328 CPSS TRANSACTIONS ON POWER ELECTRONICS AND APPLICATIONS, VOL. 4, NO.

4, DECEMBER 2019

Improved Space Vector Modulation for Neutral-Point

Balancing Control in Hybrid-Switch-Based T-Type
Neutral-Point-Clamped Inverters With Loss and
Common-Mode Voltage Reduction
Hongwu Peng, Zhao Yuan, Xingchen Zhao, Balaji Narayanasamy, Amol Deshpande,
Asif Imran Emon, Fang Luo, and Cai Chen

Abstract—This paper compares different space vector modulation

DHyS
(SVM) strategies for neutral-point voltage balancing (NPVB) HyS
control in three-level (3-L) T-type neutral-point-clamped (TNPC)
inverters, and proposes an improved SVM trimmed for NPVB
GIGBT
control in hybrid-switch-based 3-L TNPC inverter with the
GSiC
features of loss and common-mode voltage (CMV) reduction.
The proposed SVM strategy uses a new principle of small vector
selection and vector sequence, and thus, it can balance the neutral SHyS
point (NP) potential and achieve soft-switching of clamping
leg simultaneously. The paper includes detailed analysis for High-current rated Si IGBT
Low-current rated SiC MOSFET
circuit commutation mode, loss breakdown, and common-mode
voltage patterns under different operation conditions. The circuit
simulations and experiments are carried out in the last part of this Fig. 1. Structure of the hybrid switch.
paper to validate the proposed SVM strategy.
replacement for Si IGBT in power electronics systems.
Index Terms—Electromagnetic interference (EMI), neutral-
However, the state-of-the-art die size limit and the high
point (NP) voltage balance, Si-SiC hybrid switch, space vector
modulation (SVM), T-type neutral-point-clamped (TNPC) inverter, cost of SiC MOSFET are the bottlenecks for its high-current
wide bandgap (WBG) devices. commercial applications. Therefore, [13] proposed a hybrid
switch concept, as shown in Fig. 1. By adjusting the switching
sequence of the two switches, the T-type inverter with the
I. Introduction
hybrid switch (hybrid structure 1) [14]-[16], which is shown

T HREE-LEVEL T-type neutral-point-clamped (3-L TNPC)

inverter has higher efficiency and lower total harmonic
distortion (THD) compared to two-level inverter, and it has
in Fig. 2, can have low switching loss from SiC MOSFET
switching, and low conduction loss from IGBT conduction.
However, due to the long discharging time of the carriers
become popular in high-speed motor drives and all-electric aircraft inside the IGBT, the Si-SiC hybrid switch operation mode
applications [1]–[5]. Emerging silicon carbide (SiC) MOSFET has have a minimum duty cycle limitation [17], which deteriorates
lower losses and high switching speed compared to Si IGBT and the output total harmonic distortion performance. Furthermore,
enables higher efficiency and power density in power converters
the hybrid-switch-based 3-L TNPC increases the system
[6]–[11]. SiC MOSFET is becoming a major competitor and
complexity in terms of the gate driver and power loop design
[18]-[20]. Because of more paralleled semiconductors,
Manuscript received Augest 4, 2019; revised September 11, 2019; accepted more gate drivers and more gating signals are required. [12]
October 15, 2019. Date of publication December 30, 2019; date of current
version November 21, 2019. This paper was presented in part at the 2019 IEEE
mentioned that the Si-SiC hybrid switch structure should
10th International Symposium on Power Electronics for Distributed Generation ensure the safe operation area (SOA) of the SiC MOSFET.
Systems (PEDG), Xi'an, China, June 2019. Table I illustrates the current dependent operation in [12].
H. Peng, Z. Yuan, X. Zhao, B. Narayanasamy, A. Deshpande, A. I. Emon,
and F. Luo are with the Department of Electrical Engineering, University of
To reduce the system complexity as well as improve the
Arkansas, Fayetteville, USA (email: fangluo@uark.edu). output THD performance, M1, M4, Q2, Q3, D2 and D3 in hybrid
C. Chen is with the School of Electrical and Electronic Engineering, structure 1 are selected to operate, and the new structure is
Huazhong University of Science and Technology, Wuhan, Hubei, China (email:
caichen@hust.edu.cn). shown in Fig. 3 [21]. The hybrid structure 2 utilizes Si IGBT
Digital Object Identifier 10.24295/CPSSTPEA.2019.00031 and SiC Schottky diode as clamping leg switches and SiC
 H. PENG et al.: IMPROVED SPACE VECTOR MODULATION FOR NEUTRAL-POINT BALANCING CONTROL IN INVERTERS 329

the clamping leg, neutral point balancing

SiC MOSFET: (NPB) [22]–[25] and
Si IGBT:
the common-mode voltage (CMV) performance. This paper
SiC MOSFET: Si IGBT:
provides aSiC Schottkyto
criterion Diode:
choose the appropriate SVM for T-type
P P
inverter with hybrid structure 2 under different conditions.
D1Q
M1 Q1
C1 M3
Firstly, this paper obtains the switching energy of T-type
C
M2 1
T1
inverter with hybrid structure 2 by experimental double pulse
VDC/2 D1M VDC/2 D2 D3
D2 D3 L test (DPT) result, then 3 different
A SVM
L schemes are discussed
O O
in terms of switching
T2 T3 loss, NPB, and output CMV voltage
C2 Q2 Q3 M4 D4Q C2
Q4
R performance.
T4 R
VDC/2 VDC/2The organization of the paper is as follows. Section II
D4M
N
summarizes the soft-switching conditions for 3-L TNPC with
N
hybrid structure 2, then loss analysis, neutral point balancing
capability and common-mode noise voltage comparison of 3
Fig. 2. T-type inverter with hybrid switch structure 1 [13].
different SVM schemes are discussed. In Section III, a 20 kVA
hardware is built, the DPT is firstly performed to obtain exact
TABLE I
Current Dependent Operation of the Hybrid-Switch-Based Inverter [12] switching loss, and followed by the analysis and comparison
of semiconductor loss breakdown, EMI spectrum and neutral
point voltage ripple for three SVMs. Section IV presents the
conclusion of the preferred modulation scheme.

