Switched Control of Interleaved Converters

Ilse Cervantes ∗ Senior Member IEEE, Angelica Mendoza-Torres† , A.Rafael Garcia-Cuevas † ,

Francisco J. Perez-Pinal‡ Member IEEE
∗
Hybrid Systems Laboratory
Applied Mathematics Division, IPICyT
Camino a la Presa San Jos 2055 Col. Lomas 4ta 78216 San Luis Potos, S. L. P., Mexico
email: ilse@ipicyt.edu.mx;fax:+52-444-834 2010 www.ipicyt.edu.mx
† Applied Mathematics Division, IPICyT
‡ Instituto Politecnico Nacional, CITEDI

Abstract—In this paper, a time varying switching control based control techniques, among others ([6]-[9]) that concerns
of interleaved converters is proposed. The controller has the about current sharing and voltage ripple, simultaneously. Such
feature of being simple and it is able to guarantee global voltage controllers have shown good performance but they have the
regulation. The controller is based on the piece-continuous model
of the converter and it works by switching between performance main drawback that the design procedure is complex, increas-
surfaces that depends on the state vector. The switching surfaces ing with the number of cells.
are chosen differently for each cell of the interleaved converter in In this paper, a hybrid controller is proposed that is able to
order to guarantee, both maximum current and voltage ripple. regulate voltage, to ensure balanced current sharing, as well as
Furthermore, a given phase between current signals can also maximum voltage and current ripple. The control methodology
be guaranteed. The switched control is illustrated in a 2-cell
2kw boost converter: however, it can be applied directly in any used in this work is based in the so called hybrid control
interleaved converter independently of the number of cells. approach, and uses a piece-wise description of the controller
for design purposes. The proposed control switches between
I. I NTRODUCTION these behaviors by means of a set of surfaces defined by
Interleaved converters have been widely used in medium to operation restrictions (i.e. maximum ripple, voltage regulation,
low power applications, mainly to increase the output current etc). The switching surfaces are time varying in order to
while reduced input current ripple. In interleave converters, ensure the obtainment of a given phase between current
the size of dynamic components of the converter (inductances signals (360/N ), as well as appropriate current balancing.
and capacitance) are considerably decreased with respect to The controller has the advantage of being simple even if the
classical approaches. Some uses reported in the literature are number of cells in the converter is high. Experimental work
in power sources of modern microprocessors, in active power- and simulations in a 2KW two cell boost converter are used
factor correction and lately, in electric and hybrid vehicles, to illustrate advantages and features of the proposed control
among many other applications. Interleaved converters use a strategy.
parallel design of N switching legs and N inductors (cells),
operating under the action of a drive signal. Usually, if an Ln Dn
interleaved converter has N layers, the drive signal with a
phase of 360/N is used to action each switch. Such strategy
of operation has the objective of minimizing voltage ripple
without increasing switching frequency.
L1 D1
However a main drawback of these converters is to en-
sure equilibrated current (power) sharing among the cells in
presence of disturbances and other perturbations. It has been
E
+
reported that when two modules with different characteristics - S1 Sn C R
and independent control are parallel connected, one of these
converters may operate in continuous mode and the other
in discontinuous mode, depending on the duty cycle [1].
Moreover, even if the number of operation combinations in
a parallel configuration increase with N, the choice of this Fig. 1. Boost interleaved converter
configurations become complex and involve.
There exist in the literature works concerning operation and
current sharing in this topologies (see [2], [3], [4]. A method II. P IECE -C ONTINUOUS DYNAMICAL D ESCRIPTION OF
of using hysteresis current control in a pair of boost converters THE C ONVERTER
with coupled inductors has been proposed in [5]. There exists In this work, a 2-cell boost is taken as a benchmark
also modern control strategies such as sliding mode or digital to illustrate the control methodology design (see Figure 1).

