Journal of Power Electronics, Vol. 15, No. 6, pp.

1489-1498, November 2015 1489

http://dx.doi.org/10.6113/JPE.2015.15.6.1489
JPE 15-6-8 ISSN(Print): 1598-2092 / ISSN(Online): 2093-4718

Performance Evaluation of GaN-Based Synchronous

Boost Converter under Various Output Voltage,
Load Current, and Switching Frequency Operations
Di Han* and Bulent Sarlioglu†
*,†
Department of Electrical and Computer Engineering, University of Wisconsin-Madison, Madison, U.S.A

Abstract
Gallium nitride (GaN)-based power switching devices, such as high-electron-mobility transistors (HEMT), provide significant
performance improvements in terms of faster switching speed, zero reverse recovery, and lower on-state resistance compared
with conventional silicon (Si) metal–oxide–semiconductor field-effect transistors (MOSFET). These benefits of GaN HEMTs
further lead to low loss, high switching frequency, and high power density converters. Through simulation and experimentation,
this research thoroughly contributes to the understanding of performance characterization including the efficiency, loss
distribution, and thermal behavior of a 160-W GaN-based synchronous boost converter under various output voltage, load
current, and switching frequency operations, as compared with the state-of-the-art Si technology. Original suggestions on design
considerations to optimize the GaN converter performance are also provided.

Key words: Gallium nitride, Semiconductor loss, Synchronous boost converter

contribute to the optimized usage of GaN devices by

I. INTRODUCTION
providing original suggestions on the GaN converter design.
With the continuing trend toward higher power rating, SiC devices have been extensively studied in the past
higher efficiency, lighter weight, and smaller size power decade because of their relatively high level of maturity
electronic converters in various applications [1]-[6], the need [15]-[17]. In contrast, GaN devices have just started to
for better performance power switching devices has been become available to public and, hence, are less investigated
warranted. In recent years, wide bandgap (WBG) in literature. In [18], a GaN high-electron-mobility transistors
semiconductor materials, such as silicon carbide (SiC) and (HEMT) boost converter is shown to achieve 98% efficiency
gallium nitride (GaN) based devices have become available at 300-W output power at a switching frequency of 1 MHz. A
[7]-[14]. SiC- and GaN-based switching devices have low 10-W synchronous buck converter using GaN-on-SiC
semiconductor loss, high switching speed, and high HEMTs is demonstrated in [19], achieving 95% efficiency at
temperature capability because of the superior physical 10-MHz switching frequency and 90% efficiency at 40-MHz
properties of WBG materials, such as high electric switching frequency. Reference [20] presents a 3-kW 400–
breakdown field, low intrinsic carrier concentration, and large 800-V GaN-based boost converter that shows 99% efficiency
saturated electron drift velocity. This research aims to at 100-kHz switching frequency. In [21], a GaN device is
thoroughly investigate performance characteristics including used in a hybrid resonant converter with 97.5% California
the efficiency, loss distribution, and thermal behavior of the Energy Commission weighted efficiency. Reference [22]
latest GaN device in a synchronous boost converter and to characterizes the efficiency improvements by using
GaN-based source-switched field-effect transistors (SSFET)
Manuscript received Mar. 1, 2015; accepted Jun. 25, 2015
Recommended for publication by Associate Editor Joung-Hu Park.
in a PFC boost converter. However, the present research
†
Corresponding Author: sarlioglu@wisc.edu limits its focus on efficiency comparison for a few operating
Tel: +1-608-262-2703, Fax: +1-608-263-3160, University of conditions at a fixed switching frequency. At present, studies
Wisconsin-Madison
*
Department of Electrical and Computer Engineering, University of
have not evaluated the performance of GaN HEMT over a
Wisconsin-Madison, U.S.A wide range of operating voltage, current, switching
© 2015 KIPE
1490 Journal of Power Electronics, Vol. 15, No. 6, November 2015

