This article has been accepted for publication in a future issue of this journal, but has not been
fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TPEL.2014.2377738, IEEE Transactions on Power Electronics
A General Method for Analyzing Resonant and
Soft-charging Operation of Switched-Capacitor
Converters
Yutian Lei, Student Member, IEEE, Robert C.N. Pilawa-Podgurski, Member, IEEE
AbstractTraditionally, switched-capacitor (SC) converters
have suffered from high transient currents, which limit both the
efficiency and power density of such converters. Soft-charging
operation can be employed to eliminate the current transients
and greatly improve the power density of SC converters. In this
approach, a second-stage magnetic converter is cascaded with the
SC stage to act as a controlled current load. Another approach is
to use resonant SC converters with zero current switching. This
paper shows that resonant and soft-charging operations of SC
converters are closely related, and a technique will be proposed
which achieves either operation by adding a single inductor
to existing SC topologies. In addition, since most preexisting
resonant or soft-charging SC converters were devised in an ad
hoc manner, this paper formulates an analytical method that
can determine whether an existing conventional SC converter
topology is compatible with the proposed approach. A number
of common SC topologies are analyzed, including Dickson, series-
parallel, ladder, Fibonacci and doubler configurations. Through
comparison to simulated results as well as experimental work, the
proposed method is validated and a family of high performance
SC converters is obtained.
Index Termssoft-charging, switched-capacitor converter, zero
current switching, resonant.
I. I NTRODUCTION
OLTAGE-SOURCE power converters are predominantly
V implemented using magnetic components (inductors or
transformers) as energy storage elements. Due to their rel-
atively low energy density, the magnetic components are
bulky in size and costly to build. On the other hand, the
efficiency and power-density of dc-dc converters need to be
continuously improved due to increasing demands in areas
such as computing and portable devices. As a result, switched-
capacitor (SC) converters [1][10] are gaining popularity.
These converters use capacitors as the energy storage element
and consequently can achieve higher power density and greater
suitability for on-chip integration compared to magnetic-based
converters. Additionally, SC converters also tend to achieve a
higher efficiency at large voltage conversion ratios [11]. These
advantages make SC converters desirable for a broad range
of applications, including voltage balancing [2], [3], energy
buffering [4], CMOS integrated power conversion [5], [7], [12]
and renewable energy harvesting [8].
However, SC converters also have certain drawbacks, which
limit their performance in some applications [13]. Since the
The authors are with the Department of Electrical and Computer Engineer-
ing, University of Illinois at Urbana-Champaign, Urbana, IL, 61801, USA.
E-mail: {lei10, pilawa}@illinois.edu
0885-8993 (c) 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.
capacitors are directly charged/discharged by other capacitors
or voltage sources, large transient current spikes can occur,
which limit the efficiency and power density of the con-
verter. Moreover, these transient effects increase the device
stress and can cause undesirable electromagnetic interference
(EMI) problems. To reduce the current spikes, either large
capacitors or higher switching frequency has to be employed,
neither of which is a satisfactory solution. The quasi-switched-
capacitor converter given in [14] manages to reduce the
peak of the current transient, but results in the same power
loss as conventional SC converters. Resonant SC converters
(or soft-switching SC converters), which incorporate one or
more inductors, have been proposed to eliminate the current
transients [15][25]. Because of the resonant inductor, the
capacitor current becomes sinusoidal, and if the switching
takes place at the moment when the current reaches zero,
resonant SC converters can operate in zero current switching
(ZCS) mode. This mode of operation enables such converters
to operate at higher frequencies and achieve higher power
density than their hard-switched counterparts.
Another drawback of SC converters is that high efficiency is
only achieved at one or a few conversion ratios. This limits the
application of SC converters to mostly low power applications.
In higher power applications, the solution is usually to cascade
a magnetic converter to act as a post-regulation stage [26]
[29]. However, the overall converter size may increase and
the peak conversion efficiency may be reduced.
An effective method to enable lossless regulation and elim-
inate the detrimental current transients in SC converters si-
multaneously is called soft-charging operation, as first demon-
strated with a merged two-stage converter in [30], [31]. In this
architecture, a second-stage buck converter is cascaded to the
output of a step-down SC converter and acts as a controlled
current load. The buck converter operates at a low voltage
and high frequency, enabling high bandwidth regulation and
reduced magnetic component size. The difference between
soft-charging operation and the aforementioned two stage
designs is that in soft-charging operation, the output capacitor
of the SC stage is removed and the flying capacitors of the
SC converter are charged/discharged by a controlled current
source (the buck converter). This mode of operation eliminates
current transients, enabling the SC converter to operate with
large capacitor voltage ripples with high efficiency. Larger ca-
pacitor voltage ripples result in larger energy transfer through
capacitors in each switching cycle, yielding significant power
density improvements of the SC converter. Therefore, the size
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TPEL.2014.2377738, IEEE Transactions on Power Electronics

of both the buck converter as well as the SC converter can m:n Rout
be reduced. It is also possible to achieve the same regulated
+ +
soft-charging operation by adding only an inductor and utilizes
the existing switches from the SC converter to form a buck
Vin Vout
converter cell. In this approach, no additional switches are
necessary and the control of the converter is simlified [32] The
- -
switched-capacitor stage employed in [30], [31] had a series-
parallel configuration that readily lends itself to soft-charging Fig. 1: Generic model of a switched-capacitor converter.
operation. Another recent implementation of the soft-charging
two-stage design utilizes a variable ratio SC stage and achieves
100
significant power density improvements over conventional
approaches [33], [34]. There are many SC topologies that

may also be designed to operate in soft-charging mode and
yield similar benefits, but to date, no formal method exists to
evaluate SC converters for their potential use in soft-charging
architecture. One of the main contributions of this work is the
development of such a formal method.
In this paper, the origin of the transient current and associ-
ated loss are first introduced and the concept of soft-charging
revisited. The requirements imposed on the SC converters by
soft-charging operation are postulated. Resonant SC converters
and soft-charging SC converters are then analyzed, and it is
shown that their terminal behaviors are closely related. This
makes it possible to use similar techniques to analyze and
synthesize both types of converters. We present a technique
to achieve soft-charging or resonant mode of operation with
a single inductor added to existing SC converter topologies.
In addition, in order to expand the family of soft-charging SC
converters, a formal method is presented to analyze arbitrary
SC topologies to determine their suitability for the proposed
technique. Supported by simulation and experimental results,
this new family of SC converters is shown to have the
potential to achieve a higher efficiency and higher power
density compared to traditional SC converters.
II. C APACITOR C HARGE S HARING L OSS AND
S OFT- CHARGING C ONCEPT
A generic SC converter model is shown in Fig. 1, which
consists of an ideal fixed-conversion-ratio stage with an
output-referred impedance [35]. The output impedance directly
reflects the efficiency of the converter, and incorporates both
the conduction loss and the capacitor charging/discharging
loss. This impedance is usually plotted against the switching
frequency to reveal the characteristics of the SC converters. A
typical such plot is shown in Fig. 2, which shows two asymp-
totic operating regions for SC converters: the fast switching
limit (FSL) and the slow switching limit (SSL) [11] [36]
[39]. The FSL occurs at high switching frequencies, when the
dominating loss is the conduction loss due to the resistance
of the switches as well as the ESR of the capacitors. As can
be seen in Fig. 2, the output impedance in the FSL region is
independent of the switching frequency. On the other hand,
the SSL occurs at low switching frequencies, when the output
impedance is dominated by the charging/discharging loss of
the capacitors during the charge redistribution process at phase
transitions. The SSL impedance depends on the switching
frequency and capacitor values, and cannot be reduced by
Output impedance ()
SSL
101
FSL
105 106 107
Frequency (Hz)
Fig. 2: Output impedance of a typical SC converter.
lowering the series resistance. Fundamentally, the SSL power
loss is the result of charging/discharging the capacitor with a
constant voltage source or another capacitor, as illustrated in
Fig. 3a and Fig. 3b respectively. We will examine case 2 (Fig.
3b) more closely since case 1 (Fig. 3a) can be seen identical to
case 2 with C2 being infinite. When the switch closes, since the
capacitor voltage cannot change instantaneously, the mismatch
of the initial capacitor voltages will be present across the
series resistor, resulting in a large instantaneous current as
shown in Fig. 4. The power loss incurred for complete charge
redistribution for the schematic showing in Fig. 3b can be
easily calculated and is given by
1
C1 (VC1(t=0) VC2(t=0) )2 fsw
Ploss =
4
1
= C1 V(t=0)2
fsw , (1)
4
assuming C1 = C2 . This equation is valid provided that the
duration of each phase is much larger than the time-constant
of the circuit, i.e. in SSL region of operation. As can be
seen, this power loss does not depend on the value of the
series resistance. Instead, it depends on the initial voltage
difference between the capacitors. Additionally, the initial
difference in capacitor voltages is due to the charge transfer in
the previous cycle, and thus is proportional to the charge drawn
by the load and inversely proportional to the capacitor values.
Moreover, the change of charge in the capacitor is proportional
to the duration of the charge/discharge, and thus is inversely
proportional to the switching frequency. These relations are
summarized below.
1 1
V , , (2)
fsw Cf ly