978-1-4244-2601-0/09/$25.00 ©2009 IEEE 1156

However, such methodology is straight forward implementable e.g. B(i, γ), since the voltage dynamics act as a stable filter,
to other topologies also, even with large number of cells. In with time constant (RC).
this way, let us consider a 2-cells boost converter working
Inspired in [10], the main contribution of this paper is to
in continuous conduction mode(CCM), then the equations
shown that by virtue of Remark 1, it is possible to prove
describing the converter dynamics are the following:
asymptotic stability of both current and voltage to a given
bounded set. Guaranteeing in this way, maximum voltage
ẋ = Aj x + B (1) and current ripple. Furthermore, such bounded set can be
chosen as small as desired, provided that the state vector
where x ∈ R3 , constituted by the vector of inductor currents
is available for measure and the existence of a switching
i = [x1 , x2 ] and the output voltage (v = x3 ). The translation
sequence such that iref ∈ R2 is globally attractive, where
vector B ∈ R3 that accounts for the input voltage is given
T iref = iref (vref ) is a given known map. Since the set can
by B = E/L1 , E/L2 0 ,. The continuous system is
be chosen arbitrary, current balancing can also be ensured.
subjected to discrete time event S that defines the current state
In Figure 2, it is possible to observe an schematic diagram
matrix Ai ∈ R3×3 and
of the main functions of the proposed controller; how these
A1 if S =0 objectives can be obtained, will be shown below.
A2 if S =1
(2) Maximum
A3 if S =2 Current Ripple
A4 if S =3 Current
Control Current
where S is an index that stands for the different switch config- Time Varying Limitation
Switching Surface
urations between stages. Configurations S = 0, 1 correspond Maximum
to operation states (a) and (d) in [1] respectively. On the other Hybrid Voltage Ripple
(By manipulating phase f)
hand, matrices Aj are defined as: Control
⎡ ⎤
0 0 0 Voltage Voltage
A1 = ⎣ 0 0 −1/L2 ⎦, Control Regulation

⎡ 0 1/C −1/CR ⎤
0 0 −1/L1 Fig. 2. Main objectives of the proposed controller
A2 = ⎣ 0 0 −0 ⎦,
1/C 0 −1/CR
⎡ ⎤ Theorem 1: Let σ be a switching sequence such that
0 0 0 M
θ m
l=1 l j l = 0 where θl is the time length of stay in
A3 = ⎣ 0 0 0 ⎦,
subsystem j and mj is the linear trajectory described by such
0 1/C −1/CR subsystem at the event j = 0, 1, 2, 3.... Let θl  0, and a
⎡ ⎤
0 0 −1/L1 given M > 0, then current trajectories are bounded and there
A4 = ⎣ 0 0 −1/L2 ⎦, exists a time T such that for all t > T , x1 ∈ B(i1,ref , β1 ),
1/C 1/C −1/CR x2 ∈ B(i2,ref , β2 ) and x3 ∈ B(vref , γ).

where E, L1 , L2 , C, R are the input voltage, inductances, In order to sketch the proof, first notice that the quantities mj
capacitance and resistive load values respectively. denotes the linear trajectories of the system under each of the
stages of the converters (all combinations of switches on and
Remark 1: If the source voltage is constant and the nominal
off). Furthermore,
M notice also that restriction in Theorem 1 is
frequency (1/RC) is small compared to the design switching
equivalent to l=1 θl mj l = Iref = [i1,ref , i2,ref ]. Such equa-
frequency (fs ), the output voltage remains constant for short
tion implies that the state
[x1 , x2 ] = Iref can be obtained bya
times (O(1/fs )). Hence by solving Eq (1), it can be seen
switching sequence σ = (θ0 , mj 0 ), (θ1 , mj 1 )...(θM , mj M ) .
that inductor current evolves describing a piece-wise linear
Let us denote such a sequence σ( A), then observe that
trajectory as shown in Figure 4 [1], where mof f,j = E−x 3