TABLE I MOSFET.
COMPARISON OF GAN HEMT AND SI MOSFET UNDER STUDY In this study, the GaN device chosen is a 100-V 25-A
Part number EPC2001 RJK1053DPB HEMT (EPC2001) from EPC. A corresponding
Material GaN Si state-of-the-art Si device with the same ratings is the
Voltage rating VDS 100 V 100 V MOSFET (RJK1053DPB) from Renesas Electronics.
Current rating ID (Continuous) 25 A 25 A Although GaN HEMT is more expensive than Si MOSFET at
On-state resistance RDSon (Max) 7 mΩ 15 mΩ present, the price of GaN devices is believed to gradually
Total gate charge Qg 8 nC 43 nC decrease as manufacturing techniques improve. Hence, the
Gate to source charge Qgs 2.3 nC 19 nC device costs are not taken into account in this study.
Gate to drain charge Qgd 2.2 nC 12.5 nC A few important characteristic parameters of the two
Body diode reverse recovery No Yes devices are compared in Table I below. Key observations
Source-drain forward voltage VSD 2.3 V 0.83 V from the table are summarized as follows.
(ID=−25A, T=25°C)
Output capacitance Coss (VDS=50V, 450 pF 210 pF 1) The GaN HEMT and Si MOSFET chosen for the study
Vgs=0V) have the same voltage rating of 100 V and current rating
Thermal resistance, junction to case 2.1°C/W 1.92 °C/W of 25 A, which guarantee a fair comparison between the
RθJC GaN and Si technology.
2) The GaN HEMT has a maximum on-state resistance RDSon
frequencies. of only 7 mΩ, which is less than half of the value for Si
The objective of this paper is to contribute to the analysis of MOSFET (15 mΩ) and thus indicates halved conduction
important and different characteristics of GaN HEMTs in loss during operation.
power converters, as compared to conventional Si MOSFETs, 3) The gate charge values (Qgs, Qgd, and Qg) of GaN HEMT
which will be achieved through quantifying the loss, efficiency, are five to ten times smaller than those of the Si MOSFET.
and temperature rise of a GaN synchronous boost converter As a result, with the same level of gate current, the gate
over a wide range of operating conditions. The research results capacitance of a GaN HEMT will be charged within 1/5
will provide pertinent and important design suggestions, as the time needed by Si MOSFET, which further translates
well as comprehensive and valuable firsthand data, which are into the fast turn-on and turn-off of the GaN conduction
critical in reducing the knowledge gap for practical engineers channel.
and decision makers. 4) While the body diode of Si MOSFET has reverse
The paper is organized as follows. In Section II, a GaN recovery effect because of the p–n junction nature of the
HEMT device is compared to the same rated Si MOSFET in parasitic diode in a MOSFET, GaN HEMT has a “major
terms of critical device parameters. A 160-W synchronous carrier body diode” that features zero reverse recovery.
boost converter is simulated in Section III using GaN HEMTs 5) The forward voltage drop of the GaN body diode (2.3 V)
and Si MOSFETs under various operating conditions. is roughly three times that of the Si counterpart (0.83 V)
Experimental results on the GaN-based converter are presented and indicates the high reverse conduction loss of GaN
in Section IV to validate the simulation. In addition, discussion HEMT, which is a direct result of the wide energy band
and suggestions on the effective usage of GaN HEMTs are gap of GaN material.
provided. Finally, conclusions are drawn in Section V. 6) At 50-V drain-to-source voltage, output capacitance Coss
of GaN HEMT (450 pF) is more than twice that of Si
MOSFET (210 pF), which is also true for other
II. GAN-BASED POWER DEVICES
drain-to-source voltage values. As a result, the energy
Theoretically, GaN is an excellent material for making stored in output capacitance of GaN HEMT during
power switching devices targeting high-voltage, off-state is significant.
high-frequency, and high-temperature applications because of 7) Thermal resistance RθJC of GaN device is slightly higher
its large critical electric field, high electron mobility, low than that of Si MOSFET, which is mainly because
carrier concentration, and good thermal conductivity [9]. thermal contact resistance formed between GaN epilayer
However, because of the lack of commercially available and Si substrate on which it is grown.
low-cost and high-quality GaN substrates, GaN epilayers are
To clearly understand the strengths and weaknesses of
mainly grown on Si substrates and become the technical
GaN HEMTs, analyses are given from the following three
bottleneck that limits the available device type, voltage rating,
aspects, i.e., conduction loss Pc, switching loss Psw, and
and thermal conductivity. At present, the GaN-based power
temperature rise ΔT.
devices in the market are either normally-off GaN HEMT
rated at 40–200 V or 600–650 V cascode switch based on A. Conduction Loss
series connection of normally-on GaN HEMT and Si
 Performance Evaluation of … 1491

The conduction loss of a field-effect transistor (FET)

