DC-DC Converters Feedback and Control

Dc-dc converters feedback and control
presented by Christophe Basso

Product Applications Engineering Director
Pardon his
French!
Wild Bill Hickok
Chris Basso – June 2008

Course agenda
 Feedback generalities
 Building an oscillator
 Poles and zeros
 Phase
Ph margin
i andd quality
li coefficient
ffi i
 Undershoot and crossover frequency
 Compensating
p g the converter
 Current-mode converters
 Automated pole-zero placement Oooh, that looks
 Manual pole-zero placement so good!
good!…
 Compensating with a TL431

 Watch the optocoupler!
 Multi-output
M lti t t converterst
 Input filter
 A real case example
p
 Conclusion
2 Chris Basso – June 2008
Course agenda
 Poles and zeros
 Phase
Ph margin
i andd quality
li coefficient
ffi i
 Compensating
p g the converter
 Automated pole-zero placement
 Manual pole-zero placement
 Multi-output
 Input filter
p
 Conclusion
What do we expect from a dc-dc?
 A stable
t bl output
t t voltage,
lt whatever
h t loading,
l di iinput,
t ttemperature
t
and aging conditions.
 A fast reaction to a incominggp perturbation such as a load
transient or an input voltage change.
 A quick settling time when starting-up or recovering from a
transient state
state.
A stable and noiseless dc source we can trust!

Ambient temperature
TA
Input voltage Output voltage

Vin dc-dc Vout
Output current
Iout

What is feedback?
 A target is assigned to one or several state
state-variables,
variables e
e.g.
g Vout
= 12 V.
 A dedicated circuit monitors Vout deviations.
 If Vout deviates from its target, an error is created and fed-back
to the power stage for action.
 The action is a change in the control variable: duty-cycle
duty cycle (VM),
peak current (CM) or switching frequency.
Compensating for the converter shortcomings!
Input
p voltage
g Output voltage
Vin d d
dc-dc Vout Input voltage
Rth
Vth Vout
Vin
action
control

The feedback implementation
 Vout is permanently compared to a reference voltage Vref.
 The reference voltage Vref is precise and stable over temperature.
 The error,   Vref  Vout, is amplified and sent to the control input.
 The power stage reacts to reduce  as much as it can.
Power stage - H
Vout
Control
variable d
Error amplifier - G
Rupper
+
-
Vin

-
+
Vp
Modulator - GPWM Vref Rlower

Course agenda
 Poles and zeros
 Phase
Ph margin
i andd quality
li coefficient
ffi i
 Compensating
p g the converter
 Multi-output
 Input filter
p
 Conclusion
Positive or negative feedback?
 Do we want to build an oscillator?
The « Plant »
Vin(s)
() +  Vout(s)
()
H(s)

G(s)
Vout  s  H s

Vin  s  1 H sG s Open-loop gain T(s)
 H s  To sustain self-oscillations, as Vin(s)

Vout  s   lim  Vin  s  
Vin  s   0 1  G  s  H  s  goes to zero, quotient must go infinite
 
Sign is neg for: What Plant??
 = -180° =1

Observing the 0 dB point
 Create phase lag: cascading RC networks
R1 R2 R3
80.0
1k 1k 1k
4 3 2 1
Vout
40.0 AC = 1 C1 C2 C3
V1 10n 10n 10n
H(s)
vdb1 in db(volts)
plot1
-40.0
|H| = -29 dB
-80.0
100
argH(s) 38 kHz
0
es
ph_v1 in degree
plot2
-100
-200
 = -180°
180
3
-300
100 1k 10k 100k 1Meg

f i h

A constant gain with a 180° rotation
 Add gain to obtain T(s) = 0 dB where  = -180°
180° (38 kHz)
R1 R2 R3
1k 1k 1k
1
C2
2 3
parameters
t
C1 C3
10n 10n 10n
Gfc=-29
G=10^(-Gfc/20)
4
Ri 100k
Ri=100k
Rf=G*100k
X2 Vout
UM2
SUM2 R5
SU
K1 = 1 {Rf} R4
K2 = 1 {Ri}
K1
K2
6 7
Vin 5
AC = 1
E1
10k

A constant gain with a 180° rotation
 Starting the oscillator with a « crank », play with gain
Vout
R1 R2 R3
1k 1k 1k parameters
4 1 2 3
C1 C2 C3 Gfc=-29
10n 10n 10n
G=10^(-Gfc/20)
Ri=100k
Ri 100k
Rf=G*100k
X2 R55 R4
SUM2
2
SUM2 {Rf} { }
{Ri}
K1 = 1
K2 = 1
K1
K2
Yank me
6 5
Crank me!
1V
V1
0 Tran Generators = PWL
E1
10k

A simple oscillator
n°1, gain is below 1 at  = -180°
 Case n°1 180°
 Oscillations are damped, system is asymptotically stable
40.0
Vout(t)
vdboutin db(volts)
20.0 400m
plot1
-20.0
|H| = -3.6 dB 3
200m
-40.0
1
vout in volts
Plot1
0
180
ph_voutin degrees
90.0 -200m
Plot2
4
-90.0 -400m
-180
 = -180°
10.0u 30.0u 50.0u 70.0u 90.0u
time in seconds
10 100 1k 10k 100k
frequency in hertz

A simple oscillator
n°2, gain is above 1 at  = -180°
 Case n°2 180°
 Oscillations are not damped, system diverges
40.0
800m
Vout(t)
vdboutin db(volts)
20.0
plott1
4
|H| = 3 dB
400m
-20.0
-40.0
vout in volts
plot1
0
180
1
-400m
ph_voutin degrees
90.0
plot2
5
800
-800m
-90.0
-180
 = -180° 10.0u 30.0u 50.0u 70.0u 90.0u
time in seconds
10 100 1k 10k 100k

frequency in hertz

A simple oscillator
n°3, gain is equal to 1 at  = -180°
 Case n°3 180°
 Oscillations are sustained, we have an oscillator!
40.0
20.0
Vout(t)
ut in db(volts)
400m
800m
plot1
0
vdbou
-20.0
|H| = 0 dB 5 200m
400m
-40.0
vout in volts
vout in volts
Plot1
plot1
0
1
180
-200m
-400m
90.0
ph_vout in degrees
Fosc = 38 kHz
plot2
6 -400m
-800m
-90.0
 = -180° 10.0u 30.0u 50.0u 70.0u 90.0u

-180 time in seconds
10 100 1k 10k 100k

frequency in hertz

Conditions for oscillations
 when the open
open-loop
loop gain equals 1 (0 dB) – crossover point
 phase rotation is -360° in total (-180° for H and -180° for G)
 we have self-sustaining oscillating conditions
Total phase delay at fc:
180
Loop gain |H(s)| -180° H(s) power stage

Gain is 1 -180°
180° G(s)
G( ) opamp
90.0 at fc
total = -360°
0
0 dB
Loop phase arg H(s)
-90.0  = -180
-180°
-180 21
22 -180°
1 10 100 1k 10k 100k

frequency in hertz

Conditions for steady-state stability
 We do not want to create an oscillator!
 Conditions for non-permanent oscillations are:
 total phase rotation less than -360° at the crossover point
Total phase delay at fc:
180
Loop gain |T(s)| Gain is 1 112 H(s) power stage

-112°
at fc -180° G(s) opamp
90.0
total = -292°
0
Loop phase arg T(s)
 = -112° 26
Stable!
-90.0 25
m = 68°
-180
1 10 100 1k 10k 100k

frequency in hertz

The need for phase margin
 we need phase margin when T(s) = 0 dB
 we need gain margin when arg T(s) = -360°
Phase margin:
180 80.0
phase
phase The margin before the loop
margin
gain phase rotation arg T(s)
arg T(s)
90.0 40.0
= -360° reaches -360° at T(s) = 0 dB
degrees
b(volts)
0
0°
vdbout in db
ph_vout in d
2
Plot1
0 0
0 dB
gain
margin
-90.0 -40.0
Crossover
frequency
f fc
1
T(s) = 0 dB Gain margin:

The margin before the loop
-180 -80.0
gain T(s) reaches 0 dB at a
10 100 1k 10k 100k
freq. where arg T(s) = -360°

Hey, -360°, -180° or 0°?
 w2 is delayed by -155°
155° compared to w1
 w2 is also in advance by 205° compared to w1
 if the delay further shifts to -360° (2) reading goes back to 0°
1.60
4.3 µs or 155° 5.69 µs or 205°
w1 Called
modulo 2
800m
29
olts
v1, v2 in vo
Plot1
0 28
-800m
w2 F = 100 kHz
-1.60 T = 10 µs
360°  10 µ
µs
1.50u 4.50u 7.50u 10.5u 13.5u
time in seconds

Hey, -360°, -180° or 0°?
 all these plots read the same phase margin!
80.0 180 gain 0°: modulo 360° (or modulo 2) reading
40.0 90.0
phase m 4
0 0
0°
-40.0 -90.0
6
-80.0 -180
80.0 360
-180°: power stage H(s), comp. G(s)
40.0 270 gain
0 0
-40.0 -90
phase m 6
4
-80.0 -180 -180°
80.0 360 -360°:

360 popower
e stage H(s),comp.
(s),co p G(s) a
and
d op amp
a p inversion
e so
40.0 180 gain
0 0
-40.0 -180
m
6
-80.0 -360 phase -360° 7
1 10 100 1k 10k 100k

frequency in hertz

Course agenda
 Poles and zeros
 Phase
Ph margin
i andd quality
li coefficient
ffi i
 Compensating
p g the converter
 Multi-output
 Input filter
p
 Conclusion
Poles, zeros and s-plane
 A plant loop gain is defined by:
N s numerator
H s 
D s d i t
denominator
 solving for N(s) = 0, the roots are called the zeros
 solving for D(s) = 0, the roots are called the poles
5k
Two zeros f z1   796 Hz
2
sz1  5k
H s 
 s  5k  s  30k 
sz2  30k f z2 
30k
 4.77 kHz
2
s  1k 1k
s p1  1k f p1   159 Hz
O pole
One l 2

 How do the poles influence the temporal response of the plant?
 assume an input-step response is wanted:
 multiply H(s) by 1/s
 take the inverse Laplace transform
 plot the response
1

2 1 1 2
H s 
s
H s 
 s  1 s  2  s s  s  1 s  2 
Take inverse
s p1  1 s p2  2 Laplace transform
1
1{H  s  }  0.5  e 1t  0.5e 2t
s
 The roots are the exponentials exponents: -1 and -2
 If the roots are negative
negative, the signal is decaying
 If the roots are positive, the response diverges
 The roots can either be real or imaginary:
s4 j Imaginary
H s  axis
 s  0.8
0 8   s  2.5
2 5  4
2
  s p2
x
2j
sz1  4 sz1 s p1
Real
s p1  0.8 x  axis
s p2  2.5  2 j Conjugated
s p3  2.5  2 j roots x
s p3
Left Half Plane Right Half Plane
 We can place these roots in the imaginary plane

 root-locus analysis in the s-plane
 their positions in the s-plane affect the temporal response!

j j
LHP RHP x sz2

1V
Q
x x 0
 0  1
sz1 sz2
Q < 0.5
Q Vout Vout
x sz1
Q Response to a step Response to a step
j j Q  -
s z1 Q = 0.5 1V x
s z2 s z2
Q < -0.5
x 0
 0  1
sz1
Vout x Vout
Response to a step Q  - Response to a step
j j
1V
x sz2 Q > 0.5
-0.5 < Q < 0
0
 0 x x
s z1 sz2
x sz1
Vout Q = -0.5 Vout
Response to a step Response to a step
 A quick refresh on imaginary numbers…
numbers
A  a  jb b
A    a 2  b2 arg A    arctan  
a
A   e j Algebric form
Euler
A    cos   j sin   Trigonometric form
arctan(x) Im
+90° b a  jb

  
x 
a
-90°
N Nc
A c A arg  A   arg  N c   arg  Dc 
Dc Dc

 A pole lags the phase by -45°
45° at its cutoff frequency
0
Vin Cutoff
Vout 20.0
f
frequency
R2 0
- 3 dB
1k
vdbout in db(volts)
Plot1
-20.0
V1 C1
-40.0 -1 slope
-20 dB decade
AC = 1 10nF -60.0
1
-20.0
ph_vout in degrees
Vout ( s ) 1 1 - 45° at
Plot2
 
-40.0
cutoff
Vin ( s ) 1  sRC 1  s -60.0
0 -80.0
2
10 100 1k 10k 100k 1Meg 10Meg
frequency in hertz
1
0 
RC

 Its module at the cutoff frequency is -3
3 dB
 Its asymptotic phase at f = ∞ is -90°
 The pole "lags" the phase
Vout ( s ) 1 1
 
Vin ( s ) 1  sRC 1  s
0
Vout ( s p1 ) 1 1
20 log10  20 log10  20 log10  3 dB At f = fp1
Vin ( s p1 ) s p1 2
1
s p1
Vout ( s p1 )  s p1  
arg  arg 1  arg  1     arctan 1   At f = fp1
Vin ( s p1 )  sp  4
 1 
Vout ()    
arg  arg 1  arg 1     arctan      At f = ∞
Vin ()  
 s p1  2

 A zero boosts the phase by +45° at its cutoff frequency
0 0
40.0 1 20.0
+1 slope 0 1
vdbout in db(volts)
30.0
20 dB decade
vdb2 in db(volts)
plot1
-20.0
Plot1
20.0
Cutoff
10.0
-40.0 frequency
+1 slope -3 dB
Cutoff -60.0
0
frequency
20 dB decade
10 100 1k 10k 100k 1Meg 10Meg
frequency in hertz
90.0 2
90.0
70.0
70.0
_v2 in degrees
out in degrees
Plot2
50.0
plot2
+ 45° at 50.0 45° at
ph_
ph_vo
cutoff cutoff
30.0
30.0
10.0 10.0
2
10 100 1k 10k 100k 10 100 1k 10k 100k 1Meg 10Meg
frequency in hertz frequency in hertz
Vin Vout
The general form of a zero:
C1 s
10nF
s Vout ( s ) sRC 0
G (s)  1   
0 V1 R2 Vin ( s ) 1  sRC 1  s
AC = 1 1k
0
1
0 
RC
 Its module at the cuttoff frequency is +3 dB
 Its asymptotic phase at f = ∞ is +90°
 The zero "boosts" the phase
Vout ( s ) s
 1
Vin ( s ) 0
Vout ( sz1 ) sz1

20 log10  20 log10 1   20 log10 2  3 dB At f = fz1
Vin ( sz1 ) sz1
Vout ( s p1 )  sz1  
arg  arg  1    arctan 1   At f = fz1
Vin ( s p1 )  sz  4
 1 
Vout ()    
arg
g  arg
g 1    arctan      At f = ∞
Vin ()  sp  2
 1 

