IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS—I: FUNDAMENTAL THEORY AND APPLICATIONS, VOL. 44, NO.

12, DECEMBER 1997 1129

Study of Bifurcations in Current-Programmed

DC/DC Boost Converters: From Quasi-
Periodicity to Period-Doubling
William C. Y. Chan and Chi K. Tse, Senior Member, IEEE

Abstract— This paper studies the bifurcation paths exhibited identified for such circuits, namely, one that goes through a
by a simple second-order dc/dc boost converter under current- region of quasi-periodicity [6]–[8]), and one that goes through
programmed control with and without voltage feedback. Previous period-doubling [9], [10]. The work of [6] is based on a fourth-
work in this area has reported two distinct types of bifurcation
paths, namely via regions of quasi-periodic orbits and period- order Ćuk converter, whereas those of [8], [9], and [10] are
doubling. This paper demonstrates that the two different types based on a second-order boost converter.
of bifurcation paths can, in fact, be viewed as part of another Our objective in this paper is to put together the previ-
bifurcation in which the quasi-periodic sequence transmutes into ously observed bifurcation paths, and view them as part of
the period-doubling sequence, and that such a bifurcation is another bifurcation arising from variation of a suitably chosen
observed regardless of the presence of the outer voltage feedback
loop as long as a suitable set of bifurcation parameters is chosen. parameter. In other words, we will study the bifurcations
The describing iterative map is derived in closed form and is used at two abstract levels. At each level, a different bifurcation
to develop the main results via a series of computer experiments. parameter is used. We refer to the bifurcation parameter used
The characteristic multipliers are calculated and the first onset in the previous studies as the primary bifurcation parameter,
of flip-bifurcation is predicted. Computer simulation based on based on which we observe either quasi-periodicity or period-
an exact piecewise switched model confirms the predicted bi-
furcations. The exhibition of quasi-periodic orbits is confirmed doubling. We will define another bifurcation parameter, which
by computation of the Lyapunov exponent. Finally, a series of we refer to as secondary bifurcation parameter, and based
return maps are generated to provide an alternative viewpoint on which we observe a “super”-bifurcation that leads quasi-
to the reported bifurcations in terms of a transmutation from a periodicity to period-doubling.
tent-like map to a logistic-like map. We will present our results as follows. In Section II, the
Index Terms— Bifurcation chaos, boost converter, current- operation of the boost converter under current programmed
programmed control. control is explained and the exact state equation describing the
system is derived. As a prelude to the main discussion, some
I. INTRODUCTION computer-generated waveforms will be shown in Section III
to illustrate the various possible operating regimes of current-

C URRENT-PROGRAMMED control, being one of the

most commonly used control schemes in dc/dc con-
verters, has received much attention from power electronics
programmed dc/dc converters. In Section IV, an iterative map
is derived in closed form which will be used in Section V to
develop the main results regarding the bifurcations of the open-
engineers in the past two decades [1]–[5]. As is the rule rather loop system. The characteristic multipliers are computed in
than the exception for any new control application, much of Section V-B, from which the onset of the first flip bifurcation
the study reported so far has focused on the design of the is predicted. Using the same iterative map, a series of bifur-
involving feedback loops and performance evaluation on the cation diagrams are generated in Section V-C, corresponding
basis of small-signal models, presumably because such kinds to different values of the secondary bifurcation parameter.
of practical-oriented study tend to yield immediate rewards and “Exact” computer simulations presented in Section V-D yield
are usually quite tractable. Detailed nonlinear dynamics and similar bifurcation diagrams. The same exercise is repeated
potential modes of instability of current-programmed dc/dc in Section VI for a practical version of the system, in which
converters, though being important issues with subtle and long- an additional voltage feedback loop is included for load
term practical implications, do not seem to attract as much regulation. Here, with a different set of primary and secondary
attention as they deserve. Recently, chaos from dc/dc converter bifurcation parameters, we again observe similar routes to
circuits under current-programmed control has been studied chaos via regions of quasi-periodicity and period-doubling, as
[6]–[10]. Two distinct types of bifurcation paths have been well as the same bifurcation from quasi-periodicity to period-
doubling as in the case of the open-loop system. In order to
Manuscript received January 8, 1996; revised December 1, 1996. This
work was supported by the Hong Kong Research Grants Council under a confirm the exhibition of quasi-periodic orbits, the Lyapunov
competitive-bid earmarked research grant. This paper was recommended by exponents are computed in Section VII. For a certain range of
Associate Editor T. Endo. the primary bifurcation parameter lying between periodicity
The authors are with the Department of Electronic Engineering, Hong Kong
Polytechnic University, Hung Hom, Hong Kong. and chaos, the Lyapunov exponent takes the value of zero.
Publisher Item Identifier S 1057-7122(97)08045-8. Finally, in Section VIII, we generate a series of return maps
1057–7122/97$10.00  1997 IEEE

Authorized licensed use limited to: PONDICHERRY ENGG COLLEGE. Downloaded on August 9, 2009 at 01:15 from IEEE Xplore. Restrictions apply.
1130 IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS—I: FUNDAMENTAL THEORY AND APPLICATIONS, VOL. 44, NO. 12, DECEMBER 1997

Fig. 1. The boost converter.

numerically to show the transition from a tent-like map to a

logistic-like map. This provides a convenient viewpoint to the
reported phenomena.

