IEEE TRANSACTIONS ON MAGNETICS, VOL. MAG-19, NO.

2, MARCH 1983 57
A New Zero-Ripple Switching DC-to-DC Converter
and Integrated Magnetics
Abstract-A new switching dc-todc converter is synthesized which
consists of the least number of storage elements (inductive and capaci-
tive) and switches, and yet truly emulates the ideally desired dc-todc
transformer having both input and output currents as pure dc quantities
with no ripple. This result was facilitated by implementation of a new
concept termed integrated magnetics, which leads in some special
switching structures to the integration of otherwise independent and
separate magnetic components (inductors and transformers) into a
single magnetic circuit.
I. INTRODUCTION
V
IRTUALLY any electronic or electrical equipment has
some form of dc power requirements, often at multiple
dc voltage and current levels. The area of dc-to-dc power con-
version therefore constitutes a major part of the power elec-
tronics field.
The physical impossibility of a magnetic dc-to-dc transformer
has led to indirect or alternative ways of achieving the same
function through the process of switched-mode power con-
version. An input dc voltage is first converted to ac through
some switching mechanism, ac power is processed through the
usual ac transformer and finally, by some effective low-pass
filtering, the output dc voltage is recovered. However, while
this voEtage conversion approaches the ideally desired one,
except for the ever-present switching ripple voltage at the
output, the current conversion is far from the ideal one repre-
sented by constant dc input and output currents as shown
in Fig. 1. In fact, some of the major problems which limit
switchers are the large pulsating currents at either input of
output or both ports of commonly used switching dc-to-dc
converters, which in turn lead to severe conducted and radi-
ated electromagnetic interference (EMI) problems.
The search for inherently low noise and more efficient new
switching converter topologies was initiated several years ago
[l ] and was originally motivated by the desire to reduce or
possibly eliminate these formidable pulsating currents. The
ultimate objective envisioned was to synthesize the generalized
dc-to-dc transformer function with the ideal constant (dc)
current waveforms shown in Fig. l(a), in an optimum manner,
Manuscript received August 13, 1982. The paper was presented at
the IEEE Power Electronics Specialists Conference, 1980 Record,
Atlanta, GA, J une 1980. This work was sponsored in part by the Office
of Naval Research, Washington, DC,under Contract N00014-78-C-0757,
by the Naval Ocean Systems Center, San Diego, CA, under Contract
N66001-78-(2-0351 JAP, and by the International Business Machines
Corporation, Kingston, NY.
Caltech, Pasadena, CA 9 1125.
The author is with
the California Institute of Technology, 116-81
I E t i2k t
.L
I -&
I :
i2k
(a)
t t
Fig. 1. Objective of dc-to-dc conversion-the search for converter
topology realized by physical hardware (b) which will approach its
ideally desired goal: dc-to-dc voltage transformer (a) with dc currents
at both input and output.
by use of theleast number of switches and storage components
(Fig. l(b)) in order to maximize the efficiency.
Unlike some other approaches in which the emphasis is on
the switching mechanism only [ 20] , and the filtering is added
after the fact to clean upinput and output current or voltage
waveforms, we have considered storage elements (inductors,
transformer, and capacitors) together with the switches as an
integral part of the overall problem, as shown by the block
diagram of Fig. l(b). Furthermore, the utmost importance is
0018-9464/83/0300-0057$00.75 0 1983 IEEE
58 IEEE TRANSACTIONS ON MAGNETICS, VOL. MAG-19, NO. 2, MARCH 1983
attached to the topology of the component interconnection.
In Fig. I(b), for example, both minimum component selection,
topology of their electrical connection, and (considered here
for the first time) the topology of the magnetic connections
has yet to be determined in order to emulate the ideally
desired generalized dc-to-dc transformer of Fig. l(a). The
problem at first may seem formidable, but with some back-
ground material covered in Section 11, and the grasp of the
generality of the coupled-inductor method in Section 111,
the contours of the solution will begin to appear.
The first major step in that direction was the invention
of a new optimum topology (Cuk) converter, which has sub-
stantially reduced conducted EM1 problems by creating non-
pulsating input and output currents (average dc currents with
superimposed triangular ripple currents). Both the basic Cuk
converter and its dc isolated version are briefly reviewed in
Section 11, since they constitute the major building blocks for
the subsequent development.
Although originated from the investigation of the unique
topology of the Cuk converter, coupling of inductances in
switching structures is shown in Section I11to be a powerful
new concept. It not only leads to converter ripple current
improvement (with even the possibility of zero ripple current
at one end) but, in addition, has the potential for further
reduction in size, weight, and losses of the magnetic content of
switching structures, as investigated in Section 111. From the
conceptual viewpoint, it is recognized that the coupled-inductor
extension of the Cuk converter represents the first time that
the two magnetic components (inductors), which are normally
and exclusively used separately to perform their function,
have been integrated into a single magnetic circuit with two
windings.
The recognition of these facts leads in Section IV to the
search for switching configurations in which such integration
can be achieved at an even higher level, by incorporating both
separate inductors and ac transformers into a single magnetic
circuit, with additional size and weight reduction and perfor-
mance improvement (ripple current reduction and increase of
efficiency). A number of zero-ripple switching converter con-
figurations are introduced, featuring both isolated and noniso-
lated extensions. Finally, Section IV is concluded with some
multiple output extensions. Fundamental practical verification
of zero-current ripple is made in Section V through several
experiments.
The new and general method which emerges as a natural
growth of the coupled-inductor concept is termed the inte-
grated magnetics concept in clear analogy with the integrated
semiconductor terminology, as is justified in Section VI, where
this general method is postulated. A number of its applications
to other switching dc-to-dc converters as well as tq functionally
different switching dc-to-ac inverters such as the Cuk amplifier
is shown to lead to similar improvements. Finally in Section
VII, a survey of the family of new converter configurations is
made, featuring their key topologies.
Thus this paper describes the culmination of research along
two fundamental lines developed in parallel. The first is a
practical line which is concerned with synthesis of useful new
converter/inverter configurations, and the second is a theoreti-
NEW SWITCHING DC-TO-DC CONVEPTES
DC VOLTAGE
GAIN
- _ v -2
vg D
SWITCH DRIVE
rr
INPUT CURRENT ,.
OUTPUT CURRENT.
DC CURRENT
GAIN
12 D
I1 D
Fig. 2. Basic Cuk converter features nonpulsating input and output
currents in the simplest topology.
cal one which deals with new concepts of general applicability
to switching structures.
11. REVIEW OF THE BASIC CUK CONVERTER AND ITS
DC ISOLATED EXTENSION
The hardware implementation of the new converter using a
transistor/diode combination to implement the ideal single-
pole double-throw switch is shown in Fig. 2. With the transis-
tor off (open), C1 is charged by the input current through the
forward-biased diode. When the transistor turns on, the volt-
age on capacitance C1 reverse-biases the diode, turning it off,
and the previously stored charge on C1 is discharged through
the transistor to the load. Thus capacitive energy transfer in
this converter is fundamental to its operation just as inductive
energy transfer is essential for operation of the three basic con-
verter types (buck, .boost, and buck-boost). The presence of
the two inductors in series with the input and output ports, re-
spectively, leads to one of its important distinguishing features-
nonpulsating input and output currents-as seen in Fig. 2.
The important dc isolation feature, very often required in
practice, can easily be introduced in an optimum manner
following the three key steps illustrated in Fig. 3. The first
step is to separate the coupling capacitance into two series
capacitances C, and C, , and thus to make the original sym-
metrical configuration divisible into two halves; each one is a
mirror image of the other, as seen in Fig. 3(a). The second
step is to recognize that the connection point between these
two capacitances has an indeterminate dc voltage. This inde-
terminate voltage can then be set at zero, by placing an induc-
tance between this point and ground, as shown in Fig. 3(b).
The third and final step is merely the separation of the extra
inductance into two equal transformer windings, which thus
provide the desired dc isolation between input and output as
illustrated in Fig. 3(c). DC isolation obtained in this way not
only has some outstanding features when compared to other
popular isolated switching structures [4] but also has preserved
completely The overall features of the basic converter. Thus
some other important extensions applicable to the basic con-
verter configuration are equally applicable, as shown next, to
its isolated version.
