IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 29, NO.

8, AUGUST 2014 4403

An Adaptive Impedance-Matching Network Based on

a Novel Capacitor Matrix for Wireless Power Transfer
Yongseok Lim, Member, IEEE, Hoyoung Tang, Student Member, IEEE, Seungok Lim,
and Jongsun Park, Senior Member, IEEE

AbstractIn a wireless power transfer (WPT) system via the magnetic waves as a medium for power transfer, and have en-
magnetic resonant coupling, one of the most challenging design abled power transfer over several tens of meters with a low effi-
issues is to maintain a reasonable level of power transfer efficiency ciency. Although the technology has been applied to 900 MHz
(PTE), even when the distance between the transmitter and the
receiver changes. When the distance varies, the PTE drastically UHF RFID, the WPT efficiency is quite dependent on direc-
decreases due to the impedance mismatch between the resonator tionality and the atmosphere [1], [2]. Inductive coupling WPT
of the transmitter and that of the receiver. This paper presents systems exploit magnetic induction between a power transmitter
a novel serial/parallel capacitor matrix in the transmitter, where coil and a receiver coil, and the achievable transfer efficiency is
the impedance can be automatically reconfigured to track the op- over 90%, but only within several centimeters [3], [4].
timum impedance-matching point in the case of varying distances.
The dynamic WPT matching system is enabled by changing the Since it supports a longer transfer distance, the magnetic
combination of serial and parallel capacitors in the capacitor ma- resonant coupling WPT approach, which utilizes the resonance
trix. An interesting observation in the proposed capacitor matrix phenomenon between the power transmitter coil and the receiver
is that the resonant frequency is not shifted, even with capacitor- coil, is more attractive than inductive coupling WPT [5][9]. In
matrix tuning. In order to quickly find the best capacitor combina- the design of magnetic resonant coupling WPT systems, one
tion that achieves maximum power transfer, a window-prediction-
based search algorithm is also presented in this paper. The of the most significant issues is to maintain power transfer ef-
proposed resonance WPT system is implemented using a reso- ficiency (PTE) when the distance between the transmitter and
nant frequency of 13.56 MHz, and the experimental results with receiver changes. When the distance changes, the PTE between
1W power transfer show that the transfer efficiency increases up the resonant coils of the transmitter and receiver drastically de-
to 88% when the distance changes from 0 to 1.2 m. creases due to impedance mismatch. In general WPT system,
Index TermsCapacitor matrix, impedance matching, the impedance increases when the transmitter and the receiver
impedance searching, magnetic resonant coupling, wireless power are brought closer together. The increasing impedance leads to
transfer (WPT). an increasing reflection wave, which gives rise to large amount
of power reflected back to the transmitter.When the distance in-
creases, the impedance decreases to zero, which induces a large
I. INTRODUCTION
current and can damage the transmitter circuits [10].
ITH a recent increas in demand for wireless battery Many previous research works on magnetic resonant coupling
W charging and its broad applications such as charging
mobile or medical devices and electric vehicles (EVs), wire-
WPT have focused on wireless power transfer efficiency com-
pensation with dynamically changing distances [6], [12][20].
less power transfer (WPT) has attracted great attention. WPT Previous solutions include 1) tuning capacitance by using vari-
technologies are broadly divided into three categories: radio fre- able capacitor [12][14] or preset capacitor blocks correspond-
quency (RF)-based, inductive coupling, and magnetic resonant ing to a few selected distance ranges [15]. 2) Adjusting the
coupling. RF-based WPT systems utilize propagating electro- mutual inductance using four-coils system [6], [18], and 3) op-
timizing the resonant frequency for minimum reflection [19],
[20]. Although adaptive frequency tuning approaches [19], [20]
are simple methods for maintaining power transfer efficiency, it
Manuscript received July 11, 2013; revised October 1, 2013; accepted has limited applications since the resonant frequency that is de-
November 13, 2013. Date of current version March 26, 2014. This work was
partially supported by National Research Foundation Grant funded by the Korea cided after impedance tuning, should be placed within allowed
Government (2011-0020128). This work was also supported by IT R&D pro- bandwidth. The adaptive frequency tuning method also needs
gram of KEIT funded by the Korea Government (10041749). Recommended additional circuitry at the transmitter and receiver [20].
for publication by Associate Editor S. Y. (Ron) Hui.
Y. Lim is with the School of Electrical Engineering, Korea University, Seoul This paper presents a novel-matching approach and dy-
136-701, Korea. He is also with Korea Electronics Technology Institute, Seoul namic impedance tracking, where the capacitor matrix and real-
463-816, Korea (e-mail: busytom@keti.re.kr). time impedance tracking algorithm are efficiently exploited to
H. Tang and J. Park are with the School of Electrical Engineering, Ko-
rea University, Seoul 136-701, Korea (e-mail: ho-2604@korea.ac.kr; jongsun achieve the maximum PTE. In the proposed WPT system, the
@korea.ac.kr). capacitor matrix can dynamically adjust the impedance mis-
S. Lim is with Korea Electronics Technology Institute, Seoul 463-816, Korea matches without changing the resonant frequency. The orig-
(e-mail: solim@keti.re.kr).
Color versions of one or more of the figures in this paper are available online inal contributions of this paper can be summarized as fol-
at http://ieeexplore.ieee.org. lows: 1) This paper proposes an automatic impedance-matching
Digital Object Identifier 10.1109/TPEL.2013.2292596 structure based on a novel capacitor matrix. Compared to the

