HAIT Journal of Science and Engineering B, Volume 2, Issues 3-4, pp.

529-557
Copyright C 2005 Holon Academic Institute of Technology
Application of duality for derivation
of current converter topologies
Doron Shmilovitz
Faculty of Engineering, Tel-Aviv University,
Tel-Aviv 69978, Israel
e-mail: shmilo@eng.tau.ac.il
Received 1 April 2005, accepted 20 July 2005
Abstract
A unied approach is suggested for the derivation of current-to-
current converter topologies, which relies on duality principles. As
conventional duality rules have restricted utility and are dicult to ap-
ply with switched mode circuits, special attention is devoted to switch
mode converter element dualities. Also, a special means of viewing
complex switched mode topologies as a cascade combination of building
blocks is developed. This method simplies the treatment of circuits
containing transformers and switching bridges.
The suggested approach provides a universal means for derivation of
current-to-current converters by dually transforming voltage-to-voltage
converter topologies.
Examples of the utilization of this method for derivation of vari-
ous current-to-current converters, including Buck, Boost, Buck-Boost,

Cuk, Sepic, and Two-Switch Forward, are provided, as well as a practi-

cal design example for SMES (Superconducting Magnetic Energy Stor-
age) application.
Keywords: Duality, current converter, current source, switched mode
power converters.
529
1 Introduction
To a large extent, DC-DC converters may be regarded as DC transform-
ers, this is due to their high eciency which implies energy conservation, as
well as due to their output voltage being proportional to their input voltage
and vice versa: the input current of a DC-DC converter is proportional to
the output current [1]. This approach is most attractive due to a special
feature of this particular family of transformers: the so called controllable
transfer ratio. This feature provides the most important means for the im-
provement of the systems performance (through the application of various
control schemes to tune continuously the converters transfer ratio). How-
ever, unlike the ideal transformers mathematical model (which allow for the
connection of any type of source at its input as long as it complies with the
load characteristics at the output), converter realized transformer behavior
is realization dependent. Namely, the types of source and loads suitable for
connection at the converters terminals depend upon the converter imple-
mentation. Since the majority of power sources are actually voltage sources,
most converter topologies have evolved under the assumption that the con-
verter input should accommodate a voltage source at its input (this is the
reason that most of the converters are based on a switched inductor cell [1]).
This, in addition to the transformer-like behavior of these converters, yields
voltage source characteristics at the converter output as well. Thus appli-
cation of a rather sti current type of source is not possible for most of the
converters. However, current source types of power sources do exist [1-5].
They include large inductors initially charged to high currents such as su-
perconducting coils in Superconducting Magnetic Energy Storage (SMES)
systems. These systems are applied to either provide power during short
voltage sags or for the generation of high magnetic elds [2], see Fig. 1.
Large utility (or generators) source impedances (which are, to a large
extent, of inductive nature) feeding a switched mode converter would also
exhibit current source characteristics at the converters input. Additionally,
the DC input to switched-mode converters is often obtained by rectication
and ltration the AC mains. To reduce current harmonics and EMI noise,
the rectier most often includes a large lter inductor (choke), at either
side of the rectifying bridge, which might eventually introduce a current
source like behavior at the rectier output the switched mode converters
input [5].
530
Figure 1: A superconductive magnetic energy storage coil and its current
source like I-V characteristics.
In current fed converters a constant current is established through an
inductor, whose continuity is ensured by the overlapped on-states of the
