DC-DC Converter
DC-DC Converter
= = = =
} }
The fundamental, second harmonic, and third harmonic amplitudes are found from
20
20
2,1
0
0
4 2 60 60 2
15cos sin 8.27
60 60 60 2 60
s
s
S
s
V t dt t V
s s s s
t t
t
| | | |
= = =
| |
\ . \ .
}
20
20
2,2
0
0
4 2 60 60 2
15cos 2 sin 2 4.13
60 60 60 4 60
s
s
S
s
V t dt t V
s s s s
t t
t
| | | |
= = =
| |
\ . \ .
}
20
20
2,3
0
0
4 2 60 60 2
15cos 3 sin 3 0
60 60 60 6 60
s
s
S
s
V t dt t V
s s s s
t t
t
| | | |
= = =
| |
\ . \ .
}
Thus, we can write the Fourier series for this waveform as
( ) ( )
2
10 8.27 cos 2 16.7 4.13 cos 2 33.3 ...
S
V V V kHz t V kHz t t t = + +
This exposes that the waveform has a DC component, the fundamental at 16.7kHz, and
the second harmonic at 33.3kHz. Due to the waveform symmetry, the third harmonic is
not present. The fourth harmonic at 66.7kHz has an amplitude of 2.07V. The key here is
that we want to filter these signals out so that only the DC component remains.
6
This requires a low-pass filter and we propose the circuit shown in Figure 1.4 where R
represents the applied load. Clearly the circuit is low-pass since at low frequencies
( 0 e ) the inductor ( L e ) appears as a short and the capacitor (1/ C e ) is an open and
the input passes through to the output. At high frequencies (e ), the inductor looks
like an open and the capacitor a short and the input is blocked from the output.
Figure 1.4: Low-Pass Filter within a Buck Chopper
The transfer function for this circuit is given by
2 2
2
1
( ) 1
1 1
( )
1
out
S
V s
LC
L
V s
s s LCs s
RC LC R
= =
+ + + +
(1.2)
If we assume that the denominator roots are complex (the usual situation), the
approximate straight-line frequency response magnitude characteristic will have a corner
frequency at
1
o
LC
e = as shown in Figure 1.5. The fundamental switching frequency
(in rad/s) relates to the switching period and switching frequency (in Hz) by
2
2
SW SW
SW
f
T
t
e t = = (1.3)
Note in Figure 1.5 that it is desirable for the switching frequency to appear a couple of
decades higher than
o
e so that the ripple in
2 S
V is sufficiently attenuated. In this
example, the ripple at
SW
e is reduced by -80dB or by a factor of
4
1 10
absolute. What
this is starting to uncover is that there is a direct relationship between the choice of
switching frequency and the selection of L and C.
7
10
3
10
4
10
5
10
6
10
7
-120
-100
-80
-60
-40
-20
0
Frequency (rad/s)
M
a
g
n
i
t
u
d
e
(
d
B
)
Figure 1.5: Approximate Frequency Response Magnitude Plot for Buck Output Filter
Now the filter discussion is useful in uncovering that 1/
SW
LC e >> , but it does not
provide guidance in how to divide up the contributions of L and C nor does it concretely
link to desired figures of merit like output ripple voltage or inductor ripple current. Thus
we push onward by analyzing the combination of the circuits from Figures 1.2 and 1.4 as
shown in Figure 1.6. Here we have replaced switch
2
S by an ideal diode. Thus when
switch
1
S opens, any positive inductor current will clamp the diode ON and it will
remain on as long as positive current can flow or up until
1
S recloses (thereby reverse
biasing the diode and turning it OFF).
Figure 1.6: Buck Chopper
8
The preliminary analysis of this circuit utilizes the following assumptions
- C is large enough that the output voltage ripple is small relative to its
average value;
- L is large enough to ensure that the inductor current stays positive for the
switching period (this is referred to as continuous conduction mode or
CCM) this ensures that when the switch is OFF, the diode must be ON;
- All components are initially assumed ideal we will eventually find that
1
S ,
2
S , L and C all have parasitic elements that must be considered in
properly identifying parts;
- The circuit is in the steady state implying that all waveforms are in fact
periodic, ensuring that they have the same value at the beginning and end
of a switching period.
