Proceedings of the 2004 IEEE

htemational Conference on Control Applications

Taipei, Taiwan, September 24,2004

Adaptive Voltage Regulation of PWM Buck

DC-DC Converters Using Backstepping Sliding
Mode Control
Shoei-Chuen Lin Ching-Chih Tsai
Department of Electrical Engineering, National Chung Hsing University
250, Kuo-Kuang Road, Taichung 40227, Taiwan

Abstract-This paper develops novel methodologies and Currently, there exist many different approaches that
techniques for voltage -regulation and implementation of a have been proposed for the PWM switching contrnl design,
buck DC-DC converter. This type of converter is modeled e.g., PID control, optimal control, sliding mode control
as a linear averaged state-space system model with an [3,8] and f u v y control etc. It is worth noting that
adjustable load. The output voltage and the inductor different approaches for PWM switching control have the
current are employed to achieve design of an adaptive following three control objectives. The fust one is
backstepping sliding mode controller for the buck DC-DC concentrated on properly controlling the switching
converter. Using singular perturbation method, a reduced DC-DC converter to accomplish a zero steady-state error
order model is derived in order to show that the effect of between the output voltage and the desired setpoint in the
equivalent series resistance (ESR) of the capacitor can be presence of input voltage variations and load changes. In
ignored and the buck DC-DC converter can be well the second one, satisfactory transient responses can be
approximated by the reduced order model. Based on the achieved when the system is subject to perturhations,
reduced order model, adaptive sliding mode control based which will bring about instantaneous variations in load
on backstepping approach is proposed to control a current or line input voltage. Thirdly, much of the work in
stand-alone buck DC-DC converter. This proposed control the control of DC-DC converters has been focused on the
method has been verified by computer simulation and applications of linear control theories to the converters in
implemented utilizing a stand-alone digital signal the neighborhood of a fixed operating point and with small
processor (DSP) TMS320C542 from Texas Instruments. disturbances. However, when the switching DC-DC
Experimental results show that the proposed control converter has significant disturbances or the DC-DC
method is proven to be capable of giving satisfactory converter demands over wide voltage ranges, the
voltage regulation 'performance under input voltage operating points will not be fixed at one nominal position
variations and load changes. and the controller must be robust against these
Keywords: Adaptive control, backrtepping, DC-DC uncertainties.
converter: digital signal processor: pulse width modulation The singular perturbations design methodology has
(PWW, regulation, sliding mode control. been extensively used for a large class of nonlinear
systems that can be decomposed into slow dynamics and
1. INTRODUCTION fast dynamics. Those control theories and techniques
D C-DC converters are one of the simplest power
. . .
electronic clrcmts. They have been widely used in
using singular perturbations method have been widely
explored in [IO, 11, 141. In this paper we are interested in
the power supply equipment for most electronic systems. how the singular perturbations method can be applied to
As the demands for lightweight and highly efficient DC show that the effect of equivalent series resistance (ESR)
power sources, switchmode power supplies have become of the capacitor C in a buck DC-DC converter can he
vely promising solutions to these more and more stringent ignored, and to derive a order reduced slow model for the
demands. Various modeling and analysis approaches for circuit in order to develop an adaptive and efficient control
the switching DC-DC converters have been reported in policy for achieving the required performance.
[1-2] over the past decades. Many researchers have Adaptive backstepping is a systematic and recursive
focused their attentions on the study of control techniques design methodology for nonlinear feedback control. This
of switching power supplies 13-91. Among these switching type of method has been well studied and documented in
control methods pulse-width modulation ( P W M )based on [12,13.14].This approach is based upon a systematic
fast switching and duty ratio control may be the most procedure for the design of feedback control strategies
extensively useful one. Generally speaking, buck DC-DC suited for nonlinear systems with exhibiting uncertainties,
PWM converters are nonlinear systems due to their and guarantees global regulation and tracking for the
inherent nonlinear switching operations, which are nonlinear systems transformable into the parametric-strict
accomplished by activeipassive switches. feedback forms. As the authors' best understanding, this
design approach has not been applied to regulate the buck

0-7803-8633-7/04/$20.00 02004 IEEE 1382

 Fig.2 Schematics of the buck DC-DC converter with a
Fig.] Basic configuration of the buck DC-DC fixed load current i, and an adjustable current i,
converter with a changeable load

