Professional Documents
Culture Documents
converter compensation
made easy
Robert Sheehan
Systems Manager,
Power Design Services
Member Group Technical Staff
Louis Diana
Field Application,
America Sales and Marketing
Senior Member Technical Staff
Texas Instruments
Using this paper as a quick look-up reference can help
designers compensate the most popular switch-mode
power converter topologies.
Engineers have been designing switch-mode power converters for some time now.
If you’re new to the design field or you don’t compensate converters all the time,
compensation requires some research to do correctly. This paper will break the procedure
down into a step-by-step process that you can follow to compensate a power converter.
We will explain the theory of compensation and why it is necessary, examine various power
stages, and show how to determine where to place the poles and zeros of the compensation
network to compensate a power converter. We will examine typical error amplifiers as well
as transconductance amplifiers to see how each affects the control loop and work through a
number of topologies/examples so that power engineers have a quick reference when they
need to compensate a power converter.
Figure 2 shows the results of low phase margin. In passes through zero. Gain margin is the gain value
this case there is ringing after the load transient and when the phase passes through 0 degrees. This
is measured in decibels and should be a negative
the loop is underdamped. This is not a desirable
number. Phase margin is the value of the phase
response.
measured in degrees when the gain passes through
In Figure 3, the phase margin has increased;
zero. This is measured in degrees and should be a
therefore, the waveform does not show any ringing
positive number (Figure 4).
after the load transient. This response is well
damped.
Magnitude (dB)
40 90
Magnitude (dB)
Phase (°)
Phase (°)
20 Phase 45 -20 -45
margin
0 0
Gain margin -40 -90
-20 -45
Figure 4. Phase margin is the difference in degrees when the gain Figure 5. The Bode plot of a pole shows the gain decreasing by –20
crosses zero. Gain margin is the difference in decibels when the phase dB/decade, with a phase shift at higher frequencies of –90 degrees.
crosses zero. Equation 2 defines a zero where s is in the
numerator. At the frequency of the zero the gain is
Sufficient phase margin is required to prevent
3 dB up and increasing at a +20 dB/decade slope.
oscillations. A phase margin of 45 degrees or
The phase starts to increase one decade before
greater is the design goal. A gain margin of –6 dB is
the zero frequency, is 45 degrees up at the zero
the minimum, while –10 dB is considered good.
frequency, and continues to increase another 45
Although higher crossovers are generally preferable, Figure 6 degrees for one more decade. The total change is
there are practical limitations. The rule of thumb 90 degrees over two decades (Figure 6).
is one-fifth to one-tenth the switching frequency.
s
Attenuation at the switching frequency is also 1+
ωZ (2)
important for noise immunity to minimize jitter. The H ( s) =
1
gain should ideally pass through zero with a slope
of –20 dB/decade. This will maximize gain margin
and will also negate the chance of the gain turning 60 135
positive at a higher frequency where the phase may
40 90
Magnitude (dB)
Phase (°)
20 -45 frequency. This means that there are two poles at
the same frequency, and the gain changes –40 dB/
0 -90 decade with an associated phase change of –180
degrees. A current-mode buck converter does not
-20 -135
0.01 0.1 1 10 100 1000 have a complex pole at the LC resonant frequency.
Frequency (kHz) Equation 5 shows the transfer function of a complex
conjugate pole with a quality factor, Q, associated
Figure 7. The Bode plot of an inverted zero shows the gain going up
and to the left of the reference gain, shown here as 0 dB. This cancels with the LC filter. Q is a measure of the sensitivity
a pole at some lower frequency so that the phase changes from –90 or quality of the tuned circuit. A higher Q value
degrees to 0 degrees.
corresponds to a narrower bandwidth of the tuned
circuit. A high Q value is not so good for a power-
A right-half-plane zero is characteristic of boost and
supply output filter, however, because as Q increases
buck-boost power stages. The magnitude increases
the phase slope increases. This means that the
at 20 dB/decade with an associated phase lag
phase changes much more quickly over a small band
of –90 degrees. As you can see in Equation 4, s
of frequencies, whereas two regular poles would
re 8 is in the numerator, but it is negative. This makes
change with a more gradual slope over two decades.
compensating the converter more difficult (Figure 8).
1
s H ( s) =
1− s s2 (5)
ωZ 1+ + 2
( s) =
H
(4) QO ⋅ ωO ωO
1
Figure 9 Figure 9 illustrates how phase slope increases as
Q increases. Since a high Q LC filter can cause a
60 45
–180 degree phase shift in the loop Bode plot, it
40 0 is important to understand the Q of the LC filter in
Magnitude (dB)
20 -45
20 45
0 -90
Gain
0 0
Magnitude (dB)
Figure 8. The Bode plot of a right-half-plane zero shows the gain -60 -135
increasing by +20 dB/decade, with a phase shift at higher frequencies Phase
of –90 degrees. -80 -180
0.01 0.1 1 10 100 1000
Frequency (kHz)
See Equation 9.
