Micronote 1308 - (AN-19) LX1672 Design Reference Guide (258.55 kB)

LX1672
PRODUCT DESIGN GUIDE
AN19
LX1672 PR0DUCT DESIGN
GUIDE
Protected by US Patents: 6,285,571 & 6,292,378
I N T E G R A T E D
Copyright © 2000
Rev. 0.2a, 2006-03-08
P R O D U C T S
Microsemi
Integrated Products
11861 Western Avenue, Garden Grove, CA. 92841, 714-898-8121, Fax: 714-893-2570
Page 1
LX1672
PRODUCT DESIGN GUIDE
TABLE OF CONTENTS
Sections
1.0
2.0
3.0
4.0
5.0
6.0
7.0
8.0
9.0
10.0
11.0
12.0
13.0
14.0
15.0
16.0
17.0
18.0
19.0
20.0
21.0
22.0
23.0
Introduction...................................................................................................................................3
Theory Of Operation.....................................................................................................................3
Bootstrap ......................................................................................................................................7
Soft Start and Enable ...................................................................................................................7
Over Current Protection................................................................................................................8
Component Selection ...................................................................................................................8
FET Selection ...............................................................................................................................9
Upper FET Power Dissipation ......................................................................................................9
Lower FET Power Dissipation ......................................................................................................9
General FET Considerations ........................................................................................................9
FET Gate Drive...........................................................................................................................10
FET Thermal Considerations......................................................................................................10
Inductor Selection.......................................................................................................................11
Output Capacitor Selection.........................................................................................................12
Input Capacitor Selection ...........................................................................................................12
Efficiency ....................................................................................................................................13
Loop Compensation ...................................................................................................................14
Transient Response ...................................................................................................................15
Circuit Board Layout ...................................................................................................................15
Bi-Phase PWM Current Sharing.................................................................................................15
LoadSHARE ...............................................................................................................................16
Other Multiphase Methods .........................................................................................................20
DDR VTT Termination Supply ....................................................................................................20
Figures
1 Synchronous Buck Regulator ...........................................................................................................3
2 PWM Block Diagram ....................................................................................................................4, 5
3 Buck Regulator Waveforms..............................................................................................................6
4 Bootstrap Circuit ...............................................................................................................................7
5 Simplified Soft Start Circuit...............................................................................................................7
6 Over Current Protection....................................................................................................................8
7 RMS Ripple Current .......................................................................................................................13
8 Bode Plot ........................................................................................................................................14
9 Control Loop Block Diagram...........................................................................................................14
10 PWM Switch and Output Filter ....................................................................................................15
11 LoadSHARE Block diagram ...........................................................................................................16
12 LoadSHARE Using Inductor ESR Method......................................................................................16
13 LoadSHARE Using Feedback Divider Method ...............................................................................17
14 LoadSHARE Using Proportional Control .......................................................................................18
15 DDR VTT Supply ...........................................................................................................................21
Copyright © 2000
Rev. 0.2a, 2006-03-08
Microsemi
Integrated Products
11861 Western Avenue, Garden Grove, CA. 92841, 714-898-8121, Fax: 714-893-2570
Page 2
LX1672
PRODUCT DESIGN GUIDE
1.0 INTRODUCTION
This document is intended to be used with the LX1672
data sheet as a guide to designing buck regulators
with the LX1672 controller. Information is presented
on operation of the complete circuit and selection of
critical components that should allow the design of a
complete regulator.
2.0 THEORY OF OPERATION
Buck regulators are used in a variety of applications
where the input must be stepped down to a lower
output voltage. The increasing power demands of
microprocessors and their support chips require
power levels that would be impractical with linear
regulators. The demand for better efficiency is another
driving factor in the use of switching regulators. The
LX1672 contains two synchronous buck regulators in
a single package.
VCC
+5V
VCx
+12V
VIN
+5V
CIN
Q1
L
LX1671
Q2
VOUT
Phase
Mode
COUT
Figure (1) - Synchronous Buck Regulator
The inductor is used as an energy storage device and
acts as a current source. When the main FET (Q1) is
turned on, current from the input flows through the
inductor and into the load. When the output voltage
reaches the control point, the main FET (Q1) will be
turned off and the sync FET (Q2) will be driven on.
There is a short non-overlap period at all switching
transitions to insure that there can be no instance
where both FETs are on at the same time.
Copyright © 2000
Rev. 0.2a, 2006-03-08
Both FETs are constantly switching at the PWM
frequency to maintain regulation.
The control loop consists of an error amplifier with the
non-inverting input connected to a reference voltage
and the inverting input connected to the output voltage
through a resistor network that determines the output
voltage.
The error amplifier output is used to drive a
comparator with a voltage ramp on its other input. The
error amplifier output will move up and down on the
ramp voltage causing the comparator to generate a
timing signal that determines the PWM duty cycle. See
Figure (2a)
Since the inductor current cannot change instantly
when the main FET switches off the body diode in the
sync FET will conduct and current will flow up from
ground to maintain the inductor current at its previous
value. Inductor current ramps up at a linear rate when
Q1 is on and down at a linear rate when Q2 is on
because in both cases there is constant voltage
forcing function.
When the body diode is conducting there will be a
negative voltage on the drain of Q2 due to the
direction of current flow and the body diode drop. The
voltage drop across the source to drain will be much
less than across the diode. Hence, when Q2 turns on
the phase node voltage (Q1 source - Q2 drain) will be
approximately equal to ground.
When Q1 turns on, the body diode of Q2 is still
conducting so current must be commutated from the
diode to the upper FET. Due to slow recovery time and
stored charge in the Q2 body diode a current spike will
be present on the leading edge of Q1 drain current.