II. Analysis of Different Modulation on Hybrid

Switch Based 3-Phase 3-L TNPC
For switching loss reduction [21] and NPB [22]–[25] in 3
phase 3-L TNPC inverter with hybrid switch combination,
SVM 1 with NPB [23], SVM 2 with NPB [24], and improved
SVM 2 with NPB are compared, and 1st sector of the space
SiC MOSFET: Si IGBT: vector modulation hexagon is given as an example. The nearest
Si IGBT: SiC Schottky Diode:
three space vector (NTSV) [26] and discontinuous pulse width
P modulation (DPWM) [27] are adopted to track the reference
Q1 vector and further reduce the switching loss.
C1 In this chapter, switching loss and soft-switching condition
T1
L VDC/2 D2 D3 of clamping leg is discussed first in 3-L TNPC with the hybrid
A L
O switch combination. Then three different SVM schemes are
T2 T3 compared in terms of switching losses, neutral point balancing
Q4 C2
R T4 R capability, and CMV performance, which gives guidance for
VDC/2
hardware design and PWM modulation choice.
N
A. Preferred Switch Pairs in Terms of Switching Loss Reduction
Fig. 3. T-type inverter with hybrid switch structure 2 [21]. Since switching loss of Si IGBT is much higher than SiC
MOSFET, hard switching of Si IGBT on clamping leg in 3-L
MOSFET for half-bridge switch positions. With SPWM, the TNPC inverter with hybrid switch combination should be
clamping switches are soft-switching under the unity power avoided or minimized. As shown in Fig. 4(a), when the phase
factor, and thus, the utilization of Si IGBTs does not increase current is positive, and the phase output voltage is transitioning
the switching loss. Therefore, the total semiconductor cost of from positive to neutral, T1 is hard switching off, and then T2
this hybrid switch combination 3-L TNPC is lower than that and D3 are soft switched on. From Fig. 4(b) we can know that
of the all-SiC 3-L TNPC inverter. While the cost is low, the when phase current is positive, and the phase output voltage
efficiency is higher than that of the all-SiC 3-L TNPC inverter. is transiting from neutral to positive, clamping leg devices are
Although hybrid structure 2 possesses the merits of low soft switched off, and T1 is hard switched on. Moreover, when
complexity and high efficiency, it has not been fully validated the phase current is negative, and phase output is transitioning
in the 3-phase 3-L TNPC system yet. One of the main from positive to neutral or from neutral to positive, T3 is hard
challenges is to design a proper space vector modulation (SVM) switched on and hard switched off.
scheme, which needs to consider the soft switching feature of In summary, when phase current is positive, switch pair 1
330 CPSS TRANSACTIONS ON POWER ELECTRONICS AND APPLICATIONS, VOL. 4, NO. 4, DECEMBER 2019

Positive state Deadtime freewheeling Negative state

P P P
C1 ET1,OFF = EMos,OFF C C1
1
T1,ON ET2,ON = 0 V /2 T1,OFF T1,OFF
VDC/2 D2 D3 D D2 VDC/2 D 2 D3 L
A L DC 3
A L A
O From O From O
C2 T2,ON T3,OFF P to O C2 P to O C2
T2,ON T3,OFF T2,ONT3,ON R
T4,OFF R T4,OFF R T4,OFF
VDC/2 VDC/2 VDC/2
N N N

(a)

P
Negative state P
Deadtime freewheeling P Positive state
C1 C1 ET1,ON = EMos,ON C1
T1,OFF T1,OFF T1,ON
VDC/2 D2 D3 L VDC/2 D2 D3 L
ET2,ON = 0 VDC/2 D2 D3 L
O A O A O A
From From
C2 T2,ON T3,ON C2 T2,ON T3,OFF C2 T2,ON T3,OFF
T4,OFF R O to P T4,OFF R O to P T4,OFF R
VDC/2 VDC/2 VDC/2
N N N

(b)

Positive state Deadtime freewheeling Negative state

P P P
C1 ET1,OFF = 0 C1
C1
T1,ON T1,OFF ET3,ON = EIGBT,ON T1,OFF
VDC/2 D2 D3 L VDC/2 D2 D3 A L VDC/2 D2 D3 L
O A From From O A
O
C2 T2,ON T3,OFF P to O C2 P to O C2
R
T2,ON T3,OFF T2,ON T3,ON R
VDC/2 T4,OFF VDC/2 T4,OFF R VDC/2 T4,OFF

N N N

(c)

Deadtime freewheeling
P
Negative state P P Positive state
ET1,ON = 0
C1 C1 C1
T1,OFF ET3,OFF=EIGBT,OFF T1,OFF T1,ON
VDC/2 D2 D3 VDC/2 D2 D3 A L VDC/2 D2 D3 L
A L A
O From O From O
C2 O to P C2 T2,ON T3,OFF O to P C2 T2,ON T3,OFF
T2,ON T3,ON R T4,OFF R R
VDC/2 T4,OFF VDC/2 VDC/2 T4,OFF

N N N

(d)

Fig. 4. Different switching transitions when the phase output is positive or neutral. (a) Switch transition 1 (positive phase current). (b) Switch transition 2 (positive
phase current). (c) Switch transition 3 (negative phase current). (d) Switch transition 4 (negative phase current).