CUK: ZERO-RIPPLE SWITCHING DC-TO-DC CONVERTER 59
"1 I "1 2
r{+"-
&
vQ ,TI-
R
+
I CI
I
fi + 11 -]
(a) - Y , +V f
L2
"* R
+
v, R
Fig. 4. Observation of the proportional (1:l) inductor voltage wave-
forms leads to new concept of coupling inductors.
+
(C)
Fig. 3. Three key steps leading to dc isolation in original Cuk converter.
111. COUPLED-INDUCTOR METHOD
I t would seem from the above exposition that the simplest
possible converter circuit has been obtained in the form of the
basic Cuk converter (Fig. 2), but this is not the case. Consider-
ation of the voltage waveforms across the two inductors L1
and Lz over the switching cycle reveals that for the average dc
voltage across each inductor to be zero (steady-state condition,
the so-called volt-second balance on inductors), the two wave-
forms must be identical, as seen in Fig. 4. This direct 1:l pro-
portionality of the inductor voltage waveforms was the original
motivation for coupling the two inductors by putting them on
the same core. What is obtained appears to resemble a conven-
tional ac transformer, at least as far as the customary symbol
for an ac transformer suggests in the converter of Fig. 5. How-
ever it becomes rather quickly obvious that it is a quite peculiar
kind of transformer, since it must pass the dc current at both
primary (average input inductor current) and secondary (aver-
age output inductor current). Furthermore, for the polarity
marks (dots) as shown in Fig. 5, secondary current i 2 is flowing
into the dot terminal, rather than in the opposite direction as
it would for a conventional ac transformer, where the dot
designates positive voltage polarities.
Any conceptual difficulties raised by this are easily removed
by observing that it is a coupled-inductor structure. Only the
ac properties are affected by this coupling method, and not
the basic dc conversion properties which remain the same as
before the coupling. Hence the converter configuration in Fig.
5 is a perfectly valid switching configuration, and deserves
closer examination, particularly as far as its ac or inductor
ripple current properties are concerned. However before going
into detailed analysis, let us make one important observation.
+
Fig. 5 . Coupling of inductors in basic 6uk converter results in this
switching converter.
The coupled-inductor extension of the optimum topology
(Cuk) converter represents the first time that the two magnetic
components (inductors), which are nominally and exclusively
used separately to perform their function, have been integrated
into a single magnetic circuit with two windings. This single
magnetic circuit still preserves the original two inductive func-
tions but alters their effective values.
In addition to the original capacitive energy transfer between
input and output ports, the inductor coupling brings in a simul-
taneous inductive energy transfer. Thus one may intuitively
feel that this coupling method will lead to reduced ripple cur-
rents at both input and output, which is indeed the case. The
quantitative aspect of the current ripple reduction can easily
be analyzed from the subset of the original converter in Fig. 5,
featuring the coupled-inductor equivalent circuit model excited
by proportional voltage waveforms on both primary and sec-
ondary side, as illustrated in Fig. 6(c). Note that the equivalent
circuit model of the coupled-inductors is identical with the
familiar T-model of the usual ac transformer, featuring leakage
inductances L, and L, and magnetizing inductance L,, except
that in this model the secondary current direction is reversed,
leading to a magnetizing current i , which is the sum of the
primary and reflected secondary current. With the help of this
model several qualitatively different cases may be distinguished.
Here a brief and rather simdified explanation of onlv two
60 IEEE TRANSACTIONS ON MAGNETICS, VOL. MAG-19, NO. 2, MARCH 1983
cases of interest are given, while a complete more detailed and
rigorous account can be found in [21]. Shown in dotted lines
in Fig. 6 are current ripples before coupling and, in heavy lines,
after coupling. The two cases considered are balanced ripple
current reduction (Fig. 6(a)), and the limiting zero-ripple case
of the unbalanced current ripple reduction (Fig. 6(b)). There
are many alternative ways by which coupled-inductors can be
uniquely represented. One possible method is a physical de-
scription through primary and secondary leakage inductances
L, and L,, magnetizing inductance L,, and ideal N, : N, trans-
former as shown in Fig. 6. An alternative set of three param-
eters is the coupling coefficient k , secondary self-inductance
L, , and effective turns ratio n defined by n =dL7, .
A. Balanced Ripple Current Reduction
From Fig. 6(c), or using coupled-inductor equations, one ob-
tains for equal current ripples (and n =l),
dil di,
- US
dt dt (1 f k)L,
-=--
Hence both current ripples are reduced by a factor (1 t k) from
their original uncoupled value (k =0). In the limiting case
when k approaches one (tight coupling), both current ripples
are cut in half. However this may be considered as the least
beneficial improvement, since there is an even more interesting
case of unbalanced reduction of current ripples.
B. Unbalanced Ripple Current Reduction-Zero-Ripple Case
The relative change of the effective turns ratio n with respect
to coupling coefficient k results in the steering of the current
ripple from one side to the other. For the so-called matching
condition
n = k ( 2)
where
zero-current ripple is achieved in the output, as seen in Fig. 6(b).
To get a better physical grasp of how that is made possible,
the equivalent circuit model in Fig. 6(c) is quite helpful. As
seen from the accompanying waveforms, the input switched
voltage waveform us is attenuated by the inductive voltage
divider comprised of L, and L,. However if the actual turns
ratio N1 : N, of the ideal transformer is just right to step up
this attenuated voltage to its previous value (as illustrated on
the secondary of the ideal transformer), the net voltage across
the secondary leakage inductance L , is zero throu&out the
cycle, that is uL =0. Thus since L , is finite, it follows that
di,/dt =0, and the secondary current ripple is reduced to zero.
The key to the process is the recognition that the coupled
inductors are being driven by a (switched) voltage source us
at both windings, and zero ripple results from balancing the
drive-related voltages that exist at the two ends of a leakage
inductance. The detailed analysis reveals that the primary
current ripple remains at its level before the coupling.
A change of effective turns ratio n with respect to the cou-
\ I
N,:N,
(C)
Fig. 6 . Explanation of ripple current reduction in the coupled inductor
converter of Fig. 5 . Current ripple before coupling in dotted lines,
after coupling in heavy lines.
pling coefficient k has, therefore, the role of steering the
current ripple from one side to the other from its symmetrical
(balanced) reduction for n =1, and for the matching condition
(n =k) results in the complete removal of current ripple from
one side.
Zero-current ripple at the output has important practical
ramifications for applications where switching ripple needs to
be reduced as much as possible, such as the switching audio
amplifier (61 or computer power supplies, for example. From
the practical standpoint it is also invaluable that the zero-
current ripple has been demonstrated [21], to first order, to
be solely a property of the coupled-inductors design. For
example, interchange of the primary and secondary leads in
the coupled-inductor converter of Fig. 9, results again in the
zero-current ripple but now on the input side. In practice,
once we have designed the coupled-inductor to satisfy the
matching condition (2), we usually put dc current side designa-
tion on the appropriate leads! Even in the face of changes of
the operating point (duty ratio D and switching frequency f,
or line voltage V), the zero-ripple condition is maintained
since the basic proportionality of inductor voltage waveforms
still holds. Furthermore, it is not affected when the converter
moves into the discontinuous conduction mode since the volt-
age waveforms, although more complex in shape, are still pro-
portional [21].
The coupled-inductor method is not limited to the particular
switching converter topology described but may be applied to
other switching structures with similar benefits. For example,
when applied to the straightforward cascade connection of
boost and buck converters, zero-current ripple can be achieved
6UK: ZERO-RIPPLE SWITCHING DC-TO-DC CONVERTER
61
Fig. 7 . Coupled-inductordc isolated Cuk converter.
but only in the continuous conduction mode [21]. In the
discontinuous conduction mode the two inductor voltage wave-
forms lose proportionality [21] and the previously adjusted
zero ripple is suddenly lost. The coupled-inductor method
is thus a general technique and tool which can be applied to
switching structures to improve their performance. From the
several examples, a general law may be postulated:
I n any switching structure, a number of separate in-
ductors (two or more) can be integrated into a single
magnetic circuit with multiple windings, provided the in-
ductor voltage waveforms before coupling are in a fixed
ratio of proportionality to one another.
