NSC LM3753SQX

LM3753/54
Scalable 2-Phase Synchronous Buck Controllers with
Integrated FET Drivers and Linear Regulator Controller
General Description
The LM3753 and LM3754 are full featured single-output dualphase voltage-mode synchronous PWM buck controllers.
They can be configured to control from 2 to 12 interleaved
power stages creating a single high power output. Both controllers utilize voltage-mode control with input voltage feedforward for high noise immunity. An internal average current
loop forces real time current sharing between phases during
load transients.
The LM3753 supports a Tracking function, while the LM3754
supports adjustable Soft-Start. The LM3753 Tracking is always enabled, so the output is controlled both up and down.
The Soft-Start function on the LM3754 can only drive the output upwards – it will not pull it down, therefore, pre-biased
loads will not be discharged. Available in the 5 mm x 5 mm
thermally enhanced 32-lead LLP package with a thermal pad.
Features
■
■
■
■
■
Wide input voltage range of 4.5V to 18V
Up to 12 channels for 300A load
System accuracy better than 1%
0.6V to 3.6V output voltage range
Switching frequency from 200 kHz to 1 MHz
■
■
■
■
■
■
■
■
■
■
■
■
Phase current sharing ±12% max over temperature
Integrated 4.35V ±2.3% LDO
Inductor DCR or sense resistor current sensing
Interleaved switching for low I/O ripple current
Integrated synchronous NFET drivers
Programmable Soft-Start (LM3754) or Tracking (LM3753)
Pre-biased startup (LM3754)
Output voltage differential remote sensing
Minimum controllable on-time of only 50 ns
Programmable Enable and input UVLO
Power Good flag
OVP, UVP and hiccup over-current protection
Applications
■ CPUs, GPUs (graphic cards), ASICs, FPGAs, Large
■
■
■
■
■
Memory Arrays, DDR
High Current POL Converters
Networking Systems
Power Distribution Systems
Telecom/Datacom DC/DC Converters
Desktops, Servers and Workstations
Simplified Application
30091901
© 2009 National Semiconductor Corporation
300919
www.national.com
LM3753/54 Scalable 2-Phase Synchronous Buck Controllers with Integrated FET Drivers and
Linear Regulator Controller
December 14, 2009
LM3753/54
Connection Diagram
30091903
Top View
32-Lead LLP
Ordering Information
Order Number
TRACK/SS
function
Package Type
NSC Package Drawing
Supplied As
LM3753SQ
TRACK
LLP-32
SQA32A
1000 Units / Reel
LM3753SQX
TRACK
LLP-32
SQA32A
4500 Units / Reel
LM3754SQ
SS
Pre-Bias Protection
LLP-32
SQA32A
1000 Units / Reel
LM3754SQX
SS
Pre-Bias Protection
LLP-32
SQA32A
4500 Units / Reel
www.national.com
2
LM3753/54
Pin Descriptions
Pin Number
Pin Name
1
HG2
Description
Gate drive of the high-side N-channel MOSFET for Phase 2.
2
SW2
Switching node of the power stage of Phase 2.
3
LG2
Gate drive of the low-side N-channel MOSFETs for Phase 2.
4
VDD
Power supply for gate drivers. Decouple VDD to PGND with a ceramic capacitor. VDD can either
be supplied by an external 5V ±10% bus, or by the internal regulator, which uses an external
NPN pass device. If using the internal regulator, connect VDD to the emitter of the NPN pass
device.
5
PGND
6
LG1
Gate drive of the low-side N-channel MOSFETs for Phase 1.
7
SW1
Switching node of the power stage of Phase 1.
8
HG1
Gate drive of the high-side N-channel MOSFET for Phase 1.
9
BOOT1
Bootstrap of Phase 1 for the high-side gate drive power supply.
10
PGOOD
Power Good open-drain output. Active HIGH.
11
SYNCOUT
Synchronization Output. For multi-controller systems this pin should be connected to the SYNC
pin of the next controller in daisy-chain configuration
12
SYNC
Synchronization Input. SYNCOUT of one controller is connected to SYNC of the next controller
in a daisy-chain fashion. To synchronize the whole chain of controllers to an external clock, wire
the external clock to the SYNC pin of the first controller of the chain (called the Master controller).
Otherwise, connect the SYNC input of the Master controller to ground and all of the controllers
will be controlled by the internal oscillator of the Master.
13
FAULT
Input/Output. Wire the FAULT pin of all controllers together. FAULT gets pulled Low during
startup, an over-current fault, or an over-voltage fault. FAULT = Low signals all controllers to stop
switching and prepare for the next startup sequence. The first LM3753/54 in the system (the
Master) supplies the FAULT pin pull-up current for all of the controllers.
14
NBASE
Connect to the base of external series-pass NPN if using the LM3753/54 internal LDO controller
to generate VDD. Otherwise leave unconnected.
15
VIN
Input Voltage. Connect VIN to the input supply rail used to supply the power stages. This input
is used to provide the feed-forward for the voltage control of VOUT and for generating the internal
VCC voltage.
16
VCC
Supply for internal control circuitry. Decouple VCC to PGND with a ceramic capacitor. When VIN
> 5.5V, the internal LDO will supply 4.35V to this pin. When 4.5V < VIN < 5.5V, connect VIN to
VCC. In this case the internal VCC LDO will turn off and VCC current will be supplied directly by
VIN.
Power Ground. Tie PGND and SGND together on the board through the DAP.
17
SGND
Signal Ground. Tie PGND and SGND together on the board through the DAP.
18
COMP
Error Amplifier Output. For the Master, a compensation network is placed between the COMP
pin and the FB pin. The COMP pin of the Master should be connected to the SNSP pin of each
of the Slaves. The COMP pin of each of the Slaves must be connected to its VDIF pin
19
FB
Feedback Input. This is the inverting input of the error amplifier. Connect the Master FB pin to
the output voltage divider and compensation network. Connect each Slave FB pin to its own VCC
pin. This will put that controller in Slave mode and disable its error amplifier.
20
VDIF
21
SNSM
Inverting input of the remote-sense differential amplifier. Connect SNSM of the Master controller
to PGND at the load point. On Slave controllers, the differential amplifier is used to buffer COMP
from the Master controller. Connect SNSM of each Slave controller directly to the Master
controller SGND pin.
22
SNSP
Non-inverting input of the remote-sense differential amplifier. Connect the SNSP of the Master
controller to VOUT at the load point. On Slave controllers, the differential amplifier is used to buffer
COMP of the Master controller. Connect SNSP of each Slave controller to the Master controller
COMP pin.
Output of the remote-sense differential amplifier. Connect the Master VDIF pin to the output
voltage divider and compensation network. The Slave differential amplifier is used to buffer
COMP from the Master controller. Connect each Slave VDIF pin to its own COMP pin.
3
www.national.com
LM3753/54
Pin Number
Pin Name
Description
23
TRACK
(LM3753)
Tracking Input. Connect the TRACK pins of all of the controllers in the system together. Wire the
TRACK pin to the external TRACK control signal. Tracking is always enabled on power-up,
shutdown and brownout.
23
SS (LM3754)
24
FREQ
Frequency Adjust. A frequency adjust resistor and decoupling capacitor are connected between
FREQ and SGND to program the switching frequency between 200 kHz to 1 MHz (each phase).
These components must be supplied on the Master and Slaves, even if the system is
synchronized to an external clock.
25
IAVE
Current Averaging. Connect a 4.02 kΩ, 1%, resistor between each controller’s IAVE pin and
SGND. In the case where one phase is not used, connect an 8.06 kΩ resistor. Connect a filter
capacitor between IAVE and SGND at each controller,
26
EN
Enable Input. Used for VIN UVLO function, connect EN to the midpoint of a voltage divider from
VIN to SGND. The EN pins of all controllers must be wired together. For an on/off EN function,
wire the EN pins of all controllers together and control with an open drain output.
27
CS2
Positive current-sense input of Phase 2. Connect to the DCR network or the current-sense
resistor of Phase 2. The negative current-sense input is the CSM pin.
28
ILIM
Current Limit Set. Connect a resistor between ILIM and CSM. The resistance between ILIM and
CSM programs the current limit.
29
CSM
Negative current-sense input of the internal current-sense amplifiers. Connect to VOUT.
30
CS1
Positive current-sense input of Phase 1. Connect to the DCR network or the current-sense
resistor of Phase 1. The negative current-sense input is the CSM pin.
31
PH
Phase Select Input. Connect this pin to the middle of a resistor divider between VCC and SGND
to program the number of phases in the system.
32
BOOT2
DAP
www.national.com
Soft-Start. Connect the SS pins of all of the controllers in the system together. At the Master
controller, connect a soft-start capacitor between SS and SGND. Only the Master controller
supplies the pull up current to the SS capacitor.
Bootstrap pin of Phase 2 for the high-side gate drive power supply.
Die Attach Pad. Must be connected to PGND and SGND but cannot be used as the primary
ground connection; do not place any traces or vias other than GND in the outer layer under the
DAP; see AN-1187 application note.
4
If Military/Aerospace specified devices are required,
please contact the National Semiconductor Sales Office/
Distributors for availability and specifications.
VIN to SGND, PGND
SGND to PGND
VCC and VDD to VIN
VDD to PGND
PGOOD, FAULT to SGND
VCC, EN, SS, TRACK,
SYNC, CS1, CS2, CSM,
ILIM, SNSM, SNSP to SGND
FREQ, PH, FB to SGND
BOOT1, BOOT2 to PGND
(Note 2)
SW1, SW2 to PGND
(Note 2)
BOOT1 to SW1,
BOOT2 to SW2 (Note 2)
SYNCOUT
PGOOD, FAULT
VDIF
COMP
ESD Rating
Operating Ratings
−0.3V to 24V
−0.3V to 0.3V
+0.3V
−0.3V to 6V
−0.3V to 6V
−0.3V to 6V
−65°C to +150°C
(Note 1)
VIN Low Range
VIN High Range when using
integrated VCC LDO
VIN High Range when using
integrated VDD linear
regulator controller
VCC External Supply Voltage
VDD External Supply Voltage
Output Voltage Range
TRACK
SYNC, EN
SNSM
SNSP to SNSM
IAVE
CS1 and CS2 to CSM
CS1, CS2, ILIM and CSM to
SGND
ILIM to CSM
Junction Temperature Range
Thermal Data
Junction-to-Ambient Thermal
Resistance (θJA), LLP-32
Package (Note 4)
−0.3 to VCC + 0.3V
−0.3V to 24V Peak
−0.3VDC to 24V Peak
−3V for less than 40 ns
−0.3V to 6.0V Peak
HBM (Note 3)
+150°C
±20 mA
±20 mA
±5 mA
±4 mA
2 kV
4.5V to 5.5V
5.5V to 18V
6V to 18V
4.5V to 5.5V
4.5V to 5.5V
0.6V to 3.6V
0V to 5V
0V to 5.5V
−0.25V to 1.0V
0V to 3.6V
0V to 1.15V
−15 mV to 45 mV
0V to 3.6V
0V to 200 mV
−5°C to +125°C
26.4°C/W
Electrical Characteristics
Limits in standard type are for TJ = 25°C only; limits in boldface type apply over the
junction temperature (TJ) range of −5°C to +125°C. Minimum and Maximum limits are guaranteed through test, design, or statistical
correlation. Typical values represent the most likely parametric norm at TJ = 25°C, and are provided for reference purposes only.
Unless otherwise stated VVIN = 12V, VVDD = 5V, VVCC = internal LDO, VEN = 2V, RFRQ = 78.7 kΩ, VPH = 0V, VCS1 = VCS2 = VCSM =
VSS = VTRACK = VSNSP = 1.8V, VILIM − VCSM = 100 mV, VSNSM = VSYNC = 0V, VSYNCOUT floating.