0885-8993 (c) 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TPEL.2014.2377738, IEEE Transactions on Power Electronics

as a controlled current source, whose voltage is permitted

RESR RESR
to change instantaneously to accommodate the voltage mis-
match between the flying capacitors and the load during
phase transitions. In practice however, the majority of the
+ loads are voltage-source loads or current-source loads with
Vin C1 C2 C1
large decoupling capacitors. Therefore, an interfacing element
typically has to be inserted between the SC converters and
the voltage-source load. Buck converters can be such an
(a) Capacitor charged by a (b) Capacitor charged by an-
voltage source. other capacitor. interfacing element, providing controlled charging/discharging
of the capacitors while regulating the output voltage [30], [31].
Fig. 3: Basic capacitor charging scenarios. As will be shown in Section III, the buck converter can also
be replaced by an inductor, while still achieving soft-charging
1
Voltage of C1
100
operation.
0.8 Voltage of C2 80 The second requirement for soft-charging operation is that
there should be no voltage mismatch within the flying ca-

Current (A)
Voltage (V)

0.6
0.4
0.2
Current of C2
0 0
0 0.2 0.4 0.6 0.8
Time (s)
Fig. 4: Capacitor voltages and current waveform in charge
redistribution process.
where Cf ly represents the overall flying capacitor value, and
for the circuit in Fig. 3b, it is simply C1 . Substituting (2) into
the power loss equation in (1), we have
1 1
Ploss , (3)
fsw Cf ly .
For more complex SC converters, more complicated charge
sharing scenarios will arise, but the general relationship stays
the same, as shown by the analytical results given in [11].
To reduce the charge sharing loss, one may simply increase
the switching frequency so that the converter operates in the
FSL region. However, it is often not favorable to do so, since
the transistor switching losses, as well as the bottom plate
capacitance losses in integrated SC converters, increase as
the switching frequency increases. Alternatively, increasing the
flying capacitor values can push the FSL region of operation
to a lower frequency, but it inevitably increases the circuit size
and cost. To overcome this dilemma, the soft-charging tech-
nique was proposed to eliminate the capacitor charge sharing
loss [30]. In soft-charging operation, a controlled current load
is placed in the charging/discharging paths of the capacitors.
The majority of the voltage mismatch between the capacitors
and the input/output will be present across the current load,
instead of across the switch resistance. With this technique,
the capacitor charging loss that is present in conventional SC
converters is recovered through the controlled current load,
which typically is a high frequency magnetic converter. As a
result, smaller capacitance can be used without sacrificing the
efficiency, despite the resultant larger capacitor voltage ripples.
This is the key benefit of soft-charging operation.
Given the origin of the charge sharing loss, this paper
postulates two requirements for achieving soft-charging op-
eration. The first requirement is that the load must behave
60
pacitor network during phase transitions. In [30], [31], a
40
series-parallel SC stage was chosen as the SC topology,
20 since the flying capacitors are simultaneously connected to
the output either in series or in parallel, and therefore, can
1 1.2
5
always be charged/discharged in the same manner through
10
the load current. Other SC topologies tend to have more
complex switching configuration and it is not immediately
apparent whether soft-charging of all flying capacitors can
be achieved by cascading a second-stage converter (or an
inductor) alone. Section IV introduces a formal method for
evaluating an arbitrary SC topology and determining whether
soft-charging operation is possible. This method replaces the
ad-hoc approaches that have been employed to date [30], [33].
III. S OFT- CHARGING O PERATION WITH AN I NDUCTOR
Since an inductor allows instantaneous change of its ter-
minal voltage, it can also act as a controlled current load
[40] [32]. In fact, the buck converter is able to facilitate
soft-charging operation precisely because of the inductor it
contains. In this section, the technique of adding an inductor
alone to achieve soft-charging is presented. Furthermore, it
will be shown that resonant operation can also be achieved
using the same technique.
To illustrate the technique, a simple 2-to-1 SC converter
is shown in Fig. 5a, and a modified structure is shown in
Fig. 5b. As can be seen in Fig. 5b, the technique to achieve
resonant and soft-charging operation is to add an inductor at
the output of the SC converter, immediately before the output
capacitor. The simple circuit structure in Fig. 5b allows direct
circuit analysis using differential equations [40]. Figure 6 plots
the simulated output impedance as a function of frequency
for both the original SC converter (Fig. 5a), as well as the
modified converter (Fig. 5b). It can be seen that for the original
SC converter, the output impedance reduces as the frequency
increases, while leveling off at high frequencies, marking the
transition from SSL to FSL. The output impedance curve
is more complicated for the modified converter, but a few
key observations can be made. First, with the additional
inductor, the modified converter is able to reach the same
minimum impedance at a much lower switching frequency,
due to the elimination of the current transient and associated
loss. Therefore, the proposed converter can achieve the same

0885-8993 (c) 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TPEL.2014.2377738, IEEE Transactions on Power Electronics

+
Phase 1 is (C1 k C2 + C3 ) and the capacitance in Phase 2
1 2 is (C2 k C3 + C1 ). Both phases have the same equivalent
Cfly capacitance of 1.5C1 assuming C1 = C2 = C3 . For converter
+ Vload topologies that have different equivalent capacitance in each

Vin 1 Cout phase (such as the Fibonacci and series-parallel), there is a crit-
2
ical frequency for each phase, and the overall critical frequency
- is the weighted average of the individual frequencies according
(a) Example 2-to-1 SC converter. to the duty ratio. Note that the critical frequency corresponds
to the resonant frequency of the circuit. To understand the
L1 frequency dependent behavior of the modified SC converter,
+
the terminal voltage before the inductor (Vsc in Fig. 5b) as
1 2 + well as the inductor current are shown in Fig. 7 at 3 different
Cfly frequencies - the resonant frequency (fcrit ) as well as below
+ Vsc Vload (f2 ) and above (f1 ) the resonant frequency. It can be seen that,

