Magnetic Buck Converters for Portable Applications Frank De Stasi Mathew Jacob
1
Outline 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12.
Why use Switching Regulators? Common Device/Converter Specifications Buck Converter Analysis CCM/DCM modes Selection of L and C Synchronous Buck Converters Conduction and Switching Losses Efficiency improvement using PWM/PFM/LDO modes Control Approaches Current Mode Models and Compensation Guidelines Transient Measurement Techniques Layout Guidelines 2 © 2003 National Semiconductor Corporation
Efficiency Ig
Io
Power supply Vg + –
+ Vo
µP/DSP core
_
output DC power Po Vo I o η= = = input DC power Pg Vg I g
3 © 2003 National Semiconductor Corporation
Linear voltage regulator as power supply Series pass transistor
Iload
Q
+ Vg
+ –
C
Load
Vo –
+
-
Vref
Bandgap reference
• Simple, low noise, small footprint area • Output voltage lower than the battery voltage • High efficiency only if Vo is close to Vg 4 © 2003 National Semiconductor Corporation
Linear regulator power model Rs
Ig
Io +
Vg
+ –
Vo
Bias current IQ
–
I g = Io + IQ Efficiency:
Vo I o Vo I o ?= = Vg I g Vg ( I o + I Q )
Vo ?< Linear regulator efficiency cannot be greater Vg than the ratio of the output and the input voltage 5 © 2003 National Semiconductor Corporation
SMPS efficiency as a function of load 100 90
Example: • Vg = 3.6 V • Vo = 1.5 V • 0 < Io < 300 mA
Buck regulator
80
Efficiency [%]
70 60 50 40 . 30
Linear regulator
20 10 0 0.1
1
10
100
1000
Io [mA]
6 © 2003 National Semiconductor Corporation
Buck (step-down) switching power converter Ig
1
Low-pass LC filter
L
Io
+ Vg
+
2
+ –
vs (t)
C
v(t)
–
vs(t)
–
Vg
fs = 1/Ts = switching frequency
D' Ts
DTs
0 0 Switch position:
DTs 1
Load
Ts 2
t 1
D = switch duty cycle 7 © 2003 National Semiconductor Corporation
Buck converter ideal static characteristic vs(t)
Vg
〈vs〉 = DVg
area = DTsVg
0
0
V
DTs
Ts
t Conversion ratio:
Vg
0
Vo =D Vg 0
1
D
switch duty cycle
8 © 2003 National Semiconductor Corporation
Switch-Mode Power Supplies • Step-up, step-down and inverting configurations available • Switching converters are ideally 100% efficient • Real efficiency can be close to 100%; depends on operating conditions and implementation – Losses and efficiency will be discussed
• Converters generate switching noise • Discrete filter components (L, C) are required • Higher switching frequency => smaller L, C – Component selection will be discussed
• Duty cycle is the control variable • Closed-loop output voltage control is usually applied – Dynamic models and control will be discussed 9 © 2003 National Semiconductor Corporation
Impact of efficiency: a system example uP/DSP core mode % of time in this mode Load current Io [mA] Linear regulator
Total linear reg average Ig [mA] SMPS
Total SMPS average Ig [mA]
Efficiency [%] Battery current Ig [mA] Average Ig in this mode [mA]
Stand-by 90.0 0.1
Wait 4.0 1.0
Run1 3.0 10.0
Run2 2.5 100.0
FullRun 0.5 300.0
34.7 0.12 0.11
40.9 1.02 0.04
41.6 10.02 0.30
41.7 100.02 2.50
41.7 300.02 1.50
29.1 0.14 0.13
78.4 0.53 0.02
93.7 4.45 0.13
93.0 44.82 1.12
87.7 142.60 0.71
4.45 Efficiency [%] Battery current Ig [mA] Average Ig in this mode [mA] 2.12
Example: • Vg = 3.6 V • Vo = 1.5 V • 0 < Io < 300 mA 10 © 2003 National Semiconductor Corporation
