Transient and Steady-State Analysis of a SEPIC

Converter by an Average State-Space Modelling

T. Polsky, Y. Horen, S. Bronshtein, and D. Baimel
Sami Shamoon Engineering College, Beer Sheva/Ashdod, Israel
Phone: +972-8-647-5776, Fax: +972-8-647-5703, Email: [email protected]

Abstract- This study presents a simple and effective methodology transient behavior of SEPIC converter is completed, but the
for analyzing transient and steady state behavior of a Single results are too complicated for practical engineer to use.
Ended Primary Inductance Converter (SEPIC), operating in a Simple and very effective average circuit modelling for
continuous current mode (CCM). The proposed technique uses SEPIC converter is demonstrated in [8] using SPICE
an average state-space modelling approach including power
simulation software [9]. The model presented in this work
losses in converter elements. This approach makes possible
obtaining analytical solutions for the converter currents and shows excellent agreement for steady state (DC), large signal
voltages and allows examining different characteristics of (transient), and small signal (AC) responses. This behavioral
converter behavior, e.g. efficiency, output to input voltage ratio, average model is very convenient tool for the analysis and
input current, transient times, and elements stresses. This design of SEPIC converters.
methodology is valuable for keeping the converter design in Simultaneously several studies were carried out for other
compliance with design requirements, including dynamics, and fourth-order converters, Cuk and Zeta. Both of them are
reduces the need for accurate time domain simulations. This capable to operate in step up or step down modes [10]-[11].
approach can be useful for analyzing other types of high order The state-space averaging technique (SSA) [12] is applied in
converters.
these studies to find small-signal linear dynamic model of the
converter and its various transfer functions. This technique
I. INTRODUCTION operates with matrix and naturally, MATLAB is the most
suitable software for models implementation. As far, number
Single Ended Primary Inductance Converter (SEPIC) is a of studies used average model technique for obtaining transfer
switch mode converter with non-inverting topology, providing function for realizing PWM control feedback. This approach is
a wide range of output voltages that can be higher as well as not always effective, and using a non-linear control strategy
lower than the input voltage. The input current of SEPIC in might be superior solution. For choosing and applying an
steady state operation is continuous and nearly constant. These optimal control strategy, a simple analytical model of the
features make SEPIC very attractive for power conversion converter should be developed and examined in all operational
applications such as photovoltaic (PV) systems, fuel cells, modes of the converter.
battery charges, vehicle auxiliary supply, and more. Typically, In this study following the ideas of [2], [8], and [11], a
a duty cycle control is applied to a converter in order to reach comprehensive analytical model for a high order converter
a desirable output voltage, or a desirable source current, within (SEPIC) is developed. This model is built as a closed form
wide variations of source voltages and load currents. However, state-space equations that include power losses in converter
SEPIC is a fourth order time-variant converter with two elements. It gives analytical solutions to a steady state as well
switching elements – a transistor and a diode. High order as a transient behavior of the converter. It allows getting
converters have complicated mathematics for converter different characteristics of the converter's behavior, like
dynamics. Therefore, the design of such a converter and its efficiency, output voltage, input current, transient times, and
control circuit often results in a cut-and-try approach, based on stress parameters of converter elements. It can be valuable for
large numbers of simulations. This is a time-consuming practical engineers for designing and examination of a
approach, which requires running many simulations for each compliance with design requirements including dynamics.
parameter change; frequently, not ensuring optimal design Additional application of this model is to assist with a
outcomes. In order to guarantee the most suitable design and controller design.
to choose the best control strategy for better dynamics, it is In order to validate the proposed average model, theoretical
essential to build a simple and effective model. Developing results of this study were compared with simulations. A very
such model was the purpose of this study. good agreement was found between simulation results
A general method for modelling basic (single-switch) PWM obtained by Simulink/MATLAB, and a proposed average
converters, via average state-space approach, has been model.
developed in 1970s. The main idea was “in replacement of the This paper is organized as follows: chapter II is destined to
state-space descriptions of the two switched networks by their the development of the average model of SEPIC, chapter III
average over the single switching period, which results in a shows simulations results in comparison with the obtained
single continuous state-space equation” [1]. Further studies average model, chapter IV discusses possible implementations
generalized this idea for higher order converters with more and benefits of the proposed model, and finally, in chapter V
switches [2]-[4]. In works [5]-[7] a comprehensive analysis of conclusions are summarized.