Symbol
Parameter
Conditions
Min
Typ
Max
Units
–0.65
–0.11
0.45
%
–0.75
–0.134
0.6
%
–0.9
–0.165
0.7
%
–2.25
–0.4
1.25
%
12
%
System Accuracy
VOUT
Output Voltage Accuracy
VOUT = 3.6V
Includes trimmed EA and diff
VOUT = 2.5V
amplifier offset and gain errors; 0.5
VOUT = 1.8V
mA load at VDIF
VOUT = 0.6V
Phase Current Equalization
ΔIPH
Current Equalization (from average VCSM = 1.8V, VCS1 = VCS2 = VCSM + 30 mV,
per phase current)
VIAVE = 740 mV, VCOMP = 1.9V
–12
System Supplies and UVLO
VIN
IVIN
VIN Operating Current
2-phase switching gate drivers unloaded
15
IVIN-Q
VIN Quiescent Current
VFB = 650 mV, no PWM switching, NBASE
is floating (no NPN)
9
18
mA
IVIN-SD
VIN Shutdown Current
VEN = 0V
200
450
µA
5
mA
www.national.com
LM3753/54
Junction Temperature
(TJ-MAX)
Storage Temperature Range
Absolute Maximum Ratings (Note 1)
LM3753/54
Symbol
Parameter
Conditions
Min
Typ
Max
Units
VVCC
VCC Linear Regulator Output
Voltage
0 to 3 mA sourced to an external load;
VVIN = 5.5V to 18V
4.25
4.35
4.45
V
IVCC
VCC Input Current from External
Supply
VVIN = 5.5V, VVCC = 5.5V
10
20
mA
VCC
IVCC-SD
VCC Input Shutdown Current from VEN = 0V, VVIN = 12V, VVCC = 5V
External Supply
IVCC-LIM
VCC Output Current Limit
VVCC = 2.5V
260
9
VVCC = 0V
VVCC-EN
VVCC-HYS
tD-VCC
VCC UVLO Thresholds
30
µA
53
mA
V
50
VVCC Rising
4.04
4.14
4.24
VVCC Falling
3.9
4
4.1
VCC Threshold Hysteresis
VCC UVLO/UVP Debounce Time
140
mV
8
µs
VDD, NBASE, BOOT1, BOOT2, SW1, SW2
VVDD
VNBASE
VNBASE-REG
VDD Controller Regulation Voltage VVIN = 6V to 18V
VIN-to-NBASE Dropout
4.6
4.85
VVIN − 5.5V, 700 mV source connected
from VDD to NBASE, INBASE = 5 mA
330
VVIN − 5.5V, 700 mV source connected
from VDD to NBASE, INBASE = 1 mA
130
5.1
V
mV
NBASE Load Regulation
VVIN = 18V, 700 mV source connected from
VDD to NBASE, INBASE steps 1 mA to 5 mA
4
mV
VDD Operating Current from
External Power Supply
VVDD = VVIN = VVCC = 5.5V, fSW = 300 kHz,
Gate Drivers unloaded
1
mA
VDD Shutdown Current
VEN = 0V, VVIN = 12V, VVDD = 5V
2
INBASE-CL
NBASE Current Limit
VNBASE = VVDD + 0.7V, ΔVVDD = −100 mV
IBOOT-SD
BOOT1, BOOT2 Shutdown Current VEN = 0V, VSW1(2) = 0V, VBOOT − VSW = 5V
4.5
IBOOT
BOOT1, BOOT2 Operating Current VBOOT1(2) = 17.0V, VSW1(2) = 12.0V, fSW =
300 kHz, Gate Drivers unloaded
650
µA
3
µA
IVDD
IVDD-SD
5.8
VNBASE = VVDD = 0V
ISW
30
10
µA
mA
20
15
µA
SW1, SW2 Leakage Current with
Pre-Biased Output
VVCC = 0V, VEN = 0V, VSW1(2) = 3.6V
VVDD-TH
VDD UVLO Thresholds
VVDD Rising
3.8
4.02
4.28
VVDD Falling
3.37
3.71
4.03
VVDD-HYS
VDD UVLO/UVP Hysteresis
310
mV
VDD UVLO/UVP Debounce Time
11
µs
160
°C
30
°C
tD-VDD
V
V
Thermal Shutdown
TJ-SD
Thermal Shutdown Threshold
TJ-HYS
Thermal Shutdown Threshold
Hysteresis
www.national.com
Rising
6
Parameter
Conditions
Min
Typ
Max
Units
EN
VEN-H
HIGH Level Input Voltage
VEN-L
LOW Level Input Voltage
VEN-TH
EN Threshold
VEN-HYS
IEN
1.14
V
VVIN = 4.5V to 18V, VVCC = 4.5V (Rising)
1.26
1.39
1.51
V
VVIN = 4.5V to 18V, VVCC = 4.5V (Falling)
1.14
1.25
1.35
V
EN Threshold Hysteresis
EN Input Bias Current
V
1.51
140
mV
VEN = 1.5V
0.1
µA
VEN = 1.0V
0.4
1.7
0.599
0.605
Reference, Feedback & Error Amplifier: FB, COMP
FB Voltage Under Regulation
VCOMP = 1.8V
VFB-REG1
VFB
FB Voltage VIN Line Regulation
VVIN = 5.5V to 18V
±0.01
%
VFB-REG2
FB Voltage VCC Line Regulation
VVCC = VVIN = VVDD = 4.5V to 5.5V (same
source)
±0.15
%
IFB
VFB-PTH
0.593
FB Input Bias Current
45
FB Pin Master/Slave Programming
Threshold
3.2
AOL
DC Gain
fBW
Error Amplifier
Unity Gain Bandwidth
130
V
nA
V
FB to COMP, VCOMP = VFB + 1.0V
70
dB
RCOMP-SGND = 1.5 kΩ, CCOMP-SGND = 50 pF
15
MHz
VCOMP-SLEW
Error Amplifier Slew Rate
6
V/µS
VCOMP-REG
COMP Load Regulation, Sourcing VCOMP = 2.7V, ΔICOMP = +1 mA, DC Gain
= 40
−3
mV
PWM Ramp and Input Voltage Feed-Forward
DMAX
Maximum Duty Cycle Controlled by VVIN = 6V, VCOMP = 3.5V
Clock
DFF
Duty Cycle Controlled by VIN Feed- VVIN = 9V, VCOMP = 2.2V
Forward
42
%
Minimum Controllable On-Time
50
ns
Ramp Minimum
1.3
V
Ramp Maximum
2.8
V
1.5
V
3
mV
30
kΩ
tON-MIN
VRAMP-MIN
PWM Ramp Range
VRAMP-MAX
VRAMP
%
81
PWM Ramp Amplitude
Differential Amplifier: SNSP, SNSM, VDIF
VOS-INPUT
RINPUT-SNSP
Input Offset Voltage
VSNSP = 1.8V
Input Resistance of SNSP
AV-DIF
Gain
fBW-DIF
3dB Bandwidth
VSNSP = 0.6V to 3.6V
0.996
1
1.004
2
V/V
MHz
VDIF-REG1
VDIF Load Regulation, Sourcing
VVDIF = 3.6V, IVDIF = 0.5 mA
−3
mV
VDIF-REG2
VDIF Load Regulation, Sourcing
VVDIF = 0.6V, IVDIF = 0.5 mA
−3
mV
Current-Sense, Current Limit and Hiccup Mode: CS1, CS2, CSM, ILIM
VCS-OS
ICS
Current-Sense Input Offset Voltage VOUT = 1.8V
Range, VCS1(2) – VCSM
CS1, CS2 Input Bias Current
±2
mV
VCSM = 3.6V, VCS1(2) − VCSM = −15 mV and
+40 mV
−200
200
nA
VCSM = 0.6V, VCS1(2) − VCSM = −15 mV and
+40 mV
−450
450
nA
240
µA
ICSM
CSM Input Source Bias Current
VCSM = 0.6V and 3.6V, VCS1(2) − VCSM = 40
mV
150
ICSL
CS1+ CS2 + CSM + ILIM Leakage VVCC = 0V, VEN = 0V, VCSM = VCS1 = VCS2
Current with Pre-Biased Output
= VILIM = 3.6V
0.1
7
µA
www.national.com
LM3753/54
Symbol
LM3753/54
Symbol
fBW-CS
IILIM-SOURCE
Parameter
Conditions
Min
3dB Bandwidth, CS1(2) to PWM
COMPARATOR Input
Typ
Max
1.0
Units
MHz
ILIM Source Current
VILIM = 0.6V to 3.6V, VVIN = 5.5V
85
94
103
µA
VCL
Current Limit Threshold Voltage
VILIM − VCS1(2)
VILIM = 0.6V to 3.6V, VVIN = 5.5V
−2.5
0
4.6
mV
tD-CL
Current Limit Comparator
Propagation Delay
VCS1 or VCS2 stepped from 0.9V to 1.1V,
VILIM = 1V
tD-ILIM
tD-HICCUP
200
ns
Master or Slave Fast Current Limit VFB = 280 mV, 1-phase over-current:
Delay
VCS1 OR VCS2 > VILIM
7
Switch
cycles
VFB = 280 mV, 2-phase over-current:
VCS1 AND VCS2 > VILIM
3
Switch
cycles
1-phase over-current:
VCS1 OR VCS2 > VILIM
446
Switch
cycles
2-phase over-current:
VCS1 AND VCS2 > VILIM
223
Switch
cycles
6
ms
Master or Slave Over-Current
Hiccup Mode Delay
tD-COOL-DOWN Hiccup Over-Current Cool-Down
Time
Power Good: PGOOD, OVP, UVP
VOVP
OVP Threshold
VFB rising edge
125
130
tD-RESTART
OVP Restart Delay
2
NOVP-LATCH
Number of OVP Events Before
Latch-Off
7
VUVP
UVP Threshold
VUVP-HYS
UVP Threshold Hysteresis
tD-OVP/UVP
OVP/UVP Debounce Time
VFB falling edge
75
80
135
%VFB
ms
85
25
%VFB
mV
5
µs
VPG-LO
PGOOD Low Level
IPGOOD = −4 mA
0.14
0.25
V
IPG-LEAK
PGOOD Leakage Current
VPGOOD = 5.5V
5
300
nA
FAULT
IFAULT
Internal Pullup Current in Master
Mode
325
µA
VOL-FAULT
FAULT Output Low Level
IFAULT sinking 500 µA
0.21
V
VOH-FAULT
FAULT Output High Level
IFAULT sourcing 50 µA
VCC −
0.1
V
Oscillator and Synchronization (PLL): SYNC, SYNCOUT, FREQ
fSW-MIN
Minimum Switching Frequency
RFRQ = 121 kΩ
200
kHz
fSW-MAX
Maximum Switching Frequency
RFRQ = 21.3 kΩ
1000
kHz
fSW
Switching Frequency Accuracy
RFRQ = 78.7 kΩ
SYNC Frequency Capture Range
200 kHz to 1 MHz
fSYNC
VSYNC-RISE
SYNC Rising Threshold
VSYNC-FALL
SYNC Falling Threshold
tSYNC-MIN
ISYNC
300
1.12
SYNC Minimum Pulse Width
VSYNC = 0 to 5.5V
VSYNCOUT-HI
SYNCOUT Logic High Level
Sourcing 10 mA, VVCC = 4.5V external
VSYNCOUT-LO
SYNCOUT Logic Low Level
Sinking 10 mA, VVCC = 4.5V external
8
318
±25
1.46
SYNC Bias Current
(internal or external VCC)
www.national.com
282
−15
kHz
%
1.68
V
1.3
V
150
ns
25
µA
V
VCC −
0.42
0.48
V
Parameter
Conditions
PHRATIO
VPH/VVCC Divider Ratio to Set
Phase Number
2 & 4 Phases
IPH
PH Bias Current
Min
Typ
Max
0
0.138
3 Phases
0.152
3/14
0.279
5 Phases
0.294
5/14
0.418
6 Phases
0.438
7/14
0.562
8 Phases
0.587
9/14
0.703
10 Phases
0.730
11/14
0.844
12 Phases
0.874
1
VVCC = 4.5V forced, VPH = 0 to VVCC
−150
150
Units
nA
ΦHG1-N2
HG1 to HG2 Phase Shift for 2, 4, 6,
8, 10 or 12-Phase Modes
180
°
ΦHG1-N3
HG1 to HG2 Phase Shift for 3Phase Mode
240
°
ΦHG1-N5
HG1 to HG2 Phase Shift for 5Phase Mode
216
°
N>2
360/N
°
N=2
90
ΦSYNC
SYNC to SYNCOUT Phase Shift
for N-phase Operation
tSYNC-ERR
SYNC to SYNCOUT Phase Shift
Error
tSYNC-HG
SYNC to HG1(2)
ΦHG-ERR
HG1 and HG2 Controller-toController Phase Delay Error
300 kHz, 6-phase
5
ns
165
ns
5
°
Soft-Start (LM3754): SS, Pre-Biased Startup
ISS
SS Source Current
VSS = 0.3V
5.7
10
14.6
µA
RDS-SS
Soft-Start Pull-Down Resistance
750
Ohm
tLG-PW1
First LG High Pulse Width during
Soft-Start
460
ns
tLG-GT
LG Asynchronous-to-Synchronous
Gradual Transition Time
2
ms
2
ms
tD-EN-SW
EN-to-Switching Delay
Delay from EN = High to FAULT = High; no
pre-bias
Tracking (LM3753): TRACK
VTRACK
VHYS-TRACK
Tracking Range
0
TRACK Falling Voltage Hysteresis
tSS-INT
Internal SS Time during Fault
Recovery
After Fault
ITRACK
TRACK Input Bias Current
VTRACK = 0.3V
VTRACK = 5V
VREF
V
50
mV
3.8
ms
5
200
nA
0.2
mA
tLG-PW1
First LG = High Pulse Width during
Fault Recovery
460
ns
tLG-GTF
LG Asynchronous-to-Synchronous
Gradual Transition Time during
Fault Recovery
1.8
ms
tD-EN-SW
EN-to-Switching Delay
2
ms
Delay from EN = High to FAULT = High; no
pre-bias; VTRACK = 0.6V
9
www.national.com
LM3753/54
Symbol
LM3753/54
Symbol
Parameter
Conditions
Min
Typ
Max
Units
Gate Drivers
IPK-HG-SOURCE HG1 and HG2 Peak Source
Current
RHG-SOURCE
IPK-HG-SINK
RHG-SINK
Less than 100 ns
HG1 and HG2 Source Resistance VBOOT − VSW = 5V
1.9
A
2.5
Ω
HG1 and HG2 Peak Sink Current
Less than 100 ns
4
A
HG1 and HG2 Sink Resistance
VBOOT − VSW = 5V
1
Ω
IPK-LG-SOURCE LG1 and LG2 Peak Source Current Less than 100 ns
RLG-SOURCE
LG1 and LG2 Source Resistance
IPK-LG-SINK
LG1 and LG2 Peak Sink Current
A
2
Ω
4
A
1
Ω
RHG-PULLDOWN HG-SW Pull-Down Resistor
16
kΩ
RLG-PULLDOWN LG-PGND Pull-Down Resistor
16
kΩ
30
ns
28
ns
10
ns