during on-state can be expressed as
Tsw
1 2
PCF = T  pCF(t)dt = RDSon·IDrms , (1)
sw 
t=0
where subscript “CF” represents “conduction of FET,” Tsw
represents the switching period, and IDrms stands for the rms
value of the drain current.
As shown in Equ. (1), with half RDSon, the conduction
loss of GaN device will be half of that of the Si device when
Fig. 1. Schematic of the boost converter under study.
operated at the same drain current.
The conduction loss of the body diode can be expressed as TABLE II
Tsw
1 SPECIFICATIONS OF THE BOOST CONVERTER UNDER STUDY
PCD = T  pCD(t)dt ≈ VSD·IFav, (2)
sw  Specifications Value
t=0
Low side voltage Vin (V) 24
where subscript “CD” represents “conduction of diode,” and
High side voltage Vout (V) 48–80
IFav stands for the average value of the diode current.
Power rating P (W) 160
Because VSD of the GaN device is three times that of the Si
Switching frequency f (kHz) 100–300
device, as can be expected from Equ. (2), the diode
Chock inductor L (μH) 20
conduction loss of GaN device should also be three times
higher given the same length of diode conduction period. DC bus capacitor C (μF) 60
Furthermore, the length of diode conduction period is Input capacitor Cin (μF) 10
dependent on dead-time values utilized during the converter
operation. Hence, the selection of dead-time has a significant zero Qrr, GaN device generate zero Prr, whereas the reverse
effect on diode conduction loss. Fortunately, the diode recovery loss of Si MOSFET can be prominent. However,
conduction in a synchronous boost converter can be with two times larger Coss, Poss term of the GaN device is
eliminated by optimizing the dead-times [23]. expected to be larger than that of the Si device.
An assumption in this study is that the technique described The first two terms, i.e., Pvi and Prr, typically dominate the
in [23] has been adopted for both Si and GaN based converter third term in switching loss, and the total switching loss in
such that diode conduction is negligible. Hence, the total GaN converter is expected to be smaller than that of Si
conduction loss should be converter.
PC = PCF + PCD ≈ PCF. (3) C. Temperature Rise
B. Switching Loss The temperature rise from device junction to case can be
Switching loss Pswi that occurred during the switching simply estimated as
transitions can be divided into three major components as ΔTJC = (PC + Pswi) RθJC. (8)
Pswi = Pvi + Prr + Poss, (4) Although thermal resistance RθJC of the GaN device is slightly
higher than the Si device, the actual temperature rise can still
where Pvi represents the switching loss without considering
be smaller in the GaN devices because of the reduced losses.
reverse recovery and output capacitive charge, Prr represents
In the following sections, the overall effects of these
the loss induced by body diode recovery, and Poss represents
counteracting features of GaN HEMTs will be explored as
the loss induced by the energy stored in output capacitors Coss
applied to a synchronous boost converter.
of devices. These three components can be further estimated
as follows:
Pvi = Vo·IDon·(tri+tfu)/2·fsw, (5) III. SIMULATION OF SYNCHRONOUS BOOST
Prr = Qrr·VDSon ·fsw, (6) CONVERTER
Vo The synchronous boost converter used in this case study is
Poss = 
 Qoss(VDS)dV ·fsw, (7) shown in Fig. 1. Some specifications of the converter, which
t=0 is 160 W converter with a boost ratio of 24–80 V, are listed in
where Vo is the output voltage, tri and tfu are the current rise Table II. These ratings suit a typical photovoltaic application
time and voltage fall time, respectively, Qrr is the reverse as a modular DC–DC converter in the distributed MPPT
recovery charge, and fsw is the switching frequency. architecture.
As observed earlier, with five to ten times smaller gate
charge tri and tvf of GaN HEMT will be significantly smaller A. Simulation Methods
than that of Si MOSFET, leading to smaller Pvi. Similarly, with The purpose of this work is to thoroughly study the
1492 Journal of Power Electronics, Vol. 15, No. 6, November 2015

performance of the GaN-based boost converter described 60

Turn-on transient
20
above, under different operating conditions, and compare it

Drain-to-Source-Voltage (V)
with the corresponding Si-based converter.
To achieve this purpose, the proposed boost converter is

Drain Current (A)

40 10
simulated under four load currents (1, 2, 3, and 4 A), four
switching frequencies (100, 200, 300, and 400 kHz), and
three output voltages (48, 64, and 80 V) for both the GaN 20 0
HEMTs and Si MOSFETs as switching devices. Of the three
variables, (i.e., load current, switching frequency, and output
voltage,), only one is varied at a time, whereas the other two 0 -10
3.32 3.34 3.36 3.38 3.4
will be set at a default value. The default values for the three Time(s)
variables are 2 A, 300 kHz, and 48 V, respectively. Fig. 2. Drain-to-source voltage VDS and drain current ID
The converter simulation is performed using the circuit waveforms during M2 turn on, 48 V output voltage, 4A load, and
simulation tool LTSpice. The SPICE models of the two 300 kHz switching frequency.
switching devices are both from their manufacturer. Gate
Turn-off transient
driver circuitry is simplified using programmable 100 20
independent voltage sources in the simulation. A rise and fall

Drain-to-Source-Voltage (V)
time of 7 ns and 1.5 ns is used to be consistent with the
performance of a commercial gate driver chip LM5113.

Drain Current (A)

10
Although the device junction temperature is not known, it is
assumed to be 80 °C in the simulation as a best estimation. 50

In this boost converter, instead of merely implementing the 0

upper switch as a diode, a transistor is used complementarily
with the bottom switch, i.e., works in synchronous rectifying
mode. Hence, a short deadtime of only 5 ns is added between 0 -10
1.65 1.7 1.75
the complementary gate signals to prevent the shoot through Time(s)
problem, as well as minimize the diode conduction. Fig. 3. Drain-to-source voltage VDS and drain current ID
waveforms during M2 turn off, 48-V output voltage, 4-A load,
B. Switching Performances and 300-kHz switching frequency.
In this part, the switching performances of the GaN and Si
converters will be compared. The turn-on and turn-off the core material datasheet. However, the loss of gate driver
waveforms of lower switch M2 is shown in Figs. 2 and 3, is not considered because of its low value.
respectively. The converter loss and efficiency comparison under
As shown in Fig. 2, the voltage fall time tvf of GaN device different load currents are plotted in Figs. 4 and 5,
is only 2.2 ns, with a dv/dt of 22.7 kV/μs. The voltage fall respectively.
time of Si device is roughly 23 ns, with a dv/dt of 2.0 kV/μs. As shown in Fig. 4, the total losses on GaN and Si
Similarly, the current rise time of the GaN device is converters are very similar at light loads, but a significant
significantly smaller than that of the Si device. As a result, difference is shown at heavier loads. For example, at 1-A
the intersection of non-zero voltage and current of the GaN load current, the loss on the Si converter is higher than that of
device is considerably smaller than that of Si, indicating the GaN converter by only about 0.04 W, but at 4-A load
smaller switching loss of GaN during turn-on despite its current, this difference increases to about 1.3 W because as
larger Coss. the load increases, the semiconductor loss on the GaN HEMT
As shown in Fig. 3, the switching waveforms during increases by a small amount (from 0.51 to 0.83 W), whereas
turn-off are very close for the GaN- and Si-based converters. the loss on Si MOSFET almost quadrupled (from 0.53 to 2.0
Thus, the switching loss during turn-off is expected to be W). This observation indicates that the GaN HEMT has much
similar. smaller conduction loss than Si MOSFET, particularly at
large load conditions.
C. Converter Performance under Various Loads The passive component loss on both converters increases
In the simulation, the power consumed on each circuit dramatically with load current because of the current squared
component is obtained. Converter efficiency is calculated as relationship. In addition, the passive component loss always
the ration of the power absorbed on load resistance to the dominates the total loss for 2–4-A load conditions in the GaN
power supplied by the source. The losses on semiconductor converter. The main sources of passive components loss are
devices and passive components are recorded separately. The the winding loss and core loss on the choke inductor.
core loss of the inductor is also accounted for by referring to As shown in Fig. 5, the efficiency on the GaN converter is
 Performance Evaluation of … 1493

