IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 62, NO.

2, FEBRUARY 2015 735

A Switched-Capacitor Three-Phase
AC–AC Converter
Telles B. Lazzarin, Member, IEEE, Romero L. Andersen, Member, IEEE, and Ivo Barbi, Fellow, IEEE

Abstract—This paper proposes a bidirectional three- implemented successfully in chips [3], initially in the range
phase direct ac–ac converter, with only capacitors and of ultralow power, but have shown potential to be used in the
switches in its power circuit and with its operation based power range of inductive converters [4].
on the switched-capacitor principle. The converter presents
fixed gain, it keeps the frequencies of the output and input
Currently, the number of topologies that use this principle
voltages equal, and it operates in open loop with constant is increasing, and the SC is already employed in high power
switching frequency and duty cycle. The main advantages dc–dc converters for electric vehicles, photovoltaic applica-
of the proposed converter are the absence of magnetic tions [5], and circuits to balance charges in batteries [6].
elements in the power circuit, the higher efficiency, the Multilevel inverters have also been presented in the literature
higher power density, the higher specific power, the lower with series–parallel SC connections [7], different modulation
cost, and the fact that it can convert ac–ac voltages within
a wide frequency range, including dc voltage. Therefore, it methods [8], and hybrid topologies [9]. Other studies in the
is suitable to replace the conventional three-phase auto- SC field include associating coupled inductors with SC [10]
transformer in industrial, commercial, and residential ap- and modular converters [11] for high gain, as well as resonant
plications, and it can be designated as a magnetic-less converters with zero voltage switching (ZVS) [12] and phase-
three-phase solid-state autotransformer. The principle of shift interleaving [13].
operation, a qualitative and quantitative analysis, the de-
sign methodology, and a fabricated example are described
SC power converters (SCPCs) are composed only of capaci-
in this paper. In order to verify the converter in the labo- tors and switches, with no magnetic devices, and thus, they can
ratory, a prototype with the following characteristics was achieve a significant size reduction and increased efficiency in
designed and fabricated: 6 kW, 1.35 kW/kg, 380/110 V, and relation to the conventional converters. SCPC can also regulate
switching frequency of 100 kHz. The measured efficiency the output voltage, by changing the duty cycle, similarly to
at rated power was 96.3%, and other relevant experimental conventional power converters. Furthermore, the behavior of
results are reported herein.
SCs can be described as that of simple equivalent circuits [14],
Index Terms—AC–AC converter, switched capacitor (SC), which help to evaluate the impact of SC parameters on effi-
three phase. ciency [15]. Sometimes, these equivalent circuits can be used to
I. I NTRODUCTION represent subcircuits of a bigger SC system [16], [17], making
the analysis easier. In spite of the importance of these simple

T HE AC–AC power converters have been studied as an

alternative to replace autotransformers and transformers.
The use of ac–ac power converters to supply the load with
equivalent circuits on loss calculations [18], space-state models
and dynamic models of SC converters [19] were also presented.
Recently, novel proposals of single-phase direct ac–ac power
sinusoidal three-phase voltages is well known, employing both converters based on the SC principle have been presented in the
direct and indirect converter approaches. These converters can literature [20]–[22], where it has been verified for the first time
allow the control of the output voltages, and they can also be that the SC principle can be used for direct ac–ac conversion.
bidirectional [1]. Although the output voltage can be controlled This paper proposes a three-phase bidirectional direct ac–ac
with these converters, the complexity is very high, and depend- converter based on the SC principle, which has the potential
ing on the application, open-loop operation is sufficient. to replace the autotransformers in many applications. The con-
Converters based on switched capacitors (SCs) have been an verter acts as a multiplier (step-up) or a divider (step-down),
important research topic for many years. In the initial appli- and thus, the output voltage waveform follows the shape of
cations in power electronics, the SC was used in dc–dc static the input voltage, without changing the frequency. The con-
power conversion for low power [2]. These circuits have been verter operates in open loop, and therefore, it does not need
a control system. It provides fixed gain and also out-of-phase
output voltages, defined through the legs and load connections,
Manuscript received February 12, 2014; revised May 1, 2014; ac-
cepted May 28, 2014. Date of publication July 8, 2014; date of current which can be delta–delta, delta–wye, wye–delta, or wye–wye
version January 7, 2015. (similarly to an autotransformer). A detailed analysis of the pro-
T. B. Lazzarin and I. Barbi are with the Department of the Electri- posed converter, equivalent circuits, connection types, design
cal Engineering (EEL), Federal University of Santa Catarina (UFSC),
88040-970 Florianopolis, Brazil (e-mail: [email protected]; ivobarbi@ methodology, and experimental results is reported herein.
inep.ufsc.br).
R. L. Andersen is with the Department of Electrical Engineering II. P ROPOSED T HREE -P HASE B IDIRECTIONAL
(DEE), Federal University of Paraíba (UFPB), 58051-900 João Pessoa,
Brazil (e-mail: [email protected]). SC AC–AC C ONVERTER
Color versions of one or more of the figures in this paper are available
online at http://ieeexplore.ieee.org. The proposed three-phase SC ac–ac converter is introduced
Digital Object Identifier 10.1109/TIE.2014.2336625 in Fig. 1(a). It is composed of three modules, one per phase,

0278-0046 © 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Ivo Barbi. Downloaded on June 20,2022 at 02:22:16 UTC from IEEE Xplore. Restrictions apply.
736 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 62, NO. 2, FEBRUARY 2015

Fig. 1. Proposed three-phase bidirectional SC ac–ac converter in step-down configuration. (b) Gate drive signals and (c) practical implementation
of a bidirectional switch using two MOSFETs and its model.

and the module that comprises phase A is highlighted in Fig. 1.