Phase (°)
-20 -45
-40 -90 N
V OUT = VIN ⋅ D ⋅ S
(8)
-60 -135 NP
-80 -180
NS (9)
0.01 0.1 1 10 100 1000 VOUT = VIN ⋅ 2 ⋅ D ⋅
NP
Frequency (kHz)
Figure 10. An ESR zero after the complex conjugate pole causes the
Boost topologies
slope of the gain to reduce to –20 dB/decade, while the phase shift Boost topologies deliver energy to the output 180
moves toward –90 degrees.
degrees out of phase with the energy delivered to
Buck-derived topologies the input. This causes a right-half-plane zero to
appear in the transfer function. Figure 12 shows a
Buck-derived topologies deliver power to the output
boost power stage, along with its associated input/
after reception at the input. Buck-derived topologies
output voltage transfer function. In Equation 10, Dˊ
running in CCM with voltage-mode feedback have
is the off-time duty cycle given by Dˊ = 1 – D.
an LC filter response with one complex conjugate
pole and one ESR zero. Figure 11 shows a buck
gure 12
1
V OUT = VIN ⋅
(10)
D′
IN OUT
D NS
V OUT = VIN ⋅ ⋅
(12)
D′ N P
Phase (°)
0 -45
the ESR zero of the output capacitor in Equation 18, 2
-20 -90
reducing the slope to –20 dB/decade.
-40 -135
dv d ⎛ v ⎞ V -60 -180
A VC = OUT = ⋅ ⎜⎜ v IN ⋅ C ⎟⎟ = IN EQua16
(16)
dvC dvC ⎝ v RAMP ⎠ VRAMP 0.01 0.1 1 10 100 1000
Frequency (kHz)
1
ω O =
(17) Figure 15. The voltage-mode buck frequency response has the
L ⋅ COUT characteristic complex conjugate pole with ESR zero.
1
ω Z =
(18) Current-mode buck
RESR ⋅ COUT
The difference between voltage- and current-mode
The Q associated with the complex conjugate control is that for current-mode control, the PWM
pole determines the slope of the phase. See modulator uses the inductor ripple current as the
Equation 19. As we discussed in the section on ramp. Voltage-mode control uses a fixed ramp,
complex conjugate poles, Q complicates converter which contains no information about the inductor
compensation because as Q increases, the phase current. Current-mode control exhibits some
slope increases. This means that the phase changes desirable attributes because it samples the inductor
much quicker over a small band of frequencies, current. The inner current loop splits the complex
whereas two regular poles would change with a conjugate pole of the filter into two real poles, turning
K m ⋅R i
ωL =
L (25)
VIN
where K m ≈ at D = 0.5.
VSLOPE
Figure 16. The current-mode buck power stage incorporates the The transfer function also contains an ESR zero
inductor current sense as an inner control loop. The current loop
acts as a lossless damping resistor, splitting the complex conjugate associated with the output capacitor, expressed as
pole of the output filter into two real poles. It turns the modulator into Equation
26:
a voltage-controlled current source, where the inductor current is
proportional to the control voltage at VC. 1
ω
Z = R ⋅C (26)
ESR OUT
Phase (°)
Equation 31 expresses the gain as: Figure 19 shows the Bode plot of the current-mode
boost power stage.
ROUT ⋅ D′
A VC ≈ (31)
2 ⋅ Ri
40 45
P
AVC L
where Ri = A ⋅ RS and D′ = 1 − D. 20 2 0
Magnitude (dB)
2
Phase (°)
Equation 32 calculates the first pole, which is related 0 -45
(V − VIN )⋅Ri ⋅T
V SLOPE = OUT (36)
L
⎛
Figure 21 where Ri is the current-sense gain times the sense
s ⎞ ⎛ s ⎞
⎜⎜1 − ⎟⎟ ⋅ ⎜⎜1 + ⎟⎟ resistor, and T is equal to the switching period,
vˆOUT ω ω
≈ AVC ⋅ ⎝ R ⎠ ⎝ Z ⎠
(39) 1/fSW.
vˆC ⎛ s ⎞ ⎛ s ⎞
⎜⎜1 + ⎟⎟ ⋅ ⎜⎜1 + ⎟⎟
⎝ ωP ⎠ ⎝ ωL ⎠ Figure 21 shows the Bode plot of the current-mode
buck-boost power stage.
Equation 40 expresses the gain as:
ROUT ⋅ D′ 40 45
AVC ≈ P
(1 + D )⋅ Ri (40) AVC L
20 2 0
Magnitude (dB)
where Ri = A ⋅ RS and D′ = 1 − D . 2
Phase (°)
0 -45
AVC ≈ OUT ⋅ P
Ri N S (49)
Phase (°)
0 -45
This tends to limit the overall loop bandwidth due Equation 58 calculates the first pole, which is related
ure 24 to the associated phase lag of the right-half-plane
to the output capacitance and the load resistance:
zero.
Figure 24 shows the schematic of the current- 1+ D
(58)
ωP ≈
COUT ⋅ ROUT
mode flyback power stage.
T1 Equation 59 calculates the second pole, which is
NP : NS D1
VIN VOUT related to the inductance. The modulator voltage
CIN LP COUT
ROUT gain, Km, is equal to (VIN+VOUT·NP/NS)/VSLOPE at D
RESR
PWM = 0.5 and will have little variation with operating
Logic Q1
VSLOPE conditions when properly scaling VSLOPE.
A
+
RS
- K m ⋅R i
+ VC ωL =
LP
(59)
Figure 24. For the current-mode flyback power stage, the isolation
transformer also serves as the energy-storage inductor. NP
VIN + VOUT ⋅
NS
where K m ≈ at D = 0.5.
