Received: 18 July 2018 Revised: 30 November 2018 Accepted: 10 December 2018

DOI: 10.1002/etep.2813

RESEARCH ARTICLE

Modeling and analysis of complex dynamics for dSPACE

controlled closed‐loop DC‐DC boost converter

Subrata Banerjee1 | Arnab Ghosh2 | Sanjeevikumar Padmanaban3

1
Department of Electrical Engineering,
Summary
National Institute of Technology
Durgapur, Durgapur, India DC‐DC switched mode power converter circuits are time varying and nonlinear
2
Department of Electrical Engineering, in nature. This work analyzes the modeling and complex dynamics in voltage
National Institute of Technology mode controlled (VMC) of the Boost converter in continuous conduction mode
Rourkela, Rourkela, India
3 (CCM) of operation by using continuous‐time model. The switching converter is
Department of Energy Technology,
Aalborg University, Esbjerg, Denmark governed by naturally sampled constant frequency pulsed signals. Mathematical
modeling of the boost converter numerically developed by using differential
Correspondence
Arnab Ghosh, Department of Electrical
equations and tested in simulation software. The switching converter may
Engineering, National Institute of exhibit fundamental, quasiperiodic, and chaotic oscillations by the systematic
Technology Rourkela, Odisha 769008, changing of converter's variables. The stability of the system investigated
India.
Email: aghosh.ee@gmail.com
through the locus of the complex eigenvalues and the characteristic multipliers
locating the onset of Hopf bifurcation. The one‐periodic orbit loses its stability
via Hopf bifurcation, and the resulting attractor is a quasiperiodic orbit. A
dSPACE controlled boost converter prototype hardware fabricated to establish
the experimental studies in this work. Both the computational and experimental
results have been included to validate the set analysis. It is observed that the
route to chaos reached by the slow‐scale instability in this proposed work.

KEYWORDS
chaos, discrete time iterative modeling, dSPACE, phase‐plane trajectories, state‐space equations

1 | INTRODUCTION

The DC‐DC switching converters are renowned for their domestic as well as industrial applications from a few hundred
watts to several hundreds of kilowatts. Due the presence of, the switches, nonlinear elements (like diode), and control
strategies (eg, pulse width modulation), the circuits behave nonlinear and time‐varying dynamical systems. The con-
verters may show complex phenomenon with the variation of circuit parameters and that may cause to unusual mag-
nified noise, EMI problem, and nonlinear oscillation.1 The exploration and analysis of the complex dynamics of a
system had made phenomenal progression in the late 19th century. It is most striking that simple systems may behave
“random” like oscillation by changing their parameters, and system dynamics may deny “long‐term predictability” even
the initial conditions are known.2 Such behavior is well‐known as chaos, which may introduce the complexity in real‐
world systems. The mathematicians, engineers, and scientists from different disciplines have observed a similar sort of
complex phenomena in their systems. The study of nonlinear dynamics of power converters is the popular ongoing

List of symbols and abbreviations: CCM, Continuous Conduction Mode; DCM, Discontinous Conduction Mode; EMI, Electromagnetic
Interference; MOSFET, metal‐oxide‐semiconductor field‐effect transistor; PV, Photovoltaic; RHP, Right Half Plane; SSA, State Space Averaging

Int Trans Electr Energ Syst. 2019;e2813. wileyonlinelibrary.com/journal/etep © 2019 John Wiley & Sons, Ltd. 1 of 17
https://doi.org/10.1002/etep.2813
2 of 17 BANERJEE ET AL.

research area for the last decades. Nonlinear modeling and complex dynamics in switching converters have been
derived and observed by W.C. Chan et al, S. Banerjee et al, A. El Aroudi et al, C. K. Tsc etc.1-6 O. Dranga et al reported
the bifurcations in power factor corrected boost converter. Investigation of low‐frequency phenomena, fast scale insta-
bilities in switching converter carried out by H.H.C. Iu et al, A. Ghosh et al, for their versatile zone of applications.7-10
M. Arjun et al, M. Zhioua et al have been reported their recent research contributions in modeling, dynamics, bifurca-
tion behavior, and stability analysis of switching converters in applications of renewable energy.11,12
The DC‐DC switching converter is the external clock driven system, the frequency of the clock pulse is similar
to the converter's switching frequency. The dynamics of the system are fully nonautonomous and nonlinear in
nature.13-15 The nonlinear dynamics of the converter system depend upon the different circuit parameters of the con-
verter like as input voltage, inductance, capacitance, and the switching frequency, etc. The periodicity and stability of
the periodic orbit (limit cycle) have been changed because of the systematic variation of the converter parameters that
causes subharmonic oscillations in the converter.15 It is a normal habit to mention the ranges of parameters where the
switching converter will reliably operate in steady‐state condition by avoiding any occurrence of subharmonic oscilla-
tions. The parameters of the converter chosen based on the desired specifications like requiring power and voltage rat-
ing, good transient and the steady‐state response, etc. The proposed switching converter has a problem of
nonminimum phase because of a right half plane (RHP) zero in converter's plant transfer function.16,17 Therefore,
the type controllers18-21 are the most efficient to exhibit the good closed‐loop response with parametric uncertainty,
load, and line variations. It is important to estimate and check the fast‐scale and slow‐scale stability5,6,22 from simula-
tion and experimental results over a large parameter range because these will have an effect on subharmonic
oscillation.
R. S. Alishah et al discussed about the modeling and designing of boost converter for photovoltaic applications.23
J. M. Wang et al proposed a control scheme for power factor corrected boost converter, which can provide good tran-
sient response.24 A study on the nonlinear dynamical characteristic for current mode control boost converter in con-
tinuous conduction mode (CCM) has been described by N. Zamani et al. But only integral controller was used for the
modeling and implementation purpose.25 The modeling of a boost converter by using state space averaging (SSA)
method in CCM and discontinous conduction mode (DCM) mode operation have been discussed by F. M. Shahir
et al.26 M. Zhioua et al studied a boost converter, which is fed by a photovoltaic (PV) generator and supplying a
constant voltage to load. Where the overall system behavior is observed under parameter changes, and a series of
bifurcation diagrams are computed from the circuit. The first‐scale instability that takes place in the system (ie,
period‐doubling bifurcation) when primary parameter is varied.12 The synchronization and anti‐synchronization of
chaotic systems with application to boost converter in simulation were described by R. Sakthivel et al.27 L. Cheng
et al introduced a circuit‐oriented geometrical approach in predicting subharmonic oscillation of VMC switching con-
verters.28 The transient characteristics of the cascaded boost converter under a large disturbance by using nonlinear
modal series method were reported by H. Zhang et al.29 The nonlinear phenomena like chaos and other periodic
motion, which are influenced by system parameters, topological structure, load, and pulse period in a single‐stage
boost converter were discussed by R. Zhang et al.30 B. S. Tekpeti et al studied the design and modeling of boost con-
verter for photovoltaic systems in simulation.31
Although the considerable amount of works has already been done by previous researchers,12,23-31 some important
portions of the dynamics study of converters are still uncovered. It has been observed from the aforesaid literatures that
the most of the prior researchers have been used state space averaging (SSA) method for converter modeling, has been
implemented using integral controller for closed‐loop operation etc. Some advanced modeling approaches (like discrete
time iterative approach etc) are needed for fetching of all kind of nonlinearites in switching converters. Only integral
control is not sufficient for showing good transient characteristics of the converter, which have nonminimum phase
problem (like boost converter etc). Also the effect of control parameter's variation to converter dynamics has not been
studied in most of the prior works. In the present work, discrete time iterative approach is implemented for converter
modeling and the complex dynamics (ie, chaos and bifurcation) have been studied for different circuit parameters (like
as load resistance, output capacitance, controller gain, etc) in voltage mode control for DC‐DC boost converter. This
study is important for different practical implementations like switch mode power supply, renewable energy systems,
electrical vehicles, etc. In the proposed work, Type‐III controller is introduced to enhance the converter's transient per-
formances, and also the study of complex dynamics of the converter has been carried out with the changes of controller
and other circuit parameters. So the observed transient performance is better compared with the prior works.
So the present work has added additional contributions, which are indifferent from the recent published works. The
most important aim of this paper is to study, analysis, and explore the theoretical and practical chaos for Type‐III
BANERJEE ET AL. 3 of 17

