Immense thanks to our
vital staff, Peter Lönn, Jesper Friberg, Alija Cosic and Simon Nee, without your
support this work would have never happened. I would like to thank also Associate
Professor Staffan Norga, Adjunct Professor Lennart Harnefors, Associate Professor
Oskar Wallmark, Antonious, Kalle, Shuang, Naveed, Andreas, Luca, Arman, Panos,
Arash, Lebing, Hui, Rudi, Keijo, Ilka, Daniel, Matthijs, Stefanie, and more. You
have been great company at KTH but also abroad on conferences.
Thanks to my friends from Venezuela: Veluska, Monika, Gabriela, Yasmin,
Francisco M., Gilfredo, Javier, Elías, Francisco V., Carlos P., Carlos F.. Your
humor, jokes, and friendship have been essential for my work. I am also very
thankful for my flat mates during this journey, Dick and Luis, with you I shared
the daily life. Special thanks to Rafael and David, you have been my friends from
a long time. You are my brothers and I really appreciate your support along the
road.
I would like to thank Nashira. Your love, patience, support and encourage-
ment have motivated me to be better every day.
Special thanks to my friends Sara and Karl, you have become my family and
I am really thankful for receiving me with your arms wide open. You have always
made me feel like home. Thank you for your invaluable support.
Last but not least, infinite thanks my family. Gracias a mi tío Antonio,
siempre has sido mi referencia y motivación. Gracias a mis hermanos Daniel, Simon
y Alexandra por cada momento de felicidad que me han brindado. Ustedes son muy
importantes para mi y siempre los llevo en mi corazón. Por último, gracias a mis
padres Juan y Thais, gracias a ustedes soy quien soy hoy en día. No hay palabras
que puedan expresar mi gratitud por todo lo que me han dado. Los amo.
Contents xi
1 Introduction 1
1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Main Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Outline of the thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.4 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.5 Main Scientific Contributions . . . . . . . . . . . . . . . . . . . . . . 4
1.6 List of Appended Publications . . . . . . . . . . . . . . . . . . . . . 5
1.7 Related Publications . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
5 Reliability Analysis 53
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
5.2 Reliability Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
xi
xii CONTENTS
5.3 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
List of Figures 85
List of Tables 89
Bibliography 91
Chapter 1
Introduction
1.1 Background
1
2 CHAPTER 1. INTRODUCTION
electric cars, and trains. Furthermore, transistors could also be categorized by the
semiconductor material they are composed of. For years, silicon (Si) has been the
semiconductor material of excellence. However, new promising technologies have
emerged. The wide-band gap (WBG) semiconductor materials such as silicon car-
bide (SiC) and gallium–nitride (GaN) allow higher voltage ratings, lower voltage
drops, higher switching frequencies, higher thermal conductance, and higher maxi-
mum temperatures. All these advantages offered by the WBG semiconductors, in
comparison with the currently and commonly used Si, make them an attractive
choice when high power density and high efficiency is needed.
Nowadays, many companies are able to manufacture WBG power devices
with sufficiently high quality and cost that allows the introduction of these devices
on the market. Following the reduction in manufacturing cost for WBG power
semiconductors, significant advantages can be achieved in the aforementioned ap-
plications, and in some cases the introduction of WBG semiconductors might take
power electronics to new areas of development [2].
When SiC power devices were firstly introduced, various quality and perfor-
mance related issues were present such as parameters spread. In order to achieve
high production yields, these SiC power devices employed chips with comparably
small areas, which resulted in low current ratings for discrete devices. Thus, single
discrete components could not be used for high-power applications, above tens of
kilovolt-amps (kVA). In the literature, two different solutions to reach high current
ratings can be found: either to parallel-connect several single-chip discrete devices
or to build multi-chip power modules [3–7]. It is important to note that parallel
connection not only increases the total current rating but also increases the overall
efficiency of the system. Parallel-connecting devices reduces the equivalent on-state
resistance, therefore, reducing the on-state losses. However, parallel-connection of
transistors may also give rise to adverse effects. Specifically, it has been shown
that problems associated with the “Miller effect” might affect stable operation [8].
In particular, the Miller effect may cause accidental turn-ON and self-sustained
oscillations between the Miller capacitance of the parallel-connected devices and
the stray inductances of the external circuit layout and of the circuit layout of the
module as such [8]. Slowing down the switching times would avoid these issues.
However, this would mean that the full potential of the WBG technology could not
be taken as an advantage.
Another possibility to reach even higher power ratings (of the order of hun-
dreds of kVA) is the parallel-connection of power modules. This could bring the
benefits of the WBG technology to further applications. By successfully achieving
parallel-connection (either discrete components, or power modules) and fast switch-
ing performance the final design would benefit from both high switching frequencies
and a high current capability.
Unfortunately, increasing the amount of components, such as parallel connec-
tion of devices, may have a negative impact on the reliability of the system. By
introducing additional components the probability of failure increases. Therefore,
an undesired trade-off between the targeted high current ratings and high efficiency,
1.2. MAIN OBJECTIVES 3
• Study, design, propose and experimentally verify gate drivers for normally-ON
SiC JFET power modules populated with a massive number of chips.
Chapter 2 gives a brief description of the wide-bandgap devices used in this work and
their applications.
Chapter 3 summarizes the driving requirements for SiC power modules and describes
two possible solutions for these power devices.
4 CHAPTER 1. INTRODUCTION
Chapter 4 shows the experimental investigation of the static and dynamic performance
of parallel-connected SiC power modules. Additionally, it presents an appli-
cation where this concept is used.
Chapter 5 describes different failure modes of the SiC power modules and gives a life
estimation analysis of a complete system using SiC power modules.
Chapter 7 draws the conclusions of the work done in this thesis and gives ideas for future
work.
1.4 Methodology
The results of this work are mainly derived from theoretical and mathema-
tical analyses that have been demonstrated by simulations and experiments. The
experimental verifications, included in this work, have been performed by measur-
ing the voltage and current waveforms using voltage probes and Rogowski-coils,
respectively. Furthermore, the voltage, current and passive component values were
measured using high precision digital multi-meters and LCR meters. Finally, the
loss measurements have been carried out by means of electrical measurements of
powers, voltages, and currents.
• A gate driver design has been proposed, design, manufactured, built and
tested for SiC JEFT power modules. It showed that by using two totem-
pole branches, it is possible to have a fast switching performance with low
oscillations. One of these branches assists the fast switching while the other
one reduces the oscillations in the drain current and the drain-source voltage.
Simultaneous work on this subject was published by Zhang et al [15].
• It has been shown that massive parallel connection of SiC MOSFET power
modules is achievable without any rigorous sorting criteria. The circuit lay-
out and the proper selection of power modules in conjunction with the pos-
itive temperature coefficient of the ON-state resistances of the MOSFETs
contribute toward a sufficiently uniform loss distribution under steady-state
operation of the modules in a converter.
1.6. LIST OF APPENDED PUBLICATIONS 5
This paper presents a gate-drive circuit design for a SiC power module po-
pulated with parallel-connected normally-ON JFETs. This concept is able to
find a good compromise between the fast and the least oscillative switching
performance. With the suggested driver it is possible to achieve sufficiently
low switching losses such that a high-efficiency converter could be designed.
By using the presented gate driver, efficiencies of approximately 99.6 % can
be achieved for a converter rated at 125 kVA.
The main contribution to this paper are the design and construction of the
gate-drive unit, construction of the experimental setup for double pulse test-
ing, the experimental procedure, and preparation of the manuscript.
normally-ON JFETs was provided. Various gate driver designs have been
experimentally evaluated in order to find a fast switching performance with
the least oscillations. Moreover, a stability analysis of the proposed gate
driver is included. Finally, a loss estimation using the proposed gate-drive
unit in a three-phase, two-level VSC rated at 125 kVA is also presented.
Main contribution to this manuscript: design and construction of the different
generations of the gate-drive units, construction of the experimental setup,
measurements of the switching performance and switching losses, preparation
of the manuscript.
In this conference paper the design process of a 312 kVA three-phase SiC
inverter using parallel-connected MOSFET power modules for motor-drive
applications with a switching frequency of 20 kHz is presented. The de-
sign processes of the gate-drive circuits with short-circuit protection is also
presented. The experimental results show the electrical performance during
steady-state operation of the power converter with 99.3 % efficiency at the
rated power.
1.6. LIST OF APPENDED PUBLICATIONS 7
The main contributions to this paper are the reliability study, including the
failure mode identification and different cases of analysis as well as preparing
the manuscript.
VII. J. Colmenares, D.-P. Sadik, P. Hilbert, and H.-P. Nee, “Reliability Analy-
sis of a High-Efficiency SiC Three-Phase Inverter,” IEEE Trans. Emerg. Sel.
Topics Power Electron., vol. 4, no. 3, pp. 996–1006, Sept. 2016.
In peer-reviewed journals
• D.-P. Sadik, J. Colmenares, G. Tolstoy, D. Peftitsis, J. Rabkowski, and H.-
P. Nee, “Short-Circuit Protection Circuits for Silicon-Carbide Power Transis-
tors,” IEEE Trans. Ind. Electron., vol. 63, no. 4, pp. 1995–2004, Apr.
2016.
• D.-P. Sadik, K. Kostov, J. Colmenares, and H.-P. Nee, “Analysis of Para-
sitic Elements of Silicon Carbide Power Modules with Special Emphasis on
Reliability Issues,” IEEE Trans. Emerg. Sel. Topics Power Electron., vol. 4,
no. 3, pp. 988–995, Sept. 2016.
10 CHAPTER 1. INTRODUCTION
In conference proceedings
• D.-P. Sadik, J. Colmenares, D. Peftitsis, J.-K. Lim, J. Rabkowski, and H.-P.
Nee, “Experimental Investigations of Static and Transient Current Sharing of
Parallel-Connected Silicon Carbide MOSFETs,” in Proc. of 2013 15th Eur.
Conf. on Power Electron. and Appl. (EPE), Sept. 2013, pp. 1–10.
• H.-P. Nee, J. Rabkowski, D. Peftitsis, G. Tolstoy, J. Colmenares, D. Sadik,
et al., “High-Efficiency Power Conversion Using Silicon Carbide Power Electro-
nics,” in Mater. Sci. Forum, 2014, vol. 778, pp. 1083–1088.
• D.-P. Sadik, J. Colmenares, G. Tolstoy, D. Peftitsis, J. Rabkowski, and
H.-P. Nee, “Analysis of Short-Circuit Conditions for Silicon Carbide Power
Transistors and Suggestions for Protection,” in Proc. of 2014 16th Eur.
Conf. on Power Electron. and Appl. (EPE), Sept. 2014, pp. 1–10.
• G. Tolstoy, P. Ranstad, J. Colmenares, D. Peftitsis, F. Giezendanner, J.
Rabkowski, and H.-P. Nee, “An Experimental Analysis of How the Dead-
Time of SiC BJT and SiC MOSFET Impacts the Losses in a High-Frequency
Resonant Converter,” in Proc. of 2014 16th Eur. Conf. on Power Electron.
and Appl. (EPE), Sept. 2014, pp. 1–10.
• G. Tolstoy, P. Ranstad, J. Colmenares, F. Giezendanner, and H.-P. Nee,
“Dual Control Used in Series-Loaded Resonant Converter With SiC Devices,”
in Proc. of 2015 9th Int. Conf. on Power Electron. (ICPE-ECCE Asia), Jun.
2015, pp. 495–501.
• G. Tolstoy, P. Ranstad, J. Colmenares, F. Giezendanner, and H.-P. Nee,
“Experimental Evaluation of SiC BJTs and SiC MOSFETs in a Series-Loaded
Resonant Converter,” in Proc. of 2015 17th Eur. Conf. on Power Electron.
and Appl. (EPE), Sept. 2015, pp. 1–9.
• D.-P. Sadik, K. Kostov, J. Colmenares, and H.-P. Nee, “Analysis of Para-
sitic Elements of Silicon Carbide Power Modules with Special Emphasis on
Reliability Issues,” in Proc. of 31st Annual IEEE Applied Power Electron.
Conf. and Exp. (APEC), Mar. 2016, pp. 1018–1023.
• D.-P. Sadik, J.-K. Lim, J. Colmenares, M. Bakowski, and H.-P. Nee, “Com-
parison of Thermal Stress during Short-Circuit in Different Types of 1.2 kV
SiC Transistors Based on Experiments and Simulations,” 11th Eur. Conf. on
Silicon Carbide and Rel. Mater. (ECSCRM), accepted for publication.