RLG-SINK
Less than 100 ns
2.3
LG1 and LG2 Sink Resistance
tD-HG-LG
HG Falling to LG Rising CrossConduction Protection Delay
(Dead-Time)
tD-LG-HG
LG Falling to HG Rising Delay
tDS-HG-LG
HG Falling to LG Rising CrossConduction Protection Delay
(Dead-Time)
SW node not switching
SW node switching
Note 1: Absolute Maximum Ratings indicate limits beyond which damage to the device may occur, including inoperability and degradation of device reliability
and/or performance. Functional operation of the device and/or non-degradation at the Absolute Maximum Ratings or other conditions beyond those indicated in
the Recommended Operating Conditions is not implied. Operating Range conditions indicate the conditions at which the device is functional and the device should
not be operated beyond such conditions. For guaranteed specifications and conditions, see the Electrical Characteristics table.
Note 2: Peak is the dc plus transient voltage including switching spikes.
Note 3: Human Body Model (HBM) is a 100 pF capacitor discharged through a 1.5 kΩ resistor into each pin. Applicable standard is JESD22-A114C. All pins
pass 2 kV HBM except VDD, VIN and VCC which are rated for 1.5 kV.
Note 4: Tested on a four layer JEDEC board. Four vias provided under the exposed pad. See JEDEC standards JESD51-5 and JESD51-7.
www.national.com
10
LM3753/54
Typical Performance Characteristics
System Accuracy vs VOUT
fSW vs Temperature
30091904
30091907
VREF Deviation
RFRQ vs fSW
30091905
30091908
VREF vs Temperature
Load Step (High Slew)
30091909
30091906
11
www.national.com
LM3753/54
LM3754 (SS) Startup from 0V
Over-Voltage Fault
30091910
30091913
LM3754 (SS) Pre-Biased Output Startup
Repeated Over-Voltage Conditions
30091911
30091914
LM3753 (TRACK) Startup
Over-Current Fault (Soft Short)
30091912
www.national.com
30091915
12
LM3753/54
30091916
Block Diagram
13
www.national.com
LM3753/54
phases. An external resistor to ground translates this current
signal to a voltage, which all of the controllers read back.
The LM3753/54 includes an uncommitted differential amplifier. On the Master controller this amplifier is used to remotely
sense the converter’s output voltage, typically at the load. On
the Slave controllers this amplifier is used to buffer the Master
controller’s COMP signal and level shift it to the Slave
controller’s local ground.
Functional Description
GENERAL
The LM3753 and LM3754 are two-phase voltage-mode stepdown (buck) switching regulator controllers. From one to six
LM3753/54 controllers can be connected together to control
from two to twelve phases (2, 3, 4, 5, 6, 8, 10, or 12 phases).
Since external switching components can typically handle
25A per phase, a 12 phase system can supply a total of 300A.
Multiple controllers in a system communicate with each other
and work together. They will startup and shut down together,
each phase on each controller will share current equally, and
all the phases will react in unison to fault conditions. In a multicontroller system, all controllers are the same part (all
LM3753 or all LM3754). One controller functions as the Master and all the others act as Slaves. The Master and Slave are
differentiated by how they are connected in the system. The
Master controller senses the system output voltage and VIN
(as well as SS or TRACK) and sets the target duty cycle for
each phase on all of the controllers. The Master and Slave
controllers monitor the current-sense information from each
phase. Based on this current information, the controllers adjust the duty cycle on each phase up or down from the target
level, in order to achieve optimal current sharing.
Each controller incorporates a phase locked loop (PLL) that
communicates with the PLLs on the other controllers. By this
means, the switching edges of the different phases are
spread out equally within one switch period. For N phases
operating at any switching frequency, the angle in degrees
between one phase switching and the next is 360° / N. A
SYNC pin is available that can be used to lock the Master
switching frequency and phase to an external clock.
The LM3753 has a Tracking function. The output voltage will
follow the TRACK pin, both up and down, whenever it is less
than VREF. Synchronous switching is always enabled, except during fault recovery.
The LM3754 has a Soft-Start function. The Master controller
sources 10 µA out of the SS pin so that the output voltage rise
time is controlled by the size of the external SS capacitor. The
LM3754 will not pull down a pre-biased load. The synchronous NFET switch is not turned on during the soft-start
cycle until the SS ramp exceeds either the FB voltage or the
internal reference voltage VREF. At this point a gradual transition to synchronous switching is initiated.
POWER CONNECTIONS
The LM3753/54 has three supply pins, which are VIN, VCC,
and VDD. It employs two ground pins, SGND and PGND.
VDD and PGND are the power and ground for the gate driver
stage that controls the HG and LG pins. The quiescent current
drawn by VDD is very small – around 1 mA. To predict the
VDD current requirement one can assume it is mostly switching current and use the standard formula:
IVDD = (1 or 2) x fSW x QTOTAL_PHASE
QTOTAL_PHASE is the sum of the high-side switch gate charge
and the low-side gate charge. The (1 or 2) factor corresponds
to one or two phases running. The low-side driver is powered
directly from VDD. The high-side driver draws its power from
VDD through the external bootstrap Schottky diode. The rest
of the controller is powered by VCC and SGND.
The LM3753/54 has two on-board regulators, one to generate
VCC and one to generate VDD. The VCC regulator is selfcontained and only needs a 4.7 μF ceramic capacitor to
SGND. The VDD regulator uses an external NPN pass device. This device should be sized to meet the VIN to VDD
dropout requirements for the calculated IVDD. The collector of
this device goes to VIN, the base goes to NBASE and the
emitter goes to VDD. VDD also needs a 4.7 µF bypass capacitor to PGND. The internal VIN to NBASE dropout is
approximately 300 mV. The minimum VIN is calculated as:
VINMIN = VDDMIN + VBE_NPN + 300 mV
VDDMIN = MAX(VDDUVLO, VGATE-MIN)
VDDUVLO is the controller’s maximum VDD under-voltage
lockout voltage, which is 4.06V. VGATE-MIN is the minimum required gate drive voltage for the power MOSFET switches.
VINMIN is typically 5.5V to 6.0V. For VIN less than 5.5V, the
regulators are omitted and the VCC and VDD pins are connected as shown in Figure 3.
CONTROL ALGORITHM
The control architecture is primarily voltage-mode. An error
amplifier amplifies the difference between the FB pin voltage
and the internal reference voltage to generate a COMP signal.
This signal is compared against a ramp that consists of a fixed
value plus a term proportional to VIN which controls the duty
cycle. In order to facilitate current sharing there is an inner
current-sense loop. Information for the current through the inductor in each phase is sensed either with a sense resistor or
with a DCR arrangement which uses the DC resistance of the
inductor. This current-sense signal is connected to the CS pin
(CS1 or CS2). The negative reference for current-sense is
VOUT which is common for both phases and connected to the
controller’s CSM pin. The controller amplifies the (CS1(2) –
CSM) voltage difference for each phase, and compares it to
the voltage on the IAVE pin, which tracks the average current
of all phases. Any phase whose current is more than the average has its duty cycle decreased and vice versa. The IAVE
signal is common to all controllers in a system. Each controller
outputs a current onto the IAVE bus so that the total current
on the bus is the sum of the current signals from all of the
www.national.com
30091919
FIGURE 1. Power Connections Using the Internal
Regulator
14
30091924
30091921
FIGURE 4. Input Voltage UVLO with External Enable
FIGURE 2. Power Connections Using a System 5V Rail
While the EN pin has a threshold hysteresis of 140 mV typical,
a small noise-filtering capacitor may be added between the
EN pin and SGND. This is particularly useful when the controller is turning on via the resistor divider by a slowly rising
VIN rail.
STARTUP SEQUENCE
During the initial startup phase the LM3753 and LM3754 behave identically. When EN is below its threshold, the internal
regulators are off and the controller is in a low power state.
When EN crosses above its threshold the VCC regulator turns
on. When VCC rises above its under-voltage lockout threshold the VDD regulator turns on. When VDD rises above its
under-voltage lockout threshold the controller is ready to start.
If VDD or VCC is supplied externally and already sitting above
its under-voltage lockout point, then the controller is ready for
startup as soon as EN crosses above its threshold. Anytime
VCC or VDD drops below its UV threshold, switching stops
and the controller goes into a standby state. It will go through
normal startup once the supplies recover.
When the controller is ready to start, it reads the voltage on
the PH pin and determines how many phases are running in
the system. By this means the phase delay from SYNC to
SYNCOUT through the PLL is configured. Following this the
oscillator and PLL turn on and pulses will be observed on
SYNCOUT.
