AC Equivalent Circuit Modeling

Chapter 7
AC Equivalent Circuit Modeling
7 . 1 . Introduction
Converter systems invariably require feedback. For example, in a typical dc-dc
converter application, the output voltage v(t) must be kept constant, regardless of changes
in the input voltage vg(t) or in the effective load resistance R. This is accomplished by
building a circuit which varies the converter control input (i.e., the duty cycle d(t)) in such
a way that the output voltage v(t) is regulated to be equal to a desired reference value vref. In
inverter systems, a feedback loop causes the output voltage to follow a sinusoidal reference
voltage. In modern low-harmonic rectifier systems, a control system causes the converter
input current to be proportional to the input voltage, such that the input port presents a
resistive load to the ac source.
Power
Switching converter
Load
So feedback is commonly input
+
employed.
A typical dc-dc system vg(t) +
v(t) R
feedback
incorporating a buck converter
connection
and feedback loop block

transistor
gate driver
diagram is illustrated in Fig.
compensator
v
pulse-width vc
(t)
Gc(s)
7.1. It is desired to design this
modulator
feedback system in such a way
voltage
vc(t)
(t)
reference vref
that the output voltage is
dTs Ts
t
t
accurately regulated, and is
Controller
insensitive to disturbances in Fig. 7.1. A simple dc-dc regulator system, including a buck
converter power stage and a feedback network.
vg(t) or in the load current.
Additionally, the feedback
system should be stable, and properties such as transient overshoot, settling time, and
steady-state regulation should meet specifications. The ac modeling and design of
converters and their control systems such as Fig. 7.1 is the subject of part 2 of this book.
+
version 1/14/98 11:00 AM
Chapter 7. AC Equivalent Circuit Modeling
To design the system of Fig. 7.1, we need a dynamic model of the switching
converter. How do variations in the power input voltage, the load current, or the duty
cycle, affect the output voltage? What are the small-signal transfer functions? To answer
these questions, we will extend the steady-state models developed in chapters 2 and 3 to
include the dynamics introduced by the inductors and capacitors of the converter. Dynamics
of converters operating in the continuous conduction mode can be modeled using
techniques quite similar to those of chapters 2 and 3; the resulting ac equivalent circuits bear
a strong resemblance to the dc equivalent circuits derived in chapter 3.
Modeling is the representation of physical phenomena by mathematical means. In
engineering, it is desired to model the important dominant behavior of a system, while
neglecting other insignificant phenomena. Simplified terminal equations of the component
elements are used, and many aspects of the system response are neglected altogether, i.e.,
they are unmodeled. The resulting simplified model yields physical insight into the
system behavior, which aids the engineer in designing the system to operate in a given
specified manner. Thus, the modeling process involves use of approximations to neglect
small but complicating phenomena, in an attempt to understand what is most important.
Once this basic insight is gained, it may be desirable to carefully refine the model, by
accounting for some of the previously ignored phenomena. It is a fact of life that real,
physical systems are complex, and their detailed analysis can easily lead to an intractable
and useless mathematical mess. Approximate models are an important tool for gaining
understanding and physical insight.
gate
As discussed in chapter 2, the drive
switching ripple is small in a well-designed
CCM converter. Hence, we should ignore the
switching ripple, and model only the
underlying ac variations in the converter
waveforms. For example, suppose that some
actual waveform v(t)
including ripple
ac variation is introduced into the converter
averaged waveform <v(t)>Ts
duty cycle d(t), such that
with ripple neglected
d(t) = D + Dm cos mt
(7-1)
where D and Dm are constants, | Dm | << D, and

the modulation frequency m is much smaller
than the converter switching frequency s =
2fs. The resulting transistor gate drive signal
is illustrated in Fig. 7.2(a), and a typical
Fig. 7.2. Ac variation of the converter signals:

transistor gate drive signal, in which the duty
cycle varies slowly, and the resulting converter
output voltage waveform. Both the actual
waveform v(t) (including high frequency
switching ripple) and its averaged, or lowfrequency component, < v(t) >T are illustrated.
s
modulation
switching
switching
frequency and its

frequency and
harmonics
converter output voltage waveform spectrum
of v(t)
harmonics
sidebands
v(t) is illustrated in Fig. 7.2(b).
The spectrum of v(t) is illustrated in
Fig. 7.3. This spectrum contains
m
s
components at the switching

Fig. 7.3. Spectrum of the output voltage waveform v(t)
frequency as well as its harmonics
of Fig. 7.2.
and sidebands; these components
are small in magnitude if the switching ripple is small. In addition, the spectrum contains a
low-frequency component at the modulation frequency m. The magnitude and phase of
this component depend not only on the duty cycle variation, but also on the frequency
response of the converter. If we neglect the switching ripple, then this low-frequency
component remains (also illustrated in Fig. 7.2(b)). The objective of our ac modeling
efforts is to predict this low-frequency component.
A simple method for deriving the small-signal model of CCM converters is
explained in section 7.2. The switching ripples in the inductor current and capacitor voltage
waveforms are removed by averaging over one switching period. Hence, the lowfrequency components of the inductor and capacitor waveforms are modeled by equations
of the form
d i L(t)
dt
d vC(t)
Ts
= vL(t)
Ts
s
= i C(t) T
s
dt
where < x(t) >Ts denotes the average of x(t) over an interval of length Ts:
xL(t)
= 1
Ts
Ts
t + Ts
(7-2)
x() d
(7-3)
So we will employ the basic approximation of removing the high-frequency switching
ripple by averaging over one switching period. Yet the average value is allowed to vary
from one switching period to the next, such that low-frequency variations are modeled. In
effect, the moving average of Eq. (7-3) constitutes low-pass filtering of the waveform. A
few of the numerous references on averaged modeling of switching converters are listed at
the end of this chapter [1-13].
Note that the principles of inductor volt-second balance and capacitor charge
balance predict that the right-hand sides of Eqs. (7-2) are zero when the converter operates
in equilibrium. Equations (7-2) describe how the inductor currents and capacitor voltages
t
change when nonzero average inductor voltage and capacitor current are applied over a
switching period.
The averaged inductor voltage and capacitor currents of Eq. (7-2) are, in general,
nonlinear functions of the signals in the converter, and hence Eqs. (7-2) constitute a set of
nonlinear differential equations. Indeed, the spectrum in Fig. 7.3 also contains harmonics
of the modulation frequency m. In most converters, these harmonics become significant in
magnitude as the modulation frequency m approaches the switching frequency s, or as
the modulation amplitude Dm approaches the quiescent duty cycle D. Nonlinear elements
are not uncommon in electrical engineering; indeed, all semiconductor devices exhibit
nonlinear behavior. To obtain a linear model a)
b)
C
C
that is easier to analyze, we usually construct
a small-signal model that has been linearized
FiB
FiB
about a quiescent operating point, in which
iB
iB
the harmonics of the modulation or
B
B
RiB
rE
excitation frequency are neglected. For
example, the Ebers-Moll model of Fig.
E
E
7.4(a) describes the nonlinear characteristics
7.4. Equivalent circuit modeling of the bipolar
of the bipolar junction transistor. For Fig. junction
transistor: (a) the nonlinear Ebers-Moll
operation in the active region, the familiar
model; (b) a linearized small-signal model.
small-signal model of Fig. 7.4(b) can be derived. The current flowing in the reverse-biased
collector-base diode has been ignored, while resistor rE represents the slope of the forwardbiased base-emitter diode characteristic, evaluated at the quiescent operating point. The
model can be refined if necessary, by addition of capacitances and other elements which
represent other physical processes in the transistor.
An example of a nonlinear converter characteristic is the dependence of the steadyD
state output voltage V of the buck-boost converter on
0
0.5
1
0
the duty cycle D, illustrated in Fig. 7.5. Suppose that
quiescent
the converter operates with some dc output voltage,
operating
point
say, V = V g , corresponding to a quiescent duty
cycle of D = 0.5. Duty cycle variations about this Vg
linearized
function
quiescent value will excite variations in the output
actual
voltage. If the magnitude of the duty cycle variation is
nonlinear
characteristic
sufficiently small, then we can compute the resulting V
output voltage variations by linearizing the curve. The
Fig. 7.5. Linearization of the static
control-to-output characteristic of
slope of the linearized characteristic in Fig. 7.5 is
the buck-boost converter about the
chosen to be equal to the slope of the actual nonlinear
quiescent operating point D = 0.5.
4
Vg V d(t)
1:D
D' : 1
+
vg(t)
I d(t)
I d(t)
v(t)
Fig. 7.6. Small-signal ac equivalent circuit model of the buck-boost converter.
characteristic, at the quiescent operating point; this slope is the dc control-to-output gain of
the converter. The linearized and nonlinear characteristics are approximately equal in value
provided that the duty cycle variations are sufficiently small.
Although it illustrates the process of small-signal linearization, the buck-boost
example of Fig. 7.5 is oversimplified. The inductors and capacitors of the converter cause
the gain to exhibit a frequency response. To correctly predict the poles and zeroes of the
small-signal transfer functions, we must linearize the converter averaged differential
equations, Eqs. (7-2). This is done in section 7.2. A small-signal ac equivalent circuit can
then be constructed using the methods developed in chapter 3. The resulting small-signal
model of the buck-boost converter is illustrated in Fig. 7.6; this model can be solved using
conventional circuit analysis techniques, to find the small-signal transfer functions, output
impedance, and other frequency-dependent properties. In systems such as Fig. 7.1, the
equivalent circuit model can be inserted in place of the converter. When small-signal
models of the other system elements (such as the pulse-width modulator) are inserted, then
a complete linearized system model is obtained. This model can be analyzed using standard
linear techniques, such as the Laplace transform, to gain insight into the behavior and
properties of the system.
Two well-known variants of the ac modeling method, state-space averaging and
circuit averaging, are explained in sections 7.4 and 7.5. An extension of circuit averaging,
known as averaged switch modeling, is also discussed in section 7.5. Since the switches
are the only elements which introduce switching harmonics, equivalent circuit models can
be derived by averaging only the switch waveforms. The averaged switch modeling
technique is easily extended to other modes of operation such as the discontinuous
conduction mode, as well as to current programmed control and to resonant converters. In
section 7.6, it is shown that the small-signal model of any dc-dc PWM CCM converter can
be written in a standard form. Called the canonical model, this equivalent circuit describes
the basic physical functions that any of these converters must perform. A simple model of
the pulse-width modulator circuit is described in section 7.7.
These models are useless if you dont know how to apply them. So in chapter 8,
the frequency response of converters is explored, in a design-oriented and detailed manner.
Small-signal transfer functions of the basic converters are tabulated. Bode plots of
converter transfer functions and impedances are derived in a simple, approximate manner,
which allows insight to be gained into the origins of the frequency response of complex
converter systems.
These results are used to design converter control systems in chapter 9. The
modeling techniques are extended in chapters 10 and 11 to cover the discontinuous
conduction mode and the current programmed mode.
7 . 2 . The basic ac modeling approach
1
2
Let us derive a small-signal ac model
i(t)
of the buck-boost converter of Fig. 7.7. The v (t) +
C
R
g
L
analysis begins as usual, by determining the
voltage and current waveforms of the inductor
and capacitor. When the switch is in position Fig. 7.7. Buck-boost converter example.
1, the circuit of Fig. 7.8(a) is obtained. The inductor voltage and capacitor current are:
vL(t) = L
di(t)
= vg(t)
dt
i C(t) = C
dv(t)
v(t)
=
R
dt
+
v(t)
(7-4)
(7-5)
We now make the small-ripple approximation. But rather than replacing vg(t) and v(t) with
their dc components Vg and V as in chapter 2, we now replace them with their lowfrequency averaged values < vg(t) >Ts and < v(t) >Ts, defined by Eq. (7-3). Equations (74) and (7-5) then become
vL(t) = L
di(t)
vg(t)
dt
i C(t) = C
v(t)
dv(t)
R
dt
(7-6)
Ts
Ts
(7-7)
Hence, during the first subinterval, the inductor current i(t) and the capacitor voltage v(t)
a)
b)
+
i(t)
vg(t)
i(t)
C
v(t)
vg(t)
v(t)
Fig. 7.8. Buck-boost converter circuit: (a) when the switch is in position 1,
(b) when the switch is in position 2.
change with the essentially constant slopes given by Eqs. (7-6) and (7-7). With the switch
in position 2, the circuit of Fig. 7.8(b) is obtained. Its inductor voltage and capacitor
current are:
vL(t) = L
di(t)
= v(t)
dt
i C(t) = C
dv(t)
v(t)
= i(t)
R
dt
vL(t) = L
di(t)
v(t)
dt
i C(t) = C
dv(t)
i(t)
dt
(7-8)
(7-9)
Use of the small-ripple approximation, to replace i(t) and v(t) with their averaged values,
yields
Ts
(7-10)
Ts
v(t)
R
Ts
(7-11)
During the second subinterval, the inductor current and capacitor voltage change with the
essentially constant slopes given by Eqs. (7-10) and (7-11).
7.2.1.
Averaging the inductor waveforms

The inductor voltage and current waveforms are sketched in Fig. 7.9. The lowfrequency average of the inductor voltage is found by evaluation of Eq. (7-3) the
inductor voltage during the first and second subintervals, given by Eqs. (7-6) and (7-10),
are averaged:
vL(t)
= 1
Ts
Ts
t + Ts
vL() d d(t) vg(t)
Ts
+ d'(t) v(t)
Ts
(7-12)
The right-hand side of Eq. (7-12) contains no switching harmonics, and models only the
low-frequency components of the inductor voltage waveform. Insertion of this equation
into Eq. (7-2) leads to
d i(t)
L
Ts
dt
= d(t) vg(t)
Ts
+ d'(t) v(t)
a)
Ts
(7-13)
b)
i(t)
vL(t)
vg(t)
vL(t)
i(dTs)
Ts
Ts
= d vg(t)
Ts
+ d' v(t)
dTs
Ts
v(t)
i(0)
Ts
vg
Ts
t
0
Ts
dTs
i(Ts)
Ts
Ts
Fig. 7.9. Buck-boost converter waveforms: (a) inductor voltage, (b) inductor current.
This equation describes how the low-frequency components of the inductor current vary
with time.
7.2.2.
Discussion of the averaging approximation

In steady-state, the actual inductor current waveform i(t) is periodic with period
equal to the switching period Ts: i(t + Ts) = i(t). During transients, there is a net change in
i(t) over one switching period. This net change in inductor current is correctly predicted by
use of the average inductor voltage. We can show that this is true, based on the basic
inductor equation
di(t)
= vL(t)
dt
(7-14)
Divide by L, and integrate both sides from t to t + Ts:

t + Ts
di = 1
L
t + Ts
vL() d
(7-15)
The left-hand side of Eq. (7-15) is i(t + Ts) i(t), while the right-hand side can be
expressed in terms of the definition of < v L (t) > T s, Eq. (7-3), by multiplying and dividing
t
by Ts to obtain
i(t + Ts) i(t) = 1 Ts vL(t)

L
Ts
(7-16)
The left-hand side of Eq. (7-16) is the net change in inductor current over one complete
switching period. Equation (7-16) states that this change is exactly equal to the switching
period Ts multiplied by the average slope < vL(t) >Ts / L.
Equation (7-16) can be rearranged to obtain
i(t + Ts) i(t)

= vL(t)
Ts
Ts
(7-17)
Let us now define the derivative of < i(t) >Ts as
d i(t)
Ts
dt
i(t + Ts) i(t)

Ts
(7-18)
This resembles the well-known Euler formula, commonly used in the numerical analysis
field to approximate the derivative of a function. In the limit as the switching period Ts
tends to zero, this formula yields the exact derivative. Provided that variations in the
inductor current are sufficiently slow with respect to the switching period T s, a good
approximation is obtained for nonzero Ts. Substitution of Eq. (7-16) into (7-18) leads to
d i(t)
L
dt
Ts
= vL(t)
Ts
(7-19)
which coincides with Eq. (7-2).

