Feedback Control Design of Off-Line Flyback Converter: Application Note
Feedback Control Design of Off-Line Flyback Converter: Application Note
Feedback Control Design of Off-Line Flyback Converter: Application Note
Edwin Wang
AN017 – Jun 2014
Abstract
Controlling the feedback of off-line flyback converters has often perplexed power engineers because it involves the continuous
conduction mode (CCM) and discontinuous conduction mode (DCM) small signal models. Due to the unique feedback
compensation mode of the TL431 along with the optocouplers, tuning the feedback parameters still relies on trial and error. This
application note provides comprehensive design guidelines, from illustrating power circuit transfer function to designing the
circuitry for the TL431 and the optocouplers, to assisting system designers to gain a good phase margin so as to meet the
requirements of transient stability. In this note, theoretical computations are done by Mathcad software and verified by Simplis.
This method will be applicable to all applications with RT773x series off-line flyback controllers.
Contents
Most flyback converters use secondary-side peak current-mode control of the secondary-side converters to adjust feedback for
the output voltages as in Figure 1. The secondary-side output voltage is fed back through the TL431 and the optocoupler to the
primary-side. The output of the optocoupler, VCOMP, is compared with the primary-side peak current. This result serves as the
negative feedback to the loop and then determines the duty cycle of the switching component Q. R S is the resistor for the
primary-stage current detection. CTR is the current transfer ratio of the optocoupler. G FB is the small-signal gain. (Note: It is
designed to be about 1/3 of gain internally for all the RT773x ICs.) Se is added externally as the slope compensation to eliminate
sub-harmonic oscillation.
Some basic assumptions are made to facilitate the following derivations and explanations as below:
In reality, the current transfer ratio is a relatively nonlinear value, varying with the operating point (which is the current through the
diode in the optocoupler). However, to simplify the derivations, we assume it is a constant value over the currents through it. In
practical applications, currents through the optocoupler diodes are fairly low, as low as 1mA, which causes a current transfer ratio
less than 20%.
D : duty cycle
fs : switching frequency
n Vo
M : voltage transfer ratio
Vin
Np
n : transformer coil ratio
Ns
2 Lp fs
τL : time constant
n2 R
Vin
Sn Rs : the voltage slope when the primary-side current is detected on RS (unit: V/sec)
Lp
vˆRS
GFB : the small-signal gain
vˆFB
n:1 VO
D
Co
VIN LP LS R
RC
RS Q
Sn
VREG
Se
Rc3
Rd CTR
VCOMP
GFB Ra
Ca
Cb
PC817
TL431
Rb
Figure 1. The schematic of the flyback converter with the TL431 and the optocoupler.
Various small-signal models of flyback converters can be found in many references [1-3]. These models are all derived based on
the method of state averaging. There are some minor differences among them, which probably result from the different
assumptions being made. In this note, the small-signal model of Christophe Basso [1] is adapted for our feedback compensation
design. However, with other small-signal models, similar results can also be achieved.
s s
(1 ) (1 )
vˆo (s) ωz1 ωz2
G0 (1)
vˆcomp (s) s
(1 )
ωp1
where,
n R GFB 1
G0
Rs (1 D)2 S
1 2 e 2M 1
τL Sn
(1 D)3 S
(1 2 e ) D 1
τL Sn
ωp1
R Co
1
ωz1 (ESR Zero, LHP)
Rc Co
(1 D)2 n2 R
ωz2 (RHP Zero)
D Lp
This is a 1 pole 2 zeros system, as shown in Figure 2. The pole is determined by the circuit parameters and the size of the load.
The first zero is fixed because it depends on the output capacitance and the equivalent series resistance (ESR). The other zero is
on the right half s-plane, so called RHP zero. The location of RHP zero is determined by the input voltage, and the load current.
Usually, in a well-designed system, the cross-over frequency is set far below the RHP zero frequency so the system can have
sufficient phase margins. Based on this fact, this RHP zero is regarded as negligible when designing the compensation circuit.