DC-DC converters using inductor current measurement elements, we model the adjustable load current io as a
yet. fxed load current iM and a variable current i , . Note that
The objectives of the paper are to develop reduced
slow state-space models of the converters using singular the current 6 remains constant except at the time
perturbation method based on output voltage and inductor instants to change voltage setpoints.
current measurements, and to design adaptive In Figure 2, the transistor operates at a fixed
hackstepping sliding mode controllers for stand-alone frequency with a period T with a duty-ratiod . The
buck DC-DC converters with the required specifications. inductor current is assumed to be continuous, and the
To make the design feasible, computational simulation is inductance and capacitance components are lossless and
used to verify its performance, and the Lyapunov second the output capacitor ripple voltage is considered to he
theorem is used to analyze the stability of the designed negligible. The relationship between steady-state output
control system. Finally, a single-chip DSP-based voltage and duty-ratio is obtained from equating the
controller is implemented to achieve required performance positive and negative inductor volt-seconds within a
for the designed DC-DC converter. The proposed control switching cycle.
technique is expected to give robust voltage regulation in In what follows studies the two-time scale propaies
the presence of input voltage variations and load current of the standard model and gives a approximate model
fluctuations. based on the decomposition of the model into slow and
This paper is organized as follows. Section I1 fast models. In doing so, we consider a buck DC-DC
establishes the circuit's reduction using singular converter with a fxed load current, i, , and an adjustable
perturbation method. Section 111 is used to derive a current, i, , shown in Fig.2. Notice that the resistance R
reduced slow model of the circuit. In Section N, the
adaptive backstepping approach is adopted for in Fig2 is fixed. The application of linear circuit modeling
synthesizing a adaptive backstepping sliding mode method with aforementioned conditions leads to the
controller to regulate the converter's output voltage. following equations:
Computer simulations and experimental results are L-+ri,
di, =vp-v, (1)
conducted it1 Sections V and VI to show the effectiveness dl
of the proposed control method. Section VI1 concludes the d dvo dic
m
ic = C-(v, - icrc) = C--dt cr, -
dt (2)
papa.
Let E = c r, be a time constant for the capacitor C due
U. SINGULAR PERTURBATION ANALYSIS torr. Since0 < E << 1, then (2) becomes
Consider a PWM-based buck DC-DC converter with
an adjustable load as shown in Fig.1. In this paper, the (3)
buck converter is assumed to operate in continuous
conduction mode (CCM). Fig.] shows a basic By decomposing the model (3) into fast state ic,
configuration of the converter cell, consisting of a power
(boundruy) together with slow ,i state. i.e.,ic = ii, + i , ,
MOSFET transistor as a switch, a fast recovery diode and
RLC components. To simplify the notations, we defme we obtain the fast mode satisfying
that i,, he the current passing through the inductor L , i, did
he the current in the capacitor C, io be the converter
E-+i
dt '' = O (4)

and the slow state described by

output current. Moreover, v , denotes the DC input source
voltage, is the converter output voltage, r is the
equivalent series resistance of inductor L, rc is the By a simple calculation, we have
.~
equivalent series resistance of the capacitor C. In order to = icf(0)e '
iCt (6)
provide an analytical model for predicting the responses which leads to that ic, + 0 as I > 5 E . Consequently,
of the converter with respect to load changes, input
voltage variations and uncertainties caused from circuit for I > 5 ~it follows
, that

1383
 which, similarly, leads to
X=A, x + B , u
where
ro o 1
(a) Transistor is ON and diode OFF
Recalling that d is the duty ratio, and using the
well-known averaging method [7], we obtain the
following averaged circuit model
i = Ax+ Bu (12)

Transistor is OFF and diode is ON where

A=A,+(A,-A2)d =
Fig.3 Circuit configurations of a buck DC-DC converter
without

i, = id (t) + i,(t) z i,(t) = C% (7)