• D.-P. Sadik, J. Colmenares, P. Ranstad, F. Giezendanner, and H.-P. Nee,
“Introduction of SiC MOSFETs in Si IGBTs based Converters: a Reliability
and Efficiency Analysis,” 2017 IEEE Applied Power Electron. Conf. and
Exp. (APEC), submitted for review.
Chapter 2
2.1 Introduction
The power electronics field is undergoing a rapid change with the introduction
of wide-bandgap (WBG) semiconductor materials. These materials offer many
benefits compared to the currently used silicon (Si). In particular Si, as well as other
commonly used materials have a bandgap of the order of 1 to 1.5 electron–volt (eV).
WBG materials, typically have a bandgap of the order of 2 to 4 eV. This difference
allows the wide-bandgap materials to operate at much higher temperatures. Also,
most WBG semiconductors have a much higher breakdown electric field. SiC has,
for instance, approximately ten times higher breakdown electric field compared
to Si. These properties combined, allow them to operate at much higher power
with lower losses. The most common WBG materials are gallium–nitride (GaN)
and silicon carbide (SiC) and their properties compared with Si are summarized in
Table 2.1. These clear inherent qualities make the WBG semiconductors suitable
for many roles in energy conversion systems.
In general, the WBG materials are still under development, a process that
started in the 1970s. However, to date devices for high-power and high-temperature
11
12CHAPTER 2. WIDE-BANDGAP POWER DEVICES AND APPLICATIONS
SiC power devices have evolved from prototype samples to commercially avail-
able products. This is evident on the market, which now offers different SiC power
devices from discrete components to power modules. On the market it is now pos-
sible to find JFETs, BJTs, and MOSFETs made from SiC. Moreover, these devices
are available at different rated voltages from 600 V to 1.7 kV, and current ratings
from 2.6 A to 325 A. Additionally, several manufacturers are now producing SiC
power devices such as Wolfspeed, ROHM Co., Ltd., GeneSiC Semiconductor Inc.,
United Silicon Carbide, Inc., Infineon Technologies AG, and Mitsubishi Electric
Corporation. This shows signs that the WBG technology has matured to the point
that is now a viable alternative for replacing Si counterparts.
2.2. SIC POWER DEVICES 13
body diode in its structure. However, an anti-parallel SiC Schottky diode, which
will only be utilized for a short-time during the blanking times, can be externally
connected to the VTJFETs.
reliability aspects of the oxide layer. Samples of the SiC DMOSFET from Cree Inc.
were available on the market in 2010. These samples were rated at 1200 V and 10,
20 and 50 A of current ratings with with 160 mΩ, 80 mΩ and 25 mΩ of on-state
resistance respectively.
OFF-state. This makes this device very unattractive to many power conversion
applications.
It was not until 2009 when Efficient Power Conversion Corporation (EPC) in-
troduced the enhancement-mode GaN-on-silicon power transistors for the first time.
Figure 2.3 shows the structure of this device, where a depletion region under the
gate is achieved. In order to turn the FET ON, a positive voltage is applied to the
gate, as when turning on an enhancement-mode power MOSFET [64]. This device
is now available at different rated voltages from 15 V to 300 V with various current
ratings and on-state resistances. As seen from Figure 2.3, this is a lateral device
and there is not a parasitic bipolar junction or body diode. However, with this
structure a similar conduction function with no reverse recovery losses is achieved.
This device was specifically designed as a Si power MOSFET replacement. Several
manufacturers are now producing GaN power devices such as EPC, GaN Systems,
Panasonic, among others. This shows the interest from the industry and a more
mature technology is therefore expected.
• Power Factor Correction as shown in [16]. For this application the drop-in
replacement of the power switch was enough to produce an increase in total
system efficiency. Combining these results with the ability of the SiC devices
to achieve higher switching frequencies will potentially lead to smaller, and
more efficient power electronic circuits.
2.4. APPLICATIONS OF WIDE-BANDGAP POWER DEVICES 17
• Power Inverters with efficiencies well above 99 % are more than possible us-
ing SiC power devices. In [24], a 40 kVA inverter with an efficiency exceeding
99.5 % using natural convection cooling is presented.
• Data Center power supplies could also profit as presented in [26] with ultra
high-efficiency buck rectifiers.
Furthermore, the GaN power devices have a voltage rating in the range from
15 to 300 V, which make them a direct competitor with the currently used Si
MOSFETs. Even though the GaN power FET was introduced only recently to the
market, their benefits are clear in many applications, such as:
• dc/dc Conversion using GaN devices will enable high efficiencies and high
switching frequencies due to their low on-state resistance and low gate capaci-
tance, respectively. In [70] a dc/dc boost converter with a switching frequency
of 10 MHz and an efficiency above 90% is presented. Moreover, techniques
such as dead time optimization are also feasible using GaN switches [71].
• In [72] an inverter using GaN power switches operating at 10’s of MHz for
Wireless Power Transfer is presented. The combination of the low induc-
tive package, the fast switching capability, the low gate capacitance, and the
low on-state resistance make the GaN device the ideal component for this
application.
• Power Inverters used in many applications such as solar power generation
and other appliances are desired to have high efficiency and to be compact. In
[23] a 2 kW single-phase inverter is presented. The inverter presents a 7-level
flying capacitor multilevel topology using GaN switches operating at 120 kHz,
while maintaining an efficiency above 99 %. The low-loss characteristics of the
GaN power switches enable using non-typical topologies for these applications,
which could bring additional benefits not present yet.
• Other applications such as Envelope tracking, Light Distancing and
Ranging (LiDAR), and Class D audio amplifiers could benefit from
the low losses enabling higher efficiencies. These systems could produce less
heat, save space and cost, and extend the battery life in portable systems.
There are even more applications than the ones aforementioned that are tak-
ing advantage of the WBG semiconductors properties. This also shows that the
technology has matured to some extent and exploring its limits through extreme
implementations is a natural step forward.
2.5 Conclusion
A brief description of the commercially-available WBG power devices used in
this work has been presented in this chapter. SiC power devices such as JFETs and
MOSFETs are available today for evaluation and implementation in many appli-
cations. Additionally, the inherent issues with the SiC JEFTs and how to address
them as well as the limitations of the SiC MOSFETs were described. Furthermore,
the recently introduced GaN power FETs were also presented. Finally, an overview
of the reported benefits for several applications has been presented.
Chapter 3
3.1 Introduction
SiC power devices are either current controlled such as SiC BJTs, or voltage
controlled such as SiC JFETs or SiC MOSFETs. The gate or base drive unit is one
of the most important parts in every power converter because it controls the switch-
ing losses and therefore, the efficiency. The desired switching speeds are achieved
without significant oscillations by properly adjusting the circuit parameters in the
gate and base driver. Thus, in order to take advantage of the SiC characteristics,
such as fast switching performance, special considerations must be made for the
gate and base driver design.
In addition to the typical requirements for every gate or base driver, when
normally-OFF SiC JFETs and SiC BJTs are considered, a further requirement must
be taken into account. Both of these devices require a constant steady-state current
to keep them in the on-state. This introduces additional losses to the system. Thus,
an optimal design is desired in order to achieve both a fast switching performance
and a low on-state power consumption. Several studies have been presented so as
to achieve the best performance from these devices [73–77].
On the other hand, when normally-ON SiC JFETS are considered, an ad-
ditional requirement must be taken into consideration. A negative gate-to-source
voltage is necessary in order to keep the device in the OFF state, a fact that is
related to several safety and reliability issues. An special gate driver was proposed
in 2005 [4, 25, 78, 79], which is considered to be the state-of-art gate driver for SiC
JFETs. Moreover, in the literature, it is also possible to find several suggested
solutions to the normally-ON problem. One suggested solution is based on a cas-
19
20 CHAPTER 3. GATE DRIVERS FOR SIC POWER MODULES
(a) (b)
Figure 3.1: (a) Schematic diagram of the parallel-connected SiC JFET chips popu-
lating one switch position (b) Simplified schematic diagram of one switch position.
A photograph of the power module is shown in Figure 3.2. Table 3.1 summa-
rizes the parameters of the parallel-connected chips and Table 3.2 summarizes the
parameters of power module.
Table 3.1: Device parameters of the SiC JFETs and SiC Schottky diodes.
Device JFET Schottky diode
Voltage rating (V) 1200 1200
Current rating (A) @ 25o C 48 30
Current rating (A) @ 100o C 30 46
On-state resistance @ 25o C (mΩ) 45
Chip area (mm2 ) 8.12
should be done without compromising the stability of the system, i.e., with as low
oscillations as possible.
The printed circuit board (PCB) was optimized in order to obtain symmetrical
paths for the turn-ON and turn-OFF gate currents with the lowest possible stray
inductance. Moreover, the gate driver PCB also consists of two symmetrical parts.
Each one of those controls the corresponding SiC JFET switches of the upper or the
lower switch positions of the half-bridge power module. Several gate driver designs
3.2. DUAL-FUNCTION GATE DRIVER 23
have been built and experimentally tested on the SiC power module as shown in
[Publications I – II]. The design process of the gate driver started from a very
simple solution while the final version is a more complicated circuit which is able
to satisfactorily drive the power module.
Version I
The very first attempt to drive the SiC power module was performed using
a typical totem-pole configuration, as shown in Figure 3.3. Two Si MOSFETs,
T1 and T2 , were used in order to implement the totem pole. The same control
signal was provided to the gates of the MOSFETs through resistors R1 and R2 .
This control signal is fed by an optocoupler (ACPL-H342), with a high operating
voltage range and a high common-mode rejection ratio (CMRR). In order to obtain
the desired switching speeds of T1 and T2 , the values of R1 and R2 have to be
properly adjusted. The output of the gate driver is directly connected to the power
module. The positive and negative supply voltages of the totem pole, Vp and Vn
respectively, are also used to supply the optocoupler. A positive voltage of 2.5
V has been chosen. This positive voltage is high enough to reduce the RDS(ON)
as shown in [90, 91]. This positive voltage is chosen so as not to exceed the built
in potential of the p-n junction between the gate and source. Therefore, there is
not a significant amount of carriers injected to the channel. The negative voltage,
−15 V, has basically been chosen in such a way that it is unlikely for the gate-
source junctions of the JFETs to face any breakdown. It must also be noted that
an additional capacitor C1 was connected between the gates of the MOSFETs in
order to couple them and control the transition process between T1 and T2 .
Figure 3.3: Schematic diagram of version I of the gate driver for the SiC JEFT
power module.
24 CHAPTER 3. GATE DRIVERS FOR SIC POWER MODULES
Final Version
The final version of the gate driver is based on the same totem-pole configura-
tion implemented previously. Two additional resistors, R3 and R4 , were connected
between the drains of the MOSFETs and the gate pin of the module, respectively.
These resistors would make the switching speeds of the module slower than in the
previous case. It was important to develop a method to turn the JFETs OFF faster
without compromising the stability of the switching process. In order to solve this
issue, two additional MOSFET, T3 and T4 , were introduced in the gate-driver de-
sign. These MOSFETs are connected to capacitors C2 and C3 on the drain as
shown in Figure 3.4. These transistor are controlled by the same control signal
as T1 and T2 . On the gate of these MOSFETs, resistors R5 and R7 are connected
in order to control the switching speeds, similarly as in the previous case. These
additional transistors would offer a low-impedance path for the turn-off current
through the capacitors C2 and C3 . This low-impedance path is available while the
capacitor is being charged, and during this time interval a high current peak is sup-
plied to the gates of the JFETs. Once the capacitors are charged, diodes D1 and D2
would provide a fast discharging path for the capacitors. Thus, the capacitor will
be discharged for the next switching period. Moreover, a low-value resistor R9 was
connected between the output of the gate-drive unit and the gate of the module
with the purpose of damping the oscillations generated from the fast switching.
Figure 3.5 is a photograph of the gate driver PCB prototype.
Figure 3.4: Schematic diagram of the final version of the gate driver for the SiC
JEFT power module.
3.2. DUAL-FUNCTION GATE DRIVER 25
Figure 3.5: Photograph of the gate driver PCB prototype for the SiC JEFT power
module.