0
Gain (dB)
fp1
-1
+1
0
fz1 fz2
Frequency (Hz)
s s
(1 ) (1 )
ˆ
vo (s) ωz1 ωz2
G0 (2)
vˆcomp (s) s s
(1 ) (1 )
ωp1 ωp2
where,
fs R 1
G0 Vin GFB
2 Lp (Sn Se )
2
ωp1
R Co
2
1
ωp2 2 fs D
1
(1 )
M
1
ωz1 (ESR Zero, LHP)
Rc Co
n2 R
ωz2 (RHP Zero)
M (1 M) Lp
The transfer function of Equation (2) is shown in Figure 3. In small-signal model of discontinuous conduction mode, DCM, the
power circuit has two poles and two zeros. One of the poles, ωp2, is extremely high (far above the target cross-over frequency).
Therefore, designing the compensation, this pole can be neglected. As a result, with the small-signal model of either CCM or
DCM, their transfer functions can be considered as 1 pole 2 zeros. Therefore, the selection of the feedback network becomes
much easier.
From the transfer functions of Equation (1) and (2), some poles and zeros are fixed, such as the zero from the output capacitance
and the equivalent series resistance ESR. However, most poles and zeros are influenced by the operating point, which describes
the operating condition of the circuit and is specified by the input voltage and the load current condition. Next, the changes of the
poles and zeros with the operating points will be illustrated with the circuit parameters plugged in.
0
Gain (dB)
fp1 -1 fp2
+1 0
0
fz1 fz2
Frequency (Hz)
With a flyback converter as an example, given input voltage: 90V to 360V, load current: 0-3A, and output voltage as 12V. The
circuit parameters are as below: LP = 1.1mH, NP/NS = n = 7.7, CO = 1360μF, RESR = 30 mΩ, RS = 0.56Ω, fS = 65kHz, Se = 3.46 x
104 V/sec, GFB = 0.3333, where Se and GFB are provided by the controller IC。According to the operating principle of flyback
converters, in a typical design, high input voltage with light load always causes converters to operate in continuous conduction
mode; on the contrary, low input voltage with heavy load in the discontinuous conduction mode. A boundary exists between CCM
and DCM, as shown in Figure 4 and its equation is as Equation (3).
3.5
3.0
Load Current (A)
2.5
2.0
1.5
1.0
0.5
0.0
90 135 180 225 270 315 360
n2 Vo VIN2
Io (3)
2 LP fS (VIN nVo )2
Table 1 shows the DC gains and the locations of poles and zeros. Figure 5 shows the Bode plot for the different input voltages
and load currents. From the plot, it is clear that the gain is higher with high line and light load. This helps in choosing the operating
point as the criteria for feedback network design. It is better to design the feedback network under low line and heavy load
condition. Adequate phase margin under such condition, can achieve an even better phase margin at different operating
conditions.
Table 1. The DC gains and the locations of poles and zeros at different operating points
VIN (V) 90 180 270 360 90 90 90 360 360 360
IO (A) 3.0 3.0 3.0 3.0 3.0 2.0 1.0 3.0 2.0 1.0
Mode CCM CCM CCM DCM CCM CCM DCM DCM DCM DCM
G0 (dB) 13.1 16.5 17.0 17.1 13.1 15.6 17.0 17.1 18.8 21.8
fP1 (Hz) 59.0 53.0 57.0 58.5 59.0 44.0 19.5 58.5 39.0 19.5
fP2 (Hz) NA NA NA 21.7k NA NA 25k 21.7k 32.6k 65k
fZ1 (Hz) 3.9k 3.9k 3.9k 3.9k 3.9k 3.9k 3.9k 3.9k 3.9k 3.9k
fZ2 (Hz) 16.5k 44.2k 75k 106k 16.5k 24.7k 49.5k 106k 160k 319k
CCM DCM
High Line High Line
Gain (dB)
Gain (dB)
fp1 fp1
CCM DCM
Light Load Light Load
Gain (dB)
Gain (dB)
fp1 fp1
Figure 5. The variations of the gains vs. frequency with different operating conditions.
