Chapter 8. Converter Transfer Functions: 8.1. Review of Bode Plots
Chapter 8. Converter Transfer Functions: 8.1. Review of Bode Plots
L
1:D D' : 1 Output
+
–
Line +
input i(s)
(Vg – V) d(s)
vg(s) + Zin(s) I d(s) I d(s) C v(s) R Zout(s)
–
v(s) v(s)
Gvg(s) = Gvd(s) =
vg(s) d (s)
d(s) = 0 v g(s) = 0
Bode plot of control-to-output transfer function
with analytical expressions for important features
80 dBV
|| Gvd || || Gvd || Gvd
60 dBV
Gd0 = V Q = D'R C
DD' L
40 dBV f0 –40 dB/decade
D' Vg
20 dBV 2/ LC t 2D'LC DVg
0˚ 10 -1/2Q
f0 t(D') 3RC
0 dBV fz 0˚
Gvd fz /10 D' 2R –20 dB/decade
–20 dBV 2/DL –90˚
(RHP)
–40 dBV –180˚
10 1/2Q f0 10fz
–270˚
–270˚
10 Hz 100 Hz 1 kHz 10 kHz 100 kHz 1 MHz
f
Design-oriented analysis
G dB
= 20 log 10 G Actual magnitude Magnitude in dB
1/2 – 6dB
Decibels of quantities having 1 0 dB
units (impedance example):
normalize before taking log 2 6 dB
5 = 10/2 20 dB – 6 dB = 14 dB
Z
Z = 20 log 10 10 20dB
dB Rbase
1000 = 103 3 ⋅ 20dB = 60 dB
Bode plots are effectively log-log plots, which cause functions which
vary as fn to become linear plots. Given:
n
f
G =
f0 60dB 2
–40dB/decade f
Magnitude in dB is 40dB f0
n f
f f =
2
G = 20 log 10 = 20n log 10 20dB n f0
dB f0 f0 –20dB/decade
n=
1
0dB
n=
• Slope is 20n dB/decade 20 dB/decade –1 f –1
–20dB
n f0
• Magnitude is 1, or 0dB, at =
–2
–40dB –2
frequency f = f0 40dB/decade
f
f0
–60dB
0.1f0 f0 10f0
f
log scal
8.1.1. Single pole response
G(s) = 1
–
1 + sRC
with t0 = 1
RC
G(jω) and || G(jω) ||
Let s = jω:
t
1– j t Im(G(jt))
G( jt) = 1 = 0
t t 2 G(jt)
1+ j t 1+ t
0 0
|
)|
(jt
Magnitude is
G
||
2 2
G( jt) = Re (G( jt)) + Im (G( jt)) G(jt)
= 1
t 2 Re(G(jt))
1+ t
0
Magnitude in dB:
G( jt) = – 20 log 10 t
1+ t
2
dB
dB 0
Asymptotic behavior: low frequency
t 2 t 2 0dB
1+ t 5 t 0dB
0 0
–1
f
Then || G(jω) || –20dB
f0
becomes –20dB/decade
–40dB
–1
1 f
G( jt) 5 = –60dB
t 2 f0
t0 0.1f0 f0 10f0
f
t0 2
G( jt0) dB
= – 20 log 10 1+ t 5 – 3 dB
0
|| G(jt) ||dB
0dB
1dB 3dB
0.5f0 1dB
f0
–10dB
2f0
–20dB –20dB/decade
–30dB
f
Phase of G(jω)
Im(G(jt))
G(jt) t
1– j t
G( jt) = 1 = 0
t
1+ j t 1+ tt 2
|
)| 0 0
(jt
G
||
G(jt)
–1
t
Re(G(jt)) G( jt) = – tan
t0
–1
Im G( jt)
G( jt) = tan
Re G( jt)
Phase of G(jω)
–1
t
0˚ 0˚ asymptote G( jt) = – tan
G(jt) t0
-15˚
-30˚ ω ∠G(jω)
-45˚ -45˚
0 0˚
f0
-60˚
ω 0
–45˚
-75˚
–90˚ asymptote
-90˚ ∞ –90˚
0.01f0 0.1f0 f0 10f0 100f0
f
Phase asymptotes
Low frequency: 0˚
High frequency: –90˚
Low- and high-frequency asymptotes do not intersect
Hence, need a midfrequency asymptote
fa = f0 e – / / 2 5 f0 / 4.81
fb = f0 e / / 2 5 4.81 f0
Phase asymptotes
fa = f0 / 4.81
0˚
G(jt)
-15˚
-30˚
fa = f0 e–//2