Power Loss Breakdown Comparison Power Loss Breakdown Comparison

Under Different Load Current Under Different Switching Frequency
5 2.5
Semiconductor Loss Semiconductor Loss
Passive Component Loss Passive Component Loss
4 2
Power Loss [W]

Power Loss [W]

3 1.5

2 1

1 0.5

0 0
Si GaN Si GaN Si GaN Si GaN Si GaN Si GaN Si GaN Si GaN
1A 1A 2A 2A 3A 3A 4A 4A 100 100 200 200 300 300 400 400
kHz kHz kHz kHz kHz kHz kHz kHz
Fig. 4. Converter loss breakdown for different load current, 48-V
output voltage, and 300-kHz switching frequency. Fig. 6. Converter loss breakdown for different switching
frequencies, 2-A load current, and 48-V output voltage.
Overall Efficiency Comparison
Under Different Load Current Overall Efficiency Comparison
100
Under Different Switching Frequency
Si 100
99.5 GaN Si
Overall Efficiency [%]

99.5 GaN
99
Overall Efficiency [%]

99
98.5
98.5
98
98
97.5
97.5
97
97
96.5
1 2 3 4
Load Current [A] 96.5
100 200 300 400
Fig. 5. Converter efficiency under different load current, 48-V Switching Frequency [kHz]
output voltage, and 300-kHz switching frequency. Fig. 7. Converter efficiency under different switching
frequencies, 2-A load current, and 48-V output voltage.
higher than its Si counterpart by 0.08% at 1-A load and 0.59%
at 4-A load. The increased efficiency difference is because of dominant loss component in the GaN converter at the
the increased loss difference explained earlier. switching frequency of 400 kHz.
The corresponding efficiencies range from 98.13% to
D. Converter Performance under Various Switching
98.63% for the Si converter and from 98.61% to 98.85% for
Frequencies
the GaN converter.
The converter loss and efficiency data under various
switching frequencies are shown in Figs. 6 and 7, E. Converter Performance under Various Output
respectively. Voltage
As shown in Fig. 6, as the switching frequency increases The converter loss and efficiency data under various
from 100 to 400 kHz, the semiconductor loss on the Si switching frequencies are shown in Figs. 8 and 9.
MOSFET increases from 0.47 to 1.15 W, and the loss on the As shown in Fig. 8, as the output voltage increases, both
GaN HEMT increases from 0.27 to 0.71 W. The loss semiconductor loss and passive component loss increase
difference between two devices also increases from 0.2 to dramatically for the two converters. The loss on the Si
0.44 W, which indicates that the GaN HEMT has better MOSFETs is approximately twice the loss on GaN HEMTs
switching performance than the Si MOSFET. for the three voltages.
The passive component loss decreases a little with the The overall efficiency drops from 98.36% to 97.26% for
increased switching frequency because of the reduced current the Si converter and from 98.70% to 98.03% for the GaN
ripple. As a result, the semiconductor loss becomes the converter as the output voltage increases from 48 to 80 V.
1494 Journal of Power Electronics, Vol. 15, No. 6, November 2015

Power Loss Breakdown Comparison

Under Different Output Voltage
5
Semiconductor Loss
Passive Component Loss
4
Power Loss [W]

0 Fig. 10. Prototype GaN device based synchronous boost

Si GaN Si GaN Si GaN converter built and tested for experimental results.
48V 48V 64V 64V 80V 80V
TABLE III
Fig. 8. Converter loss breakdown for different output voltage,
2-A load current, and 300-kHz switching frequency. EXPERIMENTAL TEST CASES
Output Switching
Case Input voltage Load current
Overall Efficiency Comparison voltage Vout Frequency f
No. Vin (V) (A)
Under Different Output Voltage (V) (kHz)
100 1 24 48 0.5 300
Si 2 24 48 1 300
99.5 GaN 3 24 64 0.5 300
4 24 80 0.5 300
Overall Efficiency [%]