978-1-5386-4198-9/18/$31.00 ©2018 IEEE

978-1-5386-4198-9/18/$31.00 ©2018 IEEE 211

 II. DEVELOPING OF THE AVERAGE MODEL OF A SEPIC State a:
SEPIC is a switched-mode power converter with two Aa
 
switching elements. Fig.1 shows SEPIC topology. We assume  −1 
that the switching elements in the circuit (diode and transistor)  0 0 0
C1  B aU

   0 
are operating synchronously. That means that the diode  −1   v1   
changes its states in matching to position changing of the  0 (R + R )C 0 0    0 
  VC  +  1  U
.
xa = 
C
switch. Reverse biased diode is modelled by an open circuit  − R1   i1   
and forward biased diode by a constant voltage source VD.  0 0
L1
0     L1 
Transistor is supposed to be an ideal switch. Parasitic    i2   0 
 
 1 − ( RC 1 + R2 )  (4)
resistances R1, RC1, R2, and RC display conduction losses. Fig.  0 0 
1 shows all relevant voltages and currents, and positions of the  L2 L2 
switch (transistor) and the diode: state 'a' - switch ON, diode
OFF, and state 'b' - switch OFF, diode ON. Ca
  v1 
+ v C1
_  
 R  V
R1 L1 RC1 C1 vD _ Va =  0 0 0 C 
  i1 
+ R + RC

 
i1  i2 
a b R2 RC
+
+ R V
State b:
U
i2 C VC _
L2 _  Ab
 
 1 
0 0
C1
0   B
bU

   0 
Fig. 1. SEPIC topology  −1 R R   
 0 ( R+ R ) C ( R + R ) C ( R + R ) C  v1   0 
For each switching position the system is linear, therefore  C C C 
V  

 C  +U −vD 
.
it is defined by state-space equations in the standard form: xb =  RCR 
= +  − R1 + RC1 +   i1   L1 
(1)  −1 −R  R+ RC  RCR  i   −v 
= + −
 L1 ( R+ RC ) L1 L1 L1 ( R+ RC )  2   D 
where A, B, c, d are matrixes/vectors of circuit parameters, x    L2 
is vector of state-space parameters, u is input signal (input 0 −R −RRC −RRC R2 
 − (5)
voltage - U), and y is an output signal (output voltage). State  ( R+ RC ) L2 ( R+ RC) L2 L2 ( R+ RC ) L2 
variables for SEPIC as marked on Fig. 1 are: vC1, VC, - voltages
on C1 and C, i1, i2 –currents through L1 and L2.
We assume that converter operates in continuous current  Cb
   v1 
 
mode (CCM) and operation of its two switching elements  R RRC RRC VC 
Vb = 0  i 
(transistor and diode) is matched (transistor ON – diode OFF  R+ RC R+ RC R+ RC  1
 
and vice versa). We further assume that converter's switching  i2 
frequency is high compared to natural frequencies of circuit,
and an input signal U and a duty cycle are constant. These Steady-state xSS and transient x(t) solutions of average model
assumptions allow us to define an equivalent average system (2) will be as follows:
by average state-space model:
= + =− (6)
(2)
= + ( )= + ( − ) (7)
where AT, BT, cT and dT are average matrixes/vectors that = (8)
calculated by (3):
( )= ( ) (9)
=
This is the desired time domain analytical solution, which is
= valid for transient as well as steady state operation. This
(3) solution is linear with respect to input voltage, but it is not
= linear with respect to time ratio α. When converter works with
= PWM controller, α is the controller's output signal and the
converter's input parameter. As a result, we deal with non-
where Aa, Ab, Ba, Bb, ca, cb, and da, db are state-space linear system.
matrixes/vectors for each switch position, and α is a time ratio From the matrix solution (6), using MATLAB symbolic
defined as: = . These state-space matrixes/vectors are: library we obtained an explicit equations for SEPIC variables:

