CHAPTER

NVERTER

'

•
- ..
•
.. . .
•
•
•

.. ...•..., .
•
.•
• •
• • •

'

••
•
• •
•

FIGURE 3.1 Power supplies today are based on de-de converters.

94
 WHY NOT VOLTAGE DIVIDERS? 95

3.1 THE IMPORTANCE OF DC-DC CONVERSION

Ac electricity is the form of choice for the modern power grid because of the magnetic trans-
former. Convenient energy exchange among voltage levels overcame the major drawbacks
of Edison's early de system. The term transformer normally implies a magnetic device
restricted to ac signals but more generally describes a lossless two-port network with a ratio,
a, between input and output voltages. The de transformer is a device that, like its ac counter-
part, provides lossless energy transfer between circuits at different voltage or current levels.
It is less familiar, since magnetic technology used in practical ac transformers cannot handle
direct current. In power electronics, the de transformer is realized as a de-de converter.
Flexible de-de conversion is vital in electronic circuits and a host of other applications. Many
circuits use multiple voltage levels a typical mobile phone might use four or five different
voltages. It is often more convenient to convert energy from a single source rather than to
try to distribute many different supplies. This is especially true in battery-operated devices.
At one time, de power supplies took energy from an ac line source, changed its level
with a line-frequency transformer, and rectified the result. Today, most supplies, like those in
Figure 3.1, are built with de-de converters. The incoming ac is rectified with a diode bridge,
and a de-de converter processes it directly to provide the desired levels. Modern power
supplies of this type range from less than 1 V outputs for computation cores up to 700 V or
more for vehicle chargers and industrial applications. Many products are designed around
170 V inputs (the peak value of rectified 120 V ac) or 400 V inputs (above the peak values of
230 V ac, 240 V ac, and many three-phase rectified sources). Others are intended to handle
48 V as used in telephone networks or 28 V as used in military and cable communication
applications. Converters for 48 V to 2.7 V, for instance, support logic circuits in land-line
telephones. Converters for 400 V to 12 V support analog power supplies, electric automobile
equipment, and data servers.
As we explore de-de converters, we need to consider alternatives and then examine
de-de converters based on switch matrices. The basic converters lead to versions that pro-
vide input-output isolation, extreme conversion ratios, or other objectives. Certain types of
nonswitching circuits are common for de regulation, and some of these will be considered.

3.2 WHY NOT VOLTAGE DIVIDERS?

Why not just use a voltage divider for de-de conversion? After all, voltage dividers are con-
venient for creating low de voltages. However, the problems with applying dividers for power
conversion are severe; a voltage divider is inherently inefficient (at best, the efficiency is the
ratio Vou/Vi,J and does not provide control. A divider cannot provide Vout > Vin or change
polarity.
The divider circuit in Figure 3.2 can be analyzed to find its efficiency and power
conversion properties. It supplies an output load drawing current lout at voltage Vout· Circuit
analysis shows that

if/out =0, V out= Rout v.m

Rin +Rout (3.1)

if I out :;t 0,
96 DC-DC CONVERTERS

FIGURE 3.2 Resistive voltage divider as a de-de converter.

+

Rout V
out

The output depends on both Vin and lout and cannot easily be adjusted, since the resistor val-
ues must be altered to change it. Even with no load there is substantial loss. The efficiency
for lout> 0 is

(3.2)

The value can be very low, especially if a large division ratio is attempted. It will never be
higher than Vou/Vin since the input current is always at least as much as lout· The logical con-
clusion is that voltage dividers are important for sensing and measurement, but not suitable
for power. Consider the following example.

Example 3.2.1 A voltage divider is to be built to give a 5 V output from a 12 V input. The
nominal load is 5 W. Design a divider to provide 5 V ± 5% for loads ranging from 3 W to 5 W,
assuming that the input is exactly 12 V. What is the no-load output? What is the efficiency at
nominal load?
In a divider, the load interacts with the resistor set to create the output. The heavier
the load, the lower the output voltage. For a design, let us start with an arrangement that
provides the minimum output vout = 4.75 v for the rated 5 w load and the maximum allowed
output of 5.25 V for a 3 W load. Power of 5 W at 4.75 V corresponds to a load resistor, Rzoad'
of 4.51 n. Power of 3 W at 5.25 V corresponds to 9.19 n. The divider design requires

Vout _ Rout II Rzoad

- (3.3)
"in Rin +Rout II Rzoad

For the two loads, two equations can be developed in the unknowns Rin and Rout' and then
solved simultaneously. The result for a 4.75 V to 5.25 V output range is

Rin = 2.13 Q, Rout= 2.03 Q (3.4)

For the 5 W load, this provides 4.75 V and draws 3.40 A and 40.8 W from the 12 V source.
The efficiency is 12.3%. For the 3 W load, the divider draws 3.16 A and 37.9 Wand has effi-
ciency of only 7.9%. Although there are other (smaller) values of Rin and Rout that will meet the
requirements, they will give even lower efficiencies, so these values are a "best case" choice.
The no-load output will be 5.84 V, which is higher than allowed.
 LINEAR REGULATORS 97

The output varies only ±5% over this limited load range, but this small change has come
at the expense of low efficiency. The divider has been constructed to draw a much bigger cur-
rent than is needed for the load, so that any load variation is swamped out by the large current
draw of the divider itself. This divider dissipates more than 34 W even with no load at all!

Dividers are suitable for sensing applications and for creating multiple reference voltage
levels in analog-to-digital converters. They can be applied when efficiency is not an issue.

3.3 LINEAR REGULATORS

3.3.1 Regulator Circuits

Regulators are power circuits that deliver a tightly controlled output even when subject
to variations of the source and load. When switching converters are used for regulation,
feedback controllers must adjust duty ratios or phase timing to keep the output in a narrow
range. Sometimes power electronic circuits used this way are called switching regulators. In
contrast, linear regulators are nonswitching circuits that include their own internal control
and produce a tightly regulated output in spite of input and load changes. The two major
types of linear regulators are shown in Figure 3.3. The basic series pass regulator [1], [2] (or
just series regulator) in Figure 3.3 is a transistor connected as an emitter follower (or source
follower for a field-effect transistor [FET]). The transistor is designed to operate in its active
region rather than as a switch. If the base voltage is held fixed, the emitter voltage will be
about one diode drop below that set point. The emitter voltage Vout becomes a function of
Vcontrol rather than the input voltage or the load current. This represents a "pure regulation"
function: the output is independent of various disturbances. Since Vout = Vcontroz- Vbe' the
output is a linear function of a low-power control voltage, and the term linear regulator is
appropriate.
The shunt regulator [3], [4] in Figure 3.3 resembles a common-emitter amplifier
circuit. The transistor is used in its active region. The collector current will be {3/b, rather than
a function of the load current. The small base current can be adjusted as necessary to produce
the required output voltage. The output is a linear function of the control. A simpler shunt
regulator is represented by the Zener diode in Figure 3.4. As long as reverse current flows in
the device, it will maintain a nearly fixed voltage, Vz· To use a shunt regulator, it is essential
that the load never diverts too much current such that the shunt device no longer functions.

FIGURE 3.3 Basic circuits for linear (a) (b)

regulation. (a) Series regulator; (b) shunt
regulator.
vcontrol ---l

Itoad Itoad

Vcontrol
.---------<0

Series pass regulator Shunt regulator

98 DC-DC CONVERTERS

FIGURE 3.4 Zener diode in a shunt regulator function. The series

resistance Rs must be low enough to draw current for both the load
and the diode.