99
5 24 48 0.5 200
98.5 6 24 48 0.5 400

98 Limited by the maximum current rating of the DC power

97.5 supply used in the experiment, the converter is tested under
light load conditions. During the test, each of the three
97
operating variables, i.e. load current, output voltage, and
96.5 switching frequency, are varied separately while keeping the
48 64 80
Output Voltage [V]
other two fixed values to evaluate the performance of the
Fig. 9. Converter efficiency under different output voltage, 2-A converter under various operating conditions, which is
load current, and 300-kHz switching frequency. consistent with the test cases evaluated in circuit simulation,
except that smaller load current values are used for the
If comparison is made between the three variables, one can experiments. A summary of the 6 test cases are listed in Table
conclude that, the output voltage has the most influence on III.
the converter loss and efficiency, whereas the switching As shown in Table III, cases 1 and 2 compare two load
frequency has the least effect. current values while keeping the voltage and switching
frequency the same. Cases 1, 3, and 4 show a variation in the
output voltage, i.e., from 48 to 80 V. Cases 1, 5, and 6
IV. EXPERIMENTAL TESTS ON THE GAN-BASED
SYNCHRONOUS BOOST CONVERTER compare three different switching frequencies, i.e., 200, 300,
and 400 kHz.
To validate the given simulation results, a 160-W
GaN-based synchronous boost converter is built according to A. Variation in Load Current
the specifications listed in Table II. To expedite the During the experiments, four quantities, i.e., input and
prototyping process, an EPC9002 demo board is utilized. The output voltages and currents are measured and recorded for
demo board comes with two EPC2001 in a phase leg each case. The corresponding power loss and efficiency
configuration, a gate driver LM5113 to drive the two HEMTs, values are obtained based on these measurements. The scope
and a dead-time generating circuit. Two film capacitors are used for the measurement is WaveSurfer 104MXs-B from
used as the input and output bus capacitors of the converter. LeCroy with 1-GHz bandwidth and 5 GS/s sample rate. Two
The choke inductor is purchased from Coilcraft. Fig. 10 20-MHz high-voltage differential probes (ADP300) are used
shows the converter prototype with critical parts labeled. to measure the input and output voltage. Two 100-MHz
Gating signals for the two HEMTs are generated using a TI (CP031) current probes are used to measure the input and
DSP, which is not photographed. output current. In addition, to monitor the heating of the GaN
 Performance Evaluation of … 1495

Fig. 11. (C1, yellow, 5 V/div) Measured waveforms for input Fig. 13. (C1, yellow, 5 V/div) Measured waveforms for input
voltage, (C2, red, 10 V/div) output voltage, (C3, blue, 500 voltage, (C2, red, 10 V/div) output voltage, (C3, blue, 500
mA/div) inductor current, (C4, green, 100 mA/div) load current, mA/div) inductor current, (C4, green, 200 mA/div) load current,
and data for test case 1. Time scale: 5 μs/div. and data for test case 2. Time scale: 5 μs/div.

Fig. 12. Thermal image of the device on board for test case 1. Fig. 14. Thermal image of the device on board for test case 2.

devices, a thermal camera is also used to capture the thermal about 0.5 W, which corresponds to 98.77%, as shown. The
image of the board as the converter is running. To ensure the device case temperature increases to 51.9 °C.
device temperature has reached a steady-state value, the Compared to case 1, case 2 has 0.1 W higher losses. This is
converter is left running for at least 5min before a thermal due to the higher load current, thus higher conduction loss.
picture is taken. Despite the higher loss, the converter in case 2 has around 0.6%
The measured waveforms and relative data for test case 1 higher efficiency.
are shown in Fig. 11. The thermal image of case 1 is shown
in Fig. 12. In Fig. 11, the important measurement data can be B. Variation in Output Voltage
directly read from the scope screen. The input voltage from To save time and space, the waveforms and thermal
the DC power supply is very close to 24 V. The output measurements for the remaining test cases are not shown.
voltage is roughly 48 V. The load current is 0.465 A, which is For test case 3, the input voltage is about 23.9 V, and the
slightly less than 0.5 A. As shown, the overall converter output voltage is 64.1 V. The load current is 0.523 A. The
efficiency is 98.18%, and the loss is about 0.4 W. Note that total converter loss is roughly 1.06 W, and efficiency is
the loss measured here is only the loss on the main power 96.95%. The device case temperature has also increased to
circuit, and the gate driver circuit loss is not included. 66.8 °C because of the voltage rise. For test case 4, the input
In the thermal picture in Fig. 12, the square board in the voltage is about 23.8 V, and the output voltage is around 79.6
middle is the demo board EPC 9002. The hot spot indicated V. The load current is 0.513 A. The total converter loss is
by the red triangle is exactly where the two GaN devices are roughly 1.65 W, and the corresponding efficiency is 96.12%.
located. The maximum temperature value of 48.7 °C shown The case temperature of the GaN device reaches 71.4 °C in
in the picture corresponds to the hot spot temperature. Hence, this condition.
the device case temperature has reached 48.7 °C in this case. According to the comparison of cases 1, 3, and 4, the
The measured waveforms and thermal image for test case 2 converter loss increases dramatically from 0.4 to 1.65 W as
are shown in Figs. 13 and 14, respectively. As shown in Fig. the output voltage increases from 48 to 80 V because of the
13, the input and output voltage are 21.6 and 42.9 V, fact that both the current stress and voltage stress on the
respectively, which are slightly lower than the ideal values. switching device increase as the output voltage increases.
The load current is also slightly less than the ideal value; the Thus, both conduction and switching losses on the device
load current is 0.96 A instead of 1 A because of the limitation increase accordingly. The converter efficiency also decreases
of the DC power supply. The converter loss in test case 2 is from 98.18% to 96.12% as the output voltage increases.
1496 Journal of Power Electronics, Vol. 15, No. 6, November 2015