Lj σ( A) guarantees that for any x1 , x2 in a Domain D, one
E
0, and mon,j = Lj  0. where j = 1...no. of inductors. In can reach the reference value Iref , since any current value
this way, the current evolution is described by [x˙1 , x˙2 ]T = can be described as a function of values θl mj l . This is a
mj ∈ R2 for j = 1, 2, 3, 4. key observation that will be used to define a bounded set
Remark 2: Given the structure of Eq. 1 it can be observed x1 ∈ B(i1,ref , β1 ), x2 ∈ B(i2,ref , β2 ).
that the dynamics of the output voltage is affected by the At this point is worthy to notice that due to properties
current ij only if the j− switch goes off. Furthermore, if the of matrices A1 ...A4 , slopes mon,1 and mon,2 are
current vector i = [x1 , x2 ]T can be confined in a region of different if inductors L1 = L2 . This property is also
measure β e.g.B(i, β) (i.e. a ball centered in i and radius β), satisfy by mof f,1 and mof f,2 . Recall that mj =
voltage trajectories will be confined in an other bounded region {[mof f,1 , mof f,2 ]T , [mon,1 , mof f,2 ]T , [mof f,1 , mon,2 ]T ,

1157
[mon,1 , mon,2 ]T } On the other hand, it is also true that one voltage ripple without increasing the switching frequency. This
can obtain any current in a given inductor, independently on can be performed by ensuring a given phase between current
the other current. That is [x1 , x2 ] = Iref can be obtained signals, as pointed out before. Step changes in the boundaries
by a switching sequence σA that is constituted by two of the balls B(ij,ref , δj ) can be defined over the time length τ
independently (possible asynchronous) sequences σA,1 and and a distance l to ensure a suitable signal desynchronization.
σA,2 . This fact is illustrated in Figure 3. The modification of this
In the following the switching stabilizing sequence will boundary has to be performed is actually very simple and
be explicit defined, since the sequence in every switch is observe that if the increasing the length τ lead to a decrement
independent to the other, just one of them is discussed and in the required l as viceversa.
the other can be computed using the same procedure. To
this end, assume without lost of generality that the first time
the trajectories of x1 hit i1,ref subsystems mon,1 is active.
Such system will inevitably lead the current away from i1,ref
in mon,1 or mof f,1 direction. Let 0 < δ1 < β1 , then it
is clear that as time evolves, trajectories will hit the ball
x1 ∈ B(i1,ref , δ1 ) once this occurs we can apply again
sequence A (called B if initial conditions are in the boundary
B(i1,ref , δ1 )). This sequence can be repeated indefinitely to
ensure that x1 ∈ B(i1,ref , β1 ). Hence such a set consti-
tutes an invariant set where the trajectories are confined for
t > T1 and it is globally attractive since any initial condition
can be leaded to B(i1,ref , β1 ) by moving along mon,1 and
mof f,1 directions. Notice that the methodology to compute
the stabilizing sequence can be used for the other current
also, in this case, different initial condition in x2 and different
inductor value will lead to the computation of a different
sequence i.e. σA,2 that will lead to the system to the region
B(i2,ref , δ2 ), with 0 < δ2 < β2 . Once this ball is reached Fig. 3. Time evolution of currents under the proposed control scheme. Step
(at time t = T2 ), a second sequence σB,2 can be followed changes on switching surfaces can be performed in order to ensure ad hoc
current desynchronization (i.e. to introduce a given phase φ between current
to ensure that the current trajectories do not scape of the ball signals )
B(i2,ref , β2 ). At this point, one can conclude also voltage
confination in a region x3 ∈ B(vref , γ) in view of Remark
2 and the locally one-to-one map iref = Iref (vref ). On the Remark 4: Above it was pointed out that one can obtain
other hand, all the trajectories of the system will be confined any current reference in a given inductor independently on the
in a region B(xref , ) with   max{β1 , β2 , γ} for t > T value of the other current. That is, the switching sequence σB
with T  max{T1 , T2 , 2/RC} is constituted of two independently, possible asynchronous se-
Remark 3: Theorem 1 basically states that any current quences σB,1 and σB,2 . However if the inductors are equal, the
value in a domain D, can be leaded to a neighborhood of switching sequence of one inductor can be used to stabilize the
[Iref , Vr ef ]T of radius   max{β1 , β2 , γ} in finite time others. Furthermore, one can stabilize the other inductors (and
t > T with T  max{T1 , T2 , 2/RC}. In this way, the set therefore ensure confination) using i)any combination of the
B(x, ) is globally attractive and system trajectories will lead sequence of one of them (i.e. σB = ((θ0 , mon,1 ), (θ1 , mof f,1 ))
to it asymptotically. Moreover, Theorem 1 defines a family or σB = ((θ1 , mof f,1 ), (θ0 , mon,1 )) etc. or ii) any sequence
of switching sequences σ, such that trajectories of system in i) with a bounded time delay τ < τmax . If the system
(1)-(2) are confined for t  T , and ripple current converter is operating under a fixed frequency, such delay can be
is guarantee to be 2max{β1 , β2 } as maximum. The desired interpreted as the introduction of a phase ϕ in the switching
maximum ripple value can be arbitrary assigned, however if it sequence. Therefore, stabilization of the converter can be
is set very low, it may increase the switching frequency. If the performed using a feedback of one the inductors and the output
system is required to operate in fixed switching frequency, then voltage (to compute mof f,l and mon,l ). It is interesting to
it is also fixed the minimum value of β1 and β2 that can be notice that commonly, interleaved converters are operated with
reached. Observe also that the mj constitute switching surfaces a phase 360/N between driving signals. In the next section
that depends on the output voltage and system parameters. we will show that this election is only one of the possible
Furthermore, as sequence B is defined, a new definition of stabilizing sequences.
such slopes is performed every time the balls B(i1,ref , δ1 ), Remark 5: So far, it has been assumed that the map iref =
B(i2,ref , δ2 ) are reached. This idea of defining time varying iref (vref ) is known. However this is not a real assumption
switching surfaces is useful to guarantee a maximum value of in most applications. In order to overcome this limitation, a