switches that direct this current. In this kind of circuit, the current source
is practically established at an inner point of the circuit and the sub-circuit
from this point on may be regarded as a current converter. A similar situa-
tion occurs in some resonant converter topologies [1,7,8].
Other power sources exhibit current source behavior when operated in a
certain region of their I-V curve, for instance, a photovoltaic generator, as
depicted in Fig. 2.
This type of source is sometimes regarded as a current source, which, for
instance, aects the power system topology of solar array powered earth-
orbiting spacecrafts. In these systems the DC bus voltage is regulated by
dissipative shunt elements and the solar panels outputs are added together
in a parallel connection [1,4,9]. At the converters output, some loads may
be better and rather stable operated while energized by a current source
rather than voltage sources. Such characteristics are required when ener-
531
Figure 2: A solar panel and its current source like region of operation.
gizing loads with constant voltage like spark gaps and electric arcs. So are
gas discharge devices (like gas lasers) that exhibit negative impedance at
certain operation points. Application of voltage type sources to such loads
might cause instable oscillations (such applications are reported in [10,11],
where transmission line segments have been employed to form a high fre-
quency electronic ballast for HID lamps). It has been attempted to address
this problem by application of current mode control, which provides a fair
solution in some cases [12]. Yet a converter topology that exhibits inher-
ent current source characteristics at its output (the kind that results from
the power stage topology rather than control implied) should oer much
more stability and robustness. Indeed, a few attempts at the development
of current-to-current converter topologies have been reported, which incor-
porate duality principle [2,3,5,9,13-16]. The theory of duality is known to be
an important and a widely used one for analysis, as well as synthesis, within
the eld of electrical networks since long ago [17,18]. Later on, duality has
532
been employed in the eld of switched mode converters for analysis purpose.
This promoted the understanding of switched mode converter circuits and
the relations between dierent converter topologies [19,20]. The potential
of duality as a tool for the development of power converters topologies was
realized later on. Major contributions to duality in the context of power elec-
tronics topologies were made by Freeland [21,22] and by Wolfs et al. [14-16].
The former developed novel dual models for power converter building blocks
such as semiconductor switches, switching bridges, and practical inductors
and transformers [21,22]. In the second group of publications, [16], duality
has been extensively employed for the derivation of a current-sourced link
converter [14,15]. The possible application of duality to nonplanar topolo-
gies has been facilitated as well [16].
In this work, a systematic approach is suggested for derivation of current-
to-current converter topologies from known voltage-to-voltage converter
topologies, employing a simple and consistent set of rules arising from du-
ality principle. As voltage-to-voltage converters have been around for over
three decades, a large variety of such topologies already exist. Therefore the
approach suggested herein, of deriving current-to-current converter topolo-
gies from existing voltage-to-voltage converter topologies, seems reasonable.
An alternative procedure to the previously proposed -T dual transforma-
tion [5,22] based on graph transformation is proposed. This procedure might
be simpler to employ for some of the power conversion topologies. Another
simplication is oered by viewing a given circuit as a cascade combination
of few two-port-subcircuits. In most of the cases, the dual transformation
may be reached by application of duality to each of the two-port-subcircuits,
then reconnecting them in a cascade. In such a manner, the dualities of cir-
cuits including complex structures, such as switching bridges, can be rather
easily derived.
2 Duality principle
Some of the basic duality rules (applications of duality to linear components)
