Professional Documents
Culture Documents
Frank Xi
fxi@monolithicpower.com
11/1/2007
High Efficiency
Can Step-Down, Step-up, or Both, or Invert
Can Achieve Higher Output Power
11/1/2007
11/1/2007
Step-down (Buck)
VIN
IIN
VSW
S
IO
VOUT
IL
ID
VSW
IC
Basic Relationships
VOUT
IL
TON
TON
VIN = D VIN
T
= D I OUT
VOUT =
IO
IIN
CCM Mode
IL always supplies load
IC small, independent of
load
I IN
ID
DCM Mode
2
VOUT
IO
Discontinuous Conduction Mode (DCM)
11/1/2007
TON
V
=
2 I O L T IN
2
TON +
VIN
Step-up (Boost)
L
VIN
VSW
IO
VOUT
IL
IC
S
Basic Relationships
VSW
T
VOUT
TON
IL
TOFF
CCM Mode
IL only supplies load during
TOFF period
IC large and load dependent
VOUT =
T
TOFF
I IN = I L =
IC
-IO
Continuous Conduction Mode (CCM)
VIN =
1
VIN
1 D
1
I OUT
1 D
DCM Mode
IL
2IO L T
VIN
VIN
2
TON
TON +
2
IC
VOUT =
-IO
Discontinuous Conduction Mode (DCM)
11/1/2007
Modulation Scheme
PFM (Pulse-Frequency-Modulation)
PWM (Pulse-Width-Modulation)
11/1/2007
Voltage Mode
11/1/2007
Current Mode
11/1/2007
VIN
S1
VSW
VOUT
R1
RESR
CLK
RL
S2
VRAMP
R2
VEA
VSW
VOUT
D
QB R
11/1/2007
CLK
RST
VRAMP
VEA
IL
IOUT
VFB
EA
VREF
10
0 =
1
,Q =
LC
1
RL
1
L
C
+ RESR
L
C
and
d
1
= a( s)
vFB VR
where a ( s ) is the transfer function of the error amplifier
11/1/2007
11
11/1/2007
12
R2
VEA
VFB
Bandwidth: ~400kHz
Phase margin: ~35
Conditionally stable
EA
VREF
VEA = VREF +
vEA =
LG =
R2
(VREF VFB )
R1
R2
vFB
R1
R2 VIN (1 + sCRESR )
s
s2
R1 VR
1+
+ 2
Q0 0
13
11/1/2007
14
LG A
s
s
[1 + sR A C ](1 + sR C )(1 + sR C ) V
+
1+
R3=200k
UGBW R C
V LC
switching frequency for
High DC gain rolls off by dominant pole and,
high frequency noise
st
phase shift recovered by 1 zero before 0
attenuation:
nd
2 zero brings back phase shift above 0
C3=0.2pF, p2=795kHz;
2nd and 3rd pole attenuates high frequency noise
R2=10k, p3=1.5MHz
3
FB
EA
REF
1 1
EA
FB
1 1
IN
ESR
IN
11/1/2007
15
16
VIN
VSW
VOUT
R1
RESR
CLK
RL
S2
R2
VRAMP
VEA
RST
RSEN
VSW
VOUT
Slope
Comp
D
QB R
IL
CLK
RST
VRAMP
VEA
IOUT
VFB
EA
VREF
11/1/2007
17
1. D=1/3: m2/m1=1/2
ie[n]
m1
m2
ie[n+1]
m2
m1
ie[n+1]
m2
< 1 : ie attenuates over cycles
m1
m2
> 1 : ie grows over cycles
m1
2. D=2/3: m2/m1=2
ie[n]
ie[n]
11/1/2007
m2 ma
)
m1 + ma
ma is chosen so that
ie[n+1]
m2 ma
<1
m1 + ma
m2
, guaranteed stable
2
ma = m2 , 1 cycle correction
ex : ma =
18
RX
VIN
Gm
IO
VOUT
R1
RESR
RL
io
1
1
1
1
=
=
R
T
m
vEA RSEN 1 + L S [(1 + a )(1 D ) 0.5] RSEN 1 + RL
L
m1
RX
where RX =
R2
VEA
1 + 1 e sTS
1 + e sTS sTS
Gm ( s )
EA
VREF
Q=
m2 ma
m1 + ma
RSEN
1
RSEN
1
1 + 1 e sTS
sTS
RT
m
sTS
1 + L S [(1 + a )(1 D ) 0.5] 1 + e
L
m1
1
1
RL
s
s2
1+
+ 2
1+
RX
QS S
1
(1 + ma )(1 D) 0.5
m1
11/1/2007
where =
VFB
L
ma
TS [(1 +
)(1 D) 0.5]
m1
1 2 D(1 ma )
m2
19
11/1/2007
20
Example:
C=22uF, RESR=10m Ohm
RSEN = 0.5 Ohm
FSW = 1.5MHz
RX=19.4 Ohm
RL=10k, 1k, 100, 10, 1 Ohm
11/1/2007
21
a( s) =
CC
RZ
vEA