Itoad

Linear regulators sacrifice efficiency to obtain regulation. They are limited to

Vout Vin in all situations. For a series regulator implemented with a high-gain device such
that the base current can be ignored, the input and output current are the same, and

Pzoss = (Vfn - Vout )Izoad' (3.5)

so that high efficiency requires a step-down ratio as close as possible to 1. For the shunt
regulator, the loss and efficiency are

(3.6)

In the shunt circuit, there is a significant loss even with no load because /c -:t- 0. In the best
case, where lzoad >> lc, the efficiency becomes Vou/Vin- Shunt regulators are useful for
creating reference voltage signals at low power, and they are rare in other modern power
applications.
Linear regulators are not power electronic circuits, since the active device must have
a power rating not much different from the desired output power. As converters, they have
limited efficiency. However, since they bring the possibility of immunity to change, linear
regulators are frequently used as elements of larger conversion systems. They provide a
filtering function, delivering a precise output from an uncertain input. Conventional series
regulators require an input voltage at least 2 V higher than the output to provide adequate
bias for the transistor. Many engineers have designed low dropout (LDO) series regulators
[5], [6] that can operate with a little as 50 mV across the active device.

Example 3.3.1 Figure 3.5 shows an LDO regulator, set for 1.5 V output. This circuit will
produce almost exactly 1.5 V at its output for any input above 1.55 V. With 1.55 V input, it
is nearly 97% efficient, and the loss is traded off for tight output control. It is to be used in a
filtering application.
In this application, a switching converter produces a voltage slightly above the desired Vout
but with significant ripple. The switching converter, as in Figure 3.5, exhibits a triangle-wave
 LINEAR REGULATORS 99
---------------- 1
+

! _______________ _

Regulator

FIGURE 3.5 An LDO series regulator in a 1.5 V output application.

output between 1.55 V and 1.75 V. Since the voltage is always above the LDO requirement of
1.50 V +50 mV = 1.55 V, the output is held fixed at 1.50 V. In this regulator, lin provided the
controls require much less current than the load. The efficiency is the ratio of average powers,

(3.7)

With lout lin' and essentially constant because of the fixed output voltage, the currents can be
treated as coefficients in front of the averaging integrals and canceled to give

(3.8)

This 9.1% loss is the price paid for a fixed output.

3.3.2. Regulation Measures

Regulation is a general challenge in converters. Users usually seek tight regulation and out-
puts that do not change. Typically, regulation measures and numbers are based on the actual
measured change as a fraction of the intended output. For example, any de power supply has
a certain line regulation value, representing the variation above and below the average output
value. The standard definition [7] is given in percentage by

-;o L"zne reg=_

m +max(Vout)- min(Vout) X
100m
-;o (3.9)
max(Vout) + min(Vout) azzowed znput
. range

Here, the "allowed input range" refers to the rated range of input voltages, and the nominal
output is the intended voltage value. A power supply rated to deliver 12 V given input ac volt-
ages from 100 V to 265 V, and specified to have better than ±0.1% line regulation, should
show change of no more than ±12 mV (one part in a thousand) away from the average value
when actually tested across the 100 V to 265 V input range.
Another important specification for de power supplies is the load regulation, given in
percentage terms [7] as

m L d + max(Vout)- min(Vout)
-;o oa reg = _
X 100mo
7< (3.10)
max(Vout) + min(Vout) allowed toad range
100 DC-DC CONVERTERS

In this case, the measurement is to be made over the specified values of load. A 12 V con-
verter rated at 10 W to 100 W, for example, would be tested over this full range to determine
how much the output changes and whether this amount of change is acceptable. For both line
and load regulation, an ideal source gives values of 0%.
In many power supplies, temperature regulation is important, usually expressed as a
partial derivative

Temp reg=
avout (3.11)
aT
A supply might be rated, for instance, to have temperature regulation better than 0.1 mV/K.
It is also conventional to express temperature regulation as a percentage change per kelvin
(or °C) [7]. Other regulation measures, such as time variation, can also be defined in terms
of partial derivatives. Multiple-output power supplies must consider cross regulation, the
change in one output voltage linked to a change in load at other output ports. This is defined
just as in equation (3.10), with the measurement taken over allowed load ranges on the other
ports.

3.4 DIRECT DC-DC CONVERTERS AND FILTERS

3.4.1 The Buck Converter

Linear regulators and dividers are not really energy converters, although they have many
applications in sensing, control, and filtering. One of the most basic de-de converters is
the direct switching converter for de voltage to de current shown in Figure 3.6. Kirchhoff's
Voltage Law (KVL) and Kirchhoff's Current Law (KCL) constraints require that one, and
only one, switch can be on in any column at a given time. This can be written q 1, 1 + q 2 ,1 = 1
and q 1,2 + q2 ,2 = 1. Only four switch combinations avoid violations of KVL or KCL and are
given in Table 3.1. There are three possible instantaneous output voltage values (+Vin' -Vin'
and 0). Switch action selects among these three.
The bridge converter of Figure 3.6 can be simplified by allowing the input and output
sources to share a common ground reference. This can be accomplished by turning switch
2,2 on at all times and leaving switch 1,2 off. In addition, the output current source typically
consists of a large inductance in series with a load. Figure 3.7 shows the arrangement. The
switches have been relabeled simply as #1 (the previous 1,1) and #2 (the previous 2,1) for
simplicity. Switch #1 must be able to either carry or prevent the flow of lout' so a transistor

TABLE 3.1 Allowed Switch Configurations in the General De Voltage to

De Current Converter
Switch Combination Result, Voltage Result, Current

Close 1,1 and 2,2 vout = vin iin = lout

Close 2,1 and 2,2 Vout = 0 i.m = 0
Close 1,1 and 1,2 Vout = 0 i.m = 0
Close 1,2 and 2,1 Vout = - Vin iin = -lout
 DIRECT DC-DC CONVERTERS AND FILTERS 101

FIGURE 3.6 The general de voltage to de current

1,1 1,2
direct converter. / /
•
lin
-.. / /

2,1 2,2
/ /
/ /

Vout -
+
-

is appropriate. Switch #2 must be able to conduct I out and prevent reverse flow, a diode
application. In practice, the term buck converter normally refers to this common-ground
arrangement instead of the general case in Figure 3.6.
The circuit of Figure 3.7 requires q1 + q2 = 1, since a current path must be provided for
the inductor whenever the transistor is off. The output is Vin when #1 is on and 0 when #2 is
on. We can write vout(t) = q 1Vin- Thus v out looks like the switching function q 1, except that its
amplitude is Vin instead of 1. The average value of the output is

(3.12)

Since (vL ) = 0 , the resistor voltage average value must match (vout>· The input average cur-
rent will be (iin) = D 1/ out. The inductor and load act as a low-pass filter to ensure that de
is passed to the output while the unwanted ac components are attenuated. A typical output
waveform appears in Figure 3.8.
The common-ground buck converter is used in many switching de power supplies,
in high-performance de motor controllers, on computer motherboards, to drive solid-state
lamps, and in just about any appliance that includes electronic controls. This converter, along
with some closely related circuits, is sometimes called a "buck regulator" or a "step-down"
converter. Notice the regulation properties of this circuit. Since the output voltage computa-
tion does not involve the load current or resistance, the relationship ( vout) = D 1Vin holds as
long as the constraint q 1 + q2 = 1 is valid. Output is independent of load, so load regulation
is perfect in principle. On the other hand, the output is proportional to Vin' so any change

L lout
•

#1 + + +
+-...._
lout Vout V·m
#2

FIGURE 3.7 Common-ground de-de voltage-current converter, or buck converter.