This broadened generality of the coupled-inductor concept
can be easily checked on several other examples. For example,
the isolation transformer was embedded into the basic non-
isolated converter with the least disturbance to its mode of
operation (see Fig. 3(c)). The proportionality of the input and
output inductor voltage waveforms is preserved and the two
inductors can again be coupled. Furthermore, if the isolation
transformer has an N1 : N2 turns ratio, the coupled-inductors
should be designed with the same N1 : N, turns ratio in order
to preserve 1 : 1 proportionality of the inductor voltage wave-
forms, such as shown in Fig. 7. Note that the coupling capaci-
tors C, and C, are shown relocated to the other side of the
respective primary and secondary windings, which may be
beneficial in some practical applications as pointed out in [ 121.
The same opportunity for coupling the inductors exists in
the transformer-isolated multiple-output version of the Cuk
converter: any or all of the inductors can be coupled, that is,
wound on the same core. The magnetic content in Fig. 8 is
therefore reduced only to two magnetic lumps: one pro-
viding the isolation, and the other supplying the inductor func-
tion at input and all the outputs. Again, by proper design of
this coupled-inductor multiple winding (four in Fig. 8) the
ripple currents can be steered to, or away from, the input port
or any of the output ports. One of course has the choice to
couple fewer than four windings on a single core. For example,
it was demonstrated in [ 121, [ 131 that input and one of the
outputs can be coupled to obtain zero-current ripple on the
9 t +
Fig. 8. Multiple output Cuk converter with only two magnetic lumps:
one for isolation transformer, and the other with input and all output
inductors on the same core.
input side, while the coupled-inductor structure obtained from
the remaining two outputs can be adjusted for zero-current
ripple in one of them. Hence a configuration is obtained with
zero-current ripple at input and selected output simultaneously.
I t now becomes obvious that the coupled-inductor technique
leads to significant ripple current improvement over its uncou-
pled counterpart, especially for zero-ripple design. However it
is at first less transparent how it can also lead to savings in size,
weight, and reduction of losses, requirements which for many
applications are at a premium.
C. Comparative Assessment of the Uncoupled and
Coupled-Inductor Magnetic Design
So far we have concluded that if the same magnetic core
material and copper materials were used to build a coupled-
inductor instead of two separate inductors, substantial ripple-
current reductions on both sides are possible. Let us now
reverse the problem and pose the following question. For the
same given input and output ripple-current performance, how
much is the coupled-inductor design more efficient and com-
pact compared to the two separate inductor implementation
of the switching converter?
I t is apparent at the outset that the most unfavorable com-
parison for the coupled-inductor would be the case of equal or
balanced ripple reduction (current ripple reduced in half),
which we will consider here as the least beneficial case. In the
other most beneficial extreme of zero ripple, one of the two
separate inductors would have to be infinite, hence clearly
impossible to achieve by two separate inductors.
To further simplify the comparison without impairing its
generality, let us assume that each converter in Fig. 9 operates
at the same operating point D =0.5 such that input and output
inductor dc currents are I. Also the inductors in the converter
of Fig. 9(a) are assumed equal to L , while the self-inductances
of the coupled-inductor in Fig. 9(b) are assumed to be L/ 2
(which corresponds to balanced ripple current reduction).
62 IEEE TRANSACTIONS ON MAGNETICS, VOL. MAG-19, NO. 2, MARCH 1983
SOURCE
DRIVE
Fig. 9. Comparison of separate inductor versus coupled-inductor design
for equal ripple current performance.
With these assumptions, both converters exhibit the same
ripple properties at both input and output ports.
Let us now compare the optimum magnetic design in the two
converters for the two cases: a) considering that they have the
same magnetic core material, hence, core losses will be the
same, and copper losses will be compared; b) considering that
they have the same winding material, hence, their copper losses
will be the same, and core losses different.
I ) Comparison of Copper Losses for Same Core Losses:
Suppose that the inductors for the converter in Fig. 9(a) are
made of toroids with cross section A,, mean magnetic path
length I,, and permeability p , and that in each toroid we can
fit 2N turns, as shown in Fig. lO(a). The material used can be
any standard one used for inductors such as powdered iron or
molypermalloy which have a distributed air gap and some
effective p to describe it. The effective p of the core is chosen
to allow dc bias current I without saturating the core. The
inductance L then becomes
L =
p4N2A,
1, .
Let us now stack the two cores together and build coupled-
inductors with N turns primary and N turns secondary, thus
again completely filling the usable winding area and making
optimal use of the core as before. The self-inductances of the
primary and secondary now become, with twice the core area
A, 3
pN2 2A, - L
2
L , =
1,
_ -
24,
(b)
Fig. 10. Comparison of two separate inductors versus coupled-inductor
design for the converter in Fig. 9 assuming equal core losses. (a) Two
separate inductors. (b) Coupled-inductor.
Because of the mutual inductance of the order of L/ 2, the
total effective input inductance is again L,ff =L and the same
current ripple as before is obtained. Note also that the same
maximum utilization of the dc capability of the cores is used,
since in each case from Ampere's law
H1, =2NI. ( 6)
Thus in both cases the same amount of core material is used.
However the savings in terms of copper losses and size and
weight with coupled-inductor design is quite obvious from a
comparison of Fig. lO(a) and 10(b). The separate inductors
result in almost doubling the copper losses, and occupying a
space well above the double volume of the coupled-inductors.
2) Comparison of Core Losses for Same Copper Losses: We
now consider the case in which the copper losses are main-
tained the same. Let us now take another geometrical con-
figuration, say U-I cores as shown in Fig. 11 and core material
with high p (hard material). Then an air gap of length Zg has to
be inserted to give the inductance L in the converter of Fig.
9(a) the dc current capability of I. WithN turns in the usable
window area, the two inductances become
L = ,
PoN2A,
Lg
in which the usual assumption [I 51 is that the reluctance in
the air is much larger than that in the core (Zg/po >>Zm/p),
where po is the permeability of free space in the MKSA system.
From the flux distribution in Fig. 1 l(a) it becomes obvious
that the coupled-inductor can now be built by use of only the
U-core parts without the need for I-core parts, as shown in
Fig. 1 l(b). Since the dc fluxes of the two windings now add,
the effective air gap is doubled to 2Ig, to prevent saturation of
the core. The self-inductance of the primary and secondary
of the coupled-inductor then becomes
Hence, as before equal current ripples are obtained for the two
CUK: ZERO-RIPPLE SWITCHING DC-TO-DC CONVERTER
63
Fig. 1 1 . Comparison of two separate inductors versus coupled-inductor
for the converter in Fig. 9 assuming equal copper losses. (a) Two sep-
arate inductors. (b) Coupled-inductor.
converters. Also, the same two windings are used in each case,
resulting in equal copper losses. However the core losses of
the coupled-inductor configuration are significantly reduced
since the I legs are completely eliminated. Depending on the
actual geometrical shape of the original configuration, this
may even approach saving close to half of the core material,
and hence correspondingly reduce the core losses and size.
In this particular case, it becomes even more transparent
how by integrating the two separate inductors into Q single
magnetic circuit, some of the common flux paths can be shared,
saving in the process the core material. For example, the
U-core of the second inductor served the purpose of closing
the magnetic path of the U-core of the first inductor, and vice
versa, with the net result the elimination of the I-core legs.
Thus far we have demonstrated how by use of the coupled-
inductor technique the savings can be made in two extreme
cases: either all the savings are in copper losses, or all the
savings are in core losses, with size and weight reduction in
each case. This choice was purposely made to clearly present
the two limiting cases. Of course, in practice, the coupled-
inductor may be designed to fall anywhere between these two
extremes thus saving both in copper and core losses as well as
size and weight.
Finally, one should bear in mind that this is the least benefi-
cial comparison, since clearly the savings will be significantly
better when the coupled-inductor is designed for unbalanced
ripple current reduction (zero ripple current in particular)
when its performance will be hard to match with separate
inductors.
However the practical feasibility of zero-current ripple on
only one side leaves one with a feeling of incompleteness, and
a desire to accomplish this ideal goal simultaneously at both
input and output sides.
IV. ZERO-RIPPLE SWITCHING DC-TO-DC CONVERTERS
With the just-posed problem of zero-current ripple at both
input and output, there are several possible approaches to
follow. For example, two coupled-inductor Cuk converters
canbe cascaded as shown in Fig. 12. The first, input, converter
could be designed for zero-input current ripple, while the
second converter is adjusted for zero-output current ripple.