current direction change actually helps the converter achieve

zero voltage turn-on for both switches using a technique
called zero voltage resonant-transition switching (ZVRT)
[24]. This phenomenon further reduces the switching loss on
top of the lower switching frequency and explains the
ultra-low loss in test case 5.
D. Comparison between Simulation and Test Results
All test results shown in this section have been
summarized in Table IV. The observations can be concluded
as follows:
Fig. 15. (C1, yellow, 5 V/div) Measured waveforms for input 1. Converter loss increases with increased load current,
voltage, (C3, blue, 10 V/div) output voltage, (C3, red, 500 output voltage, and switching frequency.
mA/div) inductor current, (C4, green, 100 mA/div) load current, 2. Converter efficiency decreases with increasing
and data for test case 5. Time scale: 2 μs/div. output voltage and switching frequency but
increases with increasing load current within the
TABLE IV tested operating range.
EXPERIMENTAL TEST RESULTS FOR GAN CONVERTER 3. Among load current, output voltage, and switching
Switching Input Output Load Loss Efficie Device frequency, the output voltage variation has the
Case
Frequency voltage voltage current Ploss ncy η Temperatu
No. largest influence on converter efficiency.
f (kHz) Vin (V) Vout (V) (A) (W) (%) re Tc (°C)
These observed trends correlate very well with the
1 300 24 48 0.465 0.4 98.18 48.7
simulation results presented in Section III. Even though the
2 300 21.6 42.9 0.963 0.5 98.81 51.0
measured loss and efficiency values are slightly different
3 300 23.9 64.1 0.523 1.06 96.95 66.8
from the values obtained from simulation, the differences are
4 300 23.8 79.6 0.513 1.65 96.12 71.4
reasonably small. For example, for test case 2, the simulated
5 200 24 48.1 0.504 0.14 99.41 37.4
6 400 24 48 0.504 0.57 97.70 55.4
loss is 0.657 W, and the simulated efficiency is 98.64%,
which is only 0.17% less than the measured value. These
errors can be caused by several factors, such as inaccuracy of
C. Variation in Switching Frequency
the device model and datasheet, unknown device junction
For test case 5, the input voltage is 24 V, and output temperatures, and errors caused by the measurement tools.
voltage is 48.1 V. The load current is 0.504 A. In addition,
the switching frequency has been reduced to 200 kHz instead
of 300 kHz, as in the previous cases. The total converter loss
V. DISCUSSION AND SUGGESTIONS
is about 0.14 W, and the efficiency reaches 99.41%. The case
temperature of the GaN devices is 37.4 °C under this A. Discussion on GaN Converter Performance
condition. For test case 6, the measured input voltage for this According to the results presented in the last two sections,
case is 24 V, and output voltage is about 48.1 V. The load the GaN HEMT performs better than the Si MOSFET in the
current is 0.504 A. The switching frequency has been converter in all operating conditions. A GaN converter has
increased to 400 kHz. The total converter loss is around 1.65 0.08%–0.77% higher efficiency than a Si converter,
W, and the efficiency is 97.70 %. The corresponding GaN depending on the operating condition, which corresponds to a
device case temperature for this case is 55.4 °C. maximum loss reduction of 1.23 W on the switching device.
According to the comparison of cases 1, 5, and 6, as the In addition, the maximum case temperature measured on the
switching frequency increases from 200 to 400 kHz, the GaN HEMTs under all test cases is only 71.4 °C, which
converter loss increases from 0.14 to 0.57 W, which is a four provides abundant margin from the maximum allowable
times increase. The overall efficiency decreases from 99.41% operating temperature of 125 °C. Note that the temperatures
to 97.7% because the switching loss on the GaN devices are were measured when the converter was running at room
directly related to switching frequency, and higher switching temperature without any cooling methods applied. Hence,
frequency yields larger loss. low semiconductor loss on the GaN device compensates for
Another important observation is that, with lower its relatively large thermal resistance.
switching frequency, a larger current ripple on the inductor is Carefully examining the results further reveals two
present. In the case of 200-kHz switching frequency, the important observations. First, the major saving in a converter
current ripple on the inductor is about 3 A (see Fig. 15), by using a GaN device comes from the reduction in
which, because of its amplitude, causes the inductor current conduction loss, particularly under large load conditions.
to reverse its direction twice every switching cycle. This Although the switching speed of GaN device is considerably
 Performance Evaluation of … 1497