are reviewed herein. Then, an application method of duality to nonlinear
components and to complex graphs, which are of interest for converter syn-
thesis, is derived. Distinction is made between graph transformation and
network elements transformation (components transformation).
533
2.1 Network elements (components) transformation
Passive components duals are obtained through their reciprocal impedances:
Z = Y
D
Z
D
(j) = (Z (j))
1
(1)
where Z and Y represent impedance and admittance, respectively, and the
superscript
D
stands for dual. Equation (1) implies that inductors map
into capacitors of equal numerical value:
L
D
= C ; C = 1
F
Hy
L . (2)
Capacitors map into inductors of equal value:
C
D
= L ; L = 1
Hy
F
C . (3)
Ideal transformers map into transformers with inverse transfer ratio:
1 : k k : 1 . (4)
Transformers are therefore invariant to dual transformation (in terms
of component nature rather than value, similarly to pure resistances and
unlike reactive components). It should be noted that practical transform-
ers inherently contain stray inductances (leakage inductances in series with
the primary and secondary windings and magnetizing inductances in paral-
lel with them). These inductances will remain in the dual transformer and
might cause high voltages if currents are forced through them. This diculty
regarding transformer dualities has been observed and comprehensively an-
alyzed in [22] (as well as the nonlinear phenomena of magnetic saturation).
Transformers that are based upon capacitive coupling, such as piezoelectric
transformers, might enable a better realization of transformer duals.
Voltage source duals are current sources of equal value and vice versa;
current source duals are voltage sources of equal value.
(V
S
)
D
= I
S
; I
S
= 1
Amp
V olt
V
S
, (5)
(I
S
)
D
= V
S
; V
S
= 1
V olt
Amp
I
S
. (6)
Switches and switching bridges are frequently employed building blocks
of power converters. Therefore they deserve special attention. Equation
(1) indicates that the reciprocal of zero impedance is an innite impedance
(open circuit) and that of an open circuit is a 0 impedance (short circuit).
534
Therefore, a closed switch maps into an open one and vice versa, an open
switch maps into a closed one (see Fig. 3):
x
D
= x. (7)
Figure 3: Main components of switched mode converters and their duals.
535
2.2 Graph transformation
The primitive (original) network should be described as a graph combination
of branches. A dual graph may then be derived by transforming every
couple of branches connected in parallel into a series connection of the dual
branches and vice versa: transforming every couple branches connected in
series into a parallel connection of the dual branches. This process must
be carried out successively, until the entire graph is represented by basic
branches (branches that consist of no more sub-branches). For instance,
the graph in Fig. 4a transforms into that of Fig. 4b where a, b, c and d
represent three branches in the original network and a
D
, b
D
, c
D
and d
D
represent their duals, respectively.
Figure 4: Dual graph transformation.
Two-port elements (TPE), such as transformers and switching bridges
call for special treatment. Whenever possible, a network that contains such
a TPE (two-port-element) should be divided into two separate sub-networks
coupled by the two-port element, see Fig. 5a. By duality, this network is,
than transformed into that of Fig. 5b. If the TPE is a transformer, its dual
is a transformer as well, as indicated by (4).
Consider a switching bridge, implemented by two single-pole-dual-throw
switches as shown in Fig. 6.
For a planar circuit it is possible to nd its dual by graph transformation
[22] or by oriented graph as discussed in [18]. An alternative, easy to apply
method is suggested: while the switches are in position 1, the bridge input
voltage (V
G
, in this case) is applied to the bridge output, whereas while the
536
Figure 5: Two port elements coupling two subcircuits (a), and its dual
transformation (b).
Figure 6: A switching bridge.
switches are in position 2, the polarity is reversed. Therefore, its action is
equivalent to that of a virtual transformer with a controllable turns-ratio k(t)
537
k (t) =