Experimental Evaluation
The experimental verification of the proposed gate drivers and switching per-
formance of the power module with normally-ON SiC JFETs has been investigated
using different gate-driver generations in a standard double-pulse test (DPT). Fi-
gure 3.6 illustrates the schematic circuit of the experimental setup, which consists
of a voltage supply, VDC , with a capacitor, C, in parallel, and an inductor, L, which
acts as the load. Since the module is configured as a half bridge, the device under
test (DUT) is the lower switch, while the upper one operates as the freewheeling
diode. Table 3.3 summarizes the parameters of the experimental setup. Several
experiments were performed and the results showing the switching performance are
presented in this section. Figure 3.7 shows a photograph of the double-pulse test
setup.
This DPT was performed for each version of the gate driver and the different
results were attained and they are presented below. All experiments were performed
at room temperature, approximately 25o C. For the DPT a blanking time of 1 µs
was chosen in order to avoid shoot through currents and ensure a safer operation.
Moreover, due to constraints present in the experimental setup, it was possible to
measure only half of the current with a single Rogowski coil.
Table 3.3: Parameters of the double-pulse test set-up for the SiC JEFT power
module.
Capacitance, C 160 µF, 700 V
Inductance, L 35 µH, air coil
26 CHAPTER 3. GATE DRIVERS FOR SIC POWER MODULES
Figure 3.6: Schematic diagram of the double-pulse test setup for SiC the JEFT
power module.
Figure 3.7: Photograph of the gate driver PCB prototype for the SiC JEFT power
module.
achieve a design suitable for lower voltages as well as for higher voltages. Stable
conditions could be obtained at higher dc-link voltages, due to the nonlinearity of
the Miller capacitance. Nevertheless, a design for only higher dc-link voltages is
not desirable for real applications. In real applications, the dc-link voltage could
drop, and stable operation is desirable for these conditions as well.
Figure 3.8: Turn-ON and turn-OFF transients of the SiC JFET power module with
gate driver version I with VDC equal to 25 V. Measured gate-source voltage of the
SiC JFET, M1 (yellow line, 20 V/div), gate current of the SiC JFET, M1 (green
line, 20 A/div), drain-source voltage of the SiC JFET, M1 (purple line, 50 V/div),
drain current of the SiC JFET, M1 (pink line, 20 A/div), (time-base 500 ns/div).
Figure 3.8 shows the switching results for VDC = 25 V when using version I
of the gate driver. Self-sustained oscillations are observed. It is clear that the gate-
source voltage of the lower switching position oscillates around the pinch-off voltage
of the devices. These self–sustained oscillations observed during the switching of
the module are similar to the ones described in [8]. They have a frequency of
approximately 6 MHz and they are mainly caused by a resonance between the
parasitic stray inductance and parasitic capacitance. The parasitic stray inductance
results from the dc-link bus bars and the parasitic drain inductance of the module
in combination with the stray inductance of the entire gate loop. Moreover, the
parasitic capacitance is related to the the drain-gate capacitance and gate-source
capacitance of the devices. Due to the observed severe oscillations it was clear that
any attempt to investigate the performance of the power module at higher drain-
source voltages with this gate-drive circuit would probably have caused irreversible
damage to the power module. Therefore, further investigations were cancelled.
The final version of the proposed gate driver was experimentally verified fol-
lowing a similar test procedure. The dc-link voltage was increased in steps of 25 V,
until the maximum value of the voltage supply was reached, VDC , 650 V. Figure 3.9
and Figure 3.10 show the turn-ON and turn-OFF switching transients, respectively.
28 CHAPTER 3. GATE DRIVERS FOR SIC POWER MODULES
The turn-ON process takes approximately 132 ns while the corresponding turn-OFF
time is approximately 184 ns. The switching losses were estimated to be approxi-
mately 9.9 and 23.8 mJ for the turn-ON and turn-OFF processes, respectively. A
high current overshoot of approximately 260 A is observed during the turn-ON
process due to negative, discharging capacitive current of the diode and transistors.
An increase from the steady-state current is observed during the turn-OFF pro-
cess, due to the charging capacitive current. The steady-state current during the
turn-ON time equals approximately 95 A. At the turn-OFF process, the current
drops fast from its nominal value to a certain low value. However, it takes longer
time to reach the zero level. Also a clear overshoot of 800 V is shown during the
turn-OFF. This voltage spike consists of the inductive voltage which is generated
by the current slope through the parasitic stray inductance of the power circuit.
Figure 3.9: Turn-ON transient of the SiC JFET power module with the final version
of the gate driver. Measured drain-source voltage of the SiC JFET, M1 (purple line,
200 V/div), drain current of the SiC JFET, M1 (pink line, 100 A/div), (time-base
100 ns/div).
Stability Analysis
The self-sustained oscillations described in the experimental results imply a
feedback mechanism that is replacing the energy dissipated by the resistances in
the circuit. Starting from the schematic diagram in Figure 3.4, and considering the
parasitic stray inductance of the circuit and the power module, a small-signal model
for the power module FET oscillator can be developed as shown in Figure 3.11.
It is possible to distinguish two branches: an active branch and a motional branch
as it is described in the negative resistance model, which is a possible theoretical
background to describe these oscillations [8]. This small-signal model represents
an unintentional oscillator, where the active branch consists of the FET power
module with its gate driver, while the motional branch is related to the dc-link
3.2. DUAL-FUNCTION GATE DRIVER 29
Figure 3.10: Turn-OFF transient of the SiC JFET power module with final version
of the gate driver. Measured drain-source voltage of the SiC JFET, M1 (purple line,
200 V/div), drain current of the SiC JFET, M1 (pink line, 100 A/div), (time-base
100 ns/div).
bus. Applying the well-known Barkhausen Criterion to the small-signal FET power
module model, it is possible to determine the conditions for sustained oscillations.
By considering the admittance YIN , the occurrence of these oscillations re-
quires two conditions. The first one requires that the real part of YIN is equal to
−GEP , given by:
" " 2 ##−1
ωLD
GEP = RESR +1 (3.1)
RESR
where RESR is the parasitic resistance of the dc link. The second condition requires
that the imaginary part of YIN is equal to zero. The following expression describes
YIN :
X2 +ZG
X2 ZG + gm 1
YIN = X1 (X2 +ZG )
+ (3.2)
+1 XD
X2 ZG
bus, it is possible to choose properly the values of the gate driver components for any
power module, making it not fulfilling the conditions for self–sustained oscillations.
Further information on how the stability analysis was performed can be found in
[Publication I– II].
Figure 3.11: Small-signal model for a power module FET oscillator with the final
version of the gate driver [8].
Loss Analysis
Considering the results of the final version, and assuming a three-phase, two-
level voltage source converter (VSC) rated at 125 kVA (700 V and RMS current
of 147 A) an estimation of the switching losses was performed using MATLAB.
Data regarding the on-state resistance have been measured at RT and estimated at
75o C using manufacturer data, as shown in Table 3.2. Thus, a value of 3.0 mΩ at
75o C has been considered for the on-state resistance of one DUT inside the module.
Moreover, a switching frequency of 2 kHz was chosen. This choice was based on
the control system requirements of the industrial application. For the modulation,
third harmonic injection was considered. Based on the assumptions above, the
corresponding conduction and switching losses were found to be 193 and 148 W
using:
Sn − (PSW + PCD )
η = 100 · (3.5)
Sn
Further information on how the losses were estimated can be found in [Pub-
lications I – II].
(a) (b)
Figure 3.12: (a) Photograph of the power module with SiC MOSFETs and, (b)
Simplified schematic diagram of a single switch position.
32 CHAPTER 3. GATE DRIVERS FOR SIC POWER MODULES
Figure 3.13: Schematic diagram of the gate driver with the short-circuit protection.
module OFF as soon as the protection is activated. In order to avoid false trigger-
ing of the short-circuit detection during transient times, the RC filters τ1 and τ2
are connected before M3 . They are used so as to adjust the detection delay time
depending on the application. Figure 3.14 shows a photograph of the gate-drive
unit. A more detailed description regarding the operation of the proposed gate
driver can be found in [Publications IV – V]
Figure 3.14: Photograph of the gate-drive unit with the short-circuit protection.
Experimental Evaluation
Finally, an experimental evaluation of the gate driver was performed for a
single power module. This evaluation was divided into two parts. First, a short-
circuit condition test was performed, in order to determine if the driver is able to
detect a short circuit. The second part is a DPT as in Figure 3.6, which is carried
out in order to verify the switching capability of the gate drivers. The experimental
setup consists of a capacitor C = 160 µF that is connected with a voltage supply,
VDC , and a load, which is an air coil inductor, L, having an inductance of 150 µH.
For the first case, the short-circuit condition, VDC was set to 600 V and the
upper switch position of the module was continuously kept in the ON state, while
a pulse of 1 µs was applied to the gate of the lower switch position. Figure 3.15
shows the short-circuit detection performance for a single switch position. The
driver detects the short circuit within approximately 150 ns and turns OFF the
switch position in 50 ns, i.e. that the switch position is in the OFF state at
approximately 200 ns after the pulse was applied. During this period of time
the current has already increased to 800 A, which is 5 times higher than the rated
current of a single power module. Therefore, high temperatures in the power device
are expected, which could affect the reliability of the device. Thus, the faster the
detection, the lower the risk of permanent damage the power modules. An overshoot
34 CHAPTER 3. GATE DRIVERS FOR SIC POWER MODULES
of almost 400 V is observed during the turn-OFF process. This is mainly due to
the high di/dt, which generates a voltage over the stray inductance. Moreover, it
is important to note that the power device should be turned OFF at a moderate
speed in order to avoid high overshoot voltages that could also affect the power
module.
Figure 3.15: Short–circuit detection using the proposed gate driver. Measured
drain-source voltage of the SiC MOSFET, (purple line, 200 V/div), drain current
of the SiC MOSFET, (pink color, 500 A/div), logic signal from the latch (green
color, 50 V/div), gate-source voltage of the SiC MOSFET, (yellow color, 50 V/div),
(time-base 200 ns/div).
On the other hand, the switching operation of the SiC MOSFETs power mo-
dule using the proposed gate driver is shown in Figure 3.16 and Figure 3.17. The
turn-ON and turn-OFF processes take approximately 70 ns. In this case an over-
shoot of 200 V is observed during the turn-OFF. It is the hypothesis of the author
that the oscillations seen in the gate-source voltage of the power modules are due
to the stray inductance in the gate loop and not on the SiC MOSFET chip itself,
as shown in [92]. These results mean that a short-circuit protection is achieved
without compromising the switching performance of the power module. Further
information on the switching performance of the power module can be found in
[Publication IV – V].
Loss Analysis
It is found that the switching losses are approximately 5 mJ per module
(700 VDC , and 45 A per module). Assuming a VSC rated at 312 kVA, an estima-
tion of the total power losses, as well as, the efficiency has been performed using
MATLAB. Moreover, neglecting the diode losses due to the fact that they only con-
duct during the blanking time, which is significantly lower than the conduction time
of the MOSFET channel, and with an assumed on-state resistance of 1.8 mΩ per
3.3. SHORT-CIRCUIT PROTECTION GATE DRIVER 35
Figure 3.16: Turn-ON switching waveforms of the power module with the proposed
gate driver. Measured drain-source voltage of the SiC MOSFET, (purple line, 200
V/div), drain current of the SiC MOSFET, (pink color, 50 A/div), gate-source
voltage of the SiC MOSFET, (yellow color, 20 V/div), (time-base 50 ns/div).
Figure 3.17: Turn-OFF switching waveforms of the power module with the proposed
gate driver. Measured drain-source voltage of the SiC MOSFET, (purple line, 200
V/div), drain current of the SiC MOSFET, (pink color, 50 A/div), gate-source
voltage of the SiC MOSFET, (yellow color, 20 V/div), (time-base 50 ns/div).
3.4 Conclusion
In this chapter, two gate driver designs were presented. A dual-function gate
driver for custom-made power modules populated with SiC normally-ON JFETs,
which features a trade-off between fast switching speed and least oscillative per-
formance, was presented and experimentally verified. This gate-drive unit consists
of two totem-pole branches, one of which assists the fast switching speeds while
the other one reduces the oscillations in the drain current and the drain-to-source
voltage. The stability of the presented gate driver has been mathematically proven
by using the Barkhausen criterion that allows a proper selection of the values of the
components of the proposed gate driver. Finally, a gate driver with short-circuit
protection is presented and experimentally verified. The proposed gate-drive unit is
able to protect the converter against short-circuit conditions without compromising
the switching performance of the power modules.