VSLOPE
Equations 54 and 55 give the duty-cycle relationship
for the current-mode flyback in CCM: Equation 60 gives the right-half-plane zero,
which is related to the inductance and output
VOUT
D=
resistance, while Equation 61 is related to the output
N
V ⋅ S +V (54)
IN OUT
NP capacitance ESR:
2
VIN R ⋅ D′2 ⎛ N P ⎞
′=
D
(55) ω R = OUT ⋅⎜ ⎟
(60)
NP LP ⋅ D ⎜⎝ N S ⎟⎠
VIN + VOUT ⋅
NS
1
Equation 56 shows the transfer function for the ω Z =
(61)
RESR ⋅ COUT
current-mode flyback power stage. As you can see,
it contains two poles, one zero, one right-half-plane
For the peak current-mode flyback, Equation 62
zero and an associated gain. calculates the optimal value of slope compensation:
⎛ s ⎞ ⎛ s ⎞
⎜⎜1 − ⎟⎟ ⋅ ⎜⎜1 + ⎟ VOUT ⋅R i ⋅T N P
vˆ ω R ⎠ ⎝ ω Z ⎟⎠ V
SLOPE = ⋅
(62)
OUT ≈ AVC ⋅ ⎝
(56) LP NS
vˆC ⎛ s ⎞ ⎛ s ⎞
⎜⎜1 + ⎟⎟ ⋅ ⎜⎜1 + ⎟⎟
⎝ ωP ⎠ ⎝ ωL ⎠
2 − vˆC
Phase (°)
0 -45 v ˆFB =
(64)
AOL
-20 R Z -90
2 2
-40 -135 Equation 65 combines Equations 63 and 64:
-60 -180 ⎛ ⎞
0.01 0.1 1 10 100 1000 ⎜ ⎟
Frequency (kHz) vˆC Z COMP ⎜ 1 ⎟
vˆ′ = − R ⋅⎜
⎛ ⎞⎟
(65)
OUT FBT ⎜ 1 + 1 ⋅ ⎜1 + Z COMP ⎟ ⎟
Figure 25. The current-mode flyback power-stage frequency ⎜
⎝ AOL ⎜⎝ RFBT ⎟⎠ ⎟⎠
response is similar to that of the buck-boost.
Equation 66 shows that if the gain of the error
Type I error amplifier
amplifier is large enough
Figure 26 shows a Type I error amplifier
configuration, which is the simplest form of ⎛ RFBT ⎞
A OL ⋅ ⎜⎜ ⎟⎟ >> 1
(66)
R
⎝ FBT + Z COMP ⎠
compensation; it is characterized by a single pole.
You can analyze this circuit by recognizing that the
Then the closed-loop gain can be expressed by
error amplifier is inverting and has a virtual short
Equation 67 as:
between VFB and VREF. The feedback impedance
gure 26 divided by the input impedance gives you the small vˆ Z
C ≈ − COMP = −
1 ω
= − EA
(67)
signal gain. Since VREF can be viewed as an AC ′
vˆOUT RFBT s ⋅ CCOMP ⋅ RFBT s
ground, you can ignore the value of RFBB, as it does
not affect the AC transfer function. Equation 68 defines the error amplifier pole
frequency:
1
ω
EA =
(68)
RFBT ⋅ CCOMP
Examining the result reveals a single pole at the
origin. In practice, this is limited at DC by the
open-loop gain of the amplifier and is called
Figure 26. Type I error amplifier compensation has a single capacitor dominant-pole compensation.
used for a constant-current type of current-mode where AVM is defined as the mid-band voltage gain.
buck, driving a light-emitting diode (LED) load with Examining the result reveals that in Equation 70 the
RCOMP
supply, for many systems this type of A
VM ≈ (70)
RFBT
compensation does not offer the flexibility
necessary to achieve optimal performance. Equation 71 reveals a zero at:
1
60 135 ω ZEA = (71)
RCOMP ⋅ CCOMP
40 90
Magnitude (dB)
0 0
1
EA ω HF ≈
-20 -45 RCOMP ⋅ C HF
2 (72)
-40 -90
0.01 0.1 1 10 100 1000
assuming CCOMP >> CHF .
Frequency (kHz)
In practice, the open-loop gain of the amplifier
Figure 27. A Type I error-amplifier frequency response is will limit the error-amplifier gain at DC. Type II
characterized by a single pole.
gure 28 compensation is generally well suited for use with
Type II error amplifier current-mode control. In selective cases, you can
Figure 28 shows the schematic of a Type II error use it for voltage-mode control with a high value of
amplifier.
Figure 29 ESR in the output capacitor.
80 180
60 135
Magnitude (dB)
Phase (°)
40 90
Figure 28. Type II error-amplifier compensation adds a resistor and
high-frequency capacitor to the feedback. 20 AVM 45
ZEA HF
0 2 2 0
amplifiers, the bottom resistor of the input divider amplifier has a pole at the origin, a zero and a
dropped out of the transfer function due to the second high-frequency pole. Equations 75 and 76
virtual ground effect. Both pins were at the same are identical to Equations 71 and 72:
A OL = g m ⋅ REA (77)
ω s
1 + ZEA 1+
vˆC A ⋅ ω
s ≈ − VM ZEA ⋅ ω
≈ − AVM ⋅ ZEA (73)
′
vˆOUT s s s
1+ 1+
ωHF ωHF
RCOMP
A
VM ≈ (79)
Phase (°)
40 90 RFBT
20 AVM 45
ZEA HF Examining the Type III transfer function reveals
0 2 2 0
two zeros: one zero set by RCOMP and CCOMP (Equation
-20 -45
0.01 0.1 1 10 100 1000
80) and another zero set by RFBT and CFF (see
Frequency (kHz) Equation 81) to help offset the complex conjugate
poles:
Figure 31. A transconductance error amplifier with Type II
compensation has a frequency response equivalent to a standard 1
operational amplifier. ω ZEA = (80)
RCOMP ⋅ CCOMP
1
ω FP = (82)
RFF ⋅ C FF
1
ω
HF ≈ (83)
Figure 32. Type III error-amplifier compensation adds a lead network RCOMP ⋅ C HF
across the top divider resistor.