102 DC-DC CONVERTERS

Vout(t)

- - - - - - - - - - -

-
< Vout>

0 time
0 T 2T

FIGURE 3.8 Typical output waveform for converter of Figure 3.7.

in input is reflected proportionally at the output. These regulation properties are exactly the
same as those of an ideal transformer: Load regulation is limited only by internal imped-
ances, and line effects are completely unregulated. Control will be needed to provide line
regulation.
Some of the important relationships in the buck converter are:

iin = qliout

( iin ) = Dl I out

(3.13)

1 fT 2 2 /n
Vout(RMS) = T Jo ql (t)V in dt =

Ideally, no power is lost in either switch, so the input and output power must be equal at all
times. The switches each must carry the inductor current when on and must hold off the input
voltage when off:

Carrying current: i1 = I L, i 2 = I L
(3.14)
Blocking voltage: v1 = v2 =

Let us consider an application example to illustrate some of these relationships. This example
uses the previous concepts of source interfaces to demonstrate low output ripple.

Example 3.4.1 A buck converter circuit with an R-L load is shown in Figure 3.9. Let
Vin = 15 V. The switching frequency is 50 kHz. If the output is to be 5 V nominally, what duty
ratio is required? What is the actual output Vzoad(t), including ripple?
Since the circuit is a buck converter, we expect = D1 The duty ratio of switch #1
should be (5 V)/(15 V) = 1/3. The frequency is 50 kHz and the period is 20 JlS, so switch #1 should
be on for 6.67 JlS, then off for 13.33 JlS, on so on. If the output is exactly 5 V, the inductor voltage
will be Vin- Vzoad = 10 V when switch #1 is on and -Vzoad = -5 V when #1 is off. Is the output
current constant? Since Vzoad = 5 V, the resistor current must average 5 A. The inductor current
changes at a rate dildt = (10 V)/(2 mH) = 5000 A/s when #1 is on and by dildt = -2500 A/s when
#2 is on. During the on-time of switch #1, the circuit configuration is shown in Figure 3.10, and
the current increases (5 x 103 A/s) x (6.67 JlS) = 0.033 A. This is such a small fraction of the 5 A
average value that it is reasonable to treat the load current as effectively constant.
 DIRECT DC-DC CONVERTERS AND FILTERS 103

2 mH lout

+
+
V;n - - - #1 #2 1 n, 5 v nominal

FIGURE 3.9 Buck converter for Example 3.4.1.

2mH

+_ ____J__ _

15 v 1 n,5 v

FIGURE 3.10 Buck converter circuit with switch #1 on.

2mH

15 v

FIGURE 3.11 Buck converter circuit with switch #2 on.

5.1
Vout(t)
Q)

-...0
Q')
1'0
5
>
- '

cQ)
4.9
.....
.....
::J
u

0 20 40 60 80 100 120 140

Time (J!S)

FIGURE 3.12 Buck converter circuit output current and voltage.

When switch #2 is on, the circuit configuration is the R-L form in Figure 3.11. Since the
output should be 5.00 V, the inductor voltage is -5 V whenever switch #2 is on. This assump-
tion will be valid if the time constant L/R is much longer than the on-time of switch #2. In the
circuit here, L/R = 2 ms, which is 150 times the switch #2 on-time. The inductor current drops
(2.5 x 103 A/s) x (13.33 JlS) = 0.033 A. The current fall must balance the rise. Just as with an
energy analysis method, the net energy into the inductor must be zero over a cycle. If it is not,
the output average value would change accordingly. The current change of 0.033 A produces
an output voltage change of 0.033 V for this 1 Q load. Figure 3.12 shows the output waveform.
The output voltage is almost constant, with Vout = 5 ± 0.017 V.
104 DC-DC CONVERTERS

The reader is encouraged to compare the triangular waveform computed in this analysis
to the actual exponential rise and fall of iL. What is the maximum percent difference between
the exponential result and the triangular waveform in Figure 3.12?

The buck converter in Example 3.4.1 showed almost constant output current. Since
the output is so nearly a current source, the average value of Vout was given accurately as
D 1Vin- An interesting side note was that the average power was presented to a resistor. The
resistor showed nearly constant current but also nearly constant voltage. The resistor cannot
distinguish whether the source presented to it is a constant current source or a constant
voltage source.

Example 3.4.2 A buck converter is used with a de motor as its load. The converter duty ratio
provides a basis for motor control. A de motor can be associated with a circuit model, as illus-
trated in Figure 3.13. Sketch the motor current waveform and comment on the current-source
model. What is the effect of duty ratio on the motor speed?
This circuit includes an internal voltage, or back EMF, proportional to shaft speed m with
units of rad/s such that V8 = k8 m. The circuit is inductive because of the windings and magnetic
materials. The voltage applied to the inductor is either VL = vin- laRa- vg or VL =-laRa- Vg,
depending on the circuit configuration. In either case, the time-rate-of-change of inductor cur-
rent is limited, and the inductor current Ia changes little over short periods of time (such as the
time it takes a switch to turn on or turn off). If there is nonzero flow in the inductor, a current
path must be provided, according to KCL. The current waveform with D 1 = 0.5 is shown in
Figure 3.14. At any point in time, the current has a positive and slowly varying value. Even
though the waveform is not pure de, current is always flowing, so a current source has validity
as a model.
If the resistance Ra is small, as is the usual case, the voltage V8 should match the aver-
age value of Vt. Since V8 is proportional to motor speed and ("\'r) = D 1 Yin, the operating speed
is approximately

(3.15)

FIGURE 3.13 Simplified de motor

circuit model.

I FIGURE 3.14 Current from a buck

converter into an inductive load.

0 50 100 150
Time (J.ls)
 DIRECT DC-DC CONVERTERS AND FILTERS 105

The duty ratio serves as a direct speed control, with extensive application possibilities.
Many advanced de drive systems start out with this concept and then add features such as
compensation for the lfia voltage drop or corrections for minor speed variations caused by the
mechanical load.

The buck converter has ripple in its output voltage and its input current. The output
L-R load serves as a filter for the ripple. A conventional buck converter supports only
positive output current and always has an input voltage higher than the output. These
conditions are well adapted to restricted switches. To control the value of ( v out), the duty
cycle of the gate control signal to switch #1 can be altered. If the devices switch quickly
and have low loss, a near-perfect buck converter can be implemented. The circuit then
implements a de transformer with a step-down ratio D 1. One limitation of this transformer
is that it does not exhibit electrical isolation. A configuration called a forward converter
inserts a transformer into the buck circuit for isolation. It will be discussed later in
this chapter.