Owing to the second assumption, the circuit will have two admissible states as shown in
Figure 1.7: switch closed and diode OFF or switch open and diode ON. Next we will use
the first assumption to deduce the wave-shape of the inductor current. First lets re-
express the amount of time that the switch and diode are conducting in terms of the duty
cycle D
1
1
S
S SW SW
SW
T
T T DT
T
= = (1.4)
( )
2 1
1
S SW S SW SW SW
T T T T DT D T = = = (1.5)
If the output voltage ripple is very small, then we can assume that the output voltage is
constant at its average value,
, out ave
V . Thus when switch
1
S is ON (Figure 1.7a) we can
use KVL to derive an expression for the inductor voltage
,
,
L ON
L IN out ave
dI
V L V V
dt
= = (1.6)
a.
1
S ON &
2
S OFF b.
2
S ON &
1
S OFF
Figure 1.7: Two Valid Topological States for a Buck Chopper in CCM
9
Since both
IN
V and
, out ave
V are constant, the derivative is a constant and we can replace it
by the change in current over the change in time. Thus (1.6) becomes
, , L ON IN out ave
ON
I V V
t L
A
=
A
(1.7)
Substituting for the amount of time that
1
S is ON (Equation (1.4)) and solving for the
change in current, we get
,
,
IN out ave
L ON SW
V V
I DT
L
| |
A =
|
\ .
(1.8)
Note that since
, out ave IN
V V < , the current is linearly building during this interval as shown
in Figure 1.8. When switch
1
S is OFF and the diode is conducting, the circuit shown in
Figure 1.7b governs and now the inductor voltage is given by
,
L ,
V 0
L OFF
out ave
dI
L V
dt
= = (1.9)
Since we are still assuming that the output voltage is constant, the derivative is again
constant so we can rewrite (1.9) as
, , L OFF out ave
OFF
I V
t L
A
=
A
(1.10)
Substituting (1.5) for the amount of time that the diode is assumed to be ON (Equation
1.5) yields
( )
,
,
1
out ave
L OFF SW
V
I D T
L
A = (1.11)
Now the current is linearly decreasing (the derivative is negative) as shown in Figure 1.8.
For the circuit to be in the steady state and all variables periodic implies that the inductor
current must return to the same value at the end of the cycle (indicated here as
min
I ) or
mathematically that
, ,
0
L ON L OFF
I I A + A = (1.12)
If we substitute (1.8) and (1.11), we find that
10
( )
, ,
1 0
IN out ave out ave
SW SW
V V V
DT D T
L L
| |
=
|
\ .
(1.13)
After canceling L and
SW
T , we are left with
, , ,
0
IN out ave out ave out ave
DV DV V DV + = (1.14)
or simply that
, out ave IN
V DV = (1.15)
which thankfully is what we found from our low-pass filter considerations. As shown in
Figure 1.8, the steady-state inductor current signal is triangular with the peak-to-peak
current designated by
L
I A .
V
S2
V
IN
0V
t
T
SW
I
max
I
min
I
L
0A
I
L
t
I
L,ave
I
L
2
I
L
2
I
C t
+ +
- - -
0V V
cpp
V
c
DT
SW
(1-D)T
SW
V
out,ave
Figure 1.8: Steady-State Buck Chopper Waveforms Assuming CCM
11
To correctly arrive at the capacitor current waveshape, first recall that the capacitor
current constitutive relationship is given by
C
C
dV
I C
dt
= (1.16)
If the capacitor current has an average value, then
, C ave
V is not constant. Thus in the
steady state, it is necessary that
,
0
C ave
I = (think of this as if the average capacitor current
was positive, then the capacitor would on average be charging to a higher voltage value).
Furthermore, by applying KCL at the output circuit node, it follows that
, , , L ave C ave R ave
I I I = + (1.17)
Therefore
,
, ,
out ave
L ave R ave
V
I I
R
= = (1.18)
We can then determine expressions for the maximum and minimum inductor current
values since we established (from 1.11 where that expression gives a negative value so
we flip the sign) that
( )
( )
, ,
1
1
out ave out ave
L SW
SW
V V D
I D T
L Lf
A = = (1.19)
Thus
max ,
2
L
L ave
I
I I
A
= + (1.20)
min ,
2
L
L ave
I
I I
A
= (1.21)
What does this all mean? Well, the average inductor current directly depends on the
applied load resistance and the desired output average voltage (as set by the duty cycle).