1 s = 1
1 s = 2
(8)
as depicted in Fig. 7.
Figure 7: The virtual variable transformer model of the switching bridge
and its dual.
Since k is either 1 or (-1), the virtual transformer maps into a virtual
transformer with the same instantaneous turns ratio (4). This inverting
switching bridge is thus invariant to dual transformation. To ensure practical
operation, duality should be applied not only to circuit elements but also to
switching functions. If, for instance, a in Fig. 7 is a voltage source, a dead
time must be implemented in the switching bridge in order to prevent shoot-
through currents. Whereas, in the dual circuit (Fig. 7b), a
D
is a current
source, and an overlapping of the on states will automatically result by
application of duality to the switching function, ensuring current continuity.
Switching bridges are often employed in power converters in conjunction
with a transformer to form an alternating signal at the transformers input.
This is actually a cascade connection of two two-port-elements (Fig. 8).
Applying the duality rule for cascade connection successively (for each TPE)
would result in a similar cascade topology in which each of the networks in
each of the two two-port-elements is replaced by its dual element, see Fig. 8.
That is correct in principle, however, details of implementation like those
mentioned above must be considered. Similarly, switching bridges that gen-
erate zero levels in addition to their input or inversed input, can be treated
by analyzing their functionality rather than by the straightforward applica-
tion of duality. For instance, a three level voltage source inverter generates
at the bridge output a waveform as shown in Fig. 9 where x = z and y = w
in order to prevent a short across V
G
.
538
Figure 8: A cascade connection of two two-port-elements and its dual.
Figure 9: A three level inverter (a) and its output voltage waveform (b).
The varying turns ratio transformer approach cannot be employed in
this case since it would involve a turns ratio of either 0 or . In the dual
circuit, a current source should appear at the b-b terminals which has the
precisely same waveform as the voltage in the original circuit (Fig. 9b). This
539
would be accomplished by replacing the voltage source V
G
with a current
source I
G
and exercising the bridge switches dierently, as shown in Fig. 10.
The new sequence would be: x, w x, z y, z and y, w. In
this dual circuit x = y and z = w in order to ensure always a closed path
for I
G
. This approach can serve for the development of various topologies
involving switching bridges, such as current source inverters or unity power
factor rectiers, which practically implement a sinusoidal current source with
respect to the AC mains.
Figure 10: The three level inverter (a) and its output current waveform (b).
3 Derivation of some current-to-current converter
topologies
Development of current-to-current converter topologies can be either done
from scratch or accomplished by applying some symmetry principles such as
horizontally ipping of portions of a voltage converter [4] or by manipulation
of a switched capacitor cell [5]. While doing so, some basic requirements
must be ensured:
the converter topology must always provide a current path for the
input current source, regardless of the switches states;
no inductor may connect to the input;
output parameters (current or voltage) should be controllable via a
switching parameter such as the duty ratio, D.
540
Figure 11: Successive application of the method for a buck topology.
541
Additionally, the system must be stable under dynamical conditions.
When deriving a current-to-current converter topology via the duality method
proposed herein, the above mentioned requirements are automatically ful-
lled, without special care. Moreover, the same DC voltage gain that applied
for the original converter V
O
/V
G
should be valid for the DC current gain
I
O
/I
G
of the dual converter. Applying the duality rules stated in the previ-
ous paragraph, yields some current-to-current converter topologies. Herein
are some examples. Consider the voltage buck converter represented by
three branches, a, b, and c, as in Figs. 11a and 11b. The original circuits
graph is depicted in Fig. 11b. Since these branches are connected in parallel,
the graph maps into its dual as shown in Fig. 11c. Branch a consists of two
branches a
1
and a
2
connected in series, which maps therefore as a parallel
connection of the dual branches a
D
1
and a
D
2
(Fig. 11d). Similarly, branch c
consists of two branches c
1
and c
2
connected in series, which map into a
parallel connection of their duals c
D
1
and c
D
2
, as shown in Fig. 11e. c
2
itself
consists of two branches c
21
and c
22
which maps into c
D
2
according to the
same rules (Fig. 11f).
Figure 12: Graph transformation of the buck topology: (a) original circuit;
(b) dual transformation.
542
The graph of the original buck topology can now be represented in terms
of basic branches by substitution of each subgraph in the place of the equiv-
alent graph, as shown in Fig. 12a. Performing the same substitution process
for the dual sub graphs results in the total dual graph as depicted in Fig.
12b.
In this last graph, each branch is composed of a single network element,
which means that each branch can be replaced by the dual network element
according to the network elements transformations shown in Fig. 3, resulting
in the dual converter topology depicted in Fig. 13.
Figure 13: The resulting dual buck topology.
Regardless of component values, the current-to-current topology has
been derived, as shown in Fig. 14. It will be shown that it has the DC
gain of a buck topology with respect to output to input current ratio.
Figure 14: Buck current to current converter.
Similarly, the conventional boost voltage-to-voltage topology and its
graph description are shown in Fig. 15. The dual of this graph is derived
543
via the duality graph transformation rules (Fig. 16a), and nally replace-
ment of each branch by the suitable dual component results in the boost
current-to-current converter topology depicted in Fig. 16b.
Figure 15: A voltage boost topology (a) and its graph representation (b).
Figure 16: Dual boost derivation: (a) graph transformation; (b) component
placement.
A buck-boost voltage-to-voltage topology and its branch assignment are
shown in Fig. 17. It may be observed by inspection that this topology is
described by the same graph as the boost converter (Fig. 15b), thus the
resulting dual graph is also the same (Fig. 16a).
Finally, the buck-boost current-to-current converter topology is obtained
by placing the suitable component duals, see Fig. 18 (it may be noted that
the only dierence with respect to the boost topology arises from substitu-
tion of components for branches a
2
and b).
Higher order current-to-current converter topologies can be derived
through the same method as well. For instance, when continuous wave-
forms are desired at the source terminal as well as at the load terminal, a
544
boost-buck (