II. OPERATION OF THE CURRENT-PROGRAMMED Fig. 2. Schematic of current-programmed boost converter.

BOOST CONVERTER
The boost converter is a second-order circuit comprising an
inductor, a capacitor, a diode, and with the load resistance
connected in parallel with the capacitor. Fig. 1 shows the
basic circuit. When operating in continuous mode, two switch
states can be identified: 1) Switch on and diode off
and 2) Switch off and diode on. The two switch states
toggle periodically. We assume that the circuit takes switch
Fig. 3. Waveform of inductance current.
state one for , and switch state two for
, where is an integer, is the duty
cycle, and is the period. variable, which by comparing with a reference current
generates the on–off driving signal for switch . Specifically,
A. Exact State Equations switch is turned on at the beginning of the cycle, i.e., at
The boost converter circuit described above can be regarded . The inductance current increases while switch is
as a variable structure that toggles its topology according on. As climbs to the value of , switch is turned off, and
to the states of the switches. The boost converter operating remains off until the next cycle begins. A typical waveform of
in continuous mode can thus be described by the following the inductance current is shown in Fig. 3. It should be noted
sequence of state equations: that although an inner current loop always exists, the circuit
is regarded as an open-loop system when the value of is
for fixed, i.e., in the absence of an outer voltage feedback loop.
for (1) However, in practice, an outer voltage feedback loop is usually
included for the purpose of load regulation, which forces the
where denotes the state vector of the circuit, i.e., value of to be a function of the output voltage . In this
and, in particular, the ’s and ’s in (1) are given by case, the converter is said to be under a closed-loop current-
programmed control. In this paper, we will study both the
(2) cases with and without the voltage feedback loop, and show
that the bifurcation from quasi-periodicity to period-doubling
when the switch is in state one, and occurs regardless of the presence of the outer voltage loop.
Furthermore, in the authors’ other publications [11]–[13], the
(3) same bifurcation from quasi-periodicity to period-doubling is
also observed for the buck converter.
when the switch is in state two. The solution to (1) can be
obtained analytically or via a numerical algorithm such as
the Runge–Kutta algorithm. Simulation of the cycle-by-cycle III. A GLIMPSE AT THE NONLINEAR BEHAVIOR OF
time-domain waveforms is possible using the above piecewise CURRENT-PROGRAMMED CONVERTERS
switched model. We will base all our simulations on this Before embarking on a detailed study of the various routes
model. to chaos in current-programmed converters, let us take a quick
tour of some typical operating regimes of a boost converter
B. Review of Current-Programmed Control under current-programmed control. The sample waveforms
The subject of investigation is a current-programmed shown in the following subsections are derived from the exact
switching converter. The schematic diagram is shown in piecewise switched model described in Section II-A. Note that
Fig. 2. The inductance current is chosen as the programming we have included a duty cycle limiter in our simulations such

Authorized licensed use limited to: PONDICHERRY ENGG COLLEGE. Downloaded on August 9, 2009 at 01:15 from IEEE Xplore. Restrictions apply.
CHAN AND TSE: STUDY OF BIFURCATIONS IN CURRENT-PROGRAMMED DC/DC BOOST CONVERTERS 1131

(a) (b)

(c)

Fig. 4. Fundamental periodic operation. (a) Time-domain waveform, (b) phase portrait, and (c) FFT spectrum of the waveform in (a).

that . This will prevent the switch from being closed for B. Quasi-Periodic Orbits
the entire period (i.e., continuously being closed for more than Fig. 7(a) shows a quasi-periodic waveform. (This particular
one period) in the case where the reference current is large, example is a quasi-4 periodic waveform.) The corresponding
as will be manifested in the following simulated waveforms. phase portrait is shown in Fig. 7(b). In a later section, we will
Such an arrangement is also commonly found in practical compute the value of the Lyapunov exponent under this oper-
implementation. ating regime and confirm the exhibition of quasi-periodicity.
Due to its nonperiodic nature, this type of operation tends to
A. Fundamental and Subharmonic Orbits induce noises, some of which fall in the audible range, and is
The most common and perhaps the only acceptable op- thus rather undesirable for practical use [see Fig. 7(c)]. Partly
erating regime employed in practical power supplies is the for this reason, this type of operation has never been allowed
fundamental periodic regime. Fig. 4(a) shows the inductance in the final product, although it is frequently encountered by
current waveform of a typical current-programmed converter engineers during the development stage of a power supply.
under a fundamental periodic operation. The corresponding
phase portrait and the FFT spectrum are shown in Fig. 4(b)
and (c), which demonstrate the stable and periodic nature of the C. Chaotic Orbits
system. Note that a fundamental operation is possible when The current waveform, phase portrait, and corresponding
is sufficiently small. Moreover, when is not small, many FFT spectrum of the waveform for the circuit operating
other operating regimes are possible. For example, Figs. 5 and chaotically are shown in Fig. 8. The spectrum has a continuous
6 show period-two and period-four subharmonic operation, and broad-band nature. Needless to say, conventional power-
respectively. It is worth noting that subharmonic operations supply designers have always banned this type of operation
have never been considered in the practical design of power in their final products, although they could hardly avoid
supplies despite the fact that they are stable. encountering it in their workbench.