12
A small value of R corresponds to a large value of average current and vice versa. The
inductor ripple current does not depend on the load resistance, thus once the switching
frequency is set, the ripple current is set as shown in Figure 1.9. We see something
interesting in Figure 1.9, as the load power drops (
, L ave
I + and R |) the inductor current
will eventually hit zero while the diode is conducting, turning the diode off (before the
switch
1
S turns ON). This will transition the circuit into the Discontinuous Conduction
Mode (DCM) and the nice steady-state linear relationship
, out ave IN
V DV = no longer holds.
A typical DCM inductor current waveform is shown in Figure 1.10. Fortunately, a buck
chopper typically employs feedback control to compensate for component parasitics,
changes in load, and changes in the input so it will automatically find the proper D to
achieve the desired output voltage.
Figure 1.9: Inductor Currents in CCM for Varying Amounts of Load Resistance
13
t
DT
SW
(1-D)T
SW
I
L
Figure 1.10: Inductor Current in DCM
The transition point between CCM and DCM is a useful point to derive. It occurs when
the minimum value of the inductor current just hits zero or
min ,
0
2
L
L ave
I
I I
A
= = . If we
substitute (1.18) and (1.19), we come up with
, ,
(1 )
0
2
out ave out ave
SW crit
V V D
R f L
= (1.22)
We can solve this for the critical inductance (
crit
L ) value when the load resistance is the
largest value that we would like to maintain the converter in CCM and we find that
(1 )
2
big
crit
SW
D R
L
f
= (1.23)
Equation (1.23) offers one approach to designing the inductor value for the buck chopper.
The resistance
big
R is given by
2
, out ave
big
crit
V
R
P
= (1.24)
Where
crit
P is the output power level where you want your buck chopper to transition into
DCM. A smaller
crit
P will lead to a larger L and a smaller inductor ripple. A larger value
of inductance will mean a physically larger inductor with more losses, which sounds
unattractive. But if one shoots for a larger
crit
P so that L is smaller and the ripple current
is bigger, what is the downside? Well, we will find that this requires a larger capacitor
with more ripple current capability which sounds like there is an engineering tradeoff
afoot! Unfortunately equations (1.23) and (1.24) really do not illustrate this tradeoff so
we will have to figure out something better.
14
Example 1.3: Given an ideal buck chopper with 20
in
V V = and the following inductor
current waveform
Determine (a) the duty cycle and switching frequency, (b) the average output voltage, (c)
the circuit inductance, (d) the load resistance for the operating point pictured, and (e) the
load resistance that transitions the converter to discontinuous conduction mode.
a. The duty cycle is given by the ratio of ON time (ramp-up time) to switching
period. The switching frequency is the inverse of the switching period so
6
0.6
10
s
D
s
= =
1
100
10
sw
f kHz
s
= =
b. The ideal input/output relationship for a buck chopper is
, out ave in
V DV = thus
,
0.6 20 12
out ave
V V V = =
c. We can use either (1.8) or (1.11). Rearranging (1.8) to solve for the inductance
yields
,
,
20 12
0.6 10 12
8 4
in out ave
sw
L ON
V V
V V
L DT s H
I A A
| |
| |
= = =
|
|
|
A
\ .
\ .
d. The average value of the inductor current is in the center of the waveform, so it
follows that ( )
,
4 8 / 2 6
L ave
I A A A = + = . Since the average capacitor current must
be zero, the average load current must equal the average inductor current and
,
,
12
2
6
out ave
out ave
V
V
R
I A
= = = O
e. The transition to DCM occurs when
,
4
2
2 2
L
L ave
I A
I A
A
= = = . This occurs when
the output resistance equals 12 / 2 6
big
R V A = = O
15
Example 1.4: Consider that we want to design a buck chopper that will step 24V down to
12V with a maximum output power of 100W. For a switching frequency of 40kHz,
investigate the change in design if (a) 10
crit
P W = or (b) 20
crit
P W =
a. First, we evaluate the required duty cycle
,
/ 12 / 24 0.5
out ave in
D V V V V = = = , Next
we calculate the load resistance corresponding to the critical power
( )
2
12
14.4
10
big
V
R
W
= = O
The critical inductance is then given by (1.23)
( ) 1 0.5 14.4
90
2 40, 000
crit
L H
Hz
O
= =
With proper design of the output capacitor, most of this ripple current will then
flow through the capacitor (so the output current ripple is small)
b. Repeating the design for a higher critical power yields
( )
2
12
7.2
20
big
V
R
W
= = O
The critical inductance becomes
( ) 1 0.5 7.2
45
2 40, 000
crit
L H
Hz
O
= =
A | |
= = +
|
\ .
