6408 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 30, NO.

11, NOVEMBER 2015

A Comparative Study of Power Supply Architectures

in Wireless EV Charging Systems
Bryan Esteban, Member, IEEE, Maher Sid-Ahmed, and Narayan C. Kar, Senior Member, IEEE

Abstract—This paper examines two of the primary power supply CS 1 Primary coil partial series tuning capacitor for
architectures being predominantly used for wireless electric vehicle LCL topology.
(EV) charging, namely the series LC (SLC) resonant and the hy- C1 Primary resonant tuning capacitor.
brid series–parallel (LCL) resonant full-bridge inverter topologies.
The study of both of these topologies is presented in the context CB HF transformer dc-blocking capacitor.
of designing a 3-kW primary-side controlled stationary wireless L2eq Equivalent secondary inductance when partial
EV charger with nominal operating parameters of 30-kHz center series compensation is used.
frequency, a range of coupling in the neighborhood of 0.18–0.26, L2eq Equivalent primary inductance when partial se-
and a parallel secondary pick-up with partial series coil compensa- ries compensation is used.
tion. A comparison of both architectures is made in terms of their
design methodology, physical size, cost, complexity, and efficiency. ω Operating angular frequency.
It is found that the SLC architecture is 2.45% less costly than the ω0 Resonant angular frequency.
LCL topology. On the other hand, it is observed that the LCL archi- VU Utility input voltage.
tecture achieves almost 10% higher peak efficiency at rated load VS Fundamental of inverter voltage.
and minimum coupling. The study also showed that the SLC topol- Zr Reflected secondary impedance.
ogy suffers from poor light load efficiency, while the LCL topology
maintains very high efficiency over its full range of coupling and Rr (M, Req ) Reflected secondary resistance.
loading. The study also revealed that the capacitor voltage stress Cr (M ) Reflected secondary capacitance.
is significantly higher in the SLC topology. Finally, it is also shown Q2t Overall tunned secondary quality factor.
that the control complexity of the SLC architecture is higher than Q2v Secondary voltage boost quality factor.
that of the LCL architecture because of its sensitivity to changes in Q2i Secondary current boost quality factor.
the reflected secondary impedance, which result in loss of constant
current source and ZVS operation unless a suitable combination Qu Unloaded magnetic pad quality factor.
of parameters are modulated by the closed-loop controller. I1 Primary current.
IB Inverter bridge current.
Index Terms—Inductive power transfer (IPT), LCL power sup-
IC 2 LCL parallel capacitor current.
ply, resonant inverter, SLC power supply, wireless EV charging. I2 Secondary current.
IC 2 Secondary parallel capacitor current.
Isc Secondary short-circuit current.
NOMENCLATURE IL Load current.
VLB LCL bridge inductor voltage.
RL Load resistance. VC 1 LCL parallel capacitor voltage or SLC series
Req Equivalent load resistance through rectifier in a capacitor voltage.
continuous-conduction mode. VC S 1 LCL partial series tuning capacitor voltage.
P Power. VL 1 Primary coil voltage.
Pm ax Maximum power. VL 2 Secondary coil voltage.
L1 Primary pad inductance. VC S 2 Partial series tuning capacitor voltage.
L2 Secondary pad inductance. VC 2 Secondary parallel capacitor voltage. VL 2oc
M Coupled-pad mutual inductance. VL 2oc Secondary coil open-circuit voltage.
k Magnetic coupling coefficient. XL 1 Effective reactance of primary coil with partial
C2 Secondary resonant tuning capacitor. series compensation.
CS 2 Secondary coil partial series tuning capacitor. tPS Phase-shift time delay between inverter legs A
and B.
td Dead time between series bridge semiconduc-
tors.
Manuscript received September 29, 2014; revised December 21, 2014 and ton Time duration of semiconductor conduction
May 3, 2015; accepted May 18, 2015. Date of publication June 8, 2015; date interval.
of current version July 10, 2015. Recommended for publication by Associate toff Time duration of semiconductor non-
Editor Y. J. Jang.
The authors are with the Department of Electrical and Computer Engineer- conduction interval.
ing, University of Windsor, Windsor, ON N9B 3P4, Canada (e-mail: bryan. Ts Switching time period.
[email protected]; [email protected]; [email protected]). N1 Isolation transformer number of primary turns.
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. N2 Isolation transformer number of secondary
Digital Object Identifier 10.1109/TPEL.2015.2440256 turns.
0885-8993 © 2015 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution
requires IEEE permission. See http://www.ieee.org/publications standards/publications/rights/index.html for more information.
ESTEBAN et al.: COMPARATIVE STUDY OF POWER SUPPLY ARCHITECTURES IN WIRELESS EV CHARGING SYSTEMS 6409

I. INTRODUCTION The remainder of the paper is organized as follows. Section

IRELESS inductive power transfer (IPT)—also known II presents an overview of the typical wireless EV charger and
W more generally as wireless power transfer—is an emerg-
ing technology that uses time-varying magnetic fields to trans-
associated theoretical concepts. Section III outlines the design
and characterization of the magnetic pads used for the physi-
fer power over large air gaps to energize or charge one or more cal implementation completed in this study. Section IV presents
electrical loads with high efficiency. This novel and disruptive an overview of the fundamental operation, desirable features,
technology has been quietly maturing in research labs over the limitations, and design strategy of each of the two architectures
past two decades and is now poised to redefine the way electrical being compared. Section V reviews the overall strategy adopted
power is consumed by society at large. in the design of the complete wireless charging system used in
To date, IPT has been successfully adopted in clean rooms and the comparison. Section VI presents the results of the compara-
industrial manufacturing floors for powering automated guided tive study, and, finally, Section VII recaps the key outcomes of
vehicles that are constrained to travel over a track and can require the comparison.
multikilowatt levels of power to operate [1]–[5]. It has also been
used to transfer small amounts of power over small air gaps for II. WIRELESS EV CHARGER OVERVIEW
charging of portable consumer electronics, such as cell phones, As shown in Fig. 1, a typical inductive EV charging system
tablets, and electric tooth brushes [6]–[9]. is made up of two electrically isolated entities known as the
More recently though, the application of this technology for primary and the secondary sides. The primary side can be further
static and in-motion charging of electrified transportation, such partitioned into the following subsystems.
as electric vehicles, buses, and train systems over larger air 1) Utility Input: Low-frequency energy source, which could
gaps and misalignment has seen an explosion in research and be single phase as in Fig. 1 or three phase.
development in academia and private industry [10]–[27]. The 2) AC/DC: Power conditioner that in addition to converting
adoption of wireless EV charging is viewed by many as a viable the utility input to a regulated dc voltage, carries out power
mechanism for extending the range of EVs and reducing the high factor correction, and serves as one of the control mech-
cost associated with their large battery packs [14], [28]–[30]. anisms for regulating the power flow of the entire system
With an interoperability industry standard (SAEJ2954) well in by modulating the dc input voltage [23], [34].
progress toward completion by 2015, wireless charging of EVs 3) DC/AC: Voltage fed full-bridge switching network that
has gone from being merely an obscure and exotic technology efficiently converts the incoming dc energy to high-
to a tangible and appealing reality [31]–[33]. frequency (HF) energy. It produces a HF square wave
The successful deployment and mass adoption of wireless EV with fundamental component equal to the chosen oper-
charging depends greatly on achieving high end-to-end system ating frequency. Usually the bridge semiconductors are
efficiency (85–90%) so as to be on par with existing conductive switched in accord with the phase-shift (PS) scheme il-
charging technologies. This requirement means that all sub- lustrated in Fig. 1, allowing for power flow regulation
systems in the cascade must have efficiencies in the order of through modulation of the PS delay time tPS between
96–97% [34]. A critical subsystem to the overall efficiency is each leg of the inverter. The semiconductors used in the
the primary power supply [22]. implementation of this paper are MOSFETs, accordingly,
A survey of the literature reveals that two primary-side power the parallel capacitors and diodes across each switch in
supply architectures are predominantly being used for wireless Fig. 1 are the parasitic elements of each device [13], [17],
EV charging systems, namely the series LC (SLC) resonant [23], [36], [39], [40].
and the hybrid series–parallel resonant (LCL) topologies [13], 4) Isolation: HF transformer required for compliance with
[14], [17]–[23], [35]–[44]. The published works on these two regulatory safety standards. This stage could be imple-
topologies offer excellent insights into the fundamental opera- mented with a step-up or step-down transformer depend-
tion, desirable features, and limitations of each for the applica- ing on the energy-coupling network (ECN) topology be-
tion at hand, but because of differences in the system operat- ing used. The capacitor CB is a dc-blocking capacitor
ing frequency, secondary side topology, control scheme, range that prevents transformer saturation and depending on the
of coupling and loading, as well as the use of different dis- power supply topology may also be used to adjust the
crete components in the physical implementation of common effective inductance of the transformer to some desired
subsystems, a true performance comparison of the two archi- amount which is then used as one of the energy storage
tectures cannot be made. The importance and practical value of elements in the next stage [34], [36], [37], [45].
being able to compare the performance of the aforesaid architec- 5) Matching: A network of very low-loss (high Q) energy
tures is discernible from the fact that one of the key objectives storage elements used to efficiently inject the HF en-
of the J2954 task force is the recommendation of a standard ergy to the next stage and by extension lower the volt–
primary power supply topology [32]. In view of the forego- ampere (VA) rating of the power supply. This stage is also
ing, this paper has as its goal the completion of a comparative commonly referred to as an impedance matching network
study of the SLC and LCL power supplies under identical op- (IMN), a resonant tank, bandpass filter, or a compensa-
erating conditions typical in stationary wireless EV charging tion network depending on the feature being highlighted
applications. in [11], [21], [46], and [23].
6410 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 30, NO. 11, NOVEMBER 2015

Fig. 1. Wireless EV charger, simplified equivalent referenced to the primary side, and PS gate driver pulse pattern.