5 f0 / 4.81
fb = f0 e / / 2 5 4.81 f0 -45˚ -45˚
f0
-60˚
-75˚
-90˚
0.01f0 0.1f0 f0 fb = 4.81 f0 100f0
f
Phase asymptotes: a simpler choice
fa = f0 / 10
0˚
G(jt)
-15˚
-30˚
fa = f0 / 10
-45˚ -45˚
fb = 10 f0
f0
-60˚
-75˚
-90˚
0.01f0 0.1f0 f0 fb = 10 f0 100f0
f
Summary: Bode plot of real pole
0dB 1
|| G(jt) ||dB G(s) =
1 + ts
1dB 3dB
0.5f0 0
1dB
f0
2f0
–20dB/decade
0˚ f0 / 10
G(jt) 5.7˚
-45˚/decade
-45˚
f0
-90˚
5.7˚
10 f0
8.1.2. Single zero response
Normalized form:
G(s) = 1 + ts
0
Magnitude:
G( jt) = t
1+ t
2
Use arguments similar to those used for the simple pole, to derive
asymptotes:
0dB at low frequency, ω << ω0
+20dB/decade slope at high frequency, ω >> ω0
Phase:
–1
t
G( jt) = tan
t0
—with the exception of a missing minus sign, same as simple pole
Summary: Bode plot, real zero
G(s) = 1 + ts +20dB/decade
0
2f0
f0
0.5f0 1dB
0dB 1dB 3dB
|| G(jt) ||dB
10 f0 +90˚
5.7˚
f0
45˚
+45˚/decade
G(jt) 0˚
5.7˚
f0 / 10
8.1.3. Right half-plane zero
Normalized form:
G(s) = 1 – ts
0
Magnitude:
G( jt) = t
1+ t
2
G(s) = 1 – ts +20dB/decade
0
2f0
f0
0.5f0 1dB
0dB 1dB 3dB
|| G(jt) ||dB
0˚ f0 / 10
G(jt) 5.7˚
-45˚/decade
-45˚
f0
-90˚
5.7˚
10 f0
8.1.4. Frequency inversion
G(s) = 1 0dB
t0 3dB 1dB
1+ s 2f0
1dB
f0
0.5f0
|| G(jt) ||dB
+20dB/decade
+90˚ f0 / 10
G(jt) 5.7˚
-45˚/decade
+45˚
f0
0˚
5.7˚
10 f0
Inverted zero
1 + ts
0
G(s) = s
t0
t0
G(s) = 1 + s
–20dB/decade
|| G(jt) ||dB
0.5f0
f0
1dB 2f0
3dB 1dB 0dB
10 f0 0˚
5.7˚
f0
–45˚
+45˚/decade
R3(t) dB
= R1(t) dB
+ R2(t) dB
40 dB G0 = 40 32 dB
|| G || G
|| G || f1 –20 dB/decade
20 dB
100 Hz
0 dB
0 dB
f2
0˚ 2 kHz –40 dB/decade
–20 dB 0˚
G f1/10 f2/10
–40 dB 10 Hz 200 Hz –45˚
–45˚/decade
–60 dB –90˚
–90˚/decade 10f2
10f1 20 kHz –135˚
1 kHz –45˚/decade
–180˚
1 Hz 10 Hz 100 Hz 1 kHz 10 kHz 100 kHz
f
Example 2
f2 || A' ||dB
|| A ||
f1
|| A0 ||dB +20 dB/dec
10f1 f2 /10
One solution: s
1+ t
1
A(s) = A 0
s
1+ t
2
➚ t1
1 + s s
t1 s = jt t f
A0 = A0 = A0 t = A0
1+➚ s
t2
1 1 f1
s = jt
Example 2, continued
For f > f2
➚ t1
1 + s s
t1 s = jt t2 f2
A0 = A0 = A0 t = A0
➚ t2
1 + s s
t2 s = jt
1 f1
s = jt
Example L
v2(s) 1
G(s) = = +
v1(s) 1 + s L + s 2LC
R
v1(s) + C R v2(s)
Second-order denominator, of –
the form
–
G(s) = 1
2
1 + a 1s + a 2s Two-pole low-pass filter example
G(s) = 1
1 + a 1s + a 2s 2
We might factor the denominator using the quadratic formula, then
construct Bode diagram as the combination of two real poles:
1 a1 4a 2
G(s) = with s1 = – 1– 1– 2
s
1– s s
1– s 2a 2 a1
1 2
a1 4a 2
s2 = – 1+ 1– 2
2a 2 a1
• If 4a2 ≤ a12, then the roots s1 and s2 are real. We can construct Bode
diagram as the combination of two real poles.