and 0 is preferred since the clamping leg is soft-switching, can be modeled through (1).
as shown in Fig. 4(a) and (b). Switch pair 0 and -1 should
be avoided since clamping leg switch T2 is hard switching. VCM = (VAO + VBO + VCO)/3 (1)
Symmetrically when phase current is negative, switch pair 0
and -1 is preferred to obtain the soft-switching character of Space vectors in sector 1 of the modulation hexagon,
clamping leg, and switch pair 1 and 0 should be avoided or which is shown in Fig. 7, are given as an example for CMV
minimized to reduce the switching loss of Si IGBT, as shown calculation. As shown in Table II, small vector V1 and V5 have
in Fig. 4(c) and (d). Based on the aforementioned analysis, the largest CMV, while zero vector V8 and medium vector V3
different SVMs can be compared in terms of switching loss, do not contribute to CMV.
more detailed information will be given in the later section.
C. Switching Loss Reduction, NPB and CM Voltage Analysis
B. Common-Mode Voltage of 3-L TNPC Inverter for Different SVM Schemes
Assuming that only heatsink is grounded, the equivalent Firstly, space vectors of sector 1 are marked out in Fig. 7 and
model of 3 phase 3-L TNPC [28]-[33] is drawn in Fig. 5, small vectors’ influence on neutral point potential are stated in
with the consideration of all the semiconductor’s junction to Table III. For simplicity, the regions 3 and 2 are analyzed and
heatsink and output to ground capacitance. Also, a simplified compared for three kinds of SVM schemes, choices of small
model is given in Fig. 6, which indicates that the CMV noise vectors and alignment in the regions 1 and 4 are similar to the
 H. PENG et al.: IMPROVED SPACE VECTOR MODULATION FOR NEUTRAL-POINT BALANCING CONTROL IN INVERTERS 331

CSiC-H CSiC-H CSiC-H TABLE II

CMV of Space Vectors
Cline

VDC/2 Space vector V1 V2 V3 V4

VCM
CDC-link
CSi-H
A CSiC-H CDC-link 3*CL
O CMV − 1/3Vdc − 1/6Vdc 0 1/6Vdc
N
Space vector V5 V6L/3 V7 V8
CStray
CSi-H
B CSiC-H CStray 2*Clink
/3Vdc − 6*C dc //C1/6V
1/6VSiC-H Si-H dc 0C
C
CM

VDC/2 CSiC-H
CDC-link TABLE III
CSi-H CSi-H/CSiC-H: pF
Small Vector’s C
range : pF range
Influence on NP Potential
line

CL CL CLL
Cline LA LC CL: pF range CStray: pF range
LB
Small Vector Angle Discharge NP Charge NP
CCM: nF range CDC-link: μF to mF range
0o V4 V1
N CCM 60o V5 V2

CSiC-H TABLE IV
Fig. 5. Three-phase 3-L TNPC model with the parasitic capacitor. Space Vector Choice and Alignment in Sector 3

CDC-link VCM The region 3 of Sector 1 Discharge NP Charge NP

3*CL
O N SVM 1 V3 -V2 -V1 -V2 -V3 V3 -V4 -V5 -V4 -V3
CStray L/3 SVM 2 & 0 o ~30 o V1 -V2 -V3 -V2 -V1 V2 -V3 -V4 -V3 -V2
CSiC-H 2*Clink 6*CSiC-H//CSi-H
CCM SVM 2 & 30 o ~60 o V4 -V3 -V2 -V3 -V4 V5 -V4 -V3 -V4 -V5
C CSiC-H Improved SVM 2 & I b < 0 V1 -V2 -V3 -V2 -V1 V2 -V3 -V4 -V3 -V2
Improved SVM 2 & I b > 0 V4 -V3 -V2 -V3 -V4 V5 -V4 -V3 -V4 -V5
CSi-H/CSiC-H: pF range Cline: pF range
CL CLL
LC CL: pF range CStray: pF range
CCM: nF range CDC-link: μF to mF range 1) The Region 3 of Sector 1
The space vector choice and alignment for SVM 1, SVM
CCM 2, and improved SVM 2 are shown in Table IV. For SVM 1
Fig. 6. Three-phase 3-L TNPC CMV path.
scheme, both of the small vectors V1, V2, V4, and V5 are used
for balancing the neutral point potential, and B phase switching
state changes between 1 and 0 plus 0 and -1. Since SVM 1
does not consider the B phase current direction, B phase has
hard switching operations on the clamping leg. For SVM 2
scheme, the region 3 is divided into the region 3.1 and the
region 3.2 according to B phase voltage polarity. In the region
3.1, only V1 and V4 small vectors are used for balancing the
neutral point potential. In the region 3.2, only V2 and V5 small
vectors are used for balancing the neutral point potential. B
phase clamping leg is always soft-switching under unity PF.
Improved SVM 2 scheme is proposed to reduce the switching
loss under the wider power factor range. When the B phase
current is negative, V1 and V4 small vectors are used for
1: Connected to positive rail
balancing the neutral point potential. V2 and V5 small vectors
0: Connected to clamping leg are used when the B phase current is positive. Improved SVM
1: Connected to negative rail 2 scheme is basically the same as SVM 2 scheme under the
unity power factor, and since it considers the phase current
Fig. 7. Three-phase three-level space vector hexagon. direction for choosing the small vector, and thus, it has lower
losses on clamping leg under the non-unity power factor case.
regions 3 and 2. From Table II it is known that small vectors
V2 and V4 have lower CM voltage than V1 and V5, less small 2) The Region 2 of Sector 1
vectors like V2 and V4 are chosen, smaller CM voltage the For region 2, modulation strategies of SVM 1, SVM 2, and
SVM schemes can get. improved SVM 2 are the same, which is shown in Table V.
332 CPSS TRANSACTIONS ON POWER ELECTRONICS AND APPLICATIONS, VOL. 4, NO. 4, DECEMBER 2019

TABLE V
Space Vector Choice and Alignment in Sector 2

Power
The region 2 of Sector 1 Discharge NP Charge NP
Supply
Shunt Cable
SVM 1 V1 -V6 -V3 -V6 -V1 V6 -V3 -V4 -V3 -V6
Resistor
SVM 2 V1 -V6 -V3 -V6 -V1 V6 -V3 -V4 -V3 -V6
Improved SVM 2 V1 -V6 -V3 -V6 -V1 V6 -V3 -V4 -V3 -V6

3L TNPC DUT DC-link

TABLE VI Capacitor FPGA
Device Parameters
CM
Choke
Load Auxiliary Power
Inductor Supply

Fig. 8. Test setup of DPT.