In the literature, the first five blocks are commonly lumped partial series compensation on the secondary coil may
together and referred to as the IPT power supply [47], [48]. serve one or both of the following purposes: 1) reducing
Across a large and variable air gap, the secondary side can the VA rating of the components connected to the resonant
be similarly subdivided into an interconnection of the following tank, and/or 2) boosting the current out of the secondary
subsystems. coil, thereby enhancing its power delivery capabilities.
1) Matching: Another network of energy storage elements The later function is often carried out during the physical
that serves the same purpose as the one on the primary implementation’s fine-tuning stage, where there is often a
side. Though a number of different topologies are possi- need to increase I2 to achieve the design’s target power
ble, [11], [12], [49], a parallel resonant arrangement with level, but the pick-up coil’s magnetic characteristics have
partial series compensation was chosen for this study be- already been largely fixed by its construction. It is rele-
cause it is a commonly used topology for EV charging due vant to note that for the parallel secondary topology, the
of its constant current source characteristics [11], [20], maximum load current that may be delivered without the
[23], [49]. The addition of a capacitor in series with an use of partial series compensation occurs when the pick-
inductor and deliberate operation above resonance so as up is perfectly tuned and is equal to Isc , where Isc is the
to reduce or trim the overall inductive reactance to some short-circuit current that would flow in the secondary coil
smaller value is called partial series compensation. Partial if it were shorted, while coupled to the primary.
series compensation is often used in IPT systems with par- 2) AC/DC: A HF rectifier with an output LC filter that pro-
allel resonant topologies so as to trim the primary and/or vides the dc power to the vehicle battery. While other
secondary coil self-inductance to some lower value that is alternative ac/dc power conversion schemes can be im-
then resonated with a parallel capacitor [50], [51]. Capac- plemented on the secondary side so as to actively shape
itor CS 2 in Fig. 1 is used for partial series compensation the power flow, the simple HF rectifier with LC filter was
of the secondary coil. In a practical design, the use of chosen for this study because it results in a reduced cost
ESTEBAN et al.: COMPARATIVE STUDY OF POWER SUPPLY ARCHITECTURES IN WIRELESS EV CHARGING SYSTEMS 6411

and complexity on vehicle side system integration [19], by the application of the mutual inductance circuit model to
[20]. reflect the secondary to the primary [11], [46], [55].
Common to both primary and secondary is the magnetic cou- When a parallel secondary is reflected to the primary, it
pling stage. On the primary side, an inductive structure, com- presents an impedance Zr to the primary power supply. For
monly referred to as a magnetic pad or coupler, generates the ac a given frequency, the real and imaginary parts of Zr are both
magnetic flux that couples power to the secondary. Conversely, functions of the variable mutual inductance M and the load Req .
on the secondary side, another inductive structure, commonly The variability in M results from changes in alignment between
referred to as a pick-up, captures a fraction of the primary flux. the coils, while the variability of the load is due to changes in the
The size and geometry of the two coils may be identical or com- battery’s voltage and current profiles as it charges. Moreover,
pletely different. Like the primary- and secondary-side IMNs, the imaginary part of Zr is capacitive in nature [1], [56].
this stage is also designed so as to have the highest possible The reflected impedance seen at the primary can be shown as
unloaded quality factor, which generally means the use of a   1
stranded Litz conductor, with the total number of strands and Zr = Rr M, Req − j (4)
ωCr (M, Req)
gauge of each strand sized so as to minimize the ac loss at the
operating frequency [10], [42]. Req (ωM )2 [ω 2 C2 L2eq − (ω 2 C2 L2eq − 1)]
Rr (M, Req ) =
As highlighted in the magnetic coupling stage of Fig. 1, the 2 2eq − 1) + (ωL2eq )
2 (ω 2 C L
Req 2 2
mutual inductance and the primary and secondary coil self-
inductances vary as a function of position, this fact plays an (5)
 −1
important role in the design of the power supply [1], [23], [52]. ω 4 M 2 [C2 Req
2
(ω 2 C2 L2eq − 1) + L2eq ]
Fig. 1 also shows the wireless communication feedback mech- Cr (M, Req ) =
2 2eq − 1) + (ωL2eq )
2 (ω 2 C L
Req 2 2
anism used to close the control loop in order to actively regulate
the power flow due to positional changes between the source (6)
and sink coils and/or changes in loading [21], [23].
With respect to Fig. 1 provided that the output inductor Lo where L2eq is the equivalent secondary coil’s inductance after
is large enough to ensure continuous conduction of the recti- application of partial series compensation
fier diodes over the system’s full range of operation, then the 1 ω 2 L2 CS 2 − 1
following relation holds true [23] L2eq = L2 − = . (7)
ω 2 CS 2 ω 2 CS 2
π 2 RL The power that is, thus, delivered by the RMS primary current
Req = . (1)
8 to the reflected load is given by
As illustrated by the ac equivalent circuit in Fig. 1, at steady P = Re{I12 Zr }
state, the entire system may be greatly simplified for analysis
and design by replacing the primary-side power electronics with Req (ωM I1 )2 [ω 2 C2 L2eq − (ω 2 C2 L2eq − 1)]
= . (8)
2 2eq − 1) + (ωL2eq )
a sinusoidal voltage source vS and reflecting the secondary side 2 (ω 2 C L
Req 2 2

to the primary, which results in a reflected impedance Zr in

If the system is operated at or close to the secondary’s resonant
series with the primary coil.
frequency
The simplified input voltage may be expressed as
1
vS = VS × sin(2πf0 t). (2) ω0 =  (9)
L2eq C2
The RMS amplitude of vS is related to the FB’s dC-bus input
voltage VDC , the operating frequency ω, and the PS control the term ω 2 L2eq C2 will be equal or very close to one. When
delay tPS , as follows: this happens (4) and (8) become
√   ω0 M 2 Q2v ω0 M 2
2 2VDC ω0 × tPS Zr = −j (10)
VS = sin . (3) L2eq L2eq
π 2
The reduction of the primary-side power electronics to a sinu- ω0 (M I1 )2 Q2v
P = (11)
soidal voltage source is commonly referred to as fundamental- L2eq
mode analysis (FMA). The application of FMA to power elec- where Q2v is known as the voltage boost quality factor, and is
tronic converters, and in particular to IPT systems, is a very defined as
well-established practice in the literature and is possible be-
Req
cause the resonant networks that are connected to the output Q2v = . (12)
ωL2eq
of the FB have significant filtering capability, which effectively
reject all harmonics with the exception of the fundamental com- As can be discerned from (10), and as is depicted in the
ponent of the HF square-wave inverter output [13], [23], [48], simplified ac equivalent circuit of Fig. 1, when the system is
[53], [54]. operated at resonance, the dependence of the reflected capactive
The second simplification may be achieved by the application reactance on the load is eliminated. Equation (11) is conven-
of standard circuit analysis techniques at the secondary followed tionally known as the IPT power equation, and it is deemed by
6412 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 30, NO. 11, NOVEMBER 2015