• If 4a2 > a12, then the roots are complex. In Section 8.1.1, the
assumption was made that ω0 is real; hence, the results of that
section cannot be applied and we need to do some additional work.
Approach 2: Define a standard normalized form
for the quadratic case
G(s) = 1 or G(s) = 1
1 + 2c ts + ts 1 + s + ts
2 2
0 0 Qt0 0
• When the coefficients of s are real and positive, then the parameters ζ,
ω0, and Q are also real and positive
• The parameters ζ, ω0, and Q are found by equating the coefficients of s
• The parameter ω0 is the angular corner frequency, and we can define f0
= ω0/2π
• The parameter ζ is called the damping factor. ζ controls the shape of the
exact curve in the vicinity of f = f0. The roots are complex when ζ < 1.
• In the alternative form, the parameter Q is called the quality factor. Q
also controls the shape of the exact curve in the vicinity of f = f0. The
roots are complex when Q > 0.5.
The Q-factor
In a second-order system, ζ and Q are related according to
Q= 1
2c
Result: t0 1
f0 = =
2/ 2/ LC
Q=R C
L
Magnitude asymptotes, quadratic form
Qt0 0
G( jt) = 1
t 2 2 t
+ 12 t
2
1– t
0 Q 0
–40 dB/decade
Two-pole response: exact curves
0°
Q=' Q='
Q=5 Q = 10
10dB Q =5
Q=2 Q=2
Q=1 Q=1
-45°
Q = 0.7
Q = 0.7
Q = 0.5
0dB Q = 0.2
Q = 0.1
Q = 0.5 G -90°
|| G ||dB
-10dB
Q = 0.2
-135°
Q = 0.1
-20dB
0.3 0.5 0.7 1 2 3
-180°
f / f0 0.1 1 10
f / f0
8.1.7. The low-Q approximation
G(s) = 1 or 1
1 + a 1s + a 2s 2 G(s) =
1 + s + ts
2
Qt0 0
When the roots are real, i.e., when Q < 0.5, then we can factor the
denominator, and construct the Bode diagram using the asymptotes
for real poles. We would then use the following normalized form:
G(s) = 1
1 + ts 1 + ts
1 2
This is a particularly desirable approach when Q << 0.5, i.e., when the
corner frequencies ω1 and ω2 are well separated.
An example
2
L/R± L / R – 4 LC
t1 , t2 =
2 LC
This complicated expression yields little insight into how the corner
frequencies ω1 and ω2 depend on R, L, and C.
When the corner frequencies are well separated in value, it can be
shown that they are given by the much simpler (approximate)
expressions
t1 5 R , t2 5 1
L RC
Given
G(s) = 1
1 + s + ts
2
Qt0 0
t0 1 + 1 – 4Q 2 1
t2 =
Q 2 F(Q)
0.75
can be written in the form
t0
t2 = F(Q) 0.5
Q
where 0.25
F(Q) = 1 1 + 1 – 4Q 2 0
2 0 0.1 0.2 0.3 0.4 0.5
For small Q, F(Q) tends to 1. Q
We then obtain
For Q < 0.3, the approximation F(Q)=1 is
t0
t2 5 for Q << 1 within 10% of the exact value.