Chapter 4
4.1 Introduction
Depending on the design target chosen, different characteristics from the SiC
properties could be leveraged. If high efficiency is chosen as main design target for
power converters, two distinct solutions have been proposed in the literature re-
lated to parallel-connection. These solutions are: either to parallel-connect discrete
devices or to build power modules populated with multiple chips [93]. Addition-
ally, in order to increase the production yields, the first generation of SiC power
devices employed smaller chip areas compared to state-of-the-art Si IGBTs, lead-
ing to lower current ratings. Therefore, considering the current ratings of these
devices, the proposed solutions not only help increasing the efficiency but also the
total current rating of the final power converter.
The first solution has been thoroughly investigated for several SiC power de-
vices such as JFETs [3, 4], BJTs [3, 94], and MOSFETs [6, 95, 96]. These studies
have found that regardless of the SiC power device used for the parallel-connection,
the uneven distribution of the parasitic elements in the circuit layout is the main
factor affecting the current distribution among the different parallel-connected sin-
gle chips. This may affect the performance and reliability of the power converter.
On the other hand, when power modules are considered, ithe external circuit layout
characteristics has less effect if an optimized design and symmetrical placements of
the chips inside the modules is reached. Nevertheless, some other issues such as the
“Miller effect” should be taken into consideration when parallel-connection is the
design approach for the converter [8, 97].
A possible solution that has not been deeply studied for SiC power converters
37
38 CHAPTER 4. PARALLEL-CONNECTION OF SIC POWER MODULES
is the parallel-connection of power modules. Thus, dealing with the power mo-
dules as if they were discrete components. This solution will further increase the
efficiency of the system by connecting in parallel a massive total number of chips.
Additionally, higher power ratings are possible, in the order of hundreds of kVA,
which will bring the benefits of using SiC to other applications. Furthermore, this
solution has been studied for Si IGBTs [98, 99]. If parallel-connection of power mo-
dules is chosen as a strategy in order to build a power converter, special attention
must be paid to the design process of the power circuit layout in order to guarantee
a high-performance converter.
In this chapter, the design process of employing parallel-connection of SiC
MOSFET power modules in a three-phase voltage source converter (VSC) is pre-
sented. Moreover, the main objective of this investigation is to show that the
benefits introduced by SiC such as high efficiency and high switching frequencies
are feasible regardless of the power rating of the final converter.
4.2 Methodology
As mentioned above, the main target is to achieve a high efficiency (> 99 %) at
a relatively high switching frequency (20 kHz) for a three-phase two-level VSC. The
electric parameters of the converter are shown in Table 4.1. Furthermore, parallel-
connection of power modules was chosen as a solution for the design methodology of
the converter. This will increase the total chip area and decrease the total equivalent
on-state resistance. Thus, the expected conduction losses will be reduced.
Figure 4.1: Relative on-state power loss versus junction temperature for ten parallel-
connected modules (CAS100H12AM1) - calculation using datasheet values.
Taking into consideration the physical dimensions of the power modules shown
in Figure 3.12, the circuit layout of the complete three-phase converter, built with
ten parallel-connected SiC power modules per phase-leg, is shown in Figure 4.2. The
previously mentioned symmetrical placement of the power modules was achieved
using U-shaped bus bars in each phase-leg. Moreover, in order to further reduce the
parasitic stray inductances of the external circuit layout, the positive and negative
bus bars were placed on top of each other. Thus, the stray inductance introduced by
the magnetic field generated with the flow of current, will partially be canceled as
shown in Figure 4.3. The figure shows in blue and red the magnetic field generated
by the positive (bottom) and negative (top) bus bars, respectively.
Moreover, the dc-capacitance has been distributed throughout the bus bars.
This ensures the lowest stray inductance from the source of energy to each power
module. Additionally, several interconnections between the prototype and the
power supply were made in order to ensure that the distributed capacitors behave
as a single capacitor. This is important in order to ensure a uniform distribution
of the current among the distributed capacitors, including the 2nd order harmonic.
40 CHAPTER 4. PARALLEL-CONNECTION OF SIC POWER MODULES
distributing the total capacitance throughout the complete bus-bars system. Fi-
nally, it is important to note that a suitable choice of power modules is crucial
for the final design. Low-current-rated power modules that can switch faster than
higher-current-rating ones (as the SiC JFET power modules presented in the pre-
vious chapter) in combination with a successful parallel connection will enable a
fast switching performance and high current capability. A more detailed descrip-
tion of the design process and how the components were selected can be found in
[Publication III – V].
(a)
(b)
Figure 4.5: (a) Schematic diagram of the double-pulse test set-up, (b) Photograph
of the prototype.
4.3. EXPERIMENTAL VERIFICATION 43
Transient Analysis
For the DPT the dc-power supply was set to 700 V. Figure 4.6 shows the
current sharing of the power modules. It can be noted that the power modules
No. 1 and No. 8, conduct significantly higher currents than other power modules,
especially power module No. 6. There are several possible explanations to this phe-
nomenon. One possible reason is the closer placement of power modules
No. 1 and No. 8 to the output of the phase-leg prototype, which means a lower
stray parasitic inductance. Another reason could be a mismatch between the device
characteristics of the SiC MOSFET power modules, such as the transconductance
or on-state resistance. Nevertheless, except for power module No. 6, the remaining
power modules share the static current with a maximum difference of 25 %.
Figure 4.6: Current sharing of the SiC MOSFETs. Measured drain current of each
SiC power module.
(a)
(b)
Figure 4.7: Measured drain current of each SiC power module. (a) Turn-ON tran-
sient of the SiC MOSFETs. (b) Turn-OFF transient of the SiC MOSFETs.
Steady-State Analysis
transient analysis. This is mainly due to the continuous operation, which increases
the temperature of operation. The increased temperature of operation causes the
power modules to operate well within the positive-temperature-coefficient range of
the on-state resistance, which results in an auto-balancing mechanism of the module
currents.
Figure 4.8: Current sharing of the SiC MOSFETs. Measured drain current of each
SiC power module.
A closer look to the switching transients was performed. Figure 4.9a shows
the turn-ON transient of the currents. The turn-ON process takes approximately
50 ns and the current sharing is more uniform than in the transient analysis case.
Figure 4.9b shows the turn-OFF transient, which takes approximately 50 ns as well
with a current sharing that is also more uniform than in any of the previous cases.
Moreover, it is possible to note a delay in the turn-OFF process of the current in
Module No. 2. This could be due to several reasons such as a mismatch in the
MOSFET power module characteristics or a difference in the stray inductances
of the interconnections of the modules. Moreover, an additional possible reason
would be a difference in the current measurement probes. Further information
about the results and the current sharing among the power modules can be found
in [Publications III – V].
Thermal Analysis
During the steady-state analysis at the load current, a thermal analysis was
also performed. Due to physical constraints of the prototype, two thermal images
are needed in order to see the temperature of all the power modules. Therefore,
two images were recorded, one from each side of the prototype, once steady-state
conditions were reached. Figure 4.10 shows infrared images with the temperatures
46 CHAPTER 4. PARALLEL-CONNECTION OF SIC POWER MODULES
(a)
(b)
Figure 4.9: Turn-ON (a) and turn-OFF (b) transient of the phase-leg test set-
up prototype. Measured drain current of the SiC MOSFETs, M8 (purple line,
20 A/div), drain current of the SiC MOSFET, M2 (pink color, 20 A/div), drain
current of the SiC MOSFET, M6 (green color, 20 A/div), drain current of the SiC
MOSFET, M4 (yellow color, 20 A/div), (time-base 50 ns/div).
of the modules during steady-state operation. Table 4.3 presents the average tem-
perature of each module. The average temperature of the modules is 38.7o C, and
the maximum difference of temperature between two modules is 1.5o C. From these
results, it is concluded that the loss distribution among the modules is expected to
be sufficiently uniform.
4.3. EXPERIMENTAL VERIFICATION 47
(a) (b)
Figure 4.10: Infrared camera image of the dc/dc converter during steady-state
operation (a) right side and (b) left side.
(a)
(b)
Figure 4.11: (a) Photograph of the experimental set-up, (c) Photograph of the
three-phase VSC.
system would be approximately the same as using a motor load. This is mainly
due to the reverse conduction capability of the SiC MOSFETs. At these condi-
tions, (i.e. a temperature of operation of approximately TJ = 75o C, and a rated
current of approximately 45 ARMS per module) the voltage drop across the channel
4.3. EXPERIMENTAL VERIFICATION 49
Figure 4.12: Inverter waveforms at nominal power of 312 kVA and switching fre-
quency of 20 kHz. Measured line-to-line voltage, (purple line, 500 V/div), dc-link
voltage (pink color 500V/div), phase current 1, (green color, 500 A/div), phase
current 2, (yellow color, 500 A/div), (time-base 1 ms/div).
does not exceed the knee voltage of the body diode and the anti-parallel Schottky
diode. Thus, almost no current is flowing through the diodes during reverse con-
duction. Moreover, during this time, the channel of the SiC MOSFETs presents
approximately the same on-state resistance as in forward conduction. Therefore,
approximately the same losses for the inverter are expected independently of the
power factor [100].
Loss Analysis
In order to calculate the efficiency of the system, power measurements were
performed. A screen-shot of the powermeter measurements at nominal power du-
ring steady-state conditions is shown in Figure 4.13. This measurement employs
four different measurement elements, which are shown in the figure. Each element
is composed of power/voltage/current measurement. Moreover, the first element
corresponds to the total output of the three-phase system. The next two correspond
to two different output phases, while the last one, Element 5, corresponds to the
input measurement on the dc side of the VSC.
As shown in Figure 4.14, the total measured losses, PTOTAL , can be subdivided
into copper losses, PCOPP , and iron losses, PIRON , which are both dissipated in the
inductive load, and, the conduction losses, PCD , and switching losses, PSW , both
of which are dissipated in the SiC power modules.
Furthermore, based on the presented classification of the losses, the following
50 CHAPTER 4. PARALLEL-CONNECTION OF SIC POWER MODULES
Figure 4.13: Screen-shot of the power meter during operation at nominal power of
312 kVA and switching frequency of 20 kHz.
Figure 4.14: Classification of the four components of the total measured losses.
the inverter, PINV , are found to be 2.29 kW at the rated power. Moreover, the
converter efficiency, calculated using (3.5), is found to be approximately 99.3 %
at rated conditions. Thus, the total power loss of the inverter is approximately
0.7 %, which is slightly lower than estimated in the previous chapter. Further
analysis related to the accuracy of the measurements can be found in [Publications
IV – V].
4.4 Conclusion
When parallel-connection is chosen as the design methodology of high-efficiency
converters rated for high-power applications, special considerations need to be taken
into account regarding the parasitic components of the set-up. By ensuring a low
inductive set-up and properly selecting the power modules, a uniform distribution
of the currents among the power modules as well as a fast switching performance
could be achieved. These elements affect directly the switching losses and, there-
fore, the efficiency of the system. Moreover, a more uniform distribution of the
currents was recorded at high powers than at low powers. It is believed that
the positive-temperature-coefficient of the on-state resistances works as an auto-
balancing mechanism of the current sharing among the parallel-connected power
modules.
A 312 kVA three-phase SiC inverter with high efficiency has been designed,
built, and experimentally verified using parallel-connection as construction metho-
dology. From the measurements, the total losses of the main circuit were found
to be 2.29 kW at 312 kVA, which corresponds to an efficiency of 99.3 %. The key
to reach such a high efficiency was a combination of a successful massive parallel-
connection of SiC power modules, using diode-less operation, and high switching
speeds.
Chapter 5
Reliability Analysis
5.1 Introduction
If parallel-connection is chosen as the design methodology for the construction
of a high-efficiency converter using SiC power devices many issues could arise.
Design complexity, current distribution among the parallel-connected devices, and
the so-called “Miller effect” are some examples of the many considerations that have
to be taken into account in order to reach a successful design. However, one of the
most important issues that might come to surface is the reliability. Increasing the
amount of components such as in parallel-connection has a negative impact on the
total reliability of the system. Additional components increase the probability of
failure of the complete system. Thus, an undesired trade-off between high-efficiency
and reliability is introduced when using parallel-connection as design methodology.