controller‐based VMC boost converter circuit by using dSPACE‐based real time controller. The study of nonlinear
dynamics of Type‐III controller‐based boost converter by using DSP‐based real time platform is reported first time in
this work. It is noticeable that the route to chaos reached by the slow‐scale instability. This approach is conceptually
simple to implement, and the designers can easily find out the probable operating zone. By identifying the zones of
quasiperiodic and chaotic oscillation for accurate range of parametric values will assist to design a chaos free power
supply.

2 | DC‐ DC B OOS T CON VE R TER

The schematic diagram of VMC boost converter is shown in Figure 1. The power circuit comprises with a diode (D), a
MOSFET switch (SW), a capacitor (C), an inductor (L), and a load resistance (RL). The switching pulses of the MOSFET
are controlled by the output of the comparator through a driver circuit. In this present work, the main objective is to
study the nonlinear dynamics of dSPACE controlled closed‐loop converter.
It can be viewed from the schematic diagram (Figure 1) that the converter's output voltage (Vo) is sensed and com-
pared with a reference voltage (Vref). The generated error signal is passed through A/D converter to the digital control-
ler, and a control signal is generated from Type‐III controller, which is embedded in dSPACE platform. Pulse Width
Modulation (PWM) signal is generated after having comparison between control signals and high frequency triangular
waveform. It can be seen that if the control voltage is increased, the duty ratio is also increased (ie, on time increases
and off time decreases) and vice versa. The triangular waveform has fixed amplitude and frequency (switching fre-
quency of the converter), and the duty ratio is adjusted only by change in control signal. PWM signal is passed through
D/A converter, opto‐isolator, and astable multivibrator circuits before driving the gate of the MOSFET switch.

3 | S T A T E‐SPACE EQUATIONS OF DC‐DC B O O ST CO NV E R T E R

By the presence of nonlinear elements (like MOSFET switch, diode) and controlling signal (like PWM), the converter
circuit is an example of highly nonlinear and time‐varying system.5 The block diagram of the boost converter shown
by the Figure 1. A free running clock is there to control the switching of the converter. The switch is turn on at the
beginning of each clock pulse, and the clock frequency is similar to the switching frequency and operated under the
CCM. Therefore, exhibits two switching instants ie, (a) switch on and (b) switch off condition. The differential equations
of converter's states derived and elaborated below. The output capacitor voltage (vc) and the inductor current (iL) are
considered as state variables of the converter dynamical system. Figure 2

FIGURE 1 Schematic diagram of

proposed converter in closed‐loop
operation
4 of 17 BANERJEE ET AL.

FIGURE 2 A, Switch on and B, switch off circuits of boost converter

Mode 1: switch (SW) is on.

Applying Kirchhoff's current law (KCL)
Vo
i c ðt Þ ¼ − (1)
RL

dvc ðt Þ 1
or; ¼− v c ðt Þ (2)
dt ðRL þ r c ÞC

Applying Kirchhoff's voltage law (KVL)

diL ðt Þ
V in − L − r L × i L ðt Þ ¼ 0 (3)
dt

diL ðt Þ rL 1
or; ¼ − iL ðt Þ þ V in (4)
dt L L
Now, we can write both the state equations in matrix form
dX ðt Þ
¼ Ao1 X ðtÞ þ Bo1 V in for nT sw ≤ t < ðn þ dÞT sw
dt (5)
V o ðt Þ ¼ Co1 X ðt Þ


v c ðt Þ
where, X(t) is noted as the state vector of this circuit ie, X ðt Þ ¼ and Ao1 , Bo1 , Co1 can be written as
iL ð t Þ
2 3
1 " # 2 3
6 − 0 7 0 RL
ðRL þ r c ÞC
Ao1 ¼ 6
4
7 ; Bo ¼ 1 ; Co1 ¼ 4 RL þ r c 5
rL 5 1
0 − L 0
L
Mode 2: switch (SW) is off.
Applying KCL

Vo
iL ðtÞ ¼ ic ðt Þ þ (6)
RL

dvc ðt Þ 1 RL
or; ¼− vc ðt Þ þ i L ðt Þ (7)
dt ðRL þ r c ÞC ðRL þ r c ÞC

Applying KVL

diL ðt Þ
V in − L − r L × i L ðt Þ − V o ¼ 0 (8)
dt

 
diL ðt Þ RL rL r c RL 1
or; ¼− vc ðt Þ þ − − iL ðt Þ þ V in (9)
dt ðRL þ r c ÞL L ðRL þ r c ÞL L
BANERJEE ET AL. 5 of 17