many authors as the starting point for the design of a resonant TABLE I
NOMINAL MAGNETIC PAD ATTRIBUTES
IPT system [1], [4], [23]. In theory, the minimum RMS primary
current required to transfer a desired maximum amount of power
Attribute Name Value/Dimension/Material
Pm ax , when the system is operated at resonance is equal to
Inductance (L a v g ) 170.51 μH
Pm ax L2eq Quality Factor (Q u ) 535.67
I1 = . (13) AC Resistance (R A C ) 60 mΩ
ω0 M 2 Q2v Power Loss (P l o s s ) 94.61 W
Number of Turns (N) 22
The application of partial series compensation to the sec- Outer Diameter (O.D.) 482.6 mm
ondary coil results in three different quality factors [23], [50], Inner Diameter (I.D.) 101.6 mm
Pitch (p) 3 mm
[51]. The first one Q2v is the conventional loaded quality factor Conductor Diameter (d) 5.84 mm
of the standard parallel RLC network. The second quality factor Ferrite Length (F L ) 200 mm
that may be defined is known as the current boost quality factor: Ferrite Width/Height (F W , F H ) 25 mm
Ferrite Excess Length (F E ) 26.58 mm
I2 Aluminum Shield Thickness (T) 6.5 mm
Q2i = . (14) Former Dimensions (L, T) 540, 20 mm
IS c Litz Conductor Type II (16AWG 4×5X24/36)
Ferrite Material Ferroxcube 3C94
Q2i is used to quantify the increase in I2 from its uncom- No. of Ferrite Spokes 12
pensated value, Isc , to its compensated value, Q2i × Isc . The
amount of partial series compensation that may be carried out
as quantified by Q2i has been reported as being limited to less
than 3 because of the added sensitivity to the overall resonance 3) achieving self-inductance values that will ensure coil volt-
tuning of the secondary (i.e., due to component aging), as well age levels that are within established UL safety regula-
as because of saturation of the pick-up coil ferrite that results tions;
from the increased flux density [50], [51]. 4) ensuring regulatory compliance with established field
The third quality factor that may be defined is the secondary’s emission standards (ICNIRP);
overall quality factor, which takes into account the effect of both 5) having a complete understanding of how all key mag-
Q2i and Q2v ; this later quality factor is defined as netic attributes vary over the system’s range of positional
operation so as to properly design the primary and sec-
Q2t = Q2v × Q2i . (15) ondary power electronics in terms of component ratings
and closed-loop control.
For practical designs, Q2t is limited to being less than 10 [4],
Because the focus of this paper was the design of the primary-
[47], and for most high-power EV charging systems it is less
side power electronics, the approach taken in the design of the
than 6. This limitation in the overall secondary Q is due to the
magnetic pads was to forgo the initial FEA optimization process
challenges that arise in controlling the system because of the
and instead construct the pads so as to replicate some of the
narrower bandwidth at higher values of Q2t . Another factor that
better designs showcased in the literature, to this end, a circular
dictates the limit imposed on Q2t is the need to maintain the
Archimedean coil geometry was chosen [52], [57], [58]. Despite
secondary’s VA ratings within the limits prescribed by governing
having lower coupling than other similarly sized coil geometries
safety standards such as UL [1], [23], [52].
over identical air gaps and misalignment, for static EV charging
With the aforesaid definitions in mind, L2eq may also be
applications, the circular geometry is still the most widely used
expressed in terms of the three quality factors that result from
[10].
the application of partial series compensation as
The coils were designed so as to have known initial self-
Req Req Q2i inductances of 125 μH with the aid Wheeler’s formula for air-
L2eq = = . (16)
ωQ2v ωQ2t core Archimedean coils [62]. This initial value served only as a
baseline, as it would later increase due to the addition of ferrite
III. MAGNETIC PAD DESIGN AND CHARACTERIZATION spokes. To act as a shield and enhance the mechanical attributes
of the design, an aluminum backplane was added behind the
The first step in the design of an IPT system is the selection
ferrite spokes, effectively sandwiching the spokes between the
of a suitable inductive structure geometry from a number of
polycarbonate coil former and the aluminum backplane [52].
available choices [4], [11], [12]. Once a geometry that best suits
The nominal attributes of the coils built are listed in Table I.
the target application has been selected, a thorough FEA-based
All physical dimensions are defined in Fig. 2(a). The ac power
EM optimization of the design is undertaken prior to fabricating
loss of the pad as a result of skin effect is defined by (17), where
it. The goals of the FEA optimization are [51], [52], [57]–[61]:
I and V are the nominal operating primary and secondary pad
1) achieving the highest possible values of mutual inductance
voltage and current at rated power, and Qu is the coil’s unloaded
and magnetic coupling over the system’s target range of
quality factor defined with respect to its ac resistance at the target
positional operation;
operating frequency
2) reducing the design’s physical size and material cost as
much as possible without compromising its power transfer I1 V1 + I2 V2
capabilities; Ploss = . (17)
Qu
ESTEBAN et al.: COMPARATIVE STUDY OF POWER SUPPLY ARCHITECTURES IN WIRELESS EV CHARGING SYSTEMS 6413

Fig. 2. Mechanical features and measured coupled magnetic pad attributes at a 16-cm vertical offset: (a) Magnetic pad construction; (b) coupling coefficient,
mutual inductance, short- and open-circuit self-inductance; (c) expanded view of mutual- and self-inductances.

Once the coils were built their magnetic attributes were ex- ify the existence of a fixed characteristic null in the magnetic
perimentally measured and used to carry out the design of profile of circular coils [52]. Another noteworthy observation
the primary- and secondary-side power electronics. For precise is that Lsc and Loc both converge to a common value as the
measurement of the pads’ coupled self- and mutual-inductances, pads are decoupled, with the convergence starting around the
as well as coupling coefficient over their range of positional op- characteristic null point. As already noted, the operating range
eration, the test bench shown in Fig. 4(c) was built. To measure of coupling is a crucial design variable that must be known in
M and k, the coupled pad’s were energized with a known pri- order to properly design the power supply, this in fact is one of
mary current by means of an inverter with a series resonant ca- they key items that will be standardized by the J2954 task force
pacitor. The required measurements were taken, while the pads for interoperability [33]. Fig. 2 shows the operating range of
were misaligned in the vertical and/or horizontal directions in coupling being considered for this comparative study, namely
small steps. the region to the left of the vertical dashed line inclusive. This
The coupled-coil self-inductances were measured at each range corresponds to a maximum axial misalignment distance
point along the horizontal offset trajectory with an LCR me- of 12 cm, which translates to a coupling coefficient in the neigh-
ter. The mutual inductance was obtained by measuring the sec- borhood of 0.18–0.25.
ondary’s open-circuit voltage VL 2oc and the primary’s current I1 .
The coupling coefficient was obtained by two separate methods. IV. POWER SUPPLY ARCHITECTURES
The first method applied was the measurement of the primary-
coil voltage VL 1 and VL 2oc so as to use the voltage transfer ratio A. SLC Resonant Power Supply
VL 2oc /VL 1 as a measure of k [39]. The second method used was Fig. 3(a) shows the simplified equivalent circuit of the SLC
the use of the conventional ac circuit definition of k given as the power supply with all energy storage elements assumed to be
ratio of M and the square root of the product of the coupled self- lossless [45]. As can be seen, the ECN of the SLC architecture
inductances. Both methods produced very similar results. Fig. 2 is made up of a single capacitor in series with the primary coil
shows the pads measured magnetic attributes over a range of
axial misalignment spanning 64 cm for a coil-to-coil separation L2eq
C1 = (18)
distance of 16 cm. Fig. 2(b) shows all pad attributes together, ω02 [LΔ 1 L2eq − M 2 ]
while Fig. 2(c) shows an expanded view of the mutual and self- LΔ 1 = L1 − L1eq . (19)
inductances only. As can be seen, the coupled self-inductances
were measured for two different conditions: 1) a shorted sec- The value of the tuning capacitor C1 is sized according to (18)
ondary pad (Lsc ), and 2) an open secondary pad (Loc ). The so that when the system is operated at its target frequency, the
measurements are taken with the LCR meter on one pad, while resulting equivalent primary inductance L1eq seen by the FB is
the other one coupled to it has its terminals either shorted or about 10% of L1 [14]. This partial series compensation scheme
opened. Consideration of these two cases is necessary because is equivalent to choosing C1 so as to fully resonate with L1 at
depending on the resonant topology chosen for the primary and f0 = 2π √L1 C , and then, operating the system at a frequency
1 1
secondary matching networks, the self-inductance may behave f0 = 1.05 × f0 [63].
differently. For example, for the parallel resonant topology, the The reason why the SLC power supply is operated slightly
pad’s self-inductance varies according to the short-circuit mea- (5%) above perfect resonance is that this allows for the reduc-
surement profile, while for the series resonant topology, the tion of switching power losses in the FB semiconductors by
inductance varies according to the open-circuit measurement means of the circulating inductive energy associated with the
profile [45]. The plots of mutual inductance and coupling ver- residual inductance L1eq positioning zero voltage across the
6414 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 30, NO. 11, NOVEMBER 2015

Fig. 3. Simplified ac equivalent circuit of primary power supply architectures being compared: (a) SLC architecture, and (b) LCL architecture.