However it is quite obvious that this would be a brute force
solution, since not only that the input power would be pro-
cessed twice through this cascaded connection, but the part
count would be doubled and the original advantages of sim-
plicity and reliability would therefore be diminished. Although
one may attempt to further simplify this configuration by
eliminating some component or rearranging the coupling, still
it seems that a rather different and fundamental look is needed.
Hence the following objective is posed.
Synthesize a switching dc-to-dc conversion structure which
has the least number of components (switches, capacitors,
inductors, and/or transformers) and yet leads to the ideal
dc voltage and current waveforms at both input and output
simultaneously.
This problem at first may seem formidable and impossible to
achieve, but with the material outlined so far, the generaliza-
tion of the coupled-inductor concept seems to be a natural
approach.
The first idea which immediately comes to mind in order to
achieve this goal is to use two cascaded transformers in the
coupled-inductor Cuk converter, as shown in Fig. 13. I t seems
that by adjustment of the first coupled-inductor for the match-
ing conditiop n, =l /kl , the input current ripple could be made
to vanish, while with another matching condition nz =kz , the
output current ripple could be reduced to zero at the same
time. Hence, this is a potential candidate for the ultimate zero-
ripple converter. The only problem, however, is that it does
not work as described. Namely, zero-current ripple could be
still obtained only on one side. There are insufficient degrees
of freedom for the independent adjustment of zero ripple on
the two separate coupled-inductors, since their cascade con-
nection can be equivalently represented by a T model of the
transformer.
It turns out that there is another more fundamental reason
for it. One may recall that the reduction to zero-current ripple
at either end in the basic coupled-inductor converter is because
both primary and secondary of the coupled-inductor are ex-
cited by some proportional stiff switching voltage waveforms.
Namely, the switching source for the secondary is generated
by the original energy transfer path through the capacitance
C1. Thus in order to have two independent matching condi-
tions for the two coupled-inductors in Fig. 13, one would have
to apply the same switching voltage source to the connection
points A and B between the two coupled-inductors, such as
shown in Fig. 14, and resulting in both current ripples reduced
to zero. Otherwise, without that excitation source for the
secondary of the first coupled-inductor transformer, and pri-
mary of the second cascaded coupled-inductor, the cascaded
connection behaves just like a single transformer, with possible
zero-current ripple on only one side.
Even though this attempt to cascade coupled-inductor trans-
formers is not successful yet, owing to the need for some means
64 IEEE TRANSACTIONS ON MAGNETICS, VOL. MAG-19, NO. 2, MARCH 1983
Fig. 12. Cascaded connection of two coupled-inductor Cuk converters.
t t
i--
I
Q T
-L
Fig. 13. Two cascaded coupled-inductor transformers in coupled-
inductor converter configuration still lead to a zero ripple at only one
end.
of imposing a stiff switching voltage across points A and B in
Fig. 14, the approach has not failed since it has pinpointed the
crux of the problem.
The remaining puzzling question is: How can we generate
the required switching voltage source and insert it appropri-
ately in the converter?
The attempt to generate this voltage source externally would
again be a brute-force solution. However some switching volt-
ages already existing in the converter itself could perhaps be
used. Unfortunately the converter of Fig. 14 has only two
such places: the primary of the first coupled-inductor and sec-
ondary of the second coupled-inductor. Application of either
one to points A and B would short out one or the other coupled-
inductor, thus effectively reducing it to a single transformer
configuration again.
The key step then is to find a switching topology which is
abundant with such proportional inductor switching voltage
waveforms and will hence allow multiple application of the
coupled-inductor concept. However we have already seen
such a topology in Fig. 3(b) where it was just an intermediate
and key step toward dc isolation introduction. To emphasize
Fig. 14. To achieve zero-current ripple at both ends, a stiff switching
voltage source is needed between the two coupled-inductors.
the availability of the proportional switching voltage wave-
forms, this configuration is shown again in Fig. 15 with the
corresponding inductor voltage waveforms.
Note that the voltage across the additional inner inductance
L is again exactly the same as that across the other two in-
ductors L , and L2 . Namely, it switches between the input dc
voltage Vg and the output dc voltage V, since the average (dc)
capacitance voltages on C, and Cb are Vg and V, respectively.
Therefore, the configuration in Fig. 15 is in a form suitable for
multiple application of the coupled-inductor concept.
It now becomes apparent that in the converter of Fig. 15,
the inner inductance L may be coupled either with the input
inductor L , for zero-input current ripple, or it may be coupled
with the output inductor L2 for zero-output current ripple.
However with a minor modification of the converter in Fig
15, both current ripples may be reduced to zero simultaneously.
In fact, the next step is quite obvious. The original inner loop
inductance L can be replaced by two separate inductancesl,
and Lb connected in parallel as shown in Fig. 16. Then by
coupling the inductance La to the input inductance L ,, while
separately coupling the inductance Lb to the output induc-
tance L2 , both input and output current ripples may be made
to vanish (actually pushed inside the converter into the La and
Lb windings). The two inductances La and Lb are not coupled
magnetically but are only connected electrically in parallel.
Hence this configuration has two magnetic lumps (two coupled-
inductors Ma and Mb) as shown in Fig. 16(b). In fact the two
coupled-inductor configurations Ma and Mb are connected in
cascade and the required switching voltage source UL is applied
to their interconnection points A and B, just as postulated in
Fig. 14. Thus the only step required in the configuration of
Fig. 14 is to split the coupling capacitor into two series capaci-
tors and to use their common point as the source of the addi-
tional switching voltage waveform.
As seen in Fig. 16(b), by varying the coupling coefficient k ,
in the first Ma coupled-inductor through the air-gap change,
the matching condition is obtained when n, =l /kl where
n1 d m , and consequently dc input current is obtained.
CUK: ZERO-RIPPLE SWITCHING DC-TO-DC CONVERTER 65
Fig. 15. Multiple of inductor switching voltage waveforms in this switching converter permits multiple use of the coupled-
inductor concept.
" I kl
Ma
-
' I
-
T = T 4T
M A P 2
2
cb)
Fig. 16. (a) Zero-ripple duk converter. (b) Implemented by use of the two cascaded coupled-inductors.
Likewise, by the corresponding matching condition (n2 =k2 ;
n2 2 d m ) in the other coupled-inductor Mb, the output
current ripple can be adjusted to zero as well. Note that the
two coupled-inductor adjustments are independent of each
other.
It now seems that we have obtained finally the ultimate solu-
tion postulated in Fig. 14 and posed at the beginning. How-
ever, although zero-current ripple is obtained at both ends, it is
achieved by some sacrifice in the complexity. Namely, we now
have again two magnetic lumps as in the original basic Cuk
converter (Fig. 2) and even four windings instead of the origi-
nal two. Thus, in line with the original approach adopted at
the beginning of the development of new better converter
topologies [l] , one would like to further simplify the topology
of Fig. 16 and yet retain its desirable zero-ripple properties.
The sequence of the development outlined so far clearly
indicates that the electrical content is maximally simplified
and best utilized in the converter of Fig. 15. The magnetic
content of the converter in Fig. 16, however, still has some
room for improvement and simplification. For example, the
two magnetic lumps in Fig. 16(b) could be further reduced
and integrated into a single magnetic circuit by magnetically
coupling the windings L, and Lb through merging of their
cores to result in the magnetic core structure of Fig. 17(b).
But now, two windings La and Lb being electrically connected
in parallel, can be reduced to a single winding L and the sim-
66 IEEE TRANSACTIONS ON MAGNETICS, VOL. MAG-19, NO. 2 , MARCH 1983
L t
t
" 2 3 2
\
/
P
(b)
Fig. 17. (a) Zero-ripple Cuk converter. (b) Implemented by a single magnetic circuit with three windings.
plified switching converter configuration of Fig. 17(a) is ob-
tained. Note, however, how even the single winding L in the
core structure of Fig. 17(b) effectively plays the role of the
previous inductances L, and Lb . Namely, the flux that this
winding L produces couples to the input winding L , through
left-hand part of the central core producing the effect of L, ,
La coupling in the previous configuration, while it also simul-
taneously couples to the output winding L2 through its com-
plementary flux in the right-hand part of the central leg,
emulating the previous Lb, L2 coupling. Notice also that in
this magnetic configuration (Fig. 17(b)) as in the previous one
(Fig. 16(b)), the input L , and output L2 inductors are not
directly coupled, thus preventing the potential interference of
this extra coupling with the zero-ripple adjustments.