faster than the Si device, the switching loss of the GaN between the performances of two converters in terms of
converter is not significantly reduced because of the efficiency and loss breakdown. Subsequently, experimental
relatively large energy stored in the device output capacitor, tests are performed on a GaN-based converter prototype to
as discussed earlier in Section II. This stored energy is lost in verify the simulation accuracy, as well as to obtain more
the conduction channel every time the device turns on and is information on the thermal aspect.
irrelevant to the switching speed of the device. As the voltage Lastly, discussions on GaN converter performances are
stress on the GaN HEMT during the off-state increases, more given. The GaN converter shows 0.08%–0.77% higher
energy will be stored in the output capacitor of the device, efficiency than the Si converter over the whole operating
resulting higher switching loss. range, featuring a maximum loss reduction of 1.23 W and a
Second, the passive component loss, which is mainly the maximum temperature rise of approximately 51 °C. The
loss on the inductor windings, presents itself as the major loss issues of output capacitor associated switching loss and
component in the GaN converter under most operating inductor winding loss are also highlighted. Suggestions on
conditions, particularly under high current load. As a result, solving the two issues of GaN converter are provided.
further reducing the semiconductor loss on devices will be
less effective without minimizing the inductor loss, if the goal
is to optimize the overall efficiency of the converter. REFERENCES
B. Suggestion on GaN Converter Design [1] D. Ryu, B. Choi, S. Lee, Y. Kim, and C. Won, “Flyback
inverter using voltage sensorless MPPT for photovoltaic
To fully reap the benefits of GaN-based switching devices AC modules,” Journal of Power Electronics, Vol. 14, No.
and to achieve ultra-high efficiency under a wide range of 6, pp. 1293-1302, Nov. 2014.
operating points for a GaN-based converter, the two issues [2] C. Gu, H. Krishnamoorthy, P. Enjeti, Z. Zheng, and Y. Li,
“A medium-voltage matrix converter topology for wind
mentioned above, i.e., switching loss associated with output
power conversion with medium frequency transformers,”
capacitor and inductor winding loss, should be addressed Journal of Power Electron., Vol. 14, No. 6, pp.
properly. 1166-1177, Nov. 2014.
To reduce the switching loss associated with output [3] A. Lopez, D. Patino, and R. Diez, “Efficiency analysis of a
capacitor, one possible solution is to apply zero voltage ladder multilevel converter with the use of the equivalent
continuous model,” Journal of Power Electron., Vol. 14, No.
switching (ZVS) technique. A preferable method is to allow 6, pp. 1130-1138, Nov. 2014.
the converter to operate in a synchronous conduction mode. [4] H. Yoo, S. Sul, Y. Park, and J. Jeong, “System integration
as shown in experimental test case 5, achieving ZVRT, and and power-flow management for a series hybrid electric
thus eliminating the energy loss on output capacitors by vehicle using supercapacitors and batteries,” IEEE Trans.
Power Electron., Vol. 44, No. 1, pp. 108-114, Jan./Feb.
transferring them between the upper and lower switches
2008.
during each switching transition. Other ZVS techniques are [5] J. Shin, H. Shin, G. Seo, J. Ha, and B. Cho, “Low-common
also open for consideration. mode voltage H-bridge converter with additional switch
To reduce the inductor winding loss, apart from optimizing legs,” IEEE Trans. Power Electron., Vol. 28, No. 4, pp.
the inductor design, one possible solution is to further 1773,1782, Apr. 2013.
[6] J. Lee, “Design and control methods of bidirectional DC-DC
increase the switching frequency of the GaN device, with the
converter for the optimal DC-link voltage of PMSM drive,”
assumption that the switching loss mentioned above has been J. Electr. Eng. Tech., Vol. 9, No. 6, pp. 1944-1953, Nov.
reduced to a reasonably small value. With increased 2014.
switching frequency, the required choke inductance value [7] U. K. Mishra, L. Shen, T. E. Kazior, and Y.-F. Wu,
will decrease, resulting in not only a smaller inductor size, “GaN-based RF power devices and amplifiers,” Proc. IEEE,
Vol. 96, No. 2, pp. 287-305, Feb. 2008.
but also less winding turns and smaller winding resistance. [8] N. Kaminski, “State of the art and the future of wide
band-gap devices,” in Proc. IEEE Power Electron. Appl.,
pp. 1-9, Sep. 2009.
VI. CONCLUSION [9] J. Millan, P. Godignon, X. Perpina, A. Perez-Tomas, and J.
This paper contributes to the evaluation of a 160-W Rebollo, “A survey of wide bandgap power semiconductor
devices,” IEEE Trans. Power Electron., Vol. 29, No. 5, p.
GaN-based synchronous boost converter and compares with
2155-2163, May 2014.
the state-of-the-art Si technology. [10] D. Han, J. Noppakunkajorn, and B. Sarlioglu,
First, the comparison of GaN and Si is made at the device “Comprehensive efficiency, weight, and volume
level by analyzing two comparable devices. Strengths and comparison of SiC and Si-based bidirectional DC-DC
weaknesses of GaN HEMTs are pointed out. Then, a converters for hybrid electric vehicles,” IEEE Trans. Veh.
Technol., Vol. 63, No. 7, pp. 3001-3010, Sep. 2014.
simulation is performed for the GaN-based converter, as well [11] M. Hakim, L. Tan, A. Abuelgasim, C. de-Groot, W.
as the Si-based converter, for a wide range of operating Redman-White, S. Hall, and P. Ashburn, “Drive current
conditions, including various load currents, switching improvement in vertical MOSFETS using hydrogen anneal,”
frequencies, and output voltages. A comparison is made in Proc. ICECE, pp. 217-220, Dec. 2012.
1498 Journal of Power Electronics, Vol. 15, No. 6, November 2015