Let us compute how the inductor current changes over one switching period in our
buck-boost example. The inductor current waveform is sketched in Fig. 7.9(b). Assume
that the inductor current begins at some arbitrary value i(0). During the first subinterval, the
inductor current changes with the essentially constant value given by Eq. (7-6). The value
at the end of the first subinterval is
vg(t)
i(dTs)
i(0)
dTs
Ts
( final value) = (initial value) + (length of interval) (average slope)
(7-20)
During the second subinterval, the inductor current changes with the essentially constant
value given by Eq. (7-10). Hence, the value at the end of the second subinterval is
i(Ts)
i(dTs)
v(t)
L
d'Ts
Ts
( final value) = (initial value) + (length of interval) (average slope)
(7-21)
By substitution of Eq. (7-20) into Eq. (7-21), we can express i(Ts) in terms of i(0),
i(Ts) = i(0) +
Ts
d(t) vg(t)
L
Ts
+ d'(t) v(t)
vL(t)
Ts
Ts
(7-22)
Equations (7-20) - (7-22) are illustrated in Fig.

actual waveform,
averaged waveform
including ripple
7.10. Equation (7-22) expresses the final value
i(t) T
i(t)
s
v(t)
v
(t)
g
i(Ts) directly in terms of i(0), without the
L
L
intermediate step of calculating i(DTs). This
i(Ts)
i(0)
equation can be interpreted in the same manner as
d vg(t) T + d' v(t) T
s
s
Eqs. (7-20) and (7-21): the final value i(Ts) is equal
L
to the initial value i(0), plus the length of the
t
0
dTs
Ts
interval Ts multiplied by the average slope < vL(t) Fig. 7.10. Use of the average slope to
predict how the inductor current
>Ts / L. But note that the interval length is chosen to
waveform changes over one switching

period. The actual waveform i(t) and its
low-frequency components < i(t) >T
coincide with the switching period, such that the

switching ripple is effectively removed. Also, the
s
are illustrated.
use of the average slope leads to correct prediction
of the final value i(Ts). It can be easily verified that, when Eq. (7-22) is inserted into Eq.
(7-18), the previous result (7-13) is obtained.
7.2.3.
Averaging the capacitor waveforms

A similar procedure leads to the capacitor dynamic equation. The capacitor voltage
and current waveforms are sketched in Fig. 7.11. The average capacitor current can be
found by averaging Eqs. (7-7) and (7-11); the result is
i C(t)
Ts
v(t)
R
= d(t)
Ts
+ d'(t) i(t)
Ts
v(t)
R
Ts
(7-23)
Upon inserting this equation into Eq. (7-2) and collecting terms, one obtains
d v(t)
C
Ts
dt
= d'(t) i(t)
Ts
v(t)
R
Ts
(7-24)
This is the basic averaged equation which describes dc and low-frequency ac variations in
the capacitor voltage.
a)
b)
iC(t)
v(t)
Ts
R
i C(t)
v(t)
0
t
Ts
v(t)
R
dTs
Ts
dTs
v(dTs)
v(0)
Ts
i(t)
Ts
v(t)
Ts
RC
Ts
v(t)
RC
Ts
v(t)
i(t)
Ts
v(Ts)
Ts
Fig. 7.11. Buck-boost converter waveforms: (a) capacitor current, (b) capacitor voltage.
7.2.4.
The average input current

In chapter 3, it was found to be necessary to write an additional equation which
models the dc component of the converter input current. This allowed the input port of the
converter to be modeled by the dc equivalent circuit. A similar procedure must be followed
here, so that low-frequency variations at the converter input port are modeled by the ac
equivalent circuit.
For the buck-boost converter example, the current ig(t) drawn by the converter from
the input source is equal to the inductor current i(t) during the first subinterval, and zero
during the second subinterval. By neglecting the inductor current ripple and replacing i(t)
with its averaged value < i(t) >Ts, we can express the input current as follows:
i(t)
i g(t) =
Ts
during subinterval 1
during subinterval 2
(7-25)
10
The input current waveform is illustrated in Fig.

7.12. Upon averaging over one switching period,
one obtains
i g(t)
= d(t) i(t)
Ts
(7-26)
This is the basic averaged equation which describes
dc and low-frequency ac variations in the converter
input current.
Ts
ig(t)
i(t)
Ts
i g(t)
Ts
0
dTs
Ts
t
Fig. 7.12. Buck-boost converter
waveforms: input source current ig(t).
7.2.5.
Perturbation and linearization

The buck-boost converter averaged equations, Eqs. (7-13), (7-24), and (7-26), are
collected below.
d i(t)
L
C
dt
d v(t)
Ts
Ts
dt
i g(t)
Ts
= d(t) vg(t)
Ts
= d'(t) i(t)
= d(t) i(t)
+ d'(t) v(t)
Ts
v(t)
R
Ts
Ts
Ts
(7-27)
These equations are nonlinear because they involve the multiplication of time-varying
quantities. For example, the capacitor current depends on the product of the control input
d(t) and the low-frequency component of the inductor current, < i(t) >Ts. Multiplication of
time-varying signals generates harmonics, and is a nonlinear process. Most of the
techniques of ac circuit analysis, such as the Laplace transform and other frequency-domain
methods, are not useful for nonlinear systems. So we need to linearize Eqs. (7-27) by
constructing a small-signal model.
Suppose that we drive the converter at some steady-state, or quiescent, duty ratio
d(t) = D, with quiescent input voltage vg(t) = V g . We know from our steady-state analysis
of chapters 2 and 3 that, after any transients have subsided, the inductor current < i(t) >Ts,
the capacitor voltage < v(t) >Ts, and the input current < ig(t) >Ts will reach the quiescent
values I, V, and Ig, respectively, where
V = D Vg
D'
I= V
D' R
Ig = D I
(7-28)
11
Equations (7-28) are derived as usual via the principles of inductor volt-second and
capacitor charge balance. They could also be derived from Eqs. (7-27) by noting that, in
steady-state, the derivatives must equal zero.
To construct a small-signal ac model at a quiescent operating point (I, V), one
assumes that the input voltage vg(t) and the duty cycle d(t) are equal to some given
quiescent values Vg and D, plus some superimposed small ac variations vg(t) and d(t) .
Hence, we have
vg(t) = Vg + vg(t)
d(t) = D + d(t)
(7-29)
In response to these inputs, and after any transients have subsided, the averaged inductor
current < i(t) >Ts, the averaged capacitor voltage < v(t) >Ts, and the averaged input current
< ig (t) > T s waveforms will be equal to the corresponding quiescent values I, V, and I g ,
plus some superimposed small ac variations i(t) , v(t) , and i g(t) :
i(t)
v(t)
i g(t)
Ts
Ts
= I + i(t)
= V + v(t)
= I g + i g(t)
(7-30)
With the assumptions that the ac variations are small in magnitude compared to the dc
quiescent values, or
Ts
vg(t) << Vg
d(t) << D
i(t) << I
v(t) << V
i g(t) << I g
(7-31)
then nonlinear equations (7-27) can be linearized. This is done by inserting Eqs. (7-29) and
(7-30) into Eq. (7-27). For the inductor equation, one obtains
d I + i(t)
= D + d(t) Vg + vg(t) + D' d(t) V + v(t)
dt
(7-32)
It should be noted that the complement of the duty cycle is given by
d'(t) = 1 d(t) = 1 D + d(t) = D' d(t)
(7-33)
D' = 1 D
The minus sign arises in the expression for d(t) because a d(t) variation that causes d(t) to
where
increase will cause d(t) to decrease.
12
By multiplying out Eq. (7-32) and collecting terms, one obtains

0
d I + d i(t) = DV + D'V + Dv (t) + D'v(t) + V V d(t) + d(t) v (t) v(t)
L
g
g
g
g
dt
dt
Dc terms
1 st order ac terms
(linear)
2 nd order ac terms
(nonlinear)
(7-34)
The derivative of I is zero, since I is by definition a dc (constant) term. For the purposes of
deriving a small-signal ac model, the dc terms can be considered known constant
quantities. On the right-hand side of Eq. (7-34), three types of terms arise:
Dc terms: These terms contain dc quantities only.
First-order ac terms: Each of these terms contains a single ac quantity, usually
multiplied by a constant coefficient such as a dc term. These terms are linear
functions of the ac variations.
Second-order ac terms: These terms contain the products of ac quantities. Hence
they are nonlinear, because they involve the multiplication of time-varying
signals.
It is desired to neglect the nonlinear ac terms. Provided that the small-signal assumption,
Eq. (7-31), is satisfied, then each of the second-order nonlinear terms is much smaller in
magnitude that one or more of the linear first-order ac terms. For example, the secondorder ac term d(t) vg(t) is much smaller in magnitude than the first-order ac term
D vg(t) whenever d(t) << D . So we can neglect the second-order terms. Also, by
definition (or by use of Eq. (7-28)), the dc terms on the right-hand side of the equation are
equal to the dc terms on the left-hand side, or zero.
We are left with the first-order ac terms on both sides of the equation. Hence,
d i(t)
= Dvg(t) + D'v(t) + Vg V d(t)
dt
d V + v(t)
V + v(t)
= D' d(t) I + i(t)
R
dt
(7-35)
This is the desired result: the small-signal linearized equation which describes variations in
the inductor current.
The capacitor equation can be linearized in a similar manner. Insertion of Eqs. (729) and (7-30) into the capacitor equation of Eq. (7-27) yields
Upon multiplying out Eq. (7-36) and collecting terms, one obtains
13
(7-36)
0
dV + dv(t) = D'I V + D'i(t) v(t) + Id(t) +
C
R
R
dt
dt
1 st order ac terms
(linear)
Dc terms
d(t)i(t)
2 nd order ac term
(nonlinear)
(7-37)
By neglecting the second-order terms, and noting that the dc terms on both sides of the
equation are equal, we again obtain a linearized first-order equation, containing only the
first-order ac terms of Eq. (7-37):
dv(t)
v(t)
= D'i(t)
+ Id(t)
R
dt
(7-38)
This is the desired small-signal linearized equation that describes variations in the capacitor
voltage.
Finally, the equation of the average input current is also linearized. Insertion of
Eqs. (7-29) and (7-30) into the input current equation of Eq. (7-27) yields
I g + i g(t) = D + d(t) I + i(t)
(7-39)
By collecting terms, we obtain
Ig
i g(t)
DI
Di(t) + Id(t)
d(t)i(t)
2 nd order ac term
(nonlinear)
(7-40)
We again neglect the second-order nonlinear terms. The dc terms on both sides of the
equation are equal. The remaining first-order linear ac terms are
Dc term
1 st order ac term
Dc term
i g(t) = Di(t) + Id(t)
1 st order ac terms
(linear)
(7-41)
This is the linearized small-signal equation which describes the low-frequency ac
components of the converter input current.
In summary, the nonlinear averaged equations of a switching converter can be
linearized about a quiescent operating point. The converter independent inputs are
expressed as constant (dc) values, plus small ac variations. In response, the converter
averaged waveforms assume similar forms. Insertion of Eqs. (7-29) and (7-30) into the
converter averaged nonlinear equations yields dc terms, linear ac terms, and nonlinear
terms. If the ac variations are sufficiently small in magnitude, then the nonlinear terms are
much smaller than the linear ac terms, and so can be neglected. The remaining linear ac
terms comprise the small-signal ac model of the converter.
14
7.2.6.
Construction of the small-signal equivalent circuit model

Equations (7-35), (7-38), and (7-41) are the small-signal ac description of the ideal
buck-boost converter, and are collected below:
d i(t)
= Dvg(t) + D'v(t) + Vg V d(t)
dt
dv(t)
v(t)
C
= D'i(t)
+ Id(t)
R
dt
L
(7-42)
In chapter 3, we collected the averaged dc equations of a converter, and reconstructed an
equivalent circuit that modeled the dc properties of the converter. We can use the same
procedure here, to construct averaged small-signal ac models of converters.
The inductor equation of (7-42), or Eq. (7-35), describes the voltages around a
loop containing the inductor. Indeed, this equation was derived by finding the inductor
voltage via loop analysis, then averaging, perturbing, and linearizing. So the equation
represents the voltages around a loop of the
Vg V d(t)
L
small-signal model, which contains the inductor.
+ d i(t)
L
dt
The loop current is the small-signal ac inductor
+
+ D' v(t)
D vg(t)
current (t). As illustrated in Fig. 7.13, the term L

i(t)
d(t)/dt represents the voltage across the inductor
L in the small-signal model. This voltage is equal
Fig. 7.13. Circuit equivalent to the smallsignal ac inductor loop equation of
to three other voltage terms. Dvg(t) and
Eq. (7-42) or (7-35).
D'v(t) represent dependent sources as shown.