This condition can be greatly helped by putting a
Schottky diode in parallel with the synchronous FET.
Figure (3) shows some waveforms at key nodes. High
frequency details have been omitted for ease of
understanding.
The inductor also works with the output capacitor to
form a low pass filter that filters the voltage pulses
present on the phase node and results in a clean DC
output. The voltage on the top of the output capacitor
is a saw-tooth that is centered on the output voltage
and has an amplitude equal to the inductor ripple
current times the capacitor ESR.
Microsemi
Integrated Products
11861 Western Avenue, Garden Grove, CA. 92841, 714-898-8121, Fax: 714-893-2570
Page 3
LX1672
PRODUCT DESIGN GUIDE
APPLICATION INFORMATION (CONTINUED)
RSET
ISET
CS1
+12V
CS Comp
IRESET
PWM
+
VTRIP
VS1
V in
ISET
R
Q
S
Q
R2
VC1
CIN
R1
HO1
V out
L1
ESR
EO1
LO1
COUT
PG1
+5V
Error Comp
+
FB1
VCCL
Hiccup
-
+
Amplifier/
Compensation
VREF
16V
20k
Ramp
Oscillator
UVLO
16V
+5V
UVLO
VCC
S
F
FAULT S
R
S
SS1
TEMP
5.5V
SS2
SS/ENABLE
DIS1
PWGD
(MLP Only)
CSS
Figure (2a) PWM Block Diagram Phase 1
Copyright © 2000
Rev. 0.2a, 2006-03-08
Microsemi
Integrated Products
11861 Western Avenue, Garden Grove, CA. 92841, 714-898-8121, Fax: 714-893-2570
Page 4
LX1672
PRODUCT DESIGN GUIDE
APPLICATION INFORMATION (CONTINUED)
RSET
ISET
+5V
CS2
CS Comp
IRESET
VS2
VTRIP
Vin
5V Typ
PWM
+
R
Q
ISET
S
Q
VC2
CIN
HO2
L
V OUT
ESR
EO2
LO2
LPF2
COUT
PG2
+5V
Error Comp
+
FB2
VCCL
Hiccup
-
+ Amplifier/
Compensation
VREF
RF2
16V
20k
LPF1
+5V
16V
UVLO
Ramp
Oscillator
UVLO
VCC
S
F
FAULT S
R
S
SS1
TEMP
5.5V
SS2
PHASE1
SS/ENABLE
DIS2
CSS
Figure (2b) - PWM Block Diagram Phase 2 Connected in LoadSHARE mode
Copyright © 2000
Rev. 0.2a, 2006-03-08
Microsemi
Integrated Products
11861 Western Avenue, Garden Grove, CA. 92841, 714-898-8121, Fax: 714-893-2570
Page 5
LX1672
PRODUCT DESIGN GUIDE
BUCK REGULATOR WAVEFORMS
Gate Drive Voltage
Q1 Gate
VSOURCE
100ns
non-overlap
100ns
non-overlap
Q2 Gate
0V
MOSFET Current
IQ1
IQ2
Inductor Current
IRIPPLE
IL
0A
Output Capacitor Current
Q+
ICO
tON/2
tOFF/2
Output Voltage
VOUT
0V
Figure (3) - Buck Regulator Waveforms
Copyright © 2000
Rev. 0.2a, 2006-03-08
Microsemi
Integrated Products
11861 Western Avenue, Garden Grove, CA. 92841, 714-898-8121, Fax: 714-893-2570
Page 6
LX1672
PRODUCT DESIGN GUIDE
3.0 BOOTSTRAP
The bootstrap circuit will add approximately 5 volts to
the PWM input (high side FET drain) to provide
sufficient voltage to enhance the high side FET. See
Figure (4a)
The bootstrap circuit consists of a capacitor (C) that
charges to 5 volts through the bootstrap diode (D)
when the lower FET (Q2) turns on. When the upper
FET (Q1) turns on, the five volt charge on the
capacitor is coupled to the VCX pin and added to the
FET source voltage. This back biases the diode to
allow the VCX pin to float up to Q1 source potential
plus 5 volts. The diode must have a low stored
charge to prevent partial discharge of the capacitor,
Schottky diodes are quite often used for this reason.
The capacitor must store enough charge over a PWM
cycle to keep the driver supply voltage from drooping.
For most applications a .1 uf capacitor is sufficient.
As long as the bootstrap capacitor is substantially
larger than the FET input capacitance it will store
enough charge for proper operation. A value that is 50
to 100 times the FET gate capacitance (Ciss) will be
adequate.
Q1
DC Vout
Vout
Q2
Figure (4b) - Bootstrap with Rectifier / Filter
4.0 SOFT START , ENABLE
Each PWM phase has a soft start pin SSX that is used
to connect an external soft start capacitor Css. Css is
charged by an internal 20KΩ resistor (Rss) connected
to the reference voltage. The SSX pin also provides an
enable function, holding the SSX pin low disables the
corresponding phase.
The LDO has an enable pin LDDIS that disables the
LDO output when it is held high.
V REF
VIN
+5V
Vin
+5V
40m V
VCx
D
S S C o m p a ra to r
+
20K
R SS
Q1
-
a
Phase
Mode
HOx
VCCL
C
VOUT
Q2
SS/
E n a b le
a
300K
H ic c u p
LOX
F ro m C u rre n t
L im it C o m p a r a t o r
CSS
To
PW M
PGx
Figure (4a) - Bootstrap Circuit
Figure (5) - Simplified Soft Start
When using this bootstrap configuration the voltage
on the VCX pin is a square wave between the PWM
input voltage and +5 volts.