Now, we can write both the state equations in matrix form.

dX ðt Þ
¼ Ao2 X ðt Þ þ Bo2 V in
dt (10)
V o ðtÞ ¼ Co2 X ðt Þ for ðn þ dÞT sw ≤ t < ðn þ 1ÞT sw

where Ao1 , Bo1 , C o1 can be written as

2 1 RL 3 2 3
" # RL
− 0
6 ð þ r c ÞC ðRL þ r c ÞC 7 o 6 7
Ao2 ¼ 6
R L 7; B ¼ 1 ; Co ¼ 6 RL þ r c 7
4 RL rL r c RL 5 2 2 4 r c RL 5
− − − L
ðRL þ r c ÞL L ðRL þ r c ÞL RL þ r c

The Equation 5 and Equation 10 combined in only one expression by neglecting parasitic elements.
   
dX ðt Þ −1=ðRL CÞ ð1 − dÞ=C 0
¼ f ðX; t Þ ≡ X ðt Þ þ b1 SW ðt Þ (11)
dt − ð1 − dÞ=L 0 V in =L


1 if t ∈ SW
where; b
1 SW ðt Þ ¼ (12)
0 if t ∉ SW

3.1 | Discrete time iterative mapping for nonlinear modeling of the converter

In the first step of the discrete time iterative mapping, the state equations of individual switching instants written down
and finally the difference equation for the overall system derived. The converter is operating here in CCM, and two cir-
 0
cuits can be identified, first one corresponds to the “switch on” instant, t n ≤ t < t n and the another one represents
0
“switch off” interval t n ≤ t < t nþ1 . The details derivations of the state equations are discussed in the previous section.
For the sake of simplicity, the parasitic elements such as rL and rc are not considered here.
In this case, the sparseness of the matrix Ao1 can be derived from the solution of the switch on interval easily by tak-
ing integration to the right hand side of Equation 5, ie,
2 − 3
v ð t Þe ðt−t n Þ=CRL
6 c n 7 0
X ðt Þ ¼ 4 V in ðt − t n Þ 5for t n < t < t n (13)
i L ðt n Þ þ
L
0
Now, putting t ¼ t n to Equation 13 and the value of X obtained at the end of the switch on interval.

0 −
vc t n ¼ vc ðt n Þe dT sw =CRL (14)

0 V in dT sw
iL t n ¼ iL ðt n Þ þ (15)
L
0
where, d ¼ t n =T sw is known as duty cycle.
By applying Laplace transformation on Equation 10 in s‐domain to get the solution for the switch off state interval,
note that the Laplace transform of the input voltage is Vin/s.
 −1 h
0  i
X ðsÞ ¼ sI−Ao2 X t n þ Bo2 V in ðsÞ (16)

2 32
0 3
1
6 s C 76 vc t n
7
6 7
4 1 1 54  0 V in 5
− sþ iL t n þ
L CRL sL
¼
s 1
s2 þ þ
CRL LC
6 of 17 BANERJEE ET AL.

Here, X(s) represents the Laplace transform of X(t). Now, it may be written from the Equation 16 that the mathematical
expressions for the state variables in s‐domain as

V in K 1 s þ K 2
V c ðsÞ ¼ þ (17)
s H ðsÞ

V in K 3 s þ K 4
I L ðsÞ ¼ þ (18)
RL s H ðs Þ

0 1 0  0 V in s 1 1 0 1 0
where, K 1 ¼ vc t n − V in , K 2 ¼ iL t n − 2σV in , K 3 ¼ iL t n − , H ðsÞ ¼ s2 þ þ , K4 ¼ iL t n − v c t n
  C RL CRL LC CRL L
RL V in 1
þ − 2σ ,σ¼ .
L RL 2CRL
Now, the inverse Laplace transformation is used for the partial fraction expressions of Vc(s) and IL(s), the time‐
0
domain equations of vc and iL for the interval tn < t < t nþ1 can be expressed as

0
 K −K σ

1 −σ ðt−t 0n Þ
vc ðtÞ ¼ V in þ K 1 e−σðt−tn Þ cos ω t − t n þ
0 2 0
e sinω t − t n (19)
ω

V in 0
 K −K σ

3 −σ ðt−t 0n Þ
þ K 3 e−σ ðt−tn Þ cos ω t − t n þ
0 4 0
i L ðt Þ ¼ e sinω t − t n (20)
RL ω
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
where, ω ¼ − σ2.
LC
0 0
From K1, K2, K3, and K4 are the term of X t n and that X t n is also the function of X(tn). The d, difference equation
0
is involved with X (tn + 1), X (tn) and d found after putting t = tn + 1 and t nþ1 − t n ¼ ð1 − dÞT sw into Equation 19 and
Equation 20. The general form of discrete‐time iterative mapping of the converter expressed as below

X ðt nþ1 Þ ¼ f ðX ðt n Þ; dÞ (21)

where, the term f (.) is written as



f 11 f 12 g1
f ðX; dÞ ¼ Xþ V in (22)
f 21 f 22 g2

where,

8 h σ i
>
> f ¼ e −dT sw
−σ ð 1−d ÞT cos ð 1 − d ÞωT − sin ð 1 − d ÞωT
> 11
CRL
>
>
sw sw
ω
sw
>
>
>
> 1 −σ ð1−dÞT sw
>
> f 12 ¼ e sinð1 − dÞωT sw
>
> ωC
>
>
>
> 1 dT sw
>
> f 21 ¼ − e− CRL −σ ð1−dÞT sw sinð1 − dÞωT sw
>
> ωL
>
>
  
>
>
< −σ ð1−dÞT sw 1 1
f 22 ¼ e cosð1 − dÞωT sw þ − σ sinð1 − dÞωT sw
ω CRL
>
>
  
>
> 1 dT sw
>
> g1 ¼ e −σ ð1−dÞT sw
cosð1 − dÞωT sw − − σ sinð1 − dÞωT sw
>
> ω LC
>
>
>
> 
 
>
> 1 RL dT sw
>
> g2 ¼ 1 þ e−σð1−dÞT sw − 1 cosð1 − dÞωT sw
>
>
>
> RL L
>
>
   
>
> 1 1 RL dT w RL
>
: þ −σ þ − σ sinð1 − dÞωT sw
ω CRL L L
BANERJEE ET AL. 7 of 17