These terms will be combined into ideal transformers. The term Vg V d(t) is driven by
the control input d(t) , and is represented by an independent source as shown.
The capacitor equation of (7-42), or Eq. (7-38), describes the currents flowing into
a node attached to the capacitor. This equation was derived by finding the capacitor current
via node analysis, then averaging, perturbing, and linearizing. Hence, this equation
describes the currents flowing into a node of the small-signal model, attached to the
capacitor. As illustrated in Fig. 7.14, the term C dv(t) dt represents the current flowing
dv(t)
v(t)
through capacitor C in the small-signal
+
C
dt
R
model. The capacitor voltage is v(t) .
D' i(t)
According to the equation, this current is

equal to three other terms. The term D(t)
represents a dependent source, which will
15
I d(t)
v(t)
Fig. 7.14. Circuit equivalent to the smallsignal ac capacitor node equation of Eq.
(7-42), or (7-38).
eventually be combined into an ideal transformer. The term v(t) R is recognized as the
current flowing through the load resistor in the small-signal model. The resistor is
connected in parallel with the capacitor, such that the ac voltage across resistor R is v(t) as
expected. The term Id(t) is driven by the control input d(t) , and is represented by an
independent source as shown.
Finally, the input current equation of (7-42),
or Eq. (7-41), describes the small-signal ac currents
g(t) drawn by the converter out of the input voltage
source vg(t) . This is a node equation which states
i g(t)
vg(t)
I d(t)
D i(t)
that g(t) is equal to the currents in two branches, as

Fig. 7.15. Circuit equivalent to the
illustrated in Fig. 7.15. The first branch,
small-signal ac input source
current equation of Eq. (7-42), or
corresponding to the D(t) term, is dependent on the
(7-41).
ac inductor current (t). Hence, we represent this term
using a dependent current source; this source will eventually be incorporated into an
effective transformer. The second branch, corresponding to the Id(t) term, is driven by the
control input d(t) , and is represented by an independent source as shown.
The circuits of Figs. 7.13, 7.14, and 7.15, are collected in Fig. 7.16(a). As
discussed in chapter 3, the dependent sources can be combined into effective ideal
transformers, as illustrated in Fig. 7.16(b). The sinusoid superimposed on the transformer
symbol indicates that the transformer is ideal, and is part of the averaged small-signal ac
model. So the effective dc transformer property of CCM dc-dc converters also influences
small-signal ac variations in the converter signals.
a)
i(t)
Vg V d(t)
+
+
vg(t)
I d(t)
D i(t)
D' v(t)
D vg(t)
D' i(t)
I d(t)
v(t)
b)
L
Vg V d(t)
1:D
D' : 1
+
vg(t)
I d(t)
I d(t)
v(t)
Fig. 7.16. Buck-boost converter small-signal ac equivalent circuit: (a) the circuits of Figs. 7.13-7.15,
collected together; (b) combination of dependent sources into effective ideal transformers, leading
to the final model.
16
The equivalent circuit of Fig. 7.16(b) can now be solved using techniques of
conventional linear circuit analysis, to find the converter transfer functions, input and
output impedances, etc. This is done in detail in the next chapter. Also, the model can be
refined by inclusion of losses and other nonidealities an example is given in section 7.3.
7.2.7.
Results for several basic converters

The equivalent circuit models for the buck, boost, and buck-boost converters
operating in the continuous conduction mode are summarized in Fig. 7.17. The buck and
boost converter models contain ideal transformers having turns ratios equal to the converter
conversion ratio. The buck-boost converter contains ideal transformers having buck and
boost conversion ratios; this is consistent with the derivation of section 6.1.2 of the buckboost converter as a cascade connection of buck and boost converters. These models can be
solved to find the converter transfer functions, input and output impedances, inductor
current variations, etc.
a)
Vg d(t)
1:D
i(t)
vg(t)
I d(t)
v(t)
b)
V d(t)
D' : 1
i(t)
vg(t)
I d(t)
v(t)
c)
L
Vg V d(t)
+
1:D
D' : 1
+
i(t)
vg(t)
I d(t)
I d(t)
v(t)
Fig. 7.17. Averaged small-signal ac models for several basic converters operating in continuous
conduction mode: (a) buck, (b) boost, (c) buck-boost.
17
7 . 3 . Example: a nonideal flyback converter

To illustrate that the techniques of
D1
1:n
the previous section are useful for
+
ig(t)
modeling a variety of
converter
L
C
R
v(t)
phenomena, let us next derive a smallv (t) +
signal ac equivalent circuit of a converter g

containing transformer isolation and
Q1
resistive losses. An isolated flyback
converter is illustrated in Fig. 7.18. The
Fig. 7.18. Flyback converter example.
flyback transformer has magnetizing
inductance L, referred to the primary winding, and turns ratio 1:n. MOSFET Q1 has onresistance Ron. Other loss elements, as well as the transformer leakage inductances and the
switching losses, are negligible. The ac modeling of this converter begins in a manner
similar to the dc converter analysis of section 6.3.4. The flyback transformer is replaced by
an equivalent circuit consisting of the a)
D1
ig(t)
magnetizing inductance L in parallel with
+
iC(t)
1:n
i(t)
+
an ideal transformer, as illustrated in Fig.
C
R
v(t)
vL(t)
L
7.19(a).
vg(t) +
ideal
During the first subinterval, when
MOSFET Q1 conducts, diode D1 is off.

Q1
The circuit then reduces to Fig. 7.19(b).
The inductor voltage vL(t), capacitor b)
transformer model
current iC(t), and converter input current
+
ig
i +
iC
ig(t) are:
1:n
vL(t) = vg(t) i(t) Ron

v(t)
i C(t) =
R
i g(t) = i(t)
(7-43)
We next make the small ripple
approximation, replacing the voltages and
currents with their average values as
defined by Eq. (7-3), to obtain
vg
vL
Ron
c)
transformer model
ig
=0
vg
+
vL
i/n
v/n
+
iC
1:n
Fig. 7.19. Flyback converter example: (a) incorporation

of transformer equivalent circuit, (b) circuit during
subinterval 1, (c) circuit during subinterval 2.
18
vL(t) = vg(t)
Ts
v(t)
R
i g(t) = i(t) T
i C(t) =
i(t)
Ts
Ron
Ts
(7-44)
During the second subinterval, MOSFET Q1 is off, diode D1 conducts, and the circuit of
Fig. 7.19(c) is obtained. Analysis of this circuit shows that the inductor voltage, capacitor
current, and input current are given by
v(t)
vL(t) = n
i(t) v(t)
i C(t) = n
R
i g(t) = 0
(7-45)
The small ripple approximation leads to
s
v(t) T
s
n
i(t) T
v(t)
i C(t) = n s
R
i g(t) = 0
vL(t) =
Ts
(7-46)
The inductor voltage and current waveforms are sketched in Fig. 7.20. The average
inductor voltage can now be found by averaging the waveform of Fig. 7.20(a) over one
switching period. The result is
vL(t)
Ts
= d(t)
vg(t)
Ts
i(t)
Ts
v(t)
n
Ron + d'(t)
Ts
(7-47)
By inserting this result into Eq. (7-19), we obtain the averaged inductor equation,
d i(t)
L
Ts
dt
= d(t) vg(t)
Ts
d(t) i(t)
a)
Ts
Ron d'(t)
v(t)
n
Ts
(7-48)
b)
vg(t)
i(t)
vL(t)
Ts
vg iRon
vL(t)
Ron i(t)
Ts
dTs
i(t)
Ts
Ts
v(t)
Ts
Ts
nL
t
0
v/n
dTs
Ts
Fig. 7.20. Inductor waveforms for the flyback example: (a) inductor voltage, (b)
inductor current.
19
a)
b)
i(t)
v(t)
iC(t)
i v
nR
i C(t)
v(t)
Ts
RC
v(t)
Ts
Ts
nC
v(t)
dTs
Ts
RC
v/R
Ts
Ts
dTs
Ts
Fig. 7.21. Capacitor waveforms for the flyback example: (a) capacitor current, (b)
capacitor voltage.
The capacitor waveforms are constructed in Fig. 7.21. The average capacitor
current is
i C(t)
Ts
= d(t)
v(t)
R
Ts
i(t)
n
+ d'(t)
Ts
v(t)
R
Ts
(7-49)
This leads to the averaged capacitor equation
d v(t)
C
Ts
dt
= d'(t)
i(t)
n
Ts
v(t)
R
Ts
(7-50)
The converter input current ig(t) is sketched in

Fig. 7.22. Its average is
i g(t)
Ts
= d(t) i(t)
Ts
(7-51)
d i(t)
dt
d v(t)
Ts
Ts
dt
i g(t)
Ts
= d(t) vg(t)
= d'(t)
i(t)
n
= d(t) i(t)
Ts
Ts
d(t) i(t)
v(t)
R
Ts
i(t)
Ts
i g(t)
The averaged converter equations (7-48), (7-50)

and (7-51) are collected below:
ig(t)
Ts
dTs
Ts
Fig. 7.22. Input source current

waveform, flyback example.
Ron d'(t)
v(t)
n
Ts
Ts
Ts
(7-52)
This is a nonlinear set of differential equations, and hence the next step is to perturb and
linearize, to construct the converter small-signal ac equations. We assume that the converter
input voltage vg(t) and duty cycle d(t) can be expressed as quiescent values plus small ac
variations, as follows:
vg(t) = Vg + vg(t)
d(t) = D + d(t)
(7-53)
20
In response to these inputs, and after all transients have decayed, the average converter
waveforms can also be expressed as quiescent values plus small ac variations:
i(t)
v(t)
i g(t)
Ts
Ts
Ts
= I + i(t)
= V + v(t)
= I g + i g(t)
(7-54)
With these substitutions, the large-signal averaged inductor equation becomes
d I + i(t)
= D + d(t) Vg + vg(t) D' d(t)
dt
V + v(t)
D + d(t) I + i(t) Ron
n
(7-55)
Upon multiplying this expression out and collecting terms, we obtain

0
d I + d i(t) = DV D' V DR I + Dv (t) D' v(t) + V + V IR d(t) DR
L
g
on
g
g
on
on
n
n
n
dt
dt
1 st order ac terms (linear)
Dc terms
v(t)
d(t)vg(t) + d(t) n d(t)i(t)Ron
2 nd order ac terms (nonlinear)

(7-56)
As usual, this equation contains three types of terms. The dc terms contain no time-varying
quantities. The first-order ac terms are linear functions of the ac variations in the circuit,
while the second-order ac terms are functions of the products of the ac variations. If the
small-signal assumptions of Eq. (7-31) are satisfied, then the second-order terms are much
smaller in magnitude that the first-order terms, and hence can be neglected. The dc terms
must satisfy
0 = DVg D' V
n DRonI
(7-57)
This result could also be derived by applying the principle of inductor volt-second balance
to the steady-state inductor voltage waveform. The first-order ac terms must satisfy
d i(t)
v(t)
= Dvg(t) D' n + Vg + V
n IRon d(t) DRon i(t)
dt
(7-58)
This is the linearized equation that describes ac variations in the inductor current.
Upon substitution of Eqs. (7-53) and (7-54) into the averaged capacitor equation
(7-52), one obtains
21
d V + v(t)
= D' d(t)
dt
V + v(t)
I + i(t)
n
R
(7-59)
By collecting terms, we obtain
D'i(t) v(t) Id(t)

V
0 + dv(t) = D'I
C dV
n R +
n R n
dt
dt
1 st order ac terms
(linear)
Dc terms
d(t)i(t)
n
2 nd order ac term
(nonlinear)
(7-60)
We neglect the second order terms. The dc terms of Eq. (7-60) must satisfy
V
0 = D'I
n R
(7-61)
This result could also be obtained by use of the principle of capacitor charge balance on the
steady-state capacitor current waveform. The first-order ac terms of Eq. (7-60) lead to the
small-signal ac capacitor equation
dv(t) D'i(t) v(t) Id(t)

= n
n
R
dt
(7-62)
Substitution of Eqs. (7-53) and (7-54) into the averaged input current equation (752) leads to
I g + i g(t) = D + d(t) I + i(t)
(7-63)
Upon collecting terms, we obtain
Ig
Dc term
i g(t)
1 st order ac term
DI
Dc term
Di(t) + Id(t)
1 st order ac terms
(linear)
d(t)i(t)
2 nd order ac term
(nonlinear)
(7-64)
The dc terms must satisfy
I g = DI
(7-65)
We neglect the second-order nonlinear terms of Eq. (7-64), leaving the following linearized
ac equation:
(7-66)
This result models the low-frequency ac variations in the converter input current.
The equations of the quiescent values, Eqs. (7-57), (7-61), and (7-65) are collected
below:
22
0 = DVg D' V
n DRonI
V
0 = D'I
n R
I g = DI
(7-67)
For given quiescent values of the input voltage Vg and duty cycle D, this system of
equations can be evaluated to find the quiescent output voltage V, inductor current I, and
input current dc component Ig. The results are then inserted into the small-signal ac
equations.
The small-signal ac equations, Eqs. (7-58), (7-62), and (7-66), are summarized
below:
d i(t)
v(t)
= Dvg(t) D' n + Vg + V
n IRon d(t) DRon i(t)
dt
dv(t) D'i(t) v(t) Id(t)
C
= n
n
R
dt
L
(7-68)
The final step is to construct an equivalent circuit which corresponds to these equations.
The inductor equation was derived by first writing loop equations, to find the
applied inductor voltage during each subinterval. These equations were then averaged,
perturbed, and linearized, to obtain Eq. (7-58). So this equation describes the small-signal
ac voltages around a loop containing the inductor. The loop current is the ac inductor
current (t). The quantity L d i(t) dt is the
d(t) Vg IRon + V
n
DRon
L
low-frequency ac voltage across the
+ d i(t)
L
inductor. The four terms on the right-hand
dt
+
+ D' v(t)
n
side of the equation are the voltages D vg(t)
i(t)
across the four other elements in the loop.
The terms Dvg(t) and D'v(t) n are Fig. 7.23. Circuit equivalent to the small-signal ac
inductor loop equation, Eq. (7-58).
dependent on voltages elsewhere in the
converter, and hence are represented as dependent sources in Fig. 7.23. The third term is
driven by the duty cycle variations d(t) and hence is represented as an independent source.
The fourth term, DRon(t), is a voltage that is proportional to the loop current (t). Hence
this term obeys Ohms law, with effective resistance DRon as shown in the figure. So the
influence of the MOSFET on-resistance on the converter small-signal transfer functions is
modeled by an effective resistance of value DRon.
Small-signal capacitor equation (7-62) leads to the equivalent circuit of Fig. 7.24.
The equation constitutes a node equation of the equivalent circuit model. It states that the
23
capacitor current C dv(t) dt is equal to three

other currents. The current D(t)/n depends
on a current elsewhere in the model, and
hence is represented by a dependent current
source. The term v(t) R is the ac
C
D' i(t)
n
I d(t)
n
dv(t)
dt
v(t)
R
v(t)
Fig. 7.24. Circuit equivalent to the small-signal

ac capacitor node equation, Eq. (7-62).
component of the load current, which we

model with a load resistance R connected in parallel with the capacitor. The last term is
driven by the duty cycle variations d(t) , and is modeled by an independent source.
The input port equation, Eq. (7-66), also constitutes a node equation. It describes
the small-signal ac current g(t), drawn by the
i g(t)
converter out of the input voltage source vg(t) .
There are two other terms in the equation. The term
D(t) is dependent on the inductor current ac variation
(t), and is represented with a dependent source. The
term Id(t) is driven by the control variations, and is
vg(t)
I d(t)
D i(t)
Fig. 7.25. Circuit equivalent to the

small-signal ac input source
current equation, Eq. (7-66).
modeled by an independent source. The equivalent

circuit for the input port is illustrated in Fig. 7.25.
The circuits of Figs. 7.23, 7.24, and 7.25 are combined in Fig. 7.26. The
dependent sources can be replaced by ideal transformers, leading to the equivalent circuit of
Fig. 7.27. This is the desired result: an equivalent circuit which models the low-frequency
small-signal variations in the converter waveforms. It can now be solved, using
conventional linear circuit analysis techniques, to find the converter transfer functions,
output impedance, and other ac quantities of interest.
L
i g(t)
DRon
d(t) Vg IRon + V
n
i(t)
vg(t)
I d(t)
D i(t)
D' v(t)
n
D vg(t)
D' i(t)
n
I d(t) C
n
v(t)
Fig. 7.26. The equivalent circuits of Figs. 7.23-7.25, collected together.

i g(t)
1:D
d(t) Vg IRon + V
n
L
i(t)
vg(t)
D' : n
+
DRon
I d(t) C
n
I d(t)
v(t)
Fig. 7.27. Small-signal ac equivalent circuit model of the flyback converter.
24
7 . 4 . State-space averaging*
A number of ac converter modeling techniques have appeared in the literature,
including the current-injected approach, circuit averaging, and the state-space averaging
method. Although the proponents of a given method may prefer to express the end result in
a specific form, the end results of nearly all methods are equivalent. And everybody will
agree that averaging and small-signal linearization are the key steps in modeling PWM
converters.
The state-space averaging approach [1,2] is described in this optional section. This
method is often cited in the literature, and it is included in this text so that interested
students can more easily understand the technical conference and journal papers of the
field. The state-space description of dynamical systems is a mainstay of modern control
theory; the state-space averaging method makes use of this description to derive the smallsignal averaged equations of PWM switching converters. The state-space averaging method
is otherwise identical to the procedure derived in section 7.2. Indeed, the procedure of
section 7.2 amounts to state-space averaging, but without the formality of writing the
equations in matrix form. A benefit of the state-space averaging procedure is the generality
of its result: a small-signal averaged model can always be obtained, provided that the state
equations of the original converter can be written.
Section 7.4.1 summarizes how to write the state equations of a network. The basic
results of state-space averaging are described in section 7.4.2, and a short derivation is
given in section 7.4.3. Section 7.4.4 contains an example, in which the state-space
averaging method is used to derive the quiescent dc and small-signal ac equations of a
buck-boost converter.
7.4.1.
The state equations of a network