Authorized licensed use limited to: PONDICHERRY ENGG COLLEGE. Downloaded on August 9, 2009 at 01:15 from IEEE Xplore. Restrictions apply.
1132 IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS—I: FUNDAMENTAL THEORY AND APPLICATIONS, VOL. 44, NO. 12, DECEMBER 1997

(a) (b)

(c)

Fig. 5. 2T subharmonic operation. (a) Time-domain waveform, (b) phase portrait, and (c) FFT spectrum of the waveform in (a).

In the sequel we will systematically study these oper- terms of and the duty cycle :
ating regimes and, in particular, we will investigate how
the choice of parameters and their values can determine
the type of operation of a current-programmed boost con-
verter.

IV. DERIVATION OF THE ITERATIVE MAP FOR

CURRENT-PROGRAMMED BOOST CONVERTER
Our purpose in this section is to derive a difference equation (5)
for the system which takes the form
where is the transition matrix corresponding to and
(4) is given by

where and subscript denotes the value at the for (6)

beginning of the th cycle, i.e., .
As mentioned in Section II-A, the state equation for the If a truncated series is used for , an approximate iterative
circuit in any switch state can be written in the form map may be obtained which takes the form
, where and are the system matrices in
switch state , and is the input voltage. Using a successive (7)
substitution method, the value of can be expressed in

Authorized licensed use limited to: PONDICHERRY ENGG COLLEGE. Downloaded on August 9, 2009 at 01:15 from IEEE Xplore. Restrictions apply.
CHAN AND TSE: STUDY OF BIFURCATIONS IN CURRENT-PROGRAMMED DC/DC BOOST CONVERTERS 1133

(a) (b)

(c)
Fig. 6. 4T subharmonic operation. (a) Time-domain waveform, (b) phase portrait, and (c) FFT spectrum of the waveform in (a).

The expressions of the ’s and the ’s are given in the V. THE OPEN-LOOP SYSTEM
Appendix.
Since a difference equation of the form (4) is desired,
should be expressed in terms of . By inspection of the A. Overview of the Routes to Chaos and
circuit and the inductor current waveform, we have Choice of Bifurcation Parameters
In [7]–[10], has been chosen as a bifurcation parameter.
Essentially, by varying we would observe the way the
(8)
circuit changes its qualitative behavior from a fundamen-
tal stable system to a chaotic system. We refer to as
(9)
the primary bifurcation parameter. As reported previously in
[6]–[10], routes to chaos via quasi-periodic orbits and period-
Clearly (7) and (9) combine to give the required iterative doubling are possible. However, there has been no attempt to
map. We will use them to predict the chaotic behavior of find out the condition that determines the type of the route
the system. Wherever possible, we will attempt to calculate to chaos for a given set of parameters. In fact, our main
the characteristic multipliers in order to locate possible onsets result in this paper addresses this important issue. To this end
of flip bifurcation. To get better ideas of the possible routes we introduce a secondary bifurcation parameter , which is
to chaos and their relationship with the parameter values, we defined as
resort to numerically calculating the sequence of . Stable
periodic or chaotic orbits may emerge after a sufficiently large
number of iterations. Moreover, the order of periodicity can (10)
be easily checked by inspecting the sequence. In the next
section we will study the various routes to chaos and their We will show later that as increases, the route to chaos
controlling parameters for the open-loop system. The study of initially goes through quasi-4 -periodic orbits, then goes
the closed-loop system will be postponed to Section VI. through quasi-8 -periodic orbits, and continues to go through

Authorized licensed use limited to: PONDICHERRY ENGG COLLEGE. Downloaded on August 9, 2009 at 01:15 from IEEE Xplore. Restrictions apply.
1134 IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS—I: FUNDAMENTAL THEORY AND APPLICATIONS, VOL. 44, NO. 12, DECEMBER 1997

(a) (b)

(c)
Fig. 7. Quasi-4T subharmonic operation. (a) Time-domain waveform, (b) phase portrait, and (c) FFT spectrum of the waveform in (a).

quasi-periodic orbits of 16 , etc. It eventually becomes period- The characteristic multipliers of an iterative function
doubling for large values of . Fig. 9 depicts a panoramic are the roots of the characteristic equation
view of the main result to be reported in this paper. As shown
in this figure, for small values of the secondary parameter, (11)
the route to chaos (with respect to the primary parameter) is
where is the Jacobian of evaluated at the fixed
via quasi-4 -periodic orbits. As we increase the secondary
point . In the case of our boost converter under current-
parameter, the quasi-4 -periodic orbits become stable 4
programmed control, the function can be written in the
subharmonic orbits, and the route to chaos is via quasi-8 -
following form:
periodic orbits. Eventually, for large values of the secondary
parameter, the system exhibits a well-known period-doubling
(12)
route to chaos.