3.4.2 The Boost Converter

The buck converter provides de transformation with Vout Vin- The second direct de-de
converter type has a current source input and voltage source output. Since it is reversed from
the buck case, it is perhaps not surprising that IVoutl ;;::: Vin for this converter, which is shown
in Figure 3.15.
The relationships for this boost converter are the dual of those for the buck circuit.
The input current and output voltage are fixed source values, and the input voltage and
output current are determined by switch matrix action. For the common-ground version in
Figure 3.15,

ql + q2 = 1
v in ( t) = q 2 vout = (1 - q 1)vout

iout = q2/in = (1 - ql)/in (3.16)

( vin) = D2Vout = (1 - D1)Vout

Let Yin = (vin) and lout = (iout). Then the relationships can be written

and _ 1_ /
(3.17)
1- Dl out

The input and output power must always be equal, as must their averages. There are no losses
in this ideal version of the converter.
Switch #1 carries lin when on. When it is off, switch #2 must be on, and switch #1 must
block Vout· Therefore, switch #1 must be a forward-conducting forward-blocking device. Switch
#2 can be a diode. In the buck converter, we used a series inductor to keep current nearly
constant. A typical boost converter uses an inductor at the input to emulate a current source.
A capacitor is an appropriate interface at the output to provide voltage source characteristics.
Let us consider an example.
106 DC-DC CONVERTERS

(a) 11 /
I 12 /
I (b) •
1
/ / out
+ #2
V;n +
211 / 212 /
- l;n V;n #1 Vout
/ /

•
1out
+ Vout-

2 x 2 matrix version Common-ground version

FIGURE 3.15 Boost de-de converter: 2 x 2 matrix version and common-ground version.

Example 3.4.3 A boost converter intended to convert a 5 V input into a 120 V output is
shown in Figure 3.16. Rated output power is 50 W. The specifications call for an output voltage
of 120 V ± 0.1% and input current variation of no more than ±1%. The switching frequency is
20kHz. Select an inductor and capacitor to meet these specifications. What should the cur-
rent and voltage ratings of the switches be? If the on-time of switch #1, D 1T, has uncertainty of
±50 ns, how much extra uncertainty will be introduced into vou?
The combination of the input voltage source and the inductor effectively becomes a cur-
rent source. The voltage across this source is vin = q2Vout· However, since the inductor cannot
sustain de voltage, the average voltage ( v in ) = D 2 Vout must match Vin- With Vin = 5 V and
Vout = 120 V, the duty ratio D 2 = (5 V)/(120 V) = 0.0417, and switch #2 is on 2.083 JlS out of
each 50 JlS period. Switch #1 is on for 47.92 JlS each cycle. To produce high output voltage, we
need to spend a lot of time building up inductor energy and then spend a very short time dis-
charging the energy rapidly at high potential. The output is 50 W, as is the input (no losses are
modeled here). Thus lin= 10 A. The current is to vary by no more than ±1%, for a total variation
of 0.2 A. During the switch #1 on time, the inductor voltage is Vin = 5 V. The current will rise
linearly as

5V = L di = L
dt
(3.18)
= 0.2A
L

Since the on-time, 47.92 JlS, an inductor L > 1.20 mH ensures that the change in current
stays below 0.2 A.

•
L lout

v
#2 •

+ + 'c +
........ / \ V;n
#1 R=2880 120V
- c

FIGURE 3.16 Boost converter circuit for Example 3.4.3.

 DIRECT DC-DC CONVERTERS AND FILTERS 107

At the output, the capacitor must supply 0.417 A during the long period when switch #2 is
off. The voltage will fall during this interval but is not allowed to change by more than ±0.1%,
for a total change of 0.24 V. The change can be expressed approximately as

i =Cdv
c dt

= 0.24 V (3.19)
c
C F

With = 47.92 JlS and ic = 0.417 A, a capacitor value C 83.3 JlF should meet the require-
ments. The switches must carry the inductor current of 10 A when on and must block the voltage
Vout = 120 V when off. Therefore, devices need to be rated for 10 A and 120 Vat a minimum.
One significant practical problem in a boost converter is that the actual ratio Vou/Vin is rather
limited. Forward drops and resistances of real switches limit the gain. A second problem is the
difficulty to control very high or very low duty ratios precisely. In this circuit, ± 50 ns random
jitter in the switch #1 on time will place the actual on-time between 47.87 JlS and 47.97 JlS. This
seemingly insignificant on-time variation corresponds to output uncertainty over the range
117.2 V to 123.0 V, or about± 2.5%, which is far beyond the intended± 0.1% target.

3.4.3 Power Filter Design

Component selection for the boost converter design is an example of the interface problem.
When designing a power filter, we seek to eliminate unwanted ripple and retain desired de or
ac values, ultimately implementing the source conversion concept. For instance, if de current
source characteristics are needed at an output port, series inductance would be added there.
When de voltage source characteristics are desired, parallel capacitance would be the choice,
based on the capacitor's tendency to maintain fixed voltage. When ac sources are needed,
more complicated arrangements might be used. The underlying objective is to pass through a
wanted frequency component, mwanted (which could be zero for de) while attenuating all other
frequencies, the unwanted frequency components. When the wanted component is de, the
filter should show a low-pass characteristic. When the wanted component is ac, a band-pass
filter is needed. A converter usually has only one wanted component, so resonant filters are
used sometimes for ac waveforms.
Two common methods for analysis and design of power filters are

Equivalent source methods

The "ideal action" approach

Figure 3.17 shows an H-bridge inverter with a resonant L-C-R load and an equivalent
source version of the circuit. In the equivalent source circuit, shown in Figure 3.18, a
square wave (at the desired frequency) is delivered by the bridge to the resonant load. The
choice mwanted = will pass the wanted frequency while attenuating unwanted com-
ponents. The resistor will see a sinusoidal current and voltage, and the voltage will match
the fundamental Fourier component of the square wave. In the equivalent source method, a
combination of devices that will deliver a known waveform is replaced with an ideal source
that delivers that waveform.
108 DC-DC CONVERTERS

•
lout

+
L R
c

FIGURE 3.17 Square-wave inverter with resonant load and the resulting waveforms .

•
1out

'--

FIGURE 3.18 Equivalent source representation of square-wave inverter.

The ideal action approach is more profound and is based on a fundamental principle
of engineering design: the answer is known, and this should be used as much as possible to
identify the method needed to solve the problem. In a power converter, we know what a filter
is seeking to achieve, whether it is a low-ripple de output, an ideal sine wave, or some other
specific outcome. If the filter is properly designed, it will work. If it works, we know its out-
put. The known output can be used as needed to establish the design.
To understand the ideal action approach, consider the buck converter in Figure 3.19,
operating with 50% duty ratio. The input is 10 V, so the average output voltage will be 5 V.
How can the inductor value be selected to achieve a desired ripple? On the one hand, we
could solve the differential equations governing piecewise converter action and find a precise
inductor value to achieve a ripple limit. On the other hand, as in Figure 3.20, an equivalent
source can be used to eliminate switch action and provide a linear circuit to be solved. These
methods are valid but lack the insight provided by the known answer a proper choice of
inductor will make the resistor voltage 5 V with minimal ripple. The ideal action circuit,
shown in Figure 3.21, takes advantage of the known answer: a 5 V ideal source at the output.
This leads to a simplified equivalent source representation: the inductor sees a ±5 V square
wave, which in turn drives a triangular current with a peak-to-peak value given by

5V = L di = L
dt
(3.20)
=
L

and some arithmetic will give the inductor value needed to achieve any specified ripple level.
 DIRECT DC-DC CONVERTERS AND FILTERS 109

L +
+ _ _.___ #2
10V - - - R sv

#1

FIGURE 3.19 Buck converter with 10 V input and 50% duty.