Cuk) voltage-to-voltage converter might be chosen (Fig. 19).

Figure 17: A voltage buck-boost topology and its branch assignment.
Figure 18: The resulting current buck-boost topology.
Figure 19: A voltage

Cuk topology (a) and its graph description (b).
545
Following the duality rules, the boost-buck current-to-current converter
topology of Fig. 20 is reached, which exhibits continuous input and output
voltage and current (in CCM), in similarity to the original voltage-to-voltage
converter.
Figure 20: The resulting

Cuk current topology.
The SEPIC voltage-to-voltage converter shown in Fig. 21 may be de-
scribed by the same graph as the

Cuk one (Fig. 19b) and the dual current-
to-current converter topology described by the graph of Fig. 20. After
components substitution, a current-to-current converter topology dual to
the SEPIC is achieved, see Fig. 22.
Figure 21: A voltage single ended primary inductance topology.
Figure 22: The resulting single ended primary inductance current topology.
546
4 A practical example
For a certain SMES project it was required to supply a DC, constant voltage
load from a superconducting large coil, while controlling the rate of power
ow and regulating the DC output voltage. It is motivated by the replace-
ment of the storage batteries in an UPS system with a Superconducting
Magnetic Energy Storage (SMES). Other applications include energizing of
critical DC loads during voltage sags or outages.
4.1 Power stage
As the superconducting coil is a rather sti current type of source, a current-
to-current converter topology was required. The buck dual of Fig. 14 was
chosen, with the switch implementation shown in Fig. 23.
Figure 23: Power stage: (a) switch implementation; (b) two operation states;
(c) lter input current.
547
The generator current is alternately diverted from the switch Q to the
diode and vice versa. During the switch on state, the switch provides a
closed path for the source current and no energy is being supplied by the
source. During the switch o state, the source current is forced into the
capacitor-inductor junction through the diode, delivering power from the
source.
p
g
(t) =

0 s = 1
v
c
(t) i
g
(t) s = 0
. (9)
The average lter input current over a switching period is derived:
hi
f
(t)i
T
S
= h(1 d (t)) i
g
(t)i
T
S
. (10)
Since the input current hardly changes during a switching period and
the duty ratio is actually dened as the on time during a switching period,
equation (8) can be simplied:
hi
f
(t)i
T
S
= (1 d (t)) i
g
(t) . (11)
With appropriate lter components, the ripple component of i
f
ows
through the capacitor while the DC component ows through the inductor
and the load. The DC transfer function is derived:
V
o
= (1 D) I R
l
. (12)
Equation (12) shows that the load voltage is controllable via the duty
ratio and thus the topology complies with the requirement no. 3. However,
(12) also indicates that unlike as in conventional switched mode converters,
the output voltage is inherently load dependent, thus output regulation is
possible only with a closed loop mode. It can be seen from Fig. 23 that
a pulsating current i
f
is injected at the capacitor-inductor junction. As
long as the capacitor presents an impedance, that is much lower than the
inductor and load series combination, all of the current AC components may
be assumed to ow through the capacitor while the current average ows
through the inductor and load. Thus the capacitor current is a square wave
with a peak-to-peak amplitude equal to that of the source i
g
and zero DC
component, see Fig. 24.
The resulting capacitor voltage ripple can be formulated as
v
c
=
d (1 d) I
g
2 f
S
C
. (13)
548
Figure 24: Main waveforms in the converter. Top to bottom: lter input
current, capacitor current, and capacitor voltage ripple.
The maximum capacitor voltage ripple is:
v
c
=
I
g
8 f
S
C
. (14)
Not all of this ripple develops on the load; there is an additional ripple
reduction due to the inductor impedance. The maximum output ripple can
be approximated as:
v
loadmax
=
Z
load
()
Z
load
() + jL