A module has four bidirectional switches and three capacitors,
and thus, the total circuit has 12 bidirectional switches and nine
capacitors represented from S1 to S12 and from C1 to C9 in
Fig. 1(a), respectively. Bidirectional switches are implemented
in the converter employing two conventional MOSFETs, and
the details of the practical implementation and a model for the
bidirectional switches are presented in Fig. 1(c). The diodes
in antiparallel with the transistors in Fig. 1(a) and (c) are the
intrinsic diodes of the MOSFETs.
The input voltages (ABC) are connected at points 1, 4, and
7, and the output voltages (RST ) are obtained at points 2,
5, and 8. The three modules are linked through delta or wye
connections that are made employing terminals 1–3, 4–6, and
7–9. Fig. 1(a) shows an example with the connection in the
delta configuration. In each switching period, the SCs C1 , C4 ,
and C7 are first connected to upper capacitors (C2 , C5 , and C8 )
and then to lower capacitors (C3 , C6 , and C9 ), which keeps
the balance of voltages across terminals 2–3, 5–6, and 8–9
at one half of the voltage applied across terminals 1–3, 4–6,
and 7–9.
The proposed converter works in open loop, and the genera-
tion of the gate drive signals is simple due to the fact that the
same signal is applied in all odd switches and its complement
is applied in all even switches, as shown in Fig. 1(b). The duty
cycle is fixed and set close to 0.5. The frequency is also fixed,
and a deadtime has to be used between the complementary
signals from the odd and even switches.
Fig. 2 provides details on how module A (a phase from the Fig. 2. Details of module A (a phase) of the proposed three-phase
ac–ac converter. (a) Topological stage 1: Odd switches are ON.
converter) works. In Fig. 2(a), the odd switches are ON, and (a) Topological stage 2: Even switches are ON. (c) Topological stage 3:
thus, the C1 SC is connected to the capacitor C2 . In Fig. 2(b), Deadtime. (d) Proposed symbol for a module.

Authorized licensed use limited to: Ivo Barbi. Downloaded on June 20,2022 at 02:22:16 UTC from IEEE Xplore. Restrictions apply.
LAZZARIN et al.: SWITCHED-CAPACITOR THREE-PHASE AC–AC CONVERTER 737

TABLE I
C ONNECTIONS , P HASOR R EPRESENTATION , AND O UTPUT VOLTAGES

the even switches are ON, and thus, the C1 SC is connected

to the capacitor C3 . Between the changes of the topological
stages, there is a deadtime, and all switches are OFF, as
illustrated in Fig. 2(c). The C1 SC equalizes the voltages in
capacitors C2 and C3 , which ensures V23 = V13 /2, as shown
in Fig. 2(c). The other two modules have the same operation;
thus, V56 = V46 /2, and V89 = V79 /2. The symbol proposed to
represent a module can be seen in Fig. 2(d) and is employed
in Table I.
The proposed converter, which is shown in Fig. 1(a), is
configured as step-down, and it works as a voltage divider.
The main theoretical waveforms of the proposed three-phase
ac–ac converter are shown in Fig. 3, where the input voltages
of the modules are shown in Fig. 3(a), the output voltages
in each module are given in Fig. 3(b), the voltage across the
capacitors can be seen in Fig. 3(c), and the voltage across a
switch of module A is shown in Fig. 3(d). The waveforms
show that both the output voltages and voltage stresses on
the capacitors and on the switches are one-half of the input
voltages.
The voltages at points 2, 5, and 8 are called VR , VS , and VT ,
and the voltages obtained between these points are the output
line-to-line voltages of the converter in Fig. 1(a). This system
can feed three-phase loads in delta or wye configurations.
The proposed topology is a bidirectional converter. It is Fig. 3. Theoretical waveforms of the proposed three-phase ac–ac
a step-down converter, and it divides the input voltages by converter. (a) Input voltages in each module in step-down configura-
tion. (b) Output voltages in each module in step-down configuration.
two when configured, as shown in Fig. 1(a). However, it also (c) Voltage across the capacitors and (d) voltage across the switches
operates in a step-up configuration and multiplies the input of module A.
voltages by two. The only difference here is that the points
III. C ONNECTIONS OF P ROPOSED
where the source and the load are connected have to be inverted.
T HREE -P HASE C ONVERTER
Therefore, the input voltages are applied at connection points 2,
5, and 8, and the output voltages are obtained at points 1, 4, and The modular characteristic of the proposed circuit allows
7. When working as a step-up converter, it also maintains the that the legs and the loads can be either delta or wye con-
theoretical waveforms, as shown in Fig. 3. nected. Therefore, depending on the way that the legs and the

Authorized licensed use limited to: Ivo Barbi. Downloaded on June 20,2022 at 02:22:16 UTC from IEEE Xplore. Restrictions apply.
738 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 62, NO. 2, FEBRUARY 2015

Fig. 5. Proposed per-phase equivalent circuit.