4 | T Y P E‐I I I C O N T R O LL E R D Y N A M I C S

Generally, many different classical (conventional) controllers like PI, PID, lead‐lag, etc, used to achieve preferred closed‐
loop response of the converter.17 In case of CCM of operation of the boost converter, a zero is observed in the right half
plane (RHP) of the control‐to‐output transfer function of converter plant. Hence, the converter shows the poor dynamic
response for the occurrence of nonminimum phase problem.18 This RHP zeroes restrict the closed‐loop bandwidth of
the switching converter and that reason for slower response.19 It is challenging for PID controller to show fast
closed‐loop response with parametric uncertainty, load, and line changes.20 Whereas, the type controllers17-21,23
reported to show better dynamic response for this class of switching converter.
“Type‐III” controller is a lead‐lead controller with a pole at the origin. The origin pole is responsible for providing
extremely large gain at low frequencies. The other pole‐zero pairs may decrease the phase shift between the frequency
of the two zeroes and the frequency of two poles as lead controller.21 That is why this controller delivers the phase boost
of 0° to 180° with zero steady‐state error. However, converter as a problem of nonminimum, but this controller may
provide better closed‐loop dynamics. The basic structure of a Type‐III controller is given in Figure 3A. It can be noticed
from frequency domain diagram (Figure 3B) that changing the locations of controller's pole‐zero combinations maxi-
mum 180° phase boost can be achieved. The Type‐III controller plant transfer function can be written from the control-
ler's basic structure of Figure 3A.

ð1 þ s=ωz1 T−III Þð1 þ s=ωz2 T−III Þ

Tc T−III ðsÞ ¼   (23)
s=ωp0 T−III 1 þ s=ωp1 T−III 1 þ s=ωp2 T−III

It is notable, one pole ( f p0_III) at origin, and the other two poles are at f p1_T‐III = 1/2πR3C3 and f p2_T‐III = (C1 + C2)/
2πR3C3, respectively. The zeros are presented at f z1_T‐III = 1/2πR2C1 and f z2_T‐III = 1/2π(R1 + R3)C3. The two controller
gains are (i) K1_T‐III = R2/R1 and (ii) K2_T‐III = R2(R1 + R3)/R1R3. Now, the two zeros are assumed at the same point, and
likely, the two poles are considered same point. Thus, the double pole and double zero are placed at ωz1_T‐III = ωz2_T‐III
= ωz1,2_T‐III and ωp1_T‐III = ωp2_T‐III = ωp1,2_T‐III.

 2
1 þ s=ωz1;2 T−III
Tc T−III ðsÞ ¼  2 (24)
s=ωp0 T−III 1 þ s=ωp1;2 T−III

The transfer function for the designed classical Type‐III controller of boost DC‐DC converter given by the function as
3:003 × 106 ðs þ 605Þ2
.
sðs2 þ 1:31 × 105 s þ 4:26 × 109 Þ

5 | R E S U L T S AN D D I S C U S S I O N

The nonlinear phenomena of DC‐DC boost converter have been investigated both by the simulation and experiment.
The nonlinear dynamics examined by phase portraits and time‐domain waveforms due to the fact that these are popular
laboratory techniques to study the theoretical and practical chaos.

FIGURE 3 A, Basic structure and B,

bode diagram of Type‐III controller
8 of 17 BANERJEE ET AL.

TABLE 1 List of circuit parameters

Circuit Components Values

Input voltage (Vin) 5V

Output voltage (Vo) 12 V
Output capacitance (C) 220 μF
Inductance (L) 20 mH
Load resistance (RL) 25 Ω
Switching frequency ( f sw) 25 kHz

5.1 | Numerical simulation results

The VMC boost converter is numerically computed in Matrix Laboratory (MATLAB) Simulink. The computational pro-
cess is carried out cycle‐by‐cycle by solving the differential equations of switching states. The parameters of switching
circuit are given in Table 1. To study and analysis the nonlinear phenomena of the converter, the basic circuit param-
eters (like load resistance, input voltage, etc) have been varied by keeping other parameters fixed and observed the con-
verter's dynamics in the phase plane. In this study, the load resistance (RL) considered as a primary bifurcation
parameter. RL is varied from a certain range, keeping the other parameters fixed and observed the different periodicities
of the limit cycles in the phase plane.
In condition, when the load resistance (RL) equals to 25.2 Ω, the converter works in stable (Period‐I) operation. To
see the converter dynamics clearly, the sampled values of state variables like inductor current (iL) and output capacitor
voltage (vc) fetched at the start of each switching cycle in steady state. The time‐domain waveform and phase portrait of
fundamental periodicity (Period‐I) of inductor current are decorated in Figure 4A. The two state variables like vc and iL
have taken the axis of the phase portrait. It has been observed from the phase portrait that the periodicity of the limit
cycle is one in phase plane, and this kind of dynamics technically known as fundamental periodic operation ie, Period‐I
operation.
In condition, where RL is gradually increased to 35.5 Ω, the converter loses the stable operation and leads to slow
scale instability as given by Figure 4B. The quasiperiodic orbit observed in this situation, the trajectory moves on the
surface of a torus. The motion is associated with a finite number of frequencies, related to one another by irrational
ratios, and the motion appears “almost periodic” but is not exactly periodic. This kind of instability manifests itself as
Hopf bifurcation, whereby a stable fixed‐point change to a limit cycle as a certain parameter (eg, load resistance) is
changed. As the bifurcation variable (parameter) remains to change, the system avows another periodicity not in a ratio-
nal ratio to that of the first limit cycle, and the resulting system dynamics are quasiperiodic by nature. This slow‐scale
instability causes low frequency oscillation in practical power supply. It is important to note that the chaotic dynamics
observed at RL equals to 50.8 Ω (Figure 4C). At RL = 50.8 Ω, the periodicities of the phase portrait (limit cycle) turned
into infinite (practically periodicity is more than 10), so that the chaotic dynamics observed in phase portrait. The
chaotic orbits of the converter are attaining a bounded aperiodic oscillation within a definite zone in the phase plane
(ie, phase portrait). In chaotic dynamics, the same state never repeats, in every loop of phase plane, the state traverses
through a new trajectory and the resultant attractor known as a strange attractor,22 and the periodicity of the chaotic
limit cycle is infinite. The converter dynamics observed here almost on the verge of the DCM ie, the lower threshold
of the inductor current is almost 0.
In the condition, RL is increased further to 55 Ω, the converter enters into DCM of operation. The phase portrait,
respective time domains waveform shown by Figure 4D. Similarly, the chaotic operation in DCM observed at RL equals
to 70 Ω (Figure 4E). These slow‐scale instabilities and chaotic behavior may deteriorate the dynamic performance of the
closed‐loop converter. It can conclude that the fundamental periodic operation of the converter dynamics is preferred as
an operating zone for designing practical power supplies.