dt B=B,+ (B-B,)d =
(7) reveals that, after a very short transient time, the effect
of equivalent series resistance (ESR) of the capacitor C
can be ignored, and (2) can be well approximated by the
slow state (7). Iv.ADAPTIVE BACKSTEPPING SLIDING MODE
III. REDUCED SLOW MODEL CONTROL DESIGN
Fig.3 shows two circuit configurations of the buck In this section, an adaptive control strategy is
converter corresponding to the ON and OFF states of the provided to overcome the problem with unknown circuit
transistor. This figure also brings dynamic equations, parameters and load variations. Using voltage output and
expressed in the standard state-space form, for each circuit inductor current measurements, an adaptive sliding mode
configuration. The inductor current and output voltage are controller is designed to regulate the circuit's output
selected as the two state variables forming the state voltage for the buck DC-DC converter. Due to the fact that
vector x .To derive the averaging reduced slow model, we the model (12) is not suitable for deriving adaptive control
have to fmd the dynamics of the circuit for ON and OFF strategy, we need to modify the averaged reduced model.
states of the transistor, respectively. Consider the fust case In doing so, (12) can he rewritten as below:
that the transistor switch is ON and the diode is OFF, as d - ,b' v, - 7 i,.
i, = (13)
shown in Fig.3(a). The dynamic equations for the circuit a
are described hy

where a=:, P = ; ,1 7=+, and a , P , r , are

v =-I
1 . ---y 1 ---1
1.
' C L RC" C N
~ ~~~
unknown parameters. Note that the resistance R is known.
which can be re\nitten in the following vectormatrix The control goal of the converter is to fmd a controller
form x=A, a+B,u (9) with adjustable parameters and a set of parameter
adaptation laws such that the output voltage v,,tncks the
desired constant reference voltagevn, . To do so, we
defme the tracking error asv, =v, -v,., . This implies
that", =v, +v,*, anda = ;,. Hence, (13) and(1.1) can be
Next, we move to the second case that the transistor rewritten as
switch is OFF and the diode is ON. In the same way, the : d-fivo-yiL
IL = (15)
dynamic equations of the circuit shown in Fig. 3(h) can be a
exuressed by:
V# = - ( I1
L . --y1 -i ) (16)
r . 1 C R "
,L = - - I L - - Y 0
L L The detailed design procedure of the proposed
y =-I
1. --"1 : --,
1. adaptive hackstepping sliding mode controller for the
' C' RC" C N converter is stated in the sequel.

1384
Stepl: =-(k, + l / R ) v ; - k,sz - k2Is1 < 0 (24)
To facilitate the design of controller, let i, he which implies that r‘, is negative defmite. W~ththe
considered as a virtual wntrol variable such that the state Barhalat’s lemma, we imply that + o and s + O
variable ve is asymptotically stable. ‘Thus, by designing
as t + m .
iL = @ ( v e ) = L - k , g- k , b * ( r ) d r , k ,> O,k, > 0 (I7) Step 3:
R Find a set of parameters adaptation laws such
we obtain the following dynamics of the error va that s + 0 and v, + 0 as t + 00. We present the following .
adaptive control law
d = v, + f iL -6 (kpva+k,v.)-y -+k, spn(s) (25)
The subsequent method is used to show that ve
approaches zero as I tends to infmity. By choosing the
following Lyapunov function candidate
s = iL --
R
+ k,ve + k, 6 V~(I-)~T (26)

where&,jr,aadf are the estimates of the parameters .J,

1 1
‘-2
Y -
Zk,
-Cv: +-(k, va(r)dr + i, j* andy. Substituting the control (25) into (19), it follows
we obtain the time derivative of V, along its trajectoq that
6 =-(k, +l/R)ve* s. = - j-f iL td (k9Ca+k,v,)-v, - k , - k , s g n ( s )
V, (27)
a
Since 6 is negative defmite, we
apply the Barbalat’s Next, we fmd a radially, unbounded Lyapunov function
lemmatohavev,+o t+m and use it to derive a set of parameter adaptation laws.
Remark 1 The parameters k, and k, can he designed Hence we have Lyapunov function
such that (18) has a desired characteristic function. 1 1 . 1
Yo =-k,vs’+-Cv:+-asz+-+-+- 2 a’ (28)
2 2 2 24 21, 2 4
Step2 :
This step aims at fmding a controller suited for where A>O, A>O, $ > o , Z = a - d . , p = p - j ,
adaptive strategy using hackstepping and sliding mode p=y-f and a=-&, D=-D, y = - y .
-. .. -. ~

approaches. The following sliding surfaces = i, - ((v,) is

Taking the time derivative of ymandusing (27) obtains
selected to attain the controller design. Differentiating the
sliding surface function gives
j = I, - k v , ) = iL + k p v e ik,ve
-pv,-yi,+d
- + kpv*+ k,v. (19)
a