semiconductors in each leg of the FB just before turn-on. This When C1 is chosen according to the previously described
placement of zero voltage across the switches reduces the in- procedure, the value of the primary current is given as follows:
stantaneous turn-on loss that is normally associated with the √  
finite period of time during which the drain-to-source voltage VS 2 2VDC ω0 × tPS
I1 = = sin
VDS of the nonideal switch is decreasing, while its drain current |Rr + jω0 L1eq | π|Rr + jω0 L1eq | 2
ID is rising. This mode of operation is commonly known as (20)
zero voltage switching (ZVS), and the specific mechanism used using (20) along with (13) the design of the SLC topology can
to reduce VDS to zero is the complete discharge of the parasitic be finalized by determining either the required dc-bus voltage
drain-to-source capacitance of the switch CDS by means of cur- VDC or the required HF isolation transformer turns ratio needed
rent flowing through L1eq and the free-wheeling body diode of to achieve the target I1 . Selection of VDC assumes that the dc-
the MOSFET during the dead-time, td , that occurs between the bus voltage can be controlled at will, such as would be the case
turn-off and turn-on of series switches. if the system had an intermediate buck or boost converter stage
The sizing of C1 , also takes into account the nominal value of between the utility input and the FB. On the other hand, selection
Cr (M ) at the system’s target operating frequency and coupling of a suitable HF transformer turns ratio assumes that the dc-bus
point; accordingly, C1 is chosen so that the equivalent capac- voltage is limited.
itance resulting from the combination of Cr (M0 ) and C1 will
yield the desired ZVS mode of operation previously described. B. Series–Parallel Resonant (LCL) Power Supply
If the system deviates significantly away from its nominal op- Fig. 3(b) shows the simplified equivalent circuit of the LCL
erating range of coupling, then the resulting variation in the power supply neglecting all parasitic resistances. As can be
equivalent series tuning capacitance and the nonlinear change seen, the ECN of the LCL architecture has two more discrete
in L1 described in Section III will take the FB out of the de- energy storage elements than the SLC topology, namely the
sired ZVS mode and the power losses will increase significantly series bridge inductance LB and the series capacitance CS 1 .
[40]. To prevent the foregoing situation and a potential failure Moreover, the primary resonant capacitance C1 is now in paral-
of the semiconductors, the system’s switching frequency must lel with the primary coil and CS 1 . This topology is sometimes
be adjusted so as to always maintain ZVS operation. As dis- also referred to as an LCL-T topology because of the character-
cussed in [64], the frequency control scheme cannot rely on an istic T-shape of the ECN. The series capacitance CS 1 is used for
autonomous phase-locked loop-based approach due to bifurca- partial series compensation of L1 so as to reduce it to a smaller
tions that are inherent in the system because of its high order. equivalent value L1eq . The parallel capacitance C1 is chosen
Instead, a few discrete points in the frequency spectrum around so as to fully resonate with L1eq at the system’s operating fre-
the nominal operating frequency are selected and used to ensure quency. The bridge inductance is used to convert the HF voltage
the desired ZVS operating mode. The SLC architecture is at- source characteristic of the FB to an equivalent current source,
tractive primarily because of its very simple matching network, and for this reason, this ECN is sometimes spoken of as being
consisting of only a single capacitor, which tends to result in a an impedance conversion network (ICN) [35], [37].
reduced physical implementation cost and complexity. Despite A very desirable characteristic of the LCL power supply is
its reduced component count, the need to dynamically vary the that when operated at or very close to resonance, it behaves as
switching frequency of the SLC topology adds to the complexity a constant current source over a very large range of coupling
of the closed-loop controller design. In addition, the effective and loading making its control loop design easier [66]. This is
impedance seen by the FB is very sensitive to changes in both in contrast with the SLC topology, which is quite sensitive to
coupling and loading, which further increases the complexity changes in coupling and loading and requires a combination
of the controller. The foregoing sensitivity can also result in a of dc-bus voltage and frequency adjustments to maintain its
very large current stress on the bridge semiconductors when the constant current source operation [1].
coupling of the magnetic pads decreases significantly beyond The primary current of the LCL topology is fixed by the
its intended operating range [39], [65]. effective reactance X1 of the primary coil along with its partial
ESTEBAN et al.: COMPARATIVE STUDY OF POWER SUPPLY ARCHITECTURES IN WIRELESS EV CHARGING SYSTEMS 6415

series tuning capacitor and is given as TABLE II

NOMINAL DESIGN SPECIFICATIONS
√  
VS 2 2VDC ω0 × tPS
I1 = = sin . (21) Specification Name Specification Value
X1 πX1 2
Pm a x 3 kW
Another feature of the LCL architecture that is often praised E f fm i n 85%
in the literature is that it reduces the current stress of the bridge f0 30 kHz
Secondary Topology parallel–series
semiconductors by constraining the large reactive resonant cur- km in 0.1773
rents to flow only in the resonant tank [23]. The FB only sources M (k m i n ) 30.93 μH
the current associated with the active power flow and the reso- L 2 s c (k m i n ) 168.15 μH
L 1 s c (k m i n ) 168.15 μH
nant tank losses. The decrease of the bridge current results in a L 1 o c (k m i n ) 172.87 μH
reduced on-time conduction loss for the semiconductors. Cs 2 0.23 μF
The strategy in designing the LCL network is to size the C2 0.64 μF
Req 17.76 Ω
energy storage elements so that the magnitude of all the resulting Q2t 3
reactances are equal. Accordingly, the design equations that Q2v 2.14
govern the sizing of the LCL network are as follows: Q2i 1.40
VD C 0 240 V
L2eq
CS 1 = (22)
ω02 [LΔ 1 L2eq − M 2 ]
V. OVERALL SYSTEM DESIGN METHODOLOGY
where LΔ 1 is once again given by AND PHYSICAL IMPLEMENTATION

LΔ 1 = L1 − L1eq (23) The design of the primary power supply requires that the
following parameters be known:
and L1eq is once again the equivalent inductance of the partially 1) maximum power transfer level Pm ax ;
tuned primary coil 2) nominal operating frequency f0 ;
3) magnetic pad characteristics: km in , M (km in ), Lsc (km in ),
X1 and Loc (km in );
L1eq = . (24)
ω0 4) secondary-side ECN topology and energy storage element
values;
The parallel resonant tuning capacitance C1 is obtained from 5) nominal load value Req ;
1 6) nominal value of the three secondary quality factors: Q2t ,
C1 = . (25) Q2v , and Q2i ;
ω 2 L1eq
7) the nominal dc-bus input voltage of the inverter VDC 0 .
Finally, the bridge inductance is in practice set to be 10% For this study, the above stated specifications are shown in
larger than the value of L1eq , so as to ensure that the inverter Table II. The power level was selected so as to operate close
will always see a lagging/inductive load [37] to an SAE level II charging power. The nominal operating fre-
quency was set at 30 kHz primarily because at the time when
LB = 1.1 × L1eq . (26) the magnetic pads for this study were initially designed most of
the research on wireless charging was focused on frequencies
A lagging load is desirable because if the bridge were to around 20 kHz. Recently, the J2954 charging task force set the
ever become capacitively loaded, then the semiconductor losses nominal operating frequency for light duty vehicles at 85 kHz
would increase significantly due to large reverse recovery cur- [33], however, because of the limitations imposed by the large
rents in the semiconductor’s intrinsic body diode [51]. More- resonant voltages in the matching networks of the constructed
over, as noted in [66], the power stage can be operated in a coils at higher frequencies and because of limitations imposed
ZVS mode provided that the chosen switching control scheme by the value of capacitors that were available to complete the
ensures conduction in the intrinsic body diodes prior to switch physical implementation, the final frequency used to drive the
turn-on. This operating mode coupled with the previously noted system and carry out the comparison was actually 29 kHz.
reduced current stress during conduction can result in very high The nominal dc bus voltage was chosen so as to have a range
converter efficiency. of regulation voltages that would allow for both architectures to
As was the case with the SLC design strategy, the selection deliver the target power over the full operating range of coupling
of the capacitance CS 1 takes into account the reflected virtual and loading being considered as well as to readily highlight the
capacitance Cr (M0 ) [23]. The bridge inductance LB could be differences in the voltage regulation strategy needed for each
implemented with an actual inductor, but by designing the HF architecture. In a practical setting, a higher nominal dc bus
transformer so as to have a precise value of an overall equiv- voltage would likely be chosen for the LCL topology.
alent leakage inductance, it can be constructively used as a For this particular design, the extreme operating point is cho-
part of the matching network, resulting in a reduced number of sen to correspond to a maximum coil-to-coil air gap of 16
components. cm and an axial misalignment of 12 cm. This operating point
6416 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 30, NO. 11, NOVEMBER 2015