 Poles and zeros can sometimes appear "at the origin"
s
Vout ( s )  As f increases the gain increases
 0 Zero for s = 0: zero at the origin
Vin ( s ) D  s 
with a +1
1 slope ((+20
20 dB/decade)
 szo 
Vout ( szo ) s  
argg g  zo   arctan     
 arg For f > fzo
Vin ( szo )  0  2
 
 
Vout ( s ) N  s  As f increases the gain decreases
 P l for
Pole f s = 0:
0 pole
l att the
th origin
i i
with a -1 slope (-20 dB/decade)
Vin ( s ) s
0
 s po 
 
arg
Vout ( s po )
 arg 1  arg 
s po    arctan       For f > fpo
Vin ( s po )  0  2
 
 

The Right Half-Plane Zero
boost Iout is delivered during the off time: I out  I d  I L 1  D 
 In a CCM boost,
Id(t) Id(t)
IL IL
0 1
Vin Vin
L L
IL(t) IL(t)
Id
Id
0 d̂ 1
t t
D0Tsw D1Tsw
Tsw Tsw
increases D
 If D brutally increases, D' reduces and Iout drops!
d VL  t 
 What matters is the inductor current slew-rate dt
 If IL(t) can rapidly change,
change Iout increases when D goes up
200 µs
59%
d(t) Vout(t)
58.3%
IL(t)
Iout(t)
100u 300u 500u 700u 900u

 If IL(t) is limited because of a big L,
L Iout drops when D increases
10 µs 2
59%
d(t) Vout(t)
58.3% Vout drops!
IL(t)
Iout(t)
Iout drops!
100u 300u 500u 700u 900u

 Small-signal
Small signal equations can help us to formalize it
I out  I L 1  D 
 I   I 
i?out   out iL    out d?   i?L 1  D   dI L The negative sign
 I L  D  D  I L indicates a positive root!
 s 
 1  
îout  s  V D '  sL   V D ' R D ' 2
  Voltage
  
z
 out 1  2 
2
 out load
dˆ  s 
0 z mode
sL  D ' Rload  s L 2
L
0
iôut  s  D' sL  s 
 z2 
Rload D '2 Current
   G0 1   mode
d
vˆc  s  Rsense D ' Rload Rsense  z
 2 
L
 Voltage mode or current

c rrent mode
mode, the RHPZ (song) remains the same

 To limit the effects of the RHPZ
RHPZ, limit the duty
duty-cycle
cycle slew-rate
slew rate
 Chose a cross over frequency equal to 20-30% of RHPZ position
 A simple RHPZ can be easily simulated:
R1
10k
K1
C1
10n
SUM2
4 Vout(s)
Vin(s) 3 2 1
K2
X1
SUM2
K1 = 1
K2 = 1
E1
10k
R1  s 
V ( s )  Vin ( s )  Vin ( s )  Vin ( s ) 1  
1  0 
sC1

The Right-Half-Plane-Zero
 With a RHPZ we have a boost in gain but a lag in phase!
1
40.0
||Vout((s)|
)| +1
20.0
vdbout in db(volts)
Plot1
-20.0
-40.0
180
argVout(s)
90.0
egrees
ph_vout in de
Plot2
-90.0 -90° 2
-180
1 10 100 1k 10k 100k 1Meg

frequency in hertz

 A RHPZ also exists in DCM boost,
boost buck
buck-boost
boost converters
Id(t)
IL,peak(t)
D3Tsw
t
D1Tsw D2Tsw
Tsw
 When D1 increases,
increases [D1,D
D2] stays constant but D3 shrinks

 The triangle is simply shifted to the right by dd̂1
Id(t)
IL,peak(t)
D3Tsw
d̂1
t
D1Tsw D2Tsw
Tsw
 The refueling time of the capacitor is delayed and a drop occurs

 If D increases,
increases the diode current is delayed by dd̂1
7.00 340m
i(b2), i(b2)#a in amperes
5.00 300m
Id(t) D(t) 4
vduty in volts
1
Plot1
3.00 260m
1.00 220m
-1.00 180m d̂1 3

2
2.02m 2.04m 2.06m 2.09m 2.11m

time in seconds
7.05
1
6.95
voutt in volts
Plot3
6.85
P
6.75
6.65 Vout(t)
1.85m 1.99m 2.13m 2.27m 2.41m
time in seconds
7.05
Vout(t)
6.95
vout in volts
Plot2
6.85
1
v
6.75
6.65
2.02m 2.04m 2.06m 2.09m 2.11m

ti i d
 Averaged models can predict the DCM RHPZ
L1
75u 10.0V
 
1  s sz1 1  s sz 2 
11
vout vôutt  s 
 Hd
c
15 0V
15.0V
p  
1  s s p1 1  s s p 2 
Vout
10.0V
PWM switch VM d s
ˆ
R10
3
150m
Vin R11
10V
1 Rload
a
15.0V 150
d
16
X3 C5 sz1  s z2 
PWMVM
L = 75u
1m
Cout RESR M 2L
Fs = 100k
2
2M  1  11 M 
vout
1
s p1  s p2  2 Fsw  
R1 M  1 Cout RESR  D 
50k
2.50V
5
LoL
1kH 278mV
278 V Merci
2V M  1
278mV
1 8
Vatché!
6
H d  out
CoL
1kF
Verr
X2 V2
R3
10k D 2M  1
0V AMPSIMP 2.5
10
V1x
V1
AC = 1

 Averaged models can predict the DCM RHPZ
H d  28.75 dB
180 40.0
28.6 dB
f z1  1.06 kHz
90.0 20.0
f z2  141 kHz
fp1 f p1  4.2 Hz
ph_vvout in degrees
out in db(volts)
f p2  47.1 kHz
plot1
0 0
vdbo
fz2
-45°
-90.0 -20.0
fz1 7
fp2
8
-180 -40.0
1 10 100 1k 10k 100k 1Meg

frequency in hertz
4.2 Hz 1 kHz 47 kHz 141 kHz
Course agenda
 Poles and zeros
 Phase
Ph margin
i and d quality
li coefficient
ffi i
 Compensating
p g the converter
 Multi-output
 Input filter
p
 Conclusion
How much margin? The RLC filter
 let us study an RLC low-pass
low pass filter,
filter a 2nd order system
R1 L1
1
T s 
{R} {L}
Vout
3 2 1
LCs 2  RCs  1
C1
1 1
Vin {C}
T s  
s2 s s2 s
 2 1  1
r 2 r r 2 r Q
parameters
t 1
r 
LC
f0=235k
L=10u 1
C=1/(4*3.14159^2*f0^2*L) C Q
 R
w0=({L}*{C})^-0.5 4L 2
Q=10 zeta
R=1/((({C}/(4*{L}))^0
R=1/((({C}/(4 {L}))^0.5)
5)*2*{Q})
2 {Q})

The RLC response to an input step
 changing Q affects the transient response
Q=5 Q < 0.5 over damping

1.80
Q=1 Q=0 0.5

5 critical damping
Q > 0.5 under damping
Q = 0.707
vout#6, vout#5, voutt#4, vout#3, vout in vvolts
1.40
Overshoot = 65%
Plot1
1.00 11
9
10
8
7
P
Asymptotically stable
600m
Q = 0.5 Fast response and no overshoot!
200m Q = 0.1
5.00u 15.0u 25.0u 35.0u 45.0u

ti i d

The RLC response to an input step
 Q affects the poles position
1 N s Solve N(s) = 0 to obtain zeros

T s  
s2 s D s S l D(s)
Solve D( ) = 0 tto obtain
bt i poles
l
 1
r 2
r Q
j
s2

s
r r Q
2
1  0 2 roots, s1 and s2 =
r
2Q
 1  1  4Q 2
 Q=
High Q
Q < 0.5, two real negatives roots

Q = 0.5, two real coincident negative roots 
Low Q Low Q
Q > 0.5,
0 5 two complex roots
Q = , two imaginary conjugate roots Q = 0.5
High Q
LHP RHP

Where is the analogy with T(s)?
 in the vicinity of the crossover point
point, T(s) combines:
 one pole at the origin, 0
 one high frequency pole, 2
180 80.0
phase
gain
90.0 40.0
in degreess
vdbout in db(volts))
0°
1
0 0 2
0 dB
T s 
ph_vout
 s  s 
  1  
-90.0 -40.0
-1
1  0    2 
-2
-180 -80 0
-80.0
10 100 1k 10k 100k

Closed-loop gain study
 Linking the open-loop
open loop phase margin to the closed-loop
closed loop Q
50 90
-106°
1
-1
20 log T ( i  )  0 arg( T ( i  ) ) 
360
135
m
-2 2  74°°
fc fc
f 0  300 Hz
f 2  1000 Hz 50 180
3 4
10 100 1 10
3
1 10
4
10 100 1 10 1 10
 
2  2 
1 T s 1
T s 
Closed loop

 s  s  1 T s s2 s
  1    1
 0    2  02 0
1

1 solve 0
Q r  02
s2 s s2 s 2
 1  1
02 0 r 2 r Q

Closed-loop gain study
 if we plot the closed-loop
closed loop expression: T s 1
 2 > 0, low Q, no peaking 
1 T s s2 s
 2 < 0, high Q, peaking  1
r 2 r Q
0
f 0  300 Hz
 T ( i  ) 
0
 T ( i  )   360 f 2  1000 Hz
20 log  arg  90
 1 T ( i  )   1 T ( i  )  2  Q  0.55
 m  74
20 180
3 4 3 4
10 100 1 10 1 10 10 100 1 10 1 10 Open loop phase
Open-loop
 
2  2 
0
f 0  300 Hz
 T ( i  ) 
0
 T ( i  )   360 f 2  100 Hz
20 log  arg  90
 1 T ( i  )   1 T ( i  )  2 
Q  1.7
20 180
 m  32
3 4 3 4
10 100 1 10 1 10 10 100 1 10 1 10
  Closed-loop Q
2  2 

Linking m and Q
 an open-loop
open loop phase margin leads to a closed-loop
closed loop quality coeff.
coeff Q
 we have seen that Q affects the transient response (RLC filter)
 let us link the phase margin to the quality coefficient:
1. calculate the crossover frequency for which |T(s)| = 1
1
1 c 
2  1  4Q 4  1 
 jc   jc  2
  1  

 0   2 
2. substitute c into T(s), calculate its argument (phase margin)

 
2
arg T(s)@ fc = arctan  



1  4Q 4   1 
1  tan  m  cos  m 
2
4
3 extract the quality coefficient Q:
3. Q 
tan  m  sin  m 

We can now plot Q versus m
 a Q factor of 0 5 (critical response) implies a m of 76
0.5 76°
 a 45° m corresponds to a Q of 1.2: oscillatory response!
10
Q
7.5
1
 1 tan( ) 2 4 5
m
tan( )
2.5
0.5
0
0 25 50 75 100
76°
360

2 
Summary on the design criteria
 compensate the open-loop gain for a phase margin of 70°
70
 make sure the open-loop gain margin is better than 15 dB
 never accept a phase margin lower than 45° in worst case
5.12
PM = 10°
PM = 25°
PM = 45°
5.06
PM = 76°
out2#d in volts
vvout2#a, vout2, vout2#b, vo
Plot2
5.00 5
1
3
2
f  Cout , f c , I out  4.94
4.88
f  PM 
300u 900u 1.50m 2.10m 2.70m
time in seconds

Course agenda
 Poles and zeros
 Phase
Ph margin
i andd quality
li coefficient
ffi i
 Compensating
p g the converter
 Multi-output
 Input filter
p
 Conclusion
Dc-dc output impedance
 A dc-dc
d d conv. combinesbi an iinductor
d t andd a capacitor
it
 As f is swept, different elements dominate Zout
2 f0
Lout Rlf 20.0
Z0 2
 Rlf 
100u 10m 1  
Rlff Zout (dBΩ)
4 1 2 Vout  Z0 
0
Resr
1m I1 f (Hz)
vdbout in db(volts)
AC = 1
plot1
-20.0
3 Coutt
Cout
1000uF -40.0
Lout
Rlf Resr
-60.0 2
A buck equivalent circuit 1 10 100 1k

Crossover region
10k 100k 1Meg
 1 
frequency in hertz
Z out   sLout  rLf  ||  Resr   To avoid stability issues,

 sCout 
fc >> f0

Closing the loop…
 Any
A circuit
i it can b
be represented
t db
by it
its Thé
Thévenin
i model
d l
 At high frequency, Cout impedance dominates
 Once in closed-loop,
closed-loop Zout goes down as T(s) is high
Rth
Iout
Z out ,OL
Vout Vout
Vth
H(s)
d
1
Z out ,OL  G(s)
2 Cout f
1
Z out ,CL  Z out ,OL
1 T s
Closing the loop…
 At the
th crossover frequency
f Zout,CL  Zout,OL Let’s assess
« almost » :-)
100
|T(s)|
50.0
fc
out, vdberr in db(volts)
Plot1
0
vdbout#b, vdbo
|Zout,OL| 2
-50.0
5
3
|Zout,CL|| Zout,OL|
|Zout,CL|
-100
1 10 100 1k 10k 100k

frequency in hertz

Calculating the output impedance
 closed-loop output impedance is dominated by Cout
1 1
Z out ,CL 
2 f cCout 1  T  s 
 we have calculated the crossover value, |T(s)| = 1
c 
2  
1  4Q 4  1
2
b tit iinto
 substitue t 1/ (1+T( )) and
(1+T(s)) d extract
t t module
d l
1 1

1 T s  
2
 
 0  0 2
1 
  2     2

2
 2  1   2    2  1  2 
  2    2 
Calculating the output impedance
 Introduce the quality factor coefficient
1 1
0 
Q 1 T s 2Q 2  1  1  4Q 4
2 2Q
1  1  4Q 4  
1  4Q 4  1 2 2
cos  m 
 Now replace Q by its definition Q
sin  m 
1 1

1 T s 1  1  cos 2  m   
 2 cos  m      cos  m   1  1  cos  m  
2 2
cos  m    sin 2
 m 
 
1 1 I like this
 Simplify 1 T s

2  2 cos  m 
method!