A 2 ms timer is initiated so that all of the PLLs in the system
can synchronize up. As each controller times out, it stops
pulling its FAULT pin low. At the end of this sequence, the
FAULT bus rises and the controllers are ready to switch.
On both the LM3753 and LM3754, the error amplifier uses a
different input stage when TRACK/SS is below VREF. During
normal operation the error amplifier employs a low offset
bipolar input stage. At startup, the input bias current of this
stage is large enough in relation to the soft-start current to
affect the soft-start timing. A MOS input stage is used during
the soft-start or track phase which has a lower input bias current but a higher input offset voltage. A 40 mV offset is
introduced when TRACK/SS is less than 70 mV. This offset
forces the error amplifier output to be low during startup. The
offset transitions progressively to zero as TRACK/SS moves
from 0 to 70 mV.
30091923
FIGURE 3. Power Connections for VIN = 5V
UNDER-VOLTAGE LOCKOUTS and ENABLE
The LM3753/54 controller has internal under-voltage lockout
(UVLO) detection on the VCC and VDD supplies. The undervoltage lockout on VIN is set using the EN pin threshold.
Connect a voltage divider between VIN and SGND with the
midpoint going to the EN pin. The division ratio and the EN
pin threshold determine the VIN level that enables the controller. This divider should be used in all cases. If the system
does not have a particular VIN under-voltage lockout requirement, the level is set to be below the minimum VIN level at
the worst case combination of tolerances and operating conditions.
To guarantee startup at the lowest input voltage, set the divider to the VEN-TH rising max specification. For a higher
accuracy VIN UVLO operation, the resistor divider minimum
current should be 1 mA or higher. This will reduce the threshold error contribution of the EN pin bias current, which is
guaranteed to be less than 1.7 µA over temperature. The enable pin can also be used as a digital on-off. To do this, the
enable signal should be used to pull down the midpoint of the
TRACKING (LM3753)
The LM3753 implements a tracking function. The error amplifier amplifies the minimum of VREF or TRACK at the FB
15
www.national.com
LM3753/54
voltage divider using open-drain logic or a transistor. A customary implementation uses an external MOSFET.
LM3753/54
pin. By means of the closed loop regulation through the
switching stage, FB will be regulated to TRACK. When
TRACK is below VREF, the LM3753 will control FB both up
and down to follow TRACK. When TRACK is above VREF,
FB will be regulated to VREF. A pre-biased output will be
pulled down by the LM3753. Full synchronous switching is
always employed on the LM3753, except for restart after a
fault condition.
When the LM3753 is ready to switch, normally TRACK will be
grounded and COMP will be low. LG will get pulled to VDD to
turn on the synchronous switch. As TRACK slews above FB,
COMP will slew up and LG will go high for 300 ns to charge
the HG bootstrap capacitor. Following this HG begins switching. COMP will set the duty cycle with normal PWM control of
HG and LG. The loop acts to have FB follow TRACK. If
VOUT is too high, it will get pulled down. An internal timer sets
a 2 ms delay from the time of the first HG pulse, which occurs
as soon as COMP slews above the PWM ramp bottom.
When the 2 ms times out, PGOOD goes high if FB is above
the output under-voltage threshold on the Master, TRACK is
above VREF, no fault conditions are present, and SYNC is
toggling on the Slaves.
corresponding phase delays per Table 1. Each controller requires the same resistor divider at the PH pin.
30091925
FIGURE 5. Phase Selection
TABLE 1. Phase Divider Resistors
SOFT-START (LM3754)
The LM3754 implements a soft-start function, and operates
so as to prevent discharge of a pre-biased output. The error
amplifier amplifies the minimum of VREF or SS at the FB pin.
By means of the closed loop regulation through the switching
stage, FB will be regulated to SS. The Master controller
sources 10 µA onto the SS pin, while the Slaves do not source
any current. This sets the total soft-start current in a multicontroller system to 10 µA.
The SS pin is automatically pulled down to SGND prior to the
onset of switching and during a restart from a fault condition.
When SS is initially released, COMP is low and no switching
occurs. Both LG and HG are held low while SS is below FB,
which guarantees that a pre-biased load will not be pulled
down. When SS crosses above either FB or VREF, COMP
will slew up and switching will start. The first switching pulse
is a 300 ns LG pulse to charge the external HG bootstrap
capacitor. After this the LG pulse width is reduced to zero.
This insures that VOUT does not get pulled down while COMP
slews up and the system loop is settling. Pulses on HG cause
the high-side FET to turn-on so that FB tracks the SS pin as
it slews up. During the switch cycle off-time the inductor current can only flow through the body diode of the synchronous
switch. During each successive cycle the LG pulse width
gradually increases. Over the course of 0.3 ms to 2.0 ms, depending on the amount of pre-bias, LG pulses get longer until
full synchronous switching occurs. The internal timer waits 2
ms, regardless of duty cycle, for this transition in LG pulse
width to complete.
Following this PGOOD goes high if FB is above the output
under-voltage threshold on the Master, SS is above VREF,
no fault conditions are present, and SYNC is toggling on the
Slaves.
Divide Ratio
Target
RPH1
(± 1%)
RPH2
(± 1%)
2 & 4 Phases
0.000
Omit
0
3 Phases
0.214
7870Ω
2150Ω
5 Phases
0.357
6490Ω
3570Ω
6 Phases
0.5
4990Ω
4990Ω
8 Phases
0.643
3570Ω
6490Ω
10 Phases
0.786
12 Phases
1
2150Ω
0
7870Ω
Omit
OVER-CURRENT and OVER-VOLTAGE FAULTS
If any controller experiences a fault condition, it will pull the
FAULT bus low and all of the controllers will stop switching.
From the time when EN is low to the point where FAULT rises,
both HG and LG are low so that the SW node of each phase
is floating. The FAULT input may be pulled low externally
through an open drain MOSFET to disable the system.
The LM3753/54 employs cycle-by-cycle current limiting. This
occurs on each phase for both Master and Slave controllers.
The current (that is the CS1(2) − CSM voltage) is continuously
compared to the over-current set point (ILIM − CSM). Any
time that the current-sense signal exceeds current limit, the
cycle is ended.
In order to determine that a current fault has occurred, each
controller counts the number of over-current pulses. When
the sum of the counts for phase 1 and phase 2 reaches 446
an over-current fault is declared. The counter is reset after 16
consecutive switching cycles with no over-current on either
phase.
There is a second method for achieving an over-current fault,
which is meant to react to heavy shorts on VOUT. The Master
controller will determine that an over-current fault has occurred after 7 over-current cycles if the voltage at the FB pin
is less than 50% of its target value. This feature is disabled
during startup. Since the Slave controllers do not see the FB
voltage, they cannot detect this type of fault.
Any controller which sees an over-current fault will respond
by pulling the FAULT bus low. All of the controllers will react
and stop switching. Both HG and LG on each phase will be
pulled low. The inductor current in each phase will decay
through the body diodes of the low-side switches. The controller which recognized the over-current fault will hold
FAULT low for 6 ms, which determines the hiccup time. This
allows the energy stored in the inductors to dissipate. After
this, FAULT is released and all of the controllers will restart
together.
PHASE NUMBER SELECTION
The voltage at the PH pin determines the phase shift between
the two phases of each controller and also the phase shift
between the SYNC and SYNCOUT pulses in a Master-Slave
configuration. This voltage is read at startup and the resulting
phase configuration saved. The PH pin should be connected
to the center of a resistor divider between VCC and SGND to
select and program the required number of phases and the
www.national.com
Number Of
Phases
16
load current-sense signal of at least 25 mV. This is typically
a resistor in the 1 mΩ to 2 mΩ range. The current-sense resistor is inserted between the inductor and the load. The load
side of the resistor which is VOUT, is connected to CSM, the
negative current-sense input. This is the negative currentsense reference for both phases. The positive side of the
current-sense resistor goes to CS1(2).
For the DCR configuration a series resistor-capacitor combination is substituted for the current-sense resistor. The resistor connects to the switch node (SW) and the capacitor
connects to VOUT. CSM is connected to VOUT as with the
sense resistor. CS1(2) is connected to the center point of the
resistor and capacitor, so that the current-sense signal is developed across the capacitor. The voltage across the capacitor is a low pass filtered version of the voltage across the
resistor-capacitor combination, in the same way the current
through the inductor is a low pass filtered version of the voltage applied across the inductor and its intrinsic series resistance. Choose the DCR time constant (RDCR x CDCR) to be
1.0 to 1.5 times the inductor time constant (L / RL). RDCR is
selected so that the CS pin input bias current times RDCR does
not cause a significant change in the CS voltage. The inductor
time constant and the DCR time constant will skew over temperature since the components have different temperature
coefficients. Critical applications may employ a correction circuit based on a positive temperature coefficient thermistor
(PTC).
The over-current limit is set by placing a resistor between ILIM
and CSM. The value of the resistor times the ILIM current of
94 µA sets the over-current limit.
CURRENT SHARING and CURRENT AVERAGING
The current sharing works by adjusting the duty cycle of each
phase up or down to make the phase current equal to the
average current. The maximum duty cycle shift is ±20%.
To determine the average current, each phase sources a current onto the IAVE bus proportional to its load current as
measured by the current sense amplifier connected to the
CS1(2) and CSM pins. The IAVE pins of all controllers are
connected together and a resistance of 8 kΩ per phase (parallel) to SGND provides the proper voltage level for the IAVE
bus. Each phase compares its current sense output to the
IAVE bus and sums the resultant voltage into the common
COMP signal to adjust the duty cycle for optimum current
sharing.
IAVE forms the current sharing bus for the entire power converter. The IAVE pins of all controllers must be connected
together. Filter capacitors with a time constant of RAV x CAV =
1 / fSW are connected between IAVE and SGND of each controller. The parallel combination of the filter capacitors times
the summing resistors (one set per controller) forms the time
constant of the current sharing bus.
PGOOD and PGOOD DELAY
PGOOD is an open-drain logic output. It is asserted HIGH
when the output voltage level is within the PGOOD window,
which is typically −20% to +30%. In order to operate, the
PGOOD output requires a pull-up resistor to an appropriate
supply voltage. This voltage is typically the supply for an external monitoring circuit. The resistor is selected so that it
limits the PGOOD sink current to less than 4 mA.
PGOOD is delayed from either power-up or VIN under-voltage lockout, and has three primary factors:
1) A synchronization delay, set to 2 ms after the slowest
controller in the system recognizes a valid level on EN, VCC
and VDD. This delay is timed out internally and allows for the
phase lock loops to synchronize.
2) Soft-Start/Track up, in non-fault conditions.
3) Transition period from diode emulation mode to fully
synchronous operation, set to 2 ms.
CURRENT SENSE and CURRENT LIMIT
The LM3753/54 senses current to enforce equal current sharing and to protect against over-current faults. There are two
system options for sensing current; a current-sense resistor,
or a DCR configuration which uses the DC resistance of the
inductor. The current-sense resistor is more accurate but less
efficient than the DCR configuration.
The input range of the differential current-sense signal (CS1
(2) – CSM) is from −15 mV to +40 mV. The common mode
range is the same as the controller’s output range which is 0V
to 3.6V. Two considerations determine the value of the current-sense resistor. If the resistor is too large there is an
efficiency loss. If it is too small the current-sense signal to the
controller will be too low. Choose a resistor that gives a full
ERROR AMPLIFIER and LOOP COMPENSATION
The LM3753/54 uses a voltage mode PWM control method.
This requires a TYPE III or 3 pole, 2 zero compensation for
optimum bandwidth and stability. The error amplifier is a voltage type operational amplifier with 70 dB open loop gain and
unity gain bandwidth of 15 MHz. This allows for sufficient
phase boost at high control loop frequencies without degrading the error amplifier performance.