[12] H. Liu, W. Hsu, C. Lee, B. Chou, Y. Liao, and M. Chiang, Di Han received his B. S. degree in Electrical
“Investigation of temperature-dependent characteristics of Engineering from Huazhong University of
AlGaN/GaN MOS-HEMT by using hydrogen peroxide Science and Technology, Wuhan, China, in
oxidation technique,” IEEE Trans. Electron Devices, Vol. 2011. He is currently working toward the Ph.D.
61, No. 8, pp. 2760-2766, Aug. 2014. degree in Electrical Engineering at the
[13] Y. Shin, “Effects of SPS mold on the properties of sintered University of Wisconsin-Madison, Madison,
and simulated SiC-ZrB2 composites,” J. Electr. Eng. Tech., WI. He is a Research Assistant at Wisconsin
Vol. 6, No. 4, pp. 1474-1480, Nov. 2013. Electric Machines and Power Electronics
[14] I. Kang, “Electrical characteristics of enhancement-mode Consortium (WEMPEC). His research interests include
n-channel vertical GaN MOSFETs and the effects of wide-bandgap device-based power converter design and
sidewall slope,” J. Electr. Eng. Tech., Vol. 10, No. 3, pp. implementation.
1131-1137, May 2015.
[15] A. Elasser and T. P. Chow, “Silicon carbide benefits and
advantages for power electronics circuits and systems,” Bulent Sarlioglu received his B.S. degree
Proc. IEEE, Vol. 90, No. 6, pp. 969- 986, Jun. 2002. from Istanbul Technical University, Istanbul,
[16] B. Ozpineci and L. M. Tolbert, “Characterization of SiC Turkey, in 1990, his M.S. degree from the
Schottky diodes at different temperatures,” IEEE Power University of Missouri, Columbia, MO, in
Electron. Lett., Vol. 1, No. 2, pp. 54-57, Jun. 2003. 1992, and his Ph.D. degree from the
[17] J. Biela, M. Schweizer, S. Waffler, and J. W. Kolar, “SiC vs. University of Wisconsin-Madison, Madison,
Si - evaluation of potentials for performance improvement WI, in 1999, all in electrical engineering.
of inverter and DC-DC converter systems by SiC power Since 2011, he has been an Assistant
semiconductors,” IEEE Trans. Ind. Electron., Vol. 58, No. Professor at the University of Wisconsin–Madison and the
7, pp. 2872-2882, Jul. 2011. Associate Director of the Wisconsin Electric Machines and
[18] Y. Wu, M. Jacob-Mitos, M. L. Moore, and S. Heikman, “A Power Electronics Consortium (WEMPEC). From 2000 to 2011,
97.8% efficient GaN HEMT boost converter with 300-W he worked at the aerospace division of Honeywell International
output power at 1 MHz,” IEEE Electron Device Lett., Vol. Inc., Torrance, California, as a Staff Systems Engineer. He
29, No. 8, pp. 824-826, Aug. 2008. received the Outstanding Engineer Award from Honeywell in
[19] M. Rodriguez, Y. Zhang, and D. Maksimovic, 2011. His expertise includes electrical machines, drives, and
“High-frequency PWM buck converters using GaN-on-SiC power electronics and he is the inventor or co-inventor of 16 US
HEMTs,” IEEE Trans. Power Electron., Vol. 29, No. 5, pp. patents, as well as many international patents.
2462-2473, May 2014.
[20] Y. Wu, J. Gritters, L. Shen, R. P. Smith, and B. Swenson,
“kV-class GaN-on-Si HEMTs enabling 99% efficiency
converter at 800 V and 100 kHz,” IEEE Trans. Power
Electron., Vol. 29, No. 6, pp. 2634-2637, Jun. 2014.
[21] T. LaBella and J.-S. J. Lai, “A hybrid resonant converter
utilizing a bidirectional GaN AC switch for high-efficiency
PV applications,” IEEE Trans. Ind. Appl., Vol. 50, No. 5,
pp. 3468-3475, Sep./Oct. 2014.
[22] J. Yang, “Efficiency improvement with GaN-based SSFET
as synchronous rectifier in PFC boost converter,” in Proc.
PCIM Europe, pp. 1-6, May 2014.
[23] D. Han and B. Sarlioglu, “Dead-time effect on GaN-based
synchronous boost converter and analytical model for
optimal dead-time selection,” IEEE Trans. Power Electron.,
to be published.
[24] C. P. Henze, H. C. Martin, and D. W. Parsley,
“Zero-voltage switching in high frequency power
converters using pulse width modulation,” in Proc. Appl.
Power Electron. Conf., pp. 33-40, 1988.