At power on the under voltage lockout conditions, see
data sheet for values, for VCC, VCCL and VCX must
be met before Css will begin to charge. The output
voltage will then increase from zero volts towards Vout
and track the Css voltage until the soft start circuit is
switched out after three time constants of Rss-Css.
The three time constants allow Vout to be within 95%
of the correct value at the termination of soft start, an
internal comparator monitors the soft start pin and
terminates the soft start cycle.
The LX1672 LDO power comes from the VC1 pin and
if a bootstrap circuit is used for phase 1 the LDO
supply will not be a filtered DC and the LDO will not
function properly. A solution is to use the rectifier /
filter in Figure 4b which will give a clean DC output
due to the second diode and capacitor.
When using the bootstrap or rectifier / filter the voltage
on the VCX pin must not reach a level where
excessive current will flow into the internal zener
diodes on the VCX pins.
Copyright © 2000
Rev. 0.2a, 2006-03-08
During the soft start ramp period the Over Current
Protection is functional but the hiccup function is
disabled till the soft start interval is complete.
Microsemi
Integrated Products
11861 Western Avenue, Garden Grove, CA. 92841, 714-898-8121, Fax: 714-893-2570
Page 7
LX1672
PRODUCT DESIGN GUIDE
If more than one phase is using the same input rail
then the soft start timing should insure that they do not
come up at the same time to prevent loading the input
during startup. If phases 1 and 2 are used in Bi-Phase
phase 2 is controlled by the output of phase 1 so it
automatically tracks the rise in phase 1 output,
however a phase 2 soft start capacitor is still required
to insure that the internal comparator for phase 2 does
not attempt to terminate the start up sequence early.
5.0 OVER CURRENT PROTECTION AND HICCUP
The external resistor RSET along with Q1 RDS(ON) will
determine the current limit set point. See Figure (6),
the voltage drop across Q1 RDS(ON) plus the drop
developed across RSET by the internal 50uA current
source must exceed the internal 300 mv threshold to
initiate a current limit. There is a 350nSec delay in the
current limit comparator to prevent false triggering due
to ringing and noise. Note that the RDS(ON) of any FET
has a positive temperature coefficient, be sure to use
the value corresponding to the expected operating
temperature, usually at least 100 ° C, at the FET die.
Once the upper FET RDS(ON) is known the value of
RSET can be calculated.
PWM Input
+
RDS(on)
VSx
PWM
+
When the soft start interval begins the PWM is
disabled. No PWM pulses will be generated until the
ramp has reached a 300mV level where a comparator
enables the PWM. This will result in a current spike
into the output capacitor that may result in a series of
over current commands for a few cycles until the
output capacitor has charged up to the same level as
the soft start ramp.
300mV
Q1
Reset
-
CSx
350nSec
Delay
VOUT
Q2
Hiccup
50µA
+
RSET
Figure (6) - Over Current Protection
The RSET resistor and VCs pin should be as close to a
Kelvin connection to the FET source and drain
terminals as possible. Very low RDS(ON) values will
result is significant errors otherwise.
The Over Current Protection is functional during the
soft start ramp period. If currents in access of ICL are
detected the PWM will be commanded off for the
remainder of that cycle but will restart on the next
PWM cycle.
If no current limit is wanted the RSET resistor can be
left out and the CSX and VSX pins connected
together and tied to VCC. A floating VSX pin will result
in operation resembling a hiccup condition.
Where ICL is the current limit threshold.
The hiccup function is disabled during the soft start
ramp period. Once the soft start ramp has been
terminated the hiccup function will be enabled. If an
over current command is issued after the termination
of soft start the soft start capacitor will be pulled low by
the 300KΩ resistor, Figure 5, the switch in series with
RSS will open during this time to allow discharging
CSS. The voltage on CSS will drop till it reaches a
lower limit of 200mV when a comparator will begin the
soft start interval again.
Note: The minimum value for RSET is 1KΩ to prevent
damage to the LX1672 input, the maximum value is
6KΩ where VCL equals zero (50uA x 6KΩ = 300mV).
The hiccup process will continue indefinitely if the load
is greater than the current limit threshold determined
by RSET.
At low duty cycles where the upper gate drive is less
than 350nS wide the 350nS delay for current limit
enable may result in current pulses exceeding the
desired current limit set point. If the upper MOSFET
on time is less than 350nS and a short circuit
condition occurs, the duty cycle will increase since
Vout will be low. The current limit circuit will be
enabled when the upper gate drive exceeds 350nS.
The actual peak current limit value will be higher than
calculated with the above equation.
6.0 COMPONENT SELECTION
R SET =
(
300 mV − I CL × R DS(ON)
)
50 μA
Short circuit protection still exists due to the narrow
pulse width even though the magnitude of the current
pulses will be higher than the calculated value.
Copyright © 2000
Rev. 0.2a, 2006-03-08
Component selection in a DC to DC converter requires
a lot of consideration and tradeoffs since there are a
number of conflicting requirements such as size, cost,
and performance. The components that are the most
critical and require the most effort are the switching
FETs , inductor , and input / output capacitors. The
output filter is generally designed with output ripple
and load transients as the main consideration but it
has a very important role in loop compensation that
should be considered.
Microsemi
Integrated Products
11861 Western Avenue, Garden Grove, CA. 92841, 714-898-8121, Fax: 714-893-2570
Page 8
LX1672
PRODUCT DESIGN GUIDE
The LX1672 is a voltage-mode controller intended for
use in the continuous mode (i.e., inductor current
never reaches zero). All equations presented in this
document are based on these conditions.