Vin 1 Cout above the resonant frequency, the current waveform (Fig. 7a)
2
- is smooth and has small ripple, due to the filtering effect of the
- inductor. Moreover, since the flying capacitor is always in the
(b) Example 2-to-1 SC converter with an inductor. same current path as the inductor, the conventional current
spikes of the capacitor are eliminated, and the capacitors
Fig. 5: Schematics of a simple SC converter. transfer charges in soft-charging mode, with no charge transfer
loss. The effect of this can be seen directly from Fig. 6, where
for switching frequencies larger than the critical frequency,
ConventionalhSChconverter the SC converter has the minimum FSL output impedance. As
0
10 SChconverterhwithhanhoutputhinductor the switching frequency is reduced, the current waveform has
larger ripple, while having the same average value, since the
Outputhimpedanceh()

load current is kept constant. At the resonant frequency, fcrit ,

the inductor current takes the shape of a rectified sinusoid,
and the current reaches zero at moments of phase transitions,
as shown in Fig. 7b. Thus, zero current switching can be
1
10 achieved at the resonant frequency. As can be seen in Fig. 6,
the impedance of the converter at resonance is slightly larger
than the FSL impedance. This is because the sinusoidal current
h has larger RMS value than the near constant current in FSL
operation. As the switching frequency is reduced further (Fig.
4 5 6
10 10 10 7c), the inductor current drops negative during each cycle,
Frequencyh(Hz)
resulting in a much larger RMS current for the same average
Fig. 6: Simulated output impedance vs frequency. power delivered. This is why the impedance increases sharply
for fsw < fcrit . At one half of the resonant frequency defined
by (4), the current becomes nearly a full-wave sinusoid, giving
efficiency as conventional SC converters while using signifi- a peak impedance in Fig. 6. This peak repeats itself at lower
cantly lower switching frequency, or equivalently, significantly frequencies when the current waveform has multiple periods of
smaller flying capacitor values. The second observation is that, the full-wave sinusoid, at fsw = n1 fcrit , where n is an integer
at lower frequencies, the output impedance oscillates around and n 2. Therefore, fcrit given in (4) sets the lower bound
the SSL impedance of the conventional SC converter. on the switching frequency for which near FSL impedance in
The minimum frequency at which the converter is able to soft-charging operation can be achieved.
stay in FSL operation can be defined as fcrit and for the
modified converter in Fig. 5b, it is given by
It can be noted that the reduction in frequency and capaci-
1
fcrit = (4) tance with soft-charging operation can be achieved but at the
2 LC expense of adding an inductor. While the trade-off between
where L is the added inductance and C is the collective the capacitor values and inductor values should be evaluated
capacitance in series with the inductor. In the case of the on a case-by-case basis, in general, adding an inductor results
example SC converter in Fig. 5b, the capacitance is simply in better utilization of passive components than simply using
Cf ly . For more complex SC converter topologies, this equiva- larger capacitance. For a traditional SC converter circuit,
lent capacitance can be obtained by calculating the equivalent whether the high current transient takes place is determined
series and parallel connected capacitance at the inductor input by the time constant of the circuit - RESR C, where RESR
node for each phase. For instance, for the 4-to-1 Dickson is the series resistance in each conducting branch and C is
converter shown later in Fig. 10, the effective capacitance in the capacitance in each branch. Thus the critical frequency at

0885-8993 (c) 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TPEL.2014.2377738, IEEE Transactions on Power Electronics

8.5 8.5 9

(V)
Vsc (V)

Vsc (V)
8 8 8

sc
V
7.5 7.5 7
0 2 4 6 0 5 10 0 5 10 15 20
Time (s) Time (s) Time (s)
10 10 10

i (A)
iL (A)

iL (A)
5 5
0 0 0

L
5 5 5
0 2 4 6 0 5 10 0 5 10 15 20
Time (s) Time (s) Time (s)
(a) fsw = f1 > fcrit . (b) fsw = fcrit . (c) fsw = f2 < fcrit .

Fig. 7: SC stage voltage (Vsc in Fig. 5b) and inductor current of the modified 2-to-1 converter.

which the conventional SC converter enters FSL operation is

1
10
R =100 m
on
1
fcrit = (5) R =10 m
on

Power loss (W)

2RESR C 0
10

It can be seen that the switching frequency is inversely pro-

portional to the product of the equivalent series resistance and 1
10
capacitance. Thus, for a given desired critical frequency, the
capacitance must be increased if the resistance is lowered. This
limitation can be clearly seen in Fig. 8a, where the power loss 2
10
4 5 6 7
10 10 10 10
of a pure SC converter is plotted against two different switch Frequency (Hz)
Rds,on values. Even when the Rds,on of the switch is reduced (a) Conventional SC converter.
by a factor of 10, to see a factor of 10 reduction in the power
loss, one needs to increase the switching frequency by a factor 1
10
of 10, or equivalently, increase the capacitor values by a factor R =100 m
on

of 10. This is due to the inversely proportional relationship R =10 m

on
Power loss (W)

0
among the resistance, capacitance and switching frequency. On 10

the other hand, with the additional inductor presented here, the
critical frequency is decoupled from the series resistance, and 1
10
only depends on the inductance and the capacitance, as shown
in (4). The effect can be seen in Fig. 8b, where a reduction in
the series resistance instantly brings a nearly equal reduction in
2
10
4 5 6 7
10 10 10 10
power loss, without the need to increase the capacitance nor Frequency (Hz)
the frequency. Therefore, the addition of the inductor gives (b) Soft-charging SC converter.
the designer the choice of using smaller on-state-resistance
Fig. 8: Power loss at different frequencies for different Rds,on
switches and introduces a new design dimension in which
values, for the circuit in Fig. 5.
the converter can be optimized. Also can be observed from
Simulation parameters: C = 10 F , L = 0.1 H
Fig. 8b is that oscillation does not occur for the case of
Ronq= 100 m. This is because the system is over-damped
( R2 C L < 1) for large resistance values. In this case, the RC design goal of the soft-charging SC converters. In discrete
time constant starts to dominate the frequency response of the implementations, the addition of the inductors often results
system again, and soft-charging operation does not take place. in overall improvement in energy utilization of the passive
As a result, there is no change in power loss by adding the components. While the inductor is more difficult to integrate
small inductor, as can be seen by comparing the red dotted than the capacitors given the current IC technology, the energy
lines in Fig. 8a and Fig. 8b. In this case, a larger inductance density and quality of integrated inductors are improving as
would have been needed to achieve a reduction in power loss in more advanced processes are adopted [13], [41], [42], and the
the SSL region. Therefore, in addition to the critical switching proposed converter is able to take advantage of the progress
frequency requirement given in (4), the soft-charging (as well and advancement of technologies in inductors, capacitors and
as resonant) SC converters needq to be designed such that the switches simultaneously.
system is under-damped ( R2 C L < 1). Nevertheless, a lower
series resistance is one of the goals for power converters aim- IV. A NALYZING AN A RBITRARY SC T OPOLOGY FOR
ing for high conversion efficiency, especially for applications S OFT- CHARGING O PERATION
with small load resistances. This naturally coincides with the It was shown in the previous section that resonant and soft-
charging operation are closely related and both modes of op-

0885-8993 (c) 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TPEL.2014.2377738, IEEE Transactions on Power Electronics