I
g
8 f
S
C
. (15)
These expressions can serve for the selection of capacitor, inductor, and
switching frequency according to the allowable ripple.
4.2 Power manipulation and SMES coil current decay prole
With given load, regulation of the output voltage means supplying a con-
stant output power. Assuming low losses within the converter, this means a
steady power throughput and a constant decrease of the energy stored in the
549
superconducting coil reservoir. The energy stored in the superconducting
coil and the rate of discharge are
E
SMES
=
L
SMES
i
g
(t)
2
2
;
d
dt
(E
SMES
) = p
out
=
V
2
load
R
load
. (16)
which yields the superconducting coil current:
i
g
(t) =
r
i
g
(0)
2

2 P t
L
SMES
=
s
i
g
(0)
2

2 V
2
load
t
L
SMES
R
load
. (17)
The duty ratio variation over the discharge time is found by substituting
(17) into (11).
d (t) = 1
V
load
q
i
g
(0)
2

2Pt
L
SMES
R
load
= 1
V
load
q
i
g
(0)
2
R
2
load

2V
2
load
R
load
t
L
SMES
(18)
which indicates that output regulation is lost at t
cr
, shortly before all the
stored energy is exhausted.
t
cr
=
1
2

i
g
(0) L
SMES
R
l
V
2
l

L
SMES
R
l

. (19)
These equations are in agreement with the SPISE simulation results
shown in the next section.
4.3 Small signal analysis
Figure 25 shows a low-frequency model of the proposed converter in which
the switching network was replaced by two dependent sources.
Figure 25: Low frequency AC model of the converter.
Accounting to frequencies much lower than the switching frequency [1],
this model provides a fair approximation for both large and small signals.
550
For the small signal analysis an averaged switch network replaces the switch,
which is equivalent to perturbation and linearization process [1], see Fig. 26.
Figure 26: Averaged switch model (a) and the converter small signal
model (b).
In addition, the model was augmented by the equivalent series resistance
(ESR) of the capacitor, r
C
. Modeling of the inductor resistance is not needed
since it is included in the load resistance. After some circuit manipulation,
a simplied model which models the small signal behavior is reached and
shown in Fig. 26b. This model has a simple transfer function with two poles
and one zero in the LHP.
v
load
=
r
C
R
load