0885-8993 2013 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.
4404 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 29, NO. 8, AUGUST 2014

Fig. 1. Block diagram of a WPT transceiver system.

previous works offering impedance matching through the vari-

able capacitor (varicon) [13], [14], the proposed capacitor ma-
trix can cover wide range of capacitance values with higher
resolution and low resistance. Our approach also offers higher
resolution to rematch impedance with continuously varying dis-
tances compared to the discrete capacitor-based method [15]. 2)
A new efficient search algorithms for finding the best capacitor
combination that achieves PTE without frequency shifting are
presented. 3) The proposed WPT system is implemented and
verified to provide 128 searching steps using seven capacitors
and relay switches. It is shown that the proposed WPT system
works seamlessly with a varying distance.
The rest of the paper is organized as follows. Section II
shows how the power efficiency degrades with changes in dis- Fig. 2. WPT simulation setup: source coil (transmitter side), load coil (receiver
tance. Through circuit-level analysis, the conditions for per- side) and parameter table. The source and load coils have the same parameters.
fect impedance rematching are also derived in Section II.
Section III presents the proposed capacitor matrix, which can be
dynamically reconfigured for impedance rematching with vary-
ing distance. Efficient search algorithms for finding the optimal
capacitance switch configurations are presented in Section IV.
Section V shows the numerical results of the implemented WPT
matching system. Finally, conclusions are drawn in Section VI.

II. IMPEDANCE ANALYSIS WITH DISTANCE CHANGE

A block diagram of a WPT transceiver system that contains
an impedance matching mechanism, is shown in Fig. 1. The
transmitter is composed of a power amplifier, matching network,
and antenna. The receiver is composed of a receiving antenna,
matching network, rectifier, dcdc converter, and load. In this
section, we present the simulation results on how the impedance
of the transceiver and power transfer efficiencies are changed Fig. 3. S-parameter simulations with matching impedance and mismatching
cases.
with varying distances. An interesting circuit-based impedance
analysis is also presented.

Under the initial conditions, the bold line of Fig. 3 indicates that
A. Effect of Distance Change to Maximum Power Transfer the power is transferred well with S11 equal to 38 dB. The
During WPT with magnetic resonators (antennas), when the transfer distance is changed from 0.6 to 0.2 m by moving the
impedances of the transmitter and receiver are not perfectly receiver coil. As a result, the impedance (Zin ) is increased to
matched, the system never reaches the achievable maximum 785.1 , and S21 is changed to 15 dB, which means that the
PTE. Simulation results on power transmission loss due to dis- power was not properly transferred using 13.56 MHz.
tance change are in this section. The simulations were performed The power efficiency degradation can also be observed by
using Agilent advanced designsystem (ADS), and the simula- measuring the current at the transmitter side. Fig. 4 shows the
tion setup and parameters are specified in Fig. 2. current measured with the transfer distances of 0.2 and 0.6 m.
Initially, with a distance of 0.6 m between the transmitter and There are large currents under the initial impedance matching
the receiver, the impedance seen from the power AMP (Zin in conditions with a transfer distance of 0.6 m (bold line in Fig. 4),
Fig. 1) is matched to 50 , and 1 Wpower is transmitted from but when the distance is changed to 0.2 m, the measured current
the power AMP using the resonant frequency of 13.56 MHz. becomes small.
LIM et al.: ADAPTIVE IMPEDANCE-MATCHING NETWORK BASED ON A NOVEL CAPACITOR MATRIX FOR WIRELESS POWER TRANSFER 4405

Fig. 4. Current waves measured at the transmitter with distances of 0.2 and
0.6 m.

B. Impedance Matching Approaches

The impedance matching approaches in the magnetic res-
onant coupling WPT at fixed resonant frequency are broadly
categorized into two types: mutual inductance tuning and capac-
itance tuning methods. In order to compare these two matching
approaches, simulations with an ADS are performed again. As
presented in the previous section, the impedance is initially
matched with a distance of 0.6 m, and then the transfer distance
is varied from 0.2 to 0.8 m with steps of 0.2 m. As shown in
Fig. 5 (a), the even mode and odd mode frequency bands (fre-
quency split) are generated with 0.2-m distance, and the power
is not properly transferred. The space between the even mode
and odd mode band becomes smaller with well-linked resonant
coupling.
Fig. 5 (b) shows the simulation results when the impedance is
rematched using the capacitance tuning approach. Even and odd
modes disappear, and the energy is transferred at the resonant Fig. 5. S 1 1 - parameter plots with varying distances of 0.8, 0.6 (initial matching
distance), 0.4, and 0.2 m. (a) nonmatching case (b) rematching with capacitance
frequency of 13.56 MHz. S11 is improved to 60 dB with tuning, and (c) rematched with mutual inductance tuning.
a 0.2 m distance despite slight frequency shift, and power is
transferred well even with a 0.8 m distance. Fig. 5 (c) shows the
results for when the impedance is rematched using the mutual
inductance tuning approach. When the distances are changed to
0.8 and 0.4 m, the center frequencies are shifted to 13.44 and
13.99 MHz after rematching, respectively. In the case of the
0.2 m distance, S11 is around 50 dB, with serious frequency
shifting.
The frequency shift shown above occurs since the distance
change between two coils changes the inductances of the res-
onators [21]. More specifically, as the distance between two
coils decreases, mutual inductance between the coils increases, Fig. 6. Equivalent circuit model between transmitter and receiver.
which leads to the resonant frequency shift. At the same time,
since the impedance is altered, frequency split is occurred, and
power transfer efficiency is decreasing at the resonant frequency simpler for impedance rematching with varying distances. In this
of the resonators within the overcoupled region [10]. When the section, based on the capacitance tuning approach, we present
impedance is rematched, the split frequencies are merged in to an interesting circuit-level analysis on the impedance changes.
one, but on the shifted locations. Fig. 6 presents an equivalent circuit model of WPT transceiver
shown in Fig. 1. On the transmitter side, the power amplifier
and power source can be modeled as a current source parallel
C. Impedance Analysis From Circuit Perspective
with an impedance using Nortons theorem [22]. The matching
In the previous section, it is shown that the modification of network is also represented with a combination of serial (CS 1 )
the capacitance in the transmitter side is more efficient and and parallel (CP 1 ) capacitances. The impedance seen in the
4406 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 29, NO. 8, AUGUST 2014