eration can be achieved with a single additional inductor. The

C3
example used was a 2-to-1 SC converter, which easily satisfies C1
the second condition given in Section II, which states that to 1 2
achieve full soft-charging operation, there can be no voltage
mismatch among the flying capacitors during phase transitions.
The example 2-to-1 SC converter satisfies this condition easily 1 2 1 2
+ Iload
since it only has a single flying capacitance. More complicated
Vin
SC converters have multiple flying capacitors connected in 2 1
a number of different configurations. Thus, it is of great
interest to determine whether this proposed technique can be C2
broadly applied to other SC converter topologies. To answer Fig. 9: 4-to-1 Dickson topology.
this question, a general method is derived in this section to
determine if an arbitrary SC converter topology can operate
in resonance or soft-charging operation with the addition of an
output inductor [43]. Since resonance with the inductor at the
output can be viewed as soft-charging operation at a special C3 C2
switching frequency, only the term, soft-charging, is used in +
C1
this section for convenience. Vin Iload C3 Iload
C2 C1
In essence, the proposed method examines the charge flow
characteristics of an SC converter topology and observes the
(a) Phase 1. (b) Phase 2.
change in capacitor voltage subject to Kirchhoffs Voltage Law
(KVL) constraints. In each phase of the SC converter, the Fig. 10: 4-to-1 Dickson topology in each phase.
voltage across a capacitor changes according to the charge
flow in the given phase. When the converter switches to the
next phase, KVL poses new constraints on each component.
Complete soft-charging is achieved if and only if the ideal where Ai is called the reduced loop matrix of the ith phase
capacitor network satisfies KVL at all times, including at [46] and the voltage vector v is defined as
phase transitions. If during any period, the KVL constraint
v = [vin vc1 vc2 vc3 vout ]T (9)
is not satisfied, the voltage discrepancy will appear across the
series resistances, resulting in a charge transfer impulse. The
T
= [vin vc vout ] T
(10)
KVL constraint is present whether soft-charging or resonant In this analysis, the entries of the loop matrices are positive if
operation is of interest, and thus the analysis presented in this the circuit element is traversed from the negative terminal to
section applies to both operations. the positive terminal and vice versa. Combining the definitions
in (8) (9) and the KVL equations in (6) (7), the loop matrices
are found to be
A. General analysis using Dickson converter as an example    
1 0 0 1 1 0 0 1 1 1
The analysis method in this work is illustrated with a 4-to-1 A1 = and A2 = .
0 1 1 0 1 0 1 0 0 1
SC converter in Dickson configuration [5], [44], [45] shown
in Fig. 9. To simplify the analysis, a constant current source Denoting the voltage vector at the start of phase 1 as v1 , KVL
is used as the load for this and all following examples, while analysis yields
we note that a practical implementation would use a magnetic- A1 v1 = 0, (11)
based converter or an inductor. The two phases of the Dickson
which captures the KVL constraints given by (6) at the
topology are shown in Fig. 10a and Fig. 10b respectively. In
moment when the converter has begun phase 1 operation. At
each phase of Fig. 10, the circuit consists of a number of closed
the end of phase 1, the voltage vector becomes v1 + v1 ,
loops, and a KVL equation can be written for each loop. For
creating a second KVL constraint:
example, the following two independent KVL equations can
be written for phase 1 of the Dickson converter (Fig. 10a): A1 (v1 + v1 ) = 0, (12)
(
Vin VC3 Vout = 0 where v represents the change in voltage due to charge
(6)
VC2 VC1 Vout = 0 being delivered to the load, and is in the form of
[vin vc T vout ]T , similar to the voltage vector in (9)
and for phase 2 (Fig. 10b): . Since both (11) and (12) must be satisfied, a resulting
constraint is that the vector v must satisfy
(
VC3 VC2 Vout = 0
(7)
VC1 Vout = 0 A1 v1 = 0. (13)
These KVL equations can be written in a matrix-vector- From a circuit intuition point of view, (13) describes the fact
product form as while the individual node voltages can (and will) change as a
Ai vi = 0, (8) result of charge transfers, the sum of the changes in a KVL

0885-8993 (c) 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TPEL.2014.2377738, IEEE Transactions on Power Electronics

loop must be zero. Similarly for phase 2, we have
2
A2 v = 0.
Note that the vin component of vi is typically zero since
the input voltage is considered constant. This information can
be included in the loop matrices by adding a row of [1 0 0 0 0]
to both A1 and A2 , resulting in A1m and A2m respectively,
where the subscript m indicates a modified reduced loop
matrix. Correspondingly, (13) and (14) become
However, with soft-charging operation, the SC stage output
node is connected to an inductor, and the inductor voltage is
(14)
allowed to change instantaneously during phase transitions, as
opposed to the capacitor voltages, which must be continuous.
Stated in another way, the parameter vout defined previously
is no longer a state variable in a switch-linear circuit, and can
be discontinuous. As a result, the change in output voltage in
phase 1 due to the current load, vout 1
, does not necessarily
equal vout . Therefore, the inductor introduces one more
2

degree of freedom to the system. To mathematically express

A1m v1 = 0 (15) this additional degree of freedom, the basis w and u can be
A2m v = 0 2
(16) modified by removing the last element in each column (the
entry that represents vout ), resulting in the new basis w and
The solution to (15) and (16) represents the set of permissible
. Now, replacing the basis in (21) with the newly formed w
u
voltage changes that satisfy KVL and vin = 0. This solution
and u , we obtain for the soft-charging converter:
is the nullspace of A1m and A2m , by definition. Let w and
u be the collective basis for nullspaces of A1m and A2m a1
respectively. It follows that any solution to (15) and (16) can  a2
(22)

w1 w 2 u 1 u 2 b1 = 0 .

be represented by a linear combination of the basis:
b2
v1 = a1 w1 + a2 w2 (17)
Mathematically. the matrix in (22) has a reduced rank com-
v = b1 u1 + b2 u2
2
(18)
pared to the one in (21), and thus a non-zero solution can be
In the case of the 4-to-1 Dickson converter, such basis can be found. Solving (22) for the Dickson converter, we obtain
found1 as
a1 0.120

0 0 0 0
a2 0.697
0.607
0.482
0.362
0.518 b1 0.607 .
= (23)



w = 0.763 0.131 and u = 0.398 0.664 .

0.157 0.613
0.761 0.146
b2 0.364
0.157 0.613 0.362 0.518
The voltage change vectors in each phase can then be found
For conventional SC converters, we have the additional using (17) (18), yielding
constraint that
v1 = v2 , (19) 0 0
0.408 0.408
from the condition of periodic steady-state operation. This is v = 0 and v =
1 2
(24)

0 .

because in a capacitive network, the voltage changes must sum 0.4088 0.408
up to zero in a full switching cycle. Combining (17) (18) and 0.408 0.408
(19), we have
Note that the original basis are used to obtain the voltage
a1 w1 + a2 w2 + b2 u1 + b2 u2 = 0 . (20) change at each node. From (24), it can be seen that the net
change in the capacitor voltages is zero (i.e., each column adds
Note that (20) can be written in a matrix form as
to zero), except for the last entry. This entry represents vout ,
which is the node that can be discontinuous owing to the soft-

a1
 a2 charging operation. Having obtained the change in capacitor
(21)

w1 w2 u1 u2 b1 = 0 . voltage required to satisfy KVL in each phase, we can then

b2 calculate the required capacitance for soft-charging operation

For the conventional Dickson SC converter, no solution for
(21) can be found, except for the trivial case of zero. This
means that no voltage change exists for the circuit that satisfy
KVL at all times. This result is reassuring and consistent
with the behavior of conventional SC converters, where it is
well-known that this instantaneous voltage mismatch at phase
transitions is what gives rise to the power loss from charge
redistribution [11]. Hence, conventional SC converters have to
rely on high switching frequency or larger capacitor values
to minimize the voltage mismatch and the associated power
loss.
1 For example, by using the command null in Matlab, which will yield a
set of orthonormal basis.
as:
Cj = qj /vcj , (25)
for each capacitor j. Equation (25) requires the charge that
flows into each flying capacitor to be found for each phase.
For any well-posed switched-capacitor topology, a charge flow
vector can be obtained for each phase either by inspection [11]
or Kirchhoffs Current Law [36]. In this work, the charge flow
vector is defined as the vector of charge that flows into the
positive terminal of each element in the circuit and is given
in the form of
 
q = qin qc1 qc2 qc3 qout .
The charge flow vectors for the Dickson converter of this

0885-8993 (c) 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TPEL.2014.2377738, IEEE Transactions on Power Electronics

100
example are found to be Hard-charging
Soft-charging 1
q1 = 1 1 1 1 2 and q2 = 0
   