1158
voltage loop can be used. In this way, a multi-loop control using SIMULINK, as well as of experimental work. The
scheme constituted of either, continuous or discrete voltage control law that gives as result sequences of on and off times,
loop, along with the proposed switching control can be used, can be executed with FPGA or DSP commercial boards, which
as illustrated in Figure 4. In such a scheme, current reference is constitute an advantage of the proposed methodology. A 2KW
fixed by virtue of the outer loop or master loop. This reference boost converter with the following parameters was designed
is used as a center of the confinement sets B(ij,ref , δj ) for and built in the Hybrid Systems Laboratory: L1 = L2 =
j = 1, 2...N . If the inductors of the converter are equal, 450e − 6H, c = 1.88e − 6F , the nominal resistance and
then the balanced case i1,ref = i2,ref ...iN,ref is of interest. frequency are R = 52.23Ω F = 40kHz. The experimental
However, note that the result in Theorem 1, also ensure set up is shown in Figure 5. The switching controller was
stability and balancing in the case that L1 = L2 = L3 ..., which implemented using a ad hoc designed digital circuit.
is actually the real case (since uncertainty and perturbation can
be present).
Time Varying
Voltage i ref Switched Switching v 120
-+ Regulator
vref Current Surface Converter i
Control 100
R esistance (Ω)
O utputVoltage (V)

80
Current Loop

60
Voltage Loop
40
Fig. 4. Block Diagram of the proposed multi-loop controller.
20

0.000 0.002 0.004 0.006 0.008 0.010

Tim e (s)

10

8
C urrent(A)

0
0.000 0.002 0.004 0.006 0.008 0.010
Tim e (s)

Fig. 6. Simulated time evolution of the converter with the proposed feedback,
under step load changes of 50 percent.

In Figure 6 it can be observed the simulated time evolution

of current and output voltage for a two cell boost converter.
Notice that voltage regulation is performed successfully. Fur-
Fig. 5. Experimental Set up thermore, in this case, the balanced case is solved by defining
i1,ref = i2,ref . Also observe that the current ripple is kept at
less 5%, while voltage rippling fits the 1-2% variation require-
III. I LLUSTRATIVE E XAMPLE ment, In Figure 6 it can also be observed the performance of
In this section, robustness and performance of the proposed the converter in face of load disturbance of about 50
hybrid controller is evaluated during transient and steady state The controller performance shown in simulations where

1159
obtained also experimentally, in this case Figures 7 and 8
shown the results. In Figure 7, it is possible to observe
confinement of the output voltage as well as the obtainment
of the maximum current and voltage ripple. Moreover, it
is possible to observe the phase between control signals as
discussed in Remark 3. Current confinement in different values
of current can also obtained as shown in Figure 8, where the
response of the controller is tested without any compensation
of phase. As observed, even if the proposed law guarantee its
stability, it is clear that controlling phasing between currents
allow to obtain lower maximum voltage rippling.