losses caused by the intrinsic capacitance of the MOSFETs. The

Fig. 4. Proposed equivalent circuit for delta–wye connection.
equivalent capacitances Ceq process reactive power equivalent
load are connected, four possibilities are presented: delta–wye, to the original circuit. Conduction losses are represented by the
delta–delta, wye–delta, and wye–wye, as shown in Table I. series resistances Req . Finally, the load is represented by the
Each of these connections results in a particular output voltage resistances Ro which are connected according to the case.
and phase in relation to the grid voltage (similarly to the gain The parameters Rsl , Req , Ceq , and τ for these equivalent
and phase shift presented in a three-phase transformer). All circuits can be calculated by
connections are described in detail in Table I, which allows the 1
gain and phase shift in each connection to be previously known. Rsl = (1)
4 · COSS · fs
It is important to note that there is gain and there is no phase  
−1
shift in the voltages within a module. For example, in module A, 1 − e fs ·τ
1
the voltage between points 2 and 3 is in phase with the voltage Req = ·  −(D)   (2)
between points 1 and 3 and with 1/2 of the gain (as shown in 4 · fs · C −(1−D)
1 − e fs ·τ + e fs ·τ
1
+ e fs ·τ
[21]). The phase shifts and different gains are related to the delta
or wye connections. Ceq = 3 · C (3)
The load can also be connected in the delta or wye configura-
τ = 2 · Ron · C (4)
tion in both cases, and the output voltages are given in Table I.
The table also shows the phasor representation of the voltages where COSS is the intrinsic output capacitance from a
to elucidate their composition. MOSFET, fs is the switching frequency of the circuit, Ron is
The proposed converter is fed by line-to-line voltages at the conduction resistance of a bidirectional switch and its values
points A, B, and C. These input sources can also be connected are twice the resistance RDS(on) from a MOSFET, C is the
either in delta or wye configurations. capacitance from C1 to C9 (considering all capacitors equal),
Connections with the load connected in the wye configu- and τ is the time constant of the SC. Details on the derivation
ration and without the neutral point connected to the legs, as of these equations can be found in [20] and [21].
in the case of delta–wye or wye–wye configurations when the The proposed converter does not have inductors connected
load neutral is left floating, suffer from voltage unbalance in between the input and the output, as in a conventional converter.
the case of unbalanced load. The voltages in each phase will In the equivalent circuits shown in Fig. 4, only the equivalent
be dependent on the current of that phase, causing unbalance. resistance Req is connected between the input voltages and the
This problem does not occur if the load is connected in the loads. The Req value is defined by (2), and it is determined
delta configuration or if the load neutral is connected to the grid through fs , C, Ron , and D. Therefore, the appropriate choice
neutral in the wye–wye case. It is also important to note that of these parameters can supply a low value to Req , which can
this problem is not specific to this circuit and also occurs when provide a small drop voltage across Req and, thus, good regu-
a conventional autotransformer is used. lation of the output voltages. This is an important characteristic
The proposed circuit can be represented by equivalent cir- for the proposed converter due to its operation in open loop.
cuits depending on the connection chosen in Table I. The The knowledge of the three-phase equivalent circuits pre-
equivalent circuits seen by the load side for the delta–wye sented in Fig. 4 and obtaining their parameters can lead to a
connection is presented in Fig. 4. synthesized per-phase equivalent circuit, which can be used for
all cases. This per-phase equivalent circuit processes one-third
IV. Q UANTITATIVE A NALYSIS of the total output power, and its parameters can be calculated
from the same equations presented previously multiplied by a
A. Equivalent Circuit
factor which is dependent on the connection.
For simplicity, the system is considered balanced in the The per-phase equivalent circuit is shown in Fig. 5. The
equivalent circuits. The grid seen by the load side is represented conduction losses are represented by the series resistance Rs ,
by three sinusoidal voltage sources connected in wye. There- the switching losses caused by the intrinsic capacitances of the
fore, the line-to-neutral voltage from input sources is vLN /2 MOSFETs are represented by the parallel resistance Rp , and
(half the line-to-neutral voltage from the grid), and the line-to- the reactive power is represented by the parallel capacitance
line voltage from input sources is vLL /2 (half the line-to-line Cp . The factors used to determine these parameters, the voltage
voltage from the grid). The resistances Rsl change according of the model source vm , and the resistance of the load Ro
to the connection in each case and represent the switching according to the connection are shown in Table II.

Authorized licensed use limited to: Ivo Barbi. Downloaded on June 20,2022 at 02:22:16 UTC from IEEE Xplore. Restrictions apply.
LAZZARIN et al.: SWITCHED-CAPACITOR THREE-PHASE AC–AC CONVERTER 739

TABLE II An important step during the design process is to determine

C OMPONENTS OF E QUIVALENT C IRCUIT
the boundary between the CC and PC modes. A capacitor
charges or discharges totally (around 99.3%) in five times the
time constant (5τ ); thus, the charge/discharge time of the SCs
has to be lower than 5τ to avoid the CC mode. Therefore, the
maximum switching period (Ts ) is given by

D · Ts ≤ 5 · τ (5)
(1 − D) · Ts ≤ 5 · τ. (6)

Considering the duty cycle of 0.5, a minimum fs τ that

ensures the PC mode is obtained by
TABLE III
E QUIVALENT C IRCUIT E QUATIONS
fs · τ ≥ 0.1 (7)

and its critical value (boundary between CC and PC modes) is

defined by

fs τcrit = 0.1 (8)

where fs is the switching frequency. Thus, the SCs have to be de-

signed with a fs τ value higher than 0.1 to operate in PC mode.