1 1 1 2 . Soft-charging 2

Together with the voltage change vector found in (24), the

Power loss (W)

capacitor values are obtained using (25) and are simplified as
follows.
C1 1
C2 = 101
C3 1
It can be seen that the Dickson topology can achieve complete
soft-charging only when C2 = and C1 = C3 . In practice, 105 106
Frequency (Hz)
this means that it can approach soft-charging with a C2 large
enough compared to C1 and C3 . As can be seen from (24), the Fig. 11: Power loss of Dickson converter at different frequen-
output voltage ripple has the same magnitude as the voltage cies.
ripple of C1 and C3 under soft-charging operation. Thus, for
soft-charging operation with the Dickson converter, a designer
would want to minimize the charging/discharging loss by as switching frequency increases, while leveling off at high
maintaining a relatively large C2 /C1 ratio while keeping the frequency, showing the transition from SSL to FSL. In soft-
output ripple of the SC stage tolerable with a second-stage charging operation, a significant reduction in power loss at
converter or an inductor. lower frequencies is seen when the capacitors are such that
To summarize, the following steps are used to determine C1 = C2 = C3 , as in the hard-charging case. However, a
whether any given SC topology is compatible with soft- more prominent reduction is seen when the capacitor values
charging or resonant operation using a single inductor con- are chosen such that C2 /C1 = 4, plotted in Fig. 11 as Soft-
nected to the output of the SC stage: charging 2. This confirms that the Dickson converter can
1) Obtain the reduced loop matrix for each phase (A1 and approach full soft-charging by maintaining a high C2 /C1 ratio,
A2 ) using KVL analysis. as predicted by the analysis in this work.
2) Add a row of [1 0 0 ... 0] to A1 and A2 , obtaining The currents through the capacitor C2 of the Dickson SC
A1m and A2m . converter (Fig. 9) in hard-charging and soft-charging oper-
3) Find the collective nullspace basis of A1m and A2m ations with the converter switching at 250 kHz are shown
(w and u respectively). in Fig. 12. The different waveforms are shifted apart in the
4) Remove the last row of w and u to obtain w and u. time axis for clearer observation. It can be seen that under
5) Use (22), (17) and (18) to find the change in capacitor hard-charging condition, the capacitor current resembles the
voltages. exponential discharge, as expected. With soft-charging 1, both
6) Find the charge transfer vector for each capacitor [11], the magnitude and the width of the impulse are reduced,
[36]. while the tail of the exponential decay is raised. In the soft-
7) Use (25) to find the capacitance values required for soft- charging 2 case, with capacitor values selected according to
charging. the analysis result, the height and the width of the impulse
As demonstrated in this section, for a two-phase SC dc-dc is further reduced and current waveform resembles more of a
converter, if a capacitor voltage change vector, vc , can be square wave. The transient effect is not completely eliminated,
found to satisfy KVL at all times, and the resultant capacitor
values required are practical (finite and positive), the given
topology is able to perform soft-charging and resonant opera- TABLE I: Simulation parameters.
tion and will exhibit no charging/discharging loss. Otherwise,
Vin 5V
at least one loop of the circuit will not be able to perform
Iload 2A
soft-charging, and the benefit will be limited.
Ron 10 m
B. Simulation verification RESR 1 m
To verify that the analytical method is correct, the Dickson Co,hardcharging 100 F
converter shown in Fig. 9 is simulated using LTSpice with Co,sof tcharging 0.1 F
simulation parameters given in Table I and Table II. A total
capacitance of 30 F is used for the flying capacitors. In TABLE II: Flying capacitor values.
hard-charging (conventional) operation, an additional 100 F
output capacitor is added in parallel to the current load while Configuration C1 (F) C2 (F) C3 (F)
there is no output capacitance in the soft-charging simulation.
Hard-charging 10 10 10
The converters are operated at a fixed duty ratio of 0.5.
Soft-charging 1 10 10 10
The simulated power losses are plotted in Fig. 11. It can
Soft-charging 2 5 20 5
be seen that the hard-charging power loss decreases linearly

0885-8993 (c) 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TPEL.2014.2377738, IEEE Transactions on Power Electronics

15 voltage of C2 is in the opposite direction of what is required

to satisfy KVL. Thus, the single-output ladder topology is
10 not compatible with soft-charging without modification, and a
limited improvement is expected. As for the doubler converter,
5 both C1 and C2 have to be infinite for complete soft-charging,
indicating a partial soft-charging capability similar to that of
Current (A)

0 the Dickson converter.

5 13 are simulated with the same parameters as those of the
10
15
1 2 3 4 5 6
Time (s)
Fig. 12: Current waveform of capacitor C2 of the Dickson SC
converter.
TABLE III: The RMS and average current of capacitor C2 in
a single phase.
Configuration RMS current (A) Average current (A)
Hard-charging 2.19
Soft-charging 1 1.40
Soft-charging 2 1.11
due to the fact that perfect soft-charging cannot be achieved.
To quantify the change in the current waveform, the RMS
and the average currents through the capacitor for one phase
duration are calculated and tabulated in Table III. It can be
seen that while in all cases the capacitor supplies the same
average current over a phase duration, the RMS value of
the current decreases from hard-charging operation to soft-
charging operation. Again soft-charging 2 is an improvement
over soft-charging 1. Thus, the soft-charging operation reduces
the impedance in the SSL region due to the improvement in
the charging and discharging current waveform.
C. Application to other topologies
The general analysis method proposed is applied to four
additional commonly used two phase switched-capacitor con-
verter topologies - series-parallel, ladder, Fibonacci and dou-
bler. The schematics are shown in Fig. 13a, 13b, 13c and 13d
respectively. The same analysis is repeated for each of them
and the results are shown in Fig. 14.
It can be seen that for the series-parallel converter, a simple
requirement for soft-charging is that all the flying capacitors
have the same value. Under soft-charging condition, the output
voltage ripple is shown to be equal to N 1 times the change
in any of the capacitor voltages, where N is the conversion
ratio. These observations agree with the experimental work in
[30]. In addition, the Fibonacci converter is also found capable
of soft-charging operation with equal capacitors. On the other
hand, for the ladder configuration, one can see that a negative
capacitance is needed on C2 for soft-charging operation, which
is not achievable. This means that the change in capacitor
To verify the analysis results, the circuits shown in Fig.
Dickson converter. Equal flying capacitors are used in all
Hardcharging cases. Again, a constant current source is used as the load
Softcharging 1 instead of an inductor to simplify the simulation and remove
Softcharging 2
the effect of resonance, since in practice, operation below the
7 8 critical frequency is to be avoided as shown in Section III.
The corresponding power loss curves are plotted in Fig. 15.
It should be noted that the power loss values are not intended
for cross-comparison between different SC topologies. Rather,
it is the reduction of the power loss by changing from
hard-charging operation to soft-charging operation that is of
key interest here. For both the series-parallel converter and
the Fibonacci converter, soft-charging operation results in a
significantly lower power loss in SSL region than in the
1.00 hard-charging case, and the loss is almost independent of
1.00 the frequency. The ladder configuration only receives very
1.00 limited benefit from soft-charging and a strong frequency
dependency is still seen on the power loss plotted in Fig. 15b.
The doubler converter shows moderate improvement with soft-
charging. These simulation results agree with the prediction of
the analytical technique presented earlier.
Since the Fibonacci converter is shown to be able to achieve
full soft-charging operation, it is useful to examine the current
waveform of the Fibonacci converter, as shown in Fig. 16. It
can be seen that the waveform in soft-charging operation is a
square wave, confirming that the current transient associated
with capacitor charge redistribution has been eliminated. Ta-
ble IV shows the RMS and average of the absolute values of
the current through capacitor C3 of the Fibonacci converter. It
can be seen that now the RMS current is almost equal to the
average current in the soft-charging case, ensuring the lowest
power loss.
TABLE IV: The RMS and average of the absolute current of
capacitor C2 of the Fibonacci converter.
Configuration RMS current (A) Average current (A)
hard-charging 1.45 0.800
soft-charging 0.816 0.800
V. R ESONANT O PERATION WITH A S INGLE I NDUCTOR
Resonant SC converters [15][25] have been proposed as an
alternative that offers higher power density and lower switch-
ing loss compared to conventional SC converters. Resonant
SC converters achieve ZCS operation by adding one or more
inductors to a known SC topology [15][20], or in some cases
making use of distributed inductance to shape the charge and
discharge current [21][25]. While the inductor placement

0885-8993 (c) 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TPEL.2014.2377738, IEEE Transactions on Power Electronics

2 Iload
2

2
C2 Co

1 C3 1 C2 1 C1 1 1 2 1 2 1 2
+ 2 2 2
+
Vin
Co Iload Vin C3 C1

(a) 4-to-1 series-parallel. (b) 3-to-1 ladder.