Several studies related to the reliability and robustness of SiC power devices
have been performed over the last years. These studies have been focusing on
different aspects and failure modes of the currently available SiC devices. These
aspects are short-circuit behavior and protection [88,101,102], long-term reliability
[103–113], gate-oxide stability and threshold voltage instability [114–120], as well as
high-temperature conditions [121–124]. However, none of these works have focused
their studies from the system perspective. Thus, it is the main goal of this part
of the work to investigate the reliability aspects of a SiC power electronic system
using the information derived from reliability tests.
In this chapter, an estimation of the expected lifetime of a three-phase VSC
similar to the one described in the previous chapter, using parallel-connection as
main design method will be performed. Additionally, important information re-
garding which parameters govern the reliability of the system and how to improve
it will be given.
53
54 CHAPTER 5. RELIABILITY ANALYSIS
(a)
(b)
Figure 5.1: (a) Photograph of the SiC power module, (b) Layout of the proposed
VSC.
Considering the rated conditions of this application, the proposed VSC, and
extrapolating the mean time to failure (MTTF) from manufacturer accelerated
tests, available in [104,105], Figure 5.2, Figure 5.3, and Figure 5.4 were plotted. Two
5.2. RELIABILITY ANALYSIS 55
different failure modes were identified in the case of the SiC MOSFETs, the gate-
source voltage (VGS ) stress and the drain-source voltage (VDS ) stress. Temperature
stress was identified as failure mode for the power SiC diodes. Similarly, in [126],
the MTTF value of the dc-link capacitors was derived. Table 5.1 summarizes the
extrapolated MTTF values (hours and years) and the derived failure rate of the
components. Moreover, the definition used for the failure rate is the reciprocal of
the MTTF in years.
8
10
MTTF [hours]
6
10
4
10
2
10
15 20 25 30 35 40
Vgs Stress [V]
8
10
MTTF [hours]
6
10
4
10
2
10
8
10
7
10
6
10
MTTF [Hours]
5
10
4
10
3
10
2
10
150 200 250 300 350 400
Temperature [deg. C]
Figure 5.4: Extrapolated MTTF of 6x107 hours at 175o C for the SiC power diode
[104].
12
X 12
X
λM oduleX = λDiodei + λM OSF ETiX (5.1)
n=1 n=1
5.2. RELIABILITY ANALYSIS 57
where the subscript X represents each failure mode. Table 5.2 summarizes the
extrapolated MTTF values (years) and the derived failure rate of the components.
Figure 5.5: Reliability block diagram for the SiC power module.
Table 5.2: MTTF and failure rate of the SiC MOSFET power module.
Failiure Mode VGS Stress VDS Stress
MTTF [years] 79.434 296.80
Failure Rate, λ 0.0126 0.0034
Furthermore, in addition to the two different failure modes that have been
analyzed, two cases of study have been performed in this work. The first case is
called the no-redundancy, while the second case takes advantage of the parallel
connection of power modules and introduces the, so-called, active redundancy.
No-Redundancy
The no-redundancy case means that all components must be working properly
to consider that the system is in the safe state. Therefore, as for the power module
analysis, all the components of the system are connected in series from the reliability
perspective as shown in Figure 5.6. Table 5.3 shows the calculated MTTFs for each
failure mode using:
12
X 12
X
λInvX = λCapi + λM oduleiX . (5.2)
n=1 n=1
Moreover, the failure rates for the rated conditions were also calculated and
presented in Table 5.3. Finally, it is possible to note that the life expectancy of
the proposed VSC is dominated by the gate voltage stress, and it is approximately
6.57 years. This value is low compared to the industry standard. Therefore, further
analysis on how to improve this value should be performed.
58 CHAPTER 5. RELIABILITY ANALYSIS
Table 5.3: MTTF and failure rate of the high-efficiency SiC three-phase inverter
without redundancy.
Failiure Mode VGS Stress VDS Stress
MTTF [years] 6.57 24.11
Failure Rate, λ 0.1521 0.0415
Figure 5.6: Reliability block diagram for the high-efficiency SiC three-phase inverter
without redundancy.
Active Redundancy
In the case of active redundancy it is possible to consider two possible scena-
rios. In the first scenario no additional modules are considered for each phase-leg,
and the VSC has the same amount of power modules as in the no-redundancy
case. Second, additional power modules are included in the design, allowing the
inverter to continue operation at efficiencies higher than 99 %. However, in this
chapter only the first case will be analyzed. Moreover, further information includ-
ing the analysis of the second case are included in [Publications VI – VII]. It
is important to note that in order to achieve the targeted redundancy, additional
components should be included in the switching loop such as disconnectors [127].
Additional descriptions on how these disconnectors operate and how to implement
them in the proposed VSC are given in [Publications VI – VII]. By introduc-
ing these additional components, parasitic inductances that modify the switching
performance, and consequently the efficiency of the system will be added in the
switching loop. Nevertheless, for this work these additional components as well as
gate-drive units are considered to be more reliable than the implemented power
modules and capacitors.
The described VSC for this reliability analysis requires only that half of the
modules are in operation at rated conditions, i.e. two power modules per phase-leg
in order to operate at the rated conditions (see Figure 5.7). However, it must be
noted that when two power modules have failed, the inverter efficiency will drop
to 98.7 %, i.e. lower than the targeted 99 %. Moreover, in order to perform the
reliability analysis and determine the MTTF of the system, further calculations
must be performed, especially when the fact that the failure rate is not constant
is considered. For redundant systems, the failure rate is dependent on the mission
times. In particular, a higher reliability is expected for shorter mission times.
Furthermore, in order to calculate the failure rate of the phase-legs and system,
four different states are considered for each phase of the inverter. These states are:
5.2. RELIABILITY ANALYSIS 59
(a) fully functional (all power modules working), (b) one failed (one power module
failed), (c) two failed (two power modules failed), and (d) phase failed (more than
two power modules failed).
Figure 5.7: Reliability block diagram for the high-efficiency SiC three-phase inverter
with active redundancy.
Furthermore, using the reliability methodology for series and parallel systems
the failure rate of each state is calculated. For instance, in the case of the first
state (a), Figure 5.8 shows the equivalent system of a single phase of the VSC. The
failure rate for this state is calculated using Eq. 5.3.
2λM odule
λP haseX = P6 1X (5.3)
n=1 i
Similarly, the failure rates of the other three states were calculated.
Figure 5.8: Reliability block diagram of the state (a) for one phase of the high-
efficiency SiC three-phase inverter with active redundancy.
in this case are the failure rates [128, 129]. Additionally, these probabilities of
failure, which are dependent on the mission times, were calculated for each failure
mode and each state as seen in Figure 5.9a and Figure 5.9b. Moreover, from these
probabilities, the failure rate of the phase-legs could be calculated. For the total
system, the three phase-legs are connected in series from the reliability perspective.
0.8
Probability
0.6
0.4
Fully Functional
One Failed
0.2 Two Failed
Phase Failed
System Failed
0
0 20 40 60 80 100
Years
(a)
0.8
Probability
0.6
0.4
Fully Functional
One Failed
0.2 Two Failed
Phase Failed
System Failed
0
0 50 100 150 200 250 300 350 400
Years
(b)
Figure 5.9: Probability of each state of Markov Models and Probability of system
failure regarding the mission time for the (a) gate-source voltage stress, (b) drain-
source voltage stress.
Lastly, the failure rate of the system could be calculated as well as the pro-
bability density function, which contains the MTTF, as shown in Figure 5.10a and
Figure 5.10b. It must be noted that for all the reliability analyses, these plots (VDS
stress and VGS stress) are extended to 400 years and 100 years respectively in order
to illustrate how the failure rate changes with respect to the mission times. The
5.2. RELIABILITY ANALYSIS 61
probability of the system to fail and therefore, the failure rate, increases with time
as expected. The probability of the system to be in the first state, where all power
modules are properly operating, rapidly falls with time. Finally, it can be noted
that including active-redundancy in the system increases the MTTF approximately
5.5 times, to 34.38 years, as shown in Table 5.4.
Table 5.4: MTTF and failure rate of the high-efficiency SiC three-phase inverter
with active redundancy.
Failiure Mode VGS Stress VDS Stress
MTTF [years] 34.38 133.34
0.1
Failure Rate of the System
Probability Density Function (PDF)
0.08
0.06
0.04
0.02
0
0 20 40 60 80 100
Years
(a)
0.03
Failure Rate of the System
Probability Density Function (PDF)
0.025
0.02
0.015
0.01
0.005
0
0 50 100 150 200 250 300 350 400
Years
(b)
Figure 5.10: Failure rate and probability density function regarding the mission
time for the (a) gate-source voltage stress, (b) drain-source voltage stress.
62 CHAPTER 5. RELIABILITY ANALYSIS
300
No Redundancy
Active Redundancy − No Add. Modules
100
Active Redundancy − With Add. Modules
50
MTTF [Years]
10
5
1
0.5
15 20 25
Gate Voltage [V]
Figure 5.11: MTTF as function of the gate voltage, no-redundancy, (blue), active
redundancy with no additional modules (dash green).
Furthermore, reducing the gate-source voltage not only impacts the reliability
of the system but also affects different parameters of the system such as the on-
state resistance and the switching speeds. Thus, affecting the total losses of the
system and consequently, the efficiency. Therefore, an estimation of how the losses
vary depending on the gate voltage was performed using a simulation model of the
power module. The simulation model used in this part of the work was developed
by Dr. Konstantin Kostov from Acreo AB. Using ANSYS Q3D, the power module
was implemented in the software and the parasitic elements were derived. Finally,
the power module was simulated with LTSPICE including these derived parasitic
elements [92]. Using the developed simulation model, it is possible to estimate the
switching losses as well as the conduction losses with respect to the gate voltage,
as shown in Figure 5.12. As expected, the lower the VGS , the higher the losses.
By decreasing the positive gate-source voltage by 3 V, the reliability of the
system is increased approximately 4 times (solid line in Figure 5.13) and the effi-
ciency is reduced by approximately 0.1 %. Therefore, the reduction of the efficiency
is not significant compared to the MTTF improvement.
However, as described above, decreasing the gate voltage will increase the
5.2. RELIABILITY ANALYSIS 63
3000
Conduction
Switching
Total
2500
Losses [W]
2000
1500
1000
500
15 16 17 18 19 20 21 22 23 24 25
Gate Voltage [V]
Figure 5.12: Conduction losses (blue), Switching losses (green), Total losses (red)
regarding of the gate voltage.
40 99.3
35
99.2
30
MTTF [Years]
Efficiency [%]
25 99.1
20
15 99
10
MTTF at Constant Temp. 98.9
5 MTTF with Temp. Effect
Efficiency
98.8
15 20 25
Gate Voltage [V]
Figure 5.13: MTTF (blue) and Efficiency (green) regarding of the gate voltage.
losses of the system, and therefore, the junction temperature of the SiC chip will
rise. One possibility in order to maintain the validity of this study is to recalculate
the cooling system in order to keep the junction temperature within the tempera-
ture range used in this calculation. However, this might lead to a bulkier system.
Another possibility is to consider the temperature variation into the analysis and
using the so-called Arrhenius relation, an acceleration factor could be calculated
with respect to the reference temperature. Thus, estimating how the MTTF is
affected by the temperature. Finally, assuming water-cooling with a proper water
flow rate in order to guarantee a turbulent flow, the junction temperature for each
gate voltage can be calculated as
64 CHAPTER 5. RELIABILITY ANALYSIS
Pi
TJi = (TJTEST − Tw ) + Tw , (5.4)
PTEST
where
Pi = Ah(TJi − Tw ), (5.6)
where Tw is the average temperature of the cooling water between the input and
output, TJTEST is the junction temperature at the reference point, TJi the junction
temperature at each gate voltage, PTEST is the losses at the reference point, Pi
the losses at each gate voltage, A the total area for heat dissipation, and h is the
heat-transfer coefficient.
Using the Arrhenius relation [130], the acceleration factor, AF , which results
from operating a device at elevated temperatures is calculated with the following
equation:
Ea 1
k (T − T1 )
AF = e JTEST Ji
, (5.7)
where k is the Boltzmann constant, TTEST the junction temperature at the reference
point, and Ea the activation energy required to initiate a specific type of failure
mode. Finally, the new MTTF considering the temperature variations is calculated
from
1
MTTF = . (5.8)
λAF
Finally, Figure 5.13 shows the modified MTFF (dashed line) with respect to
the gate voltage. It must be noted that decreasing the gate voltage could also lead
to a lower reliability due to the high temperature of the junction. Therefore, an
optimal point can be found where both high efficiency and high reliability could be
achieved. In this case, by decreasing the gate voltage by 3 V, the MTTF increases
approximately 2.5 times and the efficiency is reduced by approximately 0.1 %.