212
 AN 1α 2 + BN 1α (10) III. AVERAGE MODEL VERIFICATION
xss1 = VC1 =
ADα 2 + BDα + CD
Model verification was carried out first, by comparison
AN 2 ⋅ α 2 + BN 2 ⋅ α (11) between analytical results of average model solutions (6), (7)
xss 2 = Vc =
ADα 2 + BDα + CD and simulations of an electric circuit on Fig.1, and then, by
BN 3α comparison between simulations of an average model (2) and
xss 3 = I1 = (12)
ADα + BDα + CD
2 simulations of an electric circuit.
Both sets of a model verification has been accomplished in a
A α 2 + BN 4α (13) transient and at a steady state. The first verification set was
xss 4 = I 2 = N 42
ADα + BDα + CD implemented for a constant time ratio α, while the second set
where: of simulations has been performed for α varying in time. For
varying time ratio, analytical solution does not exist; therefore,
AD = RR2 + RRC + R2 RC + R 2 model verification has been performed by comparison between
BD = RRC + RRC1 + RC RC1 simulations of an average model (2) and simulations of an
electric circuit.
CD = ( R + RC ) R1 Simulations of electric circuit were carried out by MATLAB
AN 1 = ( R 2 + RR2 + RRC + R2 RC ) U − ( R + RC ) R2Vd Simulink software. Fig. 2 shows simulation circuit. Converter
parameters were as follows: R1=0.1Ω, L1=330μH, R2=0.1Ω,
BN 1 = ( ( R2 + RC + RC1 ) R + R2 RC + RC RC1 )U + ( R + RC ) RV
1 d
L2=330μH, RC1=0.05Ω, C1=820μF, RC=0.5Ω, C=1mF,
AN 2 = − ( R + RC ) RVd U=12V, f=18kHz. The load was adjusted for obtaining of the
output power of 100W.
BN 2 = ( R + RC ) RU
The results of the model verification are as follows:
BN 3 = ( R + RC )(U − Vd )
• For the constant time ratio α: Fig. 3 presents simulations
AN 4 = − ( R + RC ) Vd waveforms of state variables and theoretical lines
BN 4 = ( R + RC )U obtained by analytical solutions (6), (7) from the average
converter model. Simulations and calculations on Fig. 3
Index N is for nominator and D for denominator. were performed for two cases: output voltage Vout=20V
These equations clarify the effect of each circuit parameter (higher than input voltage), (a), and output voltage
on the converter behavior in steady state operation and assist Vout=5V (lower than input voltage), (b). The waveforms
in open loop as well as in feedback design. For example, xss(α) with ripple are simulation results, and continuous lines in
is the operational line, which defines possible solutions for the middle are the analytical solutions according to the
converters' steady state without dependence on a control average model. Fig. 4 shows zoomed variables in steady
procedure. Control strategy can change the dynamics of state region. It can be seen from Figs. 3-4, that analytical
arriving to the steady state but after that, at steady state control solutions are always positioned within the ripple range
method does not significant. even in a transient stage.

+v 1
-
C1 Voltage
2
L1 Current
+ + +
RL1 L1 R C1 C1
Diode

i- +
+
+ R
+ RC
g 1 RL2 C Voltage
Pulse
Generator Ideal +v 3
U -
Switch C
L2
2
4
i+ L2 Current
-

Fig. 2. . MATLAB/SIMULINK Simulation circuit of SEPIC

213
 (a) (a)

(b)

(b)
Fig. 3. Simulation (with ripple) and average model solutions (solid lines (c)
inside) waveforms of SEPIC variables, (a) – Vout=20V (higher than Vin),
(b) – Vout=5V (lower than Vin): red – vC1(t), blue – VC(t), light blu – Fig. 5. Simulation waveforms of of state variables, (a) and output voltage, (b)
iL1(t), green – iL2(t). for varying α, (c). Waveforms with ripple are circuit simulation and
solid lines inside the ripple ranges are simulations of average model
variables.

IV. IMPLEMENTATION AND BENEFITS OF AVERAGE

MODELLING APPROACH
From analytical solutions for SEPIC variables, different
characteristics can be obtained. For example, Fig. 6 shows
output voltage Vout=ySS of the converter (8) as a function of
time ratio α for different load resistances, Fig. 7 presents
efficiency:

= = (15)

Fig. 8 demonstrates input characteristics of SEPIC, (input

current versus output voltage as a function of a time ratio α),
Fig. 4. Zoomed variables of SEPIC for Vout=5V: red – vC1(t), blue – VC(t), light (a) - ideal converter, all parasitic resistances except of R1 are
blue – iL1(t), green – iL2(t). zero, (b) – R1 and R2 are not zero, (c) – RC1 is also taken into
account, and finally (d) – all parasitic elements are included.
• For the varying time ratio α: Fig. 5 presents simulations The time ratio α is varied from 0.05 to 5 for all plots. Parasitic
results of state space variables, (a) and an output parameters of the converter were chosen as follows: R1=0.1Ω,
voltage, (b) of an average model in comparison with R2=0.1Ω, Rc1=0.05Ω, RC=0.05Ω. Forward diode voltage was
circuit simulations with α, (c) varying as: VD=0.6V, input voltage U=12V.
( ) As it can be seen from Figs. 6 and 7, sufficient output
= + (14)
voltage/power and high enough efficiency are depend on the
As it can be seen from Fig. 5, output voltage is load resistance, time ratio, and parasitic elements of the
following the time ratio changing, and average model circuits. Fig. 8 demonstrates that the same output voltage could
simulations lines are always positioned within the be obtained with different coefficients α, but input currents
ripple range even with varying time ratio. might be too high and a design ineffective. These