TABLE III expressions at a fixed coupling point, while varying the load
POWER SUPPLY MATCHING NETWORK VALUES
over the desired range. The resulting voltage and current rat-
ings for both sweeps are stored in two result arrays, one for the
LB C1 CS 1
coupling sweep and another for the loading sweep. The routine
SLC N/A 0.21 μF N/A then compares the peak ratings of both result arrays for each
LCL 49.25 μH 0.63 μF 0.28 μF circuit element and extracts the maximum values. In addition to
determining peak component ratings, the theoretical losses and
efficiency of the designed system may also be obtained over its
corresponds to km in shown in Table II and was selected so as to operating range with the foregoing MATLAB routine by inclu-
be representative of the nominal light duty EV ground clearance sion of the measured parasitic resistances of the energy storage
and the ongoing research efforts showcased in the literature for elements and of the nominal Rds(on) of the semiconductors used.
static wireless charging [23], [41], [44]. The magnetic design Much as was the case with I1 , the max ratings, losses, and effi-
characteristics for the full operating range of coupling are shown ciency obtained from sweeping the theoretical model serve only
in Fig. 2. As noted in Section III, the region to the left of the as an initial reference for the design, and may have to be ad-
dashed vertical line is the full operating range of coupling. The justed upon evaluation of the physical system’s behavior in the
idea is to ensure that the system is able to transfer the required face of higher order effects associated with component drift do
power at the boundary, with the understanding that if it is able to to prolonged operation.
do so, then it will also be able to do so at higher coupling points Fig. 4 shows the physical implementation of the complete
in the operating range. In a practical implementation, a feedback IPT system designed for this study and the operating waveforms
control system would be used to guide the user in parking so as for each power supply as well as for the vehicle side pick-up.
to ensure operation within the target nominal range of coupling Fig. 4(a) is the series resonant architecture, while Fig. 4(b) is
as well as to disable operation altogether when outside it [23]. the LCL architecture. Fig. 4(c) shows the entire IPT system
From the given specifications, the primary current needed to operating a rated conditions. The 3-kW resistive load was im-
transfer the required power can be computed using (13); the plemented with a 3-by-3 array of light bulbs consisting of three
calculated value serves as an initial design reference value. The parallel connected branches, each with three series connected
final value of I1 for the physical implementation may need to bulbs. Each bulb in the array was rated for 1 kW at 120 V. The
be larger in order to transfer the target power level. The need to core of the HF isolation transformer (HF XFRMR) between the
increase the primary current upon completion and evaluation of bridge and the ECN was implemented with the same ferrite ma-
the initial physical implementation is well reported upon in the terial used for the spokes of the magnetic pads. The constructed
literature [4], [13], [38], [52]. As highlighted in the cited refer- transformer had a one-to-one turns ratio with 21 turns per side.
ences, the VA of the tuned secondary must be able to support Moreover, its effective leakage inductance was controlled by
the desired power transfer level. adjusting the core’s air gap length until it measured 67.28 μH.
Next, the values of the energy storage elements in the match- For the SLC topology, the transformer’s leakage inductance was
ing network of each power supply architecture are obtained completely canceled by setting CB equal to 0.4 μF, whereas for
by application of the governing equations presented in Section the LCL topology it was reduced to the target value of LB by
IV. The actual values used for the physical implementation are setting CB equal to 1.56 μF. For the FB switching network,
listed in Table III. Once the component values on both the pri- the IXFN132N50P3 power MOSFET was chosen because of its
mary and secondary have been obtained, their respective voltage very low Rds(on) , fast intrinsic body diode, and low gate charge
and current ratings must be determined so as to ensure their safe requirement. The gate driver stage was implemented with two
operation. The equations governing the ratings of the primary- low-cost FAN7390 bootstrap gate driver IC’s, and the control
and secondary-side elements are summarized in Appendix A in logic was generated using a TMS320F28335 DSP from Texas
terms of RMS values. Instruments. All PCB’s are double sided and were fabricated
As is seen from the complete mathematical model presented in-house using a QC5000 prototyping system. The gate driver
thus far, the voltages and currents will vary with changes in cou- stage was implemented on a separate PCB than the power stage
pling and loading. Consequently, the establishment of the peak so as to ensure good signal integrity with the limited number of
ratings requires that all of the system’s equations be swept over copper layers used [22]. In the operating waveforms, VINV is the
their full range of coupling and loading and that the maximum inverter output voltage, and it leads the inverter current by 36◦
values of each element’s voltage and current be ascertained. for both architectures confirming the desired inductive mode of
This task is made even more challenging because as noted in operation targeted in the design. VGS is the gate-to-source volt-
Section III, when the coupling changes, so do the mutual- and age of one of the power switches. VL 1 is the primary coil voltage,
self-inductances of the coupled coils. To be able to tackle the and the 86.53 MHz ringing apparent on it at the switching in-
said computationally intensive task, a MATLAB routine was stants is a result of the transformer’s parasitics resonating with
developed that steps through the system’s range of coupling and the HF components of VINV . The stated root cause of the ring-
loading. For the coupling sweep, the routine evaluates each ex- ing was confirmed by two separate tests: First, the transformer
pression in the model with the corresponding measured values was completely removed, leaving only the ECN’s, in which case
of mutual- and self-inductance at the given coupling point for the ringing disappeared completely; second, when the value of
a fixed load value. In like manner, the routine also evaluates all the transformer’s parasitic attributes were modified by either
ESTEBAN et al.: COMPARATIVE STUDY OF POWER SUPPLY ARCHITECTURES IN WIRELESS EV CHARGING SYSTEMS 6417

Fig. 4. System implementation and operating waveforms: (a) SLC power supply, (b) LCL power supply, (c) complete system operating at rated load and minimum
coupling, (d)and (e) SLC waveforms, (f) and (g) LCL waveforms, (h) secondary-side waveforms.

changing the core’s air-gap length or the number of turns, the other hand, the LCL architecture requires the use of two more
frequency of the ringing would also change. A practical way energy storage elements, but by constructively using the para-
to eliminate the HF ringing is to place a split and balanced in- sitic inductance of the common HF isolation transformer, the
ductor between the bridge output and the transformer so as to total component count for the LCL power supply is reduced to
filter as much harmonic content as possible. All of the operating just two, namely: CS 1 and C1 . In view of the higher compo-
waveforms were captured from the complete set-up shown in nent count, the LCL topology is more complex in its physical
Fig. 4(c); consequently, VLOAD and ILOAD in Fig. 4(h) are the implementation and its volumetric space requirements.
voltage and current of the 3-kW light-bulb load bank. Video Fig. 5(c) shows the normalized open-loop power of each
of the full working system for both power supply architectures power supply as a function of coupling. The constant current
may be viewed online via the hyperlinks in Appendix B. source characteristic of the LCL architecture is clearly in evi-
dence. On the other hand, the sensitive nature of the SLC topol-
VI. PERFORMANCE COMPARISON ogy to changes in coupling is seen in the fact that as the coupling
increases, the power actually decreases. This later behavior is
A. Size and Complexity counterintuitive, but it is readily explained by the fact that the
The SLC topology has the lowest number of discrete energy SLC’s primary current is heavily affected by the variation of the
storage elements, with only one tuning capacitor C1 . On the reflected secondary load. Fig. 5(d) shows the open-loop primary
6418 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 30, NO. 11, NOVEMBER 2015

Fig. 5. Measured performance of SLC and LCL power supplies: (a) Efficiency versus percent of rated load, (b) efficiency versus coupling under constant power,
(c) open-loop normalized power versus coupling, (d) open-loop normalized primary current versus coupling, (e) ECN impedance magnitude versus coupling, (f)
SLC switching frequency for sustained ZVS operation over range of coupling (normalized to the nominal f0 ), (g) normalized ECN currents versus coupling under
constant power (normalized to the peak value of I1 ), (h) normalized ECN voltages versus coupling under constant power (normalized to the peak value of V C 1 SLC ),
(i) normalized dc bus voltage versus coupling for constant power (normalized to V DC0 ), (j) normalized ECN currents versus load voltage under constant power
(normalized to the peak value of I1 ), (k) normalized ECN voltages versus coupling under constant power (normalized to the peak value of V C 1 SLC ), (l) normalized
dc bus voltage versus load voltage for constant power (normalized to V DC0 ).
ESTEBAN et al.: COMPARATIVE STUDY OF POWER SUPPLY ARCHITECTURES IN WIRELESS EV CHARGING SYSTEMS 6419