7.0 FET SELECTION
Because the PWM FETs are the primary active power
handling components they must be carefully chosen
and properly sized for the power that will be
dissipated. In many cases the main and synchronous
FETs can be the same part number but as switching
frequencies and currents become higher it is often
desirable to use different devices that are optimized
for each circuit application.
The upper FET is
commonly called the main switch and the lower FET is
typically referred to as the synchronous switch since it
is used as a synchronous rectifier.
Eq (1) Main FET Duty cycle
D=
VOUT
VIN − VOUT
VIN
= 1−
VOUT
VIN
= 1− D
In a buck regulator topology there are no high
transients that require a drain to source voltage much
in access of the PWM input voltage. On resistance,
maximum drain current, and thermal performance are
closely related and must be considered together.
Thermal management is a key element in a buck
regulator design and requires careful component
selection and board layout. The gate threshold voltage
must be low enough to guaranty that the FET will be
fully enhanced with the available gate drive voltage.
Gate charge, Qg, should be less that 40nC.
8.0 UPPER FET POWER DISSIPATION
The power dissipation in the upper FET, Q1 is the
sum of the switching loss and conduction or I2 R loss.
As a close approximation it can be assumed that the
current through the upper FET is a linear ramp during
the switching interval.
Eq (3)
PQ1 =
⎡⎛ I D ⎞
⎤
⎢⎣⎜⎝ 2 ⎟⎠ × VIN × (t r + t f ) × f ⎥⎦ +
[I
2
D
× R DS(on) × D
]
Where ID is the FET drain current, tr and tf are the rise
and fall times, f is the PWM switching frequency, and
D is the duty cycle. For some typical numbers we will
pick:
Copyright © 2000
Rev. 0.2a, 2006-03-08
Vin = 5 V
Vout = 1.5 V
Iout = 5 A
This gives a duty cycle of (1.5V)/(5VA) = 0.3 or 30%.
Using the Si4842DY with the circuit values given
above, the switching loss in the main FET is
[(5/2)(5)(80nS + 80nS) (300KHz) = 0.60W. Note:
These turn on and off times are approximate in-circuit
values. Using RDS(ON) at 100 ° C the conduction
loss is (5A2)(0.0084Ω)(0.3) = 0.063W.
Total main FET power dissipation equals (0.60+0.063)
= 0.663W.
9.0 LOWER FET POWER DISSIPATION
The lower FET has no significant switching loss due to
the parallel diode. All power dissipation is from the I2R
loss during conduction; the formula for power
dissipation reduces to:
Eq (4)
VIN
Eq (2) Sync FET Duty Cycle
=
•
•
•
2
PQ 2 = ⎡ I D × R DS(on) × (1 − D) ⎤
⎢⎣
⎥⎦
With the previous operating conditions, the conduction
loss equals (5A2 x .0084Ω) x (1-0.3) = 0.147W. The
sync FET body diode will start to conduct immediately
after the main FET turns off and before the
synchronous FET turns on. Hence, there is always a
short non-overlap period to prevent simultaneous
conduction of both FETs. The body diode causes the
sync FET to transition with a low drain to source
voltage while the main FET transitions with a drain to
source voltage of approximately Vin. When the
synchronous FET turns off, the body diode will conduct
for a short period of time before the main FET turns
on. Due to this it is assumed that the switching losses
in the synchronous FET are negligible. There is some
power dissipation in the body diode but the conduction
time is so short that these losses are also very low.
The body diode in most FETs is a very non-ideal diode
with a long recovery time and high stored charge.
10.0 GENERAL FET CONSIDERATIONS
In the above example the I2R losses in the main FET
are less than one half that of the synchronous FET
(i.e., the switching losses are dominate). Note that the
temperature, drain current, and gate-to-source voltage
all have an effect upon the RDS(ON) value and should
be taken into account. A common practice is to pick a
main FET that is optimized for switching losses and a
sync FET that has low RDS(ON) to minimize conduction
losses. As currents get higher it is possible to parallel
several FETs to get a lower effective RDS(ON) and
spread the heat over a larger area to get better heat
sinking from the circuit board.
Microsemi
Integrated Products
11861 Western Avenue, Garden Grove, CA. 92841, 714-898-8121, Fax: 714-893-2570
Page 9
LX1672
PRODUCT DESIGN GUIDE
The reverse transfer capacitance, Crss, is another
critical parameter that effects FET switching by
requiring more gate drive. Also when the upper FET
turns on it will attempt to pull the lower gate high
through Crss of the lower FET. If this effect is to great
the lower FET can turn back on causing cross
conduction.
11.0 FET GATE DRIVE
Any FET used in a switching application should have
a low total gate charge to minimize the peak current
required to charge and discharge the gate
capacitance each cycle. If the gate charge is too high
it places a greater demand on the driver causing
possible thermal problems and slow switching.
Increasing FET transition times further degrades the
circuit’s switching losses. The FET drivers in the
LX1672 must charge and discharge the input
capacitance of the external FET switches on every
PWM cycle. Due to the very fast switching times
needed for efficiency, the peak current required to
charge the gate is significant. Charging a 2500pf gate
capacitor to 5 volts in 50ns requires an average
current of 0.25A. This current flows through the
controller output transistors and results in internal
power dissipation. It is common to use total gate
charge rather that gate capacitance to calculate drive
power. As a guideline 40nC should be considered the
maximum FET input charge to be driven. The input
capacitance of a FET is not a simple capacitor but
rather a combination of several internal capacitors that
change value with operating conditions. FET data
sheets show a total gate charge Qg that accounts for
this complexity and can be used to calculate the
power required to drive the gate.