transitions of both SiC MOSFET and IGBT devices. As

V1 and V4 are chosen to discharge and charge the neutral point mentioned in [37], switching energy in 3-L phase leg is
voltage, and alignments are V6-V3-V4-V3-V6 and V1-V6-V3-V6-V1 different from 2L half-bridge due to the device’s junction
when the neutral point voltage is lower and higher than half capacitance, so the double pulse test in this paper is performed
of the DC-link voltage. Soft switching can be achieved on the based on the single-phase 3-L TNPC platform. In this way, loss
clamping leg of B phase under the unity power factor. analysis using switching energy data from the double pulse test
on the 3-L TNPC platform gives more accurate results. The
3) Summary of Different SVM Schemes accurate switching energy information for SiC MOSFETs and
In general, the improved SVM 2 scheme has the lowest Si IGBT switch is shown in Fig. 9.
switching loss on clamping leg. Moreover, under the non-
unity power factor condition, SVM 1 has better neutral
B. Semiconductor Loss Breakdown and EMI Performance
point balancing ability than SVM 2 and improved SVM 2.
Evaluation
According to Table II, V1 and V5 have the highest CM voltage.
Since SVM 2 and improved SVM 2 exclude V1 or V5 for Switching energy from DPT and device conduction perform-
reducing the switching loss, and SVM 1 employs both of the V1 ance in the component datasheet are used, and a detailed
and V5 for NPB, SVM 2 and improved SVM 2 have relatively semiconductor loss breakdown can be obtained in simulation.
lower CMV than that of SVM 1. While the total power level is 6 kVA, and the switching
frequency is 70 kHz.
III. Experimental Test and Loss Breakdown For the near unity PF case, the loss breakdown for different
A 6 kVA 3-phase 3-L TNPC prototype is built to evaluate modulation schemes is listed in Fig. 10(a), and total semicon-
the efficiency and CM voltage noise spectrum with different ductor loss using SVM 1, SVM 2, and improved SVM 2 are
SVM, upper and lower DC-link capacitors are 150 µF each. respectively 17.1 W, 16 W, and 16 W. As shown in the diagram,
As shown in Table VI, we have chosen 1200 V/30 A SiC under the unity power factor case, total loss of the SVM 2 and
MOSFET, C3M0075120K from Wolfspeed, and 600 V/30 the improved SVM 2 are 6.4% lower than using SVM 1. All
of the SVMs have the same neutral point voltage ripple, which
A Si-IGBT IKZ50N65EH5 from ROHM, and FFSH1665A
is around 2 V due to the hysteresis control algorithm at half
SiC Schottky diode, for experiment validation. Top leg and
of the switching frequency and switching actions at switching
bottom leg switches are rated for 1.2 kV, and DC-link voltage
frequency.
is set to be 800 V to remain some safety margin. Output RMS
However, for PF = 0.8 (lead or leg) case, loss breakdown for
voltage is set to 208 V to meet one of the grid standards.
different modulation schemes are listed in Fig. 10(c), and total
Switching frequency is set to be 70 kHz to reduce the passive
semiconductor loss using SVM 1, SVM 2 and improved SVM
components’ volume. The prototype is composed of three
2 are respectively 22.7 W, 18.9 W, and 17.9 W. It can be seen
2kVA single-phase 3-L TNPC, as shown in Fig. 8.
that by adopting SVM 1, clamping leg devices have excessive
switching loss. In this condition. SVM 2 and improved SVM
A. Switching Energy Evaluation of the Three-Level Inverter 2 have respectively 4.2 times and 5.3 times lower clamping
Switching energy calculation based on the devices’ voltage leg devices’ switching loss than that of SVM 1. Neutral point
and devices’ current waveform is comprehensively evaluated ripple voltage using SVM 1, SVM 2, and improved SVM 2 are
in [34]–[36]. DPT is firstly performed to obtain the switching respectively 2.3 V, 8.5 V, and 15 V.
H. PENG et al.: IMPROVED SPACE VECTOR MODULATION FOR NEUTRAL-POINT BALANCING CONTROL IN INVERTERS 333

Switching Energy Information for SiC MOSFETs Switching Energy Information for Si IGBT
400 3500
350 3000
300
Switching loss (μJ)

Switching loss (μJ)

2500
250
2000
200
1500
150
100 1000
50 500
0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
IDS (A) IDS (A)
(a) (b)

Fig. 9. Switching loss information for SiC MOSFET position and Si IGBT position. (a) Switching energy information for SiC MOSFET position. (b) Switching
energy information for Si IGBT position.

Power Factor = 1 Power Factor = 0.9

Conduction Loss
8 Switching Loss 8 Conduction Loss
Switching Loss
6 6
Power loss (W)

Power loss (W)

4 4

2 2

0 0
D3 D3
D2 D2
T4 T4
T3 T3
T2 SVM 1 T2 SVM 1
T1 SVM 2 T1 SVM 2
Imp. SVM 2 Imp. SVM 2

(a) (b)

Power Factor = 0.8 Power Factor = 0.7

8 8 Conduction Loss
Conduction Loss Switching Loss
6 Switching Loss 6
Power loss (W)

Power loss (W)

4 4

2 2

0 0
D3 D3
D2 D2
T4 T4
T3 T3
T2 SVM 1 T2 SVM 1
T1 SVM 2 T1 SVM 2
Imp. SVM 2 Imp. SVM 2

(c) (d)

Fig. 10. Loss breakdown of semiconductor devices at different power factors. (a) Loss breakdown when PF = 1. (b) Loss breakdown when PF = 0.9. (c) Loss
breakdown when PF = 0.8. (d) Loss breakdown when PF = 0.7.

The difference of SVM schemes is the switching transitions switch does not change a lot for different SVM schemes. So
and soft-switching conditions of the clamping leg switches even the conduction loss of different modulations have some
under different conditions, the average duty ratio for each difference, this difference in conduction loss is not as much as
334 CPSS TRANSACTIONS ON POWER ELECTRONICS AND APPLICATIONS, VOL. 4, NO. 4, DECEMBER 2019

14

Neutral Point Ripple Voltage (V)

12
10
8
6
4
2
0
0.7 0.75 0.8 0.85 0.9 0.95 1 0.95 0.9 0.85 0.8 0.75 0.7
Power Factor (Leading) Power Factor (Lagging)

Fig. 11. Neutral point ripple voltage.

in switching loss, and its not obvious in the figures.