The error amplifier output COMP connections are different for
Master and Slave controllers. For the Master, a compensation
network is placed between the COMP pin and the FB pin. The
COMP pin of the Master is connected to the SNSP pin of each
Slave. The SNSM pin of each Slave is connected to the bot-
17
www.national.com
LM3753/54
The restart after fault process for the LM3754 is the same as
the initial startup process. SS is pulled low and the system will
go through a full soft-start cycle. Switching will resume when
SS crosses above FB.
The restart after fault for the LM3753 is different from the initial
startup. When an over-current fault occurs, TRACK is usually
above VREF. In order to avoid VOUT slewing up precipitously,
a fixed time internal soft-start is connected to the error amplifier to control the rise of VOUT. The low-side switch is not
turned on until the internal SS exceeds FB or VREF, which
allows V OUT to remain high. The error amp will use as a reference the minimum of VREF, TRACK or the internal SS.
Once switching ensues a gradual transition to fully synchronous operation occurs.
Over-voltage faults are only recognized by the Master controller. About 5 µs after FB crosses above the OVP threshold,
which is 30% above VREF, the Master controller declares an
over-voltage fault. It pulls the FAULT bus low and all of the
controllers stop switching, with HG being low and LG being
high. The low-side MOSFETs pull VOUT down to remove the
over-voltage condition. As soon as FB crosses below the under-voltage detect point, which is 20% below VREF, the LG
outputs go low to turn off the low-side MOSFETs. This prevents the negative inductor current from ramping too high.
The Master controller then waits 2 ms to allow any negative
inductor current to transition into the high-side MOSFETs
body diodes.
The restart from an over-voltage fault is the same as the
restart from an over-current fault. In addition there is an overvoltage fault counter. On the seventh over-voltage fault, the
system does not restart. It waits for power or EN to be cycled.
This counter is reset to zero when power goes low or EN
crosses below its threshold.
LM3753/54
tom of the Master feedback divider at SGND. The COMP pin
of each Slave is connected to its corresponding VDIF pin. This
provides sufficient buffering of the master COMP signal for
the internal summing of the current averaging circuit.
NFET SYNCHRONOUS DRIVERS
The LM3753/54 has two sets of gate drivers designed for
driving N-channel MOSFETs in a synchronous mode. Power
to the high-side driver is supplied through the BOOT pin. For
the high-side gate HG to turn on the high-side FET, the BOOT
voltage must be at least one VGS greater than VIN. This voltage is supplied from a local charge pump which consists of a
Schottky diode and bootstrap capacitor, shown in Figure 6.
For the Schottky, a rating of at least 250 mA and 30V is recommended. A dual package may be used to supply both
BOOT1 and BOOT2 for each controller.
Both the bootstrap and the low-side FET driver are fed from
VDD. The drive voltage for the top FET driver is about VDD
− 0.5V at light load condition and about VDD at normal to full
load condition.
OSCILLATOR and SYNCHRONIZATION
A resistor and decoupling capacitor are connected between
FREQ and SGND to program the switching frequency between 200 kHz to 1 MHz. These components must be supplied on each controller, even if the system is synchronized
to an external clock.
The switching frequency and synchronization are controlled
by the Master. The Master can switch in a free-running mode
or be synchronized to an external clock. To synchronize the
Master apply the external clock to the SYNC pin of the Master,
otherwise ground this pin. The amplitude of the signal on the
SYNC pin must be limited to be between 0V and VCC.
The value of the frequency setting resistor is determined as:
A 1000 pF ceramic capacitor is used to provide sufficient decoupling. If the Master is synchronized set the resistor according the nominal applied frequency. If the signal on the
SYNC pin is below 150 kHz the signal will be ignored and the
device will revert to free-running mode. The SYNCOUT signal
from the Master is applied to the first Slave’s SYNC pin. The
SYNCOUT pin of the first Slave is connected to the SYNC pin
of the second Slave, and so on, in a daisy chain configuration.
SYNCOUT of the last Slave (or the Master in a single controller system) is left unconnected.
The configuration of the system, namely the number of controllers and phases is programmed by the voltage on the PH
pin. For each controller connect the midpoint of a resistor divider between VCC and SGND to the PH pin. The division
ratios are given in the Electrical Characteristics table and
nominal resistor values in Table 1. This sets the phase shift
between SYNC and the SYNCOUT pin. Where an even number of phases (N) are employed, the phase delay from SYNC
to SYNCOUT is 360°/N. The phase difference between the
two phases on the same controller is 180°. For systems with
an odd number of the phases, the HG2 and LG2 gate drivers
on the last Slave are unconnected and the phase arrangement is set according to Table 1
30091926
FIGURE 6. Bootstrap Circuit
REMOTE SENSE DIFFERENTIAL AMPLIFIER
The differential amplifier connected internally to the SNSP,
SNSM and VDIF pins is a single stage unity gain Instrumentation amplifier. The differential gain is tightly controlled to
within 0.4%.
DUTY CYCLE LIMITATION
The minimum controllable on-time is typically 50 ns. This limits the maximum VIN , VOUT and fSW combination.
30091927
fSW < (VOUT / VIN) x 20 MHz
FIGURE 7. Differential Amplifier
The maximum guaranteed duty cycle is 81%. This limits the
minimum VIN to VOUT ratio.
On the master controller, the differential amplifier is used to
provide Kelvin sensing of the output voltage at the load. This
provides the most accurate sampling for load regulation.
On the slave controllers, the differential amplifier is used to
sense the COMP signal of the master controller with respect
to its signal ground and drive the COMP pin of that slave controller relative to its local signal ground. This allows the master
controller to accurately provide the target duty cycle of the
slave controllers.
The differential amplifier has a low output impedance to allow
it to drive the COMP pins of the Slave controllers. This is necessary because the current sense signal is internally added
(VOUT / VIN) x 1.25 < 0.81
The 1.25 term allows margin for efficiency and transient response.
THERMAL SHUTDOWN
The internal thermal shutdown circuit causes the PWM control circuitry to be reset and the NFET drivers to turn off all
external power MOSFETs. The controller remains enabled
and all bias circuitry remains on. After the die temperature
falls below the lower hysteresis point, the controller will
restart.
www.national.com
18
full accuracy of the LM3753 regulation. To meet this requirement the tracking voltage is offset by 150 mV. The output
voltage will reach its final value at 80% of the external supply
voltage. The tracking resistors are determined by:
Application Information
NUMBER of PHASES
The number of phases can be calculated by dividing the maximum output load current by 25A. Therefore a 120A load
requirement will need at least 5 phases, or 3 controllers. It
may be better to use 6 phases which will still require 3 controllers, but will reduce the maximum current/phase to 20A.
Increasing the number of phases will also reduce the output
voltage ripple and the input capacitor requirements. Note that
the 25A/phase is dictated by external components and not by
the LM3753/54. After the number of phases has been chosen,
the PH pin on each controller should be programmed as discussed in the Functional Description under PHASE NUMBER
SELECTION. The same number of phases must be selected
for each controller.
POWERING OPTIONS
The power connections will be determined by the VIN range
and the availability of an external 5V rail. This is discussed in
detail in the Functional Description under POWER CONNECTIONS. For 12V input systems, the use of an external 5V rail
to power the VDD bus can improve overall system efficiency.
30091929
FIGURE 8. Tracking an External Supply
MULTI-CONTROLLER SYSTEMS
For systems with more than 2 phases, there will be one controller configured as the Master and from 1 to 5 controllers
configured as Slave.
The Master controller uses the differential amplifier to sense
the output voltage at the load point. It also provides the common COMP signal used by all controllers, provides the loop
compensation and synchronizes the system clock to an external clock if one is provided.
The SYNCOUT of the Master is connected to the SYNC input
of the first Slave controller.
The Slave controllers are configured by tying the FB input to
the VCC pin of that controller. Each Slave uses the differential
amplifier to sense the COMP signal of the Master controller
and drive its own COMP input. The SYNCOUT of each Slave
controller is connected to the SYNC input of the next Slave
controller.
All controllers have the same parallel RC components connected from the FREQ pin to local ground corresponding to
the desired system clock even if synchronizing to an external
clock.
Common connections for all controllers:
1) IAVE (each controller will have a parallel RC filter to local
ground).
2) FAULT
3) EN
4) SS (LM3754), TRACK (LM3753)
5) PGOOD
30091930
FIGURE 9. Tracking an External Supply
For equal slew rates, the relationship for the tracking divider
is set by:
TRACKING (LM3753)
The LM3753 will track the output of an external power supply
by connecting a resistor divider to the TRACK pin as shown
in Figure 8. This allows the output voltage slew rate to be
controlled for loads that require precise sequencing.
A value of 10 kΩ 1% is recommended for RT1 as a good compromise between high precision and low quiescent current
through the divider. Note that the TRACK pin must finish at
least 100 mV higher than the 0.6V reference to achieve the
30091932
FIGURE 10. Tracking an External Supply with Equal Slew
Rates
19
www.national.com
LM3753/54
to COMP to provide the duty cycle adjustment for phase-tophase current sharing.
LM3753/54
In order to track properly, the external power supply voltage
must be higher than the LM3753 output voltage.
SOFT-START (LM3754)
To avoid current limit during startup, the soft-start time tSS
should be substantially longer than the time required to
charge COUT to VOUT at the maximum output current. To meet
this requirement:
The second criterion is inductor saturation current rating. The
LM3753/54 has an accurately programmed peak current limit.
During an output short circuit, the inductor should be chosen
so as not to exceed its saturation rating at elevated temperature. For the design example, a standard value of 440 nH is
chosen to fall within the ΔIL = (1/5 to 2/5) x IOUT range.
The dc loss in the inductor is determined by its series resistance RL. The dc power dissipation is found from:
PDC = IOUT2 x RL
Choose a soft-start capacitor according to the formula:
The ac loss can be estimated from the inductor
manufacturer’s data, if available. The ac loss is set by the
peak-to-peak ripple current ΔIL and the switching frequency
fSW.
Where CSS is the soft-start capacitor and tSS is the soft-start
time.
OUTPUT CAPACITORS
The output capacitors filter the inductor ripple current and
provide a source of charge for transient load conditions. A
wide range of output capacitors may be used with the
LM3753/54 that provides excellent performance. The best
performance is typically obtained using aluminum electrolytic,
tantalum, polymer, solid aluminum, organic or niobium type
chemistries in parallel with ceramic capacitors. The ceramic
capacitors provide extremely low impedance to reduce the
output ripple voltage and noise spikes, while the aluminum or
other capacitors provide a larger bulk capacitance for transient loading.
When selecting the value for the output capacitors the two
performance characteristics to consider are the output voltage ripple and transient response. The output voltage ripple
for a single phase can be approximated as:
External Components Selection
The following is a design example selecting components for
the Typical Application Schematic of Figure 20. The circuit is
designed for two controller 4-phase operation with 1.2V out
at 100A from an input voltage of 6V to 18V. The expected load
is a microprocessor or ASIC with fast load transients, and the
type of MOSFETs used are in SO-8 or its equivalent packages
such as PowerPAK ®, PQFN and LFPAK (LFPAK-i).
SWITCHING FREQUENCY
The selection of switching frequency is based on the tradeoff
between size, cost, and efficiency. In general, a lower frequency means larger, more expensive inductors and capacitors will be needed. A higher switching frequency generally
results in a smaller but less efficient solution. For this application a frequency of 300 kHz was selected as a good compromise between the size of the inductor and MOSFETs,
transient response and efficiency. Following the equation given for RFRQ in the Functional Description under OSCILLATOR
and SYNCHRONIZATION, for 300 kHz operation a 78.7 kΩ
1% resistor is used for RFRQ. A 1000 pF capacitor is used for
CFRQ.
With all values normalized to a single phase, ΔVO (V) is the
peak to peak output voltage ripple, ΔIL (A) is the peak to peak
inductor ripple current, RC (Ω) is the equivalent series resistance or ESR of the output capacitors, fSW (Hz) is the switching frequency, and CO (F) is the output capacitance. The
amount of output ripple that can be tolerated is application
specific. A general recommendation is to keep the output ripple less than 1% of the rated output voltage. Figure 11 shows
the output voltage ripple for multi-phase operation.