V. D ESIGN E XAMPLE
A. Specification
The delta–wye connection was chosen to verify the proposed
three-phase ac–ac converter. Hence, design specifications were
defined with the objective of simulating the converter and
implementing a prototype as follows:
1) Po = 6000 W output power;
2) vL = 380 V input line-to-line voltage;
3) vo = 110 V output line-to-neutral voltage;
4) fi = 60 Hz frequency of ac voltage;
5) η > 95% expected efficiency;
6) P F > 0.92 input power factor (capacitive).
The proposed ac–ac three-phase converter is designed to
Analysis of the equivalent circuit allows simple equations operate in PC mode, with low equivalent resistance (Req ) and
to be found, which are employed to examine and design the high efficiency. This requires that fs τ ≥ fs τcrit and D ∼ = 0.5.
proposed three-phase ac–ac converter. These equations are Thus, in an appropriate design, the duty cycle is set close to
shown in Table III, and they provide important information on 0.5, the capacitor is calculated according to the desired power
the three-phase ac–ac converter. The input power, output power, factor, the switches chosen have a low conduction resistance
and loss equations have been previously multiplied by three to (RDSon ), and the switching frequency is defined so as to
represent the entire circuit based on the per-phase version. The maintain fs τ higher than fs τcrit and to obtain a high efficiency.
variable f in the equations is the frequency of the input voltage. Furthermore, the switching frequency (fs ) has to be feasible for
practical implementation. The design methodology is based on
the equations given in Table III, and the component values are
B. Operation Modes of SC Converter shown in Table VI for a delta–wye connection. In the following
A SC converter can operate in three different modes, which sections, the choice of the components and switching frequency
are defined by the charge of the SC. These modes are complete and an analysis of the results are discussed.
charge (CC), partial charge (PC), and no charge (NC) [17],
[18]. The operation in PC and NC modes leads to both higher B. Capacitance Calculation
efficiency of the converter and lower current peaks in the The capacitors C1 to C9 are calculated to attend the specifi-
capacitors and switches. Therefore, these operation modes are cation of the power factor. Hence, based on the equations for the
preferred. reactive power and the power factor in Table III, the maximum
The NC mode usually requires a high operation frequency capacitance Cp is defined by
and large capacitances, and the advantages in relation to the PC 
mode are not significant. Hence, the suggested mode for the 1 Pi2
Cp ≤ · − Pi2 ≤ 95 μF. (9)
SCs of the converter in Fig. 1 is the PC mode [21]. 6 · v m · π · fs
2 F p2

Authorized licensed use limited to: Ivo Barbi. Downloaded on June 20,2022 at 02:22:16 UTC from IEEE Xplore. Restrictions apply.
740 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 62, NO. 2, FEBRUARY 2015

The equivalent capacitance Ceq shown in Table II for the

delta–wye connection is

Cp 1 Pi2
Ceq = ≤ 9 2 · − Pi2 ≤ 31.7 μF. (10)
3 2 · v L · π · f s F p 2

Using (3), the maximum capacitance value for capacitors C1

to C9 is

Ceq 1 Pi2
C= = 27 2 · − Pi2 ≤ 10.6 μF. (11)
2 · v L · π · fs
3 F p2

Therefore, a 9-μF/350 VAC film capacitor (polypropylene)

was chosen for capacitors C1 to C9 (all capacitors have the
same value). These capacitors have to be nonpolarized due to
the ac voltage applied on them. The expected power factor in
full load is 0.98.
Fig. 6. Efficiency versus switching frequency.
C. Switches TABLE IV
M AIN PARAMETERS OF E QUIVALENT C IRCUITS
The bidirectional switches are implemented employing
MOSFETs, as shown in Fig. 1. These switches are chosen with
the purpose of obtaining an appropriate fs τ value and a low
conduction resistance. Thus, the MOSFET CoolMOS IXKH
70N60C5, which has an RDS(on) value of 45 mΩ at 25 ◦ C (or
81 mΩ at 100 ◦ C), was selected. Its maximum drain–source
voltage is 600 V, and its drain current is 70 A at 25 ◦ C (or 48 A
at 90 ◦ C).
The total conduction resistance of a bidirectional switch
(Ron ) is twice the resistance of a MOSFET, and therefore, the
resistance Ron is 162 mΩ at 100 ◦ C.

D. Switching Frequency
The choice of switching frequency defines the fs τ value of
the converter and, consequently, the values of Req , Rs , and η.
The aim is to select the switching frequency which will provide
a high efficiency. Therefore, by applying the values of C and
Ron in the efficiency equation in Table III, the theoretical
efficiency of the ac–ac converter can be calculated for the
switching frequency selected, as shown in Fig. 6. This curve
demonstrates that the converter is most efficient between 80 and
110 kHz. Furthermore, employing the values of C and Ron in
(8), the minimum frequency for the converter to operate in PC
mode is given by
Fig. 7. Photograph of the prototype (specific power: 1.35 kW/kg).
0.1
fscrit = = 34.3 kHz. (12) VI. I MPLEMENTATION AND E XPERIMENTAL R ESULTS
2 · Ron · C
A. Verification of the Proposed Converter
This value verifies that the converter must operate in PC
mode to obtain a high efficiency. Thus, the switching frequency A prototype of 38 0 V/110 V/6 kW was built to verify the
selected was 100 kHz, as identified in the figure. At this point, operation of the proposed converter (see photograph in Fig. 7).
the expected efficiency is 97.1%. The main specifications and components used in the prototype
are presented in Table V. The specific power obtained for the
prototype was 1.35 kW/kg.
E. Parameters of Equivalent Circuits
The experiments were carried out with D = 0.4 (10% of
The definition of the values of C, Ron , D, and fs allows the deadtime), and the converter was fed by an ac power supply. As
parameters for the equivalent circuits to be calculated, as shown the converter provides a significant level of power, the inductive
in Table IV. input filter was added. The frequency of the current ripple was

Authorized licensed use limited to: Ivo Barbi. Downloaded on June 20,2022 at 02:22:16 UTC from IEEE Xplore. Restrictions apply.
LAZZARIN et al.: SWITCHED-CAPACITOR THREE-PHASE AC–AC CONVERTER 741

TABLE V
M AIN S PECIFICATIONS AND C OMPONENTS OF THE P ROTOTYPE

Fig. 9. Proposed converter with modules in wye configuration: Input

line-to-line voltage VAB , output line-to-line voltage VRS , and output line-
to-neutral voltage VRN .