current for each topology as a function of coupling. As expected, conditions, the entire primary current, in this case 30 A, is han-
the LCL topology has a perfectly constant current throughout the dled by the semiconductors in its switching network, compared
entire range of coupling, while the SLC topology’s current de- to only 23 A for the FB of the LCL topology. This fact makes the
creases significantly as coupling increases. As can been seen in required current and power rating of the switching devices for
Fig. 5(e), as the coupling increases, the reflected impedance seen the SLC topology 1.3 and 1.7 times greater, respectively, than
by the SLC architecture increases correspondingly; thereby, re- those for the LCL topology for the same power transfer level;
ducing the initial primary current from its nominal value. This moreover, as will be explained next this later factor also plays
sensitivity requires that the dc-bus voltage be increased so as to an important role in efficiency of each architecture.
maintain a constant current source operating mode.
An increase in RL beyond its nominal operating point would
translate to an increase in Q2v in (11). For the LCL topology, C. Efficiency
with its constant current characteristic, such a condition would Fig. 5(a) and (b) shows the end-to-end efficiency of the com-
result in an increase in the open-loop power sourced by the plete IPT system for both architectures as a function of the
power supply accompanied by a corresponding rise in the in- percent of rated loading and coupling, respectively. The fore-
verter current. In practice, however, the closed-loop controller going efficiency is defined as the ratio of the power at the load
would act to lower the dc-bus voltage so as to maintain the de- and the input power at the dc bus of the switching network.
sired power level with a smaller primary current. On the other Moreover, the efficiency of Fig. 5(b) was measured, while the
hand, for the SLC topology the same deviation in RL would load power was maintained constant over the full range of cou-
result in a decrease in the open-loop power delivered due to the pling via the regulation of the dc-bus voltage according to the
decrease of the primary current caused by the larger reflected characteristic shown in Fig. 5(i). For this study, the losses in
secondary impedance. In the latter case, the increase in Q2v is the FB were measured using the TPS2PWR1 power analysis
not enough to overcome the simultaneous reduction in the pri- application software for the Tektronix TPS2024B oscilloscope.
mary current due to the square dependence of the power on I1 . In the case of the SLC topology, the bridge losses at full power
So for the SLC topology, the controller would have to increase were 394 W, whereas for the LCL topology they were only 15
the dc-bus voltage so as to maintain the desired power level with W. This translates to an 88.72% and 99.52% efficiency for the
the higher value of Q2v . power stage of the SLC and LCL architectures, respectively.
Given the target range of coupling chosen for this study, it was A noteworthy observation in Fig. 5(b) is that as the coupling
determined that the FB switching frequency for the SLC archi- changes, the efficiency of the LCL architecture decreases from
tecture would have to be adjusted according to the characteristic its peak value to just above 91%, while that of the SLC architec-
shown in Fig. 5(f) so as to maintain ZVS. This requirement ture increases to 90%. This happens because as seen in Fig. 5(h),
along with the need to vary the dc-bus voltage so as to regu- as the coupling increases, the bridge current sourced by the LCL
late the primary current against changes in coupling make the increases resulting in increased semiconductor losses, whereas
overall control of the SLC architecture more complex. for the SLC the opposite takes place.
The power supply’s light load efficiency is a very important
figure of merit in EV charging applications because the power
B. Cost and Component Stress
supply is fully loaded for only one third of the total charging
The cost of a power supply is defined as the ratio of total time, namely during the bulk charging stage. However, during
active power delivered to the load and the total VA rating of the the absorption and float stages, which account for two thirds
supply. As noted in [45], the VA stored in the tuning components of the total charging time, the power supply is only partially
of the power supply is related to its size and associated cost. loaded [67]. As is seen in Fig. 5(a), the LCL architecture has
Hence, the ratio of kW/kVA is a good estimate of the kW/$ of better light-load performance than the SLC, this is so because
the power supply. In this regard, the cost of the SLC and LCL as highlighted in [65], [66], under no load, the input impedance
power supplies when operated at their rated power levels are of the LCL network becomes infinite; consequently, the inverter
0.041 and 0.042, respectively. So as might be expected, the LCL output current becomes zero and the converter is able to maintain
topology has a higher overall cost, though not by much (i.e., very high light-load efficiency. In contrast, at light loads, the
2.41%). SLC topology sources very large reactive currents due to the
Fig. 5(g) and (j) shows the normalized voltages across all very small tuned impedance seen by the inverter [39]. A way to
energy storage elements, for both architectures, over the full improve the light-load efficiency of the SLC topology is to have
operating range with the normalization done with respect to the a secondary-side regulator [42], but this improvement comes at
voltage of the SLC series capacitor. As can be observed, the volt- the expense of an increased cost and complexity on the vehicle
age stress on the single capacitor of the SLC topology is much side.
higher than that on any of the capacitors of the LCL topology As can be seen from these results, the LCL topology is clearly
over the full operating range. This fact is further corroborated superior over the entire range of loading and coupling, achiev-
in Fig. 4(c), which shows a 920-V stress on the series capacitor ing a peak end-to-end efficiency of 96%, and maintaining an
of the SLC topology, and a 624-V stress on the series capacitor overall higher efficiency even at its worst case efficiency oper-
of the LCL topology. Another factor that causes a higher stress ating point. The SLC topology, on the other hand, suffers from
on the components of the SLC topology is the fact that at rated poorlight-load efficiency, dropping to as low as 70% at light
6420 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 30, NO. 11, NOVEMBER 2015

TABLE IV
SUMMARY OF A COMPARATIVE STUDY

Number of Capacitors Capacitor Voltage Stress Cost [W/VA] Control Complexity Rated Load Efficiency Light Load Efficiency

SLC 1 HIGH 0.041 HIGH 86% 76%

LCL 2 LOW 0.042 LOW 96% 95%

load. Furthermore, the peak system efficiency at the nominal IC 1 =

operating point with the SLC supply is only 86%. This 10% (1 − ω 2 L1 e q C1 ) + jωC1 Rr − 1
difference in the peak efficiency of the two architectures is a VS (28)
Rr (1 − ω2 LB C1 ) + jω[LB (1 − ω L1 e q C1 ) + L1 e q ]
2
result of the fact that the bridge power switches of the SLC
I1 =
topology must handle the full-resonant current, which increases
the semiconductor loses even despite the sustained ZVS oper- 1
VS (29)
ation. On the other hand, the LCL topology is able to maintain Rr (1 − ω 2 LB C1 ) + jω[LB (1 − ω 2 L1 e q C1 ) + L1 e q ]
very high efficiency because the bridge stage only handles the VB =
current associated with the active power transfer and the reso- jωLB [(1 − ω 2 L1 e q C1 ) + jωC1 Rr ]
nant tank losses. The foregoing results are still applicable at 85 VS (30)
Rr (1 − ω 2 LB C1 ) + jω[LB (1 − ω 2 L1 e q C1 ) + L1 e q ]
kHz because the key reason for the higher efficiency of the LCL,
namely the lower current stress of the semiconductors, is inde- VC 1 =
pendent of frequency; consequently, while it is to be expected [(1 − ω 2 L1 e q C1 ) + jωC1 Rr − 1]VS
that the overall efficiency of both topologies would go down at jωC1 (Rr (1 − ω 2 LB C1 ) + jω[LB (1 − ω 2 L1 e q C1 ) + L1 e q ])
85 kHz due to the increased switching losses, the efficiency of (31)
the LCL would still remain higher than that of the SLC.
VC S 1 =
VII. CONCLUSION VS
jωCS 1 (Rr (1 − ω 2 LB C1 ) + jω[LB (1 − ω 2 L1 e q C1 ) + L1 e q ])
The development of wireless EV charging technologies has
gained unparalleled momentum over the latter half of the past (32)
decade. The release of a formal industry standard governing this VC S 1 =
novel technology will enable OEM’s to bring wireless EVs to
VS
market as early as 2016. A key question that remained unan-
jωCS 1 (Rr (1 − ω 2 LB C1 ) + jω[LB (1 − ω 2 L1 e q C1 ) + L1 e q ])
swered until now was: Which of the two power supply architec-
tures commonly used in wireless EV chargers is better in terms (33)
of size, complexity, cost, and efficiency? 1 1
L1 e q = L1 − 2 − 2 . (34)
This paper has compared the foregoing two topologies by ω CS 1 ω Cr (M, Re q )
applying an identical performance test. The results, which are
summarized in Table IV, show that while the SLC topology has
slightly lower component count and cost, the LCL topology is B. SLC Architecture
superior in terms of its power conversion efficiency over the full
range of coupling and loading imposed. Moreover, its constant
current source characteristic make its closed-loop control imple- VS
I1 = (35)
mentation easier than the alternative SLC architecture. Because Rr + jωL1eq
the efficiency of the power supply is a key figure of merit in a VS
wireless EV charging, the results presented in this study point to VC 1 = (36)
jωC1 [Rr + jωL1eq ]
the likely adoption of the LCL topology for a future standardized
implementation of wireless EV charging. jωL1 VS
VL 1 = (37)
Rr + jωL1eq
APPENDIX A 1 1
SYSTEM RMS VOLTAGES AND CURRENTS L1eq = L1 − − . (38)
ω 2 C1 ω 2 Cr (M, Req )
A. LCL Architecture
C. Secondary Side
IB =
(1 − ω 2 L1 e q C1 ) + jωC1 Rr Req M I1
VS (27) I2 = (39)
Rr (1 − ω 2 LB C1 ) + jω[LB (1 − ω 2 L1 e q C1 ) + L1 e q ] jωL2eq + Req (1 − ω 2 L2eq C2 ) L2eq
ESTEBAN et al.: COMPARATIVE STUDY OF POWER SUPPLY ARCHITECTURES IN WIRELESS EV CHARGING SYSTEMS 6421

