Proceedings of ECTI-CON 2008

Dynamic Modeling of a Zeta Converter with State-

Space Averaging Technique
Eng Vuthchhay and Chanin Bunlaksananusorn
Faculty of Engineering, King Mongkut’s Institute of Technology Ladkrabang (KMITL),
Chalongkrung Rd. Ladkrabang, Bangkok 10520

Abstract- A Zeta converter is a fourth-order DC-DC converter This paper presents dynamic modeling of a Zeta converter
made up of two inductors and two capacitors and capable of with the SSA technique. The paper is organized as follows.
operating in either step-up or step-down mode. Compared with a The SSA technique is reviewed in Section II. Modeling of the
Cuk or Sepic converters, the Zeta converter has received the least Zeta converter with the SSA technique is demonstrated in
attention and its dynamic properties have never been reported Section III. Simulation results are presented in Section IV.
before in the literature. This paper presents dynamic modeling of
Section V gives a conclusion of this work.
a Zeta converter with State-Space averaging (SSA) technique.
The modeling leads to a small-signal linear dynamic model of the II. OVERVIEW OF SSA TECHNIQUE
converter, from which the transfer functions used for feedback
control design can be determined. Simulation results are For the DC-DC converters operating in Continuous
presented to verify the accuracy of the obtained model. Conduction Mode (CCM), there exist two power circuit states
within one switching period, T. One is when the MOSFET is
I. INTRODUCTION
turned on for an interval dT, and another is when the
Nowadays, a DC-DC converter is widely used as a power MOSFET is turned off for an interval (1-d)T, where d is a duty
supply in electronic systems. The converter incorporates cycle. The state-space equations for these two circuit states are:
feedback control to maintain the constant output voltage.
However, changes in an input voltage and/or load current will
­° dx(t ) = A x (t ) + B u (t )
® dt 1 1 (1)
cause the converter’s output voltage to deviate from the °̄y ( t ) = C1 x(t ) + E1u( t )
desired value. It is a task of the feedback control to correct this
error and quickly bring the output voltage back into the ­° dx(t ) = A x(t ) + B u (t )
regulated level. Modeling plays a key role in revealing the ® dt 2 2 (2)
insight of the converter’s dynamic behavior as well as °̄y ( t ) = C2 x( t ) + E 2u (t )
providing a basis for feedback control design. In the last two To find the averaged behavior of the converter over one
decades, there has been a continually active research on DC- switching period, (1) and (2) are weighed average by the duty
DC converters; as a consequence, several modeling methods cycle as:
have been proposed [1]. Among them, the State-Space
Averaging (SSA) technique [2] is one of the best-known ­° dx(t ) = A x (t ) + B u (t )
methods. It provides a systematic way to model the converter ® dt s s (3)
and gains widespread acceptance. The SSA modeling consists °̄y ( t ) = Cs x (t ) + Es u( t )
of three stages: (1) Formulation of state-space equations of the where
converter for each subinterval in a switching cycle, (2) A s = A 1 d + A 2 (1 − d ) , B s = B 1 d + B 2 (1 − d ) , Cs = C1 d + C 2 (1 − d ) ,
Average of these equations to get a single averaged state-space
equation, and (3) Perturbation of the averaged equation to get and E s = E1 d + E 2 (1 − d ) .
a linear small-signal state-space equation, from which various Equation (3) is a nonlinear continuous-time equation. It can be
transfer functions can be determined. Since the SSA is a linearized by small-signal perturbation with x = X + x ,
matrix-based technique, i.e. all the steps described above are y = Y + y , u = U + u , and d = D + d , where the  symbol
done in a matrix form, formal matrix treatment can be applied
to facilitate the modeling process. represents a small signal, and the capital letter a DC value. It
The SSA technique is commonly used to model the second- should be noted that X >> x , Y >> y , U >> u , and D >> d .
order converters such as buck, boost, and buck-boost The perturbation yields the steady-state and linear small-signal
converters [1- 4]. Modeling of the fourth-order converters state-space equations in (4) and (5) respectively.
such as Cuk and Sepic converters has also been studied [5- 7]; ­° dX = AX + BU = 0
however, the techniques used were based on circuit averaging ® dt (4)
approach rather than the SSA. Like Cuk and Sepic converters, °̄ Y = CX + EU
a Zeta converter is the fourth-order converter made up of two
inductors and two capacitors and capable of working in either ­° dx (t ) = Ax (t ) + Bu (t ) + B d (t )
step-up or step-down mode. Its dynamic characteristics, ® dt d (5)
°̄  t =  t +  t + 
d
however, have never been reported before in the literatures. y ( ) Cx ( ) Eu ( ) E d (t )