1 2 1 2 1 2 1 2

C3 C2 C1 C3 C1
+ +
Vin 1 2 1 Co Iload Vin 1 C2 1 Co
2 2 Iload
2 1 2

(c) 5-to-1 Fibonacci. (d) 4-to-1 doubler.

Fig. 13: Common switched-capacitor converter topologies.

v
in " #
vc1 C1
v = vc2 , C = C2

vc3 C3
vout

(a) General.
0 0 0 0
" # " #
1 11 2 1 1
1
1
2
1 1
v = , v = 1 , C = 1 v = 1 , v = 1 , C = 2

1 1 1 1 1 1
3 1 2 1

(b) Series-parallel. (c) Ladder.

0 0 0 0
" # " #
1
2
2 2 1
1
1
2 1 1
v = , v = 1 , C = 1 v = , v = 0 , C =
1 0
1 1 1 0 0
3 2 1 1

(d) Fibonacci. (e) Doubler.

Fig. 14: Voltage change vectors and relative capacitor values for soft-charging operation.

varies, typically it is preferable to have only one additional
inductor to reduce the complexity and parameter matching
difficulty, and thus achieving ZCS operation with a single
inductor is of interest in this paper. As was demonstrated
in section III, ZCS resonant operation can be accomplished
for the 2-to-1 example converter, at a switching frequency of
fcrit , by adding one inductor at the output node. However,
the basic 2-to-1 topology is simple to analyze and has already
been exploited by many existing work [15], [17]. To date, it
has not been clear whether resonant operation is possible with
only one inductor, for a general SC converter with multiple
charge transfer paths.
It is postulated here that full ZCS on all switches can be
achieved with a single inductor if the topology is compatible
with full soft-charging operation. Since the single inductor is
placed at the output node and is connected to the capacitor
branches at all times, the same terminal voltage appears in all
the capacitor branches. Provided that there is no current tran-
sient during phase transitions among the capacitor branches,
which is guaranteed by soft-charging operation, sinusoidal
inductor current ensures sinusoidal current in the capacitors
and thus in the switches. To illustrate this concept, the currents
through all the switches of a resonant Fibonacci converter
(as shown in Fig. 17) are plotted in Fig. 18. It can be seen
that, since the Fibonacci converter is compatible with full soft-
charging operation, indeed all the switches turn on and off at
the zero crossings of the current waveforms and thus sinusoidal
inductor current ensures ZCS for all switches. A similar case
can be shown for the series-parallel SC converter. Therefore,
two new ZCS resonant two-phase SC converter based on the

0885-8993 (c) 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TPEL.2014.2377738, IEEE Transactions on Power Electronics

100
hard-charging hard-charging
Power loss (W) soft-charging 100 soft-charging

Power loss (W)

101
101

105 106 105 106

Frequency (Hz) Frequency (Hz)

(a) Series-parallel. (b) Ladder.

100 100
hard-charging hard-charging
soft-charging soft-charging
Power loss (W)

101 Power loss (W) 101

105 106 105 106

Frequency (Hz) Frequency (Hz)

(c) Doubler. (d) Fibonacci.

Fig. 15: Power loss of hard-charging and soft-charging SC converters from LTSpice simulation.

6
SW10 SW7 SW4 SW1 L
C3 C2 C1
4 +
Vin SW8 SW5 SW2 Co Iload
2 SW9 SW6 SW3
Current (A)

0 Fig. 17: Resonant Fibonacci converter with a single inductor

at the output.
2

4 or in high frequency designs, where switching loss is more

hardcharging
softcharging significant, operating at resonance with ZCS would yield the
6 highest efficiency.
0 1 2 3 4 5 6 7 8
Time (s) x 10
6

VI. E XPERIMENTAL R ESULTS

Fig. 16: Current waveform of capacitor C3 of the Fibonacci
converter. A hardware prototype is implemented for the soft-charging
Dickson SC converter, but extended to a voltage conversion
ratio of 8 to 1. The schematic is shown in Fig. 19. A total of 12
GaN switches are used, together with seven flying capacitors
series-parallel and the Fibonacci structures are possible, based of various voltages and one inductor. The design specification
on the technique presented in this work. can be found in Table V while a full component listing is
Whether one should choose soft-charging operation or ZCS provided in Table VI. A photograph of the hardware prototype
operation often depends on the load conditions. At heavy is shown in Fig. 20. All the components are placed on the
load, where conduction loss dominates, it would be beneficial top side of the PCB for clear illustration. A prototype of
to operate at higher frequencies in soft-charging mode to the conventional Dickson SC converter is also built for the
minimize the current ripple. On the other hand, at light load purpose of comparison, with the same set of switches. The

0885-8993 (c) 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TPEL.2014.2377738, IEEE Transactions on Power Electronics

2 Isw1
Isw2
1 Isw3
Current (A)

Isw4
0
Isw5
1 Isw6
Isw7
2 Isw8
Isw9
3
Isw10
4
0 2 4 6 8 10 12 14
Time (s)

Gate signal for

SW1 SW3
SW5 SW7
SW9

0 2 4 6 8 10 12 14 Fig. 20: Annotated photograph of the experimental converter

Time (s) prototype, with a U.S. penny added for perspective.

Gate signal for

SW2 SW4
SW6 SW8 output impedance. The efficiency represents an approximate
SW10 2x power loss reduction at 53 W. For both cases, the measured
0 2 4 6 8 10 12 14 efficiencies do not include the power loss due to control circuit
Time (s) and gate drivers. The combined loss of these components
Fig. 18: Current waveform of all the switches in the Fibonacci are approximately 0.5 W, which is mainly attributed to the
converter as shown in Fig. 17. poor efficiency of the level-shifting circuit used to power the
gate drivers. Overall, the experimental results demonstrate that
the soft-charging SC converter simultaneously achieves higher
C7 efficiency and higher power density than the conventional SC
C5
C3 converter.
C1 To see the similarity between resonant and soft-charging
operations, the current through the inductor is shown in Fig.
S4 S1
Co
S12 S11 S10 S9 S8 S7 S6 S5 Rload 21 at three different switching frequencies. As can be seen,
+
Vin
+ Vsc - for fsw = fcrit (Fig. 21b), the current is sinusoidal and
S3 S2 reaches zero at each phase transition, and thus ZCS operation
C2 is achieved. For fsw > fcrit (Fig. 21a), the current has a
C4
C6

Fig. 19: Schematic of the proposed soft-charging Dickson SC TABLE V: Tested specifications.
converter.
Vin 200 V DC
Conversion ratio 8:1
only difference is that in the conventional Dickson converter, Pout 53 W
there is no inductor and the flying capacitors are all of value fsw 250 kHz
2.2 F. The volume of the passive components of the two
converters are shown in Table VII. The total volume of the TABLE VI: Component listing of the proposed converter.
passive components of the SC converter with the inductor is
Component Part number Parameters
454 mm3 while that of the pure SC converter is 682 mm3 . It
can be seen that even with the additional inductor, the volume S12 , S5 - S1 EPC2014 40 V, 16 m, 10 A
S11 - S6 EPC2007 100 V, 30 m, 6 A
of the proposed converter is still smaller than that of the pure
SC converter, thanks to the improved utilization of capacitors C7 , C5 C1812X224K2RACTU 250 V, 0.22 F
C6 , C4 C2220C225MAR2CTU 250 V, 2.2 F
due to soft-charging. C3 C0805C224K1RACTU 100 V, 0.22 F
The measured efficiencies of the proposed converter as well C2 C3216X7S2A225K160AB 100 V, 2.2 F
as the conventional SC converter at various load currents C1 C1608X7R1H224K080AB 50 V, 0.22 F
Co C3216X5R1V226M160AC 35 V, 22 F
are plotted in Fig. 22. It can be seen that not only is the
Inductor XAL5030-332 3.3 H
efficiency of the soft-charging converter always higher than
that of the conventional hard-charging converter, but it also Level-shifters ADUM5210
Micro-controller STM32f051
drops at a slower pace as the current increases, due to the lower

0885-8993 (c) 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TPEL.2014.2377738, IEEE Transactions on Power Electronics

(a) fsw = 250 kHz > fcrit (b) fsw = 90 kHz = fcrit (c) fsw = 50 kHz < fcrit

Fig. 21: SC stage voltage (Vsc in Fig. 19) (upper) and inductor current (lower).