214
 characteristics can help the designer to appreciate the results When linear control strategy is selected, a designer should
and to confirm a compliance with design requirements. perform linearization of the converter model around the
For example, for designing SEPIC with efficiency above operation point for controller design. This strategy is named "a
80% and output voltages in the region (5-30)V, from Figs. 6, small signal approach" [1]-[5]. However, when output signal
and 7, follows that the load resistor should be higher than 6Ω. is expected to vary in a wide range, small signal approach does
not ensure good performance. In this case non-linear control
method is likely to be superior. In order to examine the
operation of a converter with non-linear controller, an average
model (2) can substitute it. In this way, simulations of a
converter dynamics will be simpler than when using a
complete model of a converter.
V. CONCLUSIONS
The average modelling approach for a switch-mode
converter analysis is well known and has been in use since the
1970s. In this research, we validate and employ this approach
in an extensive and different way from the usual manner, thus,
providing useful tools for a converter design. Despite the
complexity of the SEPIC mathematical description, by using
the proposed method, we obtained simple closed-form
Fig. 6. Output voltage as a function of the time ratio α
analytical expressions for different quantities in the circuit
(output voltage, power, efficiency, etc.), as a function of the
converter parameters. Furthermore, application of the
proposed method for converter design enables designing a
control system by using a simple average model without the
need for extensive and complex simulations of a high order
switch-mode power converter. The design approach of this
study could be applicable for other, similar, high order systems
or converters with duty cycle control.
REFERENCES
[1] R. Middlebrook, S.Cuk, "A general unified approach to modelling
switching converter power stages," Proc. IEEE Power Electronic
Specialists Conf., 1975, pp. 18-34
[2] V. Eng, U. Pinsopon, C. Banlaksananusorn, “Modeling of a SEPIC
converter operating in continuous conduction mode,” 2009, 6th
Fig.7. Efficiency as a function of the time ratio α international conf. on electrical engineering, electronics, computer,
telecom.,and inform. Technology.
[3] J. Carvalho, F. Tofoli, “Small-signal model validation of a SEPIC
converter based on the three-state switching cell in CCM using the
PWM switch model,”2017, COBEP
[4] B. Padhi, S. Padhy, K. Bhuyan, “Controller Design for Reduced Order
model of SEPIC Converter,” Int. conference SCOPES – 2016
[5] L.K. Wong, T.K. Man, “Small signal modeling of open-loop SEPIC
converters,” IET Power Electron., 2010, Vol. 3, Iss. 6, pp. 858-868
[6] H. Zhang, C. Yi, T. Wei, “Non-linear modal analysis of transient
interaction behaviors in SEPIC DC-DC converters,” IET Power
Electron., 2017, Vol. 10, Iss. 10, pp. 1190-1199
[7] A. Hren, P. Slibar, "Full order dynamic model of SEPIC converter,"
Proc. Of the IEEE Symp. on Ind. Elec., pp. 553-558, June 2005
[8] S. Ben-Yaakov, D. Adar, G. Rahav, "A SPICE compatible behavioral
model of SEPIC converters", Proc. IEEE PESC, 1996
[9] Pspice: MicroSim Inc., 20 Fairbanks, Irvine, California
[10] J. Hammerbauer, M. Stork, "State space study of the SEPIC converter,"
2013 Int. Conf. on Applied Electron., pp. 1-4
[11] E. Vuthchhay, C. Bunlaksananusorn, "Modelimg and control of a
Fig. 8. Input characteristics of the converter: (a) ideal case, only R1=0.1Ω is ZETA converter," Int. Power Electron. Conf., 2010, pp. 612-619
not zero, (b) R1=0.1Ω, R2=0.1Ω, (c) R1=R2=0.1Ω, RC1=0.05Ω, (d) RC=0.05Ω, [12] R.D. Middlebook and S. Cuk, "A general unified approach to modeling
other resistances with no change switching-converter power stages," Int. Journal of Electronics, vol. 42,
pp. 521-550, June 1977.

215

Powered by TCPDF (www.tcpdf.org)