power amplifier (Zin ) should be perfectly matched with Z0 (=

50 ) for maximum PTE. The matching network on the receiver
side can also be modeled as a combination of parallel (CP 2 ) and
serial (CS 2 ) capacitances, and the rest of the receiver block is
simply characterized as a load impedance (ZL ).
In the following, based on the circuit analysis, it will be
shown how the impedance (Zin ) is affected by the mutual in-
ductance (M12 ) changes mainly due to the varying distance
between the transmitter and receiver. Later, we also show that
Fig. 7. Effect of changing distance on the power transmitter side, which can
the capacitances in the transmitter matching network (CS 1 and be modeled as the change of Z M .
CP 1 ) can be efficiently reconfigured to rematch the impedance,
finally reaching the condition: Z0 = Zin = 50 under varying
distance. For the above (12) on ZM to be valid, the following two equa-
In Fig. 6, the impedance Zin is represented as follows: tions should be satisfied:

  M = M12 = M21 = k12 L1 L2 (13)
1 i1 1 V1
= = jwCP 1 1 . (1)
Zin Va Z0 Va 1 w2 L2 (CS 2 + CP 2 ) = 0 (14)
The following two equations can be derived on the receiver side: where 12 represents the coupling coefficient between two coils,
and M is the mutual inductance. (14) also show the resonant
V2 = jwL2 iL 2 + jwM iL 1 (2) condition for two coils. Accordingly, the impedance Zin can be
V2 expressed as follows:
iL 2 + jwCP 2 V2 + 1 = 0. (3)  
jw CS 2 + ZL 1 1
Zin =  (jwL1 + ZM ) + (15)
Let us refer the admittance Y2 as follows: jwCP 1 jwCS 1
  
1 1 w2 k12
2
L1 L2 1
Y2 = jwCP 2 + . (4) =  jwL1 + + . (16)
1
+ ZL jwCP 1 ZL jwCS 1
jw CS 2
From (16), Zin can be expressed using ZM , which is also a
Then, (3) can be written as
function of the mutual inductance, M12 . In other words, when
iL 2 + Y2 V2 = 0. (5) the distance changes, the mutual inductance M12 is changed,
which gives rise to changes in the impedance ZM . The modified
From the aaforementioned equations, V2 can be expressed with ZM value results in the changes in Zin , thus incurring impedance
the current of the power receiver and the admittance Y2 as mismatch and serious PTE degradation. Fig. 7 illustrates the
iL 2 equivalent circuit, where a new impedance component ZM is
V2 = . (6) added to the transmitter side meaning that the modifications of
Y2
mutual inductance M12 can be modeled as the changes in ZM
Now, (2) can be written as in the transmitter.
iL 2 In order to compensate for the changes in ZM due to distance
jwL2 iL 2 + jwM iL 1 + = 0. (7) change, we can change the capacitance CP 1 and CS 1 such that
Y2
the impedance (Zin ) is rematched to Z0 (= 50 ). When we
From (7), Z2 is defined as change (16) into the real part of the impedance R and the imag-
1 inary part X, Zin can be expressed as
Z2 = jwL2 + (8)  
Y2 1 1
Zin = R + jX = +  (jwL1 + ZM )
jwM iL 1 = Z2 iL 2 . (9) jwCS 1 jwCP 1
Using (9), V1 can be expressed as follows: (17)
    ZM
jwM =
V1 = jwL1 iL 1 + jwM iL 1 (10) 1 2w2 CP 1 L1 + w2 CP2 1 ZM
2 + w 4 C 2 L2
P1 1
Z2 
 2 2
 wL1 w3 CP 1 L21 wCP 1 ZM2
w M +j
= jwL1 + iL 1 . (11) 1 2w CP 1 L1 + w CP 1 ZM + w CP2 1 L21
2 2 2 2 4
Z2

Here, ZM can be written as 1
. (18)
w2 M 2 w2 k12
2
L1 L2 w2 k12
2
L1 L2 wCS 1
ZM = = 1w 2 L
= .
Z2 2 (C S 2 +C P 2 ) ZL
(1w 2 L 2 C P 2 )(j w C S 2 ) + ZL In (18), for perfect impedance matching, the resistive term R
(12) should be equal to 50 , and the reactance X should be zero.
LIM et al.: ADAPTIVE IMPEDANCE-MATCHING NETWORK BASED ON A NOVEL CAPACITOR MATRIX FOR WIRELESS POWER TRANSFER 4407

Fig. 8. (a) M N capacitor matrix circuit. (b) Example of a capacitor matrix with seven capacitances. (c) Equivalent circuit of (b).