From the previous analysis, poles and zeros vary with operating points as do low-frequency DC gains. There are many ways to
design compensation circuits. Typically, a Type-II compensator (with one zero frequency pole, one low-frequency zero, and one
pole) is most suitable in this case. Making a low-frequency zero to compensate the low-frequency pole and also making a
high-frequency pole to compensate the ESR zero can achieve a better phase margin. A specific mid-band gain is chosen to get
the proper cross-over frequency, and the system will therefore be stabilized.
k
loop gain = (4)
s
Such loop gain is just a straight line with the slope of -20dB/dec on the Bode plot as seen in Figure 6. At the low frequency or DC
region, the gain is extremely high (equivalent to the open-loop gain of the compensator), so the theoretical adjustment rate of the
DC regulating voltage can be set as zero. Its cross-over frequency, fC, is
k
fc (5)
2π
Since its slope is about -20dB/dec, its phase margin is about 90° near the cross-over frequency.
For an off-line flyback converter, it is best to set the cross-over frequency between 800Hz and 3kHz under the condition of low
line and full load (where switching frequency is 65kHz).
Gain (dB)
2π fc
s
0
fc
Frequency (Hz)
Figure 6. The transfer function of the power circuit (in red) and its target loop gain (in blue).
Design Procedure
As discussed above, any compensation method will work. The design procedures are listed below:
1. Design the compensation network under the condition of low line (input voltage) and full load. It usually will give a very
good phase margin even at various conditions.
2. Set the cross-over frequency fC, and its loop gain as -20dB/dec in the Bode plot. Higher cross-over frequency means
faster transient response. However, it is impossible to compensate RHP zero by the poles. Therefore, the cross-over
frequency must be far below from RHP zero. Practically, the cross-over frequency is set below 3kHz.
3. Define a 2-pole-and-1-zero compensation circuit and set the zero as the low-frequency pole of the power circuit. Set the
high-frequency pole as the ESR zero of the power circuit. Implement the design by a Type II compensator. Its transfer
function can be the “target loop gain”.
4. Calculate mid-band gain according to power circuit gain at f C.
5. Estimate phase margin.
6. The transfer function of the compensation network is determined.
s
(1 )
ωcz1
Gcomp (s) A
s
s(1 )
ωcp1
(6)
Implementation of Compensator
1. This note illustrates the design method with a typical Type II circuit, a widely used circuit block of TL431 and the
optocoupler, as shown in Figure 7.
Vreg Vo
Rd
Rc3
Vc Ra
Cb Ca
Opto
TL431
Rb
2. The small-signal transfer function of the circuit in Figure 7 is as shown below [5]:
Gain (dB)
fcz1 fcp1
Frequency (Hz)
3. From Equation (7), there are 7 parameters, Ra、Rb、Rc3、Rd、Ca、Cb and CTR to be decided, and only three of them are
known.
Rd 1
CTR A (8)
Rc3 Ca Ra
1
ωcz1 (9)
Ca Ra
1
ωcp1 (10)
Cb Rd
4. First, choose Rd. Most new controller IC’s have set the Rd value, which can be obtained from vendors.
5. Next, the reference voltage VREF can also be obtained from vendors, typically 2.5V. To make TL431 function properly,
the current through Rb (Ivd) must be at least 125μA. Ivd is usually specified as 250μA with some margins. Therefore, the
values of Ra and Rb can be determined.
VREF
Rb (11)
Ivd
6. The current transfer ratio of the optocoupler (CTR) can be estimated according to the information provided by vendors.
CTR is a nonlinear value, varying with the current through the optocoupler diode. Usually the current is on the order of
hundreds μA with CTR around 0.1 to 0.5. The exact value can be found through the measurement. Here CTR is
assumed to be 0.5.
7. Now, four of the seven parameters are decided. The rest three can be calculated by Equation (8), (9) and (10).
8. RC3 acquired from Equation (8) must be evaluated. From how the TL431 works, the cathode voltage must be higher than
2.5V and the current through the cathode (I cathode) should be greater than 1mA to get a correct regulating voltage.
Usually, a 1kΩ resistor will be added in shunt with the optocoupler diode to provide sufficient cathode current. This shunt
resistor will not change the small signal model. Therefore, we can derive the following:
Iopto _ transistor
Icathode VF 1mA (14)
CTR Rparallel
where VF is forwarded biased voltage drop of the optocoupler, around 1.0V. The maximum value of RC3 can then be
estimated.
Vo VF 2.5
Rc 3 (15)
Icathode
Assume Icathode is 1.5mA, and RC3 should be less than 5.6kΩ. Excessice RC3 will lower the mid-band gain. If RC3 is
greater than the maximum value, the cross-over frequency will be set lower or another compensation method will be
used.