1 + sCC RZ
A0
vFB
[1 + sCC (ro + g m R1ro + RZ )](1 + sC1RZ )
(CC >> C1 )
where : g m is the transconductance of the error amp
ro is the output impedance
R1
VFB
ie
EA
VEA
VREF
22
RL
1
1
1 + sCRESR
1 + sCC RZ
A0
RSEN 1 + RL 1 + s + ( s ) 2 [1 + sC ( RL || RX )]
[1 + sCC (ro + g m ro R1 + RZ )](1 + sC1RZ )
RX
QS
S
11/1/2007
1
RSEN
RZ
=1
sBW C R1
1
1 RZ 1
2 R1 RSEN C
23
VIN
S2
VSW
VOUT
CLK
R1
RESR
D
D
S1
VRAMP
RL
VEA
R2
VSW
VIN
IL
D
QB R
11/1/2007
IOUT
CLK
RST
VRAMP
VEA
VFB
EA
VREF
24
d (1 D) 2 s 2
(1 D)
,
0 =
LC
(1 + sCRESR )(1 s
L
)
2
RL (1 D)
LC
L
s
(
+
+ CRESR ) + 1
(1 D) 2
RL (1 D) 2
RHP
RL (1 D) 2
,
=
L
Q=
VIN
(1 D) 2
(1 +
)(1
RHP
s
s2
1+
+
Q0 0 2
1
1
(1 D) RL
L
C
+ (1 D) RESR
C
L
and
d
1
= a(s)
vFB VR
where a ( s ) is the transfer function of the error amplifier
11/1/2007
25
I O = I L (1 D) io = iL (1 D)
vo
sL
=
io (1 D) 2
Leff =
L
(1 D) 2
26
IO
1 D
d VO
diL
= DVIN + (1 D)(VIN VO ) iL
dt
sL
I
d VO
(1 D) |=| d O |
1 D
j RHP L
RHP
RL (1 D) 2
=
L
27
11/1/2007
28
C2
R2
VFB
R3
C1
R1
EA
VEA
VREF
vEA
(1 + sC1R1 )[1 + s ( R3 + R2 )C2 ]
A0
vFB
[1 + sR3 (1 + A0 )(C1 + C3 )](1 + sR2C2 )(1 + sR1C3 || C1 )
LG A0
11/1/2007
VIN
1
1
= 1 BW =
VR sBW LC
2
R1 C2
LC
VIN
VR
(1 +
1+
)(1
RHP
s
s2
+ 2
Q0 0
29
30
VIN
VSW
S2
VOUT
CLK
R1
RESR
D
RL
VRAMP
S1
VEA
R2
RST
VSW
VIN
RSEN
Slope
Comp
D
QB R
IL
CLK
RST
VRAMP
VEA
VFB
IOUT
EA
VREF
11/1/2007
31
Gm
IO
VOUT
R1
RESR
RL
R2
io 1 D
1
=
vEA RSEN 1 + RL
2 RX
where RX =
VEA
VFB
EA
VREF
L
1
2L
(
||
)
2
(1 D) T [(1 + ma )(1 D) 0.5] DTS
S
m1
S =
Q=
1 D
1
1
s
s2
RSEN 1 + RL
1+
+
2 RX
QS S 2
TS
2
(1 + ma )(1 D) 0.5
m1
11/1/2007
1 2 D(1 ma )
m2
32
vEA
2 RSEN 1 + RL 1 + s + ( s ) 2
S
QS
2 RX
RHP
(1 + sCRESR )(1
[1 + sC (
ZRHP
RL
|| RX )]
2
RL (1 D) 2
=
L
Example:
VIN=2.5V
VOUT= 5V, 10V, 15V, 20V
L=2.2uH, C=10uF
RESR=10m Ohm
IOUT = 100mA
RSEN = 0.5 Ohm
FSW = 1.5MHz
11/1/2007
33
2 RSEN 1 + RL 1 + s + ( s ) 2
2 RX
QS
S
(1 + sCRESR )(1 s
[1 + sC (
ZRHP
RL
|| RX )]
2
C1
CC
RZ
R1
VFB
EA
VEA
VREF
)
A0
1 + sCC RZ
[1 + sCC (ro + g m ro R1 + RZ )](1 + sC1RZ )
Generally Guideline:
To ensure loop stability, the unity-gain bandwidth is set to be
3-5x lower than the worst case RHP zero
The ESR zero and 2nd pole of the amplifier is placed higher than
the RHP zero
The current loop poles are usually much higher than RHP zero
11/1/2007
34
1 (1 D) RZ RHP
2 RSEN C R1
3
Since RHP
V
V
RL (1 D) = RL ( IN ) 2 = IN , BW (1 D)
VO
VO I O
2
35
11/1/2007
36
PWM Comparator
11/1/2007
37
INP
INN
COMP
11/1/2007
OUT
38
Error Amplifiers
OUT
INN
Good:
Input Common Mode Down
to Ground
Smaller Input Offset than
OTA
INP
BIAS
11/1/2007
But:
Difficult to get large Gm
39
I1
I1
R
OUT
INN
INP
I2
11/1/2007
I2
Good:
Constant Gm Defined by R
Scalable Gm by Current
Mirrors
But:
Higher Input Offset due to
Even More Current Mirrors
Additional Gm Regulation
Loop
40
11/1/2007
41