5.2 | Bifurcation diagrams and discussion of boundary regions of bifurcation parameters

Bifurcation is a mathematical study, and the system dynamics are observed by qualitative variation of system parame-
ters. A small smooth change made to the values of bifurcation parameters (ie, load resistance, inductance, output
BANERJEE ET AL. 9 of 17

FIGURE 4 Complex dynamics of DC‐

DC switching boost converter in voltage
mode control: A, Period‐I orbit at
RL = 25.2 Ω and respective time‐domain
waveform of inductor current, B, quasi‐
period orbit at RL = 35.5 Ω and respective
time‐domain waveform of inductor
current, C, chaotic operation at RL = 50.8
Ω and respective time‐domain waveform
of inductor current, D, Period‐I orbit in
discontinuous conduction mode at RL = 55
Ω and respective time‐domain waveform
of inductor current in Period‐I operation,
E, chaotic operation in discontinuous
mode at RL = 70 Ω and respective time‐
domain waveform of inductor current in
chaotic operation

capacitance, and gain of Type‐III controller) that may cause qualitative changes (bifurcation) of the system. The con-
verter driven by external clock pulse so that the system works like a nonautonomous system, and the clock frequency
dictates the switching frequency. Noted, the time period of clock pulse must be lesser than the load time constant RLC
10 of 17 BANERJEE ET AL.

and the state variables sampled here at the clock frequency. Numerical simulation of current mode boost converter
circuit carried out by using FORTRAN programming (Microsoft developer studio Fortran Power Station 4.0), and bifur-
cation diagrams plotted by using Origin software (Figure 5).

5.2.1 | Characteristic multipliers

In ordinary differential equations (ODE), a characteristic multiplier is an eigenvalue of a monodromy matrix and the
logarithm of a characteristic multiplier as characteristic exponent. They appear in Floquet theory of periodic differential
operators and in the Frobenius method. The characteristic multipliers of an iterative function f (.) are the roots, λ, of the
characteristic equation

detðλI − J F ðX Q ÞÞ ¼ 0 (25)

where, J F (XQ) is the Jacobian matrix of f (.) calculated at the fixed point XQ, and I is the identity matrix. In the case
boost converter, the function f (.) can be written in the following form:

"  #
f 1 vc;n ; iL;n ; κ
X nþ1 ¼ f ðX n ; κ Þ ¼  (26)
f 2 vc;n ; iL;n ; κ

From Equation 22, the expression written as

f 1 ð:Þ ¼ α11 ðdn Þvc;n þ α12 ðdn ÞiL;n þ β1 ðdn ÞV in (27)

f 2 ð:Þ ¼ α21 ðdn Þvc;n þ α22 ðdn ÞiL;n þ β2 ðdn ÞV in (28)

where, dn is a function of bifurcation parameter (κ).

FIGURE 5 Bifurcation diagrams of voltage mood controlled (VMC) boost converter for A, load resistance, B, inductance, C, output
capacitance, and D, Type‐III controller gain
BANERJEE ET AL. 11 of 17

Now, the Jacobian matrix J F (XQ) is


∂f 1 ð:Þ ∂f 1 ð:Þ∂f 2 ð:Þ∂f 2 ð:Þ
J F ðX Q Þ ¼ (29)
∂vc;n ∂iL;n ∂vc;n ∂iL;n X n ¼X Q

where,
8
> ∂f 1 ð:Þ 0 ddn 0 ddn 0 ddn
>
> ¼ f 11 ðdn Þ þ vc;n f 11 ðdn Þ þ iL;n f 21 ðdn Þ þ V in β1 ðdn Þ
>
> ∂v dv dv dv
>
>
c;n c;n c;n c;n
> ∂f ð:Þ
>
>
> 0 dd n 0 dd n 0 dd n
>
< ∂i
1
¼ vc;n f 11 ðdn Þ þ f 12 ðdn Þ þ iL;n f 12 ðdn Þ þ V in β1 ðdn Þ
L;n di L;n diL;n di L;n
> ∂f ð : Þ
> 2 ¼ f ðd Þ þ v f ðd Þ 0 dd 0 dd 0 dd
>
>
n
þ iL;n f 22 ðdn Þ
n
þ V in β2 ðdn Þ
n
> ∂v
> c;n 21 n c;n 21 n
dv dv dv
>
> c;n c;n c;n
>
>
>
> ∂f ð : Þ 0 dd 0 dd 0 dd
: 2 ¼ vc;n f 21 ðdn Þ n þ f 22 ðdn Þ þ iL;n f 22 ðdn Þ n þ V in β2 ðdn Þ n
∂iL;n diL;n diL;n diL;n

Now, the characteristic multipliers calculated at the fixed point for any bifurcation parameter (by keeping other
parameter fixed) by the above‐described equations.

5.2.2 | Load resistance as bifurcation parameter

Figure 5A shows the bifurcation diagram for taking load resistance as a bifurcation parameter in VMC of boost con-
verter. In the diagram, change of load resistance considered in X‐axis and inductor current sampled at a clock frequency
(stroboscopic sampling) taken in Y‐axis. The load resistance is varied from 1 to 50 Ω with a step of 0.1 Ω while other
parameters are kept fixed.

Stable Period‐I operation

The values of the characteristic multipliers are less than unity for RL < 25.2 Ω. So stable fundamental periodicity or
Period‐I orbit is observed in 25.2 Ω.

Stable Period‐II operation

The iterative function f (.) is unstable for RL > 25.3 Ω. But the function f ( f (.)) may be considered stable. So the period
has been doubled. Here, the fixed point of f ( f (.)) actually contains two alternate fixed points of f (.) and exhibits up to
RL ≈ 25.7 Ω.