From (7), we have

B. Calculation of Characteristic Multipliers
In this section, we examine the characteristic multipliers of (13)
the system, from which we can rigorously locate the onset of (14)
the first flip bifurcation (i.e., from period-one to period-two
orbits). Essentially, we consider the iterative function of where is, in turn, a function of , as given in (9). Hence,
(4), and compute its characteristic multipliers corresponding the Jacobian is given by
to any given fixed point . The first flip bifurcation occurs
when one of the characteristic multipliers equals 1. Likewise,
computation of the characteristic multipliers of the function (15)
can locate the onset of the second flip bifurcation (i.e.,
from period-two to period-four orbits) if it exists.

Authorized licensed use limited to: PONDICHERRY ENGG COLLEGE. Downloaded on August 9, 2009 at 01:15 from IEEE Xplore. Restrictions apply.
CHAN AND TSE: STUDY OF BIFURCATIONS IN CURRENT-PROGRAMMED DC/DC BOOST CONVERTERS 1135

(a) (b)

(c)
Fig. 8. Chaotic operation. (a) Time-domain waveform, (b) phase portrait, and (c) FFT spectrum of the waveform in (a).

where

(16)

(17)

(18)
Fig. 9. Outline of primary and secondary bifurcations.

We now attempt to use the characteristic multipliers to

(19) predict the occurrence of flip bifurcation. The general pro-
cedure will be exemplified here using the set of circuit
The characteristic multipliers can now be computed by first parameters listed in Table I, with the capacitor value chosen
finding the fixed point corresponding to any given using for . To keep our discussion concise, we will
the Newton–Raphson method or otherwise [14], and then omit details of calculations of the characteristic multipliers,
solving the characteristic equation (11). but focus mainly on the cases in the vicinity of the first flip

Authorized licensed use limited to: PONDICHERRY ENGG COLLEGE. Downloaded on August 9, 2009 at 01:15 from IEEE Xplore. Restrictions apply.
1136 IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS—I: FUNDAMENTAL THEORY AND APPLICATIONS, VOL. 44, NO. 12, DECEMBER 1997

bifurcation. Specifically we report our calculations in Table II, TABLE I

and summarize the trend of the locus as follows (see also CIRCUIT PARAMETERS
Fig. 10).
1) For small (less than 0.497), the magnitudes of the
characteristic multipliers are less than one, implying
stable period-one orbits.
2) As increases, one of the characteristic multipliers
goes toward 1, and at , the characteris-
tic multipliers are 1 and 0.752 484, implying a flip
bifurcation at this value of .
3) For , the iterative function is unstable.
We consider instead the function , denoted by
. In a likewise fashion, we can compute the fixed
point (actually two alternate fixed points) of , and
the characteristic multipliers. In this particular case, right
after the first flip, the characteristic multipliers jump to
about 1 and 0.575, and gradually move toward each
other.
4) When , the two real characteristic multipli-
ers collide and break off to become a complex conjugate
pair.
5) Calculations are no longer possible for a slight increase
of beyond 0.596. Our algorithms involve determina-
tion of the steady-state fixed points from the time series.
However, when , the series becomes a peri- Fig. 10. Locus of characteristic multipliers for = 0:125.
odic having no fixed points. Evaluation of the required
Jacobian thus becomes impossible. (Note: Difficulties in TABLE II
the computation of characteristic multipliers for other CHARACTERISTIC MULTIPLIERS FOR = 0:125
power electronics systems have also been reported by
other authors [10], [15].)
6) At the point where the 2 orbit loses stability, the
norm of the complex characteristic multipliers is ap-
proximately 0.74, which is far from the boundary of the
unit circle. Thus, we are not able to see the movement
of the characteristic multipliers across the boundary of
the unit circle, as would be expected from a typical
Neimark–Sacker bifurcation [16]. (Note: Such a jump
from stable operation to instability or possible chaos will
also be shown in Section VII in the light of Lyapunov
exponents.)
The above results give a partial picture of the bifurcation
path exhibited by the system for a particular value of the
secondary bifurcation parameter. In the following subsection
we attempt to provide (by a series of computer experiments) a
fuller picture regarding the possible routes to chaos exhibited of the bifurcation path from one that goes through quasi-4 -
by the system under study. periodic orbits via progressively longer quasi-periodic orbits
to one that goes through period-doubling is summarized as
follows.
C. Bifurcation Diagrams Derived from the Iterative Map Observation 1:
Based on (7) and (9), we can obtain the discrete time As shown in Fig. 11(a), the circuit goes through stable
values of at for all . In particular, if we allow 1 orbits, stable 2 orbits, quasi-4 orbits, and eventually
the iteration to proceed for a sufficiently long time, the exhibits chaos.
sequence may either diverge, converge to a periodic orbit, or Observation 2:
be attracted to a chaotic orbit. In our numerical experiments, The bifurcation path with respect to , after the stable 1
the values of the components are chosen to ensure that the and 2 regions, enters a region of stable 4 orbits (instead
circuit operates theoretically in the continuous mode, as listed of quasi-4 ), and then quasi-8 , and eventually chaos. The
in Table I. Our experiments with this iterative map have led bifurcation diagram is shown in Fig. 11(b), a blow-up of which
to the bifurcation diagrams shown in Fig. 11. The transition is shown in Fig. 12(a).