Fig. 9. Experimental time evolution of the inductor currents and output

voltage when no phase compensation is performed(no time varying switching
surface is used)

proposed. Performance of the proposed controller is evaluated

during transient and steady state conditions using simulations
and experimental work. The controller proposed can use a
simple structure and has the advantage of being easy to
design even if the number of cells in the converter is high,
furthermore, it the phase between current signals can also
be controlled in order to limit voltage ripple without using
excessively high frequency signals.
ACKNOWLEDGMENT
Fig. 7. Experimental time evolution of output voltage and inductor current Ilse Cervantes acknowledges financial support from CONA-
L1 CyT Grant No. 84299.
Special thanks are given to Irwin Allen Diaz for his collab-
oration in this work.
R EFERENCES
[1] P.W. Lee, Y.S. Lee, D. K. W. Cheng and X.C. Liu, Steady-State
Analysis of an Interleaved Boost Converter with Coupled Inductors,
IEEE Trans. Ind. Electron. VOL. 47, NO. 4, AUGUST 2000, pp. 787-
796
[2] M. S. Elmore, ”Input current ripple cancellation in synchronized,
parallel connected critically continuous boost converters”, in Proc. IEEE
APEC 96, vol. 1, San Jose, CA, Mar. 1996, pp. 152-158.
[3] B. A. Miwa, D. M. Otten, and M. F. Schlecht, ”High efficiency power
factor correction using interleaving techniques”, in Proc. IEEE APEC
92, vol. 1, Boston, MA, Feb. 1992, pp. 557-568.
[4] L. Balogh and R. Redl, ”Power-factor correction with interleaved
boost converters in continuous-inductor-current mode”, in Proc. IEEE
APEC, San Diego, CA, Mar. 1993, pp. 168-174.
[5] J. W. Kolar, G. R. Kamath, N. Mohan, and F. C. Zach, ”Self-
adjusting input current ripple cancellation of coupled parallel connected
hysteresis-controlled boost power factor correctors”, in Proc. IEEE
PESC’95, vol. 1, Atlanta, GA, June 1995, pp. 164-173.
Fig. 8. Experimental time evolution of the converter when phase compensa- [6] H. Bae, J. Lee; J. Yang, B. H. Cho, ”Digital Resistive Current (DRC)
tion is performed . From top to botom: control signals, inductor current L1 Control for the Parallel Interleaved DC-DC Converters”, IEEE Trans
and output voltage. Power Electron., Vol. 23, No. 5, Sept. 2008 pp.2465-2476
[7] R. Giral, L. Martinez-Salamero, R. Leyva, J. Maixe, J., ”Sliding-
mode control of interleaved boost converters”, IEEE Trans. Circuits and
Systems I: Fundamental Theory and Applications, Vol. 47, No 9, Sept.
IV. C ONCLUSIONS 2000 pp.1330-1339.
[8] L. Huber, B.T. Irving, M.M. Jovanovic, ”Open-Loop Control Methods
In this work an hybrid control that is able to ensure volt- for Interleaved DCM/CCM Boundary Boost PFC Converters”, IEEE
age regulation and maximum current and voltage rippling is Trans Power Electronics, Vol 23, No 4, July 2008 pp.1649-1657.

1160
[9] T. Kohama, T. Ninomiya, ”Automatic interleaving control for paral-
leled converter system and its ripple estimation with simplified circuit
model”, 7th Internatonal Conference on Power Electronics, 2007. ICPE
07. pp.238-242.
[10] X. Xuping, Practical stabilizability of a class of switched systems.
Proc ACC, Boston 2004,pp. 4537-4542, 2004.

1161