Hopf bifurcation
When RL ≈ 25.8 Ω, the iterative function f ( f (.)) is becoming unstable, and the stable iterative function is f ( f ( f
( f (.)))). This type of bifurcation characterized by a sudden expansion of a stable fixed point to a stable limit cycle.22 Sys-
tems that exhibit this bifurcation normalized to second‐order equation, where the system has a stable fixed point, which
is associated with a pair of complex eigenvalues, which have negative real parts. By varying the parameter, the real parts
become positive, and the complex eigenvalues move across the imaginary axis. Thus, the fixed point loses stability, and
the system has followed a stable limit cycle.

Routes to chaos via a Hopf bifurcation

By gradually increasing the value of RL, the system has entered in the chaotic region, and chaotic nature is perceived in
the complete remaining zone. One can be noted that the system is undergone Period‐I to Period‐infinity through Hopf
bifurcation.

5.2.3 | Inductance as bifurcation parameter

The diagram with inductance as a bifurcation parameter is exhibited in Figure 5B. The inductance (L) is varying from 1
to 200 mH with step of 0.1 mH by keeping other parameters fixed. The system dynamics started with Period‐I, the
12 of 17 BANERJEE ET AL.

behavior is observed from 1 to 19.74 mH, and Period‐II started at 19.75 mH from two distinguished zones. The Period‐II,
behavior last up to 29.24 mH bifurcates to Period‐IV at 29.25 mH, then enters into chaos at 31.5 mH. This chaotic zone
observed up to 49 mH and after that, the Period‐III behavior sustains up to 58 mH. Then, Period‐III has bifurcated to
Period‐VI and finally entered to chaotic zone at 71.25 mH and continues up to 200 mH.

5.2.4 | Output capacitance as bifurcation parameter

The capacitance is changed for 1 to 500 μF with a step of 0.1 μF by keeping the other parameter fixed (Figure 5C). Initial
Period‐I dynamics are observed, and Period‐I is bifurcated to Period‐II at output capacitance (C) equals to 28 μF, which
sustains up to the 152 μF and bifurcates to Period‐IV. Period‐III is observed at 188 μF after the Period‐IV, and this
Period‐III is bifurcated to Period‐VI at 416 μF and entered to chaos at 469 μF.

5.2.5 | Type‐III controller gain as bifurcation parameter

The main objective of this section is to study the nonlinear dynamics by varying the Type‐III controller gain by keeping
other converter parameter fixed.
From Figure 5D, it is observed that converter dynamics changed from fundamental periodicity to chaotic periodic-
ities through a Hopf bifurcation by gradually increasing the gain values of the Type‐III controller. Here, controller gain
varied from 0.1 × 107 to 2.6 × 107 with a step of 0.001 × 107 by keeping other converter parameters fixed. Initially, the
converter dynamics are started with Period‐I operation, and this Period‐I behavior is observed up to 0.29 × 107. After,
the slow‐scale instabilities noticed at 0.31 × 107, and these periodicities are bifurcated by gradually increasing the con-
troller gain. Finally, the converter dynamics have shown chaotic dynamics at = 0.59 × 107 and sustain up to 2.6 × 107.
Figure 6 shows the stability of the system studied by deriving the eigenvalues of the system at the equilibrium point.
The system has one negative real eigenvalue and a pair of complex poles. The real part of the complex pole may be either
positive or negative real, depending upon the values of Type‐III controller's gain. Table 2 shows the variation of the eigen-
values for various values of controller gain and the locus of the complex eigenvalues and shown in Figure 6. From the
nature of the characteristic multipliers (ie, eigenvalues) Hopf bifurcation confirmed. The movement of the locus of the left
plane to the right plane shows that the system losses its stability when a gain of the controller is increased.

5.3 | Methodology and experimental implementation VMC boost converter by using

dSPACE controller

In order to study the practical nonlinear phenomena of the VMC boost converter, a laboratory scale prototype has been
fabricated and implemented in real‐time platform using dSPACE controller. The circuit parameters of the converter are

FIGURE 6 Locus of the complex eigenvalue pair for different gains of Type‐III controller
BANERJEE ET AL. 13 of 17

TABLE 2 The nature of characteristic multipliers

Sl. Number Controller Gain Characteristics Multipliers (Eigenvalues) Remarks

1. 0.1 × 107 −0.0137, −0.0159 ± j0.2318 Stable

2. 0.6 × 10 7
−0.0199, −0.0097 ± j0.2313 Stable
3. 1.1 × 107 −0.0221, −0.0054 ± j0.2310 Stable
4. 1.6 × 10 7
−0.0233, −0.0020 ± j0.2306 Stable
5. 2.1 × 10 7
−0.0240, 0.0008 ± j0.2303 Unstable
6. 2.6 × 10 7
−0.0246, 0.0034 ± j0.2301 Unstable

already given in Table 1 for investigation. Overall, experimental setup and schematic circuit diagram are given in
Figure 7A and Figure 7B, respectively.
The experimental implementation of the closed‐loop converter is carried out by using dSPACE controlled real‐time
platform because of much improved flexibility, reduced design time, better programmability, elimination of discrete

FIGURE 7 A, Experimental setup, B, schematic circuit diagram for voltage mood controlled (VMC) closed‐loop converter
14 of 17 BANERJEE ET AL.

tuning components, improved system reliability, easier system integration, and possibility to include various perfor-
mance enhancements. Some other advantages of it are less susceptible to aging and environmental variations, less sen-
sitive to noise, changing a controller does not require an alteration in the hardware, and provide improved sensitivity to
parameter variations.
The dSPACE DS1104 is a controller board installed in the PCI slot of the PC containing two processors. The main
processor is an MPC8240 PowerPC with a clock speed of 250 MHz and 32 kB internal cache memory. It acts as the mas-
ter processor with TMS320F240 DSP as the slave, containing 4 K Word of the dual port RAM. LEM hall‐effect voltage
transducer (LV‐25P) is used to sense the output voltage of the converter. Then, the scaled and filtered signal is fed to the
ADC port of dSPACE controller. The output of voltage sensor is noisy and filtered by a low pass filter with a cutoff fre-
quency of 5 kHz. This filter cutoff frequency is less than the switching frequency (25 kHz) of the converter. The filtered
output voltage is limited to a maximum of 4.7 V by placing two zener diodes in back to back fashion. The complete
control system performed in dSPACE‐based real time interface platform. The reference input is compared with the dig-
ital output voltage coming from the ADC port of dSPACE. After comparing two signals, an error signal created, and this
error signal passed through the Type‐III controller, and control signal is obtained. Now, this control signal is compared
with a high‐frequency (similar to switching frequency) triangular waveform for generating a PWM signal within
dSPACE environment. PWM output from dSPACE is passed through the DAC, Opto‐isolator (MCT2E) and Astable
multivibrator (NE555) circuit before inputting to the gate of the MOSFET (STB55N) switch of the converter. The output
of opto‐isolator is not sufficient to drive the gate of the MOSFET switch and fed to an inverting buffer circuit using
555‐IC. The output of 555 inverting buffer connected to the gate of the MOSFET through 100 Ω resistances.