Fig. 8. Proposed converter with modules in delta configuration: Input

line-to-line voltage VAB , output line-to-line voltage VRS , and output line-
to-neutral voltage VRN .
Fig. 10. Proposed converter in delta–wye configuration, with no load
and with 380 V @ 60 Hz of input line-to-line voltage: Input line-to-line
high (100 kHz), and thus, a small filter of 5 μH was sufficient to voltage VAB and output line-to-neutral voltages VRN , VSN , and VT N .
filter the input current. In most cases, this filter can be the line
inductance itself. shows peaks of 150 V for the input voltage VAB and 75 V
The possible connections that can be implemented in the and 43 V for the output voltages VRS and VRN , respectively.
proposed converter (presented in Table I) were verified in The converter does not introduce a phase shift in VRS , and it
the prototype. Considering the converter with the modules in introduces a 30◦ phase shift in VRN .
delta connection, it supplies a line-to-line voltage to a delta- The subsequent experiments were carried out with the con-
connected load with a 1/2 gain and a 60◦ phase shift (in verter in the delta–wye configuration (modules in delta and load
advance) and √a line-to-neutral voltage to a wye-connected load in wye) and with nominal voltage. Figs. 10 and 11 demonstrate
with a 1/(2 3) gain and a 90◦ phase shift (in advance), with the converter operation when it is fed by a 380 V @ 60 Hz
both cases relating to the input line-to-line voltage. The voltages input line-to-line voltage. Three output line-to-neutral voltages
of the converter are shown in Fig. 8 to verify this operation (VRN , VSN , and VT N ) and one input line-to-line voltage (VAB )
mode, where the peak value of the input voltage VAB (line- can be seen in Fig. 10. The waveforms show that the√proposed
to-line) is 150 V and the output voltages VRS (line-to-line) converter supplies 110 V (rms) at the outputs (1/(2 3) gain)
and VRN (line-to-neutral) are 75 and 43 V, respectively. The and introduces a 90◦ phase shift, as expected.
converter introduces a 60◦ phase shift in VRS and a 90◦ phase The output voltages of the converter under a balanced three-
shift in VRN , as proposed. phase resistive load and the current in phase R are shown in
If the converter input has a wye connection, it provides a Fig. 11. The value for the line-to-neutral voltage was 106.2 V
line-to-line voltage to a delta-connected load with a 1/2 gain (96.57% of regulation). The waveforms were sinusoidal even in
and without any phase shift and √ a line-to-neutral voltage to a rated power. Furthermore, the measurement of total harmonic
wye-connected load with a 1/(2 3) gain and 30◦ phase shift distortion (THD) was 0.3%, verifying that the converter did not
(in advance). This operation mode is seen in Fig. 9, which also introduce distortion in the voltages.

Authorized licensed use limited to: Ivo Barbi. Downloaded on June 20,2022 at 02:22:16 UTC from IEEE Xplore. Restrictions apply.
742 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 62, NO. 2, FEBRUARY 2015

Fig. 11. Proposed converter in delta–wye configuration, with resistive Fig. 13. Proposed converter in delta–wye configuration, with no load
load and with 380 V @ 60 Hz of input line-to-line voltage: Output line-to- and with 380 V @ 16.66 Hz of input line-to-line voltage: Input line-to-line
neutral voltages VRN , VSN , and VT N and line-to-neutral current IRN . voltage VAB and output line-to-neutral voltages VRN , VSN , and VT N .

Fig. 12. Line-to-line voltage VAB and line current IA in one input of
Fig. 14. Proposed converter in delta–wye configuration, with no load
the proposed converter.
and with 380 V @ 400 Hz of input line-to-line voltage: Output line-to-
neutral voltages VRN , VSN , and VT N .
The line input current IA and the input line-to-line voltage
VAB are seen in Fig. 12. The power factor for rated power was
0.969. The line voltage leads from the line current at 1 ms
(approximately 21.6◦ ), which is expected for the capacitive
circuit (in a resistive circuit, the line voltage leads from the line
current at 1.36 ms or 30◦ ). The input current of the proposed
topology presents a ripple at 100 kHz (characteristic buck), and
thus, a small L filter of 5 μH was added, which was sufficient to
filter the high frequency and provided a sinusoidal waveform.
The input current THD was 179% without the L filter (value
typical in one converter with a characteristic buck), and it was
6.6% with the L filter.
The converter was tested at different frequencies to verify
this characteristic, and some results are shown in Figs. 13 and
14. These waveforms show the input voltage VAB and the out- Fig. 15. Proposed converter in delta–wye configuration, with 380 V
put voltages VRN , VSN , and VT N for input voltages of 380 V @ @ 60 Hz of input line-to-line voltage and a balanced resistive load:
Efficiency test.
16.66 Hz (see Fig. 13) and 380 V @ 400 Hz (see Fig. 14). The
efficiency and the static regulation were the same for all fre- Experimental curves were obtained for the efficiency, regu-
quencies, and the power factor decreased with an increase in the lation, and power factor, all of them with a balanced resistive
frequency. These results were as expected and were in agree- load. The experimental and theoretical efficiency curves are
ment with the equivalent circuit of the converter seen in Fig. 5. plotted in Fig. 15, and the experimental curve shows that,

Authorized licensed use limited to: Ivo Barbi. Downloaded on June 20,2022 at 02:22:16 UTC from IEEE Xplore. Restrictions apply.
LAZZARIN et al.: SWITCHED-CAPACITOR THREE-PHASE AC–AC CONVERTER 743

Fig. 16. Proposed converter in delta–wye configuration, with 380 V Fig. 17. Proposed converter in delta–wye configuration, with 380 V @
@ 60 Hz of input line-to-line voltage and a balanced resistive load: 60 Hz of input line-to-line voltage and a balanced resistive load: Output
Efficiency test in rated power for different switching frequencies. voltage regulation test.

for a wide load range, the efficiency is higher than 96%.