TABLE VII: Passive components volume comparison.
Conventional
Capacitor volume (mm ) 3
681.8
Inductor volume (mm3 ) - 75.0
Total volume (mm3 ) 681.8
much smaller ripple and the converter operates near FSL. For
fsw < fcrit (Fig. 21c), the current goes negative in each phase.
These experimentally obtained waveforms closely resemble
the simulated waveforms in Fig. 7, with some voltage spikes as
a result of switching dead-time in the practical implementation.
Therefore, the hardware not only shows that soft-charging
operation is able to achieve a high efficiency with smaller
passive component footprint, but also confirms that resonant
operation can be achieved at the specified frequency using the
same technique.
frequency. In addition, we proposed a formal and general
method to determine the conditions for soft-charging operation
Soft-charging
of an arbitrary two-phase switched-capacitor converter and
378.9 the method is illustrated with a Dickson SC converter. It
has been shown that the Dickson converter can approach
453.9 complete soft-charging by appropriately selecting the flying
capacitor values. The technique was verified with a discrete
8-to-1 Dickson converter. The hardware prototype in soft-
charging operation has shown to exhibit higher efficiency with
smaller passive component sizes compared to a conventional
Dickson converter. The analysis is also applied to four other
SC topologies and the results agree with simulation results and
published experimental work. It is found that both the series-
parallel and Fibonacci SC converters are able to achieve soft-
charging or resonant operation using the proposed technique.
The proposed method expands the family of both resonant and
soft-charging SC converters and makes SC converters suitable
for an increasing number of applications.

VII. C ONCLUSIONS R EFERENCES

In this work, we have shown that soft-charging operation
can be achieved with the addition of a single inductor. Us-
ing the proposed technique, ZCS resonant operation for SC
converters can also be achieved at an appropriate switching
95
Efficiency (%)
[1] S. Cheong, S. Chung, and A. Ioinovici, Development of power elec-
tronics converters based on switched-capacitor circuits, in Circuits
and Systems, 1992. ISCAS 92. Proceedings., 1992 IEEE International
Symposium on, vol. 4, May 1992, pp. 19071910.
[2] J. Kimball, B. Kuhn, and P. Krein, Increased performance of battery
packs by active equalization, in Vehicle Power and Propulsion Confer-
ence, 2007. VPPC 2007. IEEE, Sept 2007, pp. 323327.
[3] K. Sano and H. Fujita, Voltage-balancing circuit based on a resonant
switched-capacitor converter for multilevel inverters, Industry Applica-
tions, IEEE Transactions on, vol. 44, no. 6, pp. 17681776, Nov 2008.
[4] M. Chen, K. Afridi, and D. Perreault, Stacked switched capacitor

energy buffer architecture, Power Electronics, IEEE Transactions on,

90 vol. 28, no. 11, pp. 51835195, Nov 2013.
[5] J. Dickson, On-chip high-voltage generation in mnos integrated circuits
using an improved voltage multiplier technique, Solid-State Circuits,
IEEE Journal of, vol. 11, no. 3, pp. 374378, Jun 1976.
Vin= 150V, hardcharging
Vin= 180V, hardcharging [6] V. Ng, M. Seeman, and S. Sanders, Minimum pcb footprint point-of-
Vin= 150V, softcharging load dc-dc converter realized with switched-capacitor architecture, in
Vin= 180V, softcharging Energy Conversion Congress and Exposition, 2009. ECCE 2009. IEEE,
85 Sept 2009, pp. 15751581.
0 10 20 30 40 50 60
Output power (W) [7] Y. Ramadass and A. Chandrakasan, Voltage scalable switched capacitor
dc-dc converter for ultra-low-power on-chip applications, in Power
Fig. 22: Measured efficiency for soft-charging and conven- Electronics Specialists Conference, 2007. PESC 2007. IEEE, June 2007,
tional SC converter prototypes. pp. 23532359.