Advantages of using SMPS over Linear regulators • SMPS results in significantly lower average battery current • High efficiency over a wide range of loads and output voltages is achieved with a SMPS • SMPS with low quiescent current modes provide longer battery life for mobile systems that spend most of their time in “stand-by”
11 © 2003 National Semiconductor Corporation
Buck regulators in the system Battery
Power distribution: Vg = 2.8-5.5 V PS
PS
3.6 V
Buck SMPS regulators
PS 2.5 V 1.5 V
Display PS
Audio
2.7-5.5 V
Interface
D/A
I/O
A/D
Baseband digital
PS
Antenna
1-3.6 V
µP/DSP core
PA LO
LNA
Analog/RF
2.5 V PS
Charger
2.5 V PS
2.5 V PS
Buck regulators are often used as switch-mode power supplies for baseband digital core and the RF power amplifier (PA) 12 © 2003 National Semiconductor Corporation
Device/Converter Specifications • Static voltage regulation – DC output voltage precision, i.e., % variation with respect to the nominal value over: • input voltage range (“line regulation”) • output load range (“load regulation”) • temperature
• Dynamic voltage regulation – “Load transient response,” including peak output voltage variation and settling time for a step load transient – “Line transient response,” including output voltage variation and settling time for a step input voltage transient 13 © 2003 National Semiconductor Corporation
Device/Converter Specifications • Overvoltage protection – prevents the output voltage from rising above a specified limit
• Undervoltage shutdown – turns the device off if the input (battery) voltage drops below a specified threshold
• Current limiting (overload protection) – limits the load current
• Thermal shutdown – turns the device off if the temperature exceeds a specified threshold
14 © 2003 National Semiconductor Corporation
Device/Converter Specifications • Frequency synchronization – allows synchronization of the switching frequency to an external system clock
• Soft start – controlled output voltage increase during startup
• Shut-down and operating-mode control – enables a system controller to shut-down the device, or to select an operating mode(PWM,PFM,LDO)
• Adjustment of the output voltage using – a resistive voltage divider, – external analog control voltage, or – digital (pin-select) control 15 © 2003 National Semiconductor Corporation
Buck converter analysis iL (t)
1
L + vL (t) –
2
+ –
Vg
+ iC(t)
C
R
v(t) –
iL(t)
L
L + vL(t) –
Vg
+ –
C
+
iC(t) R
v(t)
+ v L(t) – Vg
+ –
iL(t)
C
+ iC(t) R
–
–
Switch in position 1
v(t)
Switch in position 2 16 © 2003 National Semiconductor Corporation
Switch in position 1 iL (t) Inductor voltage:
L + vL(t) –
vL = Vg − v(t )
Vg
+ –
C
+
iC(t) R
v(t)
Small-ripple approximation:
–
vL ≈ Vg − V Knowing the voltage, we can solve for the current from: Solve for the slope:
diL vL Vg − V = ≈ dt L L
diL vL = L dt
Therefore, the inductor current increases in time with an essentially constant slope. 17 © 2003 National Semiconductor Corporation
Switch in position 2 L Inductor voltage:
+ vL (t) –
vL = −v (t )
Vg
+ –
iL (t)
+
iC(t)
C
R
v(t)
Small-ripple approximation:
–
vL ≈ −V Knowing the voltage, we can solve for the current from: Solve for the slope:
diL vL = L dt
diL vL V = ≈− dt L L
Therefore, the inductor current decreases in time with an essentially constant slope. 18 © 2003 National Semiconductor Corporation
Inductor voltage and current waveforms vL(t)