In summary, with the power factor decreasing from 1 to 0.7, 21
SVM 1 has a much higher loss on clamping leg devices than 20.5
SVM 2 and improved SVM 2, which may result in device 22 20
overheat issue. And it can be seen in Fig. 10 that under different 19.5
power factor cases, SVM 2 and improved SVM 2 have more 20
19
equal loss distribution among switching devices, so SVM 2 18.5
18
and improved SVM 2 are preferred in terms of semiconductor 18
loss reduction and the semiconductor loss distribution point of 16 17.5
view. 80 17
The neutral point voltage ripple versus power factor re- 60 1 16.5
lationship is shown in Fig. 11. We can know that when the 40 0.9 16
Neutral point
power factor is in the region of 0.85 to 1 (lead or lag), the SVM ripple voltage (V)20 0.8 Power factor
1 and improved SVM 2 have higher neutral point voltage than 0 0.7 (leading)
SVM 2. When the power factor is high enough, SVM 2 also
has as good clamping leg loss reduction capability as improved
Fig. 12. Trade-off between NPB and loss performance for improved SVM 2.
SVM 2. So when the power factor is between 0.85 and 1, it is
better to use the SVM 2 modulation scheme for both clamping performance of different SVM schemes. In this simulation,
leg loss reduction as well as NPB purpose. Vasw represents the output phase voltage switching waveform,
When the power factor is in the region of 0.7 to 0.85 (lead
Iph represents the phase current, Vnp is the NP voltage, and Vcm
or lag), the SVM 1 still possesses the best NPB capability,
shows the common-mode voltage. As shown in Fig. 13, when
and in the meantime, the improved SVM 2 has better NPB
PF = 0.8, SVM 1 has a higher CM voltage ripple than SVM 2,
performance than the SVM 2. Moreover, when the power
and improved SVM 2. Under PF = 1 and PF = 0.8 cases, CM
factor is relatively lower, improved SVM 2 has better clamp-
voltage and phase leg output voltage spectrum are shown in
ing leg loss reduction capability than the SVM 2. When the
Fig. 14. As shown in Fig. 14(a) and (c), from 10 to 100 kHz
power factor is between 0.85 and 1, the improved SVM 2
serves better for both clamping leg loss reduction and NPB range which is of great significance in EMI filter design, SVM
purpose. 2 and improved SVM 2 have lower CM noise than SVM 1,
As shown in Fig. 12, comparison is made for evaluating especially at relatively low power factor case. It is shown in
the trade-off between NPB and loss performance using the Fig. 14(b) and (d), since SVM 2, and improved SVM 2 have
improved SVM 2. Hysteresis width for neutral point balancing higher neutral point unbalanced voltage under non-unity power
algorithm is controlled for obtaining different neutral point factor, their phase leg voltage has slightly larger harmonics
voltage ripple. It can be seen from Fig. 12 that, when neutral (300 Hz, 420 Hz, etc.) than SVM 1. In terms of common-
point ripple is increasing within a certain range, the total loss mode filter and output filter design, SVM 2 and improved
is decreasing due to less number of changing space vector SVM 2 are preferred due to their lower better common-mode
alignment actions. However, when neutral point voltage ripple voltage harmonics as well as comparable phase output voltage
increases, the commutation voltage of the device is higher spectrum.
due to the unbalanced neutral point, which deteriorates the In summary, SVM 2 and improved SVM 2 have better
loss performance. It is recommended that the hysteresis width performance regarding semiconductor loss and common-mode
should be kept within 30 V for a neutral point balancing control voltage reduction than SVM 1, but SVM 1 has the best neutral
algorithm. point balancing capability. When the power factor is between
MATLAB simulation has been conducted to compare EMI 0.85 and 1, SVM 2 is adopted for better overall performance,
H. PENG et al.: IMPROVED SPACE VECTOR MODULATION FOR NEUTRAL-POINT BALANCING CONTROL IN INVERTERS 335

Vasw Vasw Vasw

800 800 800
600 600 600
400 400 400
200 200 200
0 0 0
0.015 　 0.02 　0.025 　0.03 0.015 　0.02 　 0.025 　0.03 0.015 　0.02 　 0.025 　0.03
Iph Iph Iph
10 10 10
0 0 0
10 10 10

0.015 　0.02 　0.025 　 0.03 0.015 　0.02 　0.025 0.03 0.015 　0.02 　0.025 　 0.03
Vnp Vnp Vnp
410 410 410
405 405 405
400 400 400
395 395 395
390 390 390
0.015 　0.02 　0.025 　 0.03 0.015 　0.02 　0.025 0.03 0.015 　0.02 　0.025 　 0.03
Vcm Vcm Vcm
200 200 200
0 0 0
200 200 200
0.015 　0.02 　 0.025　 0.03 0.015 　0.02 　0.025 　0.03 0.015 　0.02　 0.025 　 0.03

(a) (b) (c)

Fig. 13. Phase leg output voltage, phase current, NP voltage, and CM voltage waveform when PF = 0.8. (a) Waveform for the SVM 1. (b) Waveform for the SVM 2.
(c) Waveform for improved SVM 2.