0885-8993 (c) 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TPEL.2014.2377738, IEEE Transactions on Power Electronics
[8] J. Stauth, M. Seeman, and K. Kesarwani, Resonant switched-capacitor
converters for sub-module distributed photovoltaic power management,
Power Electronics, IEEE Transactions on, vol. 28, no. 3, pp. 11891198,
2013.
[9] S. Ben-Yaakov and A. Kushnerov, Algebraic foundation of self adjust-
ing switched capacitors converters, in Energy Conversion Congress and
Exposition, 2009. ECCE 2009. IEEE, 2009, pp. 15821589.
[10] S. Ahsanuzzaman, J. Blackman, T. McRae, and A. Prodic, A low-
volume power management module for portable applications based on
a multi-output switched-capacitor circuit, in Applied Power Electronics
Conference and Exposition (APEC), 2013 Twenty-Eighth Annual IEEE,
March 2013, pp. 14731479.
[11] M. Seeman and S. Sanders, Analysis and optimization of switched-
capacitor dc-dc converters, Power Electronics, IEEE Transactions on,
vol. 23, no. 2, pp. 841851, March 2008.
[12] V. Ng and S. Sanders, A high-efficiency wide-input-voltage range
switched capacitor point-of-load dc-dc converter, Power Electronics,
IEEE Transactions on, vol. 28, no. 9, pp. 43354341, Sept 2013.
[13] S. Sanders, E. Alon, H.-P. Le, M. Seeman, M. John, and V. Ng, The
road to fully integrated dc-dc conversion via the switched-capacitor
approach, Power Electronics, IEEE Transactions on, vol. 28, no. 9,
pp. 41464155, 2013.
[14] H. Chung, B. O, and A. Ioinovici, Switched-capacitor-based dc-to-dc
converter with improved input current waveform, in Circuits and Sys-
tems, 1996. ISCAS 96., Connecting the World., 1996 IEEE International
Symposium on, vol. 1, May 1996, pp. 541544 vol.1.
[15] K. Cheng, New generation of switched capacitor converters, in Power
Electronics Specialists Conference, 1998. PESC 98 Record. 29th Annual
IEEE, vol. 2, 17-22 May 1998, pp. 15291535vol.2.
[16] Y. Yeung, K. Cheng, S. Ho, K. Law, and D. Sutanto, Unified analysis
of switched-capacitor resonant converters, Industrial Electronics, IEEE
Transactions on, vol. 51, no. 4, pp. 864873, Aug. 2004.
[17] M. Shen, F. Z. Peng, and L. Tolbert, Multilevel dc-dc power conversion
system with multiple dc sources, Power Electronics, IEEE Transactions
on, vol. 23, no. 1, pp. 420426, Jan 2008.
[18] M. Shen, A zero voltage switching switched capacitor voltage doubler,
in Industrial Electronics (ISIE), 2012 IEEE International Symposium on,
May 2012, pp. 131136.
[19] K. Sano and H. Fujita, A resonant switched-capacitor converter for
voltage balancing of series-connected capacitors, in 2009 International
Conference on Power Electronics and Drive Systems (PEDS). IEEE,
Nov. 2009, pp. 683688.
[20] E. Hamo, A. Cervera, and M. Peretz, Multiple conversion ratio resonant
switched-capacitor converter with active zero current detection, in
Energy Conversion Congress and Exposition (ECCE), 2013 IEEE, Sept
2013, pp. 805812.
[21] D. Cao and F. Z. Peng, A family of zero current switching switched-
capacitor dc-dc converters, in Applied Power Electronics Conference
and Exposition (APEC), 2010 Twenty-Fifth Annual IEEE, Feb 2010, pp.
13651372.
[22] D. Cao, X. Lu, X. Yu, and F. Peng, Zero voltage switching double-
wing multilevel modular switched-capacitor dc-dc converter with voltage
regulation, in Applied Power Electronics Conference and Exposition
(APEC), 2013 Twenty-Eighth Annual IEEE, March 2013, pp. 2029
2036.
[23] D. Cao, S. Jiang, and F. Z. Peng, Optimal design of a multilevel
modular capacitor-clamped dc-dc converter, Power Electronics, IEEE
Transactions on, vol. 28, no. 8, pp. 38163826, Aug 2013.
[24] K. Zou, M. Scott, and J. Wang, Switched-capacitor-cell-based voltage
multipliers and dc-ac inverters, Industry Applications, IEEE Transac-
tions on, vol. 48, no. 5, pp. 15981609, Sept 2012.
0885-8993 (c) 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.
[25] M. Alam and F. Khan, A high-efficiency modular switched-capacitor
converter with continuously variable conversion ratio, in Control and
Modeling for Power Electronics (COMPEL), 2012 IEEE 13th Workshop
on, June 2012, pp. 15.
[26] O. Abutbul, A. Gherlitz, Y. Berkovich, and A. Ioinovici, Step-up
switching-mode converter with high voltage gain using a switched-
capacitor circuit, IEEE Transactions on Circuits and Systems I., vol. 50,
no. 8, pp. 10981102, Aug. 2003.
[27] J. Sun, M. Xu, Y. Ying, and F. Lee, High power density, high efficiency
system two-stage power architecture for laptop computers, in Power
Electronics Specialists Conference, 2006. PESC 06. 37th IEEE, June
2006, pp. 17.
[28] F.-L. Luo, Switched-capacitorized dc/dc converters, in Industrial Elec-
tronics and Applications, 2009. ICIEA 2009. 4th IEEE Conference on,
May 2009, pp. 10741079.
[29] M. Saad, A. Shawky, and M. Orabi, Design of integrated pol dc-
dc converters based on two-stage architectures, in Applied Power
Electronics Conference and Exposition (APEC), 2013 Twenty-Eighth
Annual IEEE, March 2013, pp. 20982105.
[30] R. Pilawa-Podgurski, D. Giuliano, and D. Perreault, Merged two-
stage power converter architecture with soft-charging switched-capacitor
energy transfer, in Power Electronics Specialists Conference, 2008.
PESC 2008. IEEE, Jun. 2008, pp. 40084015.
[31] R. Pilawa-Podgurski and D. Perreault, Merged two-stage power con-
verter with soft charging switched-capacitor stage in 180 nm cmos,
Solid-State Circuits, IEEE Journal of, vol. 47, no. 7, pp. 15571567,
July 2012.
[32] Y. Lei and R. Pilawa-Podgurski, Soft-charging operation of switched-
capacitor dc-dc converters with an inductive load, in The Applied Power
Electronics Conference and Exposition (APEC), 2014 IEEE, 2014.
[33] M. Araghchini, J. Chen, V. Doan-Nguyen, D. Harburg, D. Jin, J. Kim,
M. S. Kim, S. Lim, B. Lu, D. Piedra, J. Qiu, J. Ranson, M. Sun, X. Yu,
H. Yun, M. Allen, J. del Alamo, G. DesGroseilliers, F. Herrault, J. Lang,
C. Levey, C. Murray, D. Otten, T. Palacios, D. Perreault, and C. Sullivan,
A technology overview of the powerchip development program, Power
Electronics, IEEE Transactions on, vol. 28, no. 9, pp. 41824201, 2013.
[34] S. Lim, D. M. Otten, and D. J. Perreault, Power conversion architecture
for grid interface at high switching frequency, in The Applied Power
Electronics Conference and Exposition (APEC), 2014 IEEE, 2014.
[35] P. Lin and L. Chua, Topological generation and analysis of voltage
multiplier circuits, IEEE Transactions on Circuits and Systems, vol. 24,
no. 10, pp. 517530, Oct. 1977.
[36] M. Makowski and D. Maksimovic, Performance limits of switched-
capacitor dc-dc converters, in Power Electronics Specialists Conference,
1995. PESC 95 Record., 26th Annual IEEE, vol. 2, Jun 1995, pp. 1215
1221 vol.2.
[37] B. Arntzen and D. Maksimovic, Switched-capacitor dc/dc converters
with resonant gate drive, Power Electronics, IEEE Transactions on,
vol. 13, no. 5, pp. 892902, Sep 1998.
[38] J. Kimball, P. Krein, and K. Cahill, Modeling of capacitor impedance
in switching converters, Power Electronics Letters, IEEE, vol. 3, no. 4,
pp. 136140, Dec 2005.
[39] J. Henry and J. Kimball, Practical performance analysis of complex
switched-capacitor converters, Power Electronics, IEEE Transactions
on, vol. 26, no. 1, pp. 127136, 2011.
[40] M. D. Seeman, A design methodology for switched-capacitor dc-
dc converters, Ph.D. dissertation, EECS Department, University of
California, Berkeley, May 2009. [Online]. Available: http://www.eecs.
berkeley.edu/Pubs/TechRpts/2009/EECS-2009-78.html
[41] N. Sturcken, E. OSullivan, N. Wang, P. Herget, B. Webb, L. Romankiw,
M. Petracca, R. Davies, R. Fontana, G. Decad, I. Kymissis, A. Peterchev,
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TPEL.2014.2377738, IEEE Transactions on Power Electronics

L. Carloni, W. Gallagher, and K. Shepard, A 2.5d integrated voltage

regulator using coupled-magnetic-core inductors on silicon interposer,
Solid-State Circuits, IEEE Journal of, vol. 48, no. 1, pp. 244254, Jan
2013.
[42] J. T. Dibene II, P. R. Marrow, C.-M. Park, H. W. Koertzen, P. Zou,
F. Thenus, X. Li, S. W. Montgomery, E. Stanford, R. Fite, and P. Fischer,
A 400 amp fully integrated silicon voltage regulator with in-die
magnetic coupled embedded inductors, in Proc. Appl. Power Electron.
Conf., Special Session on On-Die Voltage Regulators, 2010.
[43] Y. Lei and R. Pilawa-Podgurski, Analysis of switched-capacitor dc-
dc converters in soft-charging operation, in Control and Modeling for
Power Electronics (COMPEL), 2013 IEEE 14th Workshop on, June
2013, pp. 17.
[44] H. Greinacher, ber eine methode, wechselstrom mittels elektrischer ven-
tile und kondensatoren in hochgespannten gleichstrom umzuwandeln,
Zeitschrift fr Physik A Hadrons and Nuclei, vol. 4, pp. 195205, 1921.
[45] J.-T. Wu and K.-L. Chang, MOS charge pumps for low-voltage
operation, Solid-State Circuits, IEEE Journal of, vol. 33, no. 4, pp.
592597, Apr. 1998.
[46] L. O. Chua, C. A. Desoer, and E. S. Kuh, Linear and Nonlinear Circuits.
Mcgraw-Hill College, 1987.

Yutian Lei (S11) received the B.A. and M.Eng

degrees in Electrical and Information Science from
University of Cambridge, UK, in 2012. He is cur-
rently pursuing a Ph.D. degree at University of
Illinois, Urbana-Champaign. His research focus is
on high performance dc-dc switched-capacitor con-
verters.

Robert Pilawa-Podgurski (S06, M11) was born

in Hedemora, Sweden. He received dual B.S. de-
grees in physics, electrical engineering and computer
science in 2005, the M.Eng. degree in electrical
engineering and computer science in 2007, and the
Ph.D. degree in electrical engineering in 2012, all
from the Massachusetts Institute of Technology.
He is currently an Assistant Professor in the
Electrical and Computer Engineering Department at
the University of Illinois, Urbana-Champaign, and is affiliated with the Power
and Energy Systems group. He performs research in the area of power
electronics. His research interests include renewable energy applications,
energy harvesting, CMOS power management, and advanced control of power
converters. Dr. Pilawa-Podgurski received the Chorafas Award for outstanding
MIT EECS Masters thesis, the Google Faculty Research Award, and the
Richard M. Bass Outstanding Young Power Electronics Engineer Award of
the IEEE Power Electronics Society. He is co-author of three IEEE prize
papers.

0885-8993 (c) 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.