TABLE I
OPTIMAL CAPACITANCE SWITCH CONFIGURATION FOR VARIOUS DISTANCES (RANGING FROM 0 M TO 0.8 M) WHEN CAPACITOR MATRIX WITH SEVEN
DIFFERENT CAPACITANCE IS USED

When the resistance term is adjusted to 50 by modifying

CP 1 , the reactance X is affected, which may result in a resonant
frequency shift. Since we have a CS 1 term in the reactance term
X, the modified reactance X due to the change in CP 1 can be
compensated by changing CS 1 . More detailed relations between
CP 1 and CS 1 for making the reactance zero are presented in the
next section.

III. IMPEDANCE MATCHING BASED ON THE PROPOSED

CAPACITOR MATRIX
The previous section has shown that the impedance mismatch
due to the distance change can be compensated by controlling
the capacitance of the matching network in the transmitter. In
this section, we present a novel capacitor matrix, where the
Fig. 9. Block diagram of transmitter for real-time impedance match-
combination of serial and parallel capacitances can be dynami- ing/tracking system.
cally reconfigured to resolve the impedance mismatch. One of
the interesting observations is that the resonant frequency is not
changed when the proposed capacitor matrix is used to track the
C4 + C5 + C6 , and CP 1 has a value of C1 + C2 + C7 . Using
impedance change.
the proposed capacitor matrix, the total impedance values can
be dynamically changed by controlling the switches.
A. Proposed Capacitor Matrix for a Matching Network Based on the matrix with seven capacitors shown in Fig. 8
The M N capacitor matrix, where each of the capacitor can (b), where there are 128 combinations available, we have tried
be connected either in parallel or serial according to the switch the impedance rematch process with varying distances based on
array selection, is presented in Fig. 8 (a). For N capacitances the experimental setup specified in Section II-A. Table I shows
with switches, there are 2N different values of CP 1 and CS 1 . the optimal capacitance switch selections based on seven dif-
As an example, Fig. 8 (b) shows a capacitor matrix with seven ferent capacitance values (C1 , C2 , . . . , C7 ) for seven different
different capacitances. With the switch configuration presented distances (0 m, 0.1 m, 0.2 m, . . . , 0.8 m). For each of the capac-
in Fig. 8 (b), the equivalent circuit can be simplified into the itance combinations with different distances, Zin in Fig. 1 (the
one in Fig. 8 (c), where CS 1 has the capacitance value of C3 + impedance seen from the power amplifier) is set close to 50 .
4408 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 29, NO. 8, AUGUST 2014

our case). On the other hand, when the distance increases, ZM

and CS 1 become smaller.