L
+

10V
R

FIGURE 3.20 Equivalent source and R-L model for buck converter.

L
+ +
10V sv sv
L
ov -SV ... -

FIGURE 3.21 Ideal action design for output inductor filter and simplified equivalent source model.

In power electronics, the ideal action approach generally reduces first-order or

second-order differential equation problems to quick arithmetic results. In reality, equation
(3.20) is based on the solution of a suitable set of first-order differential equations, with the
constraint that the filter must work effectively when it is done. As it turns out, ideal action is
a fundamental tool of engineering design. Here are some other examples. If data are written
correctly on a magnetic hard disk, the appropriate magnetic variation will be present. This can
be used to locate concentric tracks and guide read-write heads. For a bridge that must be 40 m
above a river surface, the elevation of the road surface becomes a known answer that can be
used to develop a solution. In a successful jetliner, the lift on the wings will support the entire
loaded weight of the aircraft, and the trusses must support this weight with limited deflection.
Since a power filter is normally intended to serve a source interface function, ideal
action can be used to advantage to simplify computations. The individual elements can be
checked sequentially while the others maintain ideal action. Here are several examples.

Example 3.4.4 A 1 mH inductor is used in the converter shown in Figure 3.22. What is the
output ripple voltage at full load?
Since the output voltage and resistance are known, and the output current is nearly con-
stant, the output current must be 10 A. When switch #1 is on, the inductor tries to maintain
110 DC-DC CONVERTERS

1 mH

L +
+
48V #1
#2
0.360 3.6 v

Switching frequency: 100kHz

FIGURE 3.22 Buck converter for Example 3.4.4.

a fixed 10 A value, but instead the current will rise slightly since the inductor is not infinite.
The inductor voltage ideally is Vin- Vout = 44.4 V during the switch #1 on-time. The inductor
current change will be

A. 44.4 v X
44.4 V = L diL
dt
, 44.4 V = L L , tilL = L
(3.21)

The duration of the switch #1 on-time will give This duration is the duty ratio 3.6/48
times the period 10 JlS (see if you can determine why this is true). The change is therefore
= 46.3 rnA, a small fraction of the 10 A average value. The output voltage ripple will be
= = 16.7mV .

Example 3.4.5 The de-de converter in Figure 3.23 uses an inductor to provide the neces-
sary current-source characteristic for the switch matrix and a capacitor to help ensure that
the user will see voltage source behavior at the output terminals. These two can be traded
off. A large capacitor, for instance, means that more inductor current ripple can be toler-
ated. Select values based on ± 50% inductor current variation and ± % output voltage i
variation.
Solved together, this R-L-C circuit analysis problem can be formulated with
second-order differential equations. However, we know what it will do and can work
backward from the result instead. Since the output voltage is intended to be fixed, treat the
capacitor as a fixed source for the purposes of choosing the inductor, and then continue on.
If the output is treated as a 5 V source, its average current must be 20 A to deliver 100 W.
Therefore the inductor current is 20 A on average and will vary between 10 A and 30 A.
While switch #1 is on, the inductor voltage is 7 V, and the current increases by 20 A. It is
easy to show that switch #1 is on 5/12 of each period, or 2.08 JlS. The inductor can be found
as follows:

2 08
= 20A, L = ?V X · J1S = 0.73J1H (3.22)
20A

For the capacitor, we can now treat the inductor as a triangular equivalent source. The resis-
tor current is supposed to be constant, so why not treat it as an equivalent source as well? An
equivalent circuit based on these sources is given in Figure 3.24.
Given that the capacitor current follows directly from the figure, the change c over
time can be found. The current is not constant, so we cannot set dvddt to the slope
 DIRECT DC-DC CONVERTERS AND FILTERS 111

I +
#1 0.25Q 5 V,
12V
100W
c

fswitch = 200 kHz

=± 50% ±Y2%

FIGURE 3.23 A de-de buck converter for Example 3.4.5.

ic(t)

10 20A
c

10 -I-

<( 5 10 15
-.....;;..
0 time (Jis)
·-u to 2.08

-10-+-

FIGURE 3.24 Equivalent ideal action current sources imposed on the output capacitor of Example 3.4.5 and
the waveform for capacitor current based on the equivalent.

as in the examples above, but ic is a triangular function of time. The voltage increases when-
ever ic is positive. From Figure 3.24, the current becomes positive while switch #1 is on, at
time t 0 = 1.04 JlS. The current rises 10 A in 1.04 JlS, which gives a slope of 9.6 x 106 A/s.
The slope while switch #1 is off can be computed as -6.86 x 106 A/s. The voltage vc will
reach its minimum value at time t 0 (just after the current has been negative for a half
cycle), then will climb until t 1, when the current again becomes negative. The net change
is Llvc = vc(t1) - vc(t0 ). Let us integrate the current during its positive half-cycle to find
v c(t 1):

1 J2.08,us 6 1 Jt1 6
vc(t1) = vc(t0 ) +- 9.6 x 10 (t- t0 ) dt +- -6.86 x 10 (t- t1) dt (3.23)
C t0 C 2.08,us

Actually, doing an integral here is overkill. We need the area under the triangle in Figure
i
3.24, bh = (2.5 J1S)(10 A)= 12.5 x 10-6 . The value vc(t1) - vc(t0) is therefore 12.5 x 10-6/C.
i
To keep the ripple below± %, the voltage must not change by more than 0.05 V. This requires
C = 250 J.LF. Figure 3.25 shows a PSpice simulation output for the complete circuit. It meets the
requirements and shows very little voltage change.
112 DC- DC CONVERTERS

10 I I I I

0 r-- -

-10 I I I I

5.1

200 210 220 230 240 250

Time (J.lS)

FIGURE 3.25 Simulation of iL and vc for 12 V to 5 V converter, Pout = 50 W, L = 0.73 ,uH, C = 250 ,uF.

For a more general formulation based on Example 3.4.5, consider again Figure 3.23, but
with a generic set of values. The inductor has a peak-to-peak current variation given by
The capacitor current will be the difference between the inductor current and its average value,
just as in Figure 3.24, with the same peak-to-peak value. The area under the triangle bh for
switching period Twill be i and the change in capacitor voltage is given by

(3.24)

This is a general result for a de-de converter intended to have current-source characteristics
and provided with an L- C filter at the output.

3.4.4 Discontinuous Modes and Critical Inductance

The power filter designs and converters considered so far have relatively large inductor and
capacitor values that hold currents and voltages nearly fixed. What if the values are not so
big? How does the analysis change? Can we design converters even when values have wide
variation or become unpredictable? The buck converter in Figure 3.26 will be used as a basis
for discussion. The large capacitor holds a fixed output voltage, and the discussion for now
emphasizes the inductor.
What if the inductor is modest compared to previous values, allowing 20% peak-to-
peak ripple? The current and voltage waveforms for an appropriate duty ratio are shown in
Figure 3.27, including the input current waveform and the capacitor current waveform. The
diode voltage is vd = q 1 as before. Even though the current varies quite a bit, KCL requires
a current path, and q1+ q2 = 1 as before. With ( vL) = 0, the output average must match
 DIRECT DC-DC CONVERTERS AND FILTERS 113

+ +
+ #1 R
c

FIGURE 3.26 Buck converter for consideration of inductor size.