The state-space description is a canonical form for writing the differential equations
that describe a system. For a linear network, the derivatives of the state variables are
expressed as linear combinations of the system independent inputs and the state variables
themselves. The physical state variables of a system are usually associated with the storage
of energy, and for a typical converter circuit, the physical state variables are the
independent inductor currents and capacitor voltages. Other typical state variables include
the position and velocity of a motor shaft. At a given point in time, the values of the state
variables depend on the previous history of the system, rather than on the present values of
the system inputs. To solve the differential equations of the system, the initial values of the
state variables must be specified. So if we know the state of a system, i.e., the values of all
*
Optional section
25
of the state variables, at a given time t0, and if we additionally know the system inputs, then
we can in principle solve the system state equations to find the system waveforms at any
future time.
The state equations of a system can be written in the compact matrix form of Eq. (769):
dx(t)
= A x(t) + B u(t)
dt
y(t) = C x(t) + E u(t)
K
(7-69)
Here, the state vector x(t) is a vector containing all of the state variables, i.e., the inductor
currents, capacitor voltages, etc. The input vector u(t) contains the independent inputs to
the system, such as the input voltage source vg(t). The derivative of the state vector is a
vector whose elements are equal to the derivatives of the corresponding elements of the
state vector:
x1(t)
x(t) = x2(t) ,
d x1(t)
dt
d x2(t)
dx(t)
=
dt
dt
(7-70)
In the standard form of Eq. (7-69), K is a matrix containing the values of capacitance,
inductance, and mutual inductance (if any), such that K dx(t) dt is a vector containing the
inductor winding voltages and capacitor currents. In other physical systems, K may
contain other quantities such as moment of inertia or mass. Equation (7-69) states that the
inductor voltages and capacitor currents of the system can be expressed as linear
combinations of the state variables and the independent inputs. The matrices A and B
contain constants of proportionality.
It may also be desired to compute other circuit waveforms that do not coincide with
the elements of the state vector x(t) or the input vector u(t). These other signals are, in
general, dependent waveforms that can be expressed as linear combinations of the elements
of the state vector and input vector. The vector y(t) is usually called the output vector. We
are free to place any dependent signal in this vector, regardless of whether the signal is
actually a physical output. The converter input current ig(t) is often chosen to be an element
of y(t). In the state equations (7-69), the elements of y(t) are expressed as a linear
combination of the elements of the x(t) and u(t) vectors. The matrices C and E contain
constants of proportionality.
26
i(t)
iR1(t)
iin(t)
R1
iC1(t)
C1
+ + v (t)
L
iC2(t)
C2
v1(t)
+
R2
v2(t)
R3
+
vout(t)
Fig. 7.28. Circuit example.
As an example, let us write the state equations of the circuit of Fig. 7.28. This
circuit contains two capacitors and an inductor, and hence the physical state variables are
the independent capacitor voltages v1(t) and v2(t), as well as the inductor current i(t). So we
can define the state vector as
v1(t)
x(t) = v2(t)
i(t)
(7-71)
Since there are no coupled inductors, the matrix K is diagonal, and simply contains the
values of capacitance and inductance:
C1 0 0
K = 0 C2 0
0 0 L
(7-72)
The circuit has one independent input, the current source iin(t). Hence we should define the
input vector as
u(t) = i in(t)
(7-73)
We are free to place any dependent signal in vector y(t). Suppose that we are interested in
also computing the voltage vout(t) and the current iR1(t). We can therefore define y(t) as
y(t) =
vout(t)
i R1(t)
(7-74)
To write the state equations in the canonical form of Eq. (7-69), we need to express
the inductor voltages and capacitor currents as linear combinations of the elements of x(t)
and u(t), i.e., as linear combinations of v1(t), v2(t), i(t), and iin(t).
The capacitor current iC1(t) is given by the node equation
i C1(t) = C 1
v (t)
dv1(t)
= i in(t) 1 i(t)
R
dt
(7-75)
This equation will become the top row of the matrix equation (7-69). The capacitor current
iC2(t) is given by the node equation,
27
i C2(t) = C 2
dv2(t)
v (t)
= i(t) 2
R2 + R3
dt
(7-76)
Note that we have been careful to express this current as a linear combination of the
elements of x(t) and u(t) alone. The inductor voltage is given by the loop equation,
vL(t) = L
di(t)
= v1(t) v2(t)
dt
(7-77)
Equations (7-75) - (7-77) can be written in the following matrix form:
C1 0 0
0 C2 0
0 0 L
dv1(t)
dt
dv2(t)
dt
di(t)
dt
dx(t)
dt
1
R1
R2 + R3
v1(t)
1
v2(t) + 0
0
i(t)
i in(t)
x(t)
u(t)
+ B
(7-78)
Matrices A and B are now known.
It is also necessary to express the elements of y(t) as linear combinations of the
elements of x(t) and u(t). By solution of the circuit of Fig. 7.28, vout(t) can be written in
terms of v2(t) as
R3
vout(t) = v2(t)
R2 + R3
(7-79)
Also, iR1(t) can be expressed in terms of v1(t) as
i R1(t) =
v1(t)
R1
(7-80)
By collecting Eqs. (7-79) and (7-80) into the standard matrix form of Eq. (7-69), we obtain
vout(t)
=
i R1(t)
y(t)
R3
R2 + R3
1
R1
v1(t)
v2(t) + 0
0
i(t)
i in(t)
x(t)
u(t)
+ E
We can now identify the matrices C and E as shown above.
28
(7-81)
It should be recognized that, starting in chapter 2, we have always begun the

analysis of converters by writing their state equations. We are now simply writing these
equations in matrix form.
7.4.2.
The basic state-space averaged model

Consider now that we are given a PWM converter, operating in the continuous
conduction mode. The converter circuit contains independent states which form the state
vector x(t), and the converter is driven by independent sources which form the input vector
u(t). During the first subinterval, when the switches are in position 1, the converter reduces
to a linear circuit that can be described by the following state equations:
dx(t)
= A 1 x(t) + B 1 u(t)
dt
y(t) = C 1 x(t) + E 1 u(t)
(7-82)
During the second subinterval, with the switches in position 2, the converter reduces to
another linear circuit whose state equations are
dx(t)
= A 2 x(t) + B 2 u(t)
dt
y(t) = C 2 x(t) + E 2 u(t)
(7-83)
During the two subintervals, the circuit elements are connected differently; therefore, the
respective state equation matrices A 1, B 1 , C 1, E 1, and A 2, B 2 , C 2, E2 may also differ.
Given these state equations, the result of state-space averaging is the state equations of the
equilibrium and small-signal ac models.
Provided that the natural frequencies of the converter, as well as the frequencies of
variations of the converter inputs, are much slower than the switching frequency, then the
state-space averaged model which describes the converter in equilibrium is
0=AX+BU
Y=CX+EU
(7-84)
where the averaged matrices are
A = D A 1 + D' A 2
B = D B 1 + D' B 2
C = D C 1 + D' C 2
E = D E 1 + D' E 2
(7-85)
The equilibrium dc components are
29
X = equilibrium (dc) state vector

U = equilibrium (dc) input vector
Y = equilibrium (dc) output vector
D = equilibrium (dc) duty cycle
(7-86)
Quantities defined in Eq. (7-86) represent the equilibrium values of the averaged vectors.
Equation (7-84) can be solved to find the equilibrium state and output vectors:
X = A 1 B U
Y = C A 1 B + E U
(7-87)
The state equations of the small-signal ac model are
dx(t)
= A x(t) + B u(t) +
dt
A 1 A 2 X + B 1 B 2 U d(t)
y(t) = C x(t) + E u(t) +
C 1 C 2 X + E 1 E 2 U d(t)
(7-88)
The quantities x(t) , u(t) , y(t) , and d(t) in Eq. (7-88) are small ac variations about the
equilibrium solution, or quiescent operating point, defined by Eqs. (7-84) - (7-87).
So if we can write the converter state equations, Eqs. (7-82) and (7-83), then we
can always find the averaged dc and small-signal ac models, by evaluation of Eqs. (7-84) (7-88).
7.4.3.
Discussion of the state-space averaging result

As in sections 7.1 and 7.2, the low-frequency components of the inductor currents
and capacitor voltages are modeled by averaging over an interval of length T s. Hence, we
can define the average of the state vector x(t) as
x(t)
= 1
Ts
Ts
t + Ts
x() d
(7-89)
The low-frequency components of the input and output vectors are modeled in a similar
manner. By averaging the inductor voltages and capacitor currents, one then obtains the
following low-frequency state equation:
d x(t)
K
dt
Ts
= d(t) A 1 + d'(t) A 2
x(t)
Ts
+ d(t) B 1 + d'(t) B 2
u(t)
Ts
(7-90)
This result is equivalent to Eq. (7-2).

For example, let us consider how the elements of the state vector x(t) change over
one switching period. During the first subinterval, with the switches in position 1, the
converter state equations are given by Eq. (7-82). Therefore, the elements of x(t) change
with the slopes K-1 (A1x(t) + B1u(t)). If we make the small ripple approximation, e.g., that
30
x(t) and u(t) do not change much over one switching period, then the slopes are essentially
constant and are approximately equal to
dx(t)
= K 1 A 1 x(t)
dt
Ts
+ B 1 u(t)
Ts
(7-91)
This assumption coincides with the requirements for small switching ripple in all elements
of x(t) and that variations in u(t) be slow compared to the switching frequency. If we
assume that the state vector is initially equal to x(0), then we can write
x(dTs) = x(0) +
final
value
initial
value
K 1 A 1 x(t)
dTs
interval
length
Ts
+ B 1 u(t)
Ts
slope
(7-92)
Similar arguments apply during the second subinterval. With the switch in position 2, the
state equations are given by Eq. (7-83). With the assumption of small ripple during this
subinterval, the state vector now changes with slope
dx(t)
= K 1 A 2 x(t)
dt
Ts
+ B 2 u(t)
Ts
(7-93)
The state vector at the end of the switching period is
x(Ts) = x(dTs) +
final
value
initial
value
d'Ts
K 1 A 2 x(t)
interval
length
Ts
+ B 2 u(t)
Ts
slope
(7-94)
Substitution of Eq. (7-92) into Eq. (7-94) allows us to determine x(Ts) in terms of x(0):
x(Ts) = x(0) + dTsK 1 A 1 x(t)
Ts
+ B 1 u(t)
Ts
+ d'TsK 1 A 2 x(t)
Ts
+ B 2 u(t)
Ts
(7-95)
Upon collecting terms, one obtains
x(Ts) = x(0) + TsK 1 d(t)A 1 + d'(t)A 2
x(t)
Ts
+ TsK 1 d(t)B 1 + d'(t)B 2
u(t)
Ts
(7-96)
Next, we approximate the derivative of < x(t) >Ts using the net change over one switching
period:
d x(t)
dt
Ts
x(Ts) x(0)
Ts
(7-97)
31
Substitution of Eq. (7-96) into (7-97) leads to
d x(t)
K
Ts
dt
= d(t) A 1 + d'(t) A 2
which is identical to Eq. (7-91).

This is the basic averaged model
which describes the converter
dynamics. It is nonlinear because
the control input d(t) is multiplied
by < x(t) >Ts and < u(t) >Ts.
x(t)
Ts
+ d(t) B 1 + d'(t) B 2
K 1 A 1 x
x(t)
Ts
+ B1 u
u(t)
Ts
(7-98)
K 1 A 2 x
Ts
Ts
+ B2 u
x(t)
x(0)
K 1 dA 1 + d'A 2
Ts
+ dB 1 + d'B 2
Ts
Ts
x(Ts)
Ts
dT
s
s
Variation of a typical element of
Fig. 7.29. How an element of the state vector, and its
x(t) and its average are illustrated
average, evolve over one switching period.
in Fig. 7.29.
It is also desired to find the low-frequency components of the output vector y(t) by
averaging. The vector y(t) is described by Eq.
y(t)
C 1 x(t) T + E 1 u(t) T
s
s
(7-82) for the first subinterval, and by Eq. (783) for the second subinterval. Hence, the y(t)
Ts
elements of y(t) may be discontinuous at the
C 2 x(t) T + E 2 u(t) T
switching transitions, as illustrated in Fig.
s
s
0
0
dTs
Ts
t
7.30. We can again remove the switching
Fig. 7.30. Averaging an element of the
harmonics by averaging over one switching
output vector y(t).
period; the result is
y(t)
Ts
= d(t) C 1 x(t)
Ts
+ E 1 u(t)
Ts
+ d'(t) C 2 x(t)
Ts
+ E 2 u(t)
Ts
(7-99)
Rearrangement of terms yields
y(t)
Ts
= d(t) C 1 + d'(t) C 2
x(t)
Ts
+ d(t) E 1 + d'(t) E 2
u(t)
Ts
(7-100)
This is again a nonlinear equation.

The averaged state equations, (7-98) and (7-100), are collected below:
d x(t)
K
dt
y(t)
Ts
Ts
= d(t) A 1 + d'(t) A 2
x(t)
= d(t) C 1 + d'(t) C 2
x(t)
+ d(t) B 1 + d'(t) B 2
u(t)
+ d(t) E 1 + d'(t) E 2
u(t)
Ts
Ts
Ts
Ts
(7-101)
The next step is the linearization of these equations about a quiescent operating point, to
construct a small-signal ac model. When dc inputs d(t) = D and u(t) = U are applied,
converter operates in equilibrium when the derivatives of all of the elements of < x(t) >Ts
32
are zero. Hence, by setting the derivative of < x(t) > T s to zero in Eq. (7-101), we can
define the converter quiescent operating point as the solution of
0=AX+BU
Y=CX+EU
(7-102)
where definitions (7-85) and (7-86) have been used. We now perturb and linearize the
converter waveforms about this quiescent operating point:
x(t)
u(t)
y(t)
Ts
Ts
Ts
= X + x(t)
= U + u(t)
= Y + y(t)
d(t) = D + d(t) d'(t) = D' d(t)
(7-103)
Here, u(t) and d(t) are small ac variations in the input vector and duty ratio. The vectors
x(t) and y(t) are the resulting small ac variations in the state and output vectors. We must
assume that these ac variations are much smaller than the quiescent values. In other words,
U >> u(t)
D >> d(t)
X >> x(t)
Y >> y(t)
(7-104)
Here, || x || denotes the norm of the vector x.