IV. PROPOSED MATCHING TRACKING ALGORITHM BASED ON

THE CAPACITOR MATRIX

A serial/parallel capacitor matrix is presented in the previ-

ous section, where the impedance can be dynamically reconfig-
ured to track the optimum impedance matching point in case of
transfer distance change. In the capacitor matrix, if we have a
large number of capacitances, good resolution on the rematched
impedance can be achieved, however, the time taken to find the
optimum capacitance configuration increases. In this section, we
first show a hardware block diagram for real-time impedance
Fig. 10. Example of Opt capset location change when the distance is
changed from 0.1 to 0.2 m for seven capacitor matrix case shown in
tracking, and then present three different search algorithms for
Table I. finding the optimum capacitance configuration, which are linear
search, binary search, and Effective Window Prediction (EWP)
search. It is also shown that the EWP search algorithm shows
For example, for the impedance rematch with 0.2 m distance, superior performance over the other approaches.
the CS 1 and CP 1 values are supposed to be 30 and 97 pF, respec-
tively. To realize the specific CS 1 and CP 1 values, C3 , C4 , C5 ,
and C6 are serially connected, while C1 , C2 , and C7 should be A. Hardware Block Diagram for Real-Time
connected in parallel, as shown in Fig. 8 (b) and (c). Impedance Tracking
As shown in Fig. 9, the impedance matching condition can
B. Conditions for the Same Resonant Frequency be monitored by measuring the current on the power trans-
mitter side. With the current monitored, we can figure out
When the proposed capacitor -matrix is reconfigured to follow how the impedance values between the source and load coil
the impedance change, one of the key issues is that the resonant should be compensated to reach the matching impedance. In
frequency should remain the same even though the impedance our impedance rematching/tracking system, which is shown in
changes. In order to maintain the resonance, the reactance part, Fig. 9, after the current monitoring circuit detects the current
jX in (17), should be zero which can be expressed as change, the sensed current is sampled as a digital signal using
analog to digital convertor (ADC). A microcontrol unit (MCU)
ZM2
wCP 1 + wL1 w3 L21 CP 1 1
2 = wC
1 2w L1 CP 1 + w4 L21 CP2 1 + w2 CP2 1 ZM
2
(19) computes the required impedance value based on the digitized
S1
sensing current. According to the estimated impedance value
where CS 1 represents the serial capacitance, CP 1 is the paral- from the MCU, the switches in the capacitor matrix are recon-
lel capacitance, L1 is the inductance, and ZM represents the figured to reach the optimal impedance point.
impedance induced by the mutual inductance shown in Fig. 7.
To satisfy the condition shown in (19), the following equations B. Impedance Matching Tracking Algorithm
should be satisfied: In the proposed capacitor matrix, as the number of capacitors
1 in the matrix increases, the number of serial (CS 1 ) and paral-
2 = L1 (CS 1 + CP 1 ) (20)
w lel (CP 1 ) capacitor combinations are exponentially increasing,
CP 1 which incurs considerable delay to find the best capacitor com-
ZM
2
0. (21) bination that achieves the maximum PTE. This section presents
L1
capacitance search algorithms to find the optimal capacitance
As shown in (20), in order to have the same resonant frequency configuration with varying distances. In the following, capset
before and after the distance change, the sum of the total ca- (CS 1 , CP 1 ) refers to the capacitance switch setting that has
pacitance in the N CS 1 + CP 1 , should
matrix, be kept the same, the serial capacitance of CS 1 and the parallel capacitance CP 1 .
which means M i=1 C
j =1 ij = 1/ 2
L, where and L mean Opt capset means the capset(CS 1 , CP 1 ) that makes the char-
the resonant frequency and the inductance of the transmitter, acteristic impedance (Zin in Fig. 1) closest to the system charac-
respectively. As shown in Table I, the CS 1 + CP 1 values are teristic impedance (50 ). The capacitance searching problem
set to 127 pF to maintain the same resonant frequency. The is formulated as follows:For N capacitors in a capacitor ma-
physical meaning of (21) is that as the distance between the trix, find Opt_capset with varying distance between the trans-
transmitter and receiver becomes smaller, since ZM is getting mitter and the receiver. In order to understand the changes
larger, CP 1 in the capacitor matrix should be ignorably small to of the Opt_capset locations with varying distances, as an ex-
satisfy (21). Here, since CP 1 is very small, CS 1 becomes larger ample, let us consider a simple capacitance matrix with seven
(the switches configuration in the capacitor matrix is changed capacitors [see Fig. 8 (b)]. The value of each capacitance is
to increase CS 1 ) as CP 1 + CS 1 should be constant (127 pF in specified in Table I, and the distance varies from 0 to 0.8 m.
LIM et al.: ADAPTIVE IMPEDANCE-MATCHING NETWORK BASED ON A NOVEL CAPACITOR MATRIX FOR WIRELESS POWER TRANSFER 4409

Fig. 11. Opt capset points plots with varying distances from 0 to 0.8 m for
seven capacitor matrix case shown in Table I.

Initially, as shown in Fig. 10, Opt_capset is set to (61 pF, 66 pF)

with a 0.1 m distance, where the characteristic impedance is
49.72 . When the distance is changed to 0.2 m, the impedance
(Zin in Fig. 1) is changed to 12.24 , and the search algo-
rithm should be able to quickly find a new Opt capset, (30 pF,
97 pF), that has a characteristic impedance closest to 50
(50.34 ). Fig. 11 shows the Opt capset points with distance
varying from 0 to 0.8 m. The figure shows that the serial ca-
pacitance, CS 1 , becomes larger with shorter distance, while the
parallel capacitance, CP 1 , increases at greater distances. The
value of CS 1 + CP 1 is fixed to 127 pF to guarantee the same
resonant frequency of 13.56 MHz before and after the distance
change. Therefore, in this capacitor matrix, all the capset can
be expressed as (n, 127-n), n = 1, 2, . . . , 126.
Linear search: The easiest way to find the Opt_capset point
is to check all the possible points until the impedance is closest
to the system characteristic impedance (50 ). In case of the As an example, Fig. 12 (a) shows the | Zin -Z0 | plots, where
example shown in Figs. 10 and 11, we can simply start checking Zin is the impedance of capset (n,127-n), n = 1, 2, . . . , 126
from point (126 pF, 1 pF) to (1 pF, 126 pF), since we have in the case of the seven capacitor matrix shown in Table I. In
seven capacitors in the matrix. The cases of (127 pF, 0 pF) and the first stage of the example, since |Z(capset(1,126))- Z0 ||
(0 pF, 127 pF) are excluded from the search since these cases Z(capset(63,64))Z0 || Z(capset(126,1))Z0 |, the next re-
make the impedance zero or infinite. Generally, for N switchable gion of iteration starts from capset(63, 64) and ends with
capacitors in the capacitor matrix, we have a maximum of 2N capset(126,1). The iteration continues until the algorithm finds
points, and the worst case time complexity is O(2N ). opt_capset. A more complex example is shown in Fig. 12 (b).
Binary search: A binary search algorithm that iteratively dis- Since the impedance differences of the start, middle, and end
cards half of the data region to search for the Opt_capset point is points shows having a minimum in the middle type, the re-
an alternative and faster choice. The algorithm generally com- gion of the second iteration has the start, middle, and end
pares three values of the start, end, and middle points of the point of capset(32,95), capset(63,64), and capset(95,31), re-
data array, and makes a decision on which half of the array is to spectively. The new start and end points are derived from the
be searched. The iterations proceed until the remaining region following two equations: the next start point =(start+mid)/ 2
contains only one target value. = 32, and the next end point = (mid+end)/ 2 = 95. The third
In the WPT system, since the target impedance Z0 is 50 , we iteration region is determined to range from capset(1,126) to
will find the opt_capset point, where the difference between the capset(32,95) according to the continuously increasing type of
characteristic impedance Zin (see Fig. 1) and Z0 , | Zin Z0 | the second iteration, which is based on the following equa-
is minimum. Let us refer to the impedance of the start point of tions: the next start point= val(start(endmid)) = 1 and
the array as Z(start). Z(mid) and Z(end) mean the impedances the next end point= val(start(endmid)+(midstart)) = 32.
of the middle and end points, respectively. Assuming that x1 Since the third iteration shows continuously decreasing type,
= |Z(start)Z0 |, x2 =|Z(mid)Z0 | and x3 =|Z(end)Z0 |, the start and end points of the fourth iteration are capset(32,95)
the three values would be either continuously decreasing (x1 and capset(47,80), respectively. The continuously increasing
x2 x3), continuously increasing (x1 x2 x3), or having type of the fourth iteration makes the region of the fifth stage
a minimum in the middle (x1 x2, x3 x2). In the binary range from capset(25,102) to capset(32,95) in the same man-
search, the next region of the iteration can be decided according ner. The pseudocode of described binary search is presented in
to the type of x1, x2, and x3 values. Algorithm I. The time complexity of the binary search is
4410 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 29, NO. 8, AUGUST 2014