The charge on a capacitor is:
Eq (5)
2
P = QgV f
Where Qg is the total gate capacitance from the FET
data sheet, V is gate drive voltage, and f is the PWM
frequency.
The LX1672 evaluation board is typically delivered
from the factory with the lower gate drivers powered
by 5 volts on the VCCL pin. The upper gate drivers for
phase 1 and 2 (VC1 and VC2) are connected to 12
Copyright © 2000
Rev. 0.2a, 2006-03-08
Lower gate drivers = 40nC x 5V x 300KHz =
60mW each.
•
Upper drivers phase 1 and 2 = 40nC x 12V x
300KHz = 144mW each.
•
The total gate drive power is 2(60)+(2)144mW =
408mW.
With Vcc at 5 volts and an Icc of 15ma there is an
additional 75mW of package dissipation that must be
added to the 408mW for gate drive. This equals 483
mW total for the LX1672 TSSOP package. The data
sheet specifies the TSSOP package junction-toambient (θja) thermal resistance at 85°C per watt. This
value assumes that the TSSOP package is mounted to
a circuit board. The junction temperature can be
calculated from:
Eq (8)
TJ = TA + (PD × θ JA )
Assuming an ambient temperature of 23°C we get a
junction temperature of:
TJ = 23 + (.483 × 85) = 64 C
The 150°C maximum operating junction temperature
allows a maximum rise of:
150 − 64 = 86° C .
It is clear from the above that all output driver stages
should be operated on the lowest voltage that will
properly enhance the FETs.
12.0 FET THERMAL CONSIDERATIONS
P = CV f
By substituting terms the power required to drive the
FET gate is:
Eq (7 )
•
The LX1672 can be operated with up to 16 volts DC
on the VCCL pin but that would be inadvisable from a
power dissipation perspective.
Q = CV
The power required to charge a capacitor is:
Eq (6)
volts. Using these numbers we can calculate the
power required to drive the FETs. Assuming a gate
charge of 40nC and a 300KHz LX1672 we have:
Since it is always desirable to avoid the use of heat
sinks (other than the circuit board itself) careful
consideration should be given to component selection
and layout from a thermal point of view. FET package
selection is a significant part of the overall thermal
picture and has a significant effect on thermal
performance. For example, large cases like the D2
package have a much lower junction to circuit board
thermal resistance than the popular S08.
Another popular surface mount package is the D pack
slightly smaller than the D2 and has a corresponding
lower current handling capability. The tried and true
Microsemi
Integrated Products
11861 Western Avenue, Garden Grove, CA. 92841, 714-898-8121, Fax: 714-893-2570
Page 10
LX1672
PRODUCT DESIGN GUIDE
TO220 is always worth looking at if the mounting tab
is acceptable, it is a low cost package with good
performance. The TO220 is quite often used with no
heat sink on the tab but the thermal resistance to air is
fairly high so this requires careful consideration. For
surface mount components the maximum allowable
thermal resistance from the circuit board to air Rθ(CB-A)
can be calculated by.
Rθ (CB − A) ≤
(TJ max − TA max ) − Rθ
PD max
Eq (9)
(J − C)
This formula assumes that the FET case and the
circuit board are at the exact same temperature which
is valid only for surface mount components. Once the
power dissipation of each FET has been determined
and a maximum operating junction temperature
chosen the required thermal resistance of the circuit
board to air can be calculated.
If we look at the example where an Si4842DY FET
dissipates 1 watt at a maximum ambient of 85 ° C,
and allow the junction temperature to reach the
maximum data sheet spec of 150 ° C, θj-c of the
Si4842DY is 16° C per watt.
Rθ (CB − A)
(150 − 85) − 16 = 49° C/W
≤
1
The circuit board to air thermal resistance Rθ(CB-A)
must be less than 49 degrees C per watt. The
thermal resistance to air of a 1 in2 circuit board, with
properly designed copper power and ground planes to
help spread the heat, will typically be 45°C per watt.
Keep in mind that adjacent heat generating
components will add to the thermal load on the circuit
board and require more area. Small clip-on heat sinks
and airflow will help reduce component heat
dissipation considerably.
A lower inductance value will provide better transient
response and a smaller package at the expense of
increasing the ripple current. As the inductance value
increases, transient response degrades and
component size increases; however, the ripple current
decreases. Ripple current is an important factor since
it must flow into the output filter capacitor and will
result in an output ripple voltage that is the product of
the capacitor ESR and the inductor ripple current.
Inductor ripple current is usually set to be within 10%
to 30% of the output current and typical inductance
values range from 1 to 10uH. The output inductor and
capacitor play a critical role in determining overall
regulator performance. Hence, it may be necessary to
do an iterative design if the first inductor choice results
in an unacceptable capacitor value. The relationship
for voltage across an inductor takes the well
recognized form of:
V=L
Eq (10)
di
dt
In a buck regulator the forcing function is always a
constant voltage source resulting in a linear ramp of
current so we can use the ΔI and ΔT terms rather than
be concerned with integrating a nonlinear function.
The rate of change, or slew rate, that can be achieved
for an inductor is then
ΔI =
Eq (11)
VΔ t
L
Substituting terms used in a buck PWM regulator
ΔI =
Eq (12)
VIN − VOUT D
×
L
fs
and rearranging terms to calculate inductance
13.0 INDUCTOR SELECTION
The inductor is usually one of the largest components
in a buck regulator and has a large influence on
performance. There are two loss mechanisms copper
loss and core loss both of which must be considered
in order to maximize circuit efficiency.