Substitution of Eq. (7-103) into Eq. (7-101) yields
d X+x(t)
= D+d(t) A 1 + D'd(t) A 2 X+x(t)
dt
+ D+d(t) B 1 + D'd(t) B 2 U+u(t)
Y+y(t) = D+d(t) C 1 + D'd(t) C 2 X+x(t)

+ D+d(t) E 1 + D'd(t) E 2 U+u(t)
(7-105)
The derivative dX dt is zero. By collecting terms, one obtains
33
dx(t)
dt
firstorder ac
= AX + BU + Ax(t) + Bu(t) +
dc terms
A 1 A 2 X + B 1 B 2 U d(t)
firstorder ac terms
+
A 1 A 2 x(t)d(t) + B 1 B 2 u(t)d(t)
secondorder nonlinear terms
Y+y(t)
= CX + EU + Cx(t) + Eu(t) +
dc + 1st order ac
dc terms
C 1 C 2 X + E 1 E 2 U d(t)
firstorder ac terms
+
C 1 C 2 x(t)d(t) + E 1 E 2 u(t)d(t)
secondorder nonlinear terms
(7-106)
Since the dc terms satisfy Eq. (7-102), they drop out of Eq. (7-106). Also, if the smallsignal assumption (7-104) is satisfied, then the second-order nonlinear terms of Eq. (7106) are small in magnitude compared to the first-order ac terms. We can therefore neglect
the nonlinear terms, to obtain the following linearized ac model:
dx(t)
= A x(t) + B u(t) +
dt
A 1 A 2 X + B 1 B 2 U d(t)
y(t) = C x(t) + E u(t) +
C 1 C 2 X + E 1 E 2 U d(t)
(7-107)
This is the desired result, which coincides with Eq. (7-87).

7.4.4.
Example: State-space averaging of a nonideal buck-boost converter

D1
Q1
ig(t)
Let us apply the state-space
+
averaging method to model the bucki(t)
boost converter of Fig. 7.31. We will vg(t) +
L
C
R
v(t)
model the conduction loss of MOSFET
Q1 by on-resistance Ron, and the forward

Fig. 7.31. Buck-boost converter example.
voltage drop of diode D1 by an
independent voltage source of value VD. It is desired to obtain a complete equivalent circuit,
which models both the input port and the output port of the converter.
The independent states of the converter are the inductor current i(t) and the capacitor
voltage v(t). Therefore, we should define the state vector x(t) as
34
a)
b)
VD
Ron
ig(t)
+
vg(t)
ig(t)
i(t)
v(t)
vg(t)
C
i(t)
v(t)
Fig. 7.32. Buck-boost converter circuit: (a) during subinterval 1, (b) during subinterval 2.
x(t) =
i(t)
v(t)
(7-108)
The input voltage vg(t) is an independent source which should be placed in the input vector
u(t). In addition, we have chosen to model the diode forward voltage drop with an
independent voltage source of value VD. So this voltage source should also be included in
the input vector u(t). Therefore, let us define the input vector as
u(t) =
vg(t)
VD
(7-109)
To model the converter input port, we need to find the converter input current ig(t). To
calculate this dependent current, it should be included in the output vector y(t). Therefore,
let us choose to define y(t) as
y(t) = i g(t)
(7-110)
Note that it isnt necessary to include the output voltage v(t) in the output vector y(t), since
v(t) is already included in the state vector x(t).
Next, let us write the state equations for each subinterval. When the switch is in
position 1, the converter circuit of Fig. 7.32(a) is obtained. The inductor voltage, capacitor
current, and converter input current are
di(t)
= vg(t) i(t) Ron
dt
dv(t)
v(t)
C
=
R
dt
i g(t) = i(t)
L
(7-111)
These equations can be written in the following state-space form:
35
L 0
0C
Ron 0
d i(t) =
dt v(t)
0 1
R
dx(t)
dt
i g(t)
y(t)
i(t)
+ 10
v(t)
00
vg(t)
VD
A1
x(t)
B1
u(t)
10
i(t)
+ 00
v(t)
vg(t)
VD
C1
x(t)
u(t)
E1
(7-112)
So we have identified the state equation matrices A1, B1, C1, and E1.
With the switch in position 2, the converter circuit of Fig. 7.32(b) is obtained. For
this subinterval, the inductor voltage, capacitor current, and converter input current are
given by
di(t)
= v(t) VD
dt
dv(t)
v(t)
C
=
i(t)
R
dt
i g(t) = 0
L
(7-113)
When written in state-space form, these equations become
L 0
0C
d i(t) =
dt v(t)
dx(t)
dt
i g(t)
y(t)
1
1 1
R
i(t)
+
v(t)
0 1
0 0
vg(t)
VD
A2
x(t)
B2
u(t)
00
i(t)
+
v(t)
00
vg(t)
VD
C2
x(t)
E2
u(t)
(7-114)
So we have also identified the subinterval 2 matrices A2, B2, C2, and E2.
The next step is to evaluate the state-space averaged equilibrium equations (7-84) (7-86). The averaged matrix A is
A = DA 1 + D'A 2 = D
Ron
0
0
DRon D'
0 1
+ D'
=
1
D' 1
1 1
R
R
R
36
(7-115)
In a similar manner, the averaged matrices B, C, and E are evaluated, with the following
results:
B = DB 1 + D'B 2 =
D D'
0 0
C = DC 1 + D'C 2 = D 0
E = DE 1 + D'E 2 = 0 0
(7-116)
The dc state equations (7-84) therefore become
0 =
0
Ig =
DRon
D'
D'
1
R
D0
I +
V
D D'
0 0
Vg
VD
I +
V
00
Vg
VD
(7-117)
Evaluation of Eq. (7-87) leads to the following solution for the equilibrium state and output
vectors:
I =
V
1
R
D
1 + 2 on
D' R
Ig =
1
R
1 + D2 on
D' R
D
1
D' 2R D' R
D 1
D'
Vg
VD
D2 D
D' 2R D'R
Vg
VD
(7-118)
Alternatively, the steady-state equivalent circuit of Fig. 7.33 can be constructed as usual
from Eq. (7-117). The top row of Eq. (7-117) could have been obtained by application of
the principle of inductor volt-second
D'VD
DRon
1:D
D' : 1
balance to the inductor voltage
+
Ig
I
waveform. The second row of Eq. (7V
Vg +
R
117) could have been obtained by
application of the principle of capacitor

charge balance to the capacitor current Fig. 7.33. Dc circuit model for the buck-boost
converter example, equivalent to Eq. (7-117).
waveform. The ig(t) equation expresses
the dc component of the converter input current. By reconstructing circuits that are
equivalent to these three equations, the dc model of Fig. 7.33 is obtained.
The small-signal model is found by evaluation of Eq. (7-87). The vector
coefficients of d(t) in Eq. (7-87) are
37
Vg IRon + VD
Vg V IRon + VD
A 1 A 2 X + B1 B2 U = V +
=
I
I
0
C1 C2 X + E1 E2 U = I
(7-119)
The small-signal ac state equations (7-87) therefore become
DRon D'
L 0 d i(t) =
0 C dt v(t)
D' 1
R
i(t)
+ 00
v(t)
i g(t) = D 0
i(t)
+
v(t)
D D'
0 0
vg(t)
Vg V IRon + VD
+
d(t)
I
0
vg(t)
+ I d(t)
0
(7-120)
Note that, since the diode forward voltage drop is modeled as the constant value V D , there
are no ac variations in this source, and vD(t) equals zero. Again, a circuit model equivalent
to Eq. (7-120) can be constructed, in the usual manner. When written in scalar form, Eq.
(7-120) becomes
d i(t)
= D' v(t) DRon i(t) + D vg(t) + Vg V IRon + VD d(t)
dt
dv(t)
v(t)
C
= D' i(t)
+ I d(t)
R
dt
L
i g(t) = D i(t) + I d(t)
(7-121)
Circuits corresponding to these equations are listed in Fig. 7.34. These circuits can be
combined into the complete small-signal ac equivalent circuit model of Fig. 7.35.
a)
b)
L
DRon
+ d i(t)
L
dt
+
D vg(t)
d(t) Vg V + V D IRon
C
D' v(t)
D' i(t)
I d(t)
dv(t)
dt
v(t)
R
v(t)
i(t)
c)
i g(t)
vg(t)
I d(t)
D i(t)
Fig. 7.34. Circuits equivalent to the small-signal

converter equations: (a) inductor loop, (b) capacitor
node, (c) input port.
38
i g(t)
d(t) Vg V + V D IRon
1:D
i(t)
vg(t)
D' : 1
+
DRon
I d(t)
I d(t) C
v(t)
Fig. 7.35. Complete small-signal ac equivalent circuit model, nonideal buck-boost converter example.
7 . 5 . Circuit averaging and averaged switch modeling

Circuit averaging is another well-known technique for derivation of converter
equivalent circuits. Rather than averaging the converter state equations, with the circuit
averaging technique we average the converter waveforms directly. All manipulations are
performed on the circuit diagram, instead of on its equations, and hence the circuit
averaging technique gives a more physical interpretation to the model. Since circuit
averaging involves averaging and small-signal linearization, it is equivalent to state-space
averaging. However, in many cases circuit averaging is easier to apply, and allows the
small-signal ac model to be written almost by inspection. The circuit averaging technique
can also be applied directly to a number of different types of converters and switch
elements, including phase-controlled rectifiers, PWM converters operated in discontinuous
conduction mode or with current programming, and quasi-resonant converters these are
described in later chapters. However, in other cases it may lead to involuted models that are
less easy to analyze and understand. To overcome this problem, the circuit averaging and
state-space averaging approaches can be combined. Circuit averaging was developed before
state-space averaging, and is described in [4]. Because of its generality, there has been a
recent resurgence of interest in circuit averaging of switch networks [11-15].
The key step in circuit averaging is to replace the converter switches with voltage
and current sources, to obtain a time-invariant circuit topology. The waveforms of the
voltage and current generators are defined to be identical to the switch waveforms of the
original converter. Once a time-invariant circuit network is obtained, then the converter
waveforms can be averaged over one switching period to remove the switching harmonics.
Any nonlinear elements in the averaged circuit model can then be perturbed and linearized,
leading to the small-signal ac model.
In Fig. 7.36, the switching elements are separated from the remainder of the
converter. The converter therefore consists of a switch network containing the converter
switching elements, and a time-invariant network, containing the reactive and other
remaining elements. Figure 7.36 illustrates the simple case in which there are two SPST
switches; the switches can then be represented using a two-port network. In more
39
power input
load
Time-invariant network
containing converter reactive elements
vg(t)
C
vC(t)
i1(t)
v(t)
i2(t)
Switch network
+
port 2
v1(t)
port 1
iL(t)
v2(t)
control
input
d(t)
Fig. 7.36. A switching converter can be viewed as a switch network connected to a

time-invariant network.
complicated systems containing multiple transistors or diodes, such as in polyphase

converters, the switch network may contain more than two ports.
The switch terminal waveforms are v1(t), i1(t), v2(t), and i2(t). As with any two-port
network, two of these terminal quantities can be treated as independent inputs to the switch
network. The remaining two terminal quantities are then viewed as dependent signals,
which can be expressed as functions of the independent terminal inputs and the control
input.
L
For example, consider the boost
+
i(t)
converter of Fig. 7.37. The switch
network contains the transistor and vg(t) +
C
R
v(t)
diode. There are several valid ways to
define the two ports of the switch

Fig. 7.37. Ideal boost converter example.
network; two approaches are illustrated
in Fig. 7.38. These two definitions lead to equivalent, but not identical, averaged models.
Let us, somewhat arbitrarily, define the switch network as in Fig. 7.38(a). Since i1(t) and
v2(t) coincide with the converter
a)
b)
inductor current and capacitor
i1(t)
i2(t)
i1(t)
i2(t)
+
+
+
+
voltage, it is convenient to
choose these waveforms as the
v1(t)
v2(t)
v1(t)
v2(t)
independent inputs to the switch
network.
Fig. 7.38. Two ways to define the switch network terminal
quantities.
40
7.5.1.
Obtaining a time-invariant circuit

The key step in the circuit averaging technique is to replace the switch network with
voltage and current sources, such that the circuit connections do not vary with time.
Specifically, the switch network dependent terminal waveforms are modeled by dependent
sources. The waveforms of these sources are defined to be identical to the corresponding
dependent waveforms of the original circuit.
L
In the boost converter
+
+
i1(t)
i(t)
example, we can replace the switch
i2(t) v2(t) C
v (t) +
v1(t) +
R
v(t)
network with dependent voltage and g
current generators as illustrated in

Switch network
Fig. 7.39. The voltage generator v1(t)
Fig. 7.39. The switch network is replaced by dependent
models the voltage at the input port of
voltage and current sources, yielding a timethe switch network, or the transistor
invariant circuit topology.
voltage. As illustrated in Fig. 7.40,
v1(t) is zero when the transistor conducts, and is equal to v2(t) when the diode conducts:
v1(t) =
v1(t)
0, 0 < t < dTs

v2(t), dTs < t < Ts
(7-122)
When v1(t) is defined in this manner, the inductor
voltage waveform is unchanged. Likewise, i2(t) models
the current at port 2 of the network, or the diode current.
As illustrated in Fig. 7.40, i2(t) is equal to zero when the
transistor conducts , and is equal to i1(t) when diode
conducts:
0, 0 < t < dTs

i 2(t) =
i 1(t), dTs < t < Ts
v2(t)
< v1(t) >T
0
0
dTs
i2(t)
Ts
Ts
i1(t)
< i2(t) >Ts
0
0
dTs
Fig. 7.40. The waveforms of the

dependent voltage and current
sources are defined to be identical to
the corresponding waveforms of the
original circuit.
(7-123)
With i2(t) defined in this manner, the capacitor current
waveforms in the circuits of Figs. 7.37 and 7.39 are
identical.
Hence the circuit of Fig. 7.39 is electrically identical to the circuit of Fig. 7.37. So
far, no approximations have been made.
41
power input
load
Averaged time-invariant network

containing converter reactive elements
<vg(t)>T
<vC(t)>Ts
<i1(t)>T
<iL(t)>T
<v(t)>Ts
<i2(t)>T
Averaged
switch network
port 2
<v1(t)>Ts
port 1
<v2(t)>Ts
control
input
d(t)
Fig. 7.41. The switching harmonics are removed by averaging all waveforms over
one switching period.
7.5.2.
Circuit averaging
Next, we remove the switching harmonics by averaging all signals in Figs. 7.36
and 7.39 over one switching period, as in Eq. (7-3).
The basic assumption is made that the natural time constants of the converter
network are much longer that the switching period T s. This assumption coincides with the
requirement for small switching ripple. One may average the waveforms over a time
interval which is short compared to the system natural time constants, without significantly
altering the system response. Hence, when the basic assumption is satisfied, it is a good
approximation to average the converter waveforms over the switching period T s. The
resulting averaged model predicts the low-frequency behavior of the system, while
neglecting the high-frequency switching harmonics.
When all waveforms of Fig. 7.36 are averaged over one switching period, the
system of Fig. 7.41 is obtained. For the boost converter example of Fig. 7.39, averaging
the waveforms leads to Fig. 7.42. These models predict the dc and low-frequency ac
components of the converter waveforms, and neglect the high-frequency switching ripple.
L
<i(t)>Ts
<vg(t)>Ts
<i1(t)>Ts
<v1(t)>Ts
<i2(t)>Ts
<v2(t)>Ts
Averaged switch network
Fig. 7.42.
Averaging all waveforms, boost converter example.
42
+
C
<v(t)>Ts
<vg(t)>T
<i1(t)>Ts
<i(t)>Ts
+
d'(t) <v2(t)>T
d'(t) <i1(t)>T
<v2(t)>T
<v(t)>Ts
Averaged switch model
Fig. 7.43. Basic nonlinear circuit-averaged model, boost converter example.
What remains is to evaluate the averaged switch waveforms. In particular, the

averaged dependent generators < v1(t) >Ts and < i2(t) >Ts must be expressed as functions of
the averaged independent switch waveforms < i1(t) >Ts and < v2(t) >Ts, and the control input
d(t). This is accomplished by averaging the switch dependent waveforms in Fig. 7.40 over
one switching period. The results are
v1(t)
i 2(t)
Ts
Ts
= d'(t) v2(t)
= d'(t) i 1(t)
Ts
Ts
(7-124)
Here we have assumed that the switching ripples of the inductor current and capacitor
voltage are small, or at least linear functions of time. The circuit of Fig. 7.43 is now
obtained. This is the basic nonlinear circuit-averaged model for the boost converter. The
switching harmonics have been removed from all converter waveforms, leaving only the dc
and low-frequency ac components. This model is now valid for frequencies sufficiently
less than the switching frequency. The switch network has been replaced by time-invariant
generators.
7.5.3.
Perturbation and linearization