TABLE II
EXAMPLE OF LOOKUP TABLE FOR EWP SEARCH ALGORITHM (SEVEN-CAPACITOR MATRIX CASE SHOWN IN TABLE I)

Fig. 12. Examples of binary search algorithm for seven-capacitor matrix case
specified in Table I. (a) Continuously decreasing case in the first stage and
(b) having minimum in the middle case.

O(log2 (2N )) = O(N ), where N is the number of capacitances

in the matrix. algorithm in pseudocode. The time complexity of three algo-
Effective Window Prediction (EWP) search: The basic idea rithms is compared with different number of capacitors, N , and
of the EWP search algorithm is that if we can figure out the Fig. 13 shows the experimental results. The figure shows the
impedance value by checking the first capset point (2N 1, 1), performance advantage of binary search and the EWP search
the distance can be estimated, and the range of the capsets to increases exponentially with larger N .
be searched can be reduced by looking up the effective window.
An example of the effective window with the seven matrix spec-
V. IMPLEMENTATION OF THE PROPOSED WPT MATCHING
ified in Table II is presented. For example, when the measured
impedance at the first capset(126,1) is about 22 , with the SYSTEM AND EXPERIMENTAL RESULTS
lookup table, the search range is reduced from capset(81,46) to The proposed WPT system based on the capacitor matrix is
capset(88,39). Once it goes into the effective window, the binary implemented, and the system photography is shown in Fig. 14.
search algorithm can be applied iteratively, and the worst case A total of seven capacitors is used in the capacitor matrix to
time complexity is 1 + log2 (window size). The lookup table realize 128 different CS 1 and CP 1 configurations, and the sum
used for the EWP search method is calculated in advance and of all of the capacitances is around 127 pF. In our implementa-
saved to MCU memory. Algorithm II shows the EWP search tion, a negative-positive-zero (NPO) type capacitor having low
LIM et al.: ADAPTIVE IMPEDANCE-MATCHING NETWORK BASED ON A NOVEL CAPACITOR MATRIX FOR WIRELESS POWER TRANSFER 4411

Fig. 13. Worst-case searching step comparisons for different numbers of ca-
pacitors in the matrix.

Fig. 15. Comparison between the simulation and experimental results using
rematched impedance when the distance is initially set to 0.6 m (c) and moved
to (a) 0.2 m, (b) 0.4 m, (d) 0.8 m.

on reducing the number of searching steps, the number of idle

times after capacitor-matrix readjustments can be significantly
reduced.
The measured power efficiency comparisons between the
unmatched WPT system and our proposed adaptive matching
network-based WPT system for various distances are presented
in Fig. 17. Both of the WPT systems have 1 W of transmission
Fig. 14. Real-time impedance matching/tracking WPT system that we power between the transmitter AMP output and the receiver an-
implemented.
tenna. The power transfer efficiency is obtained based on the
power measured at the resonators of the transmitter and the re-
equivalent series resistance (ESR) and relay switches are used ceiver [10]. In the unmatched WPT system, when the distance
to realize the capacitor matrix. For power transfer, an RF Power is initially set to 0.6 m, the power transfer efficiency is around
MOSFET (MRF282S) [23] is employed to implement the class- 68%, however, the efficiency drops abruptly at other distances
E amplifier, and an MSP430 RISC MCU [24] is also used to due to impedance mismatch. In the proposed capacitor-matrix
monitor the impedance and control the capacitor matrix. based WPT system, the power efficiency is maintained within
In our experiment, the impedance observed in the power AMP a reasonable level. When the distance is less than 0.6 m, more
(Zin ) is initially matched to 53.4 with a 0.6 m distance. than 70% efficiency can be achieved. Though the power trans-
Then, the distance is changed from 0.2 to 0.8 m by 0.2 m step. fer efficiency is degraded with increasing distance due to power
The measured S11 values from the rematched impedance are transfer propagation properties, the improvement of the power
compared with the simulation results in Fig. 15, which shows transfer efficiency over the distance from 0 to 0.8 m is consis-
that when the impedance is first matched at 0.6 m, the return loss tently achieved as shown in the Fig. 17.
at the center frequency is close to 40 dB. Later, the receiver In order to compare the impedance tracking perfor-
coils are moved to 0.2, 0.4, and 0.8 m, where the return losses mance among the different search algorithms presented in
are changed to 50, 45, and 30 dB, respectively, after the Section IV-B, we performed experiments according to the
impedance rematch with the capacitance reconfigurations. scenarios illustrated in Fig. 18 (a). With an initial match-
The operation process when the distance between the trans- ing distance of 0.6 m, the distance is changed to 0 m at
mitter and receiver changes, is illustrated in Fig. 16. As shown in MOVE I, and it is changed again to 0.6 m at MOVE II. In
the block diagram of the transmitter (see Fig. 9), MCU periodi- the four cases (MOVE I to MOVE VI) specified in Fig. 18
cally monitors the current and calculates the impedance. When (b), the impedance is rematched using three different track-
the updated impedance becomes out of range, MCU decides to ing approaches, and the tracking performance is compared.
rematch the impedance by adjusting capsets until best possi- When the distance is very short, since the linear search
ble power transfer efficiency point is found. The time duration starts searching from capset(126,1), the searching step dif-
for impedance rematching is referred to as algorithm execution ference between linear search and EWP search algorithms is
period in Fig. 16. In the operations, the most time consuming not large. However, as the distance increases, EWP search
part is around 1 s of idle time for stabilizing the system after algorithm show better performance over linear search and binary
capacitor-matrix adjustment. Since the EWP algorithm focuses search algorithms.
4412 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 29, NO. 8, AUGUST 2014