Design the control d

d = D y + y i~-a(k,v,+k,v,)-v,-k,s-k,sgn(s) (20)
such that Thus, we propose the following parameter adaptation laws
. -ve-k,s-k s n 6 = A(koCe+ k , y ) s A, > 0
S= (’),k, 20,k, > O (21)
a
Notice that the control (20) can he reduced to a type of
sliding mode control if the parameter k,is zero. In order y = -4 i,S s>o
to prove the asymptotical stability of the overall system, Such that V , = - ( k , + I / R ) v ; - k , s * - k , I s l ~ O and Vn is
we need to rewrite (16) as below negative semidefmite. Again, the Barbalat’s lemma
implies that all the signals, ”*.s, &,B,andp,a.re bounded
... ..
and V, + 0 ,and s +0 , for t +m .This completes the
and propose the following Lyapunov function candidate
proof of which the system is globally asymptotically.
Y -
’-2
-Cv:
1 +-(k,
1
2k.
6
r)dr + iNj2+-as*
1
2
(23) stable. The following theorem summarizes the result.
Taking the time derivative of V along its trajectoq Theorem 1: Assume thaf the reference voltage v,~, is
gives = v,Cv, +v,(k, r)dr + i,)+saS . consfanf, the parameterskD,k,,k , and k , are positive
From (21) and (22), it follows that gains. Then all the frajectories o f f h e closed-loop error
p2= -(kp + 1 / R)v,‘ + v,s - V-S - k,s2- k, 1s) system composed of the sysfem(l5).(16 ) and the
adaptive control (25),(26),wifh the set of paramefer

1385
adaptation lmvs(3I),are globally uniformly bounded,
thereby imprying that all the variables,
6, p,f,~~,;~,vo,s ,are global/y uniformly bounded
for t>O. Furthermore, V, --f o ,S --t o ,as t--tm.
Remark 2 The estimates & , j , a n d j usually do not
converge to their true values, except that the reference
voltage meets the persistent excitation conditions.

V. COMPUTER SIMULATIONS
The DC-DC buck converter with the proposed adaptive
controller was simulated on a digital computer using
Matlab. The1 is chosen 2 and other parameters given in
Table 1 were adopted in this simulation. The reference
output voltage was set at 5 volts, and the input voltage at
20 volts.
The input DC voltage was increased to its 150% at the
time intervals [1,2] and [6,7], shown in Fig.4 The load was Fig.4. Input voltage vx was increased to its 150%
also changed hy its 133% ( i , from 3A to 4A) variations at in the time interva?s_[l,l] and [6,7].
the time interval [4,5] and [7,8], illustrated in Fig.5,
respectively. Fig.6 displays the behavior of the inductor
current i , due to the input voltage and load variations.
Fig.7 depicts the behavior of the duty ratio d of the
i- 7 4
converter. In order to maintain the output voltage equal to
the desired setpoint voltage, the duty ratio d varies with
input voltage and load current changes. Fig.8 shows the
output voltage output voltage remains at its setpoint. We i__ 4 , z ; ; ,“i.<.a ; , = ; ! o
observe in Fig.8 that the proposed controller exhibits Fig.5 The adjustable load current I~ was change1
excellent voltage regulation performance.
from 3A to 4A in the time intervals [ $ ~ ] ~ [ 8 _ , .
VI. EXPERIMENTATION AND I 1
DISCUSSION
An experimental DSP-based PWM buck DC-DC
converter is shown in Fig.9 in order to verify the proposed
control method. Three experiments were conducted. The
output is regulated at 4 volts and other major parameters
are listed in Table 1. The waveforms shown in Figs.10, 11
and 12 were taken from a Tektronix TDS 220 digital
oscilloscope, and the waveforms of i, were measured
from Hall current sensors. Fig.10 shows the experimental
output voltage and inductor current waveforms for a converter in Fig. 2
resistive load R = Ion at the setpoint voltage v,* = 4
The resulting controller without adaptation behaves
volts. It can be se? that the output voltage response like an integral sliding mode controller, and does not
presents a highly good result when the adaptive control have any numerical difficulties and stifiess caused
law is applied. Figure I I depicts the transient responses of by the fast mode of the capacitor current. The
i,andv0 when the load is subject to a sudden change adaptive capability of the controller has a powerful
from 5n t o i o n . Fig.11 reveals that the output voltage is strength to automatically tune its control parameters
robust against load changes. for buck DC-DC converters with unknown circuit
parameters. Moreover, simulations and experimmtal
vii. CONCLUSIONS results have shown that the proposed control method
This paper has proposed a novel adaptive backstepping is proven to be capable of giving satisfactory voltage
sliding mode control to achieve voltage regulation for a regulation performance under input voltage variations
buck DC-DC converter under input DC voltage variations and load changes. Future work will he devoted to
and load changes. The controller design is accomplished the extension of the proposed control method to
based on the circuit’s averaged reduced order model in parallel buck DC-DC converters.
which the fast mode of the capacitor current is ignored.