The model of Fig. 7.43 is nonlinear, because the dependent generators involve the
multiplication of the control input d(t) with the current i1(t) and voltage v2(t). The network
can be linearized by perturbing and linearizing the converter waveforms about a quiescent
operating point, in the usual manner. Let
vg(t)
Ts
= Vg + vg(t)
d(t) = D + d(t) d'(t) = D' d(t)

i(t) T = i 1(t) T = I + i(t)
s
v(t)
v1(t)
i 2(t)
Ts
Ts
Ts
= v2(t)
Ts
= V + v(t)
= V1 + v1(t)
= I 2 + i 2(t)
(7-125)
43
L
+
I + i(t)
Vg + vg(t)
D' d(t) V + v(t)
D' d(t) I + i(t)
V + v(t)
Fig. 7.44. Perturbation of the nonlinear circuit-averaged model about a quiescent

operating point.
The model of Fig. 7.44 is then obtained. Note that the model contains both dc and smallsignal ac terms: one equivalent circuit is used for both the dc and the small-signal ac
models.
The voltage generator has value
D' d(t) V + v(t) = D' V + v(t) V d(t) v(t) d(t)
(7-126)
The term v(t) d(t) is nonlinear, and is small in magnitude provided that the ac variations are
much small than the quiescent values (as in Eq. (7-31)). When
the small-signal assumption is satisfied, this term can be
neglected. The term V d(t) is driven by the control input, and
V d(t)
+
hence can be represented by an independent voltage source as

in Fig. 7.45. The term D' V + v(t) is equal to the constant
value D multiplied by the output voltage V + v(t) . This term
D' V + v(t)
Fig. 7.45. Linearization of the

dependent voltage source.
is dependent on the output capacitor voltage; it is represented by

a dependent voltage source in Fig. 7.45. This dependent source
will eventually become the primary winding of a transformer.
The nonlinear current generator of Fig 7.44 is treated in a similar manner. Its
current is
D' d(t) I + i(t) = D' I + i(t) I d(t) i(t) d(t)
(7-127)
The term i(t) d(t) is nonlinear, and can be neglected

provided that the small-signal assumption is satisfied. The
term I d(t) is driven by the control input d(t) , and is
represented by an independent current source in Fig. 7.46.
The term D' I + i(t) is dependent on the inductor current
I + i(t) . This term is modeled by a dependent current
D' I + i(t)
I d(t)
Fig. 7.46. Linearization of the

dependent current source.
source in Fig. 7.46; this source will become the secondary winding of an ideal transformer.
The converter equivalent circuit of Fig. 7.47 is therefore obtained. The final step is
to replace the dependent generators with an ideal D : 1 transformer, as in Fig. 7.48. The
44
transformer symbol contains both a solid line (indicating that it is an ideal transformer
capable of passing dc voltages and currents) and a sinusoidal line (which indicates that
small-signal ac variations are modeled). Figure 7.48 is the complete circuit-averaged
model. This model functions simultaneously as the dc and the small-signal ac equivalent
circuit for the boost converter.
V d(t)
I + i(t)
Vg + vg(t)
D' V + v(t)
D' I + i(t)
I d(t) C
V + v(t)
Fig. 7.47. Linearized circuit-averaged converter model.

L
V d(t)
D' : 1
I + i(t)
Vg + vg(t)
I d(t)
V + v(t)
Fig. 7.48. The dependent linear sources are replaced by an equivalent ideal transformer, yielding the
final dc and small-signal ac circuit-averaged model.
In summary, the circuit averaging method involves modeling the switch network
with equivalent voltage and current sources, such that a time-invariant network is obtained.
The converter waveforms are then averaged over one switching period to remove the
switching harmonics. The resulting low-frequency model is then perturbed and linearized
about a quiescent operating point, to obtain a small-signal model.
7.5.4.
Averaged switch modeling

Comparison of Figs. 7.48 and 7.37 reveals that, for the boost converter example,
the equivalent circuit model is very similar to the original converter circuit. The major
difference is that the switch network has been replaced by an effective ideal transformer and
independent generators, as illustrated in Fig. 7.49. It can be concluded that, for this boost
converter example, the
2
D' : 1
switch
network
+
+
i(t)
I + i(t)
V
d(t)
1
performs two basic
I d(t)
V + v(t)
v(t)
functions:
(i)
transformation of dc
switch
and small-signal ac
network
voltage and current Fig. 7.49. Boost converter ideal switch network, and its equivalent CCM dc
and small-signal ac model as predicted by the circuit averaging
levels according to the
technique.
45
a)
b)
i1(t)
i1(t)
+ vCE
iC
i2(t)
+
v1(t)
v2(t)
i 1(t)
T2
+
C
v(t)
v2(t)
T2
i2
i(t)
0
vg(t) +
i2
dTs
Ts
Ts
v1
v2(t)
Switch network
0
T2
dTs
Fig. 7.50. Buck converter example: (a) converter circuit, (b) switch waveforms.
D : 1 conversion ratio, and (ii) introduction of ac voltage and current variations into the
converter circuit, driven by the control input d(t) .
Indeed, averaging the waveforms of Fig. 7.36 modifies only the switch network;
the remainder of the converter circuit is unchanged. This suggests that, to obtain a smallsignal ac converter model, we need only to replace the switch network with its averaged
model. This procedure is called averaged switch modeling. The circuit averaging procedure
is followed, to average the switch terminal waveforms. The resulting dc and small-signal ac
averaged switch model is then inserted into the converter circuit.
For example, the CCM buck converter of Fig. 7.50(a) exhibits the terminal switch
waveforms of Fig. 7.50(b). Representation of the buck converter in the form of Fig. 7.36
leads to the network illustrated in Fig. 7.51. Let us select v1(t) and i2(t) as the independent
terminal variables of the two-port switch network, since these quantities coincide with the
applied converter input voltage vg(t) and the inductor current i(t), respectively. We then
need to express the averaged dependent terminal waveforms < i1(t) >Ts and < v2(t) >Ts as
functions of the control input d(t) and
of < v1(t) >Ts and < i2(t) >Ts. Upon
averaging the waveforms of Fig.
7.50(b), one obtains
i 1(t)
v2(t)
Ts
Ts
= d(t) i 2(t)
= d(t) v1(t)
L
+
vg(t) +
v1(t)
i(t)
+
i2(t)
i1(t)
+ v (t)
2
Ts
Ts
(7-128)
Perturbation and linearization of Eq.
(7-128) then leads to
Switch network
Fig. 7.51. Replacement of switch network by dependent

sources, buck converter example.
I 1 + i 1(t) = D I 2 + i 2(t) + I 2 d(t)

V2 + v2(t) = D V1 + v1(t) + V1 d(t)
46
v(t)
(7-129)
I1 + i1
V1 d
V1 + v1
V2 + v2
I2 d
b)
I2 + i2
1:D
I1 + i1
+
Vg + vg +
V1 + v1
I2 d
I+i
V2 + v2
V1 d
+
R
Fig. 7.52. Averaged switch modeling, buck converter example: (a) dc

and small-signal ac equivalent circuit model for the switch network;
(b) replacement of switch network by equivalent circuit (a).
I2 + i2
1:D
+
I1 + i1
V1 d
v1(t)
v2(t)
V1 + v1
i2(t)
+
V2 + v2
I2 d
I1 + i1
I2 + i2
D' : 1
+
i1(t)
V2 d
v1(t)
v2(t)
V1 + v1
i2(t)
+
I1 d
V2 + v2
I1 + i1
I2 + i2
D' : D
+
i1(t)
V +v
Switch network
i2(t)
+
I2 + i2
1:D
+
i1(t)
a)
An equivalent circuit
corresponding to Eq. (7129) is illustrated in Fig.
7.52(a). Replacement of
the switch network in
Fig. 7.50(a) with the
averaged switch model of
Fig. 7.52(a) leads to the
converter model of Fig.
7.52(b).
Three
switch
networks, and their CCM
averaged models, are
illustrated in Fig. 7.53.
v1(t)
v2(t)
V1 + v1
V1
d
DD'
+
I2
d
DD'
V2 + v2
Fig. 7.53. Three basic switch networks, and their CCM dc and small-signal ac
averaged switch models.
47
v (t)
i (t)
CE
C
Switching losses can also be
v1
modeled via averaged switch modeling.
i2
Consider again the CCM buck converter
of Fig. 7.50(a). Let us suppose that the 0
0
t
diode is ideal, and that the transistor
t1
t2
tir tvf
tvr tif
exhibits the idealized clamped inductive
Ts
load waveforms illustrated in Fig. 7.54. Fig. 7.54. Switch waveforms, buck converter
The transistor turn-on process occurs
switching loss example.
during intervals of length tir and tvf. The transistor conducts for time t1, with negligible
forward voltage drop. The transistor turn-off process occurs during intervals of length tvr
and tif. For this simple example, it is assumed that the transistor voltage and current
waveforms change linearly during the various switching intervals, as illustrated in Fig.
7.54.
Let us again select v1(t) and i2(t) as the independent terminal valiables of the twoport switch network, and derive expressions for the averaged dependent terminal
waveforms < i1(t) >Ts and < v2(t) >Ts. It can be seen from Fig. 7.50(a) that i1(t) and v2(t) are
given by
i 1(t) = i C(t)
v2(t) = v1(t) vCE(t)
(7-130)
The average value of i1(t) depends on the area enclosed by the iC(t) waveform, and can be
expressed as follows:
i 1(t)
= 1
Ts
Ts
Ts
i 1(t) dt
= i 2(t)
Ts
t 1 + t vf + t vr + 12 t ir + 12 t if
Ts
(7-131)
Let us make the definitions
d=
t 1 + 12 t vf + 12 t vr + 12 t ir + 12 t if
Ts
t vf + t vr
Ts
t ir + t if
di =
Ts
dv =
(7-132)
The quantity d(t) is the effective transistor duty cycle, equal to the transistor on-time plus
one-half of the switching times, divided by the switching period. The quantity dv is the
48
fraction of the switching period contained in the voltage rise and fall times. Similarly, di is
the fraction of the switching period consumed by the current rise and fall times. With these
definitions, Eq. (7-131) can be written
i 1(t)
Ts
= i 2(t)
Ts
d + 12 d v
(7-133)
The average value of v2(t) is
v2(t)
Ts
= v1(t) vCE(t)
= 1
Ts
Ts
Ts
vCE(t) dt + v1(t)
Ts
(7-134)
Evaluation of the integral, for the vCE(t) waveform of Fig. 7.54, yields
v2(t)
Ts
= v1(t)
Ts
t 1 + 12 t vf + 12 t vr
Ts
(7-135)
Finally, substitution of the definitions (7-132) yields the result
v2(t)
Ts
= v1(t)
Ts
d 12 d i
(7-136)
Equations (7-133) and (7-136) constitute the averaged terminal relations of the switch
network. An equivalent circuit corresponding to these relations is constructed in Fig.
7.55(a). The generators which depend on the effective transistor duty cycle d(t) can be
combined into an ideal transformer, as illustrated in Fig. 7.55(b). This is the large-signal
switch model, which accounts for the switching loss of the idealized waveforms of Fig.
7.54. If desired, this model can be perturbed and linearized in the usual manner, to obtain a
small-signal ac switch model.
a)
1
2
s
+
<v1(t)>T
1
2
dv(t) <i2(t)>T
di(t) <v1(t)>T
<i1(t)>T
d(t) <i2(t)>T
+ d(t) <v (t)>

1
Ts
<v2(t)>T
b)
1 : d(t)
+
1
2
1
2
di(t) <v1(t)>T
+
<i1(t)>T
<v1(t)>T
<i2(t)>T
<i2(t)>T
+
<v2(t)>T
dv(t) <i2(t)>T
Fig. 7.55. Switch network large-signal averaged model, buck converter switching
loss example: (a) network corresponding to Eqs. (7-133) and (7-135);
(b) averaged model incorporating ideal transformer.
49
The model of Fig. 7.55(b) has the following physical interpretation. The transistor
operates with the effective duty cycle d(t) given by Eq. (7-132). This is the turns ratio of
the ideal dc transformer, which models the first-order switch property of lossless transfer
of power from the switch input to the switch output port. The additional dv current
generator and di voltage generator model switching loss. Note that both generators
consume power. The total switching loss is
Psw = 12 d v + d i i 2(t)
v1(t)
(7-137)
These generators also correctly model how the switching loss increases the average switch
input current, and decreases the average switch output voltage. This model can be
combined with the models for other converter losses, to predict the voltages, currents, and
efficiencies of nonideal converters having significant switching loss. Other sources of
switching loss, discussed in chapter 4, can be modeled in a similar manner doing so is
left as a homework problem.
I1
Ts
Ts
+
1
2
Vg
I2
1:D
V1
1
2
I
+
Di V1
V2
Dv I2
Averaged switch network model
Fig. 7.56. Dc equivalent circuit model, buck converter switching loss

example.
By inserting the switch model of Fig. 7.55(b) into the original converter circuit of
Fig. 7.50(a), and by letting all waveforms be equal to their quiescent values, we obtain the
steady-state model of Fig. 7.56. This model predicts that the steady-state output voltage is
V = D 12 Di Vg = DVg 1
Di
2D
(7-138)
Note that DVg is the ideal (lossless) output voltage, while the term (1 Di / 2D) is the
reduction in the output voltage due to switching loss.
To find the efficiency, we must compute the average input and output powers. The
converter input power is
Pin = Vg I 1 = V1 I 2 D + 12 Dv
(7-139)
The average output power is
Pout = V I 2 = V1 I 2 D 12 Di
(7-140)
50
Hence the converter efficiency is
Di
2D
P
= out =
=
Pin
D
D + 12 Dv
1+ v
2D
D 12 Di
(7-141)
Beware, the efficiency is not simply equal to V / DVg.