Authorized licensed use limited to: PONDICHERRY ENGG COLLEGE. Downloaded on August 9, 2009 at 01:15 from IEEE Xplore. Restrictions apply.
CHAN AND TSE: STUDY OF BIFURCATIONS IN CURRENT-PROGRAMMED DC/DC BOOST CONVERTERS 1137

(a) (b)

(c) (d)

Fig. 11. Bifurcation diagrams using the iterative map. (a) Via quasi-4T orbits ( = 0:125). (b) Via quasi-8T orbits ( = 0:25). (c) Via quasi-16T
orbits ( = 0:48). (d) Via period-doubling ( = 0:625).

Observation 3: 4 .” For this particular case, we can refer to Fig. 7(a) for
The bifurcation diagram is shown in Fig. 11(c). An en- verification.
largement is shown in Fig. 12(b). Here the bifurcation path To verify quasi-periodicity, in Section V-C we will compute
goes through a region of quasi-16 orbits before it enters the the Lyapunov exponent for the entire range of the bifurcation
chaotic region. and demonstrate that the condition for quasi-periodicity, i.e.,
Observation 4: a zero largest Lyapunov exponent, is indeed satisfied.
For this value of , the circuit follows a typical period-
doubling route to chaos, as shown in Fig. 11(d).
Remarks: Up to this point, we have not provided any D. Verification by “Exact” Computer Simulations
proof to the quasi-periodicity claimed above for certain re- In this section, we verify the results reported in Section V-C
gions of the bifurcation diagrams. Indeed, judging from the by computer simulations. As mentioned in Section II, sim-
bifurcation diagrams such as the ones shown in Fig. 11, quasi- ulation based on the model of Section II-A gives the true
periodicity is not at all conclusive. Specifically, based only on waveforms of the circuit. The simulated waveforms can be
the appearance of Fig. 11(a), there is no sufficient evidence used to generate bifurcation diagrams. Specifically, a large
to claim that quasi-4 orbits exist in the parameter range number of the current waveforms are simulated for a range
around . The only conclusive claim is that of values of . The values of the current at are
values visited by the sequence of points are confined to extracted from each waveform, with the initial transient dis-
four open regions. The orbit can be chaotic if the four regions carded. A bifurcation diagram can be obtained by plotting the
are visited irregularly. Moreover, quasi-periodicity can exist discrete current values against . The bifurcation diagrams
if the sequence of visits the four regions in a regular shown in Fig. 13 agree with the predicted bifurcation diagrams
pattern, one after another in a fixed order. Thus, in order developed earlier by the iterative map, except that the predicted
to be conclusive about the exhibition of quasi-4 orbits in patterns are slightly shifted to the right. Such discrepancy is
the range of in Fig. 11(a), we need to due to the use of a truncated Taylor series in the derivation
inspect the time-domain waveform and find out if it is “nearly of the iterative map.

Authorized licensed use limited to: PONDICHERRY ENGG COLLEGE. Downloaded on August 9, 2009 at 01:15 from IEEE Xplore. Restrictions apply.
1138 IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS—I: FUNDAMENTAL THEORY AND APPLICATIONS, VOL. 44, NO. 12, DECEMBER 1997

(a) (b)
Fig. 12. (a) Zoom-in of Fig. 11(b). (b) Zoom-in of Fig. 11(c).

(a) (b)

(c) (d)
Fig. 13. Bifurcation diagrams from computer simulation of the open-loop circuit. (a) = 0:125, (b) = 0:25, (c) = 0:48, and (d) = 0:625.

VI. THE CLOSED-LOOP SYSTEM is shown schematically in Fig. 14. Similar to the open-loop
case, the inductance current, as shown in Fig. 15, ramps up
until it hits the reference current and ramps down until
A. Operation of the Closed-Loop the switch is turned on by the periodic clock signal. However,
Current-Programmed Boost Converter in this case, the reference current is controlled by the output
For power supplies that are used as regulated voltage voltage according to the following simple feedback scheme:
sources, the mandatory requirement for output voltage reg- (20)
ulation necessitates the inclusion of an outer voltage feedback
loop. In the case of the current-programmed boost converter, where is the feedback factor and can be chosen to modify
a typical arrangement for incorporating such a feedback loop the closed-loop dynamics, is the reference steady-state