(1.33)
We can rewrite this in terms of the ripple ratio r as
( ) ( )
2
2
, , 2
, , ,max , ,max
, ,max
1
1 1
1 1 2
1
2 2 2
out ave out ave
L pk L ave L ave
SW L ave SW
r
V D V D
r
E I I
rf I f r
| |
+
|
| |
\ .
= + =
|
\ .
(1.34)
This expression tells us how the energy in the inductor will vary with a given value of
ripple ratio. The other variables in (1.34) are generally specified (output voltage and
maximum load) while the switching frequency will need to be selected based on other
factors. If we plot the normalized version of (1.34) for a range of 0 2 r < s
2
, ,
1
2
L pk norm
r
E
r
| |
+
|
\ .
= (1.35)
we get the plot shown in Figure 1.12. An r = 2 implies that the rated load condition is at
the boundary of DCM. Note that below r = 0.2, the required inductor energy (think size
of inductor) increases very quickly. If r is small, the ripple current is small which
necessitates a large inductance and therefore a large peak energy capability. Beyond r =
0.6, the normalized energy decreases rather slowly, so we are not gaining much more
advantage. We will next see that there is an added disadvantage to designing with a larger
value of r.
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
2
4
6
8
10
12
14
16
18
20
22
r (ripple ratio)
Normalized Inductor Energy
Figure 1.12: Plot of Normalized Inductor Energy versus the Ripple Current Ratio
21
Why are larger values of r unattractive? Lets start with the capacitor current and
determine its RMS value. If we consider the signal
C
I shown in Figure 1.8, we must find
the RMS value of a symmetrical triangle wave. Omitting the details, this value equals its
peak value divided by the square root of three, or
,
/ 2
3 3 12
pk
L L
C rms
I
I I
I
A A
= = = (1.36)
If we would prefer to write this in terms of the ripple ratio
, ,max
L
L ave
I
r
I
A
= , then we get
, ,max
,
12
L ave
C rms
I
I r = (1.37)
This implies that we suffer a penalty in selecting r too large a larger capacitor ripple
requirement which will mean more capacitance or more capacitors.
Thus if r < 0.2, the ripple is small, so the inductance is large and since for transient
concerns we try to match the capacitor energy to the inductor energy, this implies a large
value of capacitance. For r > 0.6, the inductor energy goes down more slowly and the
capacitor ripple current requirement linearly increases, implying that more capacitance is
required. Therefore we postulate that an attractive range for the ripple ratio is
0.2 0.6 r s s . This corresponds to a critical power range of 0.1 0.3
rat crit rat
P P P s s . By
investigating this range across the design space (the set of reasonable parameter values
and components) and evaluating a figure of merit like efficiency, size, or cost, we can
then determine a best choice for r. MATLAB or EXCEL are excellent programs that
can facilitate this spread sheet approach to design!
Our analysis thus far has assumed ideal components, switches that have zero voltage
drops. If
IN
V and
out
V are above 50V, this is not a bad assumption. But what if the input
and output voltage are less than 12V? Then a 0.5V-1V switch drop will start to influence
the calculations. As a result, lets incorporate these voltage drops in our analysis and
reformulate our expression for
crit
L . The analysis proceeds as before using the circuits
shown in Figure 1.13.
The change in inductor current while the switch is closed (Figure 1.13a) is found to be
,
,
IN SW out ave
L ON SW
V V V
I DT
L
A = (1.38)
where
SW
V is the assumed constant voltage drop across the switch. Similarly using Figure
1.13b for when the diode is conducting, we can find the change in inductor current to be
22
( )
,
,
1
F out ave
L OFF SW
V V
I D T
L
A = (1.39)
where
F
V is the assumed voltage drop across the diode. In the steady state we can assert
that
, ,
0
L ON L OFF
I I A + A = (1.40)
which upon substituting and simplifying yields
, out ave F
IN SW F
V V
D
V V V
+
=
+
(1.41)
(a) (b)
Figure 1.13: Buck Chopper Analysis with Switch and Diode Replaced by Voltage Drops:
(a) Switch Closed, Diode OFF; (b) Switch Open, Diode ON
Since the change in current shown in Equation (1.39) is negative, we can flip the sign and
solve for the inductance as
( )( )
,
1
out ave F
crit
L SW
V V D
L
I f
+
=
A
(1.42)
or finally in terms of the ripple ratio
, ,max
L
L ave
I
r
I
A
=
( )( )
,
, ,max
1
out ave F
crit
SW L ave
V V D
L
rf I
+
= (1.43)
Thus upon establishing estimates of the average switch and diode voltage drops, you can
use (1.41) and (1.43) to determine the critical inductance. As can be the case, if the input
voltage varies, then you will want to determine the worst case (largest) required
inductance. From (1.43), the inductance gets bigger when D gets smaller. Therefore
from (1.41), D gets smaller when the input is its largest value.