5.4 | Experimental results and discussion

The phase portraits are a useful method to identify the different periodic orbit to chaotic behavior. A DSO (Agilent
Technologies DSO5014A) used in X‐Y mode to capture the phase portraits at certain instants to study the different peri-
odic and chaotic orbits. In this study, the output capacitor voltage is captured in Ch‐1 (X‐axis) and inductor current in
Ch‐2 (Y‐axis) and measured by a FLUKE made current probe.

5.4.1 | Phase portraits of fundamental orbit and limit cycle

The experimental result of fundamental periodicity (Period‐I) for the converter is observed in Figure 8A, and the respec-
tive time domain waveform of inductor current is observed in Figure 8A. Noted that the fundamental periodic operation
observed for load resistance (RL) equal to 21.7 Ω. From the time domain waveform, observed that the waveforms repeat
after one clock cycle with time, and the Period‐I operation is known as fundamental periodic operation. RL increased,
the many other possible limit cycles observed due to the occurrence of slow‐scale instabilities. At RL = 32.3 Ω, the limit
cycle of Period‐III observed in phase portrait by keeping other parameters remain same (Figure 8B). The respective time
domain waveform of this limit cycle has also been illustrated in Figure 8B.

5.5 | Phase portrait of chaotic orbit

In chaotic mode operation, the system dynamics are attained a bounded aperiodic oscillation within a definite zone in
phase portrait. In chaotic dynamics, the same state never repeats, in every loop of phase plane, the state traverses
through a new trajectory. Periodicities of limit cycle are infinite, and in the time domain, the waveform of the state var-
iable repeats after the time‐domain infinite clock cycle (practically above 10 clock cycles). Such situation occurs in an
electronic circuit; the system undergoes apparently random oscillations. In this work for the value of load resistance
of 43.6 Ω (keeping other variables remains constant), the above‐mentioned phenomenon occurs. Figure 9 illustrated
the phase portrait of experimental chaotic orbits and the respective time domain waveform of VMC of the converter.
Observed that the periodicities of the converter dynamics are high, and the converter works almost on the verge of
DCM operation.
Henceforth, the sequel of this study concluded that the choice of parameters and their values play a key role to
determine the dynamics of the DC‐DC boost converter, useful for designing the practical power supply.
BANERJEE ET AL. 15 of 17

FIGURE 8 A, Phase portrait of Period‐I operate at RL = 21.7 Ω and respective time‐domain waveform and FFT analysis of inductor
current (Ch‐2) and B, phase portrait of limit cycle at RL = 32.3 Ω and respective time‐domain waveform and FFT analysis of inductor
current (Ch‐2)

FIGURE 9 Phase portraits of chaotic operation at RL = 43.6 Ω and respective waveform and FFT analysis of inductor current (Ch‐2)

6 | CONCLUSION

The study of complex dynamics of a voltage mode controlled Type‐III controller‐based boost converter circuit is
observed by using a dSPACE‐based real time controller, and it is firstly reported in this manuscript. The mathematical
analysis for studying complex dynamics of this switching converter is performed by using discrete modeling. The
16 of 17 BANERJEE ET AL.

computational results observed by solving the differential equations of the switching converter, and finally, the nature
of the converter's complex dynamics verified by experimental results. Presented results suit the designers easily to find
out the probable operating zone and route to design a chaos free power supply. The slow‐scale instability observed
throughout the study of parametric variation of the converter. It is also seen that as the control parameters varied,
the nominal periodic orbit undergoes a Hopf bifurcation, quasiperiodic, and finally enters into chaotic regime. The
zones of the practical chaos, quasiperiodic oscillation easily identified from the experimental results. It is to be noted
that chaotic operations should not be accepted in the practical design of power supplies due to nonlinear oscillation,
Electromagnetic Interference (EMI) problem, etc. For designing of the power supply, one should avoid these types of
complexities in their final products, and the fundamental periodic behavior of system dynamics should desire the better
stability in practical applications.
The designers should try to make “chaos free” power supplies for sophisticated applications. Hence, the nonlinear
dynamics in switching converter are important for finding the zones of instability and useful for the design of practical
power supplies. The main reason for carrying out this work is to identify and explore the complex phenomena in power
electronic converters. This investigation, aimed to the nonlinear dynamics of power converters under certain operating
conditions to make them suitable for practical applications.
Although the work is well explained and impressive, however advanced mathematical techniques (like Filippov
method) may be used for finding better stability boundary and more accurate results for studying the dynamical behav-
iour of the converter. The use of dSPACE controller has some inherent problems like system instability due to conver-
sion of continuous data to digital data, signal delay for A/D to D/A conversion, and vice versa, unwanted loss of signal
information due to reconstruction of signal and limitation in operating frequency.

