a handy method to obtain satisfactory response of buck converter

advertisement
AN021
A Handy Method to Obtain Satisfactory Response of
Buck Converter
Introduction
Analysis:
The focus of this application note is to help users to
Fig.1 shows the unity feedback system for Buck
select a good set of components of the compensation
converter. We first
section of the Buck converter. Traditionally, engineers
identify transfer functions for each
of the corresponding block.
first concern about the power circuit for a circuit design.
After meeting the power requirement, what follows is
As to the modeling of the low-frequency behavior of
control loop design. The first priority of control design is
power switches in square-wave power converters,
the stability. And after that, the choice of components
please refer to appendix [1]. The circuit of Buck
can substantially effect the transient response. Here
converter is shown in Fig.2 and the model of its power
shows a simple method to obtain a satisfactory
switches is shown in Fig.3. Please note that the circuit
transient response with an acceptable steady state
in Fig.3 is linear. Fig.4 shows the circuit for small signal
error.
analysis.
+
dVOUT(s)=0
Gc(s)
_
Gp(s)
PWM
dVc(s)
dVOUT(s)
dD(s)
Fig.1 The Unity Feedback Control Loop for Buck Converter
R1
VOUT
+
VIN
R2
Fig. 2
Buck Converter
October, 2001
1
AN021
VOUT
+
R
VIN
D*
D* IL
Fig.3 R = DR1 + (1 − D)R 2
dVOUT(s)
L
rL
rc
R
RL
+
S
VIN *
Fig. 4
Small Signal Circuit
The transfer function of output with respect to duty ratio is:
1
dVOUT (s)
= VIN ×
dD(s)
1
+
RL
1
rC +
sL + (R + rL ) +
1
sc
1
1
+
RL
1
rC +
1
sc
… … ..(1)
= VIN ×
(CRL rC )s + R L
L(R L C + rC C)s + [L + C(RR L + rL R L + RrC + rCrL + R L rC )]s + (R L + R + rL )
2
where:
when RL is >> rL , rC and R
VIN: input voltage
Equation 1 can be simplified as below:
VOUT: output voltage
dVOUT (s)
= VIN ×
dD(s)
R: the equivalent resistance of power switches.
L: the inductor
C: the output capacitor
rL: the DC resistance of inductor.
rC: the ESR of output capacitor.
RL: the loading of Buck converter.
V ×r
= IN C ×
L
rC Cs + 1
L
LCs 2 + [
+ C(R + rL + rC )]s + 1
RL
1
rC C
R + rL + rC
1
1
s2 + [
+
]s +
CRL
L
LC
s+
… … … … … … … … … … … … ..(2)
October, 2001
2
AN021
where
the zero ZP is at
the ω n =
According to its frequency response in Fig. 7, the
1
rC C
network has high gain at low frequency, which helps to
reduce the steady state error. The attenuation of gain
1
at high frequency helps to weaken the noise
LC
disturbance.
C3
R2
dVout(s)
R1
C2
_
dVc(s)
+
Fig. 6
Fig. 5
The
Simulation
of
Typical
The Compensation Network
Frequency
REsponse of Buck Converter
Fig. 7
Fig. 5 represents the simulation of typical frequency
The
Simulation
of
Typical
Frequency
Response of the Compensation Circuit
response of Buck converter. Note that the effect of
complex conjugate poles of LC will make the gain
curve at -40dB/decade and phase curve towards -180°.
And after the region around ωn, we will meet the zero,
which is donated by the E.S.R of output capacitor. The
gain curve becomes -20dB/decade and the phase is
towards -90°.
Its transfer function is:
dVC (s)
1+ R 2C2s
×β … …
=
dVOUT (s) sR1 (C 2 + C 3 )[1 + sR 2 (C 2 //C 3 )
… … … … … … … … … (4)
Where:
the first pole P0 is at zero frequency (ω=0).
As to the PWM (Pulse Width Modulation), its transfer
function is
dD(s)
1
=
… … … … … … … … … … ..(3)
dVC (s) VM
where VM is the amplitude of ramp in PWM.
About the compensation network, we choose the type
that is shown in Fig. 6.
one zero Z1 is at
1
R 2C 2
the 2nd pole P2 is at
1
1
≈
( when C3<<C2 )
R 2 (C 2 //C 3 ) R 2C3
gain at low frequency: GC (s ) =
1
sR1(C 2 + C 3 )
gain at middle frequency:
October, 2001
3
AN021
GC (s ) =
the origin pole is already set. Then here comes the
R 2C 2
… ..(Constant)
R1 (C 2 + C3 )
strategy for the location of the zero and the second
gain at high frequency: GC (s ) =
pole.
1
sR1C 3
For the reason of stability, the zero of compensation
β is the gain of the feedback resistor divider.