Averaged switch modeling is generalized in chapter 10, where it is used to model
the dynamics of converters operating in the discontinuous conduction mode. It is also used
in later chapters to model resonant switch converters.
7 . 6 . The canonical circuit model
Having discussed several methods for deriving the ac equivalent circuit models of
switching converters, let us now pause to interpret the results. All PWM CCM dc-dc
converters perform similar basic functions. First, they transform the voltage and current
levels, ideally with 100% efficiency. Second, they contain low-pass filtering of the
waveforms. While necessary to remove the high-frequency switching ripple, this filtering
also influences low-frequency voltage and current variations. Third, the converter
waveforms can be controlled by variation of the duty cycle.
We expect that converters having similar physical properties should have
qualitatively similar equivalent circuit models. Hence, we can define a canonical circuit
model which correctly accounts for all of these basic properties [1-3]. The ac equivalent
circuit of any CCM PWM dc-dc converter can be manipulated into this canonical form.
This allows us to extract physical insight, and to compare the ac properties of converters.
The canonical model is used in several later chapters, where it is desired to analyze
converter phenomena in a general manner, without reference to a specific converter. So the
canonical model allows us to define and discuss the physical ac properties of converters.
In this section, the canonical circuit model is developed, based on physical
arguments. An example is given which illustrates how to manipulate a converter equivalent
circuit into canonical form. Finally, the parameters of the canonical model are tabulated for
several basic ideal converters.
51
7.6.1.
Development of the canonical circuit model

Converter model
The
physical
a)
1 : M(D)
elements of the canonical
+
circuit model are collected,
Vg +
V
R
one at a time, in Fig. 7.57.
The converter contains a

power input port vg(t) and a
control input port d(t), as
D
Power
Control
Load
well as a power output port
input
input
b)
and load having voltage v(t).
As discussed in chapter 3,
1 : M(D)
+
the basic function of any
V + v(s)
R
V + v (s) +
CCM PWM dc-dc converter
is the conversion of dc
voltage and current levels,

ideally
with
100%
D
Power
Control
Load
efficiency. As illustrated in
input
input
c)
Fig. 7.57(a), we have
He(s)
modeled this property with
1 : M(D)
+
an ideal dc transformer,
Effective
Zei(s)
Zeo(s)
having effective turns ratio
V + v(s)
R
V + v (s) +
low-pass
filter
1:M(D) where M is the
conversion
ratio.
This
conversion ratio is a function
D
of the quiescent duty cycle
Power
Control
Load
input
input
D. As discussed in chapter
d)
3, this model can be refined,
He(s)
e(s) d(s)
1 : M(D)
if desired, by addition of
+
Effective
resistors and other elements
Zei(s)
Zeo(s)
j(s) d(s)
V + v(s)
V + v (s) +
low-pass
that model the converter

filter
losses.
Slow
variations
vg(t) in the power input
D + d(s)
g
induce ac variations v(t) in

the converter output voltage.
As illustrated in Fig.
Power
input
Control
input
Load
Fig. 7.57. Development of canonical circuit model, based on physical arguments: (a) dc transformer model, (b) inclusion of ac variations, (c) reactive
elements introduce effective low-pass filter, (d) inclusion of ac duty cycle
variations.
52
7.57(b), we expect these variations also to be transformed by the conversion ratio M(D).
The converter must also contain reactive elements which filter the switching
harmonics and transfer energy between the power input and power output ports. Since it is
desired that the output switching ripple be small, the reactive elements should comprise a
low-pass filter having a cutoff frequency well below the switching frequency. This lowpass characteristic also affects how ac line voltage variations influence the output voltage.
So the model should contain an effective low-pass filter as illustrated in Fig. 7.57(c). This
figure predicts that the line-to-output transfer function is
Gvg(s) =
v(s)
= M(D) H e(s)
vg(s)
(7-142)
where He(s) is the transfer function of the effective low-pass filter. The effective filter also
influences other properties of the converter, such as the small-signal input and output
impedances. It should be noted that the elemental values in the effective low-pass filter do
not necessarily coincide with the physical element values in the converter. In general, the
element values, transfer function, and terminal impedances of the effective low-pass filter
can vary with quiescent operating point. Examples are given in the following subsections.
Control input variations, i.e. duty cycle variations d(t) , also induce ac variations in
the converter voltages and currents. Hence, the model should contain voltage and current
sources driven by d(t) . In the examples of the previous section, we have seen that both
voltage sources and current sources appear, which are distributed around the circuit model.
It is possible to manipulate the model such that all of the d(t) sources are pushed to the
input side of the equivalent circuit. In the process, the sources may become frequencydependent; an example is given in the next subsection. In general, the sources can be
combined into a single voltage source e(s) d(s) and a single current source j(s) d(s) as
shown in Fig. 7.57(d). This model predicts that the small-signal control-to-output transfer
function is
Gvd (s) =
v(s)
= e(s) M(D) H e(s)
d(s)
(7-143)
This transfer function is found by setting the vg(t) variations to zero, and solving for the
dependence of v(t) on d(t) . Figure 7.57(d) is the complete canonical circuit, which can
model any PWM CCM dc-dc converter.
53
7.6.2.
Example: manipulation of the buck-boost converter model into

canonical form
To illustrate the steps in the derivation of the canonical circuit model, let us
manipulate the equivalent circuit of the buck-boost converter into canonical form. A smallsignal ac equivalent circuit for the buck-boost converter is derived in section 7.2. The
result, Fig. 7.16(b), is reproduced in Fig. 7.58. To manipulate this network into canonical
form, it is necessary to push all of the independent d(t) generators to the left, while
pushing the inductor to the right and combining the transformers.
Vg + vg(s)
Vg V d
1:D
Id
D' : 1
+
Id
V + v(s)
Fig. 7.58. Small-signal ac model of the buck-boost converter, before manipulation into canonical form.
The Vg V d(t) voltage source is in series with the inductor, and hence the
positions of these two elements can be interchanged. In Fig. 7.59(a), the voltage source is
placed on the primary side of the 1 : D ideal transformer; this requires dividing by the
effective turns ratio D. The output-side I d(t) current source has also been moved to the
primary side of the D : 1 transformer. This requires multiplying by the turns ratio 1/D.
The polarity is also reversed, in accordance with the polarities of the D : 1 transformer
windings.
Next, we need to move the I d(t) D' current source to the left of the inductor. This
can be done using the artifice illustrated in Fig. 7.59(b). The ground connection of the
current source is broken, and the source is connected to node A instead. A second,
identical, current source is connected from node A to ground. The second source causes the
current flowing into node A to be unchanged, such that the node equations of Figs. 7.59(a)
and 7.59(b) are identical.
In Fig. 7.59(c), the parallel combination of the inductor and current source is
converted into Thevenin equivalent form. The series combination of an inductor and
voltage source are obtained.
54
a)
Vg V
d
D
Vg + vg(s)
1:D
D' : 1
+
I d
D'
Id
V + v(s)
b)
Vg V
d
D
Vg + vg(s)
node
A
1:D
D' : 1
+
I d
D'
Id
I d
D'
V + v(s)
Vg V
d
D
Vg + vg(s)
sLI d
D'
1:D
L
D' : 1
c)
I d
D'
Id
V + v(s)
node
B
Vg V
d
D
Vg + vg(s)
Id
DI d
D'
sLI d
D'
1:D
DI d
D'
d)
L
D' : 1
+
C
V + v(s)
Fig. 7.59. Steps in the manipulation of the buck-boost ac model into canonical form.
In Fig. 7.59(d), the I d(t) D' current source is pushed to the primary side of the
1 : D transformer. The magnitude of the current source is multiplied by the turns ratio D.
In addition, the current source is pushed through the (Vg V) d(t) / D voltage source,
using the previously described artifice. The ground connection of the source is moved to
node B, and an identical source is connected from node B to ground such that the circuit
node equations are unchanged.
Figure 7.60 is the final form of the model. The inductor is moved to the secondary
side of the D : 1 transformer, by multiplying by the square of the turns ratio as shown.
The sLI d / D' voltage source is moved to the primary side of the 1 : D transformer, by
dividing by the turns ratio D. The voltage and current sources are combined as shown, and
the two transformers are combined into a single D : D transformer. The circuit is now in
canonical form.
55
Vg V
s LI d(s)
D
DD'
D' : D
L
D' 2
Vg + vg(s) +
I d(s)
D'
V + v(s)
Effective
low-pass
filter
Fig. 7.60. The buck-boost converter model, in canonical form.
It can be seen that the inductance of the effective low-pass filter is not simply equal
to the physical inductor value L, but rather is equal to L / D2. At different quiescent
operating points, with different values of D, the value of the effective inductance will
change. In consequence, the transfer function, input impedance, and output impedance of
the effective low-pass filter will also vary with quiescent operating point. The reason for
this variation is the transformation of the inductance value by the effective D : 1
transformer.
It can also be seen from Fig. 7.60 that the coefficient of the d voltage generator is
e(s) =
Vg + V s LI
D
D D'
(7-144)
This expression can be simplified by substitution of the dc relations (7-28). The result is
e(s) = V2 1 s DL
D
D' 2 R
(7-145)
When we pushed the output-side I d(t) current source through the inductor, we obtained a
voltage source having a frequency dependence. In consequence, the e(s) d voltage
generator is frequency-dependent.
7.6.3.
Canonical circuit parameter values for some common converters

For ideal CCM PWM dc-dc converters containing a single inductor and capacitor,
the effective low-pass filter of the canonical model should contain a single inductor and a
single capacitor. The canonical model then reduces to the circuit of Fig. 7.61. It is assumed
that the capacitor is connected directly across the load. The parameter values for the basic
buck, boost, and buck-boost converters are collected in Table 7.1. Again, it should be
pointed out that the effective inductance Le depends not only on the physical inductor value
L, but also on the quiescent duty cycle D. Furthermore, the current flowing in the effective
inductance Le does not in general coincide with the physical inductor current I + (t).
56
e(s) d(s)
1 : M(D)
Le
Vg + vg(s) +
j(s) d(s)
V + v(s)
Fig. 7.61. The canonical model, for ideal CCM converters containing a single inductor and capacitor.
Table 7.1. Canonical model parameters for the ideal buck, boost, and buck-boost converters
Converter
Buck
Boost
Buck-boost
M(D)
Le
D
1
D'
L
L
D' 2
L
D' 2
D
D'
e(s)
V
D2
V 1 s 2L
D' R
V2 1 s DL
D
D' 2 R
j(s)
V
R
V
D' 2 R
V2
D' R
The model of Fig. 7.61 can be solved using conventional linear circuit analysis, to
find quantities of interest such as the converter transfer functions, input impedance, and
output impedance.
7 . 7 . Modeling the pulse-width modulator
We have now achieved the goal, stated at the beginning of this chapter, of deriving
a useful equivalent circuit model for the switching converter in Fig. 7.1. One detail
remains: modeling the pulse-width modulator. The pulse-width modulator block shown in
Fig. 7.1 produces a logic signal (t) which commands the converter power transistor to
switch on and off. The logic signal (t) is periodic, with frequency fs and duty cycle d(t).
The input to the pulse-width modulator is an analog control signal vc(t). The function of the
pulse-width modulator is to produce a duty cycle d(t) that is proportional to the analog
control voltage vc(t).
A schematic diagram of a simple pulse-width modulator circuit is given in Fig.
7.62. A sawtooth wave generator produces the voltage waveform vsaw(t) illustrated in Fig.
7.63. The peak-to-peak amplitude of this waveform is V M . The converter switching
frequency fs is determined by and equal to the frequency of vsaw(t). An analog comparator
compares the analog control voltage vc(t) to vsaw(t). This comparator produces a logic-level
output which is high whenever vc(t) is greater than vsaw(t), and is otherwise low. Typical
waveforms are illustrated in Fig. 7.63.
57
If the sawtooth waveform vsaw(t) has

minimum value zero, then the duty cycle
will be zero whenever vc(t) is less than or
equal to zero. The duty cycle will be D = 1
whenever vc(t) is greater than or equal to
V M . If, over a given switching period,
vsaw(t) varies linearly with t, then for 0
vc(t) V M the duty cycle d will be a linear
function of vc. Hence, we can write
d(t) =
vc(t)
VM
Sawtooth
vsaw(t)
wave
generator
comparator
analog
input
(t)
PWM
waveform
vc(t)
Fig. 7.62. A simple pulse-width modulator circuit.

vsaw(t)
VM
vc(t)
for 0 vc(t) VM
(7-146)
This equation is the input-output characteristic of the
pulse-width modulator [2,9].
To be consistent with the perturbed-andlinearized converter models of the previous sections, we
can perturb Eq. (7-146). Let
(t)
dTs
Ts
2Ts
Fig. 7.63. Waveforms of the

circuit of Fig. 7.62.
vc(t) = Vc + vc(t)
d(t) = D + d(t)
(7-147)
Insertion of Eq. (7-147) into Eq. (7-146) leads to
D + d(t) =
Vc + vc(t)
VM
(7-148)
A block diagram representing Eq. (7-148) is
illustrated in Fig. 7.64. The pulse-width modulator
has linear gain 1 / VM. By equating like terms on both
sides of Eq. (7-148), one obtains
Vc + vc(s)
1
VM
D + d(s)
pulse-width
modulator
Vc
VM
v (t)
d(t) = c
VM
D=
Fig. 7.64. Pulse-width modulator

block diagram.
(7-149)
So the quiescent value of the duty cycle is determined in practice by Vc.
The pulse-width modulator model of Fig. 7.64 is sufficiently accurate for nearly all
applications. However, it should be pointed out that pulse-width modulators also introduce
sampling of the waveform. Although the analog input signal vc(t) is a continuous function
of time, there can be only one discrete value of the duty cycle during every switching
period. Therefore, the pulse-width modulator samples the waveform, with sampling rate
58
equal to the switching frequency f s . Hence,

sampler
a more accurate modulator block diagram is v
d
c
1
VM
as in Fig. 7.65 [8]. In practice, this
fs
sampling restricts the useful frequencies of
the ac variations to values much less than
pulse-width modulator
the switching frequency. The designer must
ensure that the bandwidth of the control Fig. 7.65. A more accurate pulse-width
modulator model, including sampling.
system be sufficiently less than the Nyquist
rate fs / 2.
Significant high-frequency variations in the control signal vc(t) can also alter the
behavior of the pulse-width modulator. A common example is when vc(t) contains
switching ripple, introduced by the feedback loop. This phenomenon has been analyzed by
several authors [8,16], and a simple investigation of the influence of switching ripple on
the modulator gain is left as a homework problem. But it is generally best to avoid the case
where vc(t) contains significant components at the switching frequency or higher, since the
pulse-width modulators of such systems exhibit poor noise immunity.
7 . 8 . Summary of key points
1. The CCM converter analytical techniques of chapters 2 and 3 can be extended to predict
converter ac behavior. The key step is to average the converter waveforms over one
switching period. This removes the switching harmonics, thereby exposing directly
the desired dc and low-frequency ac components of the waveforms. In particular,
expressions for the averaged inductor voltages, capacitor currents, and converter
input current are usually found.
2. Since switching converters are nonlinear systems, it is desirable to construct smallsignal linearized models. This is accomplished by perturbing and linearizing the
averaged model about a quiescent operating point.
3. Ac equivalent circuits can be constructed, in the same manner used in chapter 3 to
construct dc equivalent circuits. If desired, the ac equivalent circuits may be refined
to account for the effects of converter losses and other nonidealities.
4. The state-space averaging method of section 7.4 is essentially the same as the basic
approach of section 7.2, except that the formality of the state-space network
description is used. The general results are listed in section 7.4.2.
5. The circuit averaging technique also yields equivalent results, but the derivation
involves manipulation of circuits rather than equations. Switching elements are
replaced by dependent voltage and current sources, whose waveforms are defined
59
to be identical to the switch waveforms of the actual circuit. This leads to a circuit
having a time-invariant topology. The waveforms are then averaged to remove the
switching ripple, and perturbed and linearized about a quiescent operating point to
obtain a small-signal model.
6. When the switches are the only time-varying elements in the converter, then circuit
averaging affects only the switch network. The converter model can then be derived
by simply replacing the switch network with its averaged model. Dc and smallsignal ac models of several common CCM switch networks are listed in section
7.5.4. Switching losses can also be modeled using this approach.
7. The canonical circuit describes the basic properties shared by all dc-dc PWM converters
operating in the continuous conduction mode. At the heart of the model is the ideal
1:M(D) transformer, introduced in chapter 3 to represent the basic dc-dc conversion
function, and generalized here to include ac variations. The converter reactive
elements introduce an effective low-pass filter into the network. The model also
includes independent sources which represent the effect of duty cycle variations.
The parameter values in the canonical models of several basic converters are
tabulated for easy reference.
8. The conventional pulse-width modulator circuit has linear gain, dependent on the slope
of the sawtooth waveform, or equivalently on its peak-to-peak magnitude.
REFERENCES
[1]
R. D. Middlebrook and Slobodan Cuk, A General Unified Approach to Modeling SwitchingConverter Power Stages, International Journal of Electronics, vol. 42, no. 6, pp. 521-550, June
1977.
[2]
Slobodan Cuk, Modeling, Analysis, and Design of Switching Converters, Ph.D. thesis,
California Institute of Technology, November 1976.
[3]
R. D. Middlebrook and Slobodan Cuk, Modeling and Analysis Methods for Dc-to-dc Switching
Converters, Proceedings of the IEEE International Semiconductor Power Converter Conference,
1977 Record, pp. 90-111, March 1977. Reprinted in Advances in Switched-Mode Power
Conversion, vol. 1, Irvine: Teslaco, 1983.
[4]
G. W. Wester and R. D. Middlebrook, Low-Frequency Characterization of Switched Dc-dc