This compensation network provides a pole at assuming CCOMP >> C HF and RFBT >> RFF .
the origin, two zeros and two higher-frequency Type III compensation is useful in power supplies
poles in the feedback path. The two zeros offset where the output capacitor ESR is very low, such as
the complex conjugate pole of the voltage-mode in converters with ceramic output capacitors.
buck. Type III compensation can increase both
The reason for this is that low-ESR capacitors push
the bandwidth and phase margin of a closed-loop
the ESR zero higher in frequency than high-ESR
system.
capacitors. Thus, your converter will not benefit
Equation 78 expresses the voltage-gain transfer from the phase boost at lower frequencies, but Type
function for Type III compensation as: III compensation can make up for that.
⎛ ωZEA ⎞ ⎛ s ⎞ ⎛ s ⎞ ⎛ s ⎞
⎜1 + ⎟ ⋅ ⎜1 + ⎟ ⎜⎜1 + ⎟ ⋅ ⎜1 + ⎟
vˆC ⎝ s ⎠ ⎜⎝ ω FZ ⎟⎠ A ⋅ω ⎝ ωZEA ⎟⎠ ⎜⎝ ω FZ ⎟⎠
= − AVM ⋅ = − VM ZEA ⋅
′
vˆOUT ⎛ s ⎞ ⎛ s ⎞ s ⎛ s ⎞ ⎛ s ⎞
⎜⎜1 + ⎟⎟ ⋅ ⎜⎜1 + ⎟⎟ ⎜⎜1 + ⎟⎟ ⋅ ⎜⎜1 + ⎟⎟
⎝ ω FP ⎠ ⎝ ω HF ⎠ ⎝ ω FP ⎠ ⎝ ω HF ⎠
(78)
80 270 ωZEA s
1+ 1+
60 225 vˆC s AVM ⋅ ωZEA ωZEA
≈ − AVM ⋅ ≈− ⋅
vˆ OUT
′ s s s (84)
Magnitude (dB)
40 180 1+ 1+
ωHF ωHF
Phase (°)
20 135
AVM FP HF
0 90 In Equation 85 the mid-band voltage gain is:
ZEA FZ 2 2
-20 45
2 2 RP
-40 0 VM = CTR ⋅
A
(85)
0.01 0.1 1 10 100 1000 RD
Frequency (kHz)
Equation
86 defines the current transfer ratio:
Figure 33. A Type III compensator frequency response boosts the gain
and phase in the mid band. IC
CTR =
(86)
IF
Isolated feedback with optocoupler
Figure 34 shows Type II compensation using an As you can see from Equation 84, this isolated
optocoupler and the TL431 shunt regulator. The feedback with an optocoupler is configured as a
current-transfer ratio and resistors set the mid-band Type II error amplifier, which has a pole at the origin,
gain through the optocoupler. Bias currents and the a zero and a second high-frequency pole. See
diode forward voltage can limit the dynamic range. Equations
87 and 88:
The reference voltage for a standard TL431 is 2.5
1
V, which can work for 5-V outputs and higher. For ω ZEA =
(87)
RFBT ⋅ CCOMP
lower output voltages, the TLV431 has a 1.24-V
Figure 35
reference. 1
ω
HF =
(88)
In Figure 34, CP includes the parasitic capacitance RP ⋅ C P
of the optocoupler’s output transistor. Parasitic
capacitance is often the limiting factor of the Figure 35 shows the frequency-response plot.
Figure 34 bandwidth in this configuration. The associated
phase shift can go to –180 degrees, making 80 180
compensation at higher frequencies challenging.
60 135
Magnitude (dB)
Phase (°)
40 90
20 AVM 45
ZEA HF
0 0
2 2
-20 -45
0.01 0.1 1 10 100 1000
Frequency (kHz)
2 kΩ and 200 kΩ. AVM, the mid-band voltage gain, R COMP = AVM ⋅ RFBT
(90)
is one of the design parameters; by changing this
20 2 0
product of the power-stage and error-amplifier
Phase (°)
0 -45
Z transfer functions. On the Bode plots, the overall
-20 -90
2 loop gain is the sum in decibels of the power-stage
-40 -135
gain and error-amplifier gain. Adjust the mid-band
-60 -180
0.01 0.1 1 10 100 1000 gain of the error amplifier to meet the target
Frequency (kHz) loop bandwidth.
Power stage (a)
Current-mode buck example
ZEA s Figure 38 is the complete current-mode buck
1 1 Figure 38
vˆC s FZ regulator model, showing the modulator, output filter
AVM
ˆvOUT s s and error amplifier. For Type II compensation, this
1 1
FP HF example uses a transconductance amplifier.