23
Example 1.7: Consider that we want to re-design the buck chopper from Example 1.4
that steps 24V down to 12V with a maximum output power of 100W with a switching
frequency of 40kHz and 10
crit
P W = (r = 0.2). This time lets assume that we have a
switch voltage drop of 1.8
SW
V V = and a diode drop of 1.2
F
V V = .
Equation (1.41) gives us an estimate of the new required duty cycle
12 1.2
0.564
24 1.8 1.2
V
D
V V V
+
= =
+
which is slightly larger than the ideal analysis value of 0.5. We can then evaluate the
critical inductance using (1.43)
( )( ) 12 1.2 1 0.564
86.3
100
0.2 40, 000
12
crit
V V
L H
W
V
+
= =
Here, we see that the inductance did not change appreciably (decreasing from the ideal
value of 90 H ). With the ripple ratio set to
, ,max
0.2
L
L ave
I
r
I
A
= = . The inductor ripple
current must equal
, ,max
100
0.2 0.2 1.67
12
L L ave
W
I I A
V
A = = =
as we had before. Thus, the sizing of the capacitor as performed in Example 1.5 would be
unchanged and the result in Example 1.6 would be slightly smaller since the value of L is
slightly smaller.
Example 1.8: Lets explore how a design proceeds when the input voltage can change.
Lets assume that we wish to design a DC-DC converter that will step a 12V battery
down to 5V. The terminal voltage of the battery can vary between 11V and 14V based on
the state of charge. The rated output power is specified as 15W. We are asked to use a
switching frequency of 20kHz and a ripple ratio of 0.2 ( ) /10 15 /10 1.5
crit rat
P P W W = = = .
The output ripple voltage should be no more than 1% of the average output voltage.
Establish values for the buck chopper inductance and capacitance. Assume that the
selected switch and diode have voltage drops of 0.3V and 0.5V, respectively.
OK, so the issue is which value of
in
V do we use in (1.41) to estimate the duty cycle. To
convince ourselves, lets work through the calculations using both. Starting with the
minimum input voltage we get
24
1
5 0.5
0.491
11 0.3 0.5
V V
D
V V V
+
= =
+
The corresponding value of critical inductance is then found from (1.43)
( )( )
,1
5 0.5 1 0.491
233
15
0.2 20, 000
5
crit
V
L H
W
V
+
= =
If we repeat for the larger value of input voltage, the new duty cycle is
2
5 0.5
0.387
14 0.3 0.5
V V
D
V V V
+
= =
+
And the new critical inductance is
( )( )
,2
5 0.5 1 0.387
281
15
0.2 20, 000
5
crit
V
L H
W
V
+
= =
So clearly, the worst-case design must occur at the higher input voltage which presents us
with a larger required inductance. The ripple current is fixed by the ripple ratio (or the
critical power level) ( )
, ,max
0.2 0.2 15 / 5 0.6
L L ave
I I W V A A = = = . Thus we can then
estimate the minimum capacitance for steady-state ripple from (1.27)
( )
min
0.6
75
8 8 20 0.01 5
L
sw Cpp
I A
C F
f V kHz V
A
= = =
A
If we contrast this with equating the energies in the capacitor and inductor at rated power,
we get
( )
( )
2
2
,
2 2
,
281 3 0.6
146
5
crit L pk
out ave
L I H A A
C F
V
V
+
= = =
So for transient reasons, we might like to choose a value closer to 150 F . Finally
though, we will learn in the next chapter that the capacitor must be able to handle the
inductor ripple current which is specified via the rms value found in (1.36)
,
0.6
173
12 12
L
C rms
I A
I mA
A
= = =