OUTPUT INDUCTORS
The first criterion for selecting an output inductor is the inductance itself. In most buck converters, this value is based on
the desired peak-to-peak ripple current, ΔIL that flows in the
inductor along with the load current. As with switching frequency, the selection of the inductor is a tradeoff between size
and cost. Higher inductance means lower ripple current and
hence lower output voltage ripple. Lower inductance results
in smaller, less expensive devices. An inductance that gives
a ripple current of 1/5 to 2/5 of the maximum output current is
a good starting point. (ΔIL = (1/5 to 2/5) x IOUT). Minimum inductance is calculated from this value, using the maximum
input voltage as:
By calculating in terms of amperes, volts, and megahertz, the
inductance value will come out in micro henries. The inductor
ripple current is found from the minimum inductance equation:
www.national.com
20
For D < 0.5, VL = VOUT
For D > 0.5, VL = VIN − VOUT
With RC = VP / ΔIO this reduces to:
30091936
With RC = 0 this reduces to:
FIGURE 11. Multi-Phase Output Voltage Ripple
Based on the normalized single phase ripple, the worst case
multi-phase output voltage ripple can be approximated as:
ΔVO(N) = ΔVO / N
Since D < 0.5, VL = VOUT. With RC = 3 mΩ, the minimum value
for CO is 476 μF.
The minimum control loop bandwidth fC is given by:
Where N is the number of phases.
The output capacitor selection will also affect the output voltage droop and overshoot during a load transient. The peak
transient of the output voltage during a load current step is
dependent on many factors. Given sufficient control loop
bandwidth an approximation of the transient voltage can be
obtained from:
For the design example, the minimum value for fC is 44 kHz.
Two 220 μF, 5 mΩ polymer capacitors in parallel with two 22
μF, 3 mΩ ceramics per phase will meet the target output voltage ripple and transient specification.
INPUT CAPACITORS
The input capacitors for a buck regulator are used to smooth
the large current pulses drawn by the inductor and load when
the high-side MOSFET is on. Due to this large ac stress, input
capacitors are usually selected on the basis of their ac rms
current rating rather than bulk capacitance. Low ESR is beneficial because it reduces the power dissipation in the capacitors. Although any of the capacitor types mentioned in the
OUTPUT CAPACITORS section can be used, ceramic capacitors are common because of their low series resistance.
In general the input to a buck converter does not require as
much bulk capacitance as the output.
The input capacitors should be selected for rms current rating
and minimum ripple voltage. The equation for the rms current
and power loss of the input capacitor in a single phase can
be estimated as:
With all values normalized to a single phase, VP (V) is the
output voltage transient and ΔIO (A) is the load current step
change. CO (F) is the output capacitance, L (H) is the value
of the inductor and RC (Ω) is the series resistance of the output
capacitor. VL (V) is the minimum inductor voltage, which is
duty cycle dependent.
For D < 0.5, VL = VOUT
For D > 0.5, VL = VIN − VOUT
This shows that as the input voltage approaches VOUT, the
transient droop will get worse. The recovery overshoot remains fairly constant.
The loss associated with the output capacitor series resistance can be estimated as:
Output Capacitor Design Procedure
For the design example VIN = 12V, VOUT = 1.2V, D = VOUT /
VIN = 0.1, L = 440 nH, ΔIL = 9A, ΔIO = 20A and VP = 0.12V.
To meet the transient voltage specification, the maximum
RC is:
Where IO (A) is the output load current and RCIN (Ω) is the
series resistance of the input capacitor. Since the maximum
values occur at D = 0.5, a good estimate of the input capacitor
rms current rating in a single phase is one-half of the maximum output current.
Neglecting the series inductance of the input capacitance, the
input voltage ripple for a single phase can be estimated as:
For the design example, the maximum RC is 6 mΩ. Choose
RC = 3 mΩ as the design limit.
21
www.national.com
LM3753/54
From the equation for VP, the minimum value of CO is:
LM3753/54
The converter exhibits a negative input impedance which is
lowest at the minimum input voltage:
By defining the maximum input voltage ripple, the minimum
requirement for the input capacitance can be calculated as:
The damping factor for the input filter is given by:
For multi-phase operation, the general equation for the input
capacitor rms current is approximated as:
Where RLIN is the input wiring resistance and RCIN is the series
resistance of the input capacitors. The term ZS / ZIN will always
be negative due to ZIN.
When δ = 1, the input filter is critically damped. This may be
difficult to achieve with practical component values. With δ <
0.2, the input filter will exhibit significant ringing. If δ is zero or
negative, there is not enough resistance in the circuit and the
input filter will sustain an oscillation.
When operating near the minimum input voltage, an aluminum electrolytic capacitor across CIN may be needed to
damp the input for a typical bench test setup. Any parallel
capacitor should be evaluated for its rms current rating. The
current will split between the ceramic and aluminum capacitors based on the relative impedance at the switching frequency. Using a square wave approximation, the rms current
in each capacitor is found from:
This is valid for D < 1 / N and repeats for a total of N times.
IO represents the total output current and N is the number of
phases. Figure 12 shows the input capacitor rms current as
a function of the output current, duty cycle and number of
phases.
30091948
FIGURE 12. Input Capacitor RMS Current as a Function
of Output Current
For multi-phase operation the maximum rms current can be
approximated as:
Input Capacitor Design Procedure
Ceramic capacitors are sized to support the required rms current. An aluminum electrolytic capacitor is used for damping.
Find the minimum value for the ceramic capacitors from:
ICIN(RMS)MAX ≈ 0.5 x IO / N
In most applications for point-of-load power supplies, the input voltage is the output of another switching converter. This
output often has a lot of bulk capacitance, which may provide
adequate damping.
When the converter is connected to a remote input power
source through a wiring harness, a resonant circuit is formed
by the line impedance and the input capacitors. If step input
voltage transients are expected near the maximum rating of
the LM3753/54, a careful evaluation of the ringing and possible overshoot at the device VIN pin should be completed.
To minimize overshoot make CIN > 10 x LIN. The characteristic
source impedance and resonant frequency are:
www.national.com
Allowing ΔVIN = 0.6V for the design example, the minimum
value is CIN = 34.7 μF. Find the rms current rating from:
ICIN(RMS)MAX ≈ 0.5 x IO / N
Using the same criteria, the result is 12.5A rms. Manufacturer
data for 4.7 μF, 25V, X7R capacitors in a 1210 package allows
for 4A rms with a 20°C temperature rise. For the design example, using two ceramic capacitors for each phase will meet
both the input voltage ripple and rms current target. Since the
series resistance is so low at about 4 mΩ per capacitor, a
parallel aluminum electrolytic is used for damping. A good
22
LM3753/54
IL_VL = IOUT − 0.5 x ΔIL
general rule is to make the damping capacitor at least five
times the value of the ceramic. By sizing the aluminum such
that it is primarily resistive at the switching frequency, the design is greatly simplified since the ceramic capacitors are
primarily reactive. In this case the approximation for the rms
current in the damping capacitor is:
IL_PK = IOUT + 0.5 x ΔIL
RG_ON = 5 + RG_INT + RG_EXT
Where CIN2 is the damping capacitance, RCIN2 is its series
resistance and CIN1 is the ceramic capacitance. A 470 μF,
25V, 0.06Ω, 1.19A rms aluminum electrolytic capacitor in a
10 mm x 10.2 mm package is chosen for the damping capacitor. Calculated rms current for the aluminum electrolytic is
0.67A.
RG_OFF = 2 + RG_INT + RG_EXT
Switching loss is calculated for the high-side FET only. 5 and
2 represent the LM3753/54 high-side driver resistance in the
transient region. RG_INT is the gate resistance of the high-side
FET, and RG_EXT is the extra external gate resistance if applicable. RG_EXT may be used to damp out excessive parasitic
ringing at the switch node.
For this example, the maximum drain-to-source voltage applied to either MOSFET is 18V. The maximum drive voltage
at the gate of the high-side MOSFET is 5V, and the maximum
drive voltage for the low-side MOSFET is 5V. The selected
MOSFET must be able to withstand 18V plus any ringing from
drain to source, and be able to handle at least 5V plus ringing
from gate to source. If the duty cycle of the converter is small,
then the high-side MOSFET should be selected with a low
gate charge in order to minimize switching loss whereas the
bottom MOSFET should have a low RDS(on) to minimize conduction loss.
For a typical input voltage of 12V and output current of 25A
per phase, the MOSFET selections for the design example
are SIR850DP for the high-side MOSFET and 2 x SIR892DP
for the low-side MOSFET.
A 2.2Ω resistor for the high-side gate drive may be added in
series with the HG output. This helps to control the MOSFET
turn-on and ringing at the switch node. Additionally, 0.5A
Schottky diodes may be placed across the high-side MOSFETs. The external Schottky diodes have a much faster recovery characteristic than the MOSFET body diode, and help
to minimize switching spikes by clamping the SW pin to VIN.
Another technique to control ringing at the switch node is to
place an RC snubber from SW to PGND directly across the
low-side MOSFET. Typical values at 300 kHz are 1Ω and 680
pF.
To improve efficiency, 3A Schottky diodes may be placed
across the low-side MOSFETs. The external Schottky diodes
have a much lower forward voltage than the MOSFET body
diode, and help to minimize the loss due to the body diode
recovery characteristic.
MOSFETS
Selection of the power MOSFETs is governed by a tradeoff
between cost, size and efficiency.
Losses in the high-side FET can be broken down into conduction loss, gate charge loss and switching loss. Conduction
or I2R loss is approximately:
PCOND_HI = D x (IOUT2 x RDS(on)_HI x 1.3)
(High-side FET)
PCOND_LO = (1 − D) x (IOUT2 x RDS(on)_LO x 1.3)
(Low-side FET)
In the above equations the factor 1.3 accounts for the increase in MOSFET RDS(on) due to self heating. Alternatively,
the 1.3 can be ignored and the RDS(on) of the MOSFET estimated using the RDS(on) vs. Temperature curves in the MOSFET datasheets.
The gate charge loss results from the current driving the gate
capacitance of the power MOSFETs, and is approximated as:
PDR = VIN x (QG_HI + QG_LO) x fSW
Where QG_HI and QG_LO are the total gate charge of the highside and low-side FETs respectively at the typical 5V driver
voltage. Gate charge loss differs from conduction and switching losses in that the majority of dissipation occurs in the
LM3753/54 and VDD regulator.
The switching loss occurs during the brief transition period as
the FET turns on and off, during which both current and voltage is present in the channel of the FET. This can be approximated as:
EN and VIN UVLO
For operation at 6V minimum input, set the EN divider to enable the LM3753/54 at approximately 5.5V nominal. Values
of RUV1 = 1.37 kΩ and RUV2 = 4.02 kΩ will meet the target
threshold.
CURRENT SENSE
For resistor current sense, a 1 mΩ 1W resistor is used for a
full scale voltage of 25 mV at 25A out.
For DCR sensing, RS is equal to the inductor resistance of
RL = 0.32 mΩ plus an estimated trace resistance of 0.2 mΩ..
The full scale voltage is about 13 mV at 25A. For equal time
constants, the relationship of the integrating RC is determined
by:
Where QGD is the high-side FET Miller charge with a VDS
swing between 0 to VIN; CISS is the input capacitance of the
high-side MOSFET in its off state with VDS = VIN. α and β are
fitting coefficient numbers, which are usually between 0.5 to
1, depending on the board level parasitic inductances and reverse recovery of the low-side power MOSFET body diode.
Under ideal condition, setting α = β = 0.5 is a good starting
point. Other variables are defined as:
23
www.national.com
LM3753/54
Choosing CDCR = 0.15 μF:
Using QG_HI = 2 x 10 nC and QG_LO = 4 x 21 nC per controller
with a 5V gate drive, the minimum value for CVDD = 1.04 μF.
