See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/245023293

Analysis and design of a sliding-mode strategy for start-up control and

voltage regulation in a buck converter

Article  in  IET Power Electronics · January 2013

DOI: 10.1049/iet-pel.2011.0494

CITATIONS READS

31 326

7 authors, including:

L. Martinez-Salamero Germain Garcia

Universitat Rovira i Virgili Laboratoire d’Analyse et d’Architecture des Systèmes (LAAS)
261 PUBLICATIONS   4,353 CITATIONS    269 PUBLICATIONS   4,369 CITATIONS   

SEE PROFILE SEE PROFILE

Marcos Orellana C. Alonso

Universitat Politècnica de Catalunya University of Toulouse
12 PUBLICATIONS   147 CITATIONS    148 PUBLICATIONS   1,434 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Discrete SMC View project

Sistemas Eólicos - Topologias y Control View project

All content following this page was uploaded by C.E Carrejo on 11 March 2015.

The user has requested enhancement of the downloaded file.

> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 1

Analysis and Design of a Sliding-Mode Strategy

for Start-Up Control and Voltage Regulation in a
Buck Converter
L. Martínez-Salamero, IEEE Member, G. García, IEEE Member, M. Orellana, IEEE Member,
C. E. Carrejo, C. Lahore, B. Estibals, IEEE Member, and C. Alonso.

commercial chips [3], which facilitate the use of this control in

Abstract—This paper presents the analysis and design of a new industrial applications.
sliding-mode control of a buck converter operating in continuous The name hysteresis control proceeds from the use of a
conduction mode that minimizes the energy during start-up and hysteresis comparator as a modulator in the control loop of a
provides output voltage regulation in front of input voltage
perturbations and load disturbances. A linear combination of
power converter in a clear-cut contrast with systems
inductor current and capacitor voltage errors with respect to employing a pulse with modulation (PWM). The analysis of
their corresponding equilibrium values is analyzed as a switching hysteresis control-based systems can be performed by means
surface. The analysis reveals that this surface is inherently of describing function, Tsypkin method or sliding-mode
insensitive to input voltage perturbations and offers a good control [4]. The first two methods are frequency-domain
compromise between inrush current and output voltage response techniques with different properties, namely, the describing
rapidity when it is particularized in the case of inductor current
error. The resulting controller is implemented analogically and
function is a relatively simple analytical approach that has
employs two switching surfaces, i.e., one surface for start-up and some limitations in the prediction of limit cycles, while the
another one for output voltage regulation. The theoretical Tsypkin method is an exact numerical approach requiring
predictions are verified by means of simulations and more involved calculations. Sliding-mode control, in turn, is a
experimental results. time-domain analytical technique that predicts with high
precision the dynamical behavior of power converters with a
Index Terms—DC-DC converters, buck converter, sliding- hysteresis comparator in the control loop. It is based on the
mode control, start-up, voltage regulation.
fact that input signals for switching devices are continuous
I. INTRODUCTION functions of the converter state and the corresponding models
are differential equations with discontinuous right hand sides.
T HE use of hysteresis control in the regulation of power
supplies goes back in time to early years of modern power
electronics when conditions for stable limit cycles in a buck
Such dynamic systems are known to generate sliding modes
with state trajectories in the surfaces in the system state, where
the right hand sides of the differential equations undergo
converter were established [1]. The resulting variable discontinuities.
switching frequency and the required nonlinear analysis for a Used as an analytical tool, the sliding-mode control gives a
rigorous design relegated the use of this type of control to clear insight on important design criteria such a stability or
specific applications, the electrical architecture of satellites steady-state switching frequency. An extensive review of the
being the most representative. Nevertheless, hysteresis control state of the art of sliding-mode control in power converters
has attracted the attention of many researchers in the last can be found in [5], and practical design issues for the design
decade due to its fast response and robust behavior in front of of a voltage controlled buck converter are presented in [6].
parametric changes. It constitutes nowadays the base of many Another time domain/phase-plane approach for power
voltage regulation modules (VRMs) for the new generation of converter regulation employing a hysteresis comparator is the
microprocessor power supplies [2], it having also reached a boundary control, which can be considered in certain cases
high degree of standardization with the availability of some [7]-[9] as a minimum-time control, i.e., minimum time
recovery from parameter and control disturbances, as it is
Manuscript received July 27, 2011. demonstrated in the comparative analysis reported in [10].
L. Martínez-Salamero is with the Department of Electrical, Electronic and The basic problem underlying the boundary control is
Control Engineering from the Rovira i Virgili University, Tarragona, Spain (e-
mail: [email protected]). optimizing the converter transient state without establishing a
G. García is with the Methods and Algorithms in Control Group at the clear distinction between start-up and steady state. However, it
Laboratory for Analysis and Architecture of Systems, LAAS-CNRS, could be interesting to separate both regimes and give special
Toulouse, France.
M. Orellana, C. Lahore, B. Estibals and C. Alonso are with the System emphasis to the optimization problem in the start-up in order
Integration and Energy Management group at LAAS-CNRS, Toulouse, France to make a relatively simple controller. The problem of
C. E. Carrejo is with TOTAL S.A., Dept. Solar & New Energies Tour optimizing the start-up has been formulated in [11], where it is
Coupole -2 Place Jean Miller - La Defense 6, 92400 Courbevoie, France.
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 2