60 270
40 HF 225
Magnitude (dB)
2
Phase (°)
20 180
FP
0 135
2
AVM
-20 ZEA FZ 90
&
2 2
-40 45
0.01 0.1 1 10 100 1000
Frequency (kHz)
20 45
Magnitude (dB)
compensation – you may need to lower the loop
Phase (°)
Z
bandwidth to less than fSW/20. A Type III feedback 0
2
0
Figure 39 (b)
current-mode buck: Power stage (a)
• Choose a value for RFBT based on the bias current and
1 ZEA
power dissipation. vˆC s
AVM
vˆOUT s
1
• Find the modulator transconductance in A/V. HF
Phase (°)
• Set ωZEA equal to one-tenth the target crossover frequency: 0 AVM 135
ZEA HF
ωZEA = ωC/10. 2 2
-20 90
• Set ωHF equal to the ESR zero frequency: ωHF = ωZ.
-40 45
0.01 0.1 1 10 100 1000
Solve Equations 94 through 99 to find the
component values: Figure 39 (c) Frequency (kHz)
40 90
Magnitude (dB)
20 45
0 0
AVM C
RCOMP = (gm amp)
(97) -20 2 -45
g m ⋅ K FB
-40 -90
0.01 0.1 1 10 100 1000
1
CCOMP =
(98) Frequency (kHz)
ωZEA ⋅ RCOMP
Control loop (c)
1
C
HF =
(99) Figure 39. The power-stage (a) and error-amplifier (b) plots sum to
ωHF ⋅ RCOMP produce the control-loop plot (c) for the current-mode buck frequency
response.
2
Phase (°)
0 -45
-20 R Z -90
2 2
-40 -135
-60 -180
0.01 0.1 1 10 100 1000
Frequency (kHz)
Power stage (a)
1 ZEA
vˆC s
AVM
vˆOUT s
1
HF Figure 42. The error amplifier is referenced to -VOUT for this
current-mode buck-boost.
40 270
0
2 180 compensation strategy
Phase (°)
ZEA
-20 2 135 We will now outline the compensation guidelines
-40 90 for the current-mode buck-boost with Type II
-60 45 compensation. At frequencies after the output-filter
0.01 0.1 1 10 100 1000
pole, the modulator transconductance and output-
Frequency (kHz)
filter capacitor set the power-stage gain.
Error amplifier (b)
The overall loop bandwidth is typically limited to
one-fourth the right-half-plane zero frequency.
vˆOUT vˆOUT vˆC Adjust AVM to achieve the target loop bandwidth.
vˆOUT
vˆC vˆOUT To give the maximum amount of phase boost, place
the error-amplifier zero, ωZEA, a decade below the
60 135
target crossover frequency. The high-frequency
40 90
pole, ωHF, should cancel the lower of the right-half-
Magnitude (dB)
Phase (°)
20 45
plane or ESR zero frequency.
0 0
C Some sets of operating conditions and parameter
-20 2 -45
values may require additional phase margin.
-40 -90
0.01 0.1 1 10 100 1000 In such cases, the crossover frequency may be
Frequency (kHz)
limited to one-fifth or one-sixth the right-half-plane
Control loop (c) zero frequency.
Figure 41. The power-stage (a) and error-amplifier (b) plots sum to
produce the control-loop plot (c) for the current-mode boost frequency
response.
power dissipation.
40 45
• Find the modulator transconductance in A/V. P
AVC L
20 2 0
Magnitude (dB)
• Find the right-half-plane zero frequency at the minimum 2
Phase (°)
0 -45
input voltage and maximum load current.
-20 R Z -90
• Set the target bandwidth to one-fourth the right-half-plane 2 2
-40 -135
zero frequency: ωC = 2∙π∙fC = ωR/4.
-60 -180
• Find AVM to achieve the target bandwidth. 0.01 0.1 1 10 100 1000
Frequency (kHz)
Figure 43 (b)
• Set ωZEA equal to one-tenth the target crossover
Power stage (a)
frequency: ωZEA = ωC/10.
• Set ωHF equal to the lower of the right-half-plane or ESR 1 ZEA
vˆC s
AVM
zero frequency: ωHF = ωR or ωZ.
vˆOUT 1
s
HF
Solve Equations 107 through 113 to find the
40 270
component values:
20 HF 225
D′ (107) AVM
Magnitude (dB)
Gm (mod) =
0
2 180
Ri
Phase (°)
ZEA
-20 2 135
R ⋅ Dʹ′2
ω R = OUT
(108) -40 90
L⋅D
-60 45
0.01 0.1 1 10 100 1000
ω ⋅C
Figure 43 (c)
AVM = C OUT
Gm (mod)
(109) Frequency (kHz)
40 90
1
C COMP = (112)
Magnitude (dB)
Phase (°)
ωZEA ⋅ RCOMP 20 45
0 0
C
1 (113)
CHF = -20 2 -45
ωHF ⋅ RCOMP
-40 -90
0.01 0.1 1 10 100 1000
Current-mode buck-boost Frequency (kHz)
Ri
RFBT
+
RP Use the following simplified design method for the
VC CCOMP
- current-mode forward:
+ RE
Peak current-mode forward VREF
TL431 RFBB
• Choose a value for RFBT based on the bias current and
Optimal VSLOPE = VOUT·Ri·T/L·NS/NP
power dissipation.