Indeed, the flux generated by the winding L , is almost com-
pletely closed through the inner leg (containing L winding)
owing to its low reluctance when compared to the high reluc-
tance of the parallel magnetic path containing air gap 2. Thus,
the parallel bars in Fig. 17(a) show that the inner inductor L is
coupled to the input inductor L , and to the output inductor
L2 , which are virtually not coupled to each other. This con-
vention for schematically illustrating separate inductance
couplings is used throughout the remaining figures.
As before, by adjustment of the two air gaps, zero-current
ripple can be obtained simultaneously. However unlike the
situation for the core configuration of Fig. 16(b), in the core
configuration of Fig. 17(b) the two air gap adjustments are
interdependent: adjustment of only one air gap affects the cur-
rent ripple on both sides. Nevertheless, the adjustments are
highly convergent and lead in a few iterative steps to a zero-
ripple case, as demonstrated later in Section V in the experi-
mental verification. The other alternative is the use of fured
air gaps (designed for given dc flux capabilities) of input and
output inductors) and use of turns ratio adjustments of the
outer legs with respect to the inner leg. This method of course
does not have the interdependence problem, as also illustrated
in Section V.
The converter in Fig. 17 is a good example of the generaliza-
tion of the coupled-inductor concept whereby three previously
separate inductors are merged into a single magnetic circuit. In
addition, the performance approaches that of an ideal dc-to-dc
transformer since current ripple is reduced to zero at both
input and output sides (actually pushed into the inner winding
L). From the conceptual viewpoint the converter now consists
of only two major parts: an electrical part in which electric
charges flow, and a magnetic part in which the flux flows,
which are appropriately connected to result in the useful func-
tion: ideal dc-to-dc conversion.
While approaching the properties of an ideal dc-to-dc trans-
former, the converter of Fig. 17 and all other coupled-inductor
extensions are lacking the practically very important dc isola-
tion property. Our approach again should be the same as that
for all the previous developments: how to build-in this property
with the minimum of extra elements. The isolation prop-
erty is easily recovered by a step which is completely analogous
to that shown leading from Fig. 3(b) to Fig. 3(c). Namely, by
addition of a single winding to the inner leg of the magnetic
circuit of Fig. 17(b), and then electrically separating the two
CUK: ZEROlRIPPLE SWITCHING DC-TO-DC CONVERTER 67
-
t t
MI
R
(b)
Fig. 18. (a) Zero-ripple dc isolated Cuk converter. (b) Its magnetic circuit relization.
windings on the inner leg, the isolation is obtained as shown in
Fig. 18(a), with no additional magnetic material required. Of
course, zero-current ripple properties on both ends remain.
The added isolation transformer now has a double role: it
not only provides the isolation, but also through coupling with
input and output inductances provides a means for achieving
dc currents at both ends simultaneously, and at the same time
leads to better magnetic core utilization.
From the magnetic circuit realization in Fig. 18(b), it may
seem that a rather special magnetic core configuration is needed
which even does not look like a.single-core structure. In fact,
the core configuration used consists of two sets of U-cores, as
compared to the needed three sets of U-cores if the two induc-
tors and isolation transformer were implemented with separate
magnetic circuits. However, this was done only for concep-
tual reasons to clearly explain each step of the development
and accompanying savings. There is a large number of core
shapes and configurations with two air gaps which could be
implemented. For example, a standard E1 magnetic core can
be used for the practical implementation of the magnetic cir-
cuit in Fig. 18(b), such as shown in Fig. 19. I t differs from
the standard usage only in that its outer legs contain air gaps,
as opposed to the usual implementation in which the inner
(central) leg is gapped and not the outer ones. The fact that it
consists also of three bobbins mounted on each leg of the E1
core structure is not unusual and is regularly used in three-
phase transformer applications.
A truly optimum dc converter is obtained which has both
outstanding features: true dc currents and dc isolation in the
simplest possible topology consisting of a single magnetic cir-
cuit with four windings, transistor, diode, and two capacitors.
Fig. 19. Alternative E1 core realization of the single magnetic circuit
for the zero-ripple converter of Fig. 18(a). Inner leg windings provide
dc isolation, the outer leg windings input and output inductance.
This is illustrated in Fig. 20, in which the converter hardware
realization (Fig. 20(a)) is equivalently represented by the ideal
dc-to-dc transformer (Fig. 20(b)), hence the initial problem
posed in Fig. 1 has been solved.
From the conceptual viewpoint, this is the first time that
such diversified magnetic components, such as inductors and
transformers exclusively used until now as separate components
in switching structures, are integrated into a single magnetic
circuit with multiple windings.
A natural extension of this zero-ripple converter configura-
tion (Fig. 18) to multiple outputs is shown in Fig. 2 1. There,
a single magnetic circuit M, with six windings appropriately
coupled is capable of providing zero-current ripple not only at
the input but simultaneously at both outputs as well. The
actual physical implementation of its magnetic circuit shown
in Fig. 21(b) is a simple three-dimensional extension of the
previous E1 core configuration. It required addition of only
one U-core, which shares a common leg with the central leg
of the E-core as seen in Fig. 21(b). Then, the three windings
are placed on this common central leg, while the input and
68 IEEE TRANSACTIONS ON MAGNETICS, VOL. MAG-19, NO. 2, MARCH 1983
l 2 f
5 1 I 1: t; 61 I : M( D) L-
(a) (b)
Fig. 20. Simplest realization of the ideal dc-to-dc transformer consists of the four winding magnetic circuit, transistor,
diode, and two capacitors. (a) Converter. (b) Ideal dc-to-dc transformer.
. t
t
Fig. 21. (a) Two-output zero-ripple dc isolated Cuk converter. (b) Its single magnetic circuit realization.
two output windings are placed on the three outer legs. Again,
as before the outer legs are gapped while the central leg is not.
I t is interesting to point out that a magnetic core configuration
of the same construction has been recently proposed and used
[22] , although for a different application in which the addi-
tional U-core magnetic leg serves for control winding purposes.
The extension of this zero-ripple configuration to other mul-
tiple outputs (three or more) is then quite obvious.
There are a number of extensions and variations that can be
used in multiple-output configurations, if zero-ripple is not
required at both input and all outputs. For example in the
two-output dc isolated converter of Fig. 22, two output induc-
tors can be coupled to the input inductor, but not to each
other, by use of the core configuration of Fig. 19 with the
input inductor on the inner center leg. Then, zero-current
ripple can be obtained simultaneously at both outputs, which
CUK: ZERO-RIPPLE SWITCHING DC-TO-DC CONVERTER
l 2 r-
1
i3 I -
Fig. 22. Switching dc-to-dc converters with zero-ripple currents at both
outputs implemented by E1 core of Fig. 19.
is usually of overriding practical importance. Also, the core
utilization will be best since all three windings will optimally
use air-gapped core structure for their maximum required dc
current capability. Furthermore, the remaining magnetic lump
for the isolation transformer is easy to design, since it is inde-
pendent and not constrained by additional coupling. A number
of interesting extensions and variations of the multiple-output
converter with coupled-inductors (two or three magnetic lumps)
together with detailed analysis and practical considerations
can be found in [12] -[14].
Here only a limited number of switching configurations with
zero-ripple current on both sides could have been presented, so
only a few which appear to be basic for the understanding of
the concept are shown. A number of other very interesting
zero-ripple extensions have been proposed and can be found
in [l l ].
V. EXPERIMENTAL VERIFICATION OF ZERO RIPPLE
The main purpose of the experiments was to verify the feas-
ibility of the simultaneous reduction of current ripple to zero
by use of the several switching converter topologies proposed
and their accompanying magnetic circuits. The objective was
not to go into detailed analysis of-several complex magnetic
structures suggested, nor to experimentally verify their accom-
panying design equations, since that will be the subject of
some detailed future papers. Thus the experiments were de-
signed to prove the concept and not to indicate or suggest any
preferred practical implementation.