demonstrated that the control law minimizing the energy

employed in reaching the equilibrium point from zero-initial
conditions is an on-off law. f (6)
TOFF
In this paper, we interpret the control law reported in [11]
under the perspective of sliding-mode control theory. Our goal
Expressions (5) and (6) are equal for the set of points
is to study the sliding-motions that can be introduced in a buck
satisfying
converter by a switching surface made of a linear combination
of inductor current and capacitor voltage errors with respect to (7)
their equilibrium point. Moreover, our aim is to propose a
which constitutes the equilibrium point locus of the buck
switching surface that optimizes the start-up trajectory in the
converter.
sense of [11] and presents a good compromise between inrush
On the other hand, the control law proposed in [11] can be
current and output voltage response rapidity. A last objective
summarized by
is to modify the switching surface for start-up once the
converter has reached the steady-state in order to guarantee If 0, the converter must switch to OFF state ( )
output voltage regulation.
If 0, the converter must switch to ON state ( )
II. EQUILIBRIUM POINT LOCUS where
Assuming that the set of parameters of the buck converter is (8)
given by ( , , , , ), where is the input voltage, and
are their inductance and capacitance respectively, is the and , are the coordinates of inductor current and
load resistance and is the parasitic resistance of the inductor, capacitor voltage in the desired equilibrium point ( , ).
the descriptive equations of the converter behavior during The limit between the two decision regions of the control
(subinterval for energy absorption from the input source) and law is given by 0, that can be also written as follows
(subinterval for energy transfer to the output load) are the c
vC (9)
following: c
where 0, 0 and
(1)
0.
Fig. 1 shows an interpretation of the control law for the
start-up of the buck converter. It can be expected that the state
(2)
vector trajectory follows a combination of the trajectories
corresponding to and intervals that lead the state
during and
vector from the origin of the plane at the initial instant to the
(3) desired equilibrium point ( , ).

(4)

during , where is the inductor current and is the

capacitor voltage.
We define the equilibrium point locus ( ) of the
switching converter as the set of points in the plane
that the converter state variables can take in steady-state. If
switching between subintervals and took place at
infinite frequency, there would be points in the plane wherein
the field vectors and corresponding to and
respectively would be opposite. Such points are equilibrium
points of the switching system [12].
The expressions of and in the buck converter are given Fig. 1. Start-up trajectory of the buck converter in the plane .
by
III. SLIDING-MODE CONTROL ANALYSIS
f (5) The previous graphical interpretation requires an analytical
TON justification that can be provided by studying the possible
induced sliding-modes in the converter. Equations (1)-(4) can
be written in compact form by introducing the control variable
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 3

, so that 1 during and 0 during . The where

resulting equations are the following (16)
(10) and is a positive definite matrix fulfilling
0, 1,2 (17)
(11) where is an auxiliary matrix.
The cost of the transition from to is bounded and
Introducing the invariance conditions [4] 0 and given by
0 in equations (10)-(11) yields the equivalent control
given by (18)
1 1 0 0