9. There exists an equivalent capacitance, around 2nF to 5nF, in shunt with the phototransistor of the optocoupler. The
total Cb value is this parasitic capacitance, which can be measured, and the external added capacitor. If the parasitic
capacitance is dominant, then no external capacitance is needed. Since ESR zero cannot be compensated completely,
the phase margin will become worse.
Two Mathcad computation procedures, Flyback CCM Type II Compensation” and “Flyback Loop Gain Analysis” have been used
to facilitate the calculation and analysis of the feedback network design. Simplis simulations are used to compare errors of the
simulation models. Figure 9 is the schematic of Simplis simulation circuit. Figure 10 to Figure 12 show the comparisons of
Mathcad analysis and Simplis simulations. Figure 10 and Figure 11 are the Bode plots of the transfer functions of the power
circuit and the compensation network, respectively. Figure 12 is the Bode plot of the closed-loop gain, (a) magnitude and (b)
phase. Red lines indicate the results from Mathcad calculation and blue lines from Simplis.
From the low frequency to cross-over frequency, the small signal models match very well. The model starts to show a large
mismatch at the high frequency region. However, since the loop gain is much less than 1, the mismatch can be neglected. Figure
13 shows the step response of the output voltage for the load current from 1A to 3A with 90V input voltage by Simplis simulation.
Fairly small overshoot and settling time are shown.
DIODE VOUT
VOUT
X1 IDEAL
30m
RC
90 P1 S1 4
VIN
1.36m RL
CO
U4
TX1
GND GATE
SW1
VAC
AC 1 0
FB VDD VA IN OUT VOUT
=OUT/IN
PRO CS
1.2k
RT7736_Simplified 560m VA
RCS Rc1 VFB IN OUT VOUT
VFB
=OUT/IN
U3 38k
1k Ra
Rf IN OUT
VA VFB
=OUT/IN
68n
240p Ca
Cb U2
TL431
10k
Rb
20 0
10
-45
Power Stage (deg)
Power Stage (dB)
-10 -90
-20
-135
-30
-40 -180
10 100 1000 10000 100000 10 100 1000 10000 100000
Frequency (Hz) Frequency (Hz)
Figure 10. The Bode plot of the transfer function of the power circuit (a) magnitude, (b) phase.
40 180
30
135
Compensator (deg)
Compensator (dB)
20
10 90
0
45
-10
-20 0
10 100 1000 10000 100000 10 100 1000 10000 100000
Frequency (Hz) Frequency (Hz)
Figure 11. The Bode plot of the transfer function of the compensation network (a) magnitude, (b) phase.
50 180
40 135
30 90
Loop Gain (deg)
Loop Gain (dB)
20 45 ɸm
10 0
0 -45
-10 fc -90
-20 -135
-30 -180
10 100 1000 10000 100000 10 100 1000 10000 100000
Frequency (Hz) Frequency (Hz)
Figure 12. The Bode plot of the loop gain (a) magnitude, (b) phase.
Reference
[1] Christophe P. Basso, “Switch-Mode Power Supplies Spice Simulations and Practical Designs”, McGraw_Hill, 2008.
[2] W. Kleebchampee and C. Bunlaksananusorn, “Modeling and Control Design of a Current-Mode Controlled Flyback Converter
with Optocoupler Feedback”, IEEE PEDS 2005.
[3] Yuri Panov and Milan M. Jovanovic´, “Small-Signal Analysis and Control Design of Isolated Power Supplies with Optocoupler
Feedback”, IEEE TRANSACTIONS ON POWER ELECTRONICS, JULY 2005.
Related Parts
Next Steps
Richtek products are sold by description only. Richtek reserves the right to change the circuitry and/or specifications without notice at any time. Customers should
obtain the latest relevant information and data sheets before placing orders and should verify that such information is current and complete. Richtek cannot assume
responsibility for use of any circuitry other than circuitry entirely embodied in a Richtek product. Information furnished by Richtek is bel ieved to be accurate and
reliable. However, no responsibility is assumed by Richtek or its subsidiaries for its use; nor for any infringements of patents or other rights of third parties which may
result from its use. No license is granted by implication or otherwise under any patent or patent rights of Richtek or its subsidiaries.