There are two main approaches [21] by which one can de-
sign the coupled-inductor experiment to demonstrate the cur-
rent ripple reduction with changes of the relative ratio of the
coupling coefficient k and effective turns ratio n, and to prove
that there is, indeed, the practical feasibility of the zero-ripple
condition on both ends: a) by varying the'air gap and hence
coupling coefficient k, while keeping the effective turns ratio
n constant, a continuous approach to and through zero-current
ripple is possible; b) by varying the turns ratio while keeping
69
the air gap fured, an approach to and through zero-current
ripple is made possible but now in discrete steps, whose size
depends on the incremental turns ratio change.
For verification of zero current ripple along both of the two
routes, the topological configuration of Fig. 18(a) was built
such that the interconnection points for all magnetic compo-
nents were left open so the various magnetic structures pro-
posed in the previous section could be easily attached.
A. Variation of the Air Gaps
This first laboratory experiment is actually a generalization
of the experiment originally performed in [21] to prove the
actual feasibility of zero-current ripple in the basic coupled-
inductor Cuk converter by actually going through the zero
current ripple continuously.
The magnetic circuit structure of Fig. 18(b) is made by use
of two sets of U-cores and four windings wound appropriately.
The two air gaps are then made variable by use of the two
micropositioners as shown in the set-up of Fig. 23(b). The
windings are then connected to the converter according to the
dot designation of Fig. 18(a), and the following is observed.
The variation of the air gap 1 only (left-hand air gap) results
in the continuous change of the current ripple from its usual
positive slope (triangular waveform), through zero (straight
line) to the negative slope (inverted triangular waveforms) first
time observed in the coupled-inductor structures as seen in
Fig. 23(a). As expected this variation of the air gap affected
the other output current ripple as well, since the flux distribu-
tion of the inner windings is also affected by this variation.
The air gap 1 is then adjusted for zero-input current ripple.
The subsequent variation of the air gap 2 only, leads also to
the continuous change of the output current ripple from posi-
tive through zero, to "negative" current ripple. Air gap 2 is
then adjusted for zero-ripple output current. At that position
the previously adjusted input current is slightly offset from
zero-ripple condition. However, another iteration of the air
gaps 1 and 2 adjustments leads to a perfect zero-ripple condi-
tion simultaneously on both sides.
One feature of the magnetic structure of Fig. 23 may appear
at first to be troublesome. The isolation transformer has built
in air gaps in its magnetic path and potentially increase its
leakage inductance. However it was convincingly demonstrated
in [ 101 on the coupled-inductor experimental setup analogous
to that of Fig. 23, that to the first order the leakage induc-
tances remain constant independent of the gap length. Thus
isolation transformer performance is not adversely affected
in this single magnetic circuit configuration.
This laboratory experiment convincingly demonstrates the
feasibility of a zero-current ripple converter, provided the single
magnetic circuit structure is designed appropriately. While it
very nicely demonstrates this new generalized coupled-inductor
concept, the variable gap magnetic structure very likely is not
going to be the preferred realization, even though possibilities
may exist, such as ferrite pot cores with screw adjustable air
gaps usually used for inductor turning purposes. The configu-
ration with fured air gap structure, such as the one of Fig. 19,
from practical considerations is easier to implement as shown
next.
IEEE TRANSACTIONS ON MAGNETICS, VOL. MAG-19, NO. 2, MARCH 1983
Y
(b)
Fig. 23. Variation of air gaps in the magnetic circuit (a) for the converter of Fig. 18 result in the continuous change of the
two ripple currents from positive, through zero to negative slope (b).
B. Variation of the Turns Ratio
Since the E1 core configuration of Fig. 19 was not readily
available (it comes either ungapped or with central leg gapped)
an ungapped ferrite EE core was chosen and the outer legs
of one E-core were ground to provide two air gaps (courtesy of
J et Propulsion Magnetic Laboratory and Colonel McLyman).
Note, however, that if needed in production quantities, the E1
gapped core configuration of Fig. 19 could be just as readily
obtained as the customary central gap structure, especially
with ferrite cores. Another practical feature of the core con-
figuration in Fig. 19 is that all the windings can be wound on
bobbins and slipped on core legs, which was also used for this
experiment. The two windings for the isolation transformer
are wound bifiliar on the central leg bobbin. Then, the num-
ber of windings on the outer legs were increased above that of
inner windings until zero-current ripple is achieved on both
sides. From the practical viewpoint, even with the relatively
imprecise adjustments, current ripple can be easily reduced for
two orders of magnitude (actually shifted from the input and
output side to the inner isolation transformer windings).
While the experiments qualitatively confirmed the expecta-
tion, one interesting difference with the coupled-inductor zero-
ripple result of Fig. 5 was noted. There, all the current ripple
is shifted to the primary side, which retained the same triangu,.
lar waveform (the same input current ripple as before the
coupling). In other words, primary and secondary current
flow in their respective input and output circuits, and are only
related through the transformer (or better coupled-inductor)
equivalent circuit model. In the converter of Figs. 18 and 20,
they are in addition forced to be identical during the portion
of the switching cycle. For example, during the interval when
the transistor is off, the input inductor current is the same as
the primary current of the isolation transformer. Hence, in
the zero-input current ripple case, the primary current of the
isolation transformer has a flat part during this interval. How-
ever, during the complementary interval when the transistor is
turned on, primary current does retain its familiar positive
slope which indicates that the ripple current is being pushed
inside the converter into the winding. Of course, if it were flat
that would indicate infinite magnetizing inductance of the iso-
lation transformer a physical impossibility.
A number of variations for the practical implementation of
the coupled-inductor configurations is contained in [ 191 and
can be easily updated for their implementation in zero-ripple
configurations.
VI. INTEGRATED MAGNETICS CONCEPT
Apart from the usefulness and practical advantages of a num-
ber of zero-ripple switching configurations outlined so far,
their significance goes even beyond, since they help to bring
about a new and general concept which can be applied with
success to many switching structures.
The coupled-inductor extension of the Cuk converter [21]
was the crucial first step in that direction. There, from the con-
ceptual viewpoint, the two magnetic components (inductors),
which are normally and exclusively used separately to perform
their function in switching circuits, were for the first time
integrated into a single magnetic circuit (single core) with two
windings. While it looked like a classical ac transformer from
the constructional viewpoint, it indeed performed the function
of two separate inductors, when polarity marks and actual cur-
rent directions are taken into account. Furthermore, it even
outperformed the separate inductors by significantly reducing
the size, weight, and losses as demonstrated in Section 111.
The natural outgrowth and in fact the generalization of this
concept is achieved, when in Section IV it is extended to in-
clude ac transformers as well. Thus such diversified magnetic
components as inductors and transformers, used exclusively
until now as separate magnetic components, are now integrated
into single magnetic circuit with multiple windings. Besides
CUK: ZERO-RIPPLE SWITCHING DC-TO-DC CONVERTER 71
m n
POLYSILCON
"out
Fig. 24. Analogy between the integration of the semiconductor components and the integration of magnetic components.
(a) Semiconductor. (b) Magnetic.
apparent simplification, the converter performance is signifi-
cantly. improved (ripple currents reduced to zero) with addi-
tional savings in size and weight and increased efficiency. For
this process of merging the separate magnetic components into
an integral magnetic circuit with accompanying savings in size
and weight owing to the elimination of some magnetic f l ux
paths a generic name integrated magnetics is proposed.
Even some analogy can now be drawn with the process which
led to the transition from discrete to integrated semiconductor
circuits. Integrated semiconductor technology appeared in the
late fifties when the need arose for the new technology which
could accommodate the increasing complexity of semiconduc-
tor circuits by eliminating the interconnection of discrete parts.
For example, in the integrated version of an N-type metal-
oxide semiconductor (NMOS) inverter gate (shown in Fig. 24(a)
implemented in NMOS polysilicon self-aligned gate technology
with the enhancement load) the external electrical connection
between the upper and lower transistor required in discrete
version is now eliminated, since it is automatically provided in
the semiconductor media itself. In the case of the magnetic
circuits, the transition from discrete magnetic components to
integrated magnetic circuit, leads to the elimination of mag-
netic interconnection paths in analogy with previous electric
interconnection paths. For example, in transistion from two
separate inductors to a coupled-inductor configuration (Fig.
1 l(a) and (b)) two I magnetic cores are eliminated. For an-
other example of Fig. 24(b), the transition from two separate
inductors and a transformer into a single integrated magnetic
circuit results in similar elimination of magnetic core material,
this time two E-cores.