Converters, IEEE Transactions an Aerospace and Electronic Systems, vol. AES-9, pp. 376-385,
May 1973.
[5]
Daniel M. Mitchell, Dc-dc Switching Regulator Analysis, New York: McGraw-Hill Inc., 1988.
[6]
Seth R. Sanders and George C. Vergese, Synthesis of Averaged Circuit Models for Switched
Power Converters, IEEE Transactions on Circuits and Systems, vol. 38, no. 8, pp. 905-915,
August 1991.
60
[7]
P. T. Krein, J. Bentsman, R. M. Bass, and B. C. Lesieutre, On the Use of Averaging for the
Analysis of Power Electronic Systems, IEEE Transactions on Power Electronics, vol. 5, no. 2,
pp. 182-190, April 1990.
[8]
Arthur R. Brown and R. D. Middlebrook, Sampled-Data Modeling of Switching Regulators,

IEEE Power Electronics Specialists Conference, 1981 Record, pp. 349-369, June 1981.
[9]
R. D. Middlebrook, Predicting Modulator Phase Lag in PWM Converter Feedback Loops,

Proceedings of the Eighth National Solid-State Power Conversion Conference (Powercon 8), April
1981.
[10]
A. Kislovski, R. Redl, and N. Sokal, Dynamic Analysis of Switching-Mode DC/DC Converters,

New York: Van Nostrand Reinhold, 1994.
[11]
V. Vorperian, R. Tymerski, and F. C. Lee, Equivalent Circuit Models for Resonant and PWM
Switches, IEEE Transactions on Power Electronics, vol. 4, no. 2, pp. 205-214, April 1989.
[12]
V. Vorperian, Simplified Analysis of PWM Converters Using the Model of the PWM Switch:
Parts I and II, IEEE Transactions on Aerospace and Electronic Systems, vol. AES-26, pp. 490505, May 1990.
[13]
S. Freeland and R. D. Middlebrook, A Unified Analysis of Converters with Resonant Switches,

IEEE Power Electronics Specialists Conference, 1987 Record, pp. 20-30.
[14]
Arthur Witulski and Robert Erickson, "Extension of State-Space Averaging to Resonant Switches
and Beyond," IEEE Transactions on Power Electronics, vol. 5, no. 1, pp. 98-109, January 1990.
[15]
D. Maksimovic and S. Cuk, A Unified Analysis of PWM Converters in Discontinuous Modes,

IEEE Transactions on Power Electronics, vol. 6, no. 3, pp. 476-490, July 1991.
[16]
D. J. Shortt and F. C. Lee, Extensions of the Discrete-Average Models for Converter Power
Stages, IEEE Power Electronics Specialists Conference, 1983 Record, pp. 23-37, June 1983.
P ROBLEMS
7.1.
An ideal boost converter operates in the continuous conduction mode.

(a)
Determine the nonlinear averaged equations of this converter.
(b)
Now construct a small-signal ac model. Let

vg(t)
Ts
= Vg + vg(t)
d(t) = D + d(t)
i(t) T = I + i(t)
s
v(t)
Ts
= V + v(t)
where Vg, D, I, and V are steady-state dc values, vg(t) and d(t) are small ac variations in
the power and control inputs, and i(t) and v(t) are the resulting small ac variations in
the inductor current and output voltage, respectively. Show that the following model
results:
61
Large-signal dc components
0 = D'V + Vg
0 = D'I V / R
Small-signal ac components
d i(t)
= D' v(t) + V d(t) + vg(t)
dt
dv(t)
v(t)
C
= D' i(t) I d(t)
R
dt
L
7.2.
Construct an equivalent circuit that corresponds to the boost converter small-signal ac equations
derived in problem 7.1(b).
7.3.
Manipulate your boost converter equivalent circuit of problem 7.2 into canonical form. Explain
each step in your derivation. Verify that the elements in your canonical model agree with Table
7.1.
7.4.
The ideal current-fed bridge converter of Fig. 2.31 operates in the continuous conduction mode.
(a)
Determine the nonlinear averaged equations of this converter.
(b)
Perturb and linearize these equations, to determine the small-signal ac equations of the
converter.
(c)
Construct a small-signal ac equivalent circuit model for this converter.
7.5.
Construct a complete small-signal ac equivalent circuit model for the flyback converter shown in
Fig. 7.18, operating in continuous conduction mode. The transformer contains magnetizing
inductance L, referred to the primary. In addition, the transformer exhibits significant core loss,
which can be modeled by a resistor RC in parallel with the primary winding. All other elements
are ideal. You may use any valid method to solve this problem. Your model should correctly
predict variations in ig(t).
7.6.
Modeling the Cuk converter. You may use any valid method to solve this problem.
7.7.
(a)
Derive the small-signal dynamic equations which model the ideal Cuk converter.
(b)
Construct a complete small-signal equivalent circuit model for the Cuk converter.
Modeling the inverse-SEPIC. You may use any valid method to solve this problem.
(a)
(b)
Derive the small-signal

dynamic equations which
model
the
converter
shown in Fig. 7.66.
C1
L2
+
ig(t)
vg(t) +
Construct a complete
small-signal
equivalent
circuit model for the
inverse-SEPIC.
L1
C2
v(t)
Fig. 7.66.
62
7.8.
Consider the nonideal buck converter of Fig. 7.67. The input voltage source vg(t) has internal
resistance Rg. Other component nonidealities may be neglected.
(a)
Rg
Using the state-space
averaging
method,
ig(t)
determine the smallsignal
ac
equations vg(t) +
which describe variations
in i, v, and i g , which
occur
owing
to
variations
in
the
transistor duty cycle d and input voltage vg.
L
+
C
v(t)
Fig. 7.67.
(b)
Construct an ac equivalent circuit model corresponding to your equations of part (a).
(c)
Solve your model to determine an expression for the small-signal control-to-output

transfer function.
7.9.
Use the circuit-averaging technique to derive the dc and small-signal ac equivalent circuit of the
buck converter with input filter, illustrated in Fig. 2.32. All elements are ideal.
7.10.
A flyback converter operates in the continuous conduction mode. The MOSFET switch has onresistance Ron, and the secondary-side diode has a constant forward voltage drop V D . The flyback
transformer has primary winding resistance Rp and secondary winding resistance Rs.
7.11.
(a)
Derive the small-signal ac equations for this converter.
(b)
Derive a complete small-signal ac equivalent circuit model, which is valid in the

continuous conduction mode and which correctly model the above losses, as well as the
converter input and output ports.
Circuit averaging of the bridge inverter circuit of Fig. 7.68(a).

(a)
(b)
(c)
7.12.
Show that the converter of

Fig. 7.68(a) can be written
in the electrically-identical
form shown in Fig.
7.68(b).
Sketch
the
waveforms i1(t) and v1(t).
Use the circuit-averaging
method to derive a largesignal averaged model for
this converter.
a)
C
1
vg(t)
L
+ v(t)
i(t)
b)
L
+
vg(t) +
i1(t)
Perturb and linearize your

circuit model of part (b), to
obtain a single equivalent
circuit that models dc and
small-signal ac signals in the bridge inverter.
v1(t)
v(t)
Fig. 7.68.
Use the circuit averaging method to derive an equivalent circuit that models dc and small-signal ac
signals in the buck-boost converter. You may assume that the converter operates in the continuous
conduction mode, and that all elements are ideal.
(a)
Give a time-invariant electrically-identical circuit, in which the switching elements are

replaced by equivalent voltage and current sources. Define the waveforms of the sources.
63
7.13.
(b)
Derive a large-signal averaged model for this converter.
(c)
Perturb and linearize your circuit model of part (b), to obtain a single equivalent circuit
that models dc and small-signal ac signals in the buck-boost converter.
The two-output flyback converter of Fig. 7.69(a) operates in the continuous conduction mode. It
may be assumed that the converter is lossless.
(a)
Derive a small-signal ac equivalent circuit for this converter.
(b)
Show that the small-signal ac equivalent circuit for this two-output converter can be
written in the generalized canonical form of Fig. 7.69(b). Give analytical expressions for
the generators e(s) and j(s).
b)
a)
ip
C1
L
vg(t)
e(s) d(s)
i g(s)
np : n1
R1
v1(t)
D'np : n1
1:D
i(s)
vg(s)
L
R1
C1
j(s) d(s)
+
C2
R2
C2
v2(t)
7.14.
A pulse-width modulator circuit is

constructed in which the sawtooth-wave
generator is replaced by a triangle-wave
generator, as illustrated in Fig. 7.70(a).
The triangle waveform is illustrated in
Fig. 7.70(b).
(a)
(b)
(c)
: n2
Fig. 7.69.
a)
Triangle
wave
vtri(t)
generator
comparator
Determine
the
converter
switching frequency, in Hz.
Determine the gain d(t) / vc(t) for
this circuit.
R2
analog
input
(t)
PWM
waveform
vc(t)
b)
vtri(t)
Over what range of v c is your

answer to (b) valid?
2V
50s
100s
t
-2V
Fig. 7.70.
7.15.
In the buck converter circuit of Fig. 7.50(a), the transistor voltage vCE(t) and current iC(t)
waveforms of Fig. 7.54 are observed. The diode exhibits forward voltage drop V D , and the inductor
winding has resistance RL.
(a)
v2(s)
: n2
v1(s)
Derive dc and small-signal ac equivalent circuit models for the converter, which include
the effects of switching loss, diode forward voltage drop, and inductor winding resistance.
64
(b)
Solve your dc equivalent circuit, to find analytical expressions for the converter efficiency
and steady-state output voltage.
7.16.
Modify the CCM dc and small-signal ac averaged switch models of Fig. 7.53, to account for
MOSFET on-resistance Ron and diode forward voltage drop VD.
7.17.
Use the averaged switch modeling technique to derive an ac equivalent circuit model for the buckboost converter of Fig. 7.31:
(a)
Replace the switches in Fig. 7.31 with a basic switch network model given in Fig. 7.53.
(b)
Compare your result with the model given in Fig. 7.16(b). Show that the two models
predict the same small-signal line-to-output transfer function Gvg(s).
7.18
Use the averaged switch modeling technique to derive a dc and small-signal ac equivalent circuit
model for the SEPIC.
7.19.
In the boost converter of Fig. 7.71(a), the v1(t) and i2(t) waveforms of Fig. 7.71(b) are observed.
During the transistor turn-on transition, a reverse current flows through the diode which removes
the diode stored charge. As illustrated in Fig. 7.72(b), the reverse current spike has area Qr and
duration tr. The inductor winding
a)
has resistance R L . You may
L
i1(t)
iL(t)
i2(t)
neglect all losses other than the
+
+
+
switching loss due to diode
stored charge and the conduction
v1(t)
v(t)
vg(t) +
R
v2(t) C
loss due to the inductor winding

resistance.
(a)
(b)
(c)
Derive an averaged
switch network model.
b)
v1(t)
v2
Use your result of part

(a) to derive a dc
equivalent circuit model
for the boost converter.
Derive
analytical
expressions for the
converter steady-state
output voltage and
efficiency.
v2
dTs
tr
Ts
i2(t)
i1
i1
t
area Qr
Fig. 7.71
65

AC Equivalent Circuit Modeling

Uploaded by

Document Informationclick to expand document information

Copyright:

Available Formats

AC Equivalent Circuit Modeling

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

AC Equivalent Circuit Modeling

Uploaded by

Copyright:

Available Formats

Chapter 7

AC Equivalent Circuit Modeling

and feedback loop block

version 1/14/98 11:00 AM

Chapter 7. AC Equivalent Circuit Modeling

with ripple neglected

where D and Dm are constants, | Dm | << D, and

Fig. 7.2. Ac variation of the converter signals:

Chapter 7. AC Equivalent Circuit Modeling

frequency and its

components at the switching

Chapter 7. AC Equivalent Circuit Modeling

Chapter 7. AC Equivalent Circuit Modeling

Fig. 7.6. Small-signal ac equivalent circuit model of the buck-boost converter.

Chapter 7. AC Equivalent Circuit Modeling

Chapter 7. AC Equivalent Circuit Modeling

Averaging the inductor waveforms

vL() d d(t) vg(t)

Chapter 7. AC Equivalent Circuit Modeling

Discussion of the averaging approximation

Divide by L, and integrate both sides from t to t + Ts:

i(t + Ts) i(t) = 1 Ts vL(t)

i(t + Ts) i(t)

Let us now define the derivative of < i(t) >Ts as

i(t + Ts) i(t)

Chapter 7. AC Equivalent Circuit Modeling

which coincides with Eq. (7-2).

( final value) = (initial value) + (length of interval) (average slope)

( final value) = (initial value) + (length of interval) (average slope)

Equations (7-20) - (7-22) are illustrated in Fig.

waveform changes over one switching

coincide with the switching period, such that the

Chapter 7. AC Equivalent Circuit Modeling

Averaging the capacitor waveforms

The average input current

Chapter 7. AC Equivalent Circuit Modeling

The input current waveform is illustrated in Fig.

Perturbation and linearization

Chapter 7. AC Equivalent Circuit Modeling

It should be noted that the complement of the duty cycle is given by

d'(t) = 1 d(t) = 1 D + d(t) = D' d(t)

increase will cause d(t) to decrease.

Chapter 7. AC Equivalent Circuit Modeling

By multiplying out Eq. (7-32) and collecting terms, one obtains

Chapter 7. AC Equivalent Circuit Modeling

I g + i g(t) = D + d(t) I + i(t)

By collecting terms, we obtain

i g(t) = Di(t) + Id(t)

Chapter 7. AC Equivalent Circuit Modeling

Construction of the small-signal equivalent circuit model

i g(t) = Di(t) + Id(t)

current (t). As illustrated in Fig. 7.13, the term L

D'v(t) represent dependent sources as shown.

According to the equation, this current is

Chapter 7. AC Equivalent Circuit Modeling

that g(t) is equal to the currents in two branches, as

Chapter 7. AC Equivalent Circuit Modeling

Results for several basic converters