The experimental efficiency at rated power was 96.3%, and
the theoretical expected efficiency was 97.1%. This difference
can be explained by considering the wires and connections of
the experimental setup since their resistances were not taken
into account in the theoretical analysis, and consequently, the
observed losses were greater than those considered in the
calculations. In the efficiency curves in Fig. 15, a decrease
can be observed when the load becomes smaller, which is due
to the constant loss in the converter caused by the charging
and discharging of the intrinsic capacitances of the MOSFETs.
This loss is addressed in [21] and described in Table III as
MOSFET switching losses. The expected theoretical value was
120 W, and the measured experimental value was 121.5 W, Fig. 18. Proposed converter in delta–wye configuration, with 380 V @
60 Hz of input line-to-line voltage and a balanced resistive load: Input
which represents 2% of the rated power from the converter. power factor test (curve is for a leading power factor).
This value is high due to the high intrinsic capacitance of
the CoolMOS power MOSFET chosen to implement the The experimental input power factor curve is shown in
bidirectional switch. Fig. 18 for an input voltage of 60 Hz and a resistive load.
A curve of the experimental efficiency versus switching The curve relates to a leading power factor, which is close to
frequency was also obtained, in which the points tested were one at the rated power (98%) and drops considerably for low
10, 25, 50, 75, and 100 kHz, all of them for rated power. The power. This characteristic is a consequence of the input source
test was limited to 100 kHz due to the technology of gate seeing the converter as a constant capacitive load. Therefore,
driver circuits. The experimental and theoretical curves (see the converter contributes to the system power factor with the
Fig. 6) are shown together in Fig. 16. On the experimental leading power factor (capacitive), and the value depends on the
efficiency curve, the values are lower than those on the the- capacitors employed in the power circuit.
oretical efficiency curve by around 1%, as expected, due to
resistances that were not taken into account in the theoretical
B. Comparison Between the Proposed Converter and
analysis. The graphs show that the efficiency decreases when
Other Solutions
the switching frequency decreases. This behavior is observed
because the converter starts to operate either close to or in the A new direct ac–ac power converter based on SC is proposed
complete charge/discharge mode, where its efficiency is lower. in this paper, which changes the amplitude with no change
The experimental curve also verifies that the chosen switching in frequency. It is therefore suitable for building a solid-state
frequency (100 kHz) achieves the maximum efficiency. autotransformer and thus represents an option to replace the
The experimental output regulation curve can be seen in conventional three-phase autotransformer. For this reason, a
Fig. 17, and it shows that a voltage drop of less than 2.2% quantitative comparison between the tested prototype and a
(3.8 V) was observed in the rated power compared to the no- 6-kVA autotransformer is shown in Table VI.
load output voltage (110 V). Therefore, good regulation of the The analysis showed that the proposed converter has a better
load voltages was obtained, even though this converter operates performance than the autotransformer. It presents a small gain
in open loop. This characteristic was provided due to the low in efficiency, regulation, and THD. The input power factor is
series resistance (Req ) of the proposed converter, which, in this smaller than that in the autotransformer; however, it remains ap-
design, was 0.4 Ω, as shown in Table IV. propriate and is a capacitive power factor. The main advantages

Authorized licensed use limited to: Ivo Barbi. Downloaded on June 20,2022 at 02:22:16 UTC from IEEE Xplore. Restrictions apply.
744 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 62, NO. 2, FEBRUARY 2015

TABLE VI TABLE VIII

Q UANTITATIVE C OMPARISON B ETWEEN THE P ROPOSED Q UALITATIVE C OMPARISON B ETWEEN THE
C ONVERTER AND AN E QUIVALENT AUTOTRANSFORMER P ROPOSED C ONVERTER AND THE CMC

TABLE VII
Q UANTITATIVE C OMPARISON B ETWEEN THE
P ROPOSED C ONVERTER AND THE CMC

The quantitative analysis in Table VII demonstrates a state of

balance. However, the use of a larger number of switches repre-
sents a disadvantage in the case of the proposed converter in this
comparison. The qualitative analysis in Table VIII highlights
two different points: If the goal is fixed frequency, simplicity,
and a magnetic-less converter, the proposed converter will
have advantages; if the goal is variable frequency, regulation,
and control, the CMC will have advantages. Therefore, for
certain applications, for instance, to replace an autotransformer,
the proposed converter based on SC appears to represent a
good option.