Since it is
often desirable to use off the shelf components some
compromises are usually made with the inductor.
Selecting an inductor with a low series resistance is
always desirable. Further, make sure that the core
material will perform properly at the PWM switching
frequency.
The core must be able to handle the maximum output
current plus one half of the ripple current without a
significant drop in inductance. If the core is allowed to
Copyright © 2000
Rev. 0.2a, 2006-03-08
saturate, the inductance will drop dramatically and
result in a significant increase in ripple current.
L=
Eq (13)
VIN − VOUT D
×
ΔI
fs
Where fs is the PWM switching frequency , L is the
inductance and D is the Duty Cycle.
To examine the inductor slew rate for a step change in
output current
Trise = L •
CurrentStep
Vin − Vout
Tfall = L •
CurrentStep
Vout
Eq(14a)
Eq(14b)
Microsemi
Integrated Products
11861 Western Avenue, Garden Grove, CA. 92841, 714-898-8121, Fax: 714-893-2570
Page 11
LX1672
PRODUCT DESIGN GUIDE
Equations 14 a and b are useful because they give us
the maximum slew rate that can be achieved for a
given inductor.
ESR due to the current step and
combined.
An important point on slew rate is that the above
number does not represent loop response. The final
rate of change of the output voltage in response to a
step change in current is determined by the frequency
response of the closed loop and the loop phase
margin. The control loop will always respond slower
than the inductor current ramp.
Eq (16)
When a step change is load current is known:
Capacitor ESR ≤
L=
5 − 1.5
8 × .2
×
D
fs
Once the inductor value has been determined an
output capacitor can be chosen. The ESR value is the
main selection criteria for this capacitor. Package size
at the required voltage and capacitance will also need
to be determined after the ESR is known. The output
ripple voltage and allowable error for a step change
are important since both of these influence the
required ESR of the capacitor.
If no load step is specified
Eq (15)
Output Ripple Voltage
100 mV- 25 mV
1.6 A + 4 A
= .0134 Ω
This ESR value is easily achieved. It is almost always
true that once the ESR requirement is met the bulk
capacitance will be adequate for energy storage.
Some common capacitor types are the Sanyo MV-GX,
and Sanyo OS-CON. Others like the surface mount
Panasonic SP-CAP also have low ESR. It is common
practice to parallel a number of capacitors to get the
required ESR while maintaining acceptable package
size. Do not overlook the reliability of the capacitor at
the required temperature, operating voltage, and ripple
current.
The main criteria for selecting the input capacitor is
ripple current. When the PWM main FET switches on it
places a high current demand on the input rail. The
energy must be supplied by the input capacitor
because the input path typically has sufficient
impedance such that it cannot supply the peak
demand made by the PWM.
For a single phase PWM the worst case input current
condition occurs when the duty cycle is 50% resulting
in an RMS current of Iout/2.
A two phase (or Bi-Phase) PWM improves this by a
factor of two at its worst-case point. The worst case
condition will occur at 25% and 75% duty cycle with
the RMS input current equal to Iout/4.
The RMS input current for a single phase PWM is
given by
50 mV
1.6 A
≤ .031 Ω
Eq(17)
This will not be a difficult ESR to achieve with good
quality capacitors. A 4A load step will give a drop in
the output of .031Ω x 4A =125mV which is greater
than the total 100mV allowed so we will need a lower
Copyright © 2000
Rev. 0.2a, 2006-03-08
Capacitor ESR ≤
Inductor Ripple Current
For the 50mV ripple voltage:
ESR ≤
Inductor Ripple Current + Load Step
15.0 INPUT CAPACITOR SELECTION
= 2.18 μH
14.0 OUTPUT CAPACITOR SELECTION
Capacitor ESR ≤
Total Allowable Voltage Drop
The total allowable drop will be the amount allocated
to the load step minus one half of the ripple voltage.
The following example calculations use the conditions
listed below:
Vin = 5 V
Vout = 1.5 V
D = 1.5/5 = 30%
I out = 8 A
Inductor ripple current = 20%
PWM frequency = 300 KHz
Maximum load step = 4A
Output ripple < 50mV
Total output change due to 4 A step < 100 mV
Using equation 13, the inductance value is determined
as follows:
output ripple
I IN (RMS) = I OUT
D(1 − D)
For a Bi-Phase PWM the input ripple current can be
calculated by
Eq(17a) IIN (RMS) = IOUT D(0.5 − D) for D < 0.5
Microsemi
Integrated Products
11861 Western Avenue, Garden Grove, CA. 92841, 714-898-8121, Fax: 714-893-2570
Page 12
LX1672
PRODUCT DESIGN GUIDE
As shown in Figure (7) for duty cycles greater than
50% the curve folds over around the 50% point.
careful to use the maximum input voltage when sizing
the input capacitor.
As with the output capacitor the bulk capacitance will
be more than adequate once the main criteria of ripple
current has been met. ESR should also be considered
because it results in an overall loss of efficiency, the
power loss in the capacitor is equal to (I2 x ESR).
This loss should not be ignored if maximum efficiency
is a goal. The ripple current in the capacitor will also
cause component heating, adversely effecting
reliability.
Always consider temperature and voltage stress at
the worst case conditions when picking a
16.0 EFFICIENCY
DUTY CYCLE (VOUT / VIN)
Figure (7) RMS Ripple Current
The values shown above are for a steady state
condition when the control loop is regulating Vout to
the proper value. During startup the duly cycle can
approach 100% for short periods and a short circuit
condition can result in a very low duty cycle since the
on time of the upper MOSFET will be limited by the
350nSec delay for the current limit circuit. If possible it
is advisable to pick input capacitors that are capable
of handling the worst case RMS ripple current.