Calculating the output impedance – 2nd way
T(s)
jc
~ T(s) 1  T  jc  e j
1V
  arg T  jc  j
90°

fc
0° sin  

arg T  f c  180  arg T  f c    m 180
180°
 0°
cos  
arg T  f c   m  180
m
-180° T  jc 
T  j  e j  T  j   cos    j sin   
 At crossover, |T(jc)|=1 1
T  jc  e j  cos  m  180   j sin  m  180  J’Euler de

quoi moi?
T  jc  e j   cos  m   j sin  m 

Calculating the output impedance – 2nd way
1
 Calculate
1  T  fc 
1 1 1
 
1  T  f c  1  cos  m   j sin  m  2  2 cos  m 
2
1 1 1
Z out ,CL  
2 f cCout 2  2 cos m  2 f cCout
1
1  T  fc  1 Ok if
All these
calculations 1
for this??  RESR
0
20 40 60 80
2 f cCout
360

2 

An example with a buck
 Let’s assume an output capacitor of 1 mF
 The spec states a 80-mV undershoot for a 2-A step
H
 How l
to select h crossover ffrequency?
the ?
I out I out
Vout  fc 
2 f c Cout Vout Cout 2
2 1
fc   4 kHz Z Cout @ 4 kHz   40 m
80m 1m  2 2  4k 1m
Select a 1000-µF capacitor featuring less than 40-mΩ ESR

Setting the right crossover frequency
 Compensate the converter for a 4-kHz fc
Compensated open-loop gain

180 80.0
Buck operated in voltage-mode
90.0 40.0
phase
m = 70°
n degrees
vdberr in db(volts)
ot1
0 0
Plo
ph_verr in
fc
-90.0 -40.0 3
gain
-180 -80.0
10 100 1k 10k 100k

frequency in hertz
4 kHz

Step load the output
rLf L1
10m 100u vout
a c Vout
7 3
d R10
p
 the
th lload
d varies
i
11
PWM switch VM 1m
Vin
X3 I1
10
PWMVM 16
L = 100u
Fs = 100k C5
1mF
from 100 mA to 2.1 A
H(s)
12 vout
GAIN
X1 C2
GAIN {C2}
K = 0.5 PWM R7
{R3}
gain C1
{C1}
R2
{R2}
Rupper
10k 8
13 C3
5 {C3}
Verr Rlower
X2 V2 10k
AMPSIMP 2.5
G(s)

Measure the obtained undershoot
5.00
5
I
V  40m
4.98
cos m 
2  2 cos
vout in volts
2
Plot1
4.96
V  40m   70 mV
1 14
1.14
4.94
Just
70 mV p
perfect,,
ma biche!
4.92
1.61m 2.42m 3.23m 4.05m 4.86m

time in seconds

Is my capacitor a real capacitor?
 A capacitor is made of parasitic elements
1
C = 100 µµF
70.0
RESR = 100 m
LESL = 100 nH
LESL 50.0
vdb3 in db(volts)
C
Plot1
30 0
30.0
RESR LESL
C
Z = -20 dB Z = - 4 dB
Z = 24.03 dB 
RESR @ 40.85 kHz @ 1 MHz
10.0 @ 100 Hz
-10.0
1 10 100 1k 10k 100k 1Meg 10Meg

frequency in hertz

How these elements affect the undershoot?
 The
Th output current slopel affects
ff the
h undershoot
d h
 If slope is steep, stray elements dominate the answer
Rupper
Power stage
Vout(t)
()
H(s)
RESR
Iout(t)
Zf IC(t)
Vin LESL
Cout C

How these elements affect the undershoot?
 Because
B off bandwidth
b d id h lilimits,
i RESR and
d LESL play
l alone
l
Vout
R1
100m
C1
SI
I 100u Iout
L1
100n
t
t
0 t 1
Vout  RESR S I t  LESL S I   S I t  dt
t C0
I out (t )  I  SI t SI t 2
t Vout  RESR S I t  LESL S I 
2C
The capacitor contribution is small…
200m
100m
esr in volts
plot1
-100m
V = 100 mV VESR(t)
-200m
5.04
5.00
c in volts
plot2
4.96
V = 100 mV
4.92
VC(t)
4.88
SI = 5 A/µs
µ
480m
240m
esl in voltss
plot3
-240m V = 505 mV VESL(t)

-480m
5.40
vout2 in volts
5.20
plot4
5.00
4.80
4.60
V = 606 mV Voutt(t)
1.92m 2.05m 2.18m 2.32m 2.45m
ti i d

Course agenda
 Poles and zeros
 Phase
Ph margin
i andd quality
li coefficient
ffi i
 Compensating
p g the converter
 Multi-output
 Input filter
p
 Conclusion
Compensating the converter
 Fix the current error with a proportional term (P)
 The proportional gain gives fast reaction time but also overshoot
 Fix the long-term static error with an integral term (I)
 The integral term cancels the static error but slows down the response
 Fix the immediate error by observing the slope with a derivative term (D)
 The derivative term decreases overshoot but slows down the response
d  t 
v  t   k p  t   ki    t   dt  kd
k p  t  dt
(t)
Vref(t)
+
ki    t   dt + H(s) Vout(t)

Power stage
d t  Buck, boost…
kd
dt

What kind of compensation
d  t 
v  t   k p  t   ki    t   dt  kd Time-domain equation
dt
In the Laplace domain:
d d  t  1  s
  s s  dt    t   dt 
s
s
dt dt s
 k 
v  s     s   k p  i  kd s 
 s 
The transfer function of a classical compensator
p G(s):
( )
 s  s   s 
 1   1  ...
  1  
Zeros in the denominator
   create the derivative term
G  s   G0  z1  z2   zn 
 s  s   s  Poles in the numerator

s p0 1  1 ... 1 
  p1    p 2    pn  create the integral term
    
Mid-band to Pole at the origin
crossover at fc for high dc gain
How do we stabilize the converter?
1. Select the crossover frequency fc (assume 4 kHz)
2. Provide a high dc gain for a low static error and good input rejection
3. Shoot for a 70° phase margin at fc
4. Evaluate the needed phase boost at fc to meet (3)
5. Shape the G(s) path to comply with 1, 2 and 3
Open-loop Bode plot of the power stage, H(s) Asc ,OL  s 

40 0
40.0 180
Asc ,CL  s  
1 T  s
20.0 90.0 Gain
ph_voutin degrees
vdboutin db(volts)
Plot1
0 0
Phase Arg H(s) @ fc
-20.0 -90.0
-40.0
40 0 -180
180
|H(s)| @ fc
1
10 100 1k 10k 100k
frequency in hertz

First, provide mid-band gain at crossover
1.
1 Adjust G(s) to boost the gain by +21 dB at crossover
40.0 180
Push the
20.0 90.0 Gain
gain up. 0 dB@fc
ph_voutin degrees
vdboutin db(volts)
Plot1
0 0
Phase Tailor G(s) to

exhibit a gain
|H(s)|= -21 dB of +21 dB@ fc.
-20.0 -90.0
Arg H(s)= -175°
-40.0 -180
10 100 1k 10k 100k

frequency in hertz 4 kHz

Second, provide high gain in dc
2.
2 High dc gain lowers static error and brings good input rejection
High
dc gain
Vin
40.0 180 Gain
-1
20.0 90.0 1
sRC
ph_voutin degrees
vdboutin db(volts)
Plot1
0 0
Phase
-20.0 -90.0 Voutt
-40.0 -180
Pole at the origin

g !
10 100 1k 10k 100k
frequency in hertz

2.
2 An integrator provides a high dc gain but rotates by -270
270°
C1
100n
60.0 360
60 dB 1 2
E1
R1 1k
10k
4 Vout
30.0 180
-20 dB per decade V1
AC = 1
slope -1
1
ut in db(volts)
n unknown
Plot1
0 0
p in
vdbou
-180° by inverting
op amp
-30.0 -180
90 by pole
-90° 8
at the origin 10
-60.0 -360
100m 1 10 100 1k 10k 100k

frequency in hertz

2.
2 An integrator provides a high dc gain but rotates by -270
270°
C1
100n
Vout  s  1 1 1
G s     p1  1 2
Vin  s  sR1C1 s R1C1 E1

1k
 p1 R1
10k
 j 
arg G  j    arg 
Vout
  p 0 
4
  V1
AC = 1
   
lim arg G  s   lim  arctan   
s      p0  2
 
Total phase lag brought by an origin pole is -3/2 or -270°
t
All compensators (1 2 or 3) feature
(1, f t i i pole
an origin l

Third, evaluate the phase boost at fc
0
ph_vout#a in degrees
-40.0
arg H(s) at 4 kHz
Plot1
-80.0
arg H(s)
-120
arg
g H(s)
( ) -175°
+
18
-160
160
0
arg G(s)
-90.0
arg G(s) -113°
p in unknoown
Plot3
-180
+155° 11
-270
-360 Phase boost at fc
m = 70°
arg H(s)G(s)
0
-90.0
p iin unknown
Plot2
-180
-270
m
-360 1
10 100 1k 10k 100k

frequency in hertz
arg H  f c   270  BOOST  m  360

BOOST  m  arg H  f c   90  70  175  90  155
How do we boost the phase at fc?
 The type
Th t 1 configuration
fi ti
 No phase boost, pure integral term
 Permanent phase lag of -270°270
 Ok if argH(fc) < -45° for a m of 45°
C1
10n
1 2
Vout  s  1 1
E1 G s   
R1 10k Vin  s  sR
R1C1 s
10k
1 0
4 Vout
 p1 
V1
R1C1
AC = 1
1p
pole at the origin
g
Type 1

yb  ya 0  log
l G
360 60.0
Gain   1
G(s) C1 xb  xa log f 0  log f c
355p
f 0 44.5
44 5
1 2
R1
E1
10k -1 G   11.1
180 30.0
4
10k
Vout fc 4
V1
AC = 1
Adjust f0 to
gett G att fc
vdbout in db(volts)
21 dB
p in unknown
plot1
0 0
log G 28
f0 = 44.5
44 5 kHz
v
-180 -30.0
log f
-270°
270°
27
Phase
-360 -60.0 Arg G(s)
10 100 1k 10k 100k

Type 1
frequency in hertz 4 kHz
 Th
The type
t 2 configuration
fi ti
 Phase boost up to 90°
 Ok if argH(fc) < -90°
90
C2
1  sR1C1
G s  
62pF
  C C 
sR1  C1  C2  1  sR2  1 2  
C1 R2
2nF 116k
1 2 4   C1  C2  
E1
R1 10k If C2 << C1
10k
1 1 1
3 Vout
 po   p1   z1 
V1
R1C1 R2C2 R1C1
AC = 1
1 pole at the origin
1 zero
1 pole
Type 2

Pole/zero placement and boost at fc
 Phase
Ph b
boostt appears b
between
t th
the zero and
d th
the pole
l
     
1  j  1  j 
 z1  z1  G  a  jb
G  j    arg G  j   boost  arg  b
      arg G  arctan  
1  j  1  j  a
  p1    p1 
 f   f 
arg G  f   arctan    arctan  
 f z1   f p1 
Assume 1 zero placed at 705 Hz, 1 pole at 22 kHz and a 4 kHz crossover:
 4k   4k 
arg G  4 kHz   arctan    arctan    80  10.3  70
 705   22k 
Type 2

G100 Hz = 38 dB
Gain
40.0 360
G(s)
fz = 705 Hz
C2
62pF fp = 22 kHz
20.0 180
C1 R2
2nF 116k
1 2 4
Gain at
fc = 21 dB 33
vdbout in db(volts)
E1
p in unknown
R1 10k
plot1
0 0 10k
3 Vout
V1
v
AC = 1
-20.0 -180
32
Phase boost
-40.0 -360
at fc = 71°
Phase
Arg G(s)
10 100 1k 10k 100k
frequency in hertz 4 kHz Type 2

 Th
The type
t 3 configuration
fi ti
 Phase boost up to 180°
 Ok if argH(fc) < -180°
180
sR2C1  1 sC3  R1  R3   1
G (s)  
 C C   sR R3C3  1
C2
350pF sR1  C1  C2  1  sR2 1 2 
 C1  C2 
C1 R2
11nF
1 2
20k
4
If C2 << C1 and R3 << R1
C3 R3 E1 1 1 1
22nF 321 10k  z1  z2   po 
5 R2C1 R1C3 R1C1
Vout
3
1 1
R1
10k  p1   p2 
V1 R3C3 R2C2
AC = 1
1 pole at the origin

2 zeros
2 poles
Type 3

Pole/zero placement and boost at fc
 Phase
Ph b
boostt appears b
between
t th
the zero and
d th
the pole
l
           
1  j  1  j  1  j  1  j 
 z1  z 2   z1  z 2 
G ( j )    arg G  j   boost  arg  
         
 1  j  1  j  1  j  1  j 
  p1    p 2   p1    p 2 

 f   f   f   f 
arg G  f   arctan    arctan    arctan    arctan  
 f z1   fz2   f p1   f p2 
Assume 2 zeros p
placed at 500 Hz, 2 p
poles at 50 kHz and a 4 kHz crossover:
 4k   4k 
arg G  4 kHz   2 arctan    2 arctan    166  4.6  161
 500   50k
50k 
Type 3

Gain
G100 Hz = 20 dB fp1,fp2 = 50 kHz
40.0 360
G(s)
1
20.0 180
C2
fz1, fz2, = 500 Hz
350pF
Gain at
vvdbout in db(volts)
C1 R2
p in unknown
11nF 20k fc = 21 dB
Plot1
1 2 4
0 0
C3 R3 E1
22nF 321 10k
5
3 Vout
R1
10k
V1
AC = 1
-20.0 -180 Phase boost
at fc = 158° 3
-270°
-40.0 -360
Phase
Arg G(s)
10 100 1k 10k 100k
frequency in hertz 4 kHz Type 3

Finally, we test the open-loop gain
5. Given the necessary boost of 155°,
155 , we select a type
type-3
3 amplifier
6. A SPICE simulation can give us the whole picture!
rLf L1
10m 100u vout
a c Vout
7 3
d R10
11
PWM switch VM p 1m
Vin R11
10 X3
PWMVM 16 1
L = 100u
Fs = 100k C5
1mF
Buck stage
12 vout
GAIN
X1 C2
GAIN
K = 0.5
{C2}
R7 1 pole at the origin
{R3}
2
C1
{C1}
R2
{R2}
Rupper
10k 8
2 zeros at 500 Hz
13
5
C3
{C3} 2 poles at 50 kHz
LoL
1kH
1
Verr Rlower
CoL X2 10k
1kF V2
AMPSIMP 2.5
9
Vstim
AC = 1
Type 3

Finally, we test the open-loop gain
An ac simulation gives us the open-loop
open loop Bode plot
Gain
80.0 360
T(s)
40.0 180
fc = 4 kHz
berr in db(volts)
p in unknown
plot1
0 0
vdb
-40.0 -180 15
m = 70°
-80.0 -360 Phase 14
Arg
g T(s)
( )
10 100 1k 10k 100k
frequency in hertz

What transient response?
 The crossover frequency is constant, but zeros position is changed
1. Double zero 200 Hz, m = 80°

Iout = 2 A
5.08
3
3. Hz m = 60
Double zero 700 Hz, 60°
5.04
vout#a, voutt, vout#b, vout#c in voltss
4
Plot1
5.00
18
17
19
20
3
2
4.96 1
4.92
fc = 4 kHz
Almost unchanged
1.00m 2.00m 3.00m 4.00m 5.00m
time in seconds

Zeros become poles??
 T(s) is shaped by introducing zeros/poles in G(s)
 In closed-loop conditions, the zeros of G(s) become poles…
compensator
converter
Vref(s) + G(s) k H(s) Vout(s)
–
NG (s) N H (s)
G(s)  H (s) 
DG ( s ) DH ( s )
NG (s) N H ( s)
k
Vout ( s) kG ( s) H ( s ) Vout ( s ) DG ( s ) DH ( s ) kN G ( s ) N H ( s )
  
Vref ( s) 1  kG ( s) H ( s) Vref ( s ) 1  k N G ( s ) N H ( s ) DG ( s ) DH ( s )  kN G ( s ) N H ( s )
DG ( s ) DH ( s ) For k >> 1
zeros of G(s) appear
in denominator as poles
Pushing the zeros towards low frequency slows down the response!