Authorized licensed use limited to: PONDICHERRY ENGG COLLEGE. Downloaded on August 9, 2009 at 01:15 from IEEE Xplore. Restrictions apply.
CHAN AND TSE: STUDY OF BIFURCATIONS IN CURRENT-PROGRAMMED DC/DC BOOST CONVERTERS 1139

TABLE III
VALUES OF Iref AND THE CORRESPONDING REFERENCE VOLTAGES

the bifurcation of the route to chaos with respect to variation

of .
In developing a bifurcation diagram with respect to variation
of , we need to keep fixed. In addition, the process
requires that the corresponding steady-state value of the output
voltage be known and be assigned to . Thus, in preparation
for the derivation of the bifurcation diagrams, we need to
precalculate the values of corresponding to some arbi-
trarily chosen . This can be conveniently accomplished
by performing a few simulation runs. Results are given in
Fig. 14. Schematic diagram of the closed-loop system.
Table III.

C. Bifurcation Diagrams from Computer Simulations

Similar to the analysis of the open-loop system, two ap-
proaches can be taken to derive a series of bifurcation dia-
grams, as outlined in Sections V-C and V-D. The first ap-
proach essentially makes use of an approximate iterative
map, which is derived from writing the involving transition
matrices as truncated Taylor series, to develop the bifurcation
diagrams. The second approach, which is simpler and more
Fig. 15. Waveform of inductance current in the presence of voltage feedback straightforward, simulates the required waveforms directly
loop. using the exact piecewise switched model and obtains the
bifurcation diagrams from the simulated waveforms. In order
output voltage, and is the steady-state reference current. to keep our presentation in this paper concise, we will omit the
Since , the following approximation holds: iterative map approach for the closed-loop system. Neverthe-
less, we should stress that both approaches, as demonstrated
(21) in Sections V-C and V-D for the open-loop system, will
yield essentially the same piece of information regarding the
Also, by inspection of the waveform, we have bifurcation of the route to chaos.
The bifurcation diagrams obtained from computer simu-
(22) lations corresponding to the four values of listed in
Table III are shown in Fig. 16, from which we clearly see
Finally, solving will allow the difference equation for the that the route to chaos undergoes exactly the same kind of
closed-loop system to be written as transition as we last saw in the open-loop system. In other
words, the bifurcation phenomena exhibited by the closed-
(23) loop system are found to fit the general description outlined by
Fig. 9. Moreover, we may conjecture from the above results
where is as previously defined. that Fig. 9 represents a typical bifurcation scenario for the
current-programmed dc/dc converter operating in whatever
B. Bifurcation Parameters configuration (e.g., open-loop or closed-loop) and with what-
The route to chaos can be predicted from (23) using the ever choice of bifurcation parameters that are suitable for the
analysis method described in Section V. However, in the case type of configuration concerned.
of the circuit with an outer voltage feedback loop, is no
longer a fixed value that can be selected at will and, hence, is C. Lyapunov Exponents
not qualified to serve as a primary bifurcation parameter. Thus, In the study of chaotic systems, Lyapunov exponents are
a different set of primary and secondary bifurcation parameters often used to identify and quantify chaos. In simple terms, the
must be used. In our investigation, is chosen as the primary signs of the Lyapunov exponents determine whether stretching
bifurcation parameter, and is the secondary bifurcation or contracting occurs in a certain direction in the phase
parameter. With this choice of parameters, we first will study space. For a th order system, there are Lyapunov expo-
the routes to chaos with respect to variation of , and then nents, each of which describes the extent of stretching or

Authorized licensed use limited to: PONDICHERRY ENGG COLLEGE. Downloaded on August 9, 2009 at 01:15 from IEEE Xplore. Restrictions apply.
1140 IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS—I: FUNDAMENTAL THEORY AND APPLICATIONS, VOL. 44, NO. 12, DECEMBER 1997

(a) (b)

(c) (d)
Fig. 16. Bifurcation diagrams from computer simulation of the closed-loop circuit. (a) Iref = 0:2006A, (b) Iref = 0:25A, (c) Iref = 0:2918A,
(d) Iref = 0:3528A.

contracting in one direction [17]. For stable periodic and

subharmonic motions, Lyapunov exponents for all directions
must be negative. For quasi-periodic motions, the largest
Lyapunov exponent is zero, and for chaotic motions, the largest
Lyapunov exponent is positive. Hence, Lyapunov exponents
can be used to distinguish between quasi-periodic and chaotic
behavior [17].
Mathematically, the Lyapunov exponents ( ) for a th
order dynamical system are defined by

eig (24)

where is the Jacobian of the system evaluated along Fig. 17. The largest Lyapunov exponent versus iref , corresponding to
Fig. 11(a).
the trajectory [17]. If the differential or difference equations
defining the system are known, we can apply the above for-
mula to find the Lyapunov exponents. A practical and efficient and chaotic regions, we have resorted to the time series
algorithm, based on the Gram–Schmidt orthonormalization method [19]. Our purpose of calculating the largest Lya-
technique, is often employed to calculate [18]. On the punov exponent here is mainly to verify quasi-periodicity
other hand, if the equations that describe the system are not that has been predicted from the bifurcation diagrams and
known, we can evaluate the Lyapunov exponents directly from time-domain waveforms. For brevity and to avoid presenting
the system’s time-domain waveform or sampled time series redundant results, we report here our calculations of the largest
[19]. Lyapunov exponent corresponding to the bifurcation diagram
For computational simplicity, we have calculated the largest of Fig. 11(a). As shown in Fig. 17, for this particular case,
Lyapunov exponent for fundamental and subharmonic regions the system exhibits quasi-periodicity in the region around
using the Gram–Schmidt algorithm [18]. For quasi-periodic .