978-1-4244-2101-5/08/$25.00 ©2008 IEEE 969

where (ESR) of the capacitors C1 and C2 respectively. The current
A = A1 D + A 2 (1 − D ) , B = B 1 D + B 2 (1 − D ) , C = C1 D + C 2 (1 − D ) , source, IZ, models the load current.
In CCM, the converter exhibits two circuit states. The first
E = E 1 D + E 2 (1 − D ) , B d = (A 1 − A 2 )X + ( B 1 − B 2 ) U , and state is when the MOSFET switch is turned on (Fig. 1(b)).
E d = (C1 − C 2 )X + ( E1 − E 2 ) U . During this interval (dT), the currents through L1 and L2 are
supplied by Vg, and hence iL1 and iL2 increase linearly as
The steady-state solution of the converter can be found by
shown in Fig. 2. This interval is called the charging mode. The
solving (4) which gives:
second state exists when the MOSFET switch is turned off
­ X = − A −1BU (Fig. 1(c)). During this interval ((1-d)T), L1 and L2 release the
® −1
(6)
¯ Y = ( − CA B + E)U stored energy to C1 and the output section. Thus, iL1 and iL2
The small-signal transfer function of the converter can be decrease linearly as shown in Fig. 2. This interval is known as
found by applying the Laplace transform to (5). In a matrix the discharging mode.
C1 rC1 iL2 L2
form, we get: Q

­x (s ) = ª ( sI − A ) −1 B ( sI − A ) −1 B º ª u ( s ) º C2 +
° ¬ d¼
¬« d ( s ) ¼» Vg L1 D R VO Iz
® (7) rC2 _

°y (s) = ª¬ C( s − A ) −1 B + E C( sI − A ) −1 B d + Ed º¼ ª  º
u ( s ) iL1

¯ «
¬ d ( s) »¼ (a) Zeta converter.
In DC-DC converters, the input variable u usually contains the
input voltage and load current. Hence, u is express as u= [u1 C1 rC1 iL2 L2
u2]T, the matrix B as B=[Bu1 Bu2], and the matrix E as E=[Eu1
C2 +
Eu2]. Therefore, Vg L1 R VO Iz
­ ªu1 (s ) º iL1
rC2 _
°x (s ) = ª ( sI − A )−1 Bu1 ( sI − A ) −1 Bu2 ( sI − A ) −1 Bd º «u2 ( s ) »
° «¬ »¼ « »
° ¬ d ( s ) ¼ (8)
® (b) Zeta converter when MOSFET is turned on.
° ªu1 ( s ) º
°y (s ) = ª¬C( sI − A ) Bu1 + Eu1 C( sI − A ) Bu2 + Eu2 C(sI − A ) Bd + Ed º¼ «u2 ( s ) »
−1 −1 −1

« » C1 rC1 iL2 L2
°̄ ¬ d (s ) ¼
For the fourth-order converter, (sI–A)–1Bu1, (sI–A)–1Bu2, and C2 +
(sI–A)–1Bd are the matrices that have four rows and one Vg L1
rC2
R VO
_
Iz
column. So the above equations can be extended into: iL1

­ ª G vi ( s ) G zi ( s ) G di ( s )º (c) Zeta converter when MOSFET is turned off.

° « » u ( s ) º
1 1 1

G di ( s ) ª 1
Fig. 1. Operation of Zeta converter.
°x ( s ) = « G vi ( s ) 2
G zi ( s )
2 2 » « u ( s ) »
°° « G vv ( s ) G dv ( s ) » « d ( s ) »
2 iL1 iL2

«
G zv ( s )
» ¬ ¼ IL1max IL2max
® IL1 IL2
1 1 1
(9) IL1min IL2min
° ¬« G vv ( s ) 2
G zv ( s )
2
G dv ( s ) ¼
2
» t t
° ª u1 ( s ) º 0 dT (1-d)T T 2T 0 dT (1-d)T T 2T

°y ( s ) = ª¬ G vv ( s ) G zv ( s ) G dv ( s ) º « u 2 ( s ) »
¼
(a) iL1 waveform (b) iL2 waveform
°¯ «¬ d ( s ) »¼ Fig. 2. Current waveforms.

where A. State-Space Description

−1 −1 −1 Following (1) and (2), the state-space equations of the Zeta
G ( s ) = [( sI − A ) B ] , G ( s ) = [( sI − A ) B ] , G ( s ) = [( sI − A ) B ] ,
vi1