Further information on how the losses were estimated and a detailed discussion of
the results are included in [Publications VI – VII].
5.3 Conclusion
A reliability analysis has been performed on a VSC using parallel connection
as design methodology. Two different failure modes have been studied: the VGS
stress and the VDS stress. Additionally, two possible cases were analyzed: no-
redundancy and active-redundancy. This analysis has shown that the gate-source
voltage stress determines the reliability and life expectancy of the entire system.
5.3. CONCLUSION 65
Moreover, decreasing the positive gate-source voltage could increase the relia-
bility of the system approximately 2.5 to 4 times without affecting the efficiency
significantly. Similar results could be achieved by including redundancy to the sys-
tem. Finally, an undesired trade-off between the high efficiency targeted and the
reliability of the system is introduced.
Chapter 6
6.1 Introduction
Wide-bandgap materials offer several advantages compared to the currently
used semiconductors. These advantages could be listed as higher efficiency, higher
switching frequency and higher maximum operation temperatures. These charac-
teristics match with the new design trends of power electronics. Therefore, replacing
silicon technology with WBG semiconductors seems to be a natural step.
For instance, the large bandgap that SiC offers compared to silicon, enables
the capability to operate at temperatures higher than 600o C [131]. Moreover, the
high thermal conductivity makes SiC an excellent candidate for harsh environments
such as aviation, automotive, oil and gas drilling, and space applications. In partic-
ular, for Venus explorations [9–11,132–134] there is no better option than SiC. More-
over, the high-temperature performance of several SiC power devices has been eval-
uated lately [135–138]. In the literature it is possible to find studies related to the
high-temperature performance of SiC power diodes [123,139], SiC JFETs [140,141],
SiC BJTs [142], and SiC MOSFETs [122,124,143]. Furthermore, SiC integrated cir-
cuits such as operational amplifiers as well as PWM oscillators, and linear regulators
have been demonstrated and tested at high temperatures [144–148].
On the other hand, when high power density is targeted, GaN offers simi-
lar advantages as SiC [149, 150]. However, until now GaN has not been used for
extreme high-temperature applications. Moreover, if higher efficiencies and power
densities are considered, operation at very low temperatures can be a solution. At
lower temperatures, lower on-state losses are anticipated, which could lead to an
increase in the overall power conversion efficiency [151–155]. In general, reducing
the temperature could introduce the benefits of WBG into many applications, such
as deep space explorations, and help in the electrification process of many trans-
67
68 CHAPTER 6. EXTREME TEMPERATURE APPLICATIONS
portation systems, in particular future electric aircrafts, making these systems more
powerful, efficient and environmentally friendly [156–158].
In this chapter, an evaluation of different parts including inductors as well as
power devices have been performed at different extreme temperatures.
High-Temperature Inductors
In order to fully take advantage of the high-temperature capability offered
by SiC, it is necessary for every part of the system to be able to operate at high
temperatures. Even though several studies have been performed regarding the high-
temperature performance of SiC power devices and integrated chips built with SiC
technology, passive components able to withstand such temperatures have received
relatively little attention.
Magnetic components such as inductors and transformers are considered to be
an essential part of almost every power system [159]. If high-temperature magnetic
components are targeted, powder cores have displayed a high saturation flux, a
high permeability, and a high Curie temperature. These characteristics make them
good candidates for high-temperature power converters. Therefore, in this work,
two different solutions for high-temperature inductors were designed, built, and
experimentally verified. The first solution was an inductor using a high-temperature
powder core. Second, an air coil inductor was also designed and built. In both
cases, a high-temperature cable composed of pure nickel, and a high grade mica
and ceramic isolation, was used.
Figure 6.1 shows two photographs of the powder core inductor, before and
after the high-temperature measurements. The test was performed at elevated
temperatures up to 600o C and at three different frequencies, 1 kHz, 10 kHz and
100 kHz. An inductance of 16 µH with less than 4 % variation was measured up
to 500o C, as shown in Figure 6.2. At temperatures above 500o C, which is the
Curie temperature of the used powder core, the inductance drops very rapidly as
expected.
Similarly, an air coil inductor was built and tested following the same test
methodology. Figure 6.3 shows two photograph of the air coil inductor, before and
after the high-temperature measurements. An inductance of 6 µH with less than 9
% variation was measured for the entire temperature range (up to 600o C), as shown
in Figure 6.4. Until now, no explanation to the observed dependence on temperature
could be found. It is the hypothesis of the author that a possible explanation to
the observed temperature dependence is changes in the geometry of the air coil
inductor due to the high temperature variations. A maximum inductance variation
of 10 % was measured at the three different test frequencies for each inductor.
6.2. HIGH-TEMPERATURE APPLICATIONS 69
Figure 6.1: Photograph of the powder core inductor. Before (Left) and after (Right)
the high-temperature measurements.
18
16
14
12
Inductance [μH]
10
1 kHz
4
10 kHz
100 kHz
2 1 kHz AH
10 kHz AH
100 kHz AH
0
100 200 300 400 500 600
Temperature [°C]
Figure 6.2: Inductance versus temperature for the powder core inductor.
Figure 6.3: Photograph of the air coil inductor. Before (Left) and after (Right) the
high-temperature measurements.
70 CHAPTER 6. EXTREME TEMPERATURE APPLICATIONS
5
Inductance [μH]
2
1 kHz
10 kHz
1 100 kHz
1 kHz AH
10 kHz AH
100 kHz AH
0
100 200 300 400 500 600
Temperature [°C]
Figure 6.4: Inductance versus temperature for the air coil inductor.
(a) (b)
Figure 6.5: (a) Photograph of the inside of the cryogenic probe station, (b) Photo-
graph of the cryogenic probe station.
5
On−State Resistance [mΩ]
On the other hand, following the constant current method, which shorted the
gate and the drain of the device, the threshold voltage was measured. For this
case, the drain current was set to 7 mA. At RT, the threshold voltage, VT , is equal
to 1.38 V, whereas a 16 % increase of VT was measured at −195o C, as shown in
Figure 6.7. Finally, no significant changes in VT were measured below −100o C.
Finally, the switching performance was verified using an standard double-
72 CHAPTER 6. EXTREME TEMPERATURE APPLICATIONS
1.65
1.55
1.5
1.45
1.4
1.35
−200 −175 −150 −125 −100 −75 −50 −25 0 25
Temperature [°C]
pulse test. Figure 6.8 and Figure 6.9 show the turn-ON and turn-OFF transients
respectively. The device is able to switch within the entire range of temperatures
and no significant changes in the switching waveforms were observed.
60
20°C
−50°C
50
−100°C
−150°C
40 −195°C
Voltage [V]
30
20
10
−10
10.04 10.05 10.06 10.07 10.08 10.09 10.1 10.11
Time [μs]
It is the hypothesis of the author that the decreasing delay times with tem-
perature are due to changes in the gate-drive unit at these low temperatures and
not due to changes in the power device. These changes in the delay times match
the values anticipated by the gate driver manufacturer. Further information on
how the supporting electronics performs at lower temperatures can be found in
[Publication IX].
6.3. CRYO-COOLED TEMPERATURE APPLICATIONS 73
60
50
40
Voltage [V]
30
20
10 20°C
−50°C
−100°C
0
−150°C
−195°C
−10
15.02 15.03 15.04 15.05 15.06 15.07
Time [μs]
Figure 6.11: Photograph of the 3-level switching cell (left) and GaN Inverter (right).
Note that only one single switching cell of the converter was used for this investi-
gation.
Figure 6.12 shows the GaN inverter placed inside the temperature chamber.
The dc-link voltage was 150 V and the switching frequency of the converter was
120 kHz. Moreover, Figure 6.13 shows the switching node voltage and the output
current of the inverter. The switching node voltage shows a typical PWM charac-
teristic as expected. At rated operation, the FCML converter has an output power
of 1 kW with an output current of 22.21 ARMS (i.e. 30 Apeak ) and output voltage
of 45 VRMS .
Figure 6.14 shows how the efficiency varies as a function of the temperature at
different output powers. The highest efficiency of 98.7 % was measured at −75o C
with an output power of 450 W. Moreover, a 33 % reduction in the losses was
6.3. CRYO-COOLED TEMPERATURE APPLICATIONS 75
Figure 6.13: Waveforms showing the output current and the switching node voltage
(VSW ) 3-level GaN Inverter.
98.8 220 W
98.6 320 W
450 W
98.4 550 W
660 W
98.2
Efficiency [%]
790 W
98 910 W
97.8
97.6
97.4
97.2
97
−80 −70 −60 −50 −40 −30 −20 −10 0 10 20 30
Temperature [°C]
Figure 6.14: Measured efficiency versus temperature for various power levels.
76 CHAPTER 6. EXTREME TEMPERATURE APPLICATIONS
0.78 220 W
320 W
0.76 450 W
0.74 550 W
660 W
0.72 790 W
THD [%]
910 W
0.7
0.68
0.66
0.64
0.62
−80 −70 −60 −50 −40 −30 −20 −10 0 10 20 30
Temperature [°C]
Figure 6.15: Measured THD versus temperature for various power levels.
Loss Analysis
A power-loss analysis was performed based on datasheet specifications and
the previously performed characterization of the device. The system losses could
be divided into transistor conduction, switching and capacitive losses (PTCOND ,
PTSW and PTCAP ), output inductor conduction losses (PLCOND ), output capacitor
conduction losses (PCCOND ), and the printed circuit board (PCB) resistive losses
(PPCB ). Moreover, using the loss calculator tool from the inductor manufacturer,
the core losses of the output inductor and the ac losses of the inductor were also
considered and estimated [162]. The aforementioned losses can be calculated using
(6.1) through (6.6).
1
PLCOND = · RL (T )IOUT 2 (6.4)
2
1
PCCOND = · RESR (T )IOUT 2 (6.5)
6
where VDS is the dc-link voltage, IOUT is the output RMS current, fsw is the switch-
ing frequency, and Coss is the output capacitance of the GaN power FET. Also, RL
is the equivalent resistance of the output inductor, EESR is the equivalent resis-
tance of the output capacitor, RPCB is the PCB resistance. All the aforementioned
resistances are temperature dependent and were modeled using the temperature
dependence of copper. The temperature dependence and value of RDS(ON) , the
on-state resistance of the GaN power FET, was extrapolated from the previous
characterization. Furthermore, EON and EOFF , the switching energies of the GaN
power FET, were calculated from
tON/OFF VDS IOUT
EON/OFF = , (6.7)
2
where tON/OFF are the turn-ON and turn-OFF switching times respectively. These
times were estimated using:
QGSW
tON/OFF = , (6.8)
IGON/OFF
where QGSW is the gate charge of the GaN power FET, and IGON/OFF are the gate
turn-ON and turn-OFF currents respectively during the switching process. These
currents are found from
VDR − VTH
IGON = (6.9)
RON
VTH
IGOFF = , (6.10)
ROFF
where VDR is the gate drive unit voltage, RON and ROFF the turn-ON and turn-OFF
resistances respectively, and VTH is the threshold voltage, which is temperature
dependent. The value and temperature dependence of VTH were extrapolated from
the previous characterization.
Figure 6.16 shows a comparison of the measured and combined estimated
power losses. The estimated losses agree well with the measured losses. The minor
mismatch between them is mainly due to measurement errors. Further analysis
related to the accuracy of the measurements can be found in [Publication X].
78 CHAPTER 6. EXTREME TEMPERATURE APPLICATIONS
20
Trans Cond
Trans Switch
Trans Coss
Losses [W]
15
Inductor Core Output
Inductor AC Output
Inductor Cond Output
10
Capacitor Output
PCB
Measured
5
0
−75 −50 −25 0 25
Temperature [°C]
(a)
3.5
3 Trans Cond
Trans Switch
2.5 Trans Coss
Losses [W]
0.5
0
−75 −50 −25 0 25
Temperature [°C]
(b)
Figure 6.16: Comparison between measured power losses and estimated losses
(breakdown model) versus temperature at (a) 220 W (5 ARMS output), (b) 910
W (21 ARMS output).