Q 2
Corner frequency ω1
t0 1 – 1 – 4Q 2 1
t1 =
Q 2 F(Q)
0.75
can be written in the form
Q t0 0.5
t1 =
F(Q)
where 0.25
F(Q) = 1 1 + 1 – 4Q 2 0
2 0 0.1 0.2 0.3 0.4 0.5
For small Q, F(Q) tends to 1. Q
We then obtain
For Q < 0.3, the approximation F(Q)=1 is
t1 5 Q t0 for Q << 1 within 10% of the exact value.
2
The Low-Q Approximation
|| G ||dB Q f0
f1 =
F(Q)
f0 f0F(Q)
0dB 5 Q f0
f2 =
Q
f0
5
Q
–20dB/decade
–40dB/decade
R-L-C Example
v2(s) t0 1
G(s) = = 1 f0 = =
2/ 2/ LC
v1(s) 1 + s L + s 2LC
R Q=R C
L
t1 5 Q t0 = R C 1 =R
L LC L
t0
t2 5 = 1 1 = 1
Q LC R C RC
L
8.1.8. Approximate Roots of an
Arbitrary-Degree Polynomial
P(s) = 1 + a 1 s + a 2 s 2 + + an s n
When the roots are real and well separated in value, then approximate
analytical expressions for the time constants τ1, τ2, ... τn can be found,
that typically are simple functions of the circuit element values.
Objective: find a general method for deriving such expressions.
Include the case of complex root pairs.
Derivation of method
a n = o1o2o3 on
a1 = o1 + o2 + + on
System of equations: a2 = o1 o2 + + on + o2 o3 + + on +
(from previous slide) a 3 = o1o2 o3 + + on + o2o3 o4 + + on +
a n = o1o2o3 on
Suppose that roots are real and well-separated, and are arranged in
decreasing order of magnitude:
o 1 >> o 2 >> >> o n
an
on 5
an – 1
Result
when roots are real and well separated
a2 ak ak + 1 an
a1 > > > > >
a1 ak – 1 ak an – 1
a2 ak ak + 1 2 an
P(s) 5 1 + a 1 s 1+ s 1+ s + s 1+ s
a1 ak – 1 ak – 1 an – 1
The formulas of the previous slide require a special form for the case when
the first inequality is not satisfied:
a2 a3 an
a1 > > >
a1 a2 an – 1
a3 an
P(s) 5 1 + a 1s + a 2s2 1+ s 1+ s
a2 an – 1
2
a2 a3 a4 an
> a1 > > > >
a3 a2 a3 an – 1
Example
Damped input EMI filter
L1
ig ic
L2 R Converter
vg + C
–
L1 + L2
i g(s) 1+s
G(s) = = R
i c(s) L1 + L2 2 3 L 1 L 2C
1+s + s L 1C + s
R R
Example
Approximate factorization of a third-order denominator
L1 + L2
a1 =
R
a 2 = L 1C
L 1L 2C
a3 =
R
Real roots case
Factorization as three real roots:
L1 + L2 L1 L2
1+s 1 + sRC 1+s
R L1 + L2 R
This approximate analytical factorization is justified provided
L1 + L2 L1 L2
>> RC >>
R L1 + L2 R
Note that these inequalities cannot be satisfied unless L1 >> L2. The
above inequalities can then be further simplified to
L1 L2
>> RC >>
R R
And the factored polynomial reduces to • Illustrates in a simple
way how the roots
L1 L2
1+s 1 + sRC 1 + s depend on the
R R element values
When the second inequality is violated
L1 + L2 L1 L2
>> RC ✖
>>
R L1 + L2 R
B
not
satisfied
Then leave the second and third roots in quadratic form:
a2 a3 2
P(s) = 1 + a 1s 1 + a s + a s
1 1
which is
L1 + L2 L1
1+s 1 + sRC + s 2 L 1||L 2 C
R L1 + L2
Validity of the approximation
L1
1+s 1 + sRC + s 2L 2C
R
When the first inequality is violated
L1 + L2 L1 L2
✖
>> RC >>
R L1 + L2 R
B
not
satisfied
Then leave the first and second roots in quadratic form:
2 a3
P(s) = 1 + a 1s + a 2s 1+ a s
2
which is
L1 + L2 L2
1+s + s 2L 1C 1+s
R R
Validity of the approximation
L
1:D D' : 1
+
–
+
i(s)
(Vg – V) d (s)
vg (s) + I d(s) I d (s)
– C v(s) R
–
Definition of transfer functions