VCM with PF = 1 Vph with PF = 1

170
160 SVM 1
160 SVM 2
Improved SVM 2
Magnitude (dBμV)

Magnitude (dBμV)

150 150

140 140

130
130
120
120 SVM 1 110
SVM 2
Improved SVM 2
110 100
101 102 103 104 105 106 101 102 103 104 105 106
Frequency (Hz) Frequency (Hz)
(a) (b)

VCM with PF = 0.8 Vph with PF = 0.8

170
160
160 SVM 1
SVM 2
Improved SVM 2
Magnitude (dBμV)

150
Magnitude (dBμV)

150

140 140

130
130
120
120 SVM 1 110
SVM 2
Improved SVM 2
110 100
101 102 103 104 105 106 101 102 103 104 105 106
Frequency (Hz) Frequency (Hz)
(c) (d)

Fig. 14. CMV and phase voltage spectrum at different power factors. (a) CMV when PF = 1. (b) Phase voltage spectrum when PF = 1. (c) CMV when PF = 0.8. (d)
Phase voltage spectrum when PF = 0.8.
336 CPSS TRANSACTIONS ON POWER ELECTRONICS AND APPLICATIONS, VOL. 4, NO. 4, DECEMBER 2019

and when the power factor is between 0.7 to 0.85, improved performance embedded sic power module based on a dbc-stacked hybrid
SVM 2 can be utilized for its’ overall better loss reduction as packaging structure,” in IEEE Journal of Emerging and Selected Topics
in Power Electronics, 2019.
well as neutral point balancing performance. [11] Z. Wang, M. H. Mahmud, M. H. Uddin, B. McPherson, B. Sparkman,Y.
Zhao, H. A. Mantooth, and J. R. Fraley, “A compact 250 kw
IV. Conclusion siliconcarbide mosfet based three-level traction inverter for heavy
equipment applications,” in 2018 IEEE Transportation Electrification
In this paper, semiconductor losses of different commutation Conferenceand Expo (ITEC), 2018, pp. 1129–1134.
loops in hybrid switch combination TNPC are analyzed and [12] J. He, R. Katebi, and N. Weise, “A current-dependent switching
compared. Based on different switching loss of commutation strategyfor si/sic hybrid switch-based power converters,” in IEEE
loops, SVM 1 and SVM 2 themes are utilized and compared Transactions onIndustrial Electronics, vol. 64, no. 10, pp. 8344–8352,
in this topology comprehensively in terms of their influences 2017.
on switching loss, NPB and EMI spectrum, and then improved [13] A. Deshpande, Y. Chen, B. Narayanasamy, A. S. Sathyanarayanan, and
F. Luo, “A three-level, t-type, power electronics building block using
SVM 2 is proposed to further push the converter to higher sisic hybrid switch for high-speed drives,” in 2018 IEEE Applied Power
efficiency at relatively low power factor. Electronics Conference and Exposition (APEC), 2018, pp. 2609–2616.
Comparing to SVM 1, SVM 2, and improved SVM 2 schemes [14] A. Deshpande and F. Luo, “Design of a silicon-wbg hybrid switch,” in
have better loss performance and CM noise performance, while Wide Bandgap Power Devices and Applications (WiPDA), 2015 IEEE
improved SVM 2 scheme has the lowest power loss and lowest 3rd Workshop on, 2015, pp. 296–299.
CM noise. As for neutral point balancing capability, SVM 1 [15] A. Deshpande and F. Luo, “Multilayer busbar design for a Si IGBT
and SiC MOSFET hybrid switch based 100 kW three-level T-type
is better than SVM 2 and improved SVM 2 under non-unity PEBB,” 2017 IEEE 5th Workshop on Wide Bandgap Power Devices and
power factor, and their neutral point balancing capabilities are Applications (WiPDA), Albuquerque, NM, 2017, pp. 20-24.
the same under the unity power factor. [16] A. Deshpande and F. Luo, “Comprehensive evaluation of a silicon-WBG
Hybrid-switch-based 3 phase 3-L TNPC is comprehensively hybrid switch,” 2016 IEEE Energy Conversion Congress and Exposition
studied in this paper, and it gives guidance for hybrid switch (ECCE), Milwaukee, WI, 2016, pp. 1-8.
topology design consideration and the choice of SVM strategy. [17] A. Deshpande and F. Luo, “Practical design considerations for a Si
IGBT + SiC MOSFET hybrid switch: parasitic interconnect influences,
cost, and current ratio optimization,” in IEEE Transactions on Power
References Electronics, vol. 34, no. 1, pp. 724-737, Jan. 2019.
[1] H. Peng, Z. Yuan, B. Narayanasamy, X. Zhao, A. Deshpande, and F. [18] S. Zhao, X. Zhao, A. Dearien, Y. Wu, Y. Zhao and H. A. Mantooth, “An
Luo, “Comprehensive analysis of three-phase three-level t-type neutral- intelligent versatile model-based trajectory optimized active gate driver
pointclamped inverter with hybrid switch combination,” in 2019 IEEE for silicon carbide devices,” in IEEE Journal of Emerging and Selected
10th International Symposium on Power Electronics for Distributed Topics in Power Electronics.
Generation Systems (PEDG), 2019, Xi’an, China, pp. 816–821. [19] S. Zhao, A. Dearien, Y. Wu, C. Farnell, A. U. Rashid, F. Luo, andH.
[2] J. Rosero, J. Ortega, E. Aldabas, and L. Romeral, “Moving towardsa A. Mantooth, “Adaptive multi-level active gate drivers for sic
more electric aircraft,” in IEEE Aerospace and Electronic Systems powerdevices,” in IEEE Transactions on Power Electronics, 2019.
Magazine, vol. 22, no. 3, pp. 3–9, 2007. [20] H. Gui, Z. Zhang, R. Chen, J. Niu, L. M. Tolbert, F. F. Wang, D.
[3] R. DelRosario, “A future with hybrid electric propulsion systems: Anasa Costinett, B. J. Blalock, and B. B. Choi, “Gate drive technology evaluation
perspective,” 2014. and development to maximize switching speed of sic discretedevices
[4] Y. Chen, Z. Yuan, and F. Luo, “A model-based multi-objective and power modules in hard switching applications,” in IEEE Journal of
optimizationfor high efficiency and high power density motor drive Emerging and Selected Topics in Power Electronics, 2019.
inverters foraircraft applications,” in NAECON 2018-IEEE National [21] A. Anthon, Z. Zhang, M. A. Andersen, D. G. Holmes, B. McGrath, and C.
Aerospace andElectronics Conference, 2018, pp. 36–42. A. Teixeira, “The benefits of sic mosfet s in a t-type inverter forgrid-tie
[5] Z. Yuan, H. Peng, A. Deshpande, B. Narayanasamy, A. I. Emon, F. applications,” in IEEE Transactions on Power Electronics, vol. 32,no. 4,
Luo,and C. Chen, “Design and evaluation of laminated busbar for pp. 2808–2821, 2017.
3-levelt-type npc power electronics building block with enhanced [22] H. Chen, M. Tsai, Y. Wang, and P. Cheng, “A Modulation technique for
dynamiccurrent sharing,” in IEEE Journal of Emerging and Selected neutral point voltage control of the three-level neutral-point-clamped
Topics in Power Electronics, pp. 1–1, 2019. converter,” in IEEE Transactions on Industry Applications, vol. 54, no. 3,
[6] G. Li, H. Li, A. Deshpande, X. Li, L. Xu, F. Luo, and J. pp. 2517–2524, May-June 2018.
Wang,“Comparison between 1.7 kv sic sjt and mosfet power modules,” [23] Y. Jiao, F. C. Lee, and S. Lu, “Space vector modulation for three-
in 2016 IEEE 4th Workshop on Wide Bandgap Power Devices and levelnpc converter with neutral point voltage balance and switching loss
Applications(WiPDA), 2016, pp. 17–22. reduction,” in IEEE Transactions on Power Electronics, vol. 29, no. 10,
[7] C. Chen, F. Luo, and Y. Kang, “A review of sic power module pp. 5579–5591, 2014.
packaging:Layout, material system and integration,” in CPSS [24] U.-M. Choi, J.-S. Lee, and K.-B. Lee, “New modulation strategy
Transactions on Power Electronics and Applications, vol. 2, no. 3, pp. tobalance the neutral-point voltage for three-level neutral-clamped
170–186, 2017. inverter systems,” in IEEE Transactions on Energy Conversion, vol. 29,
[8] C. Chen, Y. Chen, Y. Li, Z. Huang, T. Liu, and Y. Kang, “An sic based no. 1, pp.91–100, 2014.
half-bridge module with an improved hybrid packaging methodfor [25] J. Holtz, M. Holtgen, and J. O. Krah, “A space vector modulator forthe
high power density applications,” in IEEE Transactions on Industrial high-switching frequency control of three-level sic inverters,” in IEEE
Electronics, vol. 64, no. 11, pp. 8980–8991, 2017. Transactions on Power Electronics, vol. 29, no. 5, pp. 2618–2626, 2014.
[9] C. Chen, Z. Huang, L. Chen, Y. Tan, Y. Kang, and F. Luo, “Flexible [26] N. Babu and P. Agarwal, “Nearest and non-nearest three vector
pcb-based 3-d integrated sic half-bridge power module with three-sided modulations of npci using two-level space vector diagram—A novel
cooling using ultralow inductive hybrid packaging structure,” in IEEE approach,”in IEEE Transactions on Industry Applications, vol. 54, no. 3,
Transactions on Power Electronics, vol. 34, no. 6, pp. 5579–5593, 2018. pp. 2400–2415, 2018.
[10] Z. Huang, C. Chen, Y. Xie, Y. Yan, Y. Kang, and F. Luo, “A high [27] Y. Jiao, “High power high frequency 3-level neutral point clamped
H. PENG et al.: IMPROVED SPACE VECTOR MODULATION FOR NEUTRAL-POINT BALANCING CONTROL IN INVERTERS 337