For a single phase Buck PWM and the 8A output in
our previous example with a duty cycle of 30% the
actual input ripple current from Eq 17 would be 3.67A
RMS. Since the input voltage is always higher than
the output this puts an additional burden on the input
capacitor.
Several capacitors connected in-parallel will be
required to meet the 3.67A RMS ripple current and
voltage constraint. For example, the Sanyo MVGX
series 10 volt 1500uF capacitor has a ripple current of
1.3Arms. This would require 3 devices in-parallel to
meet the 3.67A RMS input ripple requirement.
Figure (7) shows that considerable deviation from a
50% duty cycle is required to gain any significant
reduction in RMS input current from the worst case of
0.5 x Iout. If the input voltage is not well controlled be
Copyright © 2000
Rev. 0.2a, 2006-03-08
The key components that significantly effect overall
efficiency are the power handling components in the
PWM output stage. The switching FETs have the
greatest effect on overall efficiency. In order to
maximize efficiency, the selected FETs must have the
lowest possible switching and conduction losses.
These factors are covered in the Component Selection
section. FET gate drive can require significant power,
always pick FETs with low gate charge and use the
lowest practical gate drive voltage. PWM frequency is
a significant factor in overall design. At higher PWM
frequencies the switching losses become greater.
However, at higher PWM frequencies, other
components become smaller.
A trend in FET development is to improve switching
and conduction losses hence, better components
continually become available.
The Inductor can
dissipate a significant amount of power if the I2R
losses are too high or if the inductor core has a lot of
loss at the PWM frequency. Always pick an inductor
with low ESR and be certain that it will operate
efficiently at the maximum PWM frequency. If the core
begins to saturate the core loss and ripple current will
increase. Be certain that the inductor is rated for the
full output current plus one half of the ripple current.
Remember that the ESR of input and output capacitors
dissipate power and should be taken into account.
Even with good power planes and wide traces drop
across the circuit board copper can be significant.
The MOSFET upper driver supply voltage (VCX)
should be kept as low as possible for best efficiency.
17.0 LOOP COMPENSATION
A complete mathematical analysis of a buck PWM is
possible however, due to the number of component
tolerances and variations in input voltage a simple
Microsemi
Integrated Products
11861 Western Avenue, Garden Grove, CA. 92841, 714-898-8121, Fax: 714-893-2570
Page 13
LX1672
PRODUCT DESIGN GUIDE
graphical method using a Bode plot will generally give
satisfactory results.
A buck regulator using the LX1672 controller is quite
forgiving and generally does not require any additional
compensation components to implement a stable
system with sufficient phase margin. The evaluation
board component values result in a stable system with
superior transient response. One version of the
factory delivered eval board has phase 1 and 2
operating in Bi-Phase, the feedback for the phase 1
error amplifier is a resistor that sets a fixed gain. The
phase 2 error amplifier has a capacitor in the
feedback path to form an integrator that gives a very
high gain at low frequencies so that it can track phase
1. This is one example of how the feedback
components can be used.
The output inductor and capacitor are the primary
components that determine the frequency response of
the closed loop by creating a double pole and a zero
that set the dominant break points and determine the
unity gain crossover point (unless frequency
compensation components are added). Loop gain is
another variable used to determine the unity gain
crossover frequency and overall loop stability.
A double pole at a break point frequency (fp) is
determined by the output inductor and capacitor:
f( p
)=
Eq (18)
1
2 π LC
A zero is set by the output capacitor at the frequency
(fz):
f( z
)=
Eq (19)
(
1
) ESR× COUT
2π
Figure (9) shows the blocks in the control loop that
effect frequency response and gain. The error
amplifier is the primary gain block and can be treated
like any operational amplifier. The PWM and FET
switches are considered one stage. The amplifier gain
is simply the feedback impedance between the FBX
and EOX pins divided by the input impedance. The
resistors connecting the PWM output to the amplifier
input at the FBX pin ( ZIN) must be considered as a
Thevinin equivalent to calculate the correct gain.
Zf
VIN
FBx
A common design goal is to have the unity gain
crossing occur at a slope of -20dB per decade. It is
possible to cross through zero gain at -40 dB per
decade if the ESR zero is close enough to the zero
crossing to provide sufficient phase margin. The unity
gain crossing should be below the switching
frequency pole by a factor of 5 if possible. Once the
break points have been calculated a Bode plot can be
drawn on semi log paper to obtain a graphical
representation of the loop response. A Bode plot for
a PWM regulator using the LX1672 is shown in Figure
(8)
EOx
Ref
PWM
VOUT
ESR
}
1.25V pp
ZIN
Figure (9) - LX1672 Control Loop Block Diagram
Amplifier Gain =
Eq (20)
Zf
Z IN
The gain of the PWM stage is:
Gain dB
37dB (typ)
PWM Gain =
40dB / Decade
Eq (21)
20dB / Decade
fs
Log Frequency
fp
fz
fp: is the double pole set by L & C
fz: is a zero set by the capacitor ESR
fs: is a double pole at the switching frequency
Figure (8 ) - PWM Bode Plot
Copyright © 2000
Rev. 0.2a, 2006-03-08
PWM Vin
RampVoltage
Loop gain is the product of the amplifier and PWM
gains calculated with equations 20 and 21. Note that
these equations give a unitless numerical gain (A).
The gain in db is 20log(A), gain in db is added for
cascaded stages.