6.4. CONCLUSION 79
6.4 Conclusion
In this work, the high-temperature performance of inductors has been studied.
Two different designs have been tested up to 600o C. An experimental evaluation of
a 1 kW single-phase inverter at low temperatures was performed. The inverter con-
sists of a 3-level FCML converter using GaN power FETs with a switching frequency
of 120 kHz. At −75o C, a 33 % reduction in losses was achieved at rated power.
Moreover, no significant changes in the THD was observed with respect to the tem-
perature variation. A power loss breakdown was performed and a strong agreement
between the measured and estimated losses was achieved. The conduction loss of
the converter was identified as the most temperature sensitive component of the
losses.
Chapter 7
81
82 CHAPTER 7. CONCLUSIONS AND FUTURE WORK
Finally, the recently introduced GaN power FET was also investigated. A
device characterization has been performed down to cryogenic temperatures. An
85 % reduction of the on-state resistance and a 16 % increase of the threshold
voltage were measured at −195o C. An improvement of the overall efficiency of
the system when used at extreme low-temperature environments is anticipated. In
order to verify this, an experimental evaluation of a 1 kW single-phase inverter
at low temperatures was performed. The inverter consists of a 3-level flying ca-
pacitor multi-level converter using GaN power FETs with a switching frequency of
120 kHz. At −75o C, a 33 % reduction in losses was achieved at rated power.
Moreover, no significant changes in the THD were observed with respect to the
temperature variation. A power loss breakdown was performed and strong agree-
ment between the measured and estimated losses was achieved. The conduction
losses of the converter were identified as the most temperature sensitive component
of the losses.
83
Future work
As previously described in this thesis, parallel-connection of either discrete
components or power modules is a possible solution for high-efficiency high-power
applications. However, parallel-connection of single chips for high-performance
power modules is an area of further development. New generations of power mo-
dules able to fully take advantage of the WBG characteristics are needed.
It has been shown that problems associated with the “Miller effect” might
affect the stable operation of power modules with high current ratings. Several
gate-drive circuits which aim to mitigate this effect have been proposed. However,
in a phase-leg configuration, problems related to resonance between the Miller ca-
pacitance and the parasitic stray inductance might lead to accidental turn-ON or
self-sustained oscillations if high switching speeds are used. Thus, smart gate drivers
for power modules able to control how fast to charge and discharge the parasitic
capacitance of the power module must be designed. Controlling the switching speed
is a technique that will allow to limit the unwanted effects such as electromagnetic
interference (EMI) while achieving the highest possible efficiency.
Further studies related to the feasibility of power converters operating at
extremely high temperatures must be performed. Even though several studies have
been performed related to the high-temperature performance of several WBG power
devices as well as integrated circuits, much more attention must be paid to other
aspects such as packaging as well as the integration of the different components
that comprise a power converter.
Finally, more investigations have to be performed regarding extreme low-
temperature implementations and the benefits that could be achieved by lowering
down the operational temperatures of power converters.
Appended Publications
Publication I
J. Colmenares, D. Peftitsis, J. Rabkowski and H.-P. Nee,”Dual-Function
Gate Driver for a Power Module With SiC Junction Field-Effect Transistors,” in
Proc. of 2013 Energy Conversion Cong. and Exp. Asia (ECCE Asia), Jun. 2013,
pp. 245–250.
85
86
Publication II
J. Colmenares, D. Peftitsis, J. Rabkowski, D. Sadik, and H.-P. Nee,”Dual-
Function Gate Driver for a Power Module With SiC Junction Field Transistors,”
IEEE Trans. Power Electron., vol. 29, no. 5, pp. 2367–2379, May 2014.
93
94
Publication III
J. Colmenares, D. Peftitsis, J. Rabkowski, and H.-P. Nee, ”Switching Per-
formance of Parallel-Connected Power Modules With SiC MOSFETs,” in Proc. of
2014 Int. Power Electron. Conf. (IPEC–ECCE ASIA), May 2014, pp. 3712–3717.
109
110
Publication IV
J. Colmenares, D. Peftitsis, G. Tolstoy, D. Sadik , J. Rabkowski, and H.-P.
Nee, ”High-Efficiency Three-Phase Inverter With SiC MOSFET Power Modules for
Motor-Drive Applications,” in Proc. of 2014 IEEE Energy Conversion Congr. and
Exp. (ECCE), Sept. 2014, pp. 468–474.
117
118
Publication V
J. Colmenares, D. Peftitsis, J. Rabkowski, D. Sadik, G. Tolstoy and H.-P.
Nee, ”High-Efficiency 312–kVA Three-Phase Inverter using Parallel Connection of
Silicon Carbide MOSFET Power Modules,” IEEE Trans. Ind. Appl., vol. 51, no.
6, pp. 4664–4676, Nov.-Dec. 2015.
127
128
Publication VI
J. Colmenares, D. P. Sadik, P. Hilber and H.-P. Nee, ”Reliability Analysis
of a High-Efficiency SiC Three-Phase Inverter for Motor-Drive Applications,” in
Proc. of 2016 IEEE Applied Power Electron. Conf. and Exp. (APEC), Mar. 2016,
pp. 746–753.
143
144
Publication VII
J. Colmenares, D.-P. Sadik, P. Hilbert and H.-P. Nee, ”Reliability Analysis
of a High-Efficiency SiC Three-Phase Inverter,” IEEE Trans. Emerg. Sel. Topics
Power Electron., vol. 4, no. 3, pp. 996–1006, Sept. 2016.
153
154
Publication VIII
J. Colmenares, S. Kargarrazi, H. Elahipanaht, H.-P. Nee and C.-M. Zetter-
ling, ”High Temperature Passive Components for Extreme Environments,” 2016
IEEE 4th Workshop on Wide Bandgap Power Devices and Appl. (WiPDA), ac-
cepted for publication.
167
168
Publication IX
J. Colmenares, T. Foulkes, C. Barth, T. Modeer and R. Pilawa-Podgurski,
”Experimental Characterization of Enhancement Mode GaN Power FETs at Cryo-
genic Temperatures,” 2016 IEEE 4th Workshop on Wide Bandgap Power Devices
and Appl. (WiPDA), accepted for publication.
171
172
Publication X
J. Colmenares, C. Barth, T. Foulkes, T. Modeer and R. Pilawa-Podgurski,
”Experimental Evaluation of a 1 kW, Single-Phase, 3–Level GaN Inverter at Ex-
treme Cold Environments,” 2017 IEEE Applied Power Electron. Conf. and Exp.
(APEC), submitted for review.
175
176
List of Figures
3.1 (a) Schematic diagram of the parallel-connected SiC JFET chips popu-
lating one switch position (b) Simplified schematic diagram of one switch
position. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.2 Photograph of the power module with SiC JFETs. . . . . . . . . . . . . 22
3.3 Schematic diagram of version I of the gate driver for the SiC JEFT
power module. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.4 Schematic diagram of the final version of the gate driver for the SiC
JEFT power module. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.5 Photograph of the gate driver PCB prototype for the SiC JEFT power
module. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.6 Schematic diagram of the double-pulse test setup for SiC the JEFT
power module. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.7 Photograph of the gate driver PCB prototype for the SiC JEFT power
module. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.8 Turn-ON and turn-OFF transients of the SiC JFET power module with
gate driver version I with VDC equal to 25 V. Measured gate-source
voltage of the SiC JFET, M1 (yellow line, 20 V/div), gate current of the
SiC JFET, M1 (green line, 20 A/div), drain-source voltage of the SiC
JFET, M1 (purple line, 50 V/div), drain current of the SiC JFET, M1
(pink line, 20 A/div), (time-base 500 ns/div). . . . . . . . . . . . . . . . 27
3.9 Turn-ON transient of the SiC JFET power module with the final version
of the gate driver. Measured drain-source voltage of the SiC JFET, M1
(purple line, 200 V/div), drain current of the SiC JFET, M1 (pink line,
100 A/div), (time-base 100 ns/div). . . . . . . . . . . . . . . . . . . . . 28
3.10 Turn-OFF transient of the SiC JFET power module with final version
of the gate driver. Measured drain-source voltage of the SiC JFET, M1
(purple line, 200 V/div), drain current of the SiC JFET, M1 (pink line,
100 A/div), (time-base 100 ns/div). . . . . . . . . . . . . . . . . . . . . 29
85
86 List of Figures
3.11 Small-signal model for a power module FET oscillator with the final
version of the gate driver [8]. . . . . . . . . . . . . . . . . . . . . . . . . 30
3.12 (a) Photograph of the power module with SiC MOSFETs and, (b) Sim-
plified schematic diagram of a single switch position. . . . . . . . . . . . 31
3.13 Schematic diagram of the gate driver with the short-circuit protection. . 32
3.14 Photograph of the gate-drive unit with the short-circuit protection. . . . 33
3.15 Short–circuit detection using the proposed gate driver. Measured drain-
source voltage of the SiC MOSFET, (purple line, 200 V/div), drain cur-
rent of the SiC MOSFET, (pink color, 500 A/div), logic signal from the
latch (green color, 50 V/div), gate-source voltage of the SiC MOSFET,
(yellow color, 50 V/div), (time-base 200 ns/div). . . . . . . . . . . . . . 34
3.16 Turn-ON switching waveforms of the power module with the proposed
gate driver. Measured drain-source voltage of the SiC MOSFET, (pur-
ple line, 200 V/div), drain current of the SiC MOSFET, (pink color,
50 A/div), gate-source voltage of the SiC MOSFET, (yellow color, 20
V/div), (time-base 50 ns/div). . . . . . . . . . . . . . . . . . . . . . . . 35
3.17 Turn-OFF switching waveforms of the power module with the proposed
gate driver. Measured drain-source voltage of the SiC MOSFET, (pur-
ple line, 200 V/div), drain current of the SiC MOSFET, (pink color,
50 A/div), gate-source voltage of the SiC MOSFET, (yellow color, 20
V/div), (time-base 50 ns/div). . . . . . . . . . . . . . . . . . . . . . . . 35
4.1 Relative on-state power loss versus junction temperature for ten parallel-
connected modules (CAS100H12AM1) - calculation using datasheet values. 39
4.2 Schematic diagram of a single phase set-up. . . . . . . . . . . . . . . . . 40
4.3 Layout of the bus bars for the inverter. . . . . . . . . . . . . . . . . . . 40
4.4 Photograph of the partially built three-phase inverter. . . . . . . . . . . 41
4.5 (a) Schematic diagram of the double-pulse test set-up, (b) Photograph
of the prototype. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.6 Current sharing of the SiC MOSFETs. Measured drain current of each
SiC power module. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4.7 Measured drain current of each SiC power module. (a) Turn-ON tran-
sient of the SiC MOSFETs. (b) Turn-OFF transient of the SiC MOSFETs. 44
4.8 Current sharing of the SiC MOSFETs. Measured drain current of each
SiC power module. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.9 Turn-ON (a) and turn-OFF (b) transient of the phase-leg test set-up
prototype. Measured drain current of the SiC MOSFETs, M8 (purple
line, 20 A/div), drain current of the SiC MOSFET, M2 (pink color, 20
A/div), drain current of the SiC MOSFET, M6 (green color, 20 A/div),
drain current of the SiC MOSFET, M4 (yellow color, 20 A/div), (time-
base 50 ns/div). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.10 Infrared camera image of the dc/dc converter during steady-state oper-
ation (a) right side and (b) left side. . . . . . . . . . . . . . . . . . . . . 47
List of Figures 87
4.11 (a) Photograph of the experimental set-up, (c) Photograph of the three-
phase VSC. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4.12 Inverter waveforms at nominal power of 312 kVA and switching fre-
quency of 20 kHz. Measured line-to-line voltage, (purple line, 500
V/div), dc-link voltage (pink color 500V/div), phase current 1, (green
color, 500 A/div), phase current 2, (yellow color, 500 A/div), (time-base
1 ms/div). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
4.13 Screen-shot of the power meter during operation at nominal power of
312 kVA and switching frequency of 20 kHz. . . . . . . . . . . . . . . . . 50
4.14 Classification of the four components of the total measured losses. . . . 50
5.1 (a) Photograph of the SiC power module, (b) Layout of the proposed
VSC. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
5.2 Extrapolated MTTF of 9.7x106 hours at VGS = 20 V. TDDB of gate
oxide on 20 A for the Gen2 of SiC MOSFETs [105]. . . . . . . . . . . . 55
5.3 Extrapolated MTTF of 6.5x107 hours at VDS = 1000 V. Accelerated
field testing at 150o C for the Gen2 of SiC MOSFETs [105]. . . . . . . . 55
5.4 Extrapolated MTTF of 6x107 hours at 175o C for the SiC power diode [104]. 56
5.5 Reliability block diagram for the SiC power module. . . . . . . . . . . . 57
5.6 Reliability block diagram for the high-efficiency SiC three-phase inverter
without redundancy. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
5.7 Reliability block diagram for the high-efficiency SiC three-phase inverter
with active redundancy. . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
5.8 Reliability block diagram of the state (a) for one phase of the high-
efficiency SiC three-phase inverter with active redundancy. . . . . . . . . 59
5.9 Probability of each state of Markov Models and Probability of system
failure regarding the mission time for the (a) gate-source voltage stress,
(b) drain-source voltage stress. . . . . . . . . . . . . . . . . . . . . . . . 60
5.10 Failure rate and probability density function regarding the mission time
for the (a) gate-source voltage stress, (b) drain-source voltage stress. . . 61
5.11 MTTF as function of the gate voltage, no-redundancy, (blue), active
redundancy with no additional modules (dash green). . . . . . . . . . . 62
5.12 Conduction losses (blue), Switching losses (green), Total losses (red)
regarding of the gate voltage. . . . . . . . . . . . . . . . . . . . . . . . . 63
5.13 MTTF (blue) and Efficiency (green) regarding of the gate voltage. . . . 63
6.1 Photograph of the powder core inductor. Before (Left) and after (Right)
the high-temperature measurements. . . . . . . . . . . . . . . . . . . . . 69
6.2 Inductance versus temperature for the powder core inductor. . . . . . . 69
6.3 Photograph of the air coil inductor. Before (Left) and after (Right) the
high-temperature measurements. . . . . . . . . . . . . . . . . . . . . . . 69
6.4 Inductance versus temperature for the air coil inductor. . . . . . . . . . 70
6.5 (a) Photograph of the inside of the cryogenic probe station, (b) Photo-
graph of the cryogenic probe station. . . . . . . . . . . . . . . . . . . . . 71
88 List of Figures
3.1 Device parameters of the SiC JFETs and SiC Schottky diodes. . . . . . 22
3.2 Device parameters of the SiC JFET Power Module. . . . . . . . . . . . 22
3.3 Parameters of the double-pulse test set-up for the SiC JEFT power module. 25
89
Bibliography
[2] J. Rabkowski, D. Peftitsis, and H.-P. Nee, “Silicon Carbide Power Transistors:
A New Era in Power Electronics Is Initiated,” IEEE Ind. Electron. Mag.,
vol. 6, pp. 17–26, Jun. 2012.