Fig. 16. Transient performance of the proposed WPT system.

VI. CONCLUSION
In a conventional wireless power transfer system, chang-
ing the location of either the transmitter or receiver causes an
impedance change, which drastically degrades the maximum
power transfer efficiency measured between the resonant coils
of the transmitter and the receiver. We have presented an adap-
tive impedance matching network based on a capacitor ma-
trix, which can dynamically change the impedance values to
maintain a reasonable level of maximum power transfer. The
adaptation of impedance is enabled by changing the switch con-
figuration of serial or parallel capacitances, and the number of
possible capacitance combinations increases exponentially with
the number of capacitances in the matrix. In order to quickly find
Fig. 17. Power transfer efficiency with impedance matching and tracking. the best capacitor combination that achieves maximum power
transfer efficiency, we also proposed a window prediction based
search algorithm. The adaptive impedance-matching network-
based WPT system with seven capacitors in a matrix is imple-
mented to verify the ideas presented. The WPT system works
seamlessly with up to 88% power transfer efficiency improve-
ment when the distance changes. The idea presented in this
paper can assist in the design of WPT systems for maintain-
ing maximum power transfer with varying distances between a
transmitter and receiver.

REFERENCES
[1] S. Ahson and M. Ilyas, RFID Handbook: Applications, Technology, Secu-
rity, and Privacy. Boca Raton, FL, USA: CRC Press, 2008.
[2] M. P. Theodoridis, Effective capacitive power transfer, IEEE Trans.
Power Electron., vol. 27, no. 12, pp. 49064913, Dec. 2012.
[3] Wireless Power Consortium. [Online]. Available:http://www.
wirelesspowersonsortium.com/
[4] S. L. Ho, J. Wang, W. N. Fu, and M. Sun, A comparative study between
novel witricity and traditional inductive magnetic coupling in wireless
charging, IEEE Trans. Magn., vol. 47, no. 5, pp. 15221525, May 2011.
[5] A. Kurs, A. Karalis, J. D. Joannopoulos, and M. Soljacic, Wireless power
transfer via strongly coupled magnetic resonances, Science, vol. 317,
pp. 8386, Jul. 2007.
[6] A. Karalis, J. D. Joannopoulos, and M. Soljacic, Efficiency wireless non-
radiative mid-range energy transfer, Ann. Phys., vol. 323, pp. 3448, Jan.
2008.
[7] T. Imura, T. Uchida, and Y. Hori, Experimental analysis of high efficiency
power transfer using resonance of magnetic antennas for the near field-
geometry and fundamental characteristics, in Proc. Jpn. Ind. Appl. Soc.
Conf., 2008, vol. 2, pp. 539542.
Fig. 18. (a) Different scenarios for measuring the tracking performance of lin- [8] S. Cheon, Y. Kim, S. Kang, M. Lee, J. Lee, and T. Zyung, Circuit model
ear search, binary search, and EWP search algorithm. (b) Number of steps taken based analysis of a wireless energy transfer system via coupled magnetic
to reach opt capset under different scenarios when three different algorithms resonances, IEEE Trans. Ind. Electron., vol. 58, no. 7, pp. 29062914,
are used. Jul. 2011.
LIM et al.: ADAPTIVE IMPEDANCE-MATCHING NETWORK BASED ON A NOVEL CAPACITOR MATRIX FOR WIRELESS POWER TRANSFER 4413