1386
 . d ; + , i i ; ; * % +
7.”.*.=* j Z * , , C h l
2, Ch 2
ZVOIf
z V-lr
1 -
1

g g . 7 Duty ratio dof the DCDC buck converter in

(Ch2) waveforms for v, = 4v,R = ion
. .
I 1

-
Fig.8 Output voltage response of the DCDC buck FigTll T r a n z r e s p o n z t h e inductorcurrent (El)
converter and output voltage ((32) for v,, = 4v and the load
changedfrom 5R to IOR

Thesis, Department of Electrical Engineering.

National Chung-Hsing University June 2003.
[6] A. J. Forsyth, and S. V. Mollov, “Modeling and
control of DC-DC converters,” Pro .of 1998 IEE
Power Engineering JournaI,pp229-236,0ct 1998.
I I I I-----------r [7] J. Mahdavi, A. Emaudi, M. D. Bellar and M. Ehsani,
U - “Analysis of power electronic converters using the
Fig.9 Control architecture of the proposed control method generalized state-space averaging approach,” IEEE
for the experimental buck DC-DC converter Trans. on Circuits and Systems, vo1.44. no.8, pp.
767-770, Aug 1997.
ACKNOWLEDGMENT [SI W. K. Yang. “Sliding Mode Control Design of
This study was supported by the National Science Switching Buck DC-DC Converters,” M S . 7?1esis,
Council of the Republic of China under Grant Deportment of Electrical Engineering, National
N S C I~- 2 6 2 2 - ~ - n n ~ - n n s - c ~ 3 . Chung-Hsing University June 2002.
[9] H. S. Ramirez, et al. “Adaptive input-output
REFERENCES linearization for PWM regulation of DC-DC power
[I] K. T. Ngo, “Altemate forms of the PWM switch converters,” Proc. of the American Control
models,”IEEE Trans on Aerospace and Electronic
Conference, Washington, June 1995.
~ysrems,vol.3S,no.4,pp.1283-1292,Oct. 1999.
[IO] R Ghanadan, “Nonlinear control system design via
[2] Tymerski, V. Vorperian, Lee, and W. T-Baumann dynamic order reduction,” Proc. ofrhe 33“Con. on
“Nonlinear modeling of the PWM switch,” IEEE
Trans. an Paver Electronics, vol. 4, 110.2,. .. UP. decision and Control, Dec 1994.
225-233, April 1989. [ I l l X. Xu, et al. “Modeling of generators and their
[3] J. Mahdavi, A. Emadi and H. A.Toliyat “Application of controls in power system simulations using singular
state space averaging method to sliding mode control perturbations,” IEEE Trans. on Power Systems,
of PWM DCDC converters:’ Conf Record of vo1.13, no.1, Feb 1998.
thelEEE,vol.2, pp.820 -827, 1997. [I21 K. J. Astrom and B. Wittenmark, Adaptive Control,
141 F. H. Hsieh, N. Z. Yen and H. T. Juang, “An optimal TdED,Addison Wesley, 1995.
controller design of buck DC-DC converter by [I31 M. KrstiC, 1. Kanellakopoulos, and P. KokotoviC,
regarding the uncertain load as stochastic noise,” “Nonlinear and Adaptive Control Design, John Wily
Proc. of 2001 Chinese Automatic Control Conference, and Sons,” INC.
pp.605-610,2001. [14] H. K. Khalil, “Nonlinear Systems,” 2“dED,Prentice
[5] S. C. Lin, “Adaptive Backstepping Sliding Mode Hall 1996.
Control of PWM Buck DC-DC Converters,” M.S.

1387