ORCID
Arnab Ghosh https://orcid.org/0000-0002-2223-433X

R EF E RE N C E S
1. Chan WC, Tse CK. Study of bifurcations in current‐programmed DC‐DC boost converters: from quasiperiodicity to period‐doubling. IEEE
Trans Circuits Syst I: Fundam Theory Appl. 1997;44(12):1129‐1142.
2. Banerjee S, Chakrabarty K. Nonlinear modeling and bifurcations in the boost converter. IEEE Trans Power Electron. 1998;13(2):252‐260.
3. El Aroudi A, Benadero L, Toribio E, Olivar G. Hopf bifurcation and chaos from torus breakdown in a PWM voltage‐controlled DC‐DC
boost converter. IEEE Trans Circuits Syst I: Fundam Theory Appl. 1999;46(11):1374‐1382.
4. El Aroudi A, Leyva R. Quasi‐periodic route to chaos in a PWM voltage‐controlled DC‐DC boost converter. IEEE Trans Circuits Syst I:
Fundam Theory Appl. 2001;48(8):967‐978.
5. Banerjee S, Verghese GC. Nonlinear Phenomena in Power Electronics: Bifurcations, Chaos, Control, and Applications. Wiley‐IEEE
Press; 2001.
6. Tsc CK. Complex Behavior of Switching Power Converters. New York: CRC Press; 2003.
7. Iu HHC, Tse CK. Study of low‐frequency bifurcation phenomena of a parallel‐connected boost converter system via simple averaged
models. IEEE Trans Circuits Syst I: Fundam Theory Appl. 2003;50(5):679‐685.
8. Ghosh A. Nonlinear dynamics of power‐factor‐corrected AC‐DC boost regulator: power converter, nonlinear phenomena, controlling the non-
linearity. LAP Lambert Academic Publishing; 2012.
9. Ghosh A, Banerjee S, Saha PK, Panda GK. Nonlinear modeling and bifurcations in switched power‐factor‐correction boost regulator.
IEEE Int Conf Circuits, Power Comput Technol. 2013;2013(ICCPCT‐2013):517‐522.
10. Ghosh A, Banerjee S, Basak S, Chakraborty C. A study of chaos and bifurcation of a current mode controlled flyback converter. IEEE 23rd
Int Symp Ind Electron. 2014;2014(ISIE‐2014):392‐397.
11. Arjun M, Patil V. Steady state and averaged state space modelling of non‐ideal boost converter. Int J Power Electron. 2015;7(1‐2):109‐133.
12. Zhioua M, El Aroudi A, Belghith S, et al. Modeling, dynamics, bifurcation behavior and stability analysis of a DC–DC boost converter in
photovoltaic systems. Int J Bifurcation Chaos. 2016;26(10):1650166.
13. Ghosh A, Prakash M, Pradhan S, Banerjee S. A Comparison Among PID, Sliding Mode and Internal Model Control for a Buck Converter,
40th IEEE Annual Conference of the Industrial Electronics Society 2014 (IECON 2014). Dallas, TX, U.S.A.; 2014:1001‐1006, 29th October –
1st November.
14. Ghosh A, Banerjee S. Study on chaos and bifurcation in DC‐DC flyback converter. Int J Ind Electron Drives. 2017;3(3):161‐174.
BANERJEE ET AL. 17 of 17

15. Ghosh A, Banerjee S. Study of complex dynamics of DC‐DC buck converter. Int J Power Electron. 2017;8(4):323‐348.
16. Ghosh A, Banerjee S. Control of switched‐mode boost converter by using classical and optimized type controllers. J Control Eng Appl
Informatics (CEAI). 2015;17(4):114‐125.
17. Ghosh A, Banerjee S, Sarkar MK, Dutta P. Design and implementation of type‐II and type‐III controller for DC‐DC switched‐mode boost
converter by using K‐factor approach and optimization techniques. IET Power Electron. 2016;9(5):938‐950.
18. Banerjee S, Ghosh A, Rana N. An improved interleaved boost converter with PSO based optimal type‐III controller. IEEE J Emerging Sel
Top Power Electron. 2017;5(1):323‐337.
19. Rana N, Ghosh A, Banerjee S. A comparative closed‐loop performances of a DC‐DC switched‐mode boost converter with classical and
PSO based optimized type‐II/III controllers. Int J Power Electron. 2017.
20. Rana N, Ghosh A, Banerjee S. Development of an improved tristate Buck–boost converter with optimized Type‐3 controller. IEEE J
Emerging Sel Top Power Electron. 2018;6(1):400‐415.
21. Rana N, Kumar M, Ghosh A, Banerjee S. A novel interleaved tri‐state boost converter with lower ripple and improved dynamic response.
IEEE Trans Ind Electron. 2018;65(7):5456‐5465.
22. Dranga O, Tse CK, Iu HH, Nagy I. Bifurcation behavior of a power‐factor‐correction boost converter. Int J Bifurcation Chaos.
2003;13(10):3107‐3114.
23. Shalchi Alishah R, Barzegar M, Nazarpour D. A new cascade boost inverter for photovoltaic applications with minimum number of
elements. Int Trans Electr Energy Syst. 2015;25(7):1241‐1256.
24. Wang JM, Chien HC, Wu ST, Yen SC, Lin JY. Analysis and design of a boost PFC converter with sample and hold control techniques. Int
Trans Electr Energy Syst. 2015;25(11):3122‐3138.
25. Zamani N, Ataei M, Niroomand M. Analysis and control of chaotic behavior in boost converter by ramp compensation based on
Lyapunov exponents assignment: theoretical and experimental investigation. Chaos, Solitons Fractals. 2015;81:20‐29.
26. Mohammadzadeh Shahir F, Babaei E, Sabahi M, Laali S. A new DC–DC converter based on voltage lift technique. Int Trans Electr Energy
Syst. 2016;26(6):1260‐1286.
27. Sakthivel R, Santra S, Anthoni SM, Kuppili V. Synchronisation and anti‐synchronisation of chaotic systems with application to DC–DC
boost converter. IET Gener Transm Distrib. 2017;11(4):959‐967.
28. Cheng L, Ki WH, Yang F, Mok PK, Jing X. Predicting subharmonic oscillation of voltage‐mode switching converters using a circuit‐
oriented geometrical approach. IEEE Trans Circuits Syst I: Reg Papers. 2017;64(3):717‐730.
29. Zhang H, Li W, Ding H, Luo P, Wan X, Hu W. Nonlinear modal analysis of transient behavior in cascade DC–DC boost converters. Int J
Bifurcation Chaos. 2017;27(9):1750140.
30. Zhang R, Wu A, Zhang S, Wang Z, Cang S. Dynamical analysis and circuit implementation of a DC/DC single‐stage boost converter with
memristance load. Nonlinear Dyn. 2018;1‐15.
31. Tekpeti BS, Kang X, Kheshti M, Jiao Z. Modeling and fault analysis of solar photovoltaic grid connected systems under solar radiation
fluctuation consideration. Int Trans Electr Energy Syst. 2018;28(8):e2576.

How to cite this article: Banerjee S, Ghosh A, Padmanaban S. Modeling and analysis of complex dynamics for
dSPACE controlled closed‐loop DC‐DC boost converter. Int Trans Electr Energ Syst. 2019;e2813. https://doi.org/
10.1002/etep.2813