In the integrated semiconductors, the semiconductor is a
common media into which different functions are engraved,
such as transistors, diodes, resistors and even capacitors with
the electric charge as a binding carrier. In the integrated mag-
netics, the magnetic core is a common media around which
functions of inductors and transformers are built with the flux
as a binding carrier. In both cases, however, proper voltage
and current excitation of their terminal contacts is necessary
for the desired component functions to be related.
One of the criticisms which could be raised against the single
magnetic circuit is its inflexibility that is, it is dedicated to the
particular switching converter topology, and could have only
a limited use in other configurations. In other words it is not a
general purpose component such as discrete transistors and
individual inductors and transformers. In the integrated
semiconductor circuits that flexibility is also lost with all com-
ponents interconnected performing only an overall dedicated
function. However, except at a laboratory level, one very rarely
tears apart even discrete semiconductor circuits to recover the
parts and build a new discrete circuit. In the switching power
converters, even the separate magnetic components are so
much tied to the particular design requirements, such as volt-
age, current and power rating or operating frequency and size
requirement, that it is virtually impossible to reuse it in any
72 IEEE TRANSACTIONS ON MAGNETICS, VOL. MAG-19, NO. 2 , MARCH 1983
M I
M.2
Fig. 27. Integrated magnetics applied to Cuk switching power amplifier.
other design. Therefore, in neither case the added inflexibility
appears to be the limitation.
A. Applications 0.f the Integrated Magnetics Concept
This new concept of integrated magnetics is applicable to a
broad spectrum of electrical circuits involving magnetic com-
ponents, It leads to the similar overall performance improve-
ments in other switching converters, coupled with the simpli-
fied configuration, and reduced component count, size, and
weight. To substantiate this general claim, this concept is now
illustrated on several other switching configurations with simi-
lar benefits. For example, it can be applied to the SEPIC con-
verter [23] shown in Fig. 25(a). When the input inductor La
and the isolation transformer T4 are integrated into a single
magnetic circuit M,, , a converter with zero-input current ripple
is obtained as shown in Fig. 25(b). Still another example is a
converter of Fig. 26(a), which is a dual counterpart [7] to the
previous converter. Merging the output inductor L, and isola-
tion transformer T, into a single magnetic circuit M,, a conver-
ter of Fig. 26(b) with zero-output current ripple is obtained.
However the application of integrated magnetic concept is
not limited to the switching dc-to-dc converters only. Other
switching configurations performing qualitatively different
functions, such as dc-to-ac inversion, or power amplification,
may also utilize this concept to advantage. For example, this
concept may be directly applied to a new push-pull switching
power amplifier structure [6] , [ 171 shown in Fig. 27 imple-
mented by use of two dc isolated Cuk converters. Each con-
verter's isolation transformer may be integrated with its corre-
sponding input and output inductors to result in only two
magnetic lumps and zero-current ripple on both sides.
All these examples now illustrate how in a complex switch-
ing configuration consisting of a number of storage elements
(inductors, transformers, and capacitors) and switches inter-
connected in such a way to perform some useful function
(such as dc-to-dc conversion, dc-to-ac inversion, or power am-
plification), the otherwise separate magnetic components can
be merged into an integral magnetic circuit with multiple
windings. The prerequisite for such simplification of the arbi-
trary switching configuration shown in Fig. 28(a) is the exis-
tence of synchronized and proportional voltage waveforms on
the inductors and transformers as shown in Fig. 28(b). The
voltage waveforms may even take a different shape other than
rectangular, as long as their proportionality is retained [21].
6UK: ZERO-RIPPLE SWITCHING DC-TO-DC CONVERTER 13
(b)
Fig. 28. Generalization of integrated magnetic concept to any switching
structure (a) and prerequisite for its implementation (b).
Thus, Fig. 28 indicates the generality of the integrated mag-
netics concept.
VII. EVOLUTION OF ZERO-RIPPLE SWITCHING
CONFIGURATIONS
The search for more efficient, low noise, new switching con-
verter topologies was originally motivated by the desire to
reduce or possibly eliminate some of the major problems which
limit "switchers," such as the large pulse currents at either the
input or the output port or even at both. For example, in a
conventional buck-boost (flyback) converter both input and
output-port currents are large pulses leading to severe con-
ducted and radiated electromagnetic interference (EMI). The
basic Cuk converter (nonisolated version) was a substantial
step in the direction of reducing this problem by creating nan-
pulsating input and output currents. However the unique
topology of the Cuk switching converter and in particular,
proportional voltage waveforms on their two inductors, has led
to a powerful new concept-the coupling of inductors. Besides
the obvious reduction in complexity, yet another significant
advancement has been accomplished: current ripple at either
input or output is not only reduced but completely eliminated.
The natural outgrowth of this new technique of coupling the
inductors is a switching configuration which exhibits a single
magnetic circuit with three windings and achieves zero-current
ripple at both ports simultaneously. The major cornerstones
in the development of the nonisolated converter configurations
are summarized in Fig. 29(a).
For many practical applications dc isolation between source
and load is often required. The next key step in the develop-
ment is the incorporation of an isolation transformer in an
optimum single-ended manner, which leaves the fundamental
features of the basic kuk converter intact. Hence coupling of
the input and output inductors leads again to zero-current
ripple at either end. Finally the crowning achievement of the
converter development is its final evolution into a topology
which truly emulates the ideally desired electrically control-
rTKT -
cb)
Fig. 29. Evolution of zero-current ripple topologies through (a) non-
isolated and (b) isolated versions.
lable dc-to-dc transformer. This ultimate switching dc-to-dc
converter configuration possesses both outstanding features,
true dc currents at both ports and dc isolation in the simplest
possible'' topology consisting of a single magnetic circuit with
74 IEEE TRANSACTIONS ON MAGNETICS, VOL. MAG-19, NO. 2, MARCH 1983
four windings, two capacitors, and a single switch implemented
by usual transistor diode combination. The major cornerstones
in the development of the isolated converter configurations
are summarized in Fig. 29(b).
In summary, a long journey in development of optimum
switching converter topologies has been successfully completed:
from the conventional buck-boost with large pulsating currents
at both input and output ports, through a new converter with
both currents nonpulsating to a true dc-to-dc switching con-
verter along both routes featuring nonisolated as well as iso-
lated versions.
VIII. CONCLUSION
A long and very exciting journey into the world of switching
dc-to-dc converter topologies initiated several years ago [2]
has come to its fruition in establishing a switching hardware
realization which truly emulates the ideal dc-to-dc transformer
function. After such an exhaustive trip [ 1 ] - [ 141 , [ 161 - [ 191 ,
let us pause for the moment and reminiscence about the major
milestones encountered on the road.
The traditional approach taken by many researchers and
practicing engineers was indeed very simple: let us get the re-
quired dc-to-dc conversion function first by using some simple
switching mechanism (such as buck or buck-boost) and then
we will cure its problems, such as pulsating input and output
currents later. This has even been generalized recently into a
theory 1201 , which views switching converters as being classi-
fied and related by function, not by circuit topology. I t
attaches the full emphasis on the switching mechanism alone,
while the storage components (inductors, capacitors, trans-
formers) are considered only after the fact to clean up input
and output current or voltage waveforms through additional
filtering.
Our approach, however, was just the opposite: from the very
beginning we considered that storage and energy transfer ele-
ments (inductors, capacitors, transformers), are together with
the switches an integral part of the overall problem. Further-
more, we have not only attached extreme importance to con-
verter topologies but with this paper we have extended the
previous topology of electrical connections into its comple-
mentary topology of magnetic connections through the newly
introduced integrated magnetics approach.
A few words now about the switching conversion complex-
ities. At any point in the development, we have introduced
the additional complexity by adding elements or new electrical
or magnetic connections for only two reasons: a) because it
was the simplest way to get the additional feature or function,
for example, dc isolation introduction into the basic noniso-
lated converter or its extension to a four-quadrant bidirectional
switching power amplifier; b) as an intermediate step only
toward potential further simplification (such as the reduction
of the cascaded boost-buck converter into the Cuk converter).