are related to the weight, volume, and absence of magnetic VII. C ONCLUSION
elements in the power circuit, which provides high gains in
specific power and density power. One disadvantage is the cost A three-phase magnetic-less ac–ac static power converter
since, currently, an autotransformer costs around 0.32 p.u. of based on the SC principle has been proposed and studied.
the total cost of the prototype. However, there is a tendency for The following conclusions can be drawn from the theoretical
the cost of copper to increase and of silicon to decrease. Thus, analysis and the experimental results reported in this paper.
the proposed converter is an attractive option to replace the 1) The proposed static power converter employs only capac-
autotransformer in applications where the weight and volume itors and switches.
have to be reduced. 2) The converter is able to supply power to any kind of load,
In the literature, there are attractive options for ac–ac power balanced or not, of any power factor, including nonlinear
converters that could perform the function of the proposed load.
converter, with the most traditional being the matrix converter 3) The converter operates in open loop with constant switch-
(MC). The MCs started with the conventional MC (CMC) pro- ing frequency and duty cycle.
posed by Venturini and Alesina in 1980 [23], which is a direct 4) The signal power stage is simple and inexpensive.
ac–ac converter. The most important characteristics of MCs are 5) The converter is bidirectional, and thus, it can operate as
as follows [23]–[25]: 1) generation of load voltage with arbi- a step-down converter (static gain of 0.5) or a step-up
trary amplitude and frequency; 2) a simple and compact power converter (static gain of 2), and in both modes, only one-
circuit; 3) sinusoidal input and output currents; 4) operation half of the high-side voltage is applied to the switches and
with unity power factor; and 5) regeneration capability. MCs the capacitors.
are usually employed in applications where the generation of 6) The circuit can operate within a wide range of frequen-
an output voltage with arbitrary frequency is required; how- cies, including dc voltage.
ever, they could also be employed for constant frequency. The 7) The proposed ac–ac converter offers fixed gain, and it
proposed converter has a different conception and operation does not change frequency; thus, it appears to represent
principle in comparison to MCs, but nevertheless, as the CMC a potential solution to replace the conventional three-
presents a similar structure (number and type of components), phase autotransformer in the range of 0.5–100 kVA
quantitative and qualitative comparisons between the proposed and can be considered as a magnetic-less solid-state
converter and the CMC are shown in Tables VII and VIII. autotransformer.

Authorized licensed use limited to: Ivo Barbi. Downloaded on June 20,2022 at 02:22:16 UTC from IEEE Xplore. Restrictions apply.
LAZZARIN et al.: SWITCHED-CAPACITOR THREE-PHASE AC–AC CONVERTER 745