Vg – V DTs
D'Ts
t
–V Switch position: iL(t)
1
2
1
iL(DTs)
∆iL
I iL(0)
0
DTs
Ts
t 19
© 2003 National Semiconductor Corporation
Average voltage across the inductor equals zero vL(t)
Vg – V
Total area λ t
DTs –V
vL =
Ts
1 vL dt = D (Vg − V ) + (1 − D )(−V ) = 0 ∫ Ts 0 V = DVg
The DC output voltage is directly proportional to the input voltage and the switch duty cycle 20 © 2003 National Semiconductor Corporation
Average inductor current equals the output current iL (t)
1
L
IO
+ vL(t) – Vg
+ –
+
i C(t)
2
C
R
v(t) –
v (t ) V iC (t ) = iL (t ) − ≈ iL (t ) − = iL (t ) − I o R R We know that the average capacitor current equals zero Ts
iC
1 = ∫ iC dt = iL − I o = 0 Ts 0
iL = I o
In steady state, the average inductor current equals the load current 21 © 2003 National Semiconductor Corporation
Light-load operation: CCM and DCM iL (t )
high I o
without zero cross detect
t
low I o
Inductor current reverses polarity at light loads
iL (t )
high I o
with zero cross detect
t
low I o
Inductor current drops to zero before the end of the cycle: “Discontinuous conduction mode” (DCM) 22 © 2003 National Semiconductor Corporation
Implementing Zero-cross detect iL(t)
S1
L
Io
+ vL(t) – Vg + –
iC(t)
C
+ v(t)
S2
–
S2 is turned OFF
S2 control logic
S2
• With the zero-crossing comparator the switch S2 operates as a diode, resulting in DCM and improved efficiency at light loads • All switchers in the LM26XX family have this feature 23 © 2003 National Semiconductor Corporation
CCM vs. DCM • In DCM, the inductor current is always positive • At light loads, in DCM, the duty cycle is significantly lower than in CCM • CCM operation at light loads is undesirable because the reversal of the inductor current polarity contributes to conduction losses, while it does not contribute to the output load current • With a diode rectifier, DCM operation occurs automatically because of the diode characteristic • With a synchronous rectifier, DCM operation at light loads can be accomplished by turning off the NMOS switch at the zero-crossing of the inductor current
24 © 2003 National Semiconductor Corporation
DCM/CCM boundary • Boundary between constant-frequency CCM and constant-frequency DCM depends on the circuit parameters and the load • At the CCM/DCM boundary the inductor current ripple equals the output load current:
Vg − V V I o = ∆iL = = ICCM / DCM 2 Lf s Vg • If Io > ICCM/DCM, the buck converter operates in CCM • If Io < ICCM/DCM, the buck converter operates in DCM
25 © 2003 National Semiconductor Corporation
Static characteristic in DCM V D2 = Vg D 2 + 2 Lf s I o Vg
2 VL fs I o D= Vg (Vg − V )
• As the load Io in DCM decreases, the duty cycle D must decrease to keep the output V in regulation • Minimum possible on-time tp,min of the PMOS limits the minimum load current Io,min in constantfrequency PWM mode for which the output stays in regulation: If the output load current is reduced beyond Io,min the output voltage will start to rise and over voltage protection will activate.
I o ,min =
Vg − V Vg V
2L
t 2p ,min f s 26 © 2003 National Semiconductor Corporation
Determination of the inductor current ripple magnitude iL(t)
iL(DTs)
∆iL
I iL(0)
0
Ts
DTs
(change in iL) = (slope)x(length of subinterval) Current ripple magnitude
∆iL =
Vg − V 2 Lf s
D
2∆iL =
t
Vg − V L
DTs
Basic inductance selection eq.