−1
u1 11 zi1

−1
u2 11 di1

−1
d 11
converter for the on and off states of the switch can be written
G
vi 2
( s ) = [( sI − A ) B
u1 21
] , G
zi 2
( s ) = [( sI − A ) B
u2
]
21
, G
di 2
( s ) = [( sI − A ) B ]
d 21
, from Fig. 1(b) and (c) respectively, given:
−1 −1 −1 diL1 rC1 v Vg
G
vv1
( s ) = [( sI − A ) B
u1 31
] , G
zv1
( s ) = [( sI − A ) B
u2
]
31
, G
dv 1
( s ) = [( sI − A ) B ]
d 31
, = (δ − 1)iL1 + C1 (δ − 1) + δ
dt L1 L1 L1
−1 −1 −1
G ( s ) = [( sI − A ) B ] , G ( s ) = [( sI − A ) B ] , G ( s ) = [( sI − A ) B ] , diL 2 −1 rC 2 R vC1 R Vg rC 2 R
vv 2 u1 41 zv 2 u2 41 dv 2 d 41 = (r δ + )i + δ− v + δ+ I
−1 −1 dt L2 C 1 rC 2 + R L 2 L2 L2 (rC 2 + R ) C 2 L2 L2 (rC 2 + R ) Z
G ( s ) = C ( sI − A ) B +E , G ( s ) = C ( sI − A ) B +E , and
vv u1 u1 zv u2 u2 dvC 1 iL1 iL 2
−1 = (1 − δ ) − δ (10)
G ( s ) = C ( sI − A ) B +E dt C1 C1
dv d d
dvC 2 R 1 R
= i − v − I
III. MODELING OF A ZETA CONVERTER BY SSA TECHNIQUE dt C2 ( rC 2 + R ) L 2 C2 ( rC 2 + R ) C 2 C2 ( rC 2 + R ) Z
r R R r R
A Zeta converter shown in Fig. 1(a) is made up of the vo = C 2 iL 2 + v − C2 I
rC 2 + R rC 2 + R C 2 rC 2 + R Z
MOSFET switch (Q), diode (D), two inductors (L1 and L2), It should be noted that the equations are expressed in a
and two capacitors (C1 and C2). The resistor R is a standing compact form using the switching function, . When the
load, while rC1 and rC2 are Equivalent Series Resistances switch is on,  =1, (10) will then become the on-state equation.

970
When the switch is off,  =0, (10) will then become the off- C. Linear Small-Signal State-Space Equations
state equation. From (4) and (5), the averaged matrices for the Given the averaged matrices (11) to (16), the linear small-
steady-state equations and the linear small-signal state-space signal state-space equations of the Zeta converter can be
equations are: formulated as given in (5):
ª − rC1 −1 º
« L1 (1 − D) 0 L1
(1 − D) 0
» ª − rC1 (1 − D ) 0 −1 (1 − D ) 0
º
« » « L1 L1 »
« 0 −1 (rC1D + rC 2 R ) D −R »
« L2 rC 2 + R L2 L2 (rC 2 + R ) » ª iL1 ( t ) º « −1 » ª iL1 (t ) º
A = A1 D + A 2 (1 − D ) = (11) « » « +
rC 2 R D −R »« »
«1 »
0 ( r D ) 
» = « L2
L 2 ( rC 2 + R ) » iL 2 ( t )
C1
−D d iL 2 ( t ) rC 2 + R
« « »
L2
« (1 − D ) » » « v ( t ) » +
dt « v (t ) » « 1
0 0
« C1 C1
» −D
« v ( t ) » « C1 (1 − D ) C1 »«
C1 C1

«0 R −1 »
0 0
»
0 ¬ C2 ¼ « » ¬ vC 2 (t ) ¼
¬« C2 ( rC 2 + R ) C2 ( rC 2 + R ) ¼»
« »
R −1
ª1 D º «¬ 0 C 2 ( rC 2 + R )
0
C 2 ( rC 2 + R ) ¼
»
0
« L1 » ªD Vg º
« L1 »
0
«1 D rC 2 R » (1 − D ) L1

B = B1 D + B 2 (1 − D ) = « L2 L2 ( rC 2 + R ) » (12) « »
«D rC 2 R R ( rC 1 I Z − Vg )
» ª v g ( t ) º
«0 0 » « L2 L2 ( rC 2 + R ) L2 [( DrC1 + R (1 − D )] » « »
« −R » « » « iZ ( t ) » (18)
«¬ 0 C 2 ( rC 2 + R ) »
¼ «0 0
− DVg − RI Z (1 − D )
» «¬ d ( t ) »¼
« [ DrC 1 + R (1 − D )](1 − D ) C1 »