−Req ω 2 L2eq C2 M I1 [15] S. Lee, J. Huh, C. Park, N.-S. Choi, G.-H. Cho, and C.-T. Rim, “On-line
IC 2 = (40) electric vehicle using inductive power transfer system,” in Proc. IEEE
jωL2eq + Req (1 − ω 2 L2eq C2 ) L2eq Energy Convers. Congr. Expo., Sep. 2010, pp. 1598–1601.
[16] S. Choi, J. Huh, W. Lee, S. Lee, and C. Rim, “New cross-segmented power
ωReq L2 M I1 supply rails for roadway-powered electric vehicles,” IEEE Trans. Power
VL 2 = (41)
ωL2eq − jReq (1 − ω 2 L2eq C2 ) L2eq Electron., vol. 28, no. 12, pp. 5832–5841, Dec. 2013.
[17] J. Sallan, J. Villa, A. Llombart, and J. Sanz, “Optimal design of ICPT
−Req M I1 systems applied to electric vehicle battery charge,” IEEE Trans. Ind. Elec-
VC S 2 =
ω 2 L2eq Cs2 − jωReq Cs2 (1 − ω 2 L2eq C2 ) L2eq tron., vol. 56, no. 6, pp. 2140–2149, Jun. 2009.
[18] M. Chigira, Y. Nagatsuka, Y. Kaneko, S. Abe, T. Yasuda, and A. Suzuki,
(42) “Small-size light-weight transformer with new core structure for con-
tactless electric vehicle power transfer system,” in Proc. IEEE Energy
ωReq L2eq M I1 Convers. Congr. Expo., Sep. 2011, pp. 260–266.
VC 2 = . (43) [19] O. C. Onar, J. M. Miller, S. L. Campbell, C. Coomer, C. White, and
ωL2eq − jReq (1 − ω L2eq C2 ) L2eq
2
L. E. Seiber, “A novel wireless power transfer for in-motion EV/PHEV
charging,” in Proc. IEEE 28th Annu. Appl. Power Electron. Conf. Expo.,
Mar. 2013, pp. 3073–3080.
APPENDIX B [20] O. Onar, J. Miller, S. Campbell, C. Coomer, C. White, and L. Seiber,
“Oak ridge national laboratory wireless power transfer development for
ONLINE MULTIMEDIA OF A WORKING SYSTEM sustainable campus initiative,” in Proc. IEEE Transp. Electr. Conf. Expo.,
Jun. 2013, pp. 1–8.
A. SLC Architecture [21] J. Miller, O. Onar, C. White, S. Campbell, C. Coomer, L. Seiber, R. Sepe,
and A. Steyerl, “Demonstrating dynamic wireless charging of an electric
https://www.youtube.com/watch?v=GNDM4SKEf90. vehicle: The benefit of electrochemical capacitor smoothing,” IEEE Power
Electron. Mag., vol. 1, no. 1, pp. 12–24, Mar. 2014.
B. LCL Architecture [22] O. Onar, S. Campbell, P. Ning, J. Miller, and Z. Liang, “Fabrication
and evaluation of a high performance sic inverter for wireless power
https://www.youtube.com/watch?v=wt_pU8H9GNU. transfer applications,” in Proc. IEEE Wide Bandgap Power Devices Appl.
Workshop, Oct. 2013, pp. 125–130.
[23] H. Wu, A. Gilchrist, K. Sealy, and D. Bronson, “A high efficiency 5
REFERENCES kW inductive charger for EVS using dual side control,” IEEE Trans. Ind.
Informat., vol. 8, no. 3, pp. 585–595, Aug. 2012.
[1] G. Covic and J. Boys, “Inductive power transfer,” Proc. IEEE, vol. 101, [24] K. Hall, M. Kesler, K. Kulikowski, and R. Fiorello. (2010, Nov.
no. 6, pp. 1276–1289, Jun. 2013. 4). Wireless energy transfer between a source and a vehicle. U.S.
[2] A. W. Green and J. T. Boys, “10 kHz inductively coupled power transfer- Patent 12/770 137. [Online]. Available: http://www.google.com/patents/
concept and control,” in Proc. 5th Int. Conf. Power Electron. Variable- US20100277121
Speed Drives, Oct. 1994, pp. 694–699. [25] A. Bohori, A. Kumar, B. Worden, J. Bray, S. Ramachandrapanicker,
[3] G. A. Covic, G. Elliott, O. Stielau, R. M. Green, and J. T. Boys, “The design A. Tiwari, and S. Bhat. (2011, Dec. 29). Power transfer system and
of a contact-less energy transfer system for a people mover system,” in method. U.S. Patent 12/822 232. [Online]. Available: https://www.
Proc. Power Syst. Technol., Int. Conf., 2000, vol. 1, pp. 79–84. google.com/patents/US20110315496
[4] O. Stielau and G. A. Covic, “Design of loosely coupled inductive power [26] A. Bohori, S. Ramachandrapanicker, C. Sihler, A. Tiwari, and S. Bhat.
transfer systems,” in Proc. Power Syst. Technol., Int. Conf., 2000, vol. 1, (2013, May 14). Contactless power transfer system. U.S. Patent 8 441 153.
pp. 85–90. [Online]. Available: https://www.google.com/patents/US8441153
[5] J. Boys and A. Green. (1994, Mar. 8). Inductive power dis- [27] A. Bohori, J. Bray, S. Ramachandrapanicker, and S. Bhat. (2013,
tribution system. U.S. Patent 5 293 308. [Online]. Available: Jul. 24). Contactless power transfer system and method. Eur.
http://www.google.com/patents/US5293308 Patent EP20 110 159 292. [Online]. Available: https://www.google.
[6] C.-H. Hu, C.-M. Chen, Y.-S. Shiao, T.-J. Chan, and L.-R. Chen, “Devel- com/patents/EP2369711A3?cl=en
opment of a universal contactless charger for handheld devices,” in Proc. [28] S. Lukic and Z. Pantic, “Cutting the cord: Static and dynamic inductive
IEEE Ind. Electron., Int. Symp., Jun. 2008, pp. 99–104. wireless charging of electric vehicles,” IEEE Electr. Mag., vol. 1, no. 1,
[7] S. Hui and W. C. Ho, “A new generation of universal contactless bat- pp. 57–64, Sep. 2013.
tery charging platform for portable consumer electronic equipment,” in [29] “Wireless charging research activities around the world [society news],”
Proc. IEEE 35th Annu. Power Electron. Spec. Conf., Jun. 2004, vol. 1, IEEE Power Electron. Mag., vol. 1, no. 2, pp. 30–38, Jun. 2014.
pp. 638–644. [30] N. Kar, K. Iyer, A. Labak, X. Lu, C. Lai, A. Balamurali, B. Esteban,
[8] C.-G. Kim, D.-H. Seo, J.-S. You, J.-H. Park, and B.-H. Cho, “Design of a and M. Sid-Ahmed, “Courting and sparking: Wooing consumers? In-
contactless battery charger for cellular phone,” in Proc. IEEE 15th Annu. terest in the EV market,” IEEE Electr. Mag., vol. 1, no. 1, pp. 21–31,
Appl. Power Electron. Conf. Expo., 2000, vol. 2, pp. 769–773. Sep. 2013.
[9] A. Sample, D. Meyer, and J. Smith, “Analysis, experimental results, and [31] Sae taskforce j2954 on wireless charging and positioning standards look-
range adaptation of magnetically coupled resonators for wireless power ing to have final draft of guideline this year; significant industry involve-
transfer,” IEEE Trans. Ind. Electron., vol. 58, no. 2, pp. 544–554, Feb. ment. (2012, Jan.). [Online]. Available: http://www.greencarcongress.
2011. com/2012/01/j2954-20120122.html
[10] G. Covic and J. Boys, “Modern trends in inductive power transfer for [32] J. Schneider. (2012, Jan.). Saej2954 overview and path forward. SAE.
transportation applications,” IEEE J. Emerg. Sel. Topics Power Electron., [Online]. Available: http://www.sae.org/smartgrid/sae-j2954-status_1-
vol. 1, no. 1, pp. 28–41, Mar. 2013. 2012.pdf
[11] C.-S. Wang, O. Stielau, and G. Covic, “Design considerations for a con-
[33] Sae international task force announces agreement on frequency of
tactless electric vehicle battery charger,” IEEE Trans. Ind. Electron., vol.
operation and power classes for wireless power transfer for its electric and
52, no. 5, pp. 1308–1314, Oct. 2005.
plug-in electric vehicle guideline. SAE. (2013, Nov.). [Online]. Available:
[12] C.-Y. Huang, J. Boys, G. Covic, and M. Budhia, “Practical considerations
http://www.sae.org/servlets/pressRoom?OBJECT_TYPE=PressReleases
for designing IPT system for EV battery charging,” in Proc. IEEE Veh.
&PAGE=sh owRelease&RELEASE_ID=2296
Power Propul. Conf., Sep. 2009, pp. 402–407.
[13] J. Huh, S. Lee, W. Lee, G. Cho, and C. Rim, “Narrow-width inductive [34] M. Chinthavali, O. Onar, J. Miller, and L. Tang, “Single-phase active
power transfer system for online electrical vehicles,” IEEE Trans. Power boost rectifier with power factor correction for wireless power transfer
Electron., vol. 26, no. 12, pp. 3666–3679, Dec. 2011. applications,” in Proc. IEEE Energy Convers. Congr. Expo., Sep. 2013,
[14] J. Huh, W. Lee, G.-H. Cho, B. Lee, and C.-T. Rim, “Characterization of pp. 3258–3265.
novel inductive power transfer systems for on-line electric vehicles,” in [35] C.-S. Wang, G. Covic, and O. Stielau, “Investigating an LCL load resonant
Proc. IEEE 26th Annu. Appl. Power Electron. Conf. Expo., Mar. 2011, inverter for inductive power transfer applications,” IEEE Trans. Power
pp. 1975–1979. Electron., vol. 19, no. 4, pp. 995–1002, Jul. 2004.
6422 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 30, NO. 11, NOVEMBER 2015