Figure 44. This isolated current-mode forward converter uses the • Find the modulator transconductance in A/V.
primary-side error amplifier as part of the feedback.
• Pick the target bandwidth; typically, fSW/10: ωC = 2∙π∙fC.
Current-mode forward compensation • Find AVM to achieve the target bandwidth.
strategy • Adjust RD, RP and COUT as required.
We will now outline the compensation guidelines for • Set ωZEA equal to one-tenth the target crossover
a current-mode forward with Type II compensation. frequency: ωZEA =ωC/10.
At frequencies after the output-filter pole, the • Set ωHF equal to the ESR zero frequency: ωHF = ωZ.
modulator transconductance and output-filter
Solve Equations 114 through 118 to find the
capacitor set the power-stage gain. Adjust AVM to
component values:
achieve the target loop bandwidth, typically one-
tenth the switching frequency. The selection of RD 1 NP
G
m (mod) = ⋅ (114)
Ri N S
and RP for DC bias considerations may limit AVM.
In such cases, you may need to increase COUT
ωC ⋅ COUT
A
VM = (115)
to meet the desired crossover frequency. To give Gm (mod)
the maximum amount of phase boost, place the
error-amplifier zero, ωZEA, a decade below the target R
RD = CTR ⋅ P
(116)
crossover frequency. The high-frequency pole,ωHF, A VM
Phase (°)
Z
0
2
0
example
-20
L
-45 Figure 46 Figure 46 is the complete current-mode flyback
2 regulator model showing the modulator, output
-40 -90
0.01 0.1 1 10 100 1000 filter and error amplifier. For Type II compensation,
Frequency (kHz)
this example uses a TL431 shunt regulator and
Power stage (a) optocoupler.
T1
ZEA VIN
NP : NS D1
VOUT
1
vˆC s CIN LP COUT
AVM ROUT
vˆOUT s RESR
1 PWM
HF Logic Q1
VSLOPE
A VCC
40 225 +
- Ri = A·RS RS
+
+ RP RD
20 180 VC
Magnitude (dB)
Phase (°)
CP RBIAS
0 AVM 135
ZEA HF RFBT
CCOMP
-20
2 2 90 Peak current-mode flyback TL431
Optimal VSLOPE = VOUT·Ri·T/LP·NP/NS RFBB
-40 45
0.01 0.1 1 10 100 1000 Figure 46. The optocoupler feedback for this isolated current-mode
Frequency (kHz) flyback is implemented with a minimum of components.
Phase (°)
20 45
capacitor set the power-stage gain. The overall loop
0 0 bandwidth is typically limited to one-fourth the right-
C
-20 2 -45 half-plane zero frequency. Adjust AVM to achieve the
-40 -90 target loop bandwidth. The selection of RD and RP
0.01 0.1 1 10 100 1000
Frequency (kHz)
for DC bias considerations may limit AVM. In such
cases, you may need to increase COUT to meet the
Control loop (c) desired crossover frequency. To give the maximum
amount of phase boost, place the error-amplifier
Figure 45. The power-stage (a) and error-amplifier (b) plots sum zero, ωZEA, a decade below the target crossover
to produce the control-loop plot (c) for the current-mode forward
frequency response. frequency. The high-frequency pole, ωHF, should
cancel the lower of the right-half-plane or ESR
zero frequency.
Magnitude (dB)
2
current-mode flyback:
Phase (°)
0 -45
• Choose a value for RFBT based on the bias current and -20 R Z -90
power dissipation. 2 2
-40 -135
• Find the modulator transconductance in A/V.
-60 -180
0.01 0.1 1 10 100 1000
• Find the right-half-plane zero frequency at the minimum
Frequency (kHz)
input voltage and maximum load current.
Figure 47 (b)
• Set the target bandwidth to one-fourth the right-half-plane Power stage (a)
zero frequency: ωC = 2∙π∙fC = ωR/4.
1 ZEA
• Find AVM to achieve the target bandwidth. vˆC s
AVM
vˆOUT s
• Adjust RD, RP and COUT as required. 1
HF
• Set ωZEA equal to one-tenth the target crossover
40 270
frequency: ωZEA = ωC/10.
20 HF 225
• Set ωHF equal to the lower of the right-half-plane or ESR
Magnitude (dB)
AVM 2
0 180
zero frequency: ωHF = ωR or ωZ.
Phase (°)
ZEA
-20 2 135
Solve Equations 119 through 124 to find the
-40 90
component values:
-60 45
(mod) = D′ ⋅ N P
G (119) 0.01 0.1 1 10 100 1000
m
Ri N S Figure 47 (c) Frequency (kHz)
A VM
Phase (°)
20 45
1 0 0
C COMP = (123)
RFBT ⋅ ωZEA -20
C
-45
2
1 -40 -90
C P = (124) 0.01 0.1 1 10 100 1000
RP ⋅ ω HF
Frequency (kHz)
ΔI
Phase (°)
20 45 VP =
8 ⋅ f C ⋅C OUT
49 0
-20 fC
0
-45
VP =
5A
= 140mV
(128)
8 ⋅10kHz ⋅ 440µF
-40 -90
0.01 0.1 1 10 100 1000
Frequency (kHz)
Practical limitations of error
Figure 48. The current-mode bandwidth is set at 10 kHz for this amplifiers
example.