L=
Vg − V 2∆iL f s
D 27
© 2003 National Semiconductor Corporation
iC(t) Total charge q
Output capacitor voltage ripple 1 Ts q = C ( 2∆v ) q = ∆iL 2 2 ∆i t ∆iL T /2 (neglecting esr) ( 2 ∆v ) ≈ D'T 4Cf s ∆iL ( 2 ∆v ) ≈ [2D − 1 + 8ResrCf s ] 4Cf s ∆v L
0
s
DTs
vC(t) V
s
∆v
(including esr)
t
The peak to peak output voltage ripple is the larger of the two values in the equations above.The equations can be used as capacitance selection equations if a target peak to peak output voltage ripple is known. 28 © 2003 National Semiconductor Corporation
Practice problem: selection of L and C • LM2612 is used to generate the output voltage of V = 1.5V at the max. DC output current of Io = 300 mA • The input voltage is between Vg = 2.8V and Vg = 5.5V • Select L and C so that: – the worst-case peak current ripple is ∆iL = 120 mA, and – the worst-case peak-to-peak output voltage ripple is 2∆v = 5 mV
29 © 2003 National Semiconductor Corporation
Inductor selection Vg − V V V 1 V 1 − L= D= = 2∆iL f s 2∆iL f s Vg 2∆iL f s Vg Vg − V
• LM2612 datasheet: Switching frequency is between fsmin = 468 kHz and fsmax = 732 kHz
V 1 L≥ 2∆iL f s min
1 − V V g max
= 9.7 µH
A 10µH inductor is chosen in the datasheet 30 © 2003 National Semiconductor Corporation
Output filter capacitor selection
1 ∆iL C= 4 f s ( 2 ∆v ) 1 ∆iL max C≥ = 12.8µF 4 f s min (2∆v) A 22µF ceramic capacitor is chosen in the datasheet. A 10µF capacitor can also be used with slightly higher output ripple, in case the load transient requirements are not demanding. 31 © 2003 National Semiconductor Corporation
Input current waveform iL (t)
1
+ vL (t) –
ig (t ) Vg
+ –
L +
iC(t)
2
C
R
v(t)
Cg – ig (t)
t
• Input current is pulsating, with large switching-noise component • Input filter (“decoupling”) capacitor is mandatory • to reduce the input voltage noise and ensure proper operation of the device • to prevent propagation of the switching noise to other system components 32 © 2003 National Semiconductor Corporation
Capacitor ripple currents 2∆iL Ripple Current Ratio = =r IO
(Vg − Vo) D r= L fs min IO
L = 10 µH
Vg = 3.6 V
Irms Input Capacitor = IO
Irms Output Capacitor = IO
Vo = 1.8 V
Io = 200 mA
r2 D(1 − D + ) 12 r 12
fs min = 468 kHz Io = 400 mA
r = 0.962
r = 0.481
Irms Input Capacitor = 107 mA
Irms Input Capacitor = 204 mA
Irms Output Capacitor = 56 mA
Irms Output Capacitor = 56 mA 33 © 2003 National Semiconductor Corporation
Capacitors : How small can I go ? Output Capacitor Input Capacitor
When reducing the value of output capacitors ensure proper gain and phase margins and evaluate line/load transient performance and whether it meets requirements. 34 © 2003 National Semiconductor Corporation
Switch realization with a synchronous rectifier
“Synchronous Buck”
NMOS: synchronous rectifier PMOS
PMOS: main switch
Vg
+ –
vp
iL(t)
ip(t)
Io
+ + v (t) – L
in(t) NMOS
vn
vsw(t) –
drivers
L
C
+ iC(t) v(t) –
n p
Switch control signals
Dead times are used to prevent short-circuit current through PMOS/NMOS
p n td1 td2 “dead” times
35 © 2003 National Semiconductor Corporation
Switch currents Average and RMS values
I p = i p (t ) ≈ DI o
i p (t )
I p ,rms =
i 2p (t ) ≈ D I o
t in (t )
I n = in (t ) ≈ (1 − D) I o t
I n,rms =
in2 (t ) ≈ 1 − D I o
Switch on-resistance and forward voltage drops result in switch conduction losses 36 © 2003 National Semiconductor Corporation
Conduction-loss models _
+
vON body diode
Ron,p
i p(t) + vSG
vp
ON
_
vn
ON
_
vGS
OFF
vON
+
OFF
vON
vON
NMOS: On-resistance Ron,n
+ in(t)
VD
RD _
i n(t)
_
PMOS: On-resistance Ron,p _
Ron,n
i n(t)
i p(t)
vON
+
_
+ vON body diode
_
L
in(t) +
Diode: Forward voltage drop VD in series with on-resistance RD
iL(t)
ideal L
RL winding + resistance
vL
i L(t) _
Winding resistance RL 37 © 2003 National Semiconductor Corporation
+