powerconversion system,” Ph.D. dissertation, Virginia Polytechnic Xingchen Zhao was born in Jiangsu, China,
Institute and State University, 2015. in1993. He received the B.S. degree and M.S.
[28] H. Zhang, L. Yang, S. Wang, and J. Puukko, “Common-mode emi degree in Electrical Engineering from Nanjing
noisemodeling and reduction with balance technique for three-level University of Aeronautics and Astronautics,
neutralpoint clamped topology,” in IEEE Transactions on Industrial Nanjing, China, in 2015 and 2018, respectively.
Electronics, vol. 64, no. 9, pp. 7563–7573, 2017. He is currently pursuing the Ph.D. Degree with the
[29] A. Deshpande, B. Narayanasamy, and F. Luo, “Analysis of conducted Department of Electrical Engineering, University
emi in si igbt + sic mosfet hybrid switch based converters,” in 2018 IEEE of Arkansas, Fayetteville, AR, USA. His research
International Symposium on Electromagnetic Compatibility and2018 interests include motor drive, multi-level inverters,
IEEE Asia-Pacific Symposium on Electromagnetic Compatibility (EMC/ and applications of wide bandgap devices.
APEMC), 2018, pp. 127–127.
[30] B. Narayanasamy, F. Luo, and Y. Chu, “Modeling and stability analysisof
voltage sensing based differential mode active emi filters for ac-dc power
converters,” in 2018 IEEE Symposium on ElectromagneticCompatibility,
Signal Integrity and Power Integrity (EMC, SI & PI), 2018, pp. 322–328.
[31] B. Narayanasamy, A. S. Sathyanarayanan, A. Deshpande, and F. Luo, Balaji Narayanasamy received his bachelor’s
“Analysis and mitigation of reflected wave voltages and currents inwbg degree in Electrical and Electronics Engineering
devices based motor drives,” in 2016 IEEE 4th Workshop on WideBandgap from Amrita University, India in 2012, and master’s
Power Devices and Applications (WiPDA), 2016, pp.297–301. degree in Electrical Engineering from The Ohio
State University, Columbus, Ohio in 2016. He is
[32] B. Narayanasamy, A. S. Sathyanarayanan, A. Deshpande and F. Luo, “Impact
currently pursuing his Ph.D. degree in Electrical
of cable and motor loads on wide bandgap device switching and reflected
Engineering at the University of Arkansas,
wave phenomenon in motor drives,” 2017 IEEE Applied Power Electronics
Fayetteville, Arkansas. He was a Research Assistant
Conference and Exposition (APEC), Tampa, FL, 2017, pp. 931-937.
from August 2015 to July 2018 at The Ohio State
[33] B. Narayanasamy, F. Luo and Y. Chu, “High density EMI mitigation
University and the University of Arkansas. He is
solution using active approaches,” in 2017 IEEE International currently serving as a Teaching Assistant at the University of Arkansas.
Symposium on Electromagnetic Compatibility & Signal/Power Integrity Before joining his master’s degree, he worked as a Product Engineer at
(EMCSI), Washington, DC, 2017, pp. 813–818. L&T Kobelco, India from 2012 to 2014.
[34] D. Christen and J. Biela, “Analytical switching loss modeling based His research interests include analysis and mitigation of EMI in Wide
ondatasheet parameters for mosfets in a half-bridge,” in IEEE Bandgap devices-based power converters. He has authored and co-authored
Transactionson Power Electronics, vol. 34, no. 4, pp. 3700–3710, 2019. twelve papers which were presented in different international conferences.
[35] Z. Zhang , “Characterization and realization of high switching-speed
capability of SiC power devices in voltage source converter.” Ph.D.
dissertation, University of Tennessee, 2015.
[36] Z. Chen, “Characterization and modeling of high-switching-speed
behaviorof sic active devices,” Ph.D. dissertation, Virginia Tech, 2009.
[37] B. Liu, R. Ren, E. A. Jones, H. Gui, Z. Zhang, R. Chen, F. Wang, and
D. Costinett, “"Effects of Junction Capacitances and Commutation Amol Deshpande was born in Vadodara, India, in
Loops Associated With Line-Frequency Devices in Three-Level AC/DC 1991. He received a B.Tech degree in Electrical
Converters,” in IEEE Transactions on Power Electronics, vol. 34, no. 7, Engineering from Nirma University in India
pp. 6155-6170, July 2019. and an M.S. degree in Electrical and Computer
Engineering from The Ohio State University,
Columbus, OH, in 2013 and 2016, respectively.
Hongwu Peng Received the B.S. degree in Electrical He is currently working toward the Ph.D. degree
Engineering from the Huazhong University of in Electrical Engineering with the University of
Science and Technology, Wuhan, China, in 2018. Arkansas, Fayetteville, AR. He is currently a
He is currently working toward the Ph.D. degree graduate research assistant and involved in research
in the University of Arkansas at Fayetteville, AR, projects with the National Science Foundation (NSF) Engineering Research
USA. Center for Power Optimization of Electro-thermal Systems (POETS). In
2018, he was a research intern at Kilby Labs, Texas Instruments, Santa
His research interests include high-efficiency
Clara, CA, USA.
motor drive with WBG devices and active EMI
His current research interests include high-voltage power module packaging
cancellation. and electro-thermal design of high-power density inverters for traction and
aircraft applications using wide-bandgap power semiconductor devices.He
Zhao Yuan was born in Jiaozuo, China, in received a Best Paper Award at the 2016 IEEE Energy Conversion Congress
1992. He received the B.S. degree in Electrical and Exposition (ECCE).
Engineering from Huazhong University of Science
and Technology and the master degree in Electrical
Engineering from Arizona State University, Tempe,
USA, in 2015 and 2017 respectively. He is currently
working toward the Ph.D. degree at University
of Arkansas, Fayetteville, Arkansas, USA. He Asif Imran Emon was born in Chittagong, Bangladesh
is a graduate researching assistant involved in in 1993. He received his B.Sc. degree in Electrical
the research project with the National Science and Electronic Engineering from Chittagong
Foundation (NSF) Engineering Research Center for Power Optimization University of Engineering and Technology in 2015.
of Electro-thermal Systems (POETS). In 2019, he was a research intern at He is currently a Ph.D. student at the University of
United Technologies Research Center, East Hartford, CT, USA. Arkansas, Fayetteville, Arkansas, USA.
His research interests include design and electro-thermal optimization His research interests include power module
of high-power converters for traction systems, high-voltage power module packaging of wide band gap devices, Electro-
packaging. He received 2nd place student paper award at the 2018 National magnetic Interference and Compatibility in high
Aerospace & Electronics Conference (NAECON). power application.
338 CPSS TRANSACTIONS ON POWER ELECTRONICS AND APPLICATIONS, VOL. 4, NO. 4, DECEMBER 2019