11 Oo/o 11 Oo/o -r- 11 Oo/o -,..

90% 90% I""" 90% -r-
........... ........... ..........
<( <(
......... .........
iL(t) iin (t)
c: c: c:
.... .... ....
::s ::s ::s
u u u
I
0 0 I
' '
0 5 10 15 20 0 5 10 15 20 0 5 10 15 20
time (j.Js) time (j.Js) time (j.Js)

1
v.m
q, (t) >
......... vd(t) >
......... 1 II
VL(t)
I
<1J <1J I I
01 01
ro ro
-0 -0
> >
I I
0 1

' 0 I

' '
I

'
0 5 10 15 20 0 5 10 15 20
time (j.Js) time (j.Js)

FIGURE 3.27 Buck converter waveforms based on 20% peak-to-peak inductor current ripple.

\ vd), so the converter still has Vout = D1"in . The average current in the capacitor is zero, so
the average inductor current is l out· Notice from Figure 3.27 that the average input current
\ iin) is given by D 1/out· Even though in general (x(t)y(t) ) -:t (x(t))(y(t)) (try cosines to con-
firm this), for the piecewise waveform for iin in Figure 3.27, the product of averages holds and
the average input current is the same as for infinite inductance.
Now consider an even smaller inductor, allowing 100% peak-to-peak current ripple.
The currents and voltages are shown in Figure 3.28. Even though the current is not at all
constant, KCL still requires a current path. The combination of KVL and KCL requires
the switches to work in alternation such that q 1 + q 2 = 1. The average requires D 1 + D 2 = 1.
Following from above, we still have vout = Dl"in ' (iL) = I out ' and \ iin) = Dl/ out ' and so on.
These average relationships are unchanged and remain the same as for the ideal converter
with large inductance.
How far can this go? The extreme case, exemplified in Figure 3.29, allows 200%
peak-to-peak current ripple, such that iLjust reaches zero. Now the inductor current is trian-
gular with = 2 (iL ) = 2/out, but still current always flows and KCL requires q 1+ q 2 = 1
and D 1 + D 2 = 1. The average relationships remain in place. For analysis and design, this
limiting case is especially useful, and it is associated with a special value called the critical
inductance. Any larger value will ensure that the average relationships for this converter will
114 DC-DC CONVERTERS

150% 150% 150%

-
<(
.........
+-'
i;n (t)
-
<(
.........
+-'
c: c: id(t)
50%
......
Q)
......
Q)

::J ::J
(t}> u u
0 0 0
0 5 10 15 20 0 5 10 15 20 0 5 10 15 20
time t ime time (J.Ls)

1
VL (t)
I
q l(t) vd(t) ' ' '
Q) Q) '
0'1 0'1
ro ro
-g
+-'
-0
+-'

>
I I
0 ' 0 '
'
0 5 10 15 20 0 5 10 15 20
time time

FIGURE 3.28 Buck converter waveforms based on 100% peak-to-peak inductor current ripple.

200% 200% 200%

iL (t)
-
<(
.........
+-'
i;n (t) -
<(
.........
+-'
c: id (t)
./ ...
Q)
...c:
Q)

::J ::J
(t)> u u
0 0 0
0 5 10 15 20 0 5 10 15 20 0 5 10 15 20
time (J.Ls) time (J.Ls) time (JlS)

1 VL (t)
ql (t) vd(t) I I
Q) Q)
I
'
0'1 0'1
ro ro
-0
+-'
-0
+-'

> >
0 '' '' '' ''
' 0 '
0 5 10 15 20 0 5 10 15 20
time (J.Ls) time (J.Ls)

FIGURE 3.29 Buck converter waveform based on critical inductance, with 200% peak-to-peak ripple.

hold, but smaller values are a different story. As long as i L > 0 and a current path is required,
the relationships are the same. A definition follows:

Definition: Critical inductance is the smallest inductor value that always maintains
iL > 0.

From the analysis and definitions, it is not necessary that a current source maintains a fixed
value but only that it maintains nonzero current flow under all conditions. Critical induc-
tance Lcrit is not difficult to compute, since it is the value associated with a specific 200%
peak-to-peak ripple value. Notice that the amount of ripple is inversely proportional to induc-
tance. The inductor for Figure 3.28 with 100% peak-to-peak ripple has a value L = 2Lcrit·
The inductor for Figure 3.27 with 20% peak-to-peak ripple is L = 10Lcrit· To achieve peak-
to-peak ripple of 0.1%, it will be necessary to use L = 2000Lcrit' and so on. If a designer
 DIRECT DC-DC CONVERTERS AND FILTERS 115

........ ........
iL (t) +-' iin (t) +-' id(t)
c: c:
........CIJ ........CIJ
::l ::l
u u

0 5 10 15 20 0 5 10 15 20 0 5 10 15 20
time (Jls) time (Jls) time (Jls)

1
vd(t) VL (t)
q, (t) r-
I I J
• • •
CIJ CIJ
Cl Cl
ro ro
-0
+-'
-0
+-'

> >
0 I • 0 I I I I
I I • I I I I

0 5 10 15 20 0 5 10 15 20
time (Jls) time (Jls)

FIGURE 3.30 Buck converter with subcritical inductance.

determines Lcrit' the value to meet a specification is immediately evident. The result for the
relative ripple in terms of inductor ratio is

_ 2Lcrit
(3.25)
(iL ) L

What if L < Lcrit? In this case, the current will rise, but then fall all the way back to
zero, as in Figure 3.30. Once iL = 0, there is no need to provide a current path. The configu-
ration is no longer consistent with a diode on-state condition, so the diode turns off. The
"both off" configuration no longer violates KCL and is valid for part of the cycle, and the
converter is said to enter discontinuous conduction mode (DCM). There are times when
q 1 + q 2 < 1, so D 1 + D 2 < 1. What happens to the output? Look at the three circuit configura-
tions in Figure 3.31, emphasizing the voltage vd that changes most during switching. In the
third configuration, valid while the two switches are off, i L = 0 and vd will be the open-circuit
voltage Vout· With switching functions, this can be represented as

(3.26)

Even though the inductor is small, it does not support de voltage and requires (vL ) = 0. The
average of vd therefore must match Vout' and

(3.27)
V = Dl V
out D+D m
1 2

This is a generalization of the more basic result, since previously D 1 + D 2 = 1 while now D 1 +
D 2 < 1. The output is effectively higher than before, since the denominator is less than one.
116 DC-DC CONVERTERS

#1 + #1 +
R R
#2
c c

#1 on #2on

L
#1 +
R
Neither switch on V;n #2
c

FIGURE 3.31 Three configurations for a buck converter in DCM.