[9] A. P. Sample, D. T. Meyer, and J. R. Smith, Analysis experimental re- Hoyoung Tang (S12) received the B.S. degree in
sults, and range adaptation of magnetically coupled resonators for wireless electrical engineering from Korea University, Seoul,
power transfer, IEEE Trans. Ind. Electron., vol. 58, no. 2, pp. 544554, Korea, in 2012, where he is is currently working to-
Feb. 2011. ward the integrated masters and Ph.D degrees in the
[10] S. Y. R. Hui, W. X. Zhong, and C. K. Lee, A critical review of recent VLSI Signal Processing Research Lab, Korea Uni-
progress in mid-range wireless power transfer, IEEE Trans. Power Elec- versity.
tron., doi:10.1109/TPEL.2013.2249670, to be published His current research interests include efficient
[11] Y. Moriwaki, T. Imura, and Y. Hori, Basic study on reduction of reflected wireless power transfer and low-power VLSI digi-
power using DC/DC converters in wireless power transfer system via tal signal processing design.
magnetic resonant coupling, in Proc. Telecommun. Energy Conf., 2011,
pp. 15.
[12] B. Waters, A. Sample, and J. Smith, Adaptive impedance matching for
magnetically coupled resonators, in PIERS Proc., 2012, pp. 694701.
[13] B. K. Eplett, On-chip impedance matching using a variable capacitor,
U.S. Patent 2008/0211598 A1, Sep. 04, 2008
[14] Z. Wang, S. Gao, and C. Park, A simple method for tunable load
impedance matching network of power amplifier, in Proc. Microw. Mil- Seungok Lim received the B.S., M.S., and Ph.D.
limeter Wave Technol., Int. Conf., Chengdu, China, 2010, pp. 484487. degree in electrical engineering from KonKuK Uni-
[15] W. Lee, H. Lee, K. Oh, and J. Yu, Switchable distance-based impedance versity, Seoul, Korea, in 1997, 1999, and 2005,
matching networks for a tunable HF system, Progr. Electromagn. Res., respectively.
vol. 128, pp. 1934, 2012. He is currently a Director at the Network Con-
[16] M. Zargham and P. Gulak, Maximum achievable efficiency in near-field vergence Research Center, Korea Electronics Tech-
coupled power transfer systems, IEEE Trans. Biomed. Circuits Syst., nology Institute, Seoul, Korea. His research interests
vol. 6, no. 3, pp. 228245, Jun. 2012. include wireless communication systems for harsh
[17] J. Lee and S. Nam, Fundamental aspects of near-field coupling antennas environments such as underground and underwater
for wireless power transfer, IEEE Trans. Antennas Propag., vol. 58, environments, and wireless power transfer systems
no. 11, pp. 34423449, Nov. 2010. based on magnetic coupling mechanism.
[18] B. L. Cannon, J. F. Hoburg, D. D. Stancil, and S. C. Goldstein, Magnetic
resonant coupling as a potential means for wireless power transfer to
multiple small receivers, IEEE Trans. Power Electron., vol. 24, no. 7,
pp. 18191825, Jul. 2009.
[19] N. Kim, K. Kim, J. Choi, and C. Kim, Adaptive frequency with power-
level tracking system for efficient magnetic resonance wireless power
transfer, IEEE Power Electron. Lett., vol. 48, no. 8, pp. 452454, Apr.
12, 2012. Jongsun Park (M05SM13) received the B.S. de-
[20] Z. Pantic and S. M. Lukic, Framework and topology for active tuning of gree in electronics engineering from Korea Univer-
parallel compensated receivers in power transfer systems, IEEE Trans. sity, Seoul, Korea, in 1998, and the M.S. and Ph.D.
Power Electron., vol. 27, no. 11, pp. 45034513, Nov. 2012. degrees in electrical and computer engineering from
[21] C. K. Lee, W. X. Zhong, and S. Y. R. Hui, Effects of magnetic coupling Purdue University, West Lafayette, IN, USA, in 2000
of nonadjacent resonators on wireless power domino-resonator systems, and 2005, respectively.
IEEE Trans. Power Electron., vol. 27, no. 4, pp. 19051916, Apr. 2012. He joined the Electrical Engineering faculty of
[22] M. F. Moad, On Thevenins and Nortons Equivalent Circuits, IEEE the Korea University, Seoul, Korea, in 2008, where
Trans. Edu., vol. 25, no. 3, pp. 99102, Mar. 1982. he is an Associate Professor. From 2005 to 2008, he
[23] Motorola, Inc. (1997). Product. [Online]. Avaliable:http://www. was with the Signal Processing Technology Group,
datasheetcatalog.com/datasheets pdf/M/R/F/2/MRF282S.shtml Marvell Semiconductor Inc., Santa Clara, CA, USA.
[24] Texas Instriments. (2011). Product. Catalogue Archive. Avaliable: http:// He was also with the Digital Radio Processor System Design Group, Texas
www.ti.com/general/docs/lit/getliterature.tsp?literatureNumber=slas590 Instruments, Dallas, TX, USA, in Summer of 2002. His research interests in-
lfileType=pdf clude variation-tolerant, low-power and high-performance VLSI architectures,
and circuit designs for digital signal processing and digital communications.

Yongseok Lim (M12) received the B.S. degree

in electronics engineering from Korea University,
Seoul, Korea, in 2001, and the M.S. degree in ul-
tra large scale integration digital system from Korea
University in 2003.
He was with the ASIC Lab, Central Research Insti-
tute, Samsung Electro-Mechanics Inc., from 2003 to
2005. He is currently a Senior Researcher of Network
Convergence Research Center of the Korea Electron-
ics Technology Institute and working toward the Ph.D
degree in electrical engineering from Korea Univer-
sity. His areas of expertise include wireless power transfer and wireless/network
SOC system architecture.