Fang Luo received the bachelor’s and Ph.D. degrees Cai Chen is an Associate Research Fellow at
in Electrical Engineering from the Huazhong Huazhong University of Science and Technology.
University of Science and Technology,Wuhan, He received the B.S. degree and Ph.D. degree
China, in 2003 and 2010, respectively. He is in Electrical and Electronic Engineering from
currently an Assistant Professor with the University Huazhong University of Science and Technology,
of Arkansas,Fayetteville, AR, USA. From 2007 to Wuhan, China, in 2008 and 2014, respectively.
2010, he was a joint Ph.D. student at the Center for From March 2013 to December 2013, he was an
Power Electronics Systems (CPES), Virginia Tech, Intern in GE Global Research Center, Shanghai,
supported by the Chinese Scholarship Council and China. From 2014 to 2016, he joined the Advanced
CPES. From 2010 to 2014, he was with CPES, Semiconductor, Packaging and Integration Lab,
Virginia Tech, first as a Postdoctoral Researcher and then as a Research Huazhong University of Science and Technology, Wuhan, Hubei, China as a
Scientist. From 2014 to 2017, he was a Research Assistant Professor Postdoctoral Researcher. From 2016 to Oct. 2017, he was a visiting scholar
with The Ohio State University. In July 2017, he joined the University of in the Center for High Performance Power Electronics, The Ohio State
Arkansas as a Tenure Track Assistant Professor. University, Columbus, OH, USA. From 2017 to Oct. 2018, he was a visiting
His research interests include turbo electric propulsion converters, high- scholar in the College of Engineering, University of Arkansas, Fayetteville,
power-density converter design, high-density electromagnetic interference AR, USA. In 2019, Dr. Chen joined the Huazhong University of Science
filter design, and power module packaging/integration for wide-bandgap and Technology, Wuhan, China, as an Associate Research Fellow.
devices. His research interests include WBG devices packaging, integration,
pack-aging EMI issues, packaging reliability and high-density applications.