If we roll-off the BW in the low frequencies?
 Could we avoid the resonance with a 10 Hz crossover point?
40.0 180
Open-loop Bode plot of the power stage, H(s)
|H(s)| @ fc
14 dB
20.0 90.0
ph_voutin degrees
vdboutin db(volts)
Gain
Plot1
0 0
Phase
-20.0 -90.0
-40.0 -180
Arg H(s) @ fc
0
0°
1
10 100 1k 10k 100k
frequency in hertz
 We have almost no phase rotation at 10 Hz

Hz, a type 1 could do??

If we roll-off the BW in the low frequencies?
 A simple SPICE ac simulation gives us the open-loop
open loop gain
rLf L1
10m 100u vout
a c vout
7 3
parameters
Vout=5V d R10
Rupper=10k
11
PWM switch VM p 1m
Vin R11
fc=10 10 X3
Gfc=14 PWMVM 16 1 ac 1
Vcp(t)
L = 100u
Fs = 100k C5
G=10^(-Gfc/20) 1m
pi=3.14159
C1 1/(2 pi fc G Rupper)
C1=1/(2*pi*fc*G*Rupper)
fp0=1/(2*pi*C1*Rupper) vout
8
GAIN
X1
GAIN
K = 0.5
-6 dB
PWM gain C1
{C1}
Rupper
2
10k
LoL
1kH
Vin
Verr Rlower
CoL X2 10k
1kF V2
AMPSIMP 2.5
9
Vstim
AC = 1
 An ac current source sweeps the output impedance

The open-loop gain looks good…
 The type 1 gives our 0 dB crossover frequency
vout
30
29
|T(s)| C1
8uF Rupper
40.0 180
10k
3
20.0 90.0
2
arg T(s) Rlower

m = 90° V2
2.5
10k
ph_verr in degrees
vdberr in db(volts)
Plot1
0 0
m = 0°
-20.0 -90.0
fc = 10 Hz -24 dB
-40.0 -180 f0
100m 1 10 101 1k 10k

frequency in hertz
 We have plenty of phase margin, a slow system, ok…

Oh no, it is ringing!
 The load step reveals a ringing ac output
Iout = 2 A
5.60
1/f0
5.20
Vout(t)
48
47
v(7), vout in volts
Plot1
4.80
4.40
Vcp(t)
4.00
Vcp(t) is first order

5 50m
5.50m 16 5m
16.5m 27 5m
27.5m 38 5m
38.5m 49 5m
49.5m
time in seconds
Munch

There is no gain to compensate the peaking!
 The gain, when the resonance occurs, is below 0 dB
20.0
f0 of output filter
0
b(volts)
vdboutt2, vdbout#a, vdbout in db
Gain action
Plot1
-20.0
Zout,OL
-40.0
Zout,CL 25
2
23
245
-60.0
Properly compensated, fc = 4 kHz
Zout,CL
1 10 100 1k 10k 100k 1Meg
frequency in hertz
 The system does not reduce the Q at the resonance frequency

The RLC network rings alone…
 H1 is
i stable
t bl per B
Bode
d analysis,
l i b butt H2 is
i outt off the
th loop…
l
Vin Vout(s)
H1(s) H2(s)
T(s)
d(s)
dc
f << fc Loose coupling
in ac…
G(s) ac
f >> fc
 The dc is fed back via the loop but not the ac…
“Fast Analytical Techniques for Electrical and Electronic Circuits”
Vatché Vorpérian, Cambridge Press, 2002

The crossover must be above f0
 There still must be gain at the resonance to damp the filter
 fc should be far from fo to reduce phase stress at resonance
40.0 180
Open-loop Bode plot of the power stage
stage, H(s)
fo = 450 Hz
20.0 90.0 Gain

ph_voutin degrees
vdboutin db(volts)
Plot1
0 0
Phase
-20.0 -90.0
-40.0 -180
> 3fo
1
10 100 1k 10k 100k
frequency in hertz
Place fc at least three times above resonance!

CCM to DCM, the transfer function changes
 In DCM, the buck voltage-mode
voltage mode becomes a first order system
40.0 |H(s)|
db(volts)
20.0
dbout2#a, vdbout2 in d
CCM
Plot1
-1
-20.0 -2
vd
DCM
-40.0
11
8
180 arg H(s)

ut2 in degrees
90.0
Plott3
vphout2, ph_vou
DCM
-90.0
10
9
-180 CCM
1 10 100 1k 10k 100k
frequency in hertz

CCM to DCM, the transfer function changes
 In DCM, the compensated system becomes slower than in CCM
80.0 |T(s)|
bout, vdbout#a in db(volts)
40.0
fc
CCM
Plot1
fc
-40.0 DCM
vdb
9
7
-80.0
180 arg T(s)

DCM
#a in degrees
90.0 CCM
ph_vout, ph_vout#
Plot2
2
0 10
8
m m
Potential
-90.0
stability 45° 70°
Issue.
-180
180
1 10 100 1k 10k 100k 1Meg

f i h t

The DCM response is slower than in CCM
 A 100 Hz crossover in DCM versus a 4 kHz crossover in CCM
5.006
Iout = 100 mA
5.002 Vout = 3 mV
voutt2#a, vout2 in volts
Plot1
4.998 25
24
CCM
4.994
DCM
Vout = 10 mV
4.990
2.89m 8.68m 14.5m 20.3m 26.0m

time in seconds

General methods for compensation
 For
F a buck
b k in
i CCM voltage-mode:
lt d
• Place a double zero at the LCout resonance
• If the
th ESR zero fz is
i bbelow
l fc, putt a pole
l fp1=ffz
• If the ESR zero fz is above fc, put a pole fp1=Fsw/2
• Put
P a second d pole
l fp2 at fp2=FFsw/2
 s 
 1  
This zero is brought by
Vout ( s ) Vin   z1 
the output capacitor ESR
 Kc 2
Verr ( s ) V peak s  s 
1  
PWM gain
Q 0   0 
Rload 1 1 1
Kc   z1  0  Q
rCf  rLf || Rload
rLf  Rload rCf C R  rCf Zo
LC 
R  rLf rLf  Rload Zo Type 3

 For
F a buck
b k in
i DCM voltage-mode:
lt d
• Place a zero below the selected cross-over
• If the
th ESR zero fz is
i bbelow
l fc, putt a pole
l fp1=ffz
• Put
P a second d pole
l fp2 at fp2=FFsw/2
 s 
K1 1  
This zero is brought
g by y
Vout ( s ) Vin   z1 
the output capacitor ESR

Verr ( s ) V peak  s 
1  
  p1 
PWM gain
1 2M 1 2 1  M  1  M K
2L
M
Vout
 z1   p1  K1 
Rload Tsw Vin
rCf C 1  M Rload C 2M K
Type 2

Course agenda
 Poles and zeros
 Phase
Ph margin
i andd quality
li coefficient
ffi i
 Compensating
p g the converter
 Multi-output
 Input filter
p
 Conclusion
The buck in current-mode
 A CCM current
current-mode
mode converter acts as a third order system
 It looks like a first order system at fc << Fsw/2
 No LC peaking anymore!
 But a subharmonic peaking at Fsw/2 now appears!
2
3
180 40.0
90.0 20.0
|H(s)|
ph_vout2 in degrees
vdbout2 in db(volts)
Subharmonic
Plot1
0 0
Arg H(s) -1 peaking now!

-90.0 -20.0
-180 -40.0
-3
1 10 100 1k 10k 100k 1Meg
f i h t
Fsw/2
CCM operation, current instabilities
IL
I peak  a  S1t
IL(0) Ipeak  IL(Tsw) Verr
Ri b  I peak  S 2 t
S1 S2 Solving
a t
b IL(Tsw) I peak  a I peak  b

IL(0) S1 S2
IL(0) I L ((0)) I L (Tsw )
 IL(Tsw) 
t S1 S2
S2 d
t 
S1 d '
dTsw d’Tsw
Tsw n
 d 
I L (nTsw )  I L (0)   
 d '

CCM operation, current instabilities
IL
clock
Ipeak
IL(0)
IL(0)
Duty-cycle < 50%
t
Asymptotically stable Perturbation has gone…
IL
clock
Ipeak
Duty clamp
IL(0)
Duty-cycle > 50%
IL(0)
t
Asymptotically unstable

CCM operation, curing current instabilities
IL n
 Sa 
 1 S 
Verr I L (nTsw )  I L (0)   2
  I L (0)  a 
n
b c Ri  d '  Sa 
 d S 2 
Sa
S1 Must stay
IL(0)
S2 IL((Tsw) below 1
IL(0) Sa
1
t IL(Tsw)
Up to S2
d = 100% 1
S
0 a
t S2
dTsw d’T
d Tsw
Tsw
Inject ramp compensation Sa  50% S 2

A buck CCM in current-mode
 A SPICE model can predict subharmonic instabilities
dc
505mV
L1 R6
duty-cycle 100u 10m vout
5.00V
a c Vout2
2 12
5 05V
5.05V
5.05V
512mV
10.0V vc Resr
3 5
PWM switch CM p 1m
Vin Rload
10 PWMCM 1
16 5.00V
AC = 0 X4 X2
GAIN L = 100u Cout
GAIN
K = 0.3333
0 3333 Fs = 100k 1m
N
1.54V
Ri = 0.1
Se = 2.5k Ramp
7
Compensation Se
vout
parameters
Vout=5V
C1
Rupper=10k {C1} Rupper
fc=10 {Rupper}
Gfc=5.5
9 2.50V
LoL
G=10^(-Gfc/20)
( ) 1kH
1 54V
1.54V
pi=3.14159 4
2.50V
14 Rlower
CoL Verr
V12 10k
fp0=1/(2*pi*C1*Rupper) 1kF X3
AMPSIMP 2.5
0V
10 VHIGH = 3
V1
AC = 1

Injecting ramp damps the double pole
 Too much ramp turns the converter into voltage-mode!
voltage mode!
6
5
4
3
1
Sn
Mc  1
Se
40.0
Fsw/2
dbout2 in db(volts)
0
out2#2, vdbout2#1, vd
Mc=1
f0
Plot1
-40.0
Voltage-mode
g Mc=1.5
vdboutt2#4, vdbout2#3, vdbo
response
Mc=30
-80.0
Mc=10
10 Mc=2
Vout  s 
Vc  s 
-120
1 10 100 1k 10k 100k 1Meg

f i h t

The right way to inject ramp compensation
1 8 Low
2 7 i
impedance
d
4 3 6
4 5
NCP1200 1
X1
NCP1200
R1 D1 3
18k 1N4148
2
RComp
R4
1k C1
1nF Rsense
S S = generated ramp
Rcomp  1k  S’ = reflected sensed ramp
S ' M M = amount of ramp, 0.5-0.75

Damping the CCM buck subharmonic poles
 Evaluate the inductor downslope:

Vout 5 x1 L
Soff    50 kA s Vout
L 100u
 Select 50% of itit, 25 kA/s

R
x2 C
 Reflect over the sense resistor
Soff '  Soff Rsense  25  0.1  2.5 kV s
 Feed
F d the
th model,
d l run th
the simulator…
i l t

Would the type 1 give a stable response?
6
5
40.0 180
Looks like a simple integrator could roll-off the gain…
20.0 90.0
|H(s)|
ph_vout2 in degrees
s
plot1
0 0
Arg H(s)
Peaking is gone
thanks to ramp
-20.0 -90.0
compensation
-40.0 -180
1 10 100 1k 10k 100k 1Meg

g
frequency in hertz

All looks ok, no hidden peaking…
8
7
vout
180 40.0
C1
5uF Rupper
10k
arg T(s) 1
90.0 20.0
m = 90
90° 2
vdberr in db(volts)
Rlower
vpherr in degrees
|T(s)| 10k
V2
Plot1
0 0 2.5
-90.0 -20.0
fc = 10 Hz
-180 -40.0
1 10 100 1k 10k 100k 1Meg

frequency in hertz

The transient step is slow, but stable…
5.00 14
Iout = 2 A
4.00
vout2 in volts
plot1
3.00
2.00
1.00
V = 3.8 V
Vout(t)
10.0m 30.0m 50.0m 70.0m 90.0m
ti
time in
i seconds
d

The open-loop impedance of a CM converter
 The current-mode
current mode output impedance depends on the ramp level
0 10
13
11
9
8
Soff
-5
5.00
00
V lt d
Voltage-mode
7Soff peaking
#d in db(volts)
Vout
1
30Soff -15.0
Zoutt
Rload
#a, vdbout2#b, vdbout2#
Cout
Z out  s   Rload || Cout f0

Plot1
-25.0
ut2, vdbout2#c, vdbout2#
-35.0
dBΩ
vdbou
-45.0
1 10 100 1k 10k 100k 1Meg

frequency in hertz

The closed-loop output impedance of CM and VM
 Zout CM is naturally larger than Zout VM
Open-loop output impedances

40 0
40.0
20.0
vdbout2, vvdbout2#a in db(volts)
 Rload
Plot1
Current mode
-20.0  Cout dBΩ
 rLf
-40.0
Voltage mode
3
5
1 10 100 1k 10k 100k 1Meg
f i h t

 In closed
closed-loop,
loop Zout in VM is still smaller
smaller…
14
Closed-loop output impedances

80.0
dBΩ
40.0
vdbout2#a, vdbout2 in db(volts)
Plot1
fc
-40.0
Current mode
17
-80.0
Voltage mode
1 10 100 1k 10k 100k 1Meg
frequency in hertz

 The dc gain in VM is smaller because of zeros
80.0
Current mode Open-loop gains
fc
bout, vdberr in db(volts)
40.0
Voltage mode
Plot1
-40
40.0
0
vdb
18
-80.0
|T(s)|
21
180 arg T(s)

ph_vout, ph_verrr in degrees
90.0
Plot2
2
0 24
Same shape in
-90.0
the vicinity of fc 22
-180
1 10 100 1k 10k 100k 1Meg

frequency in hertz

With similar crossover frequency…
 The voltage mode is slightly slower than current
current-mode
mode
Output response to a 2 A step
5.08
5.04
out2#a in volts
Current mode
Plot1
5.00 29
31
vout2, vo
4.96
Voltage mode
4.92
1.00m 3.00m 5.00m 7.00m 9.00m

time in seconds

Transition from CCM to DCM in current-mode
 Current-mode remains a 1st order system in both cases
60.0
|H(s)|
out2#a, vdbout2 in db(vvolts)
30.0
Plot1
0
-1
DCM CCM
-30.0 -1
vdbo
-60.0
17
15
180
arg H(s)
ut2 in degrees
90.0
CCM
ph_vout2#a, ph_vou
Plot2
-90.0 DCM
18
16
180
-180
10 100 1k 10k 100k 1Meg

frequency in hertz

Transition from CCM to DCM in current-mode
 Once compensated, fc still changes but m is still ok
80.0
dberr#a, vdberr in db((volts)
40.0
fc
DCM
Plot1
fc
-40.0 CCM
vd
-80.0 |T(s)| 19
21
180
ph_verr#a, ph_verrr in degrees
90.0
CCM
Plot2
0
DCM 20
m m 22
-90.0
80° 60°
-180
arg T(s)
10 100 1k 10k 100k 1Meg
frequency in hertz