There is a problem the duty ratio D 2 is not known; previously it was just 1- D 1, but
not now. The average relationships will not allow us to find the output voltage or power. How
can a second equation be generated? Energy analysis can do the job. In DCM, the inductor
starts and ends each cycle with no stored energy. All injected energy is consumed in the load.
The load energy consumed per period is

(3.28)

The inductor current is triangular, and the converter input energy can be computed. Current
iL is drawn from the input Vin only while switch #1 is on, so the input energy is

(3.29)

and this must match the load energy consumed over a cycle. The current rises with slope

_ iLpeak (3.30)
D 1T

What is the peak current? With vL = L dildt, the current rises from zero to its peak while
switch # 1 is on, so

d" i
V -V =L l =L Lpeak
(3.31)
m out dt DT '
1

which means

(3.32)

Equation (3.29) now can be written out:

(3.33)
 DIRECT DC-DC CONVERTERS AND FILTERS 117

This must match the energy delivered to the load, so

2 2
YinVout (DlT) _ Vout T (3.34)
L 2 Rzoad

The expression can be solved for Vout with the quadratic formula. The quadratic form implies
two possible solutions, but this is a buck converter with positive output less than the input,
so the result is unique,

2
V - D1 VinRzoadT + D V (3.35)
out = 4L 1 zn

Although this expression is complicated, it follows directly from energy analysis. The term
RzoadT/(2L) is prominent here. It is the ratio of half the period, T/2, to the time constant
LIRzoad·
A similar situation holds for a boost converter like the one of Figure 3.32. There is a
critical inductance value Lcrit that will enforce iL > 0. Provided the inductor is larger than this
value (given a large capacitance), the average relationships hold just as if the inductor were
infinite. If L < Lcrit' there will be times when both switches are off and a third configuration
will be valid, as in Figure 3.33. Now the voltage vt can be represented as

(3.36)

Since (vL) = 0, we have

V = Dl + D2 V (3.37)
out D zn
2

with D 2 as an additional unknown. The current and voltage waveforms for a typical case with
L < Lcrit are shown in Figure 3.34.
In the DCM boost converter, duty ratio D 2 is unknown, and energy analysis can be
used to create a second equation to permit an independent solution for Vout· Just as in the buck
case, a subcritical inductor will start and end each cycle with zero stored energy. All injected
energy is delivered to the load, so the input energy must match. However, now energy flows

FIGURE 3.32 Boost converter for L

critical inductance analysis.
#2 +

#1
c
118 DC- DC CONVERTERS

L L

#2 + #2 +
+
R R
c c

#1 on #2 on

+ vL - #2
+ +
Neither on Vin R
c

FIGURE 3.33 Valid boost circuit configurations based on subcritical inductance.

iL (t) ........ ........

<( <(
........ ........
+J iin (t) +J
c: c:
id(t)
.... ....
:::l :::l
u u
0 0 0
0 5 10 15 20 0 5 10 15 20 0 5 10 15 20
time time time

1 --
VL (t)
vt (t)
q l (t) I I _j

' ' '

Q)
.... Ol
.... vin
-0
0 I
I
II '' ''
>
0 5 10 15 20 5 10 15 20
time time

FIGURE 3.34 Circuit waveforms for boost converter with subcritical inductance.

from the input source unless both switches are off, so it is nonzero when switches #1 or #2
are on. Following from the buck converter, this means

(3.38)

For the boost converter, the input voltage is imposed on the inductor while switch #1 is on, so
•
V = L di = L l Lpeak (3.39)
m dt DT'
1

and

(3.40)
 DIRECT DC-DC CONVERTERS AND FILTERS 119

The integral of the input energy can be written in terms of a triangular area, since the induc-
tor current rises and falls linearly. This leads to

2 2
W = (D1T + D2T) = W Vout T (3.41)
m 2L out
Rzoad

Now (3.37) and (3.41) can be solved simultaneously. This leads again to two solutions, but
this boost converter must have positive input larger than Vin to be useful, so only one is valid.
The result is

Vout = + ----"!..!:... (3.42)

2 2

Consider the following example.

Example 3.4.6 A boost converter with 8 V input and 24 V output has a nominal load of
200 Wand a large output capacitor. Find the critical inductance, given a 40kHz switching
frequency. Set L = L crit for this case. Now the load decreases to 24 W. What value of duty ratio
will be needed to maintain 24 V output?
When the inductor matches its critical value, the conventional boost converter average
relationships hold, and the inductor current ripple is 200% peak to peak (±100% of the aver-
age value). In a boost converter, as in Figure 3.32, the inductor carries the input current. For
8 V input and 200 W output, the average input current must be 200/8 = 25 A. Therefore, if the
inductor is equal to Lcrit' the current ripple will be double this, or 50 A. While switch #1 is on,
the current changes from 0 A to 50 A, and we have

V = L di = L 8V = L . 50A (3.43)
m dt cnt DT
1

Since the average relationships hold, the duty ratio is such that 8 V/(1- D 1) = 24 V, so D 1 = 2/3.
The period is 25 JlS, so D 1T = 16.67 JlS. From equation (3.43), this gives Lcrit = 2.67 J.LH. As
an aside, notice that this provides an immediate design result. There is no input ripple current
specification here, but if it was 10% peak to peak, for instance, an inductor value of 53.33 JlH
would do the job, and so on.
Now drop the load to 24 W, with an average output current of 1 A and a load resistance
of 24 Q. Since the input current is much lower, an inductance value of 2.67 JlH is no longer suf-
ficient to maintain nonzero current flow and we expect L < Lcrit in this scenario. From equation
(3.42), we have

24V = 8V + 8V 1 + = 4 + + 450D2 V (3.44)

1
2 2 2.67 JLH

and D 1 = 0.231 to maintain 24 V output (much less than the 2/3 value needed when L Lcrit).
Some of the waveforms for this case are shown in Figure 3.35. Even though the output
current is 1 A and the average input current is 3 A, the peak inductor current is more than
17 A here!
120 DC-DC CONVERTERS

30 I I I

20 f- -
Q)
Q) Vt
co
-
+-'

g 10 f- -

0
20 I I I

•
• • • IL • • • • •
1- •
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
-
• • • • • • • •
• • • • • • • •
• • • • • • • •
• • • • • • • •
• • • • • • • •
• • • • • • • •
• • • • • • • •
• • • • • • • •
• • • • • • • •
•
50 100 150 •
• 200
time (jls)

FIGURE 3.35 Current and voltage waveforms for Example 3.4.6.

The value of Lcrit is usually straightforward to compute, since when L = Lcrit' the inductor
is a triangle with minimum current i L = 0+ and both the DCM and conventional duty ratio
relationships are valid simultaneously. For a buck converter, the value can be found by setting
Vout = D 1Vin in equation (3.34) or (3.35). This gives

buck: L . = RzoadT (1 - D) (3.45)

crzt 1
2

For the boost case, Vout = Vin/(1-D1) at the critical value, and

boost: L . = RzoadT D (1 - D )2 (3.46)

crzt 1 1
2

Given the extreme peak currents and the fact that the converter does not transfer
energy part of the time, why consider DCM? A typical converter will enter DCM when
lightly loaded, so it is important to recognize. The duty ratio becomes linked to the load
value, degrading load regulation and complicating operation. Even so, there are some cir-
cumstances that favor DCM over continuous conduction mode (CCM). Table 3.2 lists some

TABLE 3.2 Advantages and Disadvantages of Discontinuous Conduction Mode in

De-De Converters
Advantages Fast action. Every cycle starts from zero energy in the subcritical device,
driving to periodic steady state quickly.
Some switch actions occur at zero current. This reduces loss for some devices.
Small inductors with low values and low energy storage are used.
Disadvantages The converter is less well used, with no energy transfer part of the time.
Current peaks can be extreme.
Operation depends on load.
 INDIRECT DC-DC CONVERTERS 121

advantages and disadvantages of DCM. Some will be explored later in this chapter with
respect to grid power flow.
There is also a dual concept, critical capacitance, that will be explored later:

Definition: Critical capacitance is the lowest value of capacitor that always main-
tains vc > 0.