To use common component values, CVDD1 and CVDD2 are also
selected as 4.7 μF ceramic.
A general purpose NPN transistor is sized to meet the requirements for the VDD supply. Based on the gate charge of
104 nC per controller, the required current is found from:
RDCR = 440 nH / (0.15 μF x 0.52 mΩ) = 5.64 kΩ.
Using a standard value of 5.90 kΩ, the average current
through RDCR is calculated as 203 μA from:
IDCR = VOUT / RDCR
IDCR is sufficiently high enough to keep the CS input bias current from being a significant error term.
IGC = QG_TOTAL x fSW
At 300 kHz, IGC = 31.2 mA per controller. For a two controller
system, the minimum HFE for the transistor is determined by:
CURRENT LIMIT
For the design example, the desired current limit set point is
chosen as 34.5A peak per phase, which is about 25% above
the full load peak value. Using DCR sense with RS = 0.52
mΩ:
HFEMIN = IGC_TOTAL / 5 mA
The power dissipated by the transistor is:
PR = (VIN − VDD) x IGC_TOTAL
RILIM = 34.5A x 0.52 mΩ / 94 μA = 191Ω
The transistor must support 62.4 mA with an HFE of at least
12.5 over the entire operating range. At 18V in the power dissipated is 0.8W. A CJD44H11 in a DPAK case is chosen for
the design example. A 0.047 μF capacitor from base to PGND
will improve the transient performance of the VDD supply.
CBOOT provides power for the high-side gate drive, and is
sized to meet the required gate drive current. Allowing for
ΔVBOOT = 100 mV of ripple, the minimum value for CBOOT is
found from:
For resistor sense, the relatively low output inductor value
forms a voltage divider with the intrinsic inductance of the
sense resistor. When the MOSFETs switch, this adds a step
to the otherwise triangular current sense voltage. The step
voltage is simply the input voltage times the inductive divider.
With L = 440 nH and LS = 1 nH, the step voltage is:
VLS = 12V x 1 nH / 441 nH = 27.2 mV
Using the same method as DCR sense, an RC filter is added
to recover the actual resistive sense voltage. Choosing C = 1
nF the resistor is calculated as:
R = 1 nH / (1 nF x 1 mΩ) = 1 kΩ
The current limit resistor is then calculated as:
Using QG_HI = 10 nC per phase with a 5V gate drive, the minimum value for CBOOT = 0.1 μF. CBOOT is selected as 0.22 μF
ceramic per phase for the design example. A 0.5A Schottky
diode is used for DBOOT at each controller.
RILIM = 34.5A x 1 mΩ / 94 μA = 367Ω
The closest standard value of 365Ω 1% is selected for the
design example.
TRACK (LM3753)
For the design example, an external voltage of 3.3V is used
as the controlling voltage. The divider values are set so that
both voltages will rise together, with VEXT reaching its final
value just before VOUT. Following the method in the Application Information under TRACKING (LM3753) and allowing for
a 120 mV offset between FB and TRACK, standard 1% values
are selected for RT1 = 10 kΩ and RT2 = 35.7 kΩ.
PRE-LOAD RESISTOR
For normal operation, a pre-load resistor is generally not required. During an abnormal fault condition with the output
completely disconnected from the load, the output voltage
may rise. This is primarily due to the high-side driver off-state
bias current, and reverse leakage current of the high-side
Schottky clamp diode.
At room temperature with 12V input, the reverse leakage of
each 0.5A Schottky diode is about 15 μA. With the EN pin high
and the FAULT pin low, the bias current in each high-side
driver is about 105 μA. Allowing for a 2 to 1 variation, the
maximum value of resistor to keep the output voltage from
rising above 5% of its nominal value is found from:
SOFT-START (LM3754)
To prevent over-shoot, the soft-start time is set to be longer
than the time it would take to charge the output voltage at the
maximum output current. Following the equations in the Application Information under SOFT-START (LM3754):
R = 0.05 x 1.2V / 330 µA = 182Ω
tSS(MIN) = (1.2V x 484 μF) / (34.5A − 25A) = 61 μs
A value of 120Ω is selected for the design example. This represents a 10 mA pre-load at the rated output voltage, which
is 0.01% of the 100A full load current.
Choosing a value of CSS = 0.1 μF, the soft-start time is:
tSS = (0.1 μF x 0.6V) / 10 μA = 6 ms
CONTROL LOOP COMPENSATION
The LM3753 uses voltage-mode PWM control to correct
changes in output voltage due to line and load transients. Input voltage feed-forward is used to adjust the amplitude of the
PWM ramp. This stabilizes the modulator gain from variations
due to input voltage, providing a robust design solution. A fast
inner current sharing circuit ensures good dynamic response
to changes in load current.
VCC, VDD and BOOT
VCC is used as the supply for the internal control and logic
circuitry. A 4.7 μF ceramic capacitor provides sufficient filtering for VCC.
CVDD provides power for both the high-side and low-side
MOSGET gate drives, and is sized to meet the total gate drive
current. Allowing for ΔVVDD = 100 mV of ripple, the minimum
value for CVDD is found from:
www.national.com
24
fier with a typical dc gain of 70 dB and a unity gain frequency
of 15 MHz. Figure 13 shows the power stage, error amplifier
and current sharing components.
30091962
FIGURE 13. Power Stage, Error Amplifier and Current Sharing
The simplified power stage transfer function (also called the
control-to-output transfer function) for the LM3753/54 can be
written as:
Km is the dc modulator gain and Ri is the current-sharing gain.
KFF is the input voltage feed-forward term, which is internally
set to a value of 0.232 V/V. The IAVE filter is accounted for
by Ha(s), which provides additional damping of the modulator
transfer function.
RAV sets the gain of the current averaging amplifier. A fixed
value of 8 kΩ/phase must be used for proper scaling. Since
the effective resistance is in parallel, each LM3753/54 should
have a 4.02 kΩ 1% resistor at IAVE for 2-phase/controller operation. CAV sets the IAVE filter time constant of the current
sharing amplifier. For optimal performance of the current
sharing circuit, the IAVE filter is designed to settle to its final
value in five switching cycles. The optimal IAVE time constant
is defined as:
Where:
T = CAV x RAV
A value of CAV = 1/(RAV x fSW) per phase must be used for the
optimal time constant. Each LM3753/54 should have a value
of two times the normalized single phase value of CAV at IAVE
for 2-phase/controller operation. In this manner, the IAVE
time constant maintains a fixed value of T for any number of
phases.
Typical frequency response of the gain and the phase for the
power stage are shown in Figure 14 and Figure 15. It is designed for VIN = 12V, VOUT = 1.2V, IOUT = 25A per phase and
a switching frequency of 300 kHz. For 2-phase operation
RAV = 4.02 kΩ and CAV = 1000 pF. The power stage component values per phase are:
L = 0.44 μH, RL = 0.52 mΩ, CO1 = 440 μF, RC1 = 2.5 mΩ,
CO2 = 44 μF, RC2 = 1.5 mΩ, RS = RL = 0.52 mΩ and RO =
VOUT / IOUT = 48 mΩ.
With:
25
www.national.com
LM3753/54
The control loop is comprised of two parts. The first is the
power stage, which consists of the duty cycle modulator, current sharing circuit, output filter and load. The second part is
the error amplifier, which is a voltage type operational ampli-
LM3753/54
In general, the goal of the compensation circuit is to give high
gain, a bandwidth that is between one-fifth and one-tenth of
the switching frequency, and at least 45° of phase margin.
Control Loop Design Procedure
Once the power stage design is complete, the power stage
components are used to determine the proper frequency
compensation. Knowing the dc modulator gain and assuming
an ideal single-pole system response, the mid-band error amplifier gain is set by the target crossover frequency. Based on
the ideal amplifier transfer function, the zero-pair is set to
cancel the complex conjugate pole of the output filter. One
pole is set to cancel the ESR of the output capacitor. The
second pole is set equal to the switching frequency. A correction factor is used to accommodate the modulator damping
when the output filter pole is within a decade of the target
crossover frequency.
The compensation components will scale from the feedback
divider ratio and selection of the bottom feedback divider resistor. A maximum value for the divider current is typically set
at 1 mA. Using a divider current of 200 μA will allow for a
reasonable range of values. For the bottom feedback resistor
RFBB = VREF / 200 μA = 3 kΩ. Choosing a standard 1% value
of 3.01 kΩ, the top feedback resistor is found from:
30091966
FIGURE 14. Power Stage Gain
For VOUT = 1.2V and VREF = 0.6V, RFBT = 3.01 kΩ.
Based on the previously defined power stage values, calculate general terms:
For the design example D = 0.1, Ri = 0.026Ω, T = 3.33 μs and
Km = 3.22.
Calculate the output filter pole frequency and the ESR zero
frequency from:
30091967
FIGURE 15. Power Stage Phase
Assuming a pole at the origin, the simplified equation for the
error amplifier transfer function can be written in terms of the
mid-band gain as:
For the output filter pole using CO = CO1 + CO2, ωP = 68.5 krad/
sec. Since CO1 >> CO2, the ESR zero is calculated using
CO1 and RC1 as ωZ = 909 krad/sec.
Choose a target crossover frequency fC greater than the minimum control loop bandwidth from the OUTPUT CAPACITORS section. The optimum value of the crossover frequency
is usually between 5 and 10 times the filter pole frequency.
With fP = ωP / (2 x π) = 10.9 kHz, this places fC between 54.5
kHz and 109 kHz. The upper limit for fC is typically set at 1/5
of the switching frequency.
Where:
Choosing fC = 60 kHz for the design example ωC = 377 krad/
sec. The switching frequency is ωSW = 1.88 Mrad/sec.
www.national.com
26
LM3753/54
For output capacitors with very low ESR, if the target
crossover frequency is more than 10 times the filter pole frequency, bandwidth limiting of the error amplifier may occur.
See the Comprehensive Equations section to incorporate the
error amplifier bandwidth into the design procedure.
For reference, the parallel equivalent CO and RC at any frequency can be calculated from:
30091976
FIGURE 16. Error Amplifier Gain
At the target crossover frequency X1 = 0.00603, X2 = 0.0603,
Z = 0.00592 and A = 1.213. The parallel equivalent CO = 478
μF and RC = 2.1 mΩ.
Calculate the error amplifier gain coefficient and the compensation component values. The (1 − ωP/ωC) term is the correction factor for the modulator damping.
30091977
FIGURE 17. Error Amplifier Phase
The complete control loop transfer function is equal to the
product of the power stage transfer function and error amplifier transfer function. For the Bode plots, the overall loop gain
is the equal to the sum in dB and the overall phase is equal
to the sum in degrees. Results are shown in Figure 18 and
Figure 19. The crossover frequency is 57 kHz with a phase
margin of 73°.
For the design example, the calculated values are GC = 1.71,
CHF = 103 pF, CCOMP = 2236 pF, RCOMP = 6527Ω, RFF = 245
and CFF = 4483 pF.
Using standard values of CHF = 100 pF, CCOMP = 2200 pF,
RCOMP = 6.2 kΩ, RFF = 240Ω and CFF = 4700 pF, the error
amplifier plots of gain and phase are shown in Figure 16 and
Figure 17.
27
www.national.com
LM3753/54
The total power dissipated in the power components can be
obtained by adding together the loss as mentioned in the
OUTPUT INDUCTORS, OUTPUT CAPACITORS, INPUT
CAPACITORS and MOSFETS sections.
The highest power dissipating components are the power
MOSFETs. The easiest way to determine the power dissipated in the MOSFETs is to measure the total conversion loss
(PIN − POUT), then subtract the power loss in the capacitors,
inductors, LM3753/54 and VDD regulator. The resulting power loss is primarily in the switching MOSFETs. Selecting
MOSFETs with exposed pads will aid the power dissipation
of these devices. Careful attention to RDS(on) at high temperature should be observed.