[36] J. T. Boys, C.-Y. Huang, and G. A. Covic, “Single-phase unity power- [61] M. Budhia, G. Covic, and J. Boys, “A new IPT magnetic coupler for
factor inductive power transfer system,” in Proc. IEEE Power Electron. electric vehicle charging systems,” in Proc. IEEE 36th Annu. Conf. Ind.
Spec. Conf., Jun. 2008, pp. 3701–3706. Electron. Soc., Nov. 2010, pp. 2487–2492.
[37] H. Hao, G. Covic, M. Kissin, and J. Boys, “A parallel topology for in- [62] H. Wheeler, “Simple inductance formulas for radio coils,” Radio Eng.,
ductive power transfer power supplies,” in Proc. IEEE 26th Annu. Appl. Proc. Inst., vol. 16, no. 10, pp. 1398–1400, Oct. 1928.
Power Electron. Conf. Expo., Mar. 2011, pp. 2027–2034. [63] C. Park, S. Lee, G. Cho, and C. Rim, “Innovative 5-m-off-distance in-
[38] P. Ning, J. M. Miller, O. C. Onar, C. P. White, and L. D. Marlino, “A com- ductive power transfer systems with optimally shaped dipole coils,” IEEE
pact wireless charging system development,” in Proc. IEEE 28th Annu. Trans. Power Electron., vol. 30, no. 2, pp. 817–827, Feb. 2015.
Appl. Power Electron. Conf. Expo., Mar. 2013, pp. 3045–3050. [64] C. Tang, X. Dai, Z. Wang, Y. Sun, and A. Hu, “Frequency bifurcation
[39] J. Miller, C. White, O. Onar, and P. Ryan, “Grid side regulation of wire- phenomenon study of a soft switched push-pull contactless power transfer
less power charging of plug-in electric vehicles,” in Proc. IEEE Energy system,” in Proc. IEEE 6th Ind. Electron. Appl. Conf., Jun. 2011, pp.
Convers. Congr. Expo., Sep. 2012, pp. 261–268. 1981–1986.
[40] R. Mecke and C. Rathge, “High frequency resonant inverter for contactless [65] C.-Y. Huang, A Low Cost ICPT Power Supply. Auckland, New Zealand:
energy transmission over large air gap,” in Proc. IEEE 35th Annu. Power Electrical and Computer Engineering, Univ. of Auckland, 2006
Electron. Spec. Conf., Jun. 2004, vol. 3, pp. 1737–1743. [66] M. Borage, S. Tiwari, and S. Kotaiah, “Analysis and design of an LCL-T
[41] J. Miller, O. Onar, and M. Chinthavali, “Primary-side power flow control resonant converter as a constant-current power supply,” IEEE Trans. Ind.
of wireless power transfer for electric vehicle charging,” IEEE J. Emerg. Electron., vol. 52, no. 6, pp. 1547–1554, Dec. 2005.
Sel. Topics Power Electron., vol. 3, no. 1, pp. 147–162, Mar. 2015. [67] F. Musavi, M. Edington, W. Eberle, and W. Dunford, “Energy efficiency
[42] R. Bosshard, J. Kolar, J. Muhlethaler, I. Stevanovic, B. Wunsch, and in plug-in hybrid electric vehicle chargers: Evaluation and comparison of
F. Canales, “Modeling and η-α-Pareto optimization of inductive power front end ac-dc topologies,” in Proc. IEEE Energy Convers. Congr. Expo.,
transfer coils for electric vehicles,” IEEE J. Emerg. Sel. Topics in Power Sep. 2011, pp. 273–280.
Electron., vol. 3, no. 1, pp. 50–64, Mar. 2015.
[43] H. Hao, G. Covic, and J. Boys, “An approximate dynamic model of LCL-
T-based inductive power transfer power supplies,” IEEE Trans. Power
Electron., vol. 29, no. 10, pp. 5554–5567, Oct. 2014.
[44] C.-Y. Huang, J. James, and G. Covic, “Design considerations for variable
coupling lumped coil systems,” IEEE Trans. Power Electron., vol. 30, no. Bryan Esteban (M’15) received the B.A.Sc. and
2, pp. 680–689, Feb. 2015. M.A.Sc. degrees in electrical and computer engineer-
[45] H. Wu, A. Gilchrist, K. Sealy, P. Israelsen, and J. Muhs, “Design of ing from the University of Windsor, Windsor, ON,
symmetric voltage cancellation control for LCL converters in inductive Canada, in 2011 and 2014, respectively.
power transfer systems,” in Proc. IEEE Int. Electr. Mach. Drives Conf., His research interests include high power and effi-
May 2011, pp. 866–871. ciency power electronics, resonant power conversion,
[46] C.-S. Wang, G. Covic, and O. Stielau, “Power transfer capability and bifur- and inductive power transfer. The work described in
cation phenomena of loosely coupled inductive power transfer systems,” this paper is a first of its kind in Canada and was
IEEE Trans. Ind. Electron., vol. 51, no. 1, pp. 148–157, Feb. 2004. completed by him with the support of his co-authors
[47] J. T. Boys, G. A. Covic, and A. W. Green, “Stability and control of over a period spanning January 2013 to August 2014.
inductively coupled power transfer systems,” Electr. Power Appl., IEE During this time, he worked as a Lead Researcher of
Proc., vol. 147, no. 1, pp. 37–43, Jan. 2000. wireless EV charging systems at the Center for Hybrid Automotive Research
[48] Z. Pantic, S. Bai, and S. Lukic, “ZCS LCC—Compensated resonant in- and Green Energy.
verter for inductive-power-transfer application,” IEEE Trans. Ind. Elec-
tron., vol. 58, no. 8, pp. 3500–3510, Aug. 2011.
[49] G. Covic, J. T. Boys, A. M. W. Tam, and J. C. H. Peng, “Self tuning pick-
ups for inductive power transfer,” in Proc. IEEE Power Electron. Spec.
Conf., June 2008, pp. 3489–3494. Maher Sid-Ahmed received the B.A.Sc. degree from
[50] N. Keeling, G. Covic, and J. Boys, “A unity-power-factor IPT pickup for the University of Alexandria, Alexandria, Egypt, in
high-power applications,” IEEE Trans. Ind. Electron., vol. 57, no. 2, pp. 1968, and the M.A.Sc. and Ph.D. degrees from the
744–751, Feb. 2010. University of Windsor, Windsor, ON, Canada, in
[51] G. Elliott, S. Raabe, G. Covic, and J. Boys, “Multiphase pickups for large 1971 and 1973, respectively.
lateral tolerance contactless power-transfer systems,” IEEE Trans. Ind. He is the Author of a book entitled “Image Pro-
Electron., vol. 57, no. 5, pp. 1590–1598, May 2010. cessing, Theory, Algorithms and Architectures” (New
[52] M. Budhia, G. Covic, and J. Boys, “Design and optimization of circular York, NY, USA: McGraw-Hill, 1995). He holds four
magnetic structures for lumped inductive power transfer systems,” IEEE U.S. patents and has published 80 journal papers. Su-
Trans. Power Electron., vol. 26, no. 11, pp. 3096–3108, Nov. 2011. pervised and graduated more than 60 M.A.Sc. and
[53] R. Steigerwald, “A comparison of half-bridge resonant converter topolo- Ph.D. students. His research interest includes image
gies,” IEEE Trans. Power Electron., vol. 3, no. 2, pp. 174–182, Apr. 1988. processing, machine vision, meteorology using stereo vision, systolic struc-
[54] H. Ma and W. Zhou, “Modeling a current source push-pull resonant con- tures, hardware realization for real-time processing, OCR including hand writ-
verter for loosely coupled power transfer systems,” in Proc. IEEE 30th ing recognition, and now wireless electricity.
Annu. Conf. Ind. Electron. Soc., Nov. 2004, vol. 2, pp. 1024–1029.
[55] C.-S. Wang, O. Stielau, and G. Covic, “Load models and their application
in the design of loosely coupled inductive power transfer systems,” in
Proc. Power Syst. Technol., Int. Conf., 2000, vol. 2, pp. 1053–1058.
[56] C.-S. Wang, G. Covic, and O. Stielau, “General stability criterions for zero
phase angle controlled loosely coupled inductive power transfer systems,” Narayan C. Kar (S’97–M’00–SM’07) received the
in Proc. IEEE 27th Annu. Conf. Ind. Electron. Soc., 2001, vol. 2, pp. B.Sc. degree in electrical engineering from the
1049–1054. Bangladesh University of Engineering and Technol-
[57] M. Budhia, J. Boys, G. Covic, and C.-Y. Huang, “Development of a single- ogy, Dhaka, Bangladesh, in 1992, and the M.Sc. and
sided flux magnetic coupler for electric vehicle IPT charging systems,” Ph.D. degrees in electrical engineering from the Ki-
IEEE Trans. Ind. Electron., vol. 60, no. 1, pp. 318–328, Jan. 2013. tami Institute of Technology, Hokkaido, Japan, in
[58] A. Zaheer, D. Kacprzak, and G. Covic, “A bipolar receiver pad in a 1997 and 2000, respectively.
lumped IPT system for electric vehicle charging applications,” in Proc. He is currently an Associate Professor at the Elec-
IEEE Energy Convers. Congr. Expo., Sep. 2012, pp. 283–290. trical and Computer Engineering Department, Uni-
[59] M. Kissin, J. Boys, and G. Covic, “Interphase mutual inductance in versity of Windsor, Windsor, ON, Canada, where he
polyphase inductive power transfer systems,” IEEE Trans. Ind. Electron., holds the Canada Research Chair position at the Elec-
vol. 56, no. 7, pp. 2393–2400, Jul. 2009. trified Transportation Systems. His research interests include the analysis, de-
[60] M. Budhia, G. Covic, and J. Boys, “Design and optimization of circular sign, and control of electrical machines for electric and hybrid electric vehicles,
magnetic structures for lumped inductive power transfer systems,” IEEE charging systems, testing, and performance analysis of batteries and develop-
Trans. Power Electron., vol. 26, no. 11, pp. 3096–3108, Nov. 2011. ment of optimization techniques for hybrid energy management system.