ª rC 2 R R º « »
C = C1 D + C 2 (1 − D ) = 0 0 (13) «0 −R
»
¬« rC 2 + R »
rC 2 + R ¼ ¬ C 2 ( rC 2 + R )
0
¼

E = E1 D + E 2 (1 − D ) = 0
ª − rC 2 Rº (14)
ª iL 1 (t ) º
ª vg (t ) º
«¬ rC 2 + R »
¼ R º « iL 2 ( t ) » ª

ª
vo ( t ) = 0
rC 1 R − rC 2 R º « i ( t ) »
«¬
0
» «
+ R ¼ vC 1 ( t )
» + «0 0
»¼ « Z » (19)
» ¬
rC 1 + R rC 2 + R
«
rC 2
ª Vg º ¬ d (t ) ¼
« » ¬ vC 2 ( t ) ¼
(1 − D ) L1
« R ( rC 1 I z − Vg ) » D.Finding Transfer Functoins
« » From (9), there are altogether fifteen transfer functions that
Bd = ( A1 − A 2 ) X + ( B1 − B 2 ) U = « L2 [( DrC1 + R (1 − D )] » (15) can be determined from (18) and (19). However, only a few of
« − DV g − RI z (1 − D ) » them are significant for feedback control design purpose.
« [ DrC1 + R (1 − D )](1 − D )C1 » These transfer functions are:
« »
¬ 0 ¼
The duty ratio-to-output voltage transfer function
E d = ( C1 − C 2 ) X + ( E1 − E 2 ) U = [0] (16)
vo ( s ) −1
G dv ( s ) = = C ( sI − A ) B d + Ed
B. Steady-State Equations d ( s )
Given the averaged matrices (11) to (16), the steady-state R
2
( a1 s + b1 s + c1 )( d1 s + 1)
solution of the converter can be obtained from (6): = (20)
4 3 2
(1 − D )[( DrC 1 + (1 − D ) R ] a 2 s + b2 s + c 2 s + d 2 s + e 2
­ ª D º
° 1 The input voltage-to-output voltage transfer function
« R (1 − D ) »
° ª I L1 º « 1 1− D» V
vo ( s ) −1
°« I L2 » = RD
«R ª gº
D »« »
G vv ( s ) =
v g ( s )
= C( sI − A ) B u1 + E u1
° « C1 » rC1 D + R (1 − D ) «
V
» ¬IZ ¼ , , , ,
®¬«VC 2 ¼» (17) 2
− rC1 » ( a1 s + b1 s + c1 )( d 1 s + 1)
«1 = DR (21)
° «¬1 − rC1 »
4 3 2
¼ a 2 s + b2 s + c 2 s + d 2 s + e2
°
° (Vg − rC1I Z ) DR
The output impedance transfer function
°Vo = vo ( s ) −1
¯ DrC1 + R (1 − D ) G zv ( s ) = = C ( sI − A ) B u2 + E u2
iz ( s )
If rC1 and rC2 are assumed to be zero, the conversion ratio of ,, 4 ,, 3 ,, 2 ,, ,,
a1 s + b1 s + c1 s + d1 s + e1
Vo D = −R (22)
the Zeta converter will become = , which is the 4
a2 s + b2 s + c 2 s + d 2 s + e2
3 2
Vg 1− D
The coefficients in (20) to (22) are listed in TABLE I.
same as Cuk and Sepic converters.