If a snubber is used, the power loss can be estimated with an
oscilloscope by observation of the resistor voltage drop at
both the turn-on and turn-off transitions. Assuming that the
RC time constant is << 1 / fSW:
30091978
FIGURE 18. Control Loop Gain
P = ½ x C x (VP2 + VN2) x fSW
VP and VN represent the positive and negative peak voltage
across the snubber resistor, which is ideally equal to VIN.
LM3753/54 and VDD REGULATOR OPERATING LOSS
These terms accounts for the currents drawn at the VIN and
VDD pins, used for driving the logic circuitry and the power
MOSFETs. For the LM3753/54, the VIN current is equal to the
steady state operating current IVIN. The VDD current is primarily determined by the MOSFET gate charge current IGC,
which is defined as:
IGC = QG_TOTAL x fSW
PD = (VIN x IVIN) + (VDD x IGC)
QG_TOTAL is the total gate charge of the MOSFETs connected
to each LM3753/54. PD represents the total power dissipated
in each LM3753/54. IVIN is about 15 mA from the Electrical
Characteristics table. The LM3753/54 has an exposed thermal pad to aid power dissipation.
The power dissipated in the VDD regulator is determined by:
PR = (VIN − VDD) x IGC_TOTAL
30091979
IGC_TOTAL is the sum of the MOSFET gate charge currents for
all of the controllers.
FIGURE 19. Control Loop Phase
For the small-signal analysis, it is assumed that the control
voltage at the COMP pin is dc. In practice, the output ripple
voltage is amplified by the error amplifier gain at the switching
frequency, which appears at the COMP pin adding to the
control ramp. This tends to reduce the modulator gain, which
may lower the actual control loop crossover frequency. This
effect is greatly reduced as the number of phases is increased.
Layout Considerations
To produce an optimal power solution with a switching converter, as much care must be taken with the layout and design
of the printed circuit board as with the component selection.
The following are several guidelines to aid in creating a good
layout.
KELVIN TRACES for GATE DRIVE and SENSE LINES
The HG and SW pins provide the gate drive and return for the
high-side MOSFETs. These lines should run as parallel pairs
to each MOSFET, being connected as close as possible to
the respective MOSFET gate and source. Likewise the LG
and PGND pins provide the gate drive and return for the lowside MOSFETs. A good ground plane between the PGND pin
and the low-side MOSFETs source connections is needed to
carry the return current for the low-side gates.
The SNSP and SNSM pins of the Master should be connected
as a parallel pair, running from the output power and ground
sense points. Keep these lines away from the switch node
Efficiency and Thermal
Considerations
The buck regulator steps down the input voltage and has a
duty ratio D of:
Where η is the estimated converter efficiency. The efficiency
is defined as:
www.national.com
28
LM3753/54
and output inductor to avoid stray coupling. If possible, the
SNSP and SNSM traces should be shielded from the switch
node by ground planes.
SGND and PGND CONNECTIONS
Good layout techniques include a dedicated ground plane,
usually on an internal layer adjacent to the LM3753/54 and
signal component side of the board. Signal level components
connected to FB, TRACK/SS, FREQ, IAVE, EN and PH along
with the VCC and VIN bypass capacitors should be tied directly to the SGND pin. Connect the SGND and PGND pins
directly to the DAP, with vias from the DAP to the ground
plane. The ground plane is then connected to the input capacitors and low-side MOSFET source at each phase.
ERROR AMPLIFIER TRANSFER FUNCTION
Using a single-pole operational amplifier model, the complete
error amplifier transfer function is given by:
MINIMIZE the SWITCH NODE
The copper area that connects the power MOSFETs and output inductor together radiates more EMI as it gets larger. Use
just enough copper to give low impedance for the switching
currents and provide adequate heat spreading for the MOSFETs.
Where the open loop gain AOL = 3162 (70 dB) and the unity
gain bandwidth ωBW = 2 x π x fBW.
The ideal transfer function is expressed in terms of the midband gain as:
LOW IMPEDANCE POWER PATH
In a buck regulator the primary switching loop consists of the
input capacitor connection to the MOSFETs. Minimizing the
area of this loop reduces the stray inductance, which minimizes noise and possible erratic operation. The ceramic input
capacitors at each phase should be placed as close as possible to the MOSFETs, with the VIN side of the capacitors
connected directly to the high-side MOSFET drain, and the
PGND side of the capacitors connected as close as possible
to the low-side source. The complete power path includes the
input capacitors, power MOSFETs, output inductor, and output capacitors. Keep these components on the same side of
the board and connect them with thick traces or copper
planes. Avoid connecting these components through vias
whenever possible, as vias add inductance and resistance. In
general, the power components should be kept close together, minimizing the circuit board losses.
The feedback gain is then:
Where:
Comprehensive Equations
POWER STAGE TRANSFER FUNCTION
To include all terms, it is easiest to use the impedance form
of the equation:
Where:
ERROR AMPLIFIER BANDWIDTH LIMIT
When the ideal error amplifier gain reaches the open loop
gain-bandwidth limit, the phase goes to zero. To incorporate
the amplifier bandwidth into the design procedure, determine
the boundary limit with respect to the ESR zero frequency:
Based on the relative ESR zero, the crossover frequency is
set at 1/3 of the bandwidth limiting frequency.
If ωZ > ωZB, calculate the optimal crossover frequency from:
With:
29
www.national.com
LM3753/54
Using this method, the maximum phase boost is achieved at
the optimal crossover frequency.
In either case, the upper limit for fC is typically set at 1/5 of the
switching frequency.
If ωZ < ωZB, calculate the optimal crossover frequency from:
www.national.com
30
LM3753/54
Typical Application
30091902
All controllers in the system are the same part. The Master and Slave are differentiated by how they are connected in the system.
FIGURE 20. Typical Application
31
www.national.com
FIGURE 21. Master with DCR Sense
30091989
LM3753/54
Design Examples
www.national.com
32
FIGURE 22. Slave with DCR Sense
30091990
LM3753/54
33
www.national.com
www.national.com
34
FIGURE 23. Master with Resistor Sense
30091991
LM3753/54
FIGURE 24. Slave with Resistor Sense
30091992
LM3753/54
35
www.national.com
LM3753/54
Physical Dimensions inches (millimeters) unless otherwise noted
32-Lead LLP Package
NS Package Number SQA32A
www.national.com
36
LM3753/54
Notes
37
www.national.com
LM3753/54 Scalable 2-Phase Synchronous Buck Controllers with Integrated FET Drivers and
Linear Regulator Controller
Notes
For more National Semiconductor product information and proven design tools, visit the following Web sites at:
www.national.com
Products
Design Support
Amplifiers
www.national.com/amplifiers
WEBENCH® Tools
www.national.com/webench
Audio
www.national.com/audio
App Notes
www.national.com/appnotes
Clock and Timing
www.national.com/timing
Reference Designs
www.national.com/refdesigns
Data Converters
www.national.com/adc
Samples
www.national.com/samples
Interface
www.national.com/interface
Eval Boards
www.national.com/evalboards
LVDS
www.national.com/lvds
Packaging
www.national.com/packaging
Power Management
www.national.com/power
Green Compliance
www.national.com/quality/green
Switching Regulators
www.national.com/switchers
Distributors
www.national.com/contacts
LDOs
www.national.com/ldo
Quality and Reliability
www.national.com/quality
LED Lighting
www.national.com/led
Feedback/Support
www.national.com/feedback
Voltage References
www.national.com/vref
Design Made Easy
www.national.com/easy
www.national.com/powerwise
Applications & Markets
www.national.com/solutions
Mil/Aero
www.national.com/milaero
PowerWise® Solutions
Serial Digital Interface (SDI) www.national.com/sdi
Temperature Sensors
www.national.com/tempsensors SolarMagic™
www.national.com/solarmagic
PLL/VCO
www.national.com/wireless
www.national.com/training
PowerWise® Design
University
THE CONTENTS OF THIS DOCUMENT ARE PROVIDED IN CONNECTION WITH NATIONAL SEMICONDUCTOR CORPORATION
(“NATIONAL”) PRODUCTS. NATIONAL MAKES NO REPRESENTATIONS OR WARRANTIES WITH RESPECT TO THE ACCURACY
OR COMPLETENESS OF THE CONTENTS OF THIS PUBLICATION AND RESERVES THE RIGHT TO MAKE CHANGES TO
SPECIFICATIONS AND PRODUCT DESCRIPTIONS AT ANY TIME WITHOUT NOTICE. NO LICENSE, WHETHER EXPRESS,
IMPLIED, ARISING BY ESTOPPEL OR OTHERWISE, TO ANY INTELLECTUAL PROPERTY RIGHTS IS GRANTED BY THIS
DOCUMENT.
TESTING AND OTHER QUALITY CONTROLS ARE USED TO THE EXTENT NATIONAL DEEMS NECESSARY TO SUPPORT
NATIONAL’S PRODUCT WARRANTY. EXCEPT WHERE MANDATED BY GOVERNMENT REQUIREMENTS, TESTING OF ALL
PARAMETERS OF EACH PRODUCT IS NOT NECESSARILY PERFORMED. NATIONAL ASSUMES NO LIABILITY FOR
APPLICATIONS ASSISTANCE OR BUYER PRODUCT DESIGN. BUYERS ARE RESPONSIBLE FOR THEIR PRODUCTS AND
APPLICATIONS USING NATIONAL COMPONENTS. PRIOR TO USING OR DISTRIBUTING ANY PRODUCTS THAT INCLUDE
NATIONAL COMPONENTS, BUYERS SHOULD PROVIDE ADEQUATE DESIGN, TESTING AND OPERATING SAFEGUARDS.
EXCEPT AS PROVIDED IN NATIONAL’S TERMS AND CONDITIONS OF SALE FOR SUCH PRODUCTS, NATIONAL ASSUMES NO
LIABILITY WHATSOEVER, AND NATIONAL DISCLAIMS ANY EXPRESS OR IMPLIED WARRANTY RELATING TO THE SALE
AND/OR USE OF NATIONAL PRODUCTS INCLUDING LIABILITY OR WARRANTIES RELATING TO FITNESS FOR A PARTICULAR
PURPOSE, MERCHANTABILITY, OR INFRINGEMENT OF ANY PATENT, COPYRIGHT OR OTHER INTELLECTUAL PROPERTY
RIGHT.
LIFE SUPPORT POLICY
NATIONAL’S PRODUCTS ARE NOT AUTHORIZED FOR USE AS CRITICAL COMPONENTS IN LIFE SUPPORT DEVICES OR
SYSTEMS WITHOUT THE EXPRESS PRIOR WRITTEN APPROVAL OF THE CHIEF EXECUTIVE OFFICER AND GENERAL
COUNSEL OF NATIONAL SEMICONDUCTOR CORPORATION. As used herein:
Life support devices or systems are devices which (a) are intended for surgical implant into the body, or (b) support or sustain life and
whose failure to perform when properly used in accordance with instructions for use provided in the labeling can be reasonably expected
to result in a significant injury to the user. A critical component is any component in a life support device or system whose failure to perform
can be reasonably expected to cause the failure of the life support device or system or to affect its safety or effectiveness.
National Semiconductor and the National Semiconductor logo are registered trademarks of National Semiconductor Corporation. All other
brand or product names may be trademarks or registered trademarks of their respective holders.
Copyright© 2009 National Semiconductor Corporation
For the most current product information visit us at www.national.com
National Semiconductor
Americas Technical
Support Center
Email: [email protected]
Tel: 1-800-272-9959
www.national.com
National Semiconductor Europe
Technical Support Center
Email: [email protected]
National Semiconductor Asia
Pacific Technical Support Center
Email: [email protected]
National Semiconductor Japan
Technical Support Center
Email: [email protected]