(12) The first term of (18) is directly related to the time integral
of the square error and can be interpreted as the energy wasted
in the trajectory during the start-up. This waste is an indirect
and the corresponding ideal dynamics can be expressed as measure of the trajectory in terms of overshoot, oscillations or
(13) stress in the passive components rather than a dissipative
iL t energy in itself. The second term in (18) corresponds to the
weighted Euclidean distance between start point and
where is the coordinate of the inductor current at equilibrium point, matrix being the weighting factor.
Interpreting (14) leads to two possible cases
the equilibrium point, and is a constant that allows to adjust
the initial value of the current. 1 ξ is minimum (19)
It has to be pointed out that (13) express the averaged
behavior of the current around the straight line 0. 2 ξ is minimum (20)
Reaching this line from zero-initial conditions is ensured by
the state trajectory in the first and the subsequent From (19) and (20) it is derived
switching policy of the control law. It can be also observed ξ (21)
that (13) corresponds to a stable dynamic behavior because
parameters , , and are positive. where
(22)
IV. START-UP CONTROL
Choosing the coefficients of brings the problems of Since in the buck converter and 0,
establishing a criterion that allows minimizing the energy expression (22) becomes
wasted during start-up and ensures at the same time the
(23)
stability of the equilibrium point. Although the stability is
guaranteed in the buck converter for any set of positive values Therefore,
of and compatibles with ( , ), a criterion for
optimizing the start-up should be proposed. 1 ξ 0
(Converter in ON state) (24)
If we review the control law reported in [11], we will find
that this law makes a system that switches between affine
systems present an equilibrium point ( , ) that is 2 ξ 0
globally and asymptotically stable. Besides, the time integral (Converter in OFF state) (25)
between zero and infinity of the square error associated to the
transition from the initial state to the equilibrium point is (24) and (25) can be expressed in equivalent form
bounded, and the corresponding bound can be minimized.
Such control law has been summarized in section II but its ξ
(26)
formal expression is reproduced as follows 0, 1 (Converter in ON state)

arg , 1,2 (14) ξ

(27)
where is the unit step signal, and , correspond to the 0, 2 (Converter in OFF state)
description of equations (1)-(4) in matrix form Note that (26) and (27) represents the on-off law that has
been interpreted under the optics of sliding-mode control
, 1,2 1 , 2
(15) theory in section III.
0 , ,
Also, assuming that matrix in (18) has the form
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 4

0 , which would represent a current control

(28) strategy for the start-up. The corresponding experimental
0
results for 3 are depicted in Fig. 3
So that coefficients and are used to weight the influence
of current or voltage in the mentioned waste of energy. In the
case of 1, a small coefficient of will allow a high peak
of current and a fast voltage response. In the opposite case of a
high value of will result in a smaller current peak and in a
slower convergence of the output voltage to its coordinate at
the equilibrium point. iL
Moreover, calculating the coefficients of the switching
surface requires knowing the elements of matrix as it
is shown in (23). Matrix , in turn, must satisfy the set of
inequalities given by (17), which leads to the formulation of
vC
the following convex programming problem [11]
: 0, 0, 1,2 (29)
where stands for matrix trace. Fig. 3. Start-up experimental response corresponding to
Fig. 2 shows MATLAB simulations of the converter start- 3 .
up for the set of parameters 20 , 100 ,
172 , 0.1Ω, 5Ω and equilibrium point coordinates V. OUTPUT VOLTAGE REGULATION
15 , 3 . Three different weighting cases are Is has to be pointed out that the switching surface employed
considered, which are characterized by the respective matrices for start-up can be also used to reject input voltage
0.15 0 0.075 0 perturbations. Note that an eventual variation in would
0 0.01667 0 0.01667 (30) modify the upper limits of given by , , as it is
0.001 0
0 0.01667 shown in Fig. 1, but it would not change the coordinates of the
equilibrium point. However, this is not the case of a load
perturbation, which results in a change of the slope of .
Fig. 4 shows the trajectory of the state vector in the plane
when a load perturbation appears. If no dynamic
correction is used, a decrease of will result in a trajectory
from to where will be the new equilibrium point
with a different voltage coordinate than . Modifying the
switching surface for start-up by

, 0 (31)

makes the state vector follow the trajectory ,

so that the new equilibrium point keeps the same voltage
coordinate than the original point. The required increase of
current in the output port due to the decrease of load resistance
is ensured by operating in the new equilibrium point where
the current coordinate satisfies the consuming required by the
load. A symmetrical situation occurs for an increase of the
load resistance as illustrated in the same figure where the
correction effects yields the transition .
The analysis of the buck converter in sliding-mode with the
switching surface given by (31) leads to the analytic
description of the output voltage recovery in the transient-state
Fig. 2. MATLAB simulation of buck converter start-up for different by means of the characteristic equation
values of , (top), (bottom).
1
Note that a good compromise between inrush current and 0 (32)
output voltage response rapidity is observed in case . Such
case corresponds practically to a switching surface of the form Therefore, and can be found by pole assignment in
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 5