The error-amplifier bandwidth can limit the maximum
loop-crossover frequency. As you can see from
Figure 50, where the closed-loop gain intersects the
∆I open-loop gain plot, the phase drops quickly. You will
need a wider-bandwidth op amp for voltage mode
due to Type III compensation.
vP
tP
Phase (°)
40 180
damping. Higher crossover frequency results in higher
20
0 90
Figure 52
135
noise sensitivity and possible switching jitter. A general
rule of thumb is one-fifth to one-tenth the switching
-20 45
0.01 0.1 1 10 100 1000 frequency. The right-half-plane zero usually limits boost
Frequency (kHz)
and buck-boost converters operating in CCM.
Figure 50: The error amplifier’s open-loop gain and bandwidth set a
hard limit for the closed-loop response.
Voltage-mode control loop
80 180
Practical limitations of optocouplers 60 135
Magnitude (dB)
ure 51 no external capacitor the roll-off approaches –40 dB/ Figure 52. The voltage-mode control loop shows the phase going
to zero near the switching frequency. The available gain and phase
decade, with phase shift at higher frequencies of margin set an upper limit on the control-loop bandwidth.
–180 degrees. This is more of an issue for forward- or
other buck-derived topologies with higher crossover Discontinuous versus continuous
frequencies. conduction-mode
DCM occurs when the inductor current dwells at zero
Opto frequency response
20 45 before the end of the switching cycle. DCM is a normal
Gain condition for diode-rectified circuits when the load drops
0
Figure 53
0
Magnitude (dB)
-20 -45
which you can clearly see in Figure 53 for the current-
-40 RP = 1 kΩ -90
RP = 5 kΩ mode response. In general, if the control loop is stable in
-60 RP = 20 kΩ -135
Phase CCM, it will also be stable in DCM.
-80 -180
1 10 100 1000
Frequency (kHz) Current-mode buck power stage
40 90
Figure 51. The optocoupler pole moves toward lower frequencies at 20 45
Magnitude (dB)
-20 -45
Practical limitations due to the
-40 -90
switching frequency CCM Phase
-60 -135
DCM
The maximum crossover frequency is some fraction -80 -180
of the switching frequency. Figure 52 shows the 0.01 0.1 1 10 100 1000
Frequency (kHz)
control-loop phase margin going quickly negative at
the switching frequency. With a sufficient error-amplifier Figure 53. The current-mode power-stage bandwidth is reduced in
DCM.
re 54
For the voltage-mode response shown in Figure
54, the order of the filter is reduced in DCM, since
the inductor now acts as a current source.
Gain
0 0
Phase (°)
-20 -45
Figure 55. The converter acts as a load on the input filter. For stability,
filter ZOUT << converter ZIN.
-40 -90
-60 CCM Phase -135
DCM Equations 132 and 133 calculate the characteristic
-80 -180
0.01 0.1 1 10 100 1000 source impedance and resonant frequency,
Frequency (kHz) respectively:
LIN
Figure 54. The voltage-mode power-stage bandwidth is also reduced Z S = (132)
in DCM. CIN
1
DCM duty cycle f S = (133)
2 ⋅ π ⋅ LIN ⋅ CIN
Equations 129 through 131 show the duty-cycle
equations for DCM. To determine the mode The converter exhibits a negative input impedance,
boundary, you can set the CCM and DCM duty- which is lowest at the minimum input voltage
cycle equations equal to each other. (Equation 134):
2
2 ⋅ L ⋅ f SW ⋅ I OUT ⋅VOUT VIN V
Buck: D= (129) Z IN = − = − IN (134)
VIN ⋅ (VIN − VOUT ) I IN POUT
harness, as shown in Figure 55. Typical wiring there is not enough resistance in the circuit and the
inductance is on the order of 0.5 μH/m. An input input filter will sustain an oscillation.
across CIN to damp the input. You should evaluate three times larger than the crossover frequency.
any parallel capacitor for its root-mean-square • Damping – damp the second-stage filter to a Q of 1.
(RMS) current rating. The current will split between
the ceramic and aluminum capacitors based on the Primary-side compensation
relative impedance at the switching frequency. considerations
Primary-side compensation
Second-stage filter
Figure 57 illustrates a method that uses the
Figure 56 shows the power stage of a buck
Figure 57
regulator followed by a second-stage filter. The first-
primary-side inverting amplifier to implement
frequency compensation. Though not often
stage capacitor is sized to handle the output RMS
used alone, it can complement secondary-side
current, while the second-stage capacitor reduces
compensation. If the output-voltage accuracy is not
the ripple voltage and provides energy storage
critical, a series Zener and resistor can directly drive
for the load transient. This method is particularly
the optocoupler from the output voltage.
useful for boost- or buck-boost-type outputs, which
have high ripple voltage due to the pulsing rectifier
VC VCC
current.
You must include the effect of the second-stage
RF
filter in the control-loop analysis. Account for the RI
total output capacitance and use the capacitance -
to determine the loop bandwidth. A second-stage
+ VREF RE
resonance at too low a frequency can cause the
Figure 57. Primary-side compensation can provide additional control
control loop to go unstable, even when taking
of the frequency response.
feedback at the first-stage output.
58 This configuration minimizes the pole due to the bandwidth. This method cancels the inherent pole
optocoupler’s parasitic capacitance, since the of the optocoupler.
collector-to-emitter voltage does not change. This
V′OUT
configuration is also particularly useful in forward RD CFF
converters to achieve high bandwidth; we used it
RFF
in the section covering the isolated current-mode
forward example.