The converter contains two inputs, d(s) and vg(s) and one output, v(s)
Hence, the ac output voltage variations can be expressed as the
superposition of terms arising from the two inputs:
vg (s) +
– C v(s) R
–
Derivation of transfer functions
+
Use voltage divider formula L
to solve for transfer function: D' 2
vg(s) – D +
D' – C v(s) R
R || 1
v(s) sC
Gvg(s) = =– D –
vg(s) D' sL 1
d(s) = 0 + R ||
D' 2 sC
Gvg(s) = Gg0 1
1 + s + ts
2
Qt0 0
Salient features of the line-to-output transfer function
Gg0 = – D
D'
1 = LC
t 20 D' 2 t0 = D'
LC
1 = L C
2
Qt0 D' R Q = D'R
L
Derivation of
control-to-output transfer function Gvd(s)
L
D' : 1
+
–
In small-signal model, +
(Vg – V) d (s)
set vg source to zero:
I d (s) C v(s) R
+
L
Push all elements to Vg – V – D' 2
output side of d (s) + I d(s) C v(s) R
D'
transformer:
–
There are two d sources. One way to solve the model is to use superposition,
expressing the output v as a sum of terms arising from the two sources.
Superposition
Vg – V R || 1
sC
Total: Gvd(s) = – + I sL2 || R || 1
D' sL + R || 1 D' sC
D' 2 sC
Control-to-output transfer function
1–s LI
v(s) Vg – V Vg – V
Gvd(s) = = –
d(s) D' 2 L + s 2 LC
vg(s) = 0 1+s
D' 2 R D' 2
1 – ts
z
Gvd(s) = Gd0
1 + s + ts
2
Qt0 0
Salient features of control-to-output transfer function
Vg – V Vg V
Gd0 = – =– 2 =
D' D' DD'
Vg – V D' R
tz = = (RHP)
LI DL
t0 = D'
LC
Q = D'R C
L
80 dBV || Gvd ||
|| Gvd || Gvd
60 dBV Gd0 = 187 V
45.5 dBV Q = 4 12 dB
40 dBV f0 –40 dB/decade
400 Hz
20 dBV
10 -1/2Q f0
0˚ 300 Hz
0 dBV fz 0˚
Gvd fz /10 2.6 kHz
RHP –20 dB/decade
–20 dBV 260 Hz –90˚
f
Bode plot: line-to-output transfer function
20 dB
|| Gvg || Gg0 = 1.5
Q = 4 12 dB Gvg
3.5 dB
0 dB
|| Gvg || f0
400 Hz –40 dB/decade
–20 dB
–40 dB
10 –1/2Q 0 f0
0˚ 300 Hz
–60 dB 0˚
Gvg
–80 dB –90˚
–180˚
–180˚
10 1/2Q 0 f0
533 Hz
–270˚
10 Hz 100 Hz 1 kHz 10 kHz 100 kHz
f
8.2.2. Transfer functions of
some basic CCM converters
Table 8.2. S alient features of the small-signal CCM transfer functions of some basic dc-dc converters
1 – ts
z
Gvg(s) = Gg0 1
Gvd(s) = Gd0
1 + s + ts
2
1 + s + ts
2
Qt0 0
Qt0 0
8.2.3. Physical origins of the right half-plane zero
G(s) = 1 – ts
1
0
uin(s) + uout(s)
–
s
tz
• phase reversal at
high frequency
• transient response:
output initially tends
in wrong direction
Two converters whose CCM control-to-output
transfer functions exhibit RHP zeroes
iD Ts
= d' iL Ts
L 2 iD(t)
+
Boost iL(t)
1
vg + C R v
–
iD(t)
Buck-boost 1 2 +
iL(t)
vg + C R v
– L
–
Waveforms, step increase in duty cycle
iL(t)
iD Ts
= d' iL Ts
• Increasing d(t) t
causes the average iD(t)
diode current to iD(t)T
s
initially decrease
• As inductor current
increases to its new t
equilibrium value, | v(t) |
average diode
current eventually
increases
t
d = 0.4 d = 0.6
Impedance graph paper
60dB1 10H
1nF 1k1
40dB1 1H
10n 1001
F
m H
1 0 0
20dB1 100 101
nF
H
0dB1 10m 11
1µF
H
–20dB1 1m 100m1
10µ
F
0 µ H
–40dB1 10 10m1
100
100 µF
1F mF n H 10m 1m
1 0µ H 1µH 10 0 F 1 0nH F 1nH
–60dB1 1m1
10Hz 100Hz 1kHz 10kHz 100kHz 1MHz
Transfer functions predicted by canonical model
He(s)
e(s) d(s)
+ 1 : M(D)
–
+ +
Le
+ j(s) d(s)
Zin Zout
vg(s) ve(s) C v(s) R
–
– –