Even when seemingly a very simple configuration was achieved
in the basic converter, desire to still further simplify and im-
prove its performance led to the new coupled-inductor concept,
which is here generalized into a useful integrated magnetics
approach. They now become useful tools for converter im-
provement, independent of technology. Although component
manufacturers will always come up with better and improved
components such as transistors, magnetic cores, capacitors,
etc., their utilization in different converter topologies may
significantly differ. For example, the coupled-inductorimple-
mentation will result in substantial savings in size, weight and
losses when compared with the uncoupled case, regardless of
the state-of-the art of magnetic core technology.
Along the development we have strived for the fundamental
conceptual solutions. This has been very beneficial in two
respects: it not only resulted in the host of new very useful
and practical switching configurations but also gave birth to
some fundamental techniques and methods such as the coupled-
inductor concept, the integrated magnetics approach as well as
duality theory for switching structures. All these can now be
used as tools to improve performance of many other switching
configurations.
Finally, dc-to-dc conversion topologies are extended along
the same objective to include bidirectional power flow (two
quadrant converters) [5] and dc-to-ac inversion and power
amplification (four-quadrant converters) [6] . Thus the ap-
proach has been generalized to include the whole spectrum of
power electronics applications.
REFERENCES
[I]. S. &k, Modelling, analysis, and design of switching converters,
Ph.D. dissertation, California Institute of Technology, Nov. 1976.
(Al>o, NASA Report CR-135174.)
21 S.Cuk and R. D.Middlebrook, A new optimum topology switch-
ing dc-to-dc converter, in ZEEE Power Electronics Specialists
Conf., 1977 Record, pp. 160-179 (IEEE Publication 77CH 1213-8
AES).
31 -, Coupled-inductor and other extensions of a new optimum
topology switching dc-to-dc converter, in ZEEE Znd. Appl. SOC.
Annu. Meeting, 1977 Record, pp. 1110-1126 (IEEE Publication
41 R. D. Middlebrook and S. Cuk, Isolation and multiple output
verter, in IEEE Power Electronics Specialists Conf., 1978 Rec-
extensions of a new optimum topology switchin:: dc-to-dc con-
ord, pp. 256-264 (IEEE Publication 78CH1337-5 AES).
51 R. D. Middlebrook, S. Cuk, and W. Behen, A new battery charger/
discharger converter, in IEEE Power Electronics Specialists
Conf., 1978 Record, pp. 251-255 (IEEE Publication 78CH1337-5
AES).
61 S. Cuk and R. W. Erickson, A conceptually new high-frequency
switched-mode amplifier technique eliminates current ripple, in
Proc. Fifth National Solid-State Power Conversion Con5 (Power-
con,5),pp. G3.143.22, May 1978.
71 S. Cuk, General topological properties of switching structures,
in IEEE Power Electronics Specialists Conf., 1979 Record, pp.
-. .
77CH1246-8-IA).
109-130.
[8] L. Rensink, A. Brown, S. P. Hsu, and S. duk, Design of a kilo-
watt off-line switcher using a Cuk converter, presented at Proc.
Sixth National Solid-state Power Conversion Conf. (Power-
con,6).
[9] S. Cuk and R. D. Middlebrook, Advances in switched-mode
power conversion, survey paper, Robotics Age, vol. 1, no. 2,
pp. 6-19, Winter 1979.
[l o] S. P. Hsu, Problems in analysis and design of switching regula-
tors, Ph.D. dissertation, California Institute of Technology,
Sept. 1979.
[ 111 L. Rensink, Switching regulator configurations and circuit reali-
zations, Ph.D. dissertation, California Institute of Technology,
Dec. 1979.
[12] G. E. Bloom and A. Eris, Practical design considerations of a
multi-output Cuk converter, in ZEEE Power Electronics Spe-
cialists Conf., 1979 Record, pp. 133-146 (IEEE Publication
79CH1461-3 AES).
IEEE TRANSACTIONS ON MAGNETICS, VOL. MAG-19, NO. 2, MARCH 1983 15
[13] G. E. Bloom, A. Eris, and R. Ruble, Modelling, analysis; and duced ripple without need for adjustments, U.S. Patent Applica-
design of a multi-output Cuk converter, in Proc. Seventh Na- cation, J une 15,1979.
tional Solid-state Power Conversion Conf., pp. 11.1-11.19, Mar. [20] P. Wood! General theory of switching power converters, in
1980. IEEE Power Electronics Specialists Conf., 1979 Record, pp. 3-10
output Cuk converter, presented at IEEE Power Electronics [21] S. Cuk, Switching dc-to-dc converter with zero input or output
Specialists Conf., J une 1980, Atlanta, GA. current ripple, in IEEE Ind. Appl. Soc. Annu. Meeting, 1978
[ 151 W. T. McLyman, Transformer and Inductor Design Handbook. Record, pp. 1131-1145, (IEEE Publication 78CH1346-61A).
New York: Marcel Dehker. [ 221 H. Hirayama, Simplifying switched mode converter design with
[16] S. Cuk and R. D. Middlebrook, Dc-to-dc switching converter, a new variable leakage transformer topology, in Proc. Seventh
US. Patent 4,184,197, J anuary 15, 1980. Foreign patents National Solid-state Power Conversion Conf: (Powercon 7), pp.
pending. El.1-El.10, Mar. 1980.
[17] S. Cuk, Push-pull switching power amplifier, U.S. Patent [23] R. P. Massey and E. C. Snyder, High-voltage singleended dc-dc
4,186,437, J an. 29,1980. Foreign patents pending. converter, in IEEE Power Electronics Specialists Conf., 1977
[18] -, Dc-to-dc switching converter with zero input and output
Record, pp. 156-159.
current ripple and integrated magnetics circuits, U.S. Patent Ap- [ 241 R. D. Middlebrook, Modelling and Design of the Cuk Converter,
plication, March 30,1979.
in Proc. Sixth National Solid-state Power Conversion Coni
[14] G. E. Blpom and A. Eris, Modelling and analysis of a multi- (IEEE Publication 79CH1461-3 AES).
[19] S. Cuk and R. D. Middlebrook, Dc-to-dc converter having re- (Powercon 6), pp. G3.143.14, May 1979.
New Magnetic Structures for Switching Converters
have been centered around a very few standard switching converter designs have been so notorious in noise generation that the
topologies. Recently, a number of new converter topologies have been
proposed in order to find the best possible electrical interconnection of
whole field of switching power supplies received a bad name
.~
power processing ,elements: switches, storage components, and trans-
formers, that would yield the highest efficiency and best performance. Early attempts to cure some of these noise performance
However the equally important and complementary problem of their
best magnetic interconnection has been completely overlooked. In
some new converter structures, the nature of the switching process and
existing waveforms allows integration of previously separate inductors
and transformers into a single magnetic structure. Several such mag-
netic core structures are proposed and analyzed, which lead to further
converter simplifications and performance improvements.
U
I . I NTRODUCTI ON
P UNTIL two decades ago, only a small number of
standard switching configurations have been in wide-
spread use, such as the basic buck, boost, buck-boost, flyback,
buck-forward, push-pull, half-bridge, and full-bridge switch-
ing converters. However each of these configurations had both
theoretical and practical disadvantages, often requiring addi-
tion of some extra components (such as input filtering in the
buck or buck-boost derived family of converters, flux sensing
to prevent saturation in push-pull converters, etc.) to cure
problems by adding brute-force filtering and heavy shielding
resulted in designs almost as heavy and bulky as their linear
power supplies predecessors. Fortunately, space applications
and their requirements for small size, weight, and highest ef-
ficiency maintained a strong interest and sponsored innova-
tions in switching power supplies. J ust within the last decade
a number of new switching configurations mushroomed, pri-
marily.because of the need to come up with more efficient and
quiet converter topologies.
The prior effort in devising new, better converter topologies
has been documented in several Ph.D. dissertations [l ] , [2],
in a two-volume book Advances in Switched-Mode Power Con-
version [ 3 ] , [4], and,in a series of patents [SI -[8]. Through-
out this research effort the emphasis was on the topology of
the electrical interconnections among various components
(switches, capacitors, inductors, and transformers), as illus-
trated on the general block diagram of Fig. l(a). Several new
converter configurations [ 11, [4] have been proposed and
thoroughly analyzed [ 1 ] , [ 3 ] .
Manuscript received August 13, 1982; revised November 23, 1982.
Recognition of some of the unique features of the charac-
This paper was presented at the Power Conversion International Con- teristic switching waveforms in these converters led to the
001 8-9464/83/0300-0075$00.75 0 1983 IEEE