Buck circuit when the PMOS is ON Ron,p vON
+ + –
ideal L
RL
ip (t) _
winding + resistance
vL
iL(t) _
Io + V
Vg
_
v L = Vg − ( Ron , p + RL )iL − v ≈ Vg − ( Ron , p + RL ) I o − V
iL ≈ I o
38 © 2003 National Semiconductor Corporation
Buck circuit when the NMOS is ON ideal L
RL winding + resistance Vg
Io
_ +
R on,n
+ –
vL
iL (t)
V _
v L = −( Ron, n + RL )iL − v ≈ −( Ron , n + RL ) I o − V
ig = 0 39 © 2003 National Semiconductor Corporation
Steady-state model with conduction losses Inductor volt-second balance:
vL = 0
V = DVg − ( DRon , p + (1 − D )Ron , n + RL ) I o
Input current: I g = i g = DI o
V + Io ( Ron , n + RL ) Duty cycle considering losses D = Vg + Io ( Ron , n − Ron , p ) Equivalent steady-state circuit model with conduction losses: DRon,p + (1-D)Ron,n + RL
Ig
Io +
Vg
+ –
DIo
+ DV g –
R
V – 40
© 2003 National Semiconductor Corporation
Switching losses • Switching losses are proportional to the switching frequency • Switching loss mechanisms: – Charging/discharging of capacitance at MOSFET gates and switch node – Inductive switching transitions – Body-diode reverse recovery – Oscillator and other misc. controller losses – Inductor eddy-current and core losses
41 © 2003 National Semiconductor Corporation
Improving light-load efficiency • In PWM mode, light-load efficiency is reduced because a significant portion of switching losses does not scale with load • In PWM mode, the oscillator and the power switches are always switching at high switching frequency • Low-power modes are based on the idea of reducing the switching frequency in proportion to the load • If the switching frequency is proportional to load, high efficiency can be maintained over a very wide range of loads 42 © 2003 National Semiconductor Corporation
Switching frequency in PFM mode tp = Ipeak iL I o = iL
tp
tn
tn =
LI peak Vg − Vo LI peak Vo
Ts
1 I o = I peak (t p + t n ) f s 2
2Vo I o Vo fs = 2 1 − LI peak Vg
In PFM, the switching frequency is directly proportional to the load current 43 © 2003 National Semiconductor Corporation
Output voltage ripple in PFM Ipeak iL I o = iL
tp
( 2∆v ) ≈
I peak 2C
tn
Ts
(t p + t n ) =
2 LI peak
Vg
2C Vo (Vg − Vo )
The output voltage ripple is typically higher in PFM than in constant-frequency PWM mode 44 © 2003 National Semiconductor Corporation
PWM/PFM Combination 100% 90% 80%
Efficiency
70% 60% 50% 40% 30% 20%
Vin = 3.6V 10% 0% 0.1
1
10
100
1000
Iout in mA
LM2618 PFM
LM2618 PWM
• High efficiency over very wide range of loads • Low IQ
45
© 2003 National Semiconductor Corporation
Discussion of Operating Modes
PWM
PFM
LDO • • • •
LDO: linear regulator Low-noise Very low IQ Simple controller
• Best efficiency at moderate to heavy load • Constant-frequency, low switching noise • Synchronization to external clock possible • Relatively high IQ and poor lightload efficiency
• • • •
High efficiency over very wide load range Very low IQ Simple controller Increased output voltage ripple 46 © 2003 National Semiconductor Corporation
PWM/LDO Combination
PWM
Example: LM2608
LDO
PFM
• High efficiency (moderate-to-heavy load) • Low noise: • Constant-frequency operation • No switching noise at very light loads (LDO) • Very low IQ 47 © 2003 National Semiconductor Corporation
PWM/PFM Combination
Examples: LM2612/LM2614
PWM
LDO
PFM
• High efficiency over very wide range of loads • Low IQ 48 © 2003 National Semiconductor Corporation
Selection Guide
49 © 2003 National Semiconductor Corporation
Control approaches in constant-frequency PWM mode
• Voltage-mode control – The switch duty cycle is controlled based on output voltage sensing
• Current-mode control – The switch duty cycle is controlled based on output voltage and switch current sensing
50 © 2003 National Semiconductor Corporation
Voltage-Mode Control Architecture Power input