CCOMP RFBT
VCC
VC TL431 RFBB
RP
Figure 60. Placing a lead network across the optocoupler’s
- gain-setting resistor adds phase boost.
+ VREF RE
Zener bias
Figure 58. The optocoupler emitter is biased at VREF of the primary- Figure 61 shows how Zener bias for RD eliminates
Figure 61
side amplifier in the high-bandwidth configuration.
the high-frequency feedback path for secondary-
side compensation. This method is common for
Secondary-side compensation higher output voltages to prevent exceeding the
considerations TL431’s voltage rating. RS and ZS can also bias the
ESR zero compensation optocoupler from the rectified secondary voltage.
Magnitude (dB)
igure 62 bandwidth. Bench measurements of the control
Phase (°)
loop agreed with the transient response shown 0 -45
0.1 μF
Figure 64 (b) -40 -135
-60 -180
3.3 μH
0.1 1 10 100 1000
VIN BOOT VOUT
VIN PH Frequency (kHz)
15 μF
31.6 kΩ
GND U1 47 μF (a)
VSENSE
EN
e 63
COMP 10 kΩ
SS/TR 5 kΩ Error amplifier
RT/CLK
22 μF 60 270
10 nF 100 kΩ 1 nF effective
capacitance 40 225
Magnitude (dB)
Phase (°)
Figure 62. A synchronous buck regulator exhibits poor transient 20 180
performance. 0 135
Figure 64 (c) -20
-40
90
45
0.1 1 10 100 1000
(b)
VOUT
Control loop
60 135
40 90
Magnitude (dB)
Phase (°)
Magnitude (dB)
Figure 65 shows the revised circuit, while
Phase (°)
0 -45
Figure 66 shows the improved transient response.
-20 -90
0.1 μF
-40 -135
3.3 μH
Figure 67 (b)
VIN BOOT VOUT
VIN PH -60 -180
15 μF
31.6 kΩ 0.1 1 10 100 1000
GND U1 47 μF
VSENSE Frequency (kHz)
66 EN
SS/TR
COMP
RT/CLK
10 kΩ
1 kΩ (a)
22 μF
10 nF 100 kΩ 10 nF effective
capacitance
Error amplifier
Figure 65. Adjusting only two compensation components improved 60 270
the performance.
Magnitude (dB) 40 225
Phase (°)
20 180
0 135
IOUT -20 90
Figure 67 (c) -40 45
0.1 1 10 100 1000
VOUT Frequency (kHz)
(b)
Figure 66. A critically damped transient response indicates good Control loop
phase margin. 60 135
40 90
Magnitude (dB)
20 45
Figure 67, you can analyze the performance.
0 0
In Figure 67a, there is no external adjustment for
-20 -45
slope compensation. All power-stage components
-40 -90
are the same. In Figure 67b, decreasing RCOMP
0.1 1 10 100 1000
lowers the mid-band gain, which will lower the Frequency (kHz)
crossover frequency. Increasing CCOMP to move the
error-amplifier zero to a lower frequency increases (c)
2. Dixon, L.H. “Current-Mode Control of Switching Power Electronics Conference and Exhibition – APEC 2014,
Seminar SEM400, 1985. 12. Meeks, D. “Loop Stability Analysis of Voltage Mode
3. Dixon, L.H. “The Right-Half-Plane Zero – A Simplified Buck Regulator with Different Output Capacitor Types
Explanation.” Texas Instruments Power Supply Design – Continuous and Discontinuous Modes.” Texas
Internet
TI Semiconductor Product Information Center
Home Page
support.ti.com
The platform bar and E2E are trademarks of Texas Instruments. All other trademarks
are the property of their respective owners.
TI PROVIDES TECHNICAL AND RELIABILITY DATA (INCLUDING DATASHEETS), DESIGN RESOURCES (INCLUDING REFERENCE
DESIGNS), APPLICATION OR OTHER DESIGN ADVICE, WEB TOOLS, SAFETY INFORMATION, AND OTHER RESOURCES “AS IS”
AND WITH ALL FAULTS, AND DISCLAIMS ALL WARRANTIES, EXPRESS AND IMPLIED, INCLUDING WITHOUT LIMITATION ANY
IMPLIED WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT OF THIRD
PARTY INTELLECTUAL PROPERTY RIGHTS.
These resources are intended for skilled developers designing with TI products. You are solely responsible for (1) selecting the appropriate
TI products for your application, (2) designing, validating and testing your application, and (3) ensuring your application meets applicable
standards, and any other safety, security, or other requirements. These resources are subject to change without notice. TI grants you
permission to use these resources only for development of an application that uses the TI products described in the resource. Other
reproduction and display of these resources is prohibited. No license is granted to any other TI intellectual property right or to any third
party intellectual property right. TI disclaims responsibility for, and you will fully indemnify TI and its representatives against, any claims,
damages, costs, losses, and liabilities arising out of your use of these resources.
TI’s products are provided subject to TI’s Terms of Sale (www.ti.com/legal/termsofsale.html) or other applicable terms available either on
ti.com or provided in conjunction with such TI products. TI’s provision of these resources does not expand or otherwise alter TI’s applicable
warranties or warranty disclaimers for TI products.
Mailing Address: Texas Instruments, Post Office Box 655303, Dallas, Texas 75265
Copyright © 2020, Texas Instruments Incorporated