The DCM response is not too much affected
 The step response is almost unchanged…
5.004
Iout = 100 mA
5.002
#b, vout2 in volts
Plot1
5.000 28
26
vout2#
CCM
4.998
Vout = 4 mV
4.996
DCM
500u 1.50m 2.50m 3.50m 4.50m

time in seconds

CM brings superior results in ripple rejection
 The response to a step input is better in current
current-mode
mode
 The controller fixes the peak current whatever Vin
 In VM,, the output
p must first change
g before duty y cycle
y is adjusted
j
5.40
Comp. 1 Voltage-mode Compensation 1:
fz1 = 300 Hz, fz2 = 1 kHz
out2 in volts
5.20
fp1,fp2 = 26 kHz
Polarity
P l it iis Comp. 2
vout2#b, vout2#a, vo
Plot1
5.00 4
3
5
positive Compensation 2
4.80
300 mV 100 mV fz1, fz2 = 2.2 kHz
fp1,fp2 = 40 kHz
20 V 4.60 Vin = 10 V
Vin(t)
10 V 5.0004
Polarity is 5.0000
5
vout2#b in volts
negative
Plot2
4.9996
4.9992  2 mV
4 9988
4.9988 Current mode
Current-mode
500u 1.50m 2.50m 3.50m 4.50m
time in seconds

 For
F a buck
b k in
i DCM/CCM current-mode: t d
• Place a zero below the selected cross-over
• If the
th ESR zero fz is
i bbelow
l fc, putt a pole
l fp1=ffz
• Put
P a second d pole
l fp2 at fp2=FFsw/2
Vout ( s ) R 1
 Fp ( s ) Fh ( s ) CCM
Verr ( s ) Ri 1  RTsw m D ' 0.5
L
 c 
s 1
1  z1 
 z1 rCf C
Fp ( s )  1 Double pole brought
s 1 T Fh ( s ) 
1  p1   sw  mc D ' 0.5 by subharmonic
 p1 Rload C LC s s2 oscillations.
1 
 nQp  n 2
Type 2

Course agenda
 Poles and zeros
 Phase
Ph margin
i andd quality
li coefficient
ffi i
 Compensating
p g the converter
 Multi-output
 Input filter
p
 Conclusion
The k factor in an automated type 2
 Poles
P l and d zeros b boostt th
the phase
h att th
the crossover ffreq.
 How to place poles/zeros for a precise boost at fc?
 Use
U ththe k ffactor
t technique
t h i introduced
i t d db by D
Dean V
Venablebl
From a 1 zero/ 1 pole compensation circuit, we have:
 fc  If only I had
 1     
listened
fz0   arctan  f c   arctan  f c 
during these
arg T ( f c )   boost  arg  classes!
 fc   
 fz0   f p0 
 1 f p0 
 
If we place one pole at kfc and one zero at fc/k, we have:
1
boost  arctan  k   arctan  
k
If we remember our past trigonometric classes (I don’t!):
1
arctan  x   arctan    90
 boost 
k  tan   45 
 x  2  Van Gogh

 Suppose
S we have
h th
the ffollowing
ll i specs:
 fc = 1 kHz, |H(1kHz)|= 10 dB
 Arg
A H(1kHz)
H(1kH ) = -100°
100°
 m = 70° BOOST  m  arg H  f c   90  70  100  90  80
 For
F a 80° phase
h b
boost, select
l a type 2
 Calculate k
 80 
k  tan   45   11.4
 2 
 Place a zero at 1k/11
1k/11.4
4 = 90 Hz
 Place a pole at 1k×11.4 = 11.4 kHz
 Adjust mid-gain
mid gain to reach 10 dB@1 kHz
Type 2

 Calculate
C l l t ththe elements
l t ffor a ttype 2 amplifier
lifi
1 1
C2    442 pF
2 f c GkRupper 6.28  1k  3.16  11.4  10k
   
Vout
C1  C2 k  1  442 p  11.4  1  57 nF
2 2
C2
k 11.4 Rupper
R2    31.8 k  10k
2 f c C1 66.28
28  1k  57 n 1
C1
2
R2
Gf c
G  10 20
 1010 20  3.16 4 Vout
6
Rlower
10k Vref
D. Venable, “The k-factor: a new mathematical tool for stability analysis and synthesis”,
proceedings of Powercon 10, 1983, pp. 1-12
Type 2

 Macros
M can be
b used
d tto automate
t t the
th calculation
l l ti
parameters C2
{C2}
Rupper=10k
C1 R2
fc=1k {C1} {R2}
pm=70 1 3 2
Gfc=-10
Gfc 10
E1
pfc=-100 10k
R1
10k **************************
G=10^(-Gfc/20) 4 Vout
* DEFAULT PARAMETERS
*
boost=pm-(pfc)-90
boost pm (pfc) 90 **************************
pi=3.14159 V1 ***** mainckt

AC = 1
K=tan((boost/2+45)*pi/180) RUPPER = 1.000e+004
FC = 1.000e+003
C2=1/(2*pi*fc*G*k*Rupper) PM = 7.000e+001
C1=C2*(K^2-1)
C1 C2 (K 2 1) GFC = -1.000e+001
PFC = -1.000e+002
R2=k/(2*pi*fc*C1) G = 3.162e+000
BOOST = 8.000e+001
PI = 3.142e+000
K = 1.143e+001
fp=1/(2*pi*R2*C2) C2 = 4.403e-010
results
fz=1/(2*pi*R2*C1)
fz=1/(2 pi R2 C1) C1 = 5.709e-008
R2 = 3.187e+004
FP = 1.134e+004
FZ = 8.749e+001
Type 2

 The
Th numbers
b exactly
tl match
t h th
the predictions
di ti
180 20 0
20.0
arg G(s)
80°
2
90.0 10.0
|G(s)|
10 dB
ph_voutt in degrees
vdbout in db(volts)
Plot1
0 0
P
1
-90.0 -10.0
-180 -20.0
10 100
f
1k
i h t
10k 100k
Type 2

 Adjusting k modulates the resulting phase boost
20.0
k=1
10.0 k=2
gain
0
k = 10
-10.0
-20.0 1
3
5
7
19 17 15 13 11 9
270 k = 10
Boost = 90°
180
90.0 6
4
2
10
8
16
14
12
18
20
phase
0 k=1
90 0
-90.0
10 100 1k
frequency in hertz
10k 100k
Type 2

 2 poles
l and d 2 zeros give
i a hi
higher
h bboostt att crossover
 These 2 poles and 2 zeros are coincident
 How
H tto place
l th
them ffor a precise
i b boostt att fc?
From a 1 zero pair/ 1 pole pair compensation circuit, we have:
 2 
 1  fc  
  
fz0    fc   fc 
arg T ( f c )   boost  arg  2   2 arctan    2 arctan  
 fc   f
 z0  f
 p0 
 1   
 f p 0  
If we place two poles at k f c and two zeros at f c k ,we have:
  1 
boost  2 arctan

  k  arctan  
 k 
2
boost  2 arctan
  k   arctan  k   90  4 arctan  k   180   boost
k   tan
t 
  4

 45  


 Suppose
S we have
h th
the ffollowing
ll i specs:
 fc = 1 kHz, |H(1kHz)|= 15 dB
 Arg
A H(1kHz)
H(1kH ) = -140°
140°
 m = 70° BOOST  m  arg H  f c   90  70  140  90  120
 For
F a 120° phaseh b
boost, select
l a type 3
 Calculate k
2
  120 
k   tan   45    13.9
  4 
 Place a zero at 1k/3

1k/3.7
7 = 268 Hz
 Place a pole at 1k×3.7 = 3.7 kHz
 Adjust mid-gain
mid gain to reach 15 dB@1 kHz
Type 3

 Calculate
C l l t ththe elements
l t ffor a ttype 3 amplifier
lifi
1 1
C2    2.85
2 85 nF Vout
2 f c GR1 6.28  1k  5.6  10k
C1  C2  k  1  2.85n  13.9  1  36.7 nF C3 C2

Rupper
5
10k
k 13.9
R2    16.2 k  R3 1 2
R2
2 f c C1 6.28  1k  36.7 n C1
R1 10k
R3    775  Vout
k  1 13.9  1
4
6
Rlower
1 1
C3    55 nF 10k Vref
2 f c k R3 6.28 1k  13.9  775
Gf c
G  10 20
 1015 20  5.6
Type 3

Macros can be used to automate the calculation
parameters C2
{C2}
Rupper=10k
fc=1k
C1 R2
pm=70 {C1} {R2}
Gfc=-15
1 3 2
ps=-140 C3 R3 E1
{C3} {R3} **************************
10k * DEFAULT PARAMETERS
G=10^(-Gfc/20) 5 *
boost=pm-(ps)-90 4 Vout **************************
pi=3.14159 R1
***** mainckt
{Rupper}
( (( )p ))
K=(tan((boost/4+45)*pi/180))^2 V1
AC = 1 1 000 004
RUPPER = 1.000e+004
C2=1/(2*pi*fc*G*Rupper) FC = 1.000e+003
C1=C2*(K-1) PM = 7.000e+001
GFC = -1.500e+001
R2=sqrt(k)/(2*pi*fc*C1) PS = -1.400e+002
G = 5.623e+000
R3=Rupper/(k-1) BOOST = 1.200e+002
C3 1/(2*pi*fc*sqrt(k)*R3)
C3=1/(2*pi*fc*sqrt(k)*R3) PI = 3.142e+000
3 142 000
K = 1.393e+001
C2 = 2.830e-009
fp1=1/(2*pi*R2*C2) C1 = 3.659e-008
R2 = 1.623e+004
fp2=1/(2*pi*R3*C3) results R3 = 7.735e+002
C3 = 5.513e-008
fz1=1/(2*pi*R2*C1)
fz1 1/(2 pi R2 C1) FP1 = 3.464e+003
3 464 +003
fz2=1/(2*pi*Rupper*C3) FP2 = 3.732e+003
FZ1 = 2.680e+002
FZ2 = 2.887e+002

 The
Th numbers
b exactly
tl match
t h th
the predictions
di ti
180 40.0
arg G(s)
120°
90 0
90.0 20 0
20.0 2
|G(s)|
15 dB
ph_vout in degrees
vdbout in db(volts)
Plot1
0 0
P
-90.0 -20.0
-180 -40.0
10 100
f
1k
i h t
10k 100k
Type 3

 Adjusting k modulates the resulting phase boost
20.0
k=1
10.0
k=2
gain
0
k = 15
-10
10.0
0
k = 20 12
4
-20.0 1
3
5
7
19 17 15 13 11 9
k = 20
270
Boost = 125°
180
90.0 16
4
2
10
12
14
6
8
16
18
20
0
phase
k=1
90 0
-90.0
10 100 1k 10k 100k

frequency in hertz Type 3

Course agenda
 Poles and zeros
 Phase
Ph margin
i andd quality
li coefficient
ffi i
 Compensating
p g the converter
 Multi-output
 Input filter
p
 Conclusion
Is the k factor the panacea?
 The tool is a quick and easy means to stabilize the converter
 However, it solely focuses on the crossover frequency
 What is happening before and beyond fc?
 Th k factor
The f t technique
t h i is
i blind
bli d tto th
these effects.
ff t
 In resonant systems, conditional stability can occur.
360 40.0
|T(s)|
180 20.0
arg T(s)
ph_verr in degrees
vdbout in db(volts)
plot1
0 0
-180 -20.0
8
-360 -40.0
Isolated Ćuk converter

10 100 1k 10k 100k
f i h t

The k factor can lead to conditional stability
 In CCM voltage-mode
voltage-mode, conditional stability occurs
180 40.0
f0=1.2 kHz
|H(s)|
90.0 20.0 fc
ph_vout2 in degrees
k factor
Plot2
0 0
-90.0 -20.0
arg H(s) gives
12
-180 -40.0
11
fp1,2= 40 kHz
180 80.0 |T(s)| fz1,2= 2.3 kHz
m
90.0 40.0
degrees
vdberr in db(volts)
arg T(s)
85
85° 10
Plot11
ph_verr in d
0 0
-90.0 -40.0
-180 -80.0
Conditional stability
1 10 100 1k 10k 100k
frequency in hertz

Manual pole/zero placement is a solution
 Avoid coincident poles and zeros brought by k factor
 Place them wherever you wish to shape G(s)
Vout
 1  1 
  1   1 
R2  sR1C3   sR2C1 
G(s)  C3 C2
R3 1  sR2C2   1  Rupper
f z1 
1
  1  5
10k 2 R1C3
sR C
 3 3  R3 1 2
fz2 
1
C1 R2 2 R2C1
 s z1   s z 2 
1   1 
R2  s  s 

G( f )  Vout
R3  s  1  s p 2 
4
 1     6
s  s  Rlower
 p1  10k Vref 1
f p1 
2 R2C2
1
  
f p2 
f p12  f c 2 f p 2 2  f c 2 Gf c R3 2 R3C3
R2 
f z1
2
 fc 2  
f z 22  fc 2 f p1
Type 3