iL(t) +
vg (t) + –
Io Load
L
+
+ vL(t) –
iC(t)
C
vsw(t)
v(t) –
– Gate drivers p
Feedback connection
n
Dead-time
Compensator Pulse-width modulator
p(t)
vc
Voltage reference Vref
vc(t)
dTs Ts
t
v
G c (s)
t
Controller chip 51 © 2003 National Semiconductor Corporation
Current-Mode Control Architecture Power input
L
ip(t)
+ vL(t) – + C vsw(t)
vg(t) + –
Io Load
iL(t)
+
iC(t)
v(t)
–
–
Gate drivers p
Feedback connection
n Rsip (t)
Dead-time
Current-mode modulator p(t)
Compensator vc
Voltage reference Vref
v c(t)
dTs Ts
t
v
Gc(s)
t
Controller chip 52 © 2003 National Semiconductor Corporation
Current-mode summary • Advantages of current-mode control – Simpler, approximately single-pole responses – Inherent rejection of line disturbances – Built-in over-current protection
• LM26XX family is based on current-mode architecture • LM2608/12/18 feature internal compensation • LM2614/19 require external compensation 53 © 2003 National Semiconductor Corporation
Important definitions Cross-over frequency fc is the frequency where the magnitude response of the loop gain drops to 1, i.e. 0 dB T ( jwc ) = 1 → 0 dB Phase margin PM is the difference between the phase of the loop gain at the cross-over frequency and -180o
PM = phase [T ( jwc )] + 180o Gain margin GM (in dB) is the negative of the loop-gain magnitude response (in dB) at the frequency fm where the phase of the loop gain equals -180o
GM = −20 log T ( jwm ) , phase [T ( jwm )] = −180o 54 © 2003 National Semiconductor Corporation
Example of finding phase and gain margins 1
10.
100.
1000.
10000.
100 50 0
GM
-50 -100
PM
-150
10.
f c ≈ 10 KHz
100.
PM = 71o
1000.
10000.
100000.
GM = 24 dB 55 © 2003 National Semiconductor Corporation
Current Mode Power Stage Model L Vˆo Ro gm
Vˆc
ESR
R
C
Vˆo gm ⋅ ( Ro // R ) ⋅ (1 + s ⋅ ESR ⋅ C ) ≈ Vˆc (1 + s ⋅ ( Ro // R ) ⋅ C ) ⋅ (1 + s ⋅ L ) Ro 56 © 2003 National Semiconductor Corporation
Closed Loop Regulator Model L
Vˆo ESR
Ro
gm
R3 ⋅ C 4 ≈ ( Ro // R) ⋅ C
IF:
1 ≥ Fc 2 ⋅π ⋅ ESR ⋅ C
R
C
1 Vˆc
Ao
L 2 ⋅π ⋅ Ro
Vref
+ -
Rp R3 Loop Gain ≈
Fc ≈
C4
≥ Fc
gm ⋅ R o // R 2 ⋅ π ⋅ Rp ⋅ C 4
( g m ⋅ Ro // R ⋅ Ao) ⋅ (1 + s ⋅ ESR ⋅ C ) ⋅ (1 + s ⋅ R 3 ⋅ C 4 ) L (1 + s ⋅ Ro // R ⋅ C ) ⋅ (1 + s ⋅ ) ⋅ (1 + s ⋅ Rp ⋅ C 4 ⋅ Ao) Ro 57 © 2003 National Semiconductor Corporation
Compensation Example Objective: To compensate a LM2614 to get a stable system R = 10 Ω Load resistance C = 10 µF Output capacitor ESR = 10 mΩ ESR of output capacitor L = 10 µH Inductor
The load pole =
1 = 4.8 kHz 2 ⋅ π ⋅ ( R // Ro) ⋅ C
The high frequency pole = The ESR zero =
Ro = 80 kHz 2 ⋅π ⋅ L
1 = 1.6 MHz (high enough to ignore ) 2 ⋅ π ⋅ ESR ⋅ C
Ro = 5 Ω Small signal output resistance gm = 1 mho Transconductance of power stage Rp = ( R1 // R2) + 5 kΩ R1,R2 are external feedback resistor dividers,5 kO is internal
Rp = 33 kΩ Ao = 10000
Open loop gain of error amplifier
58 © 2003 National Semiconductor Corporation
Compensation Example We now need to choose the values of R3 and C4 to give a stable regulator response. If we set the zero frequency of R3 and C4 equal to the load pole frequency, and we choose a loop gain crossover frequency, Fc, much lower than the high frequency pole, then we can assume that the loop gain has a first order response. By choosing Fc = 30 kHz, the 80 kHz pole will contribute only 20 degrees of phase lag at Fc. This should give us a phase margin of about 90-tan-1(30/80) = 90-20 = 70 degrees. C4 =
gm ⋅ ( Ro // R ) = 536 pF ≈ 680 pF 2 ⋅ π ⋅ Fc ⋅ Rp
R3 =
( Ro // R ) ⋅ C = 49 k Ω ≈ 47 k Ω C4
This should give a stable regulator. Of course the real circuit should be checked under all conditions to ensure a stable system. This is only one of the methods to stabilize a regulator. Any other small signal methods that apply to feedback systems, will work here as well.