{
{
Z1 Z2
Output impedance Zout: set sources to zero
Zout
Le C R
{
Z1
{ Z2
Zout = Z1 || Z2
Graphical construction of output impedance
1 || Z1 || = tLe
tC
R
Q = R / R0
R0
f0
|| Zout ||
Graphical construction of
filter effective transfer function
tL e Q = R / R0
=1
tL e
f0 1 /tC 1
=
tL e t 2L eC
Z out
He =
Z1
2
Boost and buck-boost converters: Le = L / D’
1 tL
tC increasing
D 2
D'
R
Q = R / R0
R0
f0
|| Zout ||
8.4. Measurement of ac transfer functions
and impedances
Network Analyzer
Injection source Measured inputs Data
vz vz Data bus
magnitude frequency vy to computer
17.3 dB
vx
vz vx vy
output input input
+ – + – vy
– 134.7˚
+
–
vx
Swept sinusoidal measurements
Network Analyzer
Injection source Measured inputs Data • Potentiometer
vz vz Data bus
magnitude frequency vy
vx
–4.7 dB to computer establishes correct
vz vx vy quiescent operating
output input
+ –
input
+ – vy
– 162.8˚
point
+
–
vx
• Injection sinusoid
coupled to device
DC vy(s)
blocking = G(s) input via dc blocking
capacitor vx(s) capacitor
VCC • Actual device input
and output voltages
DC are measured as vx
bias
adjust and vy
output
input
G(s)
• Dynamics of blocking
capacitor are irrelevant
Device
under test
Measurement of an output impedance
v(s)
Z(s) =
i(s)
VCC Zs
Device
{
under test DC blocking
DC i out capacitor R
source
bias
adjust current
output
input
voltage
probe
vy(s)
Z out(s) =
i out(s) amplifier
ac input
=0 + – + –
vy vx
Measurement of output impedance
{
return to analyzer via
two paths. Injection
current which returns (1 – k) i out
Measured
via voltage probe ground inputs
voltage
induces voltage drop in probe +
vx
voltage probe, corrupting the voltage –
probe
measurement. Network return Zprobe +
analyzer measures connection
Z + (1 – k) Z probe = Z + Z probe || Z rz
For an accurate measurement, require
+
{ –
(1 – k) i out Z probe
– vy
Z >> Z probe || Z rz
Improved measurement: add isolation transformer
{
through
transformer. 0
No additional Measured
inputs
voltage is voltage
induced in probe +
vx
–
voltage probe voltage
probe
ground return Zprobe +
connection
{
connection – vy
+ –
0V
8.5. Summary of key points
8. Salient features of the transfer functions of the buck, boost, and buck-
boost converters are tabulated in section 8.2.2. The line-to-output
transfer functions of these converters contain two poles. Their control-
to-output transfer functions contain two poles, and may additionally
contain a right half-pland zero.
9. Approximate magnitude asymptotes of impedances and transfer
functions can be easily derived by graphical construction. This
approach is a useful supplement to conventional analysis, because it
yields physical insight into the circuit behavior, and because it
exposes suitable approximations. Several examples, including the
impedances of basic series and parallel resonant circuits and the
transfer function He(s) of the boost and buck-boost converters, are
worked in section 8.3.
10. Measurement of transfer functions and impedances using a network
analyzer is discussed in section 8.4. Careful attention to ground
connections is important when measuring small impedances.