A Unified Model for Filter Capacitor Constraints
in DC to DC Switching Converters Design
A. Cantillo, A. De Nardo, N. Femia, W. Zamboni
Dipartimento di Ingegneria dell’Informazione ed Ingegneria Elettrica (DIIIE)
Università degli Studi di Salerno
Fisciano (SA), Italy
{acantillo,adenardo,femia,wzamboni}@unisa.it
Abstract— This paper discusses a unified model for the
determination of the constraint equations to be used in the
selection of feasible capacitors for DC to DC converters. The
model is based on a generalized waveform of the current flowing
through capacitors, which provides the maximum acceptable
ESR as a function of the capacitance C ensuring the compliance
with ripple and constraints determined by the application.
Design examples are presented regarding the application of the
proposed model to boost, buck-boost and SEPIC topologies.
I. INTRODUCTION
The selection of filter capacitors is one of the major steps in
the design of the power section of switching converters. The
choice of filter capacitors may highly influence the size, the
cost and the overall static and dynamic behavior of the
converter. One of the main goals is to keep the filter capacitor
as small as possible, compatibly with reliability and
robustness needs. The maximum value ESR
max
and the
minimum value C
min
can be determined by means of
simplified approaches, like the ones indicated in textbooks [1-
2], scientific papers [3-4], and IC manufacturers technical
literature [5], which provide very simple formulas, where
ESR
max
and C
min
are independent. For example, referring to
boost converter, if Δv
opp
and Δv
ipp
are respectively the
maximum allowed output and input ripple, Δi
Lpp
and I
o
are
ripple inductor current and load current, f
s
is the switching
frequency, D is the duty-cycle, the following constraints are
usually adopted for selecting input capacitors:
Lpp
ipp
i
i
v
ESR
Δ
Δ
=
max
;
ipp s
Lpp
i
v f
i
C
Δ
Δ
=
8
min
(1)
and output capacitor:
2 / ' /
2
1
max
Lpp o
opp
o
i D I
v
ESR
Δ +
Δ
= ;
opp s
o
i
v f
DI
C
Δ
=
2 / 1
min
(2)
Some inconveniencies may derive from the use of such
formulas. As an example, for the input capacitor, formulas (1)
are determined by assuming that the full voltage ripple is
sustained both by the ESR and by the capacitance C as the
two components of the ripple determined by ESR and C are
orthogonal. Such assumption may lead to an undersizing of
the capacitor. For the output capacitor, instead, formulas (2)
are based on the assumption that the total voltage ripple is
equally divided between the ESR and the capacitance C. This
may lead to an oversizing of the capacitor. In general,
simplified formulas imply an a priori allocation of a certain
fraction of ripple to the ESR and to the capacitance C. In [6-8]
it was shown that an accurate ripple analysis allows to
consider the correlation existing between ESR and C to get
more accurate and reliable design formulas wherein ESR
max
is
a function of C. In this paper, that approach is generalized by
introducing a unified model based on a general definition of
the current waveform flowing in the capacitor which covers
all the applications related to DC to DC converters. Thus, the
peak-to-peak ripple voltage is determined as a function of
ESR and C. Inverting the ripple equation provides the exact
expression of ESR
max
as a function of C. Examples of such
modeling approach are presented in the paper, regarding
buck, buck-boost and SEPIC topologies.
II. GENERALIZED CAPACITOR WAVEFORMS
Let us consider DC to DC converters in steady-state
continuous conduction mode, characterized by a switching
period T
s
and a duty-cycle D. Let T
1
and T
2
be the time
intervals [0, DT
s
] and [DT
s
, T
s
], respectively. It is reasonable
to assume that:
i) the currents flowing through the inductors are
continuous and piecewise linear functions of time;
ii) the currents flowing through the switches vanish either
in T
1
or in T
2
.
Additionally, we make the following hypothesis:
iii) the switches are ideal (they are equivalent to a short-
circuit in T
1
and to an open-circuit in T
2
, or vice-
versa);
iv) input voltage source is ideal and the converter input
current is piecewise linear;
978-1-4244-5091-6/09/$25.00 ©2009 IEEE 295