59 © 2003 National Semiconductor Corporation
Compensation guidelines Typically we like to choose a crossover frequency as high as possible. This gives a regulator with a fast transient response. However, if Fc is too close to the high frequency pole, of the power stage, the phase margin will be degraded. If we chose a Fc in the previous example of 75 kHz, then the phase margin would only be 47 degrees. Given the fact that these equations are only approximate, the phase margin of the real circuit will probably be smaller. This will give a “ringy” transient response. Lower crossover frequencies give a slower regulator, but tend to be more stable, and more “on-the-safeside”. The size of the output capacitor is also a compromise. Smaller gives more under/overshoot during a load transient and slightly higher output voltage ripple. However, with regulators that are internally compensated, smaller values of output capacitor will tend to increase Fc and therefore decrease phase margin. Large values of output capacitor will give small under/over-shoot and ripple, but are physically larger. Parts such as the LM2614, with external compensation, are much more flexible with regards to output capacitor value. In any case, it is always best to stay within the range given in the datasheet. 60 © 2003 National Semiconductor Corporation
Line Transient Measurements +15V
1000µ F
220pF(C) 10µ F 50KΩ
Pulse Generator Output
50KΩ
0.5µ H (L)
LM12CL 30Ω
DUT
10µF
50Ω
-15V
1000µ F
Adjust L and C to minimise overshoots 600mV
30µ s
61 © 2003 National Semiconductor Corporation
DUT
IRF 510
Function Generator Output
Constant Load
Pulse Load
Load Transient Measurements
50Ω
62 © 2003 National Semiconductor Corporation
Layout guidelines • Electrical guidelines – component placement and length of traces – width of traces – curling of critical current loops – routing of sensitive traces – ground pins and ground plane – voltage regulator placement on the system board
• Mechanical guidelines
63 © 2003 National Semiconductor Corporation
Critical current loops in a buck regulator Ig
L
1
Io
+ Vg
+ –
+
2
vs (t) –
C
v(t)
Load
–
The critical current loops carry large currents with significant switching ripples 64 © 2003 National Semiconductor Corporation
Component placement and length of traces
Loop 1 Loop 2
• The two critical loops carry large switching currents and act as antennas that radiate switching noise • Place C1, chip, L, and C2 as close as possible, to minimize the area of the two critical current loops 65 © 2003 National Semiconductor Corporation
Routing of sensitive traces
Route noise sensitive traces, such as the voltage feedback path, away from the critical current loops with noisy traces between power components
66 © 2003 National Semiconductor Corporation
Ground pins and ground plane Connect the chip ground pins and the filter capacitor ground pins using a large component-side fill Connect this area to the ground plane using several vias This approach prevents large switching currents from circulating through the ground plane, and reduces ground bounce to the chip 67 © 2003 National Semiconductor Corporation
Voltage regulator placement
sensitive analog/RF
Place switching regulator away from sensitive analog/RF subsystems
68 © 2003 National Semiconductor Corporation
References and Acknowledgements
• R.W.Erickson, D.Maksimovic, Fundamentals of Power Electronics, 2nd Edition, Kluwer Academic Publishers, 2000, ISBN 0-7923-7270-0 • LM26XX Data Sheets, National Semiconductor Corporation – LM2608,LM2612,LM2614,LM2618,LM2619
• Dragan Maksimovic, Associate Professor, ECE Dept, University of Colorado, Boulder, CO
69 © 2003 National Semiconductor Corporation
70