Note that due to the origin pole, the phase at low
network must be set lower than the zero of Buck
frequency is -90° and the gain curve is at -20dB/
converter. Without compensation zero, the zero of
decade. When the frequency approaches to zero, the
Buck converter would face three poles before it. The
phase increases toward 0° with a flat gain curve. When
phase of the loop transfer function falls down toward
it moves toward the second pole, the frequency will be
-270° and passes through -180° between the poles of
towards -90° and the gain curve goes back to -20dB/
LC and the zero of Buck converter. This situation can
decade again.
lead to an unstable condition. To avoid that, the
compensation circuit zero is advised to be lower than
Control Strategy
that of Buck converter. However the location of the
The loop transfer function of Fig.1 is shown as below:
compensation zero will affect the phase margin around
T(s) = G C (s) × PWM × G P (s)
the crossover frequency. To improve the phase margin
1
R 2 C 2 (s +
)
R 2C2
=
sR 1 (C 2 + C 3 )R 2 (C 2 //C 3 )(s +
of the loop transfer function and keep high gain at low
frequency, the compensation zero is suggested
1
R s (C 2 //C 3 )
between ωn and 0.1 × ωn. (ωn=
1
rC (s +
)
rC C
1
×ß ×
× VIN ×
R + rC + rL
1
1
VM
]
)s +
L[s 2 + (
+
CR L
L
LC
1
).
LC
The second pole of compensation network is
suggested far from the zero of Buck converter. In this
way, we can obtain the –20dB/decade slope around
.… … … (5)
the crossover frequency. According to the sampling
After meeting the power circuit requirement, we start
to design the control loop. Therefore the LC poles and
output capacitor zero are already allocated. What we
can do is to distribute the location of pole and zero of
compensation
network
to
achieve
the
dVOUT(s)=0
Fig. 8
Gc(s)
_
suggested to be under one fifth of switching frequency
of Buck converter (
desired
response demand. To obtain high gain at low frequency,
+
theorem, the second pole of compensation network is
high frequency.
Gp(s)
PWM
dVc(s)
1
f Switching ) to reduce the noise at
5
dVOUT(s)
dD(s)
The Unity Feedback Control Loop for Buck Converter
October, 2001
4
AN021
Design Example
response are shown in Fig. 10 and Fig. 11, respectively.
Obviously, the rising time is too long and the overshoot
Design example:
VIN=8V,VOUT=5V, the operation point of output current
is 1A
is too large. To increase the speed of response and to
have more phase margin, we try to extend the
crossover frequency.
L=27µH, rL=38.5mΩ
C=1000µH, rC=52mΩ
MOSFET=IR3103, RDS-ON=14 mΩ
RL=5Ω
VM=1.3V
Design Procedure:
1. Construct the model of Buck converter and obtain
the transfer function.
dVOUT(s)
27µH
38.5m
52m
14m
+
S
VIN *
Fig. 9
Fig. 10
5
The Transient Response (Ch2 is Current
Curve, 1A/div)
1000µF
The Model of the Given Buck Converter
dVOUT (s)
1.93k (s + 19.2k )
= 8× 2
… … … … … ..(6)
dD(s)
s + 4.07ks + 37M
From equation (6), the location of zero and the
complex conjugate poles can be easily obtained.
2. Locate
the
zero
and
the
second
pole
of
compensation network.
Fig. 11
The Corresponding Simulation of
Frequency Response
Since ωn= 6.08k,
zero of Buck= 19.2k,
fSwitching= 200kHz= 1.25M(rad/s)
According to equation (4), we can add the extra gain
Let
which is provided by the compensation network at the
zero of compensation network=2.04k
Pole of compensation network=250k
middle frequency without changing the location of zero
and pole of the original loop transfer function.
3. Choose the components:
According to the equation (4),
Let R2=75k then C2=6800p, and C3=56p
Let R2=560k, then
R2C2=
1
2.04k
C2=875 ≈ 820p
The transient response and the simulation of frequency
October, 2001
5
AN021
R2C3=
1
… … … … … … . C3=7.14 ≈ 8p
250k
idea to verify the steady state error by the load
regulation. From Fig.12, it shows that the steady state
error of two settings is small, and the two curves are
almost identical.
5.1
Vout(V)
5.05
5
C2=820p R2=560k
C3=8p
C2=6800p R2=75k
C3=56p
4.95
4.9
0
Fig.12
2
4
6
Iout(A)
8
10
12
The Transient Response
(Ch2 is current curve, 1A/div)
Fig.14
The Comparison of Load Regulation of
Different Components (VIN=8V, VOUT=5V)
From the results of transient response and the steady
state error, the second set of components can achieve
a satisfactory response.
Summary
By taking the advantage of a high gain at low
frequency with a low gain at high frequency, the
best performance of the Buck Converter can be
Fig.13
The Corresponding Simulation of
Frequency Response
achieved. The objectives are to improve the steady
state error at low frequency and reduce noise
disturbance at high frequency.
Referring to Fig. 13, the crossover frequency of open
loop transfer function is raised. And we could also
expect the bandwidth of the closed loop transfer
function being increased. Indeed, in Fig 12, the set of
control components, which we just used, shows high
speed of response and less overshoot.
The key points of the compensation are:
1) Locate the zero around the LC resonant
frequency for the issue of stability.
2) Locate the pole around
1
f Switching for the noise
5
reduction at high frequency.
3) The combination of the above 2 procedures can
Although the pole at origin (according to equation (5))
make the crossover frequency at the -20dB/
means the infinitive gain at DC (ω=0), it is still a good
decade situation. Thus a satisfactory phase
October, 2001
6
AN021
margin is achieved.
Reference:
[1] Yim-Shu Lee, Computer-Aided Analysis and Design
of Switch-Mode Power Supplies., Marcel Dekker, Inc.
Hong Kong,1993
October, 2001
7
Download