In the analysis and examples so far, it has been assumed that the capacitor is large. Based
on this definition, more formally we have been treating the capacitor as substantially greater
than its critical value.

3.5 INDIRECT DC-DC CONVERTERS

3.5.1 The Buck-Boost Converter

The direct de-de converters offer basic conversion functions. The limitations that Vout Vin
for the buck and V out Vin for the boost are the most obvious restrictions. How can we create
a more complete de transformer? One way is to cascade the two direct converters to form
an indirect converter. The two converters can be adjusted independently to give any desired
output ratio. This cascade arrangement is developed in Figure 3.36.
The converter shown in Figure 3.36 is a known circuit, called the four-switch
buck-boost converter, and is used sometimes because of its flexibility [8]. However, it has two
controlled switches, and operation can be complicated. Some of the switches in the cascade
can be eliminated based on redundancy and the zero average power target in the transfer
current source. Only two switches are essential in the final reduced version, as shown in
Figure 3.37. The reduced common-ground circuit is referred to as a buck-boost converter or
an up-down converter since outputs of any magnitude are possible, but the polarity is nega-
tive; this is the polarity-reversing circuit introduced in Chapter 1. The four-switch version
avoids the reversal at the cost of extra complexity.
Let us analyze this circuit with emphasis on the transfer source, since its voltage
changes substantially during switch action. The KVL and KCL restrictions require a current

#4 +
#1 #2 R
#3
c

FIGURE 3.36 Cascading buck and boost converters to deliver any voltage ratio.
122 DC-DC CONVERTERS

(a) (b)

Buck and boost Boost upside down

(c) (d) (e)

Share current source Remove redundancy Final reduced version

FIGURE 3.37 Buck-boost converter, showing reduction from the four-switch version.

path and avoidance of voltage shorting, so one and only one switch can be on at a time. The
voltage across the transfer source, vt, is Vin when switch #1 is on and Vout when #2 is on.
The transfer current source value is Is. To meet the basic objectives of power electronics, the
transfer source must have (pt) = 0. The relationships are

vt = + q2Vout (3.47)
Pt = vtls = qlV. Is + q2Voutls
m

\Pt) = D1 + D2Voutls = 0

In any case with Is > 0 (which implies L LcriJ, KVL and KCL require D 1 + D 2 = 1. The last
part of equation (3.47) can be satisfied for nonzero Is if D 1Vin = -D2Vout· Simplifying,

(3.48)

The other variables determined by switch action include the input and output currents. These
bring about the relationships

(3.49)
( iin) + (iout) = Is

Dl (iout) = D2 (iin)

The switch reduction process required the polarity of the output boost converter to
be inverted: the buck-boost converter produces a negative voltage with respect to the input,
 INDIRECT DC-DC CONVERTERS 123

as seen previously in polarity reversal examples. The output in principle can range from 0
to -oo. It is zero if D 1 = 0 and goes to infinity as D 1 goes to one. When D 1 = the output i,
magnitude is equal to the input. The buck-boost converter provides a de transformer func-
tion except for the polarity reversal. Notice that switch #1 carries Is when on and blocks
IVinl + IVoutl when off. A forward-carrying forward blocking device is needed. The second
switch can be a diode with these same ratings. The output load requires a capacitor to give it
voltage source properties. What about the transfer current source? It must maintain constant
current without power loss the function of an inductor. Let us implement a converter simi-
lar to that in Example 3.3.3 to evaluate the buck-boost converter.

Example 3.5.1 Figure 3.38 shows a buck-boost converter, intended to provide -120 V
output at 50 W with +5 V input. What are the switch duty ratios for this function? What are
the switch ratings for voltage and current? What values of Land C will keep the output varia-
tion below ±0.1% and the transfer current variation below ±1%? The switching frequency is
20kHz.
Given that Is and Vout are nearly constant and the polarity reversal has been taken
into account, we expect D 1Vin = D 2Vout· Then D/D 2 = 24 and D 1 + D 2 = 1. This requires
D 1 = 24/25 while D 2 = 1/25. Since switch #1 is on almost the full cycle, energy in the inductor
is built up over a long time interval, and then released quickly into the load. The average input
current must be 10 A to provide the 50 W input, while the average output current must be
0.417 A for this power level. The transfer source value is Is = (iin) + (iout) = 10.417 A. Each
switch must carry 10.417 A when on and must block 125 V when off. The transfer source
variation is to be less than ±1%, or ±0.104 A. This translates to a total current change of
0.208 A. While switch #1 is on, the transfer voltage vt is 5 V, and the current rises since
vL = L dildt is positive. The result is

5 V = L di = L lli lli 0.208 A

dt flt' (3.50)

For the capacitor, the allowed voltage fall is ±0.1%, or 0.24 V. This drop will occur dur-
ing the switch #1 on interval, or 24/25 of a period. The result is

i = cdv c llv llv 0.24 V

c dt flt' (3.51)

C 83.3 J.lF

•
lout

+ #2
#1 L
sv R 120V
c +
R=2880

FIGURE 3.38 A realization of the buck-boost converter.

124 DC-DC CONVERTERS

One common mistake in solving buck-boost problems involves the value of Is. This value
is not the same as either lin or lout' but instead is their sum. The following example illustrates
this issue.

Example 3.5.2 A buck-boost converter transfers energy from a +12 V source to a -12 V
source. The rated output load is 30 W. What are the switch duty ratios and what is the transfer
current source value? What are the switch rating requirements?
Since IVoutl = Vin' the duty ratios should be D 1 = D 2 and D 1 = i.
The average value of lin
is Pi/Vin = 30/12 = 2.5 A. The value of Is is 2.5/D1 = 5 A. The transfer current source here is
double the input and output average currents. The switches must carry 5 A when on and block
IVinl + IVoutl = 24 V when off.

3.5.2 The Boost-Buck Converter

Just as a buck converter can be cascaded with a boost converter for conversion, so a boost
converter can supply a buck circuit. The idea is shown in Figure 3.39. As in the buck-boost
case, there is a four-switch version that is rarely used, but in the final analysis, only two
switches are required. The final reduced circuit, with energy storage elements in place, is
shown in Figure 3.40. The center capacitor serves as a transfer voltage source. It takes energy
in from the input side, stores it briefly, and then sends it on to the load. The transfer source
should exhibit (pt) =O.
Switch action will determine the input voltage, the output voltage, and the current in
the transfer source. Some of the major relationships are

Vin(t) = q2Vr,

V ou/t) = qlVr,

it = q2/ in + ql/out (3.52)

Pt = q2/in Vr + ql/ outVr
\ Pt J = 0 = D2/in + Dl/out
D2/in = - Dl/ out

provided the transfer voltage Vt is nonzero. Like the buck-boost circuit, the boost-buck
converter exhibits a polarity reversal between input and output. In the literature, the
two-switch arrangement is called a Cuk converter, after the developer who patented it in
the mid-1970s.
Figure 3.41 shows the configurations of this converter in CCM. Each switching device
must carry the current sum 1/inl + 1/outl when on, and each must block the transfer voltage Vt

l -vt .. t
t vt -
+ +
v-
t +

FIGURE 3.39 Boost converter supplying a buck converter to yield a four-switch boost-buck circuit.