91
92 BIBLIOGRAPHY
[24] J. Rabkowski, D. Peftitsis, and H.-P. Nee, “Design Steps Toward a 40–kVA
SiC JFET Inverter With Natural-Convection Cooling and an Efficiency Ex-
ceeding 99.5 %,” IEEE Trans. Ind. Appl., vol. 49, pp. 1589–1598, Jul. 2013.
[26] F. Xu, B. Guo, L. M. Tolbert, F. Wang, and B. J. Blalock, “An All-SiC Three-
Phase Buck Rectifier for High-Efficiency Data Center Power Supplies,” IEEE
Trans. Ind. Appl., vol. 49, pp. 2662–2673, Nov. 2013.
[35] G. Kelner, S. Binari, K. Sleger, and H. Kong, “SiC MESFET’s and Buried-
Gate JFET’s,” IEEE Electron Device Lett., vol. 8, pp. 428–430, Sept. 1987.
[37] Y. Chang and J. Kuo, “SiC vs. Si: Two-Dimensional Analysis of Quasi-
Saturation Behavior of DMOS Devices Operating at Elevated Temperatures,”
in 4th Int. Conf. on Solid-State and Integrated Circuit Tech., pp. 298–300,
Oct. 1995.
[38] P. Friedrichs and R. Rupp, “Silicon Carbide Power Devices – Current Devel-
opments and Potential Applications,” in IEEE Eur. Conf. on Power Electron.
and Appl., Sept. 2005.
[39] P. Friedrichs, “Silicon Carbide Power Devices – Status and Upcoming Chal-
lenges,” in IEEE Eur. Conf. on Power Electron. and Appl., pp. 1–11, Sept.
2007.
[42] J. K. Lim, M. Bakowski, and H. P. Nee, “Design and Gate Drive Considera-
tions for Epitaxial 1.2 kV Buried Grid N-On and N-Off JFETs for Operation
at 250o C,” Mater. Sci. Forum, vol. 645-648, pp. 961–964, Apr. 2009.
[65] P. Ranstad, H.-P. Nee, J. Linner, and D. Peftitsis, “An Experimental Eval-
uation of SiC Switches in Soft-Switching Converters,” IEEE Trans. Power
Electron., vol. 29, pp. 2527–2538, May 2014.
[68] P. Ranstad and H.-P. Nee, “On Dynamic Effects Influencing IGBT Losses in
Soft-Switching Converters,” IEEE Trans. Power Electron., vol. 26, pp. 260–
271, Jan. 2011.
[72] J. Choi, D. Tsukiyama, Y. Tsuruda, and J. Rivas, “13.56 MHz 1.3 kW Reso-
nant Converter With GaN FET for Wireless Power Transfer,” in IEEE Wire-
less Power Transfer Conf. (WPTC), pp. 1–4, May 2015.
[73] H. Yu, J. Lai, J. H. Zhao, and B. H. Wright, “Gate Driver Based Soft Switch-
ing for SiC BJT Inverter,” in IEEE 34th Annual Power Electron. Specialist
Conf. (PESC), vol. 4, pp. 1857–1862, Jun. 2003.
98 BIBLIOGRAPHY
[76] L. Liao, S. Tang, J. Wang, Z. Shuai, X. Yin, and Z. J. Shen, “A New Propor-
tional Base Driver Technique for SiC Bipolar Juction Transistor,” in IEEE
Energy Conversion Congr. and Exp. (ECCE), pp. 942–946, Sept. 2015.
[80] R. Siemieniec and U. Kirchner, “The 1200V Direct-Driven SiC JFET Power
Switch,” in Proceedings of the 14th Eur. Conf. on Power Electron. and Appl.
(EPE), pp. 1–10, Aug. 2011.
[81] K. Norling, C. Lindholm, and D. Draxelmayr, “An Optimized Driver for SiC
JFET-Based Switches Delivering More than 99 % Efficiency,” in IEEE Int.
Solid-State Circuits Conf., pp. 284–286, Feb. 2012.
[82] D. Peftitsis, J. Rabkowski, and H.-P. Nee, “Self-Powered Gate Driver for Nor-
mally ON Silicon Carbide Junction Field-Effect Transistors Without External
Power Supply,” IEEE Trans. Power Electron., vol. 28, pp. 1488–1501, Mar.
2013.
and Safe Operation,” in 6th Int. Conf. on Integrated Power Electron. Systems
(CIPS), pp. 1–6, Mar. 2010.
[85] F. Guedon, S. K. Singh, R. A. McMahon, and F. Udrea, “Gate Driver for
SiC JFETs With Protection Against Normally-On Behaviour Induced Fault,”
Electron. Lett., vol. 47, pp. 375–377, Mar. 2011.
[86] R. Lai, F. Wang, R. Burgos, D. Boroyevich, D. Zhang, and P. Ning, “A
Shoot-Through Protection Scheme for Converters Built With SiC JFETs,”
IEEE Tran. Ind. Appl., vol. 46, pp. 2495–2500, Nov. 2010.
[87] J. H. Kim, B. D. Min, J. W. Baek, and D. W. Yoo, “Protection Circuit
of Normally-On SiC JFET Using an Inrush Current,” in 31st Int.Telecom.
Energy Conf. (INTELEC), pp. 1–4, Oct. 2009.
[88] D. P. Sadik, J. Colmenares, G. Tolstoy, D. Peftitsis, M. Bakowski,
J. Rabkowski, and H.-P. Nee, “Short-Circuit Protection Circuits for Silicon-
Carbide Power Transistors,” IEEE Trans. Ind.l Electron., vol. 63, pp. 1995–
2004, Apr. 2016.
[89] E. Velander, L. Kruse, S. Meier, A. Löfgren, T. Wiik, H.-P. Nee, and D.-
P. Sadik, “Analysis of Short Circuit Type II and III of High Voltage SiC
MOSFETs With Fast Current Source Gate Drive Principle,” in IEEE 8th Int.
Power Electron. and Motion Control Conf. (IPEMC-ECCE Asia), pp. 3392–
3397, May 2016.
[90] B. Wrzecionko, S. Käch, D. Bortis, J. Biela, and J. W. Kolar, “Novel AC
Coupled Gate Driver for Ultra Fast Switching of Normally-Off SiC JFETs,”
in 36th Annual Conf. on IEEE Ind. Electron. Soc. (IECON), pp. 605–612,
Nov. 2010.
[91] J. B. Casady, D. C. Sheridan, R. L. Kelley, V. Bondarenko, and A. Ritenour,
“A Comparison of 1200 V Normally-Off & Normally-On Vertical Trench SiC
Power JFET Devices,” in Mater. Sci. Forum, vol. 679, pp. 641–644, Trans
Tech Publ, 2011.
[92] D. P. Sadik, K. Kostov, J. Colmenares, F. Giezendanner, P. Ranstad, and
H.-P. Nee, “Analysis of Parasitic Elements of SiC Power Modules With Spe-
cial Emphasis on Reliability Issues,” IEEE Trans. Emerg. Sel. Topics Power
Electron., vol. 4, pp. 988–995, Sept. 2016.
[93] H.-P. Nee, J. Rabkowski, and D. Peftitsis, “Multi-Chip Circuit Designs for
Silicon Carbide Power Electronics,” in 8th Int. Conf. on Integrated Power
Systems (CIPS), pp. 1–10, Feb. 2014.
[94] J. Rabkowski, D. Peftitsis, and H.-P. Nee, “Parallel-Operation of Discrete SiC
BJTs in a 6–kW/250–kHz DC/DC Boost Converter,” IEEE Trans. Power
Electron., vol. 29, pp. 2482–2491, May 2014.
100 BIBLIOGRAPHY
[96] R. Fu, A. Grekov, J. Hudgins, A. Mantooth, and E. Santi, “Power SiC DMOS-
FET Model Accounting for Nonuniform Current Distribution in JFET Re-
gion,” IEEE Trans. Ind. Appl., vol. 48, pp. 181–190, Jan. 2012.
[104] A. Ward, “SiC Power Diode Reliability,” CREE Inc., Mar. 2006.
[105] S. Allen, “Silicon Carbide MOSFETs for High Powered Modules,” CREE Inc.,
Mar. 2013.
BIBLIOGRAPHY 101
[113] T. Ueda, “Reliability Issues in GaN and SiC Power Devices,” in IEEE Inter-
national Reliability Phys. Symp., pp. 3D.4.1–3D.4.6, Jun. 2014.
[123] D. Watt, “Cree SiC Diodes High Temperature Reliability Trials,” 2012.
[141] F. Xu, D. Jiang, J. Wang, F. Wang, L. Tolbert, T. Han, and S. Kim, “Char-
acterization of a high temperature multichip SiC JFET-based module,” in
IEEE Energy Conversion Congress and Exposition (ECCE), pp. 2405–2412,
Sept. 2011.
[143] W. Zhou, X. Zhong, and K. Sheng, “High Temperature Stability and the
Performance Degradation of SiC MOSFETs,” IEEE Trans. Power Electron.,
vol. 29, no. 5, pp. 2329–2337, 2014.
[149] L. F. Eastman and U. K. Mishra, “The Toughest Transistor Yet [GaN tran-
sistors],” IEEE spectrum, vol. 39, no. 5, pp. 28–33, 2002.
[150] Y. Lei, C. Barth, S. Qin, W.-C. Liu, I. Moon, A. Stillwell, D. Chou, T. Foulkes,
Z. Ye, Z. Liao, et al., “A 2 kW, Single-Phase, 7-Level, GaN Inverter With
an Active Energy Buffer Achieving 216 W/in3 Power Density and 97.6 %
Peak Efficiency,” in 2016 IEEE Applied Power Electronics Conference and
Exposition (APEC), pp. 1512–1519, 2016.