L1 R6
100m vout
75u
parameters a c Vout2
3 2 12
Rupper=10k
Rlower=Rupper d Resr
5
PWM switch VM p 70m
Vin Rload
fc=10k 10 X1 16 2.5
AC = 0
Gf 20
Gfc=-20 PWMVM2
L = 75u Cout
Fs = 100k 220u
G=10^(-Gfc/20)
pi=3.14159
fz1=1k vout
fz2=1k C2
fp1=26k {C2}
R3
fp2=26k
GAIN
XPWM {R3}
GAIN R2 C1
Rupper
K = 0.5 {R2} {C1}
C3=1/(2*pi*fz1*Rupper) {Rupper} 13
R3 =1/(2*pi*fp2*C3)
=1/(2 pi fp2 C3) 15 C3
6 14 {C3}
LoL
C1=1/(2*pi*fz2*R2) 1kH
C2=1/(2*pi*(fp1)*R2) 4
9
Rlower
a=fc^4+fc^2*fz1^2+fc^2*fz2^2+fz1^2*fz2^2 CoL Verr
{{Rlower}}
X2 V2
c=fp2^2*fp1^2+fc^2*fp2^2+fc^2*fp1^2+fc^4
f 2^2*f 1^2 f ^2*f 2^2 f ^2*f 1^2 f ^4 1kF AMPSIMP 2.5
7
R2=sqrt(c/a)*G*fc*R3/fp1 V1
AC = 1
fz1x=1/(2*pi*C1*R2)
fz2x=1/(2*pi*C3*(Rupper+R3))
( p ( pp ))
fp1x=1/(2*pi*(C1*C2/(C1+C2))*R2)
fp2x=1/(2*pi*C3*R3)
 Automated calculations help iterations!
 Conditional stability is gone
 Fine tuning is now possible
180 40.0
f0=1.2 kHz
|H(s)|
90.0 20.0
fc
ph_vout2 in degrees
Plot2
0 0
arg H(s)
-90.0 -20.0 12
fp1,2= 26 kHz
-180 -40.0
11
zeros
fz1,2= 1 kHz
80.0 180
|T(s)|
40.0 90.0
m
vdberr#a in db(volts)
es
ph_verr in degree
arg T(s)
90° 14
Plot1
0 0
13
-40.0 -90.0
-80.0 -180
1 10 100 1k 10k 100k

frequency in hertz

The zeros position affects the response
 Transient response changes as the zeros position moves
5 01
5.01
fz1 = 3 kHz
m = 75°
4
3
2
4.99
fz1 = 1 kHz
out2#1, vout2 in volts
m = 85°
plot1
4.97
vout2#2, vo
fz1 = 0.3 kHz

4.95 m = 90°
4.93
fz2 = 1 kHz
300u 900u 1.50m 2.10m 2.70m
ti i d

The zeros position affects the response
 Splitting
S litti the
th zeros can fix
fi DCM stability
t bilit iissue iin b
buck
k VM
Settling Phase
Frequency Overshoot
time margin
faster
fz1
slower
fz2 faster

By decreasing fz1, DCM stability is improved
|T(s)| CCM
40.0 180
DCM
arg T(s)
20.0 90.0
m
m
ees
vdberrr#4, vdberr in db(volts)
90° 12
ph_verrr#4, ph_verr in degre
80° CCM 11
Plot1
0 0
DCM
-20.0 -90.0
fc 13
10
fc
-40.0 -180  Conditional stability is gone in DCM
 Crossover frequency is improved in DCM
1 10 100 1k 10k 100k
frequency in hertz
CCM/DCM voltage mode buck

Improving DCM stability slows down the buck
 fz1 moving
i llow, it h
hampers th
the response titime…
5.01
2
1
4.99 k factor compensation

fz1, fz2 = 2.2 kHz
fp1,fp2 = 40 kHz Poles / zeros compensation
ut2#a in volts
fz1 = 300 Hz, fz2 = 1 kHz

Plot1
4.97
fp1,ffp2 = 26 kHz
kH
vout2, vou
P
4.95
4.93
236u 707u 1.18m 1.65m 2.12m

ti i d

Course agenda
 Poles and zeros
 Phase
Ph margin
i andd quality
li coefficient
ffi i
 Compensating
p g the converter
 Multi-output
 Input filter
p
 Conclusion
Type 2 and 3 with a TL431
 Litterature
Litt t examplesl use op amps to t close
l th
the lloop.
 Reality differs as the TL431 is widely implemented.
 How
H tto convertt type
t 2 and
d 3 tto a TL431 circuit?
i it?
 Question: who hides behind the TL431 anyway?
K
R
K
TL431A R
A
2.5V
R
A A shunt regulator! A
K

Feedback with a TL431
 A TL431 implements
i l t a two-loop
t l configuration
fi ti
FB signal
Rpullup FB D2
L1
MBR20100CT
2.2u
Vout
solution A Vdd
Rbias RLED Rupper
G d
Gnd
fast slow
lane lane
C2 C3
1mF Czero 100uF
Vcc
X1
solution B TL431A
FB Rlower
FB signal
g
Rpulldown
Gnd

Feedback with a TL431
 When Czero is a short circuit,
circuit the slow lane is off
 The TL431 turns into a static zener diode
RLED
VFB ( s )  CTR  R pullup  I1 I1
Vout LED
RdLED
V ( s)
Ic
RLED
I1  out
I1
RLED FB Vout
Vfz
Rpullup
Rpullup
VFB ( s ) R pullup
Vdd FB  CTR  Ic Rdz zener
Vout ( s ) RLED
Vz Vdd CTR x I1
Vz

The TL431 and its equivalent schematic
 The fast lane presence adds a zero in the equation
Vout
Czero
K1 = 1
K2 = -1
Rupper
K1
FB GAIN SUM2
K2
CTR*Rpullup/RLED
Rlower
 1  R pullup
VFB ( s )   Vout ( s )  Vout ( s )  CTR
 sRupper C zero 
  RLED
VFB ( s )  1  R pullup  sRupper C zero  1  R pullup No high

   1 CTR     CTR
Vout ( s )    frequency pole?
 sRupper C zero  RLED  sRupper C zero  RLED
Type 2

Adding a pole for a type 2 circuit
 The pole is a simple capacitor on the collector
Rpullup
Vdd FB Vdd
FB
Rpulldown
Cpole
p Cpole Or on the
emitter
VFB ( s )  sRupper C zero  1   1  R pullup

G (s)       CTR

Vout ( s )  sRupper C zero   1  sR pullup C pole  RLED
1 1 R pullup 1
f po  fz  G CTR fp 
2 Rupper Czero 2 Rupper Czero RLED 2 R pullup
ll C pole
l
Pole at the origin Low frequency zero High frequency pole Type 2
Mid-band gain
The type 2 final implementation
 The LED resistor fixes the mid
mid-band
band gain
Vout
RLED
Rupper
Vdd
1 3
U2A
Rpullup Czero
2
Cpole U2B X1 Rlower

TL431

Testing the ac response
B5
g
Voltage
parameters C2 V(ac)+4.99985
Vout=5 {C2} 5.00
1
R2 C1
Rupper=(Vout-2.5)/250u
The type 2 based on an

{R2} {C1} Rupper
fc=1k 20 {Rupper}
pm=100 2.47 2.50
Gfc=-20
op
p ampp is the reference
2
2.47
pfc=-55
2 50
2.50
G=10^(-Gfc/20) 3 Rlower
boost=pm-(pfc)-90 Vopamp X4 V3 10k
pi=3.14159 AMPSIMP 2.5
K=tan((boost/2+45)*pi/180)
C2=1/(2*pi*fc*G*k*Rupper)
C1=C2*(K^2-1)
R2=k/(2*pi*fc*C1) B3
Voltage
V(ac)+5.00004
Fzero=fc/k ac
0
Fpole=k*fc
V11 5.00
AC = 1
Rpullup=20k Czero1
7
RLED=CTR*Rpullup/G {Czero} R9
Czero=1/(2*pi*Fzero*Rupper) 5 {Rupper}
2.50
Cpole=1/(2*pi*Fpole*Rpullup)
16
CTR = 1
2.06
B1
Voltage
V(ac)+5.00004 X1
K1
K2
SUM2 2.50
F2 X5
SUM2
18
{CTR} RLED K1 = 1
VTL431 4.99 AMPSIMP R10
Controlling Vsource = V6 {RLED} K2 = -1 V9
4.83 21
2.5 10k
3.41 2.94
25 {Rupper}
26 15
Rupper
V6 Czero
Rpullup 2.49 X2
4.83 {Czero}
{Rpullup} 9
GAIN
GAIN
5.00 8
K = -Rpullup*CTR/RLED
27
V7 Cpole
X7 -29.4
5 {Cpole}
17
TL431_G
Rlower R8
10k {{Rpullup}
p p}
-29.4
19
C6
VoutEq
{Cpole}
Type 2

Comparison is perfect!
 The ac plot shows no significant differences
Phase
180 40.0
b
boost
phase
90.0 20.0 1
2
outeq, ph_vtl431 in degrees
3
eq, vdbopamp in db(volts)
plot1
0 0
1 kHz crossover
vdbtl431, vdboute
ph_vopamp, ph_vo
gain 6
5
4
Gain = 20 dB
-90.0 -20.0
-180 -40.0
10 100 1k 10k 100k

frequency in hertz
Type 2

A type 3 is also possible
 The type 3 is less flexible given the RLED role
Vout
RLED Rpz
Rupper
Cpz
Vdd
1 5 3
U2A
Rpullup Czero
2
C
Cpole U2B U1 Rlower
TL431
Type 3

The type 3 with a TL431
 RLED affects the gain and the zero position
VFB ( s)  sRupper Czero1  1  1   sC pz  RLED  R pz   1 R pullup

G(s)      CTR
Vout ( s )  
 sRupper Czero1  1  sR pullup C pole 2   sRpzC pz  1 RLED
1 1 1
f po  f z1  f p1 
2 Rupper
pp C zero1
2 Rupper Czero1 2 R pz C pz
1 R pullup 1
fz2  G CTR f p2 
2  RLED  R pz  C pz RLED 2 R pullup C pole 2
Pole at the origin Two zeros Mid-band gain Two poles
Type 3

Testing the type 3 with a TL431
B5
Voltage
5.00 V(ac)+4.99985
parameters 1
Vout=5 C2
{C2}
Rupper=(Vout-2.5)/250u R3
fc=1k {R3}
R2 C1
pm=100 Rupper
The type 3 based on an

5.00
{R2} {C1}
Gfc= 20
Gfc=-20 {Rupper} 20
34 C3
2.50
pfc=-55 2.04
{C3}
op amp is the reference

2
2.04
G=10^(-Gfc/20)
boost=pm-(pfc)-90 4 2.50
pi=3.14159 3
ac
K=(tan((boost/4+45)*pi/180))^2 Vopamp Rlower 0
X4 V3 10k
AMPSIMP 2.5 V11
C1=C2*(K-1)
C1=C2 (K-1)
R2=sqrt(k)/(2*pi*fc*C1) AC = 1
R3=Rupper/(k-1)
C3=1/(2*pi*fc*sqrt(k)*R3)
Fzero=fc/sqrt(k)
Fpole=sqrt(k)*fc
B1
pu up 0
Rpullup=20k Voltage
V(ac)+4.988
a=(fpole^2+fc^2)*(fc^2+fzero^2) * (fpole^2+fc^2)*(fc^2+fzero^2)
b=fpole^2*fpole^2+fpole^2*fc^2+fc^2*fpole^2+fc^4
Rled=(sqrt(a)/b)*Rpullup*fpole*fpole/(fzero*fc*G) Rled Rpz
F2 {Rled} Cpz 4.99
{Rpz}
1 {Cpz}
Czero1=1/(2*pi*Fzero*Rupper) Controlling Vsource = V6 4.99
10
Cpole2=1/(2*pi*Fpole*Rpullup) 4.70
16 Rupper2
p (p )( p p)
Cpz=(fpole-fzero)/(2*fzero*fpole*Rled*pi) 3.43
9
{Rupper}
VTL431 5
Rpz=1/(2*pi*Fpole*Cpz) 2.49
V6
CTR =1 Czero1 12
Rpullup 4.70 {Czero1}
{Rpullup} 11
5.00
Cpole2
14
{Cpole2} X7
V7 TL431_G
5
Rlower2
10k
Type 3

Final type 3 results
 The ac plot shows no significant differences
0 1
Phase
180 40.0
boost
phase
90 0
90.0 20 0
20.0 3
4
opamp in degrees
btl431 in db(volts)
Plot1
0 0
ph_vtl431, ph_vo
vdbopamp, vdb
1 kHz crossover 1
2
Gain = 20 dB gain
-90.0 -20.0
-180 -40.0
10 100 1k 10k 100k Type 3

frequency in hertz

The problem is the fast lane…
 The difference with the TL431 comes from the fast lane
 Can we get rid of it?
D2
L1
MBR20100CT BAS16 R5
2.2u D3 1k
Vout Q2
C4
100nF 2N3904
RbiasZ
D1
Dz 6.8 V
D2 L1
MBR20100CT 2.2u
Rbias Rupper Vout
RLED C5
100nF
slow 6V
lane
Rbias slow Rupper
C2 C3 RLED lane
1mF Czero 100uF
C2 C3
1mF 100uF
X1
TL431A
Czero
Rlower X1
TL431A
Rlower
 Rather costly solutions…

The implementation becomes simpler with a bias
 The TL431 acts as an open-collector op amp
Vbias Vout
No ac link
RLED Rupper R3
Vdd 1 3 9
C3
C2
Rpullup U2A C1 R2
2 8
Cpole U2B U1 Rlower

TL431
sR2C1  1 sC3  R1  R3   1 R pullup

G (s)   CTR
 C C   sR3C3  1 RLED
sR1  C1  C2   1  sR2 1 2 
 C1  C2  Mid-band gain Type 3

Watch for the TL431 bias!
 A TL431 requires 1 mA minimum to operate within specs
 A TLV431, 100 µA only…
16
15.8 with bias
15.6
15.4
15.2
Vout (V)
15
without bias
14.8
14 6
14.6
14.4
14.2
14
0 1 2 3 4 5 6
Iout (A)
 Changes in Zout implies a change in T(0)

How to provide more bias?
 Connect a resistor to Vout
 Use the LED to form a constant current source
L1
D2 2.2u
L1
MBR20100CT
2.2u
Vout RLED Rupper
Rbias RLED Rupper Vout
Ibias IL Ib
C3
R1 100uF
Vf R5
C2 C3 1k
1V 1k
1mF Czero 100uF
IL 1 mA Czero
X1
TL431A
IL + 1 mA X1
TL431A
Rlower
Rlower
Vf LED  1 V

What TL431?
 The TL431 is available under several grades
 TL431AI, 2.495 V, ± 2.2% TA = -25 °C to +85 °C
TL431AC 2.495
 TL431AC, 2 495 V 1 6% TA = -25
V, ± 1.6% 25 °CC to +85 °C
C
 TL431BI, 2.495 V, ± 0.8% TA = -25 °C to +85 °C
• BV = 37 V
V, IK,max = 100 mA and IK,min = 1 mA
 The TLV431 can regulate to a lower output

 TLV431A, 1.24 V, ± 2% TA = -25 °C to +85 °C
 TLV431B,
TLV431B 1.24
1 24 V
V, ± 1% TA = -25
25 °C to +85 °C
• BV = 18 V, IK,max = 20 mA and IK,min = 100 µA
NCP100 down to 0.9 V

Why Rlower never appears in the ac functions?
 Because of the virtual ground, Rlower disappears in ac!
Vin
dc equation
Rf
Rupper
pp  Rf  Rf
Vout  Vref   1  V
0  Rupper || Rlower  Rupper in
to the 4
in ac  
PWM Vout 2
1
ac equation
X1 Rlower
Vout  s 
AMPSIMP Vref
Rf

Vin  s  Rupper
pp
Does not play

a role in ac!