D
0

d
0

L

S +
1
r
C
C
S
2
+

r
C
+R
load
L

S +
1
LC
i
g
. (20)
As it can be observed, this topology is unconditionally stable for duty
ratios in the range of 01.
4.4 Simulation and experimental results
The circuit is meant to connect to a UPS DC bus at its output, therefore
the converter output is regulated at 400 Volt. The power stage linearized
transfer function (19) was employed for the derivation of the controller.
551
Figure 27: The average model simulation scheme.
Figure 28: The average model simulation results.
552
The circuit was simulated by SPICE software, rst with averaged models
and later using switching models. The average model is shown in Fig. 27,
the focus is on the converter section while the rest of the scheme is about
charging and protecting the SMES coil.
The SMES coil is assumed to have an inductance of 0.5H and be initially
charged to almost 50A. The average mode simulation results are shown in
Fig. 28.
The satisfactory load voltage regulation can be noticed as well as the
duty ratio and current variation, in accordance with (17) and (18). For the
switching detailed simulation, a switching circuit with realistic component
models replaced the average model of Fig. 28. A simple voltage mode
modulator was incorporated. The output is well regulated until all the
superconductor energy is exhausted (at t 86ms). The parabolic prole of
current decrease may be noticed, in accordance with (17) and (18) and with
the average model simulation, see Fig. 29.
Figure 29: Converter input voltage (ch1 20V/div), and input current (ch2
2A/div).
Experimental results are presented for a partial loading of 110 Watt,
down sized prototype. In the experimental setup a 470F capacitor and
20H inductor were used and the switching frequency was 70kHz. The
input current I
G
is 4A. The output is regulated at 50V and the load is a
22 resistor. Typical waveforms are presented in Fig. 30.
553
Figure 30: Simulation results in the switched mode.
Figure 31: Experimental steady state transfer ratio of the dual buck at
I
g
= 4A.
554
The steady state transfer ratio from input current to output voltage
is shown in Fig. 31, which agrees with (12). Due to simplicity and low
component count, high eciency is achieved even with hard switching, see
Fig. 31.
A dual buck-boost current converter was build and tested at the same
power level as well.
Comparison between theoretical and experimental results is provided in
Fig. 32.
Figure 32: Experimental steady state transfer ratio of the dual buck-boost
converter at I
g
= 4A.
5 Conclusions
A unied approach for the derivation of current-to-current converter topolo-
gies was presented. In this method, previously known duality principles
are augmented by the application of cascaded two port elements and func-
tional analysis of circuits, as well as by the successive application of se-
ries/parallel transformation. Switching bridges are modeled by a virtual
variable-transformer, which greatly simplies the treatment of circuits con-
taining transformers and switching bridges. Thus, a universal tool for
555
derivation of current-to-current converters by dually transforming voltage-
to-voltage converter topologies was developed. Some examples for the ap-
plication of this method show its consistency.
Experimental results for a practical design of a current-to-current con-
verter for SMES (Superconducting Magnetic Energy Storage) application
were also provided.
References
[1] R. Erickson, Fundamentals of Power Electronics (Chapman and Hall,
N.Y., 1997).
[2] M. Ehsani and R.L. Kustom, Converter Circuits for Superconductive
Magnetic Energy Storage (Texas A&M University Press, 1988).
[3] M. Ehsani, R.L. Kustom and R.E. Fuja, IEEE Trans. on Industry Ap-
plications 19, 690 (1983).
[4] D. Shmilovitz and S. Singer, 17-th Annual Applied IEEE Power Elec-
tronics Conference and Exposition, APEC2002, p.630 (2002).
[5] R. Rabinovici and V. Tentser, 8th Int. Conference on Optimization
of Electric and Electronic Equipment (Optim2002), Brasov, Romania,
16-17 May, 2002, p.269 (2002).
[6] M. Ehsani, J. Mahdavi, I. Pitel, J.E. Brandenberg, and F.E. Little,
IEEE Aerospace & Electronic Systems Magazine 13, 37 (1998).
[7] M. Milanovic, I. Godec, and F. Mihalic, 27-th Annual IEEE Power Elec-
tronics Specialists Conference (PESC96), Baveno, Italy, 23-27 June
1996, vol.2, p.1294 (1996).
[8] M. Kazimierczuk, and D. Czarkowski, Resonant Power Converters (Wi-
ley, N.Y., 1995).
[9] M. Ehsani, M.O. Bilgic, A.D. Patton, and J. Mitra, IEEE Aerospace
& Electronic Systems Magazine 10, 3 (1995).
[10] S. Singer, D. Shmilovitz, Y. Ifrah, and I. Cohen, IEEE Trans. on Power
Electronics 10, 239 (1995).
[11] H. Ohguchi, M.H. Ohsato, T. Shimizu, and G. Kimura, IEEE Trans.
on Power Electronics 13, 1023 (1998).
556
[12] L. Martinez-Salamero, J. Calvente, R. Giral, A. Poveda, and E. Fossas,
IEEE Transactions on Circuits & Systems I 45, 355 (1998).
[13] R. Rabinovici and B.Z. Kaplan, IEEE Electronics Letters 27, 1948
(1991).
[14] P.J. Wolfs, K.C. Kwong, and G.F. Ledwich, IEEE Transactions on
Power Electronics 7, 683 (1992).
[15] P.J. Wolfs, IEEE Transactions on Industrial Electronics 40, 139 (1993).
[16] P.J. Wolfs, G.F. Ledwich, and K.C. Kwong, IEEE Transactions on
Power Electronics 8, 104 (1993).
[17] C.A. Desoer and E.S. Kuh, Basic Circuit Theory. New York, p.444
(McGraw Hill, N.Y., 1969).
[18] N. Balabanian and T.A. Bickart, Electrical Network Theory. p.122 (Wil-
ley, N.Y., 1969).
[19] S.

Cuk, IEEE Power Electronics Specialists (PESC79), p.109 (1979).
[20] R.P. Severns and G.E. Bloom, Modern DC to DC Switched Mode Power
Converter Circuits (Van Nostrad Reinhold, N.Y., 1985).
[21] S.D. Freeland, IEEE Power Electronics Specialists (PESC89) p.114
(1989).
[22] S.D. Freeland, IEEE Transactions on Power Electronics 7, 374 (1992).
557