971
 TABLE I V. CONCLUSION
COEFFICIENTS OF Gdv(s), Gzv(s), AND Gvv(s).
In this paper, dynamic modeling of a Zeta converter has
2
a1 = (Vg − rC 1 I Z )(1 − D ) RL1C1 , c1 = (Vg − rC 1 I Z )(1 − D ) R , d 1 = C 2 rC 2 , been performed with the State-Space Averaging (SSA)
2 2 2 technique. The results yield an insight into the steady-state and
b1 = [ C1 RrC 1 (1 − D ) − L1 D ]Vg − [( L1 D + C1 rC 1 (1 − D )](1 − D ) RI Z . small-signal dynamic properties of the converter as shown by
, , , , (17), (18), and (19) respectively. To provide a basis for
a1 = L1 C1 , b1 = rC 1C1 (1 − D ) , c1 = 1 − D , d1 = C1 rC 2 .
feedback control design, three relevant transfer functions
,, ,, (Gdv(s), Gvv(s), and Gzv(s)) were derived and their Bode plots
a1 = L1 C1 L2 C 2 rC 2 , b1 = L1 C1 ( L2 + rC 1 rC 2 C 2 D ) + rC 1 rC 2 C1 L 2 C 2 (1 − D ) ,
presented. Finally, simulation results are given to validate the
,, 2
small-signal dynamic model of the Zeta converter.
c1 = [ rC 2 L2 C 2 (1 − D ) + rC 1 C1 L2 + rC 1 rC 2 C1 C 2 D ](1 − D ) + L1 D ( rC 1C1 + rC 2 C 2 D ) ,
,, 2 2
d 1 = L2 (1 − D ) + L1 D + ( rC 1 C1 + rC 2 C 2 ) rC 1 D (1 − D ) ,
,,
e1 = rC 1 D (1 − D ) .
REFERENCES
[1] R. W. Erickson and D. Maksimovi, Fundamentals of Power
a 2 = ( R + rC 2 ) L1C1 L2 C 2 , e 2 = [(1 − D ) R + DrC 1 ](1 − D ) , Electronics, 2nd ed., Kluwer Academic Publishers, 2001.
[2] R. D. Middlebook and S. Cuk, "A General Unified Approach to
b2 = ( rC 2 + R )(1 − D ) rC 1C1 L 2 C 2 + L1 C1 [ L 2 + RrC 2 C 2 + ( R + rC 2 ) rC 1C 2 D ] ,
Modeling Switching-Converter Power Stages," International Journal
of Electronics, vol. 42, pp. 521-550, June 1977.
2
c 2 = [ L2 C 2 ( R + rC 2 )(1 − D ) + rC 1 C1 L 2 + C1 C 2 rC 1 rC 2 D + C1C 2 RrC 1 rC 2 +
[3] N. Mohan, T. M. Undeland, and W. P. Robbins, Power Electronics,
Converter, Applications, and Design, 3rd ed., John Wiley and Sons
2 2
C1 C 2 RrC 1 D ](1 − D ) + [( R + DrC 1 ) C1 + ( R + rC 2 ) C 2 D ] L1 , Inc, 2003.
[4] M. H. Rashid, Power Electronics Handbook: Devices, Circuits, and
d 2 = [( rC 2 RC 2 + L2 )(1 − D ) + ( rC 2 + R ) rC 1 C 2 D + ( R + rC 1 D ) rC 1 C1 ](1 − D ) + L1 D
2
. Applications, 2nd ed., Elsevier Inc, 2007.
[5] R. Ridley, "Analyzing the Sepic Converter," Power Systems Design
Europe Magazine, pp. 14-18, November 2006.
IV. SIMULATION RESULTS [6] A. Hren and P. Slibar, "Full Order Dynamic Model of SEPIC
Converter," Proc. of the IEEE International Symposium on Industrial
Fig. 3 shows Bode plots of Gdv(s) in (20), Gvv(s) in (21), and Electronics, pp. 553-558, June 2005.
Gzv(s) in (22) generated with MATLAB. To plot these transfer [7] P. R. K. Chetty, "Modeling and Analysis of Cuk Converter Using
functions, the following converter parameters are used: Current-Injected Equivalent Circuit Approach," IEEE trans. on
Industrial Electronics, pp. 56-59, February 1983.
L1=100μH, C1=100μF, rC1=0.19, L2=55μH, C2=200μF,
rC2=0.095, Vg=15V, VO=5V, IZ=0, and R=1. All the three
transfer functions have the same four complex poles: p1,2=(–
2.5234±9.4385i)x103 and p3,4=(–1.7028±7.0112i)x103. Gdv(s)
has one real and two complex zeros: z1=–5.2632x104 and
z2,3=(–0.0301±0.8655i)x104. Gvv(s) has one real and two
Vo (v)

complex zeros: z1=–5.2632x104 and z2,3=(–

4
0.0717±0.8630i)x10 . Gzv(s) has two real and two complex
zeros: z1=–5.2632x104, z2=–0.0969x104, and z3,4=(–
0.0670±0.8193i)x104.
Fig. 4(a) shows a start-up transient of the converter
simulated with MATLAB using (18) and (19). The output
voltage is settled to about 5V, approximately 4ms after Vg was Time (ms)
applied. This result agrees well with the PSPICE simulation (a)
result of the Zeta converter using an ideal switch shown in Fig.
4(b), confirming the accuracy of the derived model.
Vo (v)

Time (ms)
(b)
Fig. 4. Output voltage start-up transient simulation:
(a) MATLAB and (b) PSPICE.
Fig. 3. Frequency responses of Gvv(s), Gzv(s), and G dv(s).

972