In absence of virtual ground, Rlower comes back
 In an OTA
OTA-based
based circuit,
circuit there is no virtual ground
 Rlower now affects the gain
vin
Rlower
k
Rlower  Rupper
Rupper
G1
gm
3 2 Verr
C1
1 C2
Rlower
R1 Verr  s  Rlower gmR2 1  sR2C1

Vin  s  Rlower  Rupper sR2C1 1  sR2C2 

Course agenda
 Poles and zeros
 Phase
Ph margin
i andd quality
li coefficient
ffi i
 Compensating
p g the converter
 Multi-output
 Input filter
p
 Conclusion
The optocoupler is the treator here!
 You need galvanic isolation between the prim. and the sec.
 An optocoupler transmits light only, no electrical link
Silicone
LED Creepage path
dome
c a
Detector
Ic IF
500 µm
Clearance
Detector LED Luigi Galvani, 1737-1798
Italian physician and physicist
e k
Ic
CTR   100
IF
French
Current Transfer Ratio specimen

The internal pole should be known
 The photons are collected by a collector-base area.
 This area offers a large parasitic capacitance.
Vdd
Vdd Vout
Rpullup RLED
Rpullup RLED
VFB
CTR C
VFB  s  R pullup CTR 1 If fp is above 5 times fc, its effect is negligible


Vout  s  RLED 1  sR pullup C If fp is close to fc, phase margin degradation

Assess the CTR variations
 CTR changes with the operating current!
 Try to select collector bias currents around 2-5 mA
CTR between 0.63 and 1.25

Normalized to 1 (0 dB) Watch out for
0.63 gives –4 dB fc
crossover frequency
1.25 gives +2 dB
changes and phase margin
at CTR extremes!
SFH-615

Find the pole position on your optocoupler
 Once your optocoupler is selected, characterize it:
Cdc Rled
Ic 10uF 20k
2 5 Vin
Rpullup
20k
Rbias
VFB 1 3
Vdd
5
X1 4 6
SFH615A-4
Vbias Vac
IF
 Adjust Vbias to VFB = 2.5 V (or Vdd/2) then ac sweep

 Observe VFB with a scope or a network analyser

Select the bias point as in your circuit
 Identify the pole position. Here it is located at 10 kHz
Vce = 2.5 V
Vce = 1 V
Pole at 10 kHz

Common collector or common emitter?
 The picture is unchanged. Pole stays the same.
Common collector Common emitter
Vdd Vdd
VFB
VFB

Changing the pullup affects the pole position
 A low
l pullup
ll resistor
i t offers
ff b
better
tt b bandwidth!
d idth!
30 kHz 10 kHz
Rpullup = 1 kΩ Rpullup = 4.7 kΩ
 Changing the bias point affects the CTR

CTR 5 kHz VFB  s  R pullup
ll
 CTR
Vout  s  RLED
Rpullup = 15 kΩ
 If Rpullup = RLED, then |G0| = 0 dB…?

A simple optocoupler model
 Read
R d th
the pole
l position
iti and
d th
the CTR value
l
Vsat
A 100m
C
DLED
N = 1.8 4
Cpole D2
2 {Cpole} BV = 30
N = 0.01
VILED
E
K F2
{CTR}
Controlling Vsource = VILED
1
C ppole 
2 R pullup f pole

Including the pole in the open-loop sweep
 We know the pole is there
there, then include it in the sweep!
parameters 9
vc
Vout=19
PWM switcch CM
duty-cycle
DC 10 X2x
Rupper=(Vout-2.5)/250u XFMR
D1
MBR20100CT
fc=4k RATIO = -166m vout
pm=70 2 16 12
vout
p
Gfc=-11.3 Vin
c
pfc=-92.3 120 X3
PWMCM R10
L = 770u 70m
G=10^(-Gfc/20) 5 X5
Fs = 65k
POLE Rload
Rl d
S+A
boost=pm-(pfc)-90 Ri = 0.33 L1 15
K
FP = pole C5 6
pi=3.14159 Se = 24k 770u K = 0.33 500u
K=tan((boost/2+45)*pi/180)
C2=1/(2*pi*fc*G*k*Rupper)
C1=C2*(K^2-1)
R2=k/(2*pi*fc*C1)
Vdd
vout
5
Fzero=fc/k
11
Fpole=k*fc
Verr Rpullup RLED Rupper
LoL {Rpullup} {RLED} {Rupper}
Rpullup=20k 1kH
RLED=CTR*Rpullup/G 3 1 4 8
Czero=1/(2*pi*Fzero*Rupper) CoL
Cpole=1/(2*pi*Fpole*Rpullup) 1kF
Czero
14 {Czero}
CTR = 2 6
Cpole
Pole=6.8k {Cpole}
18
X9 Rlower
Vac TL431_G 10k
AC = 1
Xopto
SFH615A-4

Plot the open-loop gain and compensate
 Compensate using the k factor for a 4 kHz fc
0 1 0 1
180 20 0
20.0 Gain 180 80 0
80.0
Gain
PM = 71°
90.0 10.0 90.0 40.0
ph_vout in degrees
vdbout in db(volts)
ph_verr in degrees
Phase = - 92°
vdberr in db(volts)
Phase
Plot1
Plot1
0 0 0 0
3.8 kHz crossover 1
-90.0
90 0 -10.0
10 0 -90.0
90 0 -40.0
40 0
Phase
2
4 kHz crossover
-180 -20.0
Gain = - 11 dB -180 -80.0
2 1
10 100 1k 10k 100k 10 100 1k 10k 100k
frequency in hertz frequency in hertz
H(s) T(s)
Opto pole included Opto in normal position

Course agenda
 Poles and zeros
 Phase
Ph margin
i andd quality
li coefficient
ffi i
 Compensating
p g the converter
 Multi-output
M lti t t converters
t
 Input filter
p
 Conclusion
Compensating multi-output converters
 A multi-output
multi output converter can be simplified as a single output system
system.
 All loads and capacitors have to be reflected to the regulated winding.
FB
1:N2 R1
2
reflected N 
FB
C1
Ceq  C2  3 
 N2 
Req 2
R1 N 
RL
RLeq RLeq  RL  2 
Vout
 N3 
C1 Ceq 2
N 
Req  R2  2 
R2 1:N2  N3 
RL
1:N3
R1 || R
e
q
C2
If time-constants
are close: C1 || C
e
q
Assumed to
be a short-circuit

The diode is not always a short-circuit!
 The dynamic resistance varies depending on the load
dV f 50m
I = 1 A Rd    50 m @10 A
dI F 1
Not really a short-circuit

in light load conditions!
dV f 50m
Rd    1  @100 mA
dI F 50m
I = 50 mA
The simple reflection seen

before no longer works!

Fortunately, SPICE does the reflection for you
 The dynamic resistance is calculated during bias point
X1 D1
RATIO = 0.2 MR852 V18
123 24.6 23.8
10 1
18V
315 R1
IN OUT
12 500
500m
DON GND X4 23.8 R2
Vin 0.188 FLYBACKVM 3 18
315 17 FS = 75k C1
L = 180u 1mF
• Multi-output
Vstim
AC = 1
• different lanes connection
0
18
MC44608 monitor application X2 • Different diodes.
COL RATIO = 0.071 7.86
1kF
8.74
4 5
8V • SPICE can do it!
D2
R3
LOL V18 MR852
500m
1kH R4
7 86
7.86
0.188 6 5
19
Vcc R9 V112 C2
14 270 1mF
14.0 23.2
14 15
Rupper
I to D
240k
X7 X3 D4
44608 Cf RATIO =1 MR854 V112
100n 123 122
5.07 7
122V
16
13
R5
X5 2.50 400m
11
TL431 122 R6
Rlower 9 180
5k C3
47uF

SPICE can predict the small-signal response
 An ac sweep on the prototype confirms the modeling approach
+180° 30dB
+20 144.0
PHASE PHASE
0dB
0 72.00
Open-loop phase (Degrees)

0° -20 0
GAIN
GAIN 2
2
-40 -72.00
-60 -144.0 1
-180° 100 1K 10K

Frequency (Hz)
Bode Plot of the 44608 EVK when stimulated from the PWM input
Frequency (Hz)
Measured with an network analyzer Simulation results

A multi-output forward converter
 It delivers all voltage to power a motherboard
Icoil1
D9
Main
L1 RL1
mbr6045wt vout
Vbulk {L1} 9m
V1 390 V12V
3 8 4
31
X13
D8
1N4148 R10x
47
1
X8
R14
30m
g
regulation
VdsUpper

R1 XFMR-AUX D3
{Rpullup} X1 19
IRF350 Lleakx
RATIO_POW = 2 mbr6045wt 6
C7 I1
PWMCM C8 X3 1p
RATIO_AUX = 2 2.2m
CMP 0.22u R15 22 XFMR-AUX IC = 0
1k IdUpper RATIO_POW = N1
FB OUT D1
.
5 12 13 10 33
osc OSC GND 18 MUR460 RATIO_AUX = N2
SENS
VgsU
.

23
7 X7
D6 XFMR
1N4148 Imag
11 RATIO = N2/N1
L4x D4
D7
. mbr6045wt RL2
296u 14
1n5818 Vprim Lmag 4m

D10 Lleak
{Lleak} {Lp} 24 3 V5V
1N4148
VgsL 36 20
R20
17 16
47 X9 34
R6
9 IRF350 D5 50m
R2
mbr6045wt
{RL2}
35 25
C6
D2 2.2m
MUR460 IC = 0
R21 IdLower
1k
VdsLow

15
Coupled inductors
R18
1k
VCS 2
C9 Rsense
100p {Rsense}
vout
Verr Rled
{RLED}
26 28 {Rupper}
Rupper
X4 R8
SFH610A 1k Czero 27
{Czero}
Cpole 29
{{Cpole}
p }
A multi-output forward X2
TL431_G
Rlower
10k
Cycle-by-cycle model
Average modeling of the multi-output forward
D1
mbr6045wt L1 RL1 L5 R8
12.0V vout
12.1V {L1}
12.1V 9m 0.27u 2m 12.0V
9 6 22 25 Vout1 12V
12.5V 12.0V
15
X1
R1 R10
XFMR 30m 45m
RATIO = N1 12.0V 12.0V
7 26
I1
X2 C1 C3
RATIO = N2/N1 2.2m 1200u
4.90V
X4 29
PWMCM2
L = Leq
Fs = 75k
Ri = Rsense
Se = Sp
Lleak
1p
dc
285mV
RL2 L6 R11
duty-cycle 1.5m 0.27u 2m
4.90V 4.90V 4.85V
a c Vout2 5V
3 14 28 2
108V 5.38V D2 4.88V 4.88V
8 5 27
mbr6045wt
380V vc X4x R13
10
1
PWM switch CM p XFMR R2 45m
RATIO = N2 50m Rload2
V1608mV 4.85V
4.88V 30 333m
{Vin}
11 C4
B1 C2 1200u
Voltage 2.2m
V(err2)/3>1 ? V2
1 : V(err2)/3 5
5.00V
21
vout
R4
{Rpullup}
Verr RLED
{RLED}
Auto-toggling err2
X5
11.5V
4 {Rupper}
Rupper
CM model
Optocoupler R5 2.50V
Cpole = 2.2n 1k Czero 16
1.82V CTR = CTR 10.3V {Czero}
13
Cpole
{Cpole}
A multi-output forward X6
TL431_G
Rlower
10k
Averaged model
Compensate the Bode plot with a type 2
 The converter is compensated for a 10 kHz crossover
180 20.0 Power stage only
90.0 10.0 |H(s)|

ph_vout1 in degrees
vvdbout1 in db(volts)
Plot1
0 0
arg H(s)
90 0
-90.0 10 0
-10.0
3
-180 -20.0
180 60.0
|T(s)|
90.0 30.0
arg T(s)
egrees
vdberr in db(volts)
ph_verr in de
Plot2
0 0
-90.0 -30.0
5
fc 4
-180
180 -60.0
60 0
10 100 1k 10k 100k

f i h t

Averaged response versus cycle by cycle
 Very good agreement between both models
12.20
Averaged
vout1#a, v12v in volts
12.10
Plot1
12.00
7
9
11.90
11.80 12 V output – regulated
5.02 5.04
Cycle-by-cycle
4.98 5.00
8
10
vout2 in volts
v5v in voltts
Plot2
4.94 4.96
4.90 4.92
4.86 4.88
5 V output – non regulated
4.93m 5.57m 6.21m 6.85m 7.49m
ti i d

Simulation versus reality
 The converter was built and load stepped
4.0V
Feedback pin 100 µs / division
3.0V
272 mV Feedback pin

12 V output SEL>>
2.0V
V(FB)
12.3V
12.2V
8A 250 mV
12 V output current 12.1V
12 V output
12.0V
200 µs / division
di i i
11.9V
0.8ms 1.2ms 1.6ms 1.8ms
V(V12V)
Time
Measured on the bench Simulation results

Weighted feedback
 With multiple outputs
outputs, we have two loops!
24 V  Assign weights to outputs
 e.g. 30% tto Vout1, 70% tto Vout2
 Select a bridge current Ib,250 µA
12 V
Vout1  Vref
R1 
R2 R1
I bW1
weights
Vout 2  Vref
R2 
I bW2
Vref
2.5 V R3 
Ib
R3

How do we measure the loop gain?
 With linear systems
systems, we can apply superposition (1)
B plot B 
20 log10  1 
Ac
A
 A1 
modulation
12 V
12 V
R2 R1 bias
 Dc bias the 12-V

12 V input
 Ac modulate the second one
R3
 Store the Bode plot

 Study the second loop by biasing the previous one
24 V
12 V B plot B 
Ac A
20 log10  2 
modulation  A2 
24 V Watch out, high
R2 R1
bias gain input!
 Dc bias the 24-V

24 V input
 Ac modulate the second one
R3  Store the Bode plot

 Combining the signals requires vector manipulations:
1.20
j
90°

600m
A
sin  
180° 
0 1
 0
0°
cos  
-600m
A
Ae j  A  cos    j sin   
-1.20
1 20
1.02m 1.07m 1.12m 1.17m 1.22m

time in seconds
 
Vout1  A1  cos 1  j sin 1  Vout 2  A2  cos 2  j sin 2 
 
 
Re VFB  A1 cos 1  A2 cos  2  Y  
Im VFB  A1 sin 1  A2 sin 2  X
   X
VFB  Y  jX VFB  Y 2  X 2 arg VFB  tan 1  
Y 

Course agenda
 Poles and zeros
 Phase
Ph margin
i andd quality
li coefficient
ffi i
 Compensating
p g the converter
 Multi-output
 Input filter
p
 Conclusion
The need for an input filter
 The dc-dc
Th d d converter iis supplied
li d ffrom a d
dc source
 A capacitor locally decouples the line. In theory:
 The capacitor supplies the ac current
 The source only sees a dc current
Idc Idc + Iac
Iac
Iac
 In reality, the capacitor is not perfect (limited cap., ESR and ESL)
 Some ac current manages to enter the source
 Source
S pollution,
ll ti adjacent
dj t converters
t disturbance,
di t b radiated
di t d noise
i

 A front-end filter has to be installed
 The ac current will only flow in C as L opposes its circulation
 The remaining ac current in the source must pass the specs!
L rLf Iout
Iin Switch
Vin C Mode Rload Vout
Converter

 O
Our dc-dc
d d ensures Pout is i constant regardless
dl off Vin
 If Vin increases, Iin reduces to keep Vin.Iin constant
 = 100%
 If Vin decreases, Iin increases to keep Vin.Iin constant
Vin Pin
Rin  Vin  Pin  Rload I out 2
I in I in
Neg.
N sign!
i !
2
dVin d Rload I out 2
 I out 
   Rload   SPICE simulation
dI in dI in I in I
 in  constant p
power sink
4
.TF V(4) V1 ; transfer function analysis
V1 Bpower
***** SMALL SIGNAL DC TRANSFER FUNCTION 100 Current
output_impedance_at_V(4) 0.000000e+000 60/(V(4)+1u)
v1#Input_impedance -1.66667e+002
Transfer_function 1.000000e+000
P = 60 W
Neg. sign!