expression (32).
vC

iL

Fig. 4. State vector trajectory in the plane for load Fig. 6. Experimental response to load perturbations of step type.
perturbations. between inrush current and output response rapidity is the
The experimental results in Fig. 5 shows the insensitivity of current control given by the switching surface ,
the output voltage to input perturbations when the output where is the current coordinate of the equilibrium point.
voltage is regulated by the same switching surface This surface has been proposed for start-up and for rejecting
1.5 employed for starting-up. Step perturbation of 6 input voltage perturbations. Output voltage regulation has
are superposed to the nominal value of 24 but they been accomplished by modifying once the converter is in
hardly change the desired output voltage of 15 . steady-state. The new switching surface is a modification of
that includes a correction for output voltage
regulation.
Vg
REFERENCES
[1] I. M. H. Bábaá, T. G. Wilson and Y. Yu, “Analytic Solutions of Limit
Cycles in a Feedback-Regulated Converter System with Hysteresis”,
IEEE Trans. Automat. Contr., vol. AC-13, no. 5, Oct. 1968, pp. 524-531.
vC [2] R. Miptakhutdinov, “Analysis of synchronous buck converter with
hysteretic controller at high slow-rate load current transients”, Proc.
iL High Frequency Power Conv. Conf., 1999, pp. 55-69.
[3] LM3489, Hysteretic PFET Buck controller with Enable Pin, Data Sheet,
National Semiconductor, 2006,
[4] V. Utkin, J. Guldner and J. Shi, “Sliding mode control in
electromechanical systems”, CRC Press, 2nd edition, Taylor and Francis
Group, 2009.
[5] S. C. Tan, Y. M. Lai and C. K. Tse, “General Design Issues of Sliding-
Mode Controllers in DC-DC Converters”, IEEE Trans. on Indust.
Electron., vol. 55, no. 3, March 2008, pp. 1160-1174.
Fig. 5. Experimental response to input voltage perturbations of step
[6] S. C. Tan, Y. M. Lai, M. K. H. Cheung, and C. K. Tse, ‘On the practical
type. design of a sliding mode voltage controlled buck converter’, IEEE
Trans. on Power Electr., vol. 20, no. 2, march 2005, pp. 425-437.
Fig. 6 shows the experimental response to load perturbation [7] A. Soto, A. de Castro, P. Alon, J. A. Cobos, J. Uceda, and A. Lofti,
of step type when the switching surface is given by (31). In “Analysis of the buck converter for scaling the supply voltage of digital
this case, the output load is pulsating and changes circuits”, IEEE Trans. Power Electron., vol 22, no. 6, Nov. 2007, pp.
3432-2443.
periodically between 10Ω and 5Ω, what results in a very small [8] K. K. S. Leung, and H. S. H. Chung, “Derivation of Second-Order
overshoot of the output voltage and a fast recovery of the Switching Surface in the Boundary Control of Buck Converters”, IEEE
steady-state value. The load perturbations are mainly absorbed Power Electron. Lett., vol. 2, no. 2, Jun. 2004, pp. 63-67.
[9] K. K. S. Leung, and H. S. H. Chung, “A Comparative Study of
by the inductor current, which changes periodically between Boundary Control with First and Second Order Switching Surfaces for
1.5 and 3.2 approximately. Buck Converters Operating in DCM”, IEEE Trans. Power Electron., vol
22, no. 4, Jul. 2007, pp. 1196-1209.
[10] G. E. Pitel and P. T. Krein, “Minimimum-time Transient Recovery for
VI. CONCLUSIONS DC-DC Converters Using Raster Control Surfaces”, IEEE Trans. on
Start-up in a buck converter has been analyzed in sliding- Power Electr., vol. 24, no. 12, December 2009, pp. 2692-2702.
[11] G. S. Daecto, J. C. Geromel, F. S. Garcia, and J. A. Pomilio, “Switched
mode operation with a switching surface made of a linear affine systems control design with application to DC-DC converters”,
combination of inductor current and capacitor voltage errors IET Control Theory and Applications, vol. 4, iss. 7, 2010, pp. 1201-
with respect to their corresponding equilibrium values. The 1210.
[12] L. Bernadero, A. El Aroudi, E. Toribio, G. Olivar and L. Martinez-
coefficients of the linear combination have been obtained
Salamero, “Characteristic curves for analyzing limit cycle behaviour in
using the LMI procedure reported in [11] in order to minimize switching converters”, Electron. Lett., vol 35, no. 9, April 1999, pp. 687-
the wasted energy during the start-up. A good compromise 689.

View publication stats