Authorized licensed use limited to: PONDICHERRY ENGG COLLEGE. Downloaded on August 9, 2009 at 01:15 from IEEE Xplore. Restrictions apply.
CHAN AND TSE: STUDY OF BIFURCATIONS IN CURRENT-PROGRAMMED DC/DC BOOST CONVERTERS 1141

(a) (b)

(c)
Fig. 18. Return map for (a) Fig. 11(a), (b) Fig. 10(b), and (c) Fig. 11(c).

VII. AN ALTERNATIVE VIEWPOINT USING RETURN MAPS In the process of constructing return maps, we employ (7)
Return maps represent very useful tools for analyzing and (9) to generate discrete-time values of at ,
chaotic systems [20], in particular high-order chaotic systems with the first 1000 transient points discarded. We then plot
and experimental chaos for which analytical or closed-form the points , for to , i.e., 15 000
points plotted. To maintain a concise presentation, we report
iterative functions are difficult to obtain. Essentially, the
in Fig. 18 the results for the case of . Here, we
method allows the dynamics of the system to be summarized
clearly observe the qualitative change from a tent-like map
by a 1-dim map. Here, we make use of this method to examine
to a smooth logistic-like map [21].
the transition of the route to chaos, subject to variation of the
secondary bifurcation.
In the previously derived bifurcation diagrams, we have VIII. CONCLUSION
observed that the route to chaos undergoes a region of quasi- Although very few formal reports have been written about
periodicity for small values of the secondary bifurcation pa- chaos in dc/dc converters, the power-supply engineers have
rameter, and becomes period-doubling for large values of lived with chaos ever since the introduction of dc/dc converters
the secondary bifurcation parameter. It would be interesting and their overwhelming use in power-supply design. Despite
to see how the appearance of the return map varies as the its importance and frequent occurrence, chaos in dc/dc con-
secondary bifurcation parameter changes, and it would be quite verters is still rarely studied, presumably because the kind of
enlightening if we could relate the return maps of the second- approach required for the study of chaotic dynamics is rather
order system under study to some familiar 1-dim maps. For unfamiliar to the power electronics engineers. Up to now, only
instance, period-doubling is often associated with a logistic a few papers have been published on the subject of chaos in
type of 1-dim maps, and the bifurcation via period-doubling is dc/dc converters. A recent survey of published work in this
suppressed in a tent type of 1-dim maps. In the following, we area can be found in [22].
will show that the return maps corresponding to the current- This paper focuses in particular on the current-programmed
programmed converter in question appear at first as a tent-like boost converter and attempts to put together the few isolated
map, and as the secondary bifurcation parameter increases, results reported previously on the routes to chaos exhibited
the return maps approach a logistic-like map with an obvious by such systems. Specifically, previous studies have identified
unimodal appearance. two types of route to chaos, namely via regions of quasi-

Authorized licensed use limited to: PONDICHERRY ENGG COLLEGE. Downloaded on August 9, 2009 at 01:15 from IEEE Xplore. Restrictions apply.
1142 IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS—I: FUNDAMENTAL THEORY AND APPLICATIONS, VOL. 44, NO. 12, DECEMBER 1997