R EFERENCES [22] T. B. Lazzarin, M. P. Moccelini, and I. Barbi, “Direct buck-type ac/ac con-
verter based on switched-capacitor,” in Proc. Brazilian Power Electron.
[1] J. W. Kolar, T. Friedli, J. Rodriguez, and P. W. Wheeler, “Review of three-
Conf., 2013, vol. 1, pp. 1–6.
phase PWM ac-ac converter topologies,” IEEE Trans. Ind. Electron.,
[23] M. Venturini, “A new sine wave in sine wave out, conversion tech-
vol. 58, no. 11, pp. 4988–5006, Nov. 2011.
nique which eliminates reactive elements,” in Proc. Powercon 7, 1980,
[2] A. Ioinovici, “Switched-capacitor power electronics circuits,” IEEE
pp. E3/1–E3/15.
Circuits Syst. Mag., vol. 1, no. 3, pp. 37–42, 2001.
[24] J. Rodriguez et al., “Matrix converter controlled with the direct transfer
[3] Y.-S. Hwang, H.-H. Chou, and J.-J. Chen, “A high-efficiency dc-dc con- function approach: Analysis, modelling and simulation,” Int. J. Electron.,
verter with wide output range using switched-capacitor front-end tech-
vol. 92, no. 2, pp. 63–85, Feb. 2005.
niques,” IEEE Trans. Ind. Electron., vol. 61, no. 5, pp. 2244–2251,
[25] J. Rodriguez, M. Rivera, J. W. Kolar, and P. W. Wheeler, “A review of
May 2014.
control and modulation methods for matrix converters,” IEEE Trans. Ind.
[4] G. Villar-Pique, H. J. Bergveld, and E. Alarcon, “Survey and benchmark Electron., vol. 59, no. 1, pp. 58–70, Jan. 2012.
of fully integrated switching power converters: Switched-capacitor ver-
[26] S. Bernet, S. Ponnaluri, and R. Teichmann, “Design and loss comparison
sus inductive approach,” IEEE Trans. Power Electron., vol. 28, no. 9,
of matrix converters voltage-source converters for modern ac drives,”
pp. 4156–4167, Sep. 2013.
IEEE Trans. Ind. Electron., vol. 49, no. 2, pp. 304–314, Apr. 2002.
[5] J. T. Stauth, M. D. Seeman, and K. Kesarwani, “Resonant switched- [27] F. Schafmeister, C. Rytz, and J. W. Kolar, “Analytical calculation of the
capacitor converters for sub-module distributed photovoltaic power man-
conduction and switching losses of the conventional matrix converter and
agement,” IEEE Trans. Power Electron., vol. 28, no. 3, pp. 1189–1198,
the (very) sparse matrix converter,” in Proc. 30th IEEE APEC, 2005,
Mar. 2013.
vol. 2, pp. 875–881.
[6] M. Kim, C. Kim, J. Kim, and G. Moon, “A chain structure of switched
capacitor for improved cell balancing speed of lithium-ion batteries,”
IEEE Trans. Ind. Electron., vol. 61, no. 8, pp. 3989–3999, Aug. 2014.
[7] Y. Hinago and H. Koizumi, “A switched-capacitor inverter using series/
Telles B. Lazzarin (S’09–M’12) was born in
parallel conversion with inductive load,” IEEE Trans. Ind. Electron.,
Criciúma, Brazil, in 1979. He received the B.Sc.,
vol. 59, no. 2, pp. 878–887, Feb. 2012.
M.Sc., and Ph.D. degrees in electrical engi-
[8] A. Tsunoda, Y. Hinago, and H. Koizumi, “Level and phase shifted
neering from the Federal University of Santa
PWM for seven-level switched-capacitor inverter using series/parallel
Catarina (UFSC), Florianópolis, Brazil, in 2004,
conversion,” IEEE Trans. Ind. Electron., vol. 61, no. 8, pp. 4011–4021,
2006, and 2010, respectively.
Aug. 2014.
He is currently an Adjunct Professor with the
[9] E. Babaei and S. S. Gowgani, “Hybrid multilevel inverter using switched-
UFSC. In 2007, he was a Research Assistant
capacitor units,” IEEE Trans. Ind. Electron., vol. 61, no. 9, pp. 4614–4621,
for the University Center of Jaragua do Sul
Sep. 2014.
(UNERJ), Brazil. From 2010 to 2011, he was
[10] Y.-P. Hsieh, J.-F. Chen, T.-J. Liang, and L.-S. Yang, “Novel high step-
a Postdoctoral Fellow at the Power Electronics
up dc-dc converter with coupled-inductor and switched-capacitor tech-
Institute (INEP), UFSC. From 2011 to 2013, he was a Professor at the
niques,” IEEE Trans. Ind. Electron., vol. 59, no. 2, pp. 998–1007,
Federal Institute of Santa Catarina (IFSC), Florianópolis. His research
Feb. 2012.
interests include inverters, parallel operation of inverters, uninterruptible
[11] H. Liangzong, “High-performance bridge modular switched-capacitor
power supplies, high-voltage dc–dc converters, direct ac–ac power con-
converter with small component requirement based on output impedance
verters, switched capacitor (SC) converters, and hybrid SC converters.
analysis for low loss,” IEEE Trans. Power Electron., vol. 28, no. 10,
Dr. Lazzarin is a member of the Brazilian Power Electronics Society
pp. 4668–4680, Oct. 2013.
(SOBRAEP), IEEE Power Electronics Society, and IEEE Industrial Elec-
[12] D. Siemaszko, C. Rod, and A. C. Rufer, “Single-phase resonant LC circuit
tronics Society.
using a bank of self-switched capacitors,” IEEE Trans. Ind. Electron.,
vol. 58, no. 9, pp. 4175–4184, Sep. 2011.
[13] S. Kiratipongvoot, S.-C. Tan, and A. Ioinovici, “Phase-shift interleaving
control of variable-phase switched-capacitor converters,” IEEE Trans. Romero L. Andersen (M’11) was born in
Ind. Electron., vol. 60, no. 12, pp. 5575–5584, Dec. 2013. Florianópolis, Brazil, in 1980. He received the
[14] J. W. Kimball, P. T. Krein, and K. R. Cahill, “Modeling of capacitor B.S., M.S., and Ph.D. degrees in electrical en-
impedance in switching converters,” IEEE Power Electron. Lett., vol. 3, gineering from the Federal University of Santa
no. 4, pp. 136–140, Dec. 2005. Catarina, Florianópolis, in 2003, 2006, and
[15] C.-K. Cheung, S.-C. Tan, C. K. Tse, and A. Ioinovici, “On energy effi- 2010, respectively.
ciency of switched-capacitor converters,” IEEE Trans. Power Electron., He is currently a Professor with the Federal
vol. 28, no. 2, pp. 862–876, Feb. 2013. University of Paraíba, João Pessoa, Brazil. His
[16] M. Evzelman and S. Ben-Yaakov, “Average-current-based conduction research interests include dc–dc power con-
losses model of switched capacitor converters,” IEEE Trans. Power version, switched-capacitor converters, power
Electron., vol. 28, no. 7, pp. 3341–3352, Jul. 2013. converter modeling, and renewable energy
[17] S. Ben-Yaakov, “Behavioral average modeling and equivalent circuit sim- sources.
ulation of switched capacitors converters,” IEEE Trans. Power Electron.,
vol. 27, no. 2, pp. 632–636, Feb. 2012.
[18] S. Ben-Yaakov, “On the influence of switch resistances on switched-
capacitor converter losses,” IEEE Trans. Ind. Electron., vol. 59, no. 1, Ivo Barbi (M’78–SM’90–F’11) was born in
pp. 638–640, Jan. 2012. Gaspar, Brazil, in 1949. He received the B.S.
[19] L. Muller and J. Kimball, “A dynamic model of switched-capacitor power and M.S. degrees in electrical engineering
converters,” IEEE Trans. Power Electron., vol. 29, no. 4, pp. 1862–11869, from the Federal University of Santa Catarina
Apr. 2014. (UFSC), Florianopolis, Brazil, in 1973 and 1976,
[20] T. B. Lazzarin, R. L. Andersen, G. B. Martins, and I. Barbi, “A 600 W respectively, and the Dr. Ing. degree from the
switched-capacitor ac-ac converter for 220 V/110 V and 110 V/220 V Institut National Polytechnique de Toulouse,
applications,” IEEE Trans. Power Electron., vol. 27, no. 12, pp. 4821– France, in 1979.
4826, Dec. 2012. He is currently a Professor with the Power
[21] R. L. Andersen, T. B. Lazzarin, and I. Barbi, “A 1-kW step-up/step- Electronics Institute, UFSC.
down switched-capacitor ac-ac converter,” IEEE Trans. Power Electron., Prof. Barbi founded the Brazilian Power Elec-
vol. 28, no. 7, pp. 3329–3340, Jul. 2013. tronics Society and the Power Electronics Institute of the UFSC.

Authorized licensed use limited to: Ivo Barbi. Downloaded on June 20,2022 at 02:22:16 UTC from IEEE Xplore. Restrictions apply.