 Ohmic paths in the RLC filter are losses that damp the network
R1 L
R2
R3
C
 If these losses are compensated, the damping factor is zero

1 1
T s  Q
s2 s 2 If ohmic losses are gone, the
 1 damping factor is zero.
02  0Q
 In a lossy RLC
C filter,
f the damping factor
f is:
L  C  R2 R3  R1 R3  R2 R1  R1 R2C  L
  0 R3  
2  R1  R3  =0 C  R1  R2 

A tunnel-based oscillator?
 Loading an RLC filter with a negative impedance:
 You build a negative resistance oscillator!
R1 R2C  L 100m  500m 1u  100u
R3     166
66 
C  R1  R2  1u   600m 
1.00k
R3 = – 150 
600
200 Diverging oscillations

-200 <0
-600
300
R3 = – 166 
200
100 Steady-state oscillations

0 =0
-100
180
140 Damped oscillations

100 >0
60.0
20.0 R3 = – 175 
500u 1.50m 2.50m 3.50m 4.50m

When the damping factor changes…
 A simple experiment using SPICE
 Take the constant power source
 Load the RLC filter with it
 Ch
Change the
th iinputt voltage
lt b
by steps
t
150 Input voltage
Req = -166 
130
Output voltage
v4, v1 in volts
plot1
110
90.0
70.0
>0 0
<0
3.64m 10.9m 18.2m 25.5m 32.7m

Watch the output/input impedances
 A possibility
ibili to llook
k ffor oscillations
ill i is
i to ac sweep Zout off RLC
70.0
Rin = 166 Ω = 44.4 dBΩ <0
vvdbout#ain db(volts)
40.0
2
Z02 R 
Z out  1  1 
Plot1
10.0
dB max
R1
-20 0
-20.0
 Z0 
R2 = 0
-50.0 ZoutFILTER max = 60 dB 
To be safe
70.0
Rin = 166 Ω = 44.4 dBΩ = 0
40.0
vdboutin db(volts)
Z out  Z in
Plot2
2
d
10.0
dB
R2 = 500 m 
-20.0
-50.0 Z outFILTER max = 44.4 dB 

10m 100m 1 10 100 1k 10k 100k 1Meg 10Meg
frequency in hertz

The filter effect also appears on the Bode plot
 The
Th filter
fil peaking
ki b brings
i iinstabilities…
bili i
180 40.0
Without filter
|T(s)|
90.0 20.0
gain
ph_vout2#1, ph_vout2 in degrees
With filter
vdbout2, vdbout2#1 in db(volts)
Plot1
0 0
argT(s)
phase
-90.0 -20.0
Without filter
3
Damp that
-180 -40.0
RLC filter!
With filter 1
4
2
10 100 1k 10k 100k
frequency in hertz

How to damp the RLC network
 Damping
D i can b be obtained
b i d through
h h diff
different arrangements:
 a resistor in parallel with C1
 a resistor in parallel with L1
 Numerous other possible combinations!
Vin
R1 L1 dc-block
0.1 100u
2 4 Vout
1
C1 Cdamp
1u
V1 B1
AC = 1 3 5 Current
60/(V(4)+1)
R2
Rdamp
0.5

Buck, a design example
 Th
The damping
d i resistor
i iis calculated
l l d to 99.5
5 Ω iin this
hi example
l
 The dc-block capacitor is swept from 2 to 10 times C1
44.0
Cdamp = 0
28.0 Cdamp = 2 µF
Cdamp = 4 µF
dB
12.0
Cdamp = 6 µF 1
4
3
2
5
Cdamp = 8 µF
4.00
ZoutFILTER
20.0
1 10 100 1k 10k 100k

frequency in hertz

Buck, a design example
 Once
O damping
d i iis adequate,
d oscillations
ill i cease
150 Input
Input
voltage
130
ut, vin in volts
110
vou
90.0
Output
voltage
70.0
4.00m 12.0m 20.0m 28.0m 36.0m

time in seconds

Course agenda
 Poles and zeros
 Phase
Ph margin
i andd quality
li coefficient
ffi i
 Compensating
p g the converter
 Multi-output
 Input filter
 A real case examplep
 Conclusion
Stabilizing a UC3843 converter
 A 19 V/3 A converter is built around an UC3843
T1
86H-6232
0.18 : 1 : 0.25
HV-bulk
R19 R3 R13
47k 47k 47k C2
10n D5
400V MBR20100 L2
2.2u
+
. .
-
C5a C5b C7
Vout
IC4
KBU4K
D2 . 1.2mF
25V
1.2mF
25V
220uF
25V
MUR160
C11 R6
100p
Vref C13
6k Gnd
R17 2.2nF
IN
47k Type = Y1
1 CMP Ref 8 R1 D8 R8 R14

R7 330 1N4937 1k 4.7k
C4 10k
2 x 10mH 100uF
2 FB Vcc 7
L1
Schaffner RN122-1.5/02 400V M1

R12
3 CS DRV 6
SPP11N60S5 10k
V f
Vref R16
R11 4 Rt GND 5 10 R18
10k 47k
C10
U1
470n X2
UC3843
U3B R10
56k
R23 R24
1Meg 1Meg
U3A C6
R5 100n
1k
R15 R6a R6b

85-260 Vac 4.7k 1 1
C12 C16 C15 C3 IC2 R9

220p 4.7nF 10nF 220uF TL431 10k
Gnd

Opening the loop to get H(s)
 A simple dc bias is ok for H(s) alone
T1 Put the nominal load
86H-6232
0.18 : 1 : 0.25 Adjust dc bias to reach Vout
HV-bulk
R19
Ac sweep p the board
R3 R13
47k 47k 47k C2
10n D5
400V MBR20100 L2
2.2u
. . C5a C5b C7
Vout
D2 . 1.2mF
25V
1.2mF
25V
220uF
25V
MUR160
C11 R6
100p
Vref C13
6k Gnd
R17 2.2nF
47k Type = Y 1
1 CMP Ref 8 R1 D8 R8 R14

R7 1N4937 1k 4.7k
C4 330
10k 2 FB Vcc 7
Dc
100uF
400V M1
3 CS DRV 6 R12
SPP11N60S5 10k
block R11 4 Rt GND 5
Vref R16
10 R18
10k 47k
U1
UC3843 R10
U3B
56k
C10
100u
U3A C6
R5 100n
R21 1k
100k R15 R6a R6b
4.7k R20 1 1
4.7k
C12 C16 C3 IC2 R9
220p 4.7nF 220uF TL431 10k
C15
10 F
10nF Gnd
Dc bias
Ac mod.

Run the same circuit in SPICE
 A simple dc bias is ok for H(s) alone
witch CM
vc
a
y-cycle
DC 6
X2x
duty
PWM sw
XFMR D1A
mbr20200ctp vint L1 R1
RATIO = -250m vout
2.2u 20m
vout
2 3 4 5
p
Vin c
150
AC = 0 R10 R15
8 15m 85m
X9
Rload
PWMCM L
Lp
1 10 6.3
L = Lp {Lp} (V(errXX)-1.2)/3 > 1 ?
Fs = Fs 1 : (V(errXX)-1.2)/3 C5 C1
9
Ri = Rsense 2m 220u
B3
Se = Se
Voltage
R2
47k
CCM operation
X3
OP384X1 R3 Low line voltage
47k
13 15
errXX UC384X
gnd
2V5
V3
1.08
AC = 1
Dc + ac modulation

Compare model and real board results
simulated
2

Another simple method
 H(s) alone can be measured without loop opening
R19 R3 R13
47k 47k 47k C2
10n D5
400V MBR20100 L2 ( )
Vout(s)
2.2u
. . C5a C5b C7
Vout
D2 . 1.2mF
25V
1.2mF
25V
220uF
25V
MUR160
C11 R6
100p
Vref C13
6k
R17 22 F
2.2nF
Gnd
47k Type = Y1
1 CMP Ref 8 R1 D8 R8 R14 B

C4
R7 330 1N4937 1k 4.7k
B
100uF
10k 2 FB Vcc 7 20 log10  
400V M1  A
3 CS DRV 6
Vref R16
SPP11N60S5
R12
10k A
R11 4 Rt GND 5 10 R18
10k 47k
U1
UC3843
U3B R10
56k
Vc(s)
U3A C6 C10
R5 100n 10uF
1k
R15 R6a R6b

4.7k 1 1
C12 C16 C3 IC2 R9

220p 4.7nF 220uF TL431 10k
C15
10nF Gnd
Watch out for capacitor connection (short-circuit to GND when discharged)

Open-loop study, a possible solution?
 Opening the loop offers several
se eral advantages:
ad antages
 Easy access to the output stage transfer function
 Works if the power supply is not stabilized
 Can almost pass dc if no ac capacitor is use
 It also brings
g drawbacks:
 Possible runaway in high-gain systems
 Needs to adjust the bias point if Rload or Vin change
 Almost impossible to stabilize the error amp. output
 Operate in closed loop and inject signal via a transformer!
1 3
Floating Ground referenced

ac source ac source

Transformer injection, a safer way
 The loop is closed in dc but open in ac!
Gain decreases
Vout
1 8
2 7
5 mV / div Vref
0 dB 3 6
Vin = Vout Vin
1.5 kHz / V 4 5
R2
20
R20
4.7k
High
gain A B
Vsource ac
source
VB
T  s   20 log10
2.13 6.38 10.6 14.9 19.1 VA

Change the operating conditions easily
simulated
CCM operation, Rload = 6.3 Ω

Reduce the load to enter in DCM
simulated
DCM operation, Rload = 20 Ω

From the open-loop Bode plot, compensate
RLED = 270 Ω
40.0 180 CZERO = 8.2 nF
argT(fc) = -90° CPOLE = 10 nF
20.0 90.0
ph_vout in degrees
vdbout in db(volts)
Plot1
0 0
-20.0 -90.0 |T(fc)| = -18 dB

21
-40.0 -180
20
10 100 1k 10k 100k
frequency in hertz

Verify in the lab. the open-loop gain
 Sweep extreme voltages and loads as well!
Simulated
CCM operation, Rload = 6.3 Ω, Vin = 150 Vdc

Si l t d
Simulated
0 100 1k 10k 100
CCM operation, Rload = 6.3 Ω, Vin = 330 Vdc

Si l d
Simulated
DCM operation, Rload = 20 Ω, Vin = 330 Vdc

As a final test, step load the output
 Good agreement between curves!
Vin = 150 V
Simulated CCM
2 to 3 A
1 A/µs

As a final test, step load the output
 DCM operation at high line is also stable
Vin = 330 V
Simulated DCM
0.5 to 1 A
1 A/µs

Agenda
 Poles and zeros
 Phase
Ph margin
i andd quality
li coefficient
ffi i
 Compensating
p g the converter
 Multi-output
 Conclusion

Conclusion
 We now understand the origins of phase margin needs
 The crossover frequency value is analytically derived
Current mode technique simplifies the compensation
 Current-mode
 Operating mode transition is not a problem for CM
 Type 2 and type 3 are also available with a TL431
 The optocoupler brings a pole degrading the phase margin
 Do NOT forget the influence of the EMI filter
 SPICE eases the design with multi-output converters
 A real-case example confirmed the validity of the approach!
Merci !
Thank you!
Xiè i !
Xiè-xie!

DC-DC Converters Feedback and Control

Caricato da

Informazioni sul documento

Titolo originale

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

DC-DC Converters Feedback and Control

Caricato da

Copyright:

Formati disponibili

Dc-dc converters feedback and control

presented by Christophe Basso

Wild Bill Hickok

Chris Basso – June 2008

 Compensating with a TL431

A stable and noiseless dc source we can trust!

Input voltage Output voltage

4 Chris Basso – June 2008

Compensating for the converter shortcomings!

5 Chris Basso – June 2008

Modulator - GPWM Vref Rlower

6 Chris Basso – June 2008

 H s  To sustain self-oscillations, as Vin(s)

8 Chris Basso – June 2008

100 1k 10k 100k 1Meg

9 Chris Basso – June 2008

10 Chris Basso – June 2008

11 Chris Basso – June 2008

12 Chris Basso – June 2008

10 100 1k 10k 100k

13 Chris Basso – June 2008

 = -180° 10.0u 30.0u 50.0u 70.0u 90.0u

10 100 1k 10k 100k

14 Chris Basso – June 2008

Loop gain |H(s)| -180° H(s) power stage

1 10 100 1k 10k 100k

15 Chris Basso – June 2008

Loop gain |T(s)| Gain is 1 112 H(s) power stage

Loop phase arg T(s)

1 10 100 1k 10k 100k

16 Chris Basso – June 2008

T(s) = 0 dB Gain margin:

17 Chris Basso – June 2008

18 Chris Basso – June 2008

80.0 360 -360°:

1 10 100 1k 10k 100k

19 Chris Basso – June 2008

 solving for N(s) = 0, the roots are called the zeros

 solving for D(s) = 0, the roots are called the poles

21 Chris Basso – June 2008

Left Half Plane Right Half Plane

 We can place these roots in the imaginary plane

23 Chris Basso – June 2008

LHP RHP x sz2

25 Chris Basso – June 2008

Vin ( s ) 1  sRC 1  s -60.0

26 Chris Basso – June 2008

27 Chris Basso – June 2008

Vout ( sz1 ) sz1

29 Chris Basso – June 2008

30 Chris Basso – June 2008

100u 300u 500u 700u 900u

32 Chris Basso – June 2008

58.3% Vout drops!

33 Chris Basso – June 2008

 Voltage mode or current

34 Chris Basso – June 2008

35 Chris Basso – June 2008

1 10 100 1k 10k 100k 1Meg

36 Chris Basso – June 2008

37 Chris Basso – June 2008

 The refueling time of the capacitor is delayed and a drop occurs

38 Chris Basso – June 2008