periodicity and via period-doubling [6]–[10]. In this paper, we [2] F. C. Lee and R. A. Carter, “Investigations of stability and dynamic per-
have shown that the bifurcation path exhibited by a current- formances of switching regulators employing current-injected control,”
in IEEE Power Electron. Specialists Conf. Rec., June 1981, pp. 3–16.
programmed dc/dc converter with respect to variation of a [3] R. D. Middlebrook, “Topics in multiple-loop regulators and current-
primary bifurcation parameter undergoes a bifurcation with mode programming,” in IEEE Power Electron. Specialists Conf. Rec.,
respect to variation of a secondary bifurcation parameter, Toulouse, France, June 1985, pp. 716–732.
[4] D. M. Mitchell, “An analytical investigation of current-injected control
starting from one that goes through quasi-4 -periodic orbits for constant frequency switching regulators,” in IEEE Power Electron.
via progressively longer quasi-periodic orbits, eventually to Specialists Conf. Rec., 1985, pp. 225–233.
[5] R. Redl and N. Sokal, “Current-mode control, five different types,
period-doubling. We conclude this paper by reiterating our used with the three basic classes of power converters,” in IEEE Power
conjecture that such a bifurcation is characteristic of current- Electron. Specialists Conf. Rec., 1985, pp. 771–785.
programmed dc/dc converters, and as long as a suitable set [6] C. K. Tse and W. C. Y. Chan, “Chaos from a current programmed Ćuk
converter,” Int. J. Circuit Theory Appl., vol. 23, pp. 217–225, May–June
of bifurcation parameters is chosen, the route to chaos will 1995.
exhibit this typical bifurcation. [7] , “Instability and chaos in a current-mode controlled Ćuk con-
verter,” in IEEE Power Electron. Specialists Conf. Rec., Atlanta, GA,
June 1995, pp. 608–613.
APPENDIX [8] I. Zafrany and S. Ben-Yaakov, “A chaos model of subharmonic oscilla-
tions in current mode PWM boost converters,” in IEEE Power Electron.
Expressions for the matrix elements in the iterative map Specialists Conf. Rec., Atlanta, GA, June 1995, pp. 1111–1117.
derived in Section IV are summarized below. For brevity, we [9] J. H. B. Deane and D. C. Hamill, “Chaotic behavior in current-mode
controlled dc/dc converters,” Inst. Elect. Eng. Electron. Lett., vol. 27,
define and . All other symbols are no. 13, pp. 1172–1173, June 1991.
consistent with those given in the paper: [10] J. H. B. Deane, “Chaos in a current-mode controlled boost dc/dc
converter,” IEEE Trans. Circuits Syst. I, vol. 39, pp. 680–683, Aug.
1992.
[11] W. C. Y. Chan and C. K. Tse, “A universal bifurcation path conjectured
from current-programmed switching converters,” in Proc. Int. Symp.
Nonlinear Theory Appl., Rochi, Japan, Oct. 1996, pp. 121–124.
[12] , “Studies of routes to chaos current-programmed dc/dc convert-
ers,” in IEEE Power Electron. Specialists Conf. Rec., Baveno, Italy,
June 1996, pp. 789–795.
[13] , “Bifurcations in current-programmed dc/dc buck switching
regulators—Conjecturing a universal bifurcation path,” Int. J. Circuit
Theory Appl., to be published.
[14] T. S. Parker and L. O. Chua, Practical Numerical Algorithims for
Chaotic Systems. Berlin, Germany: Springer-Verlag, 1989.
[15] S. Jalali, I. Dobson, R. H. Lasseter, and G. Venkataramanan, “Switching
time bifurcations in a thyristor controlled reactor,” IEEE Trans. Circuits
Syst. I, vol. 43, pp. 209–218, Mar. 1996.
[16] Y. A. Kuznetsov, Elements of Applied Bifurcation Theory. Berlin,
Germany: Springer-Verlag, 1996.
[17] R. C. Hilborn, Chaos and Nonlinear Dynamics. London, U.K.: Oxford
University Press, 1994.
[18] H. Peitgen, H. Jürgens, and D. Saupe, Chaos and Fractals—New
Frontiers of Science. Berlin, Germany: Springer-Verlag, 1992.
[19] A. Wolf, J. B. Swift, H. L. Swinney, and J. A. Vastano, “Determining
Lyapunov exponents from a time series,” Physics D, vol. 16, pp.
285–317, 1985.
[20] R. H. Simoyi, A. Wolf, and H. L. Swinney, “One dimensional dynamics
in a multicomponent chemical reaction,” Phys. Rev. Lett., vol. 49, pp.
245–248, 1982.
[21] S. N. Rashand, Chaotic Dynamics of Nonlinear Systems. New York:
Wiley, 1990.
[22] D. C. Hamill, “Power electronics: A field rich in nonlinear dynamics,”
presented at the Int. Workshop Nonlinear Dynamics Electron. Syst.,
Dublin, Ireland, July 1995.

William C. Y. Chan received the B.S. degree in

engineering physics with first class honors from
Hong Kong Polytechnic University, Hung Hom,
Hong Kong, in 1994, and is currently working
ACKNOWLEDGMENT toward the Ph.D. degree in electronic engineering.
The authors wish to thank the anonymous reviewers for His main research interests include the identifi-
cation and theoretical study of chaotic behavior of
their comments and suggestions which helped to improve the power electronics. He has co-authored 10 technical
content of this paper. papers in this area.
Mr. Chan was the recipient of the Sir Edward
Youde Memorial Fellowship awarded by the Sir
REFERENCES Edward Youde Memorial Council, Hong Kong, in 1997.
[1] A. Capel, G. Ferrante, D. O’Sullivan, and A. Weinberg, “Application
of the injected current model for the dynamic analysis of switching
regulators with the new concept of LC3 modulator,” in IEEE Power Chi K. Tse (M’90–SM’97), for a photograph and biography, see p. 675 of
Electron. Specialists Conf. Rec., Syracuse, NY, June 1978, pp. 135–147. the August 1997 issue of this TRANSACTIONS.

Authorized licensed use limited to: PONDICHERRY ENGG COLLEGE. Downloaded on August 9, 2009 at 01:15 from IEEE Xplore. Restrictions apply.