A very import point is that the input voltage is a gain
term in the loop transfer function so that varying the
voltage on the main FET drain will have an effect on
loop gain and stability. The difference in gain going
Microsemi
Integrated Products
11861 Western Avenue, Garden Grove, CA. 92841, 714-898-8121, Fax: 714-893-2570
Page 14
LX1672
PRODUCT DESIGN GUIDE
from 3.3 to 12 volts input is 20log12/3.3 = 11.2db, a
considerable change in loop gain.
sized in accordance with the current requirements, be
as large as possible, and connected as shown.
18.0 TRANSIENT RESPONSE
VIN
+5V
VCC
A useful check of performance is to apply a step
change of current to the output using a FET switch
with a series load resistor and observe the response
of the output voltage and inductor current. A large
overshoot that takes a number of cycles to settle
indicates low phase margin. A system with greater
phase margin will have less overshoot but take longer
to correct the error introduced into the feedback loop.
CIN
CVCC
Q1
L
LX1671
Q2
COUT
19.0 CIRCUIT BOARD LAYOUT
Switching regulators operating at multiple amps of
output current are very sensitive to layout due to the
high current pulses and fast rise times when the FETs
are switched. It is important to keep the currents in the
FET switching and filtering section confined to their
own ground plane and use a single point ground
connection to the analog and digital ground pins on
the LX1672.
The connection between the lower FET drain and
upper FET source should be as short as possible and
properly sized for current. Inductance in this path will
result in unwanted transients that will impair proper
switching operation. The input and output capacitor
ground leads should be returned to the source of the
lower FET. The PGX pins should also be connected at
the lower FET source. Large traces are required in all
output filter connections. A good layout keeps all
circulating currents produced by the output switching
and filter sections away from the low signal level
analog circuitry. Feedback lines going to the LX1672
for voltage and current should be routed on a lower
layer protected from the output section by a ground or
power plane. Separate bypass capacitors should be
used on the VCC and VCCL pins to prevent possible
UVLO problems during startup.
All filtering
components on lines bringing signals back to the
controller should be located near the controller and
tied to the ground plane for the analog ground pin.
The evaluation board is a six-layer design with
separate power planes for each supply voltage and a
ground plane. It would be possible to reduce this
without significant compromise in performance if very
careful attention is paid to critical high current paths.
The evaluation board uses a separate analog ground
plane located on the +5V PWB layer, this is an
important feature that should be included in any
layout.
Figure (10)
shows the output filter
components and FETs, the heavy traces should be
Copyright © 2000
Rev. 0.2a, 2006-03-08
VOUT
PGx
Figure (10) - PWM Switch and Output Filter
20.0 BI-PHASE PWM CURRENT SHARING
METHODS
There are several methods used to control the current
sharing in multiphase PWM regulators. Multiphase
controllers can have as few as two phases and in
many cases four or even six. The goal of any
multiphase controller is to proportion the current
between multiple phases to achieve a specified ratio
from L1 to Lx. By dividing the output current among
several phases it is possible to get much higher total
output current than could be achieved with a single
phase. The trend towards lower supply voltages and
higher currents has made the multiphase PWM
controller popular. Multi-phase designs offer significant
thermal advantages by spreading the heat out over a
larger area by using a number of FETs. An additional
advantages is reduced ripple current in the input and
output capacitors. The reduced ripple current in the
input capacitor is especially important since there is
less chance of disturbing the input rail and finding
input capacitors with low ESR and small size can be
difficult.
21.0 LOADSHARE
The LX1672 uses a Ratio Current Sharing method
(LoadSHARETM) that forces the currents through the
inductors to be proportional to the inductor DC
resistance. For Bi-Phase operation Phase 1 and 2 are
always the two that are used. Phase 1 is the control
phase and phase 2 is forced to follow. Phase 1 works
in a typical PWM configuration with the output side of
the inductor divided down and used as feedback to the
Microsemi
Integrated Products
11861 Western Avenue, Garden Grove, CA. 92841, 714-898-8121, Fax: 714-893-2570
Page 15
LX1672
PRODUCT DESIGN GUIDE
negative input of the error amplifier, the positive side
of the error amplifier is tied to the reference so the
loop is closed.
The phase 2 error amplifier has a filtered feedback
from the phase 1 phase node connected to its positive
input to act as a reference and a filtered feedback
from the phase 2 phase node connected to its
negative input. The corner frequency of the filters is
much lower that the PWM frequency so a DC voltage
is present on both error amplifier inputs. With phase 1
acting as a reference the filtered feedback from phase
2 closes the loop and forces the duty cycle of phase 2
to produce an average voltage that is identical to that
produced by phase 1. The output ends of the two
inductors are connected together and the input ends
have equal average voltages impressed on them so
the average current in each inductor is proportional to
the inductor resistance. If identical inductors are used
the current contribution for both phases will be equal.
This method has good feedback for current balancing
and can be modified to imbalance the average
currents if desired. By using different input voltages
and unbalancing the currents in the two phases the
power drawn from each input rail can be controlled to
fit within the available limits. Figure (11) shows this
configuration.
current to flow through one inductor than the other.
The inductor with the lower ESR value will have the
larger current. The inductor currents are directly
proportional to the ratio of the inductor’s ESR value.
The following circuit description shows how to select
the inductor ESR for each phase where a different
amount of power is taken from two different input
power supplies. A typical configuration will have a
+5V rail connected to the phase 1 half bridge driver
and a +3.3V rail connected to the phase 2 half bridge
driver. The combined output power for this core
voltage is 18W (+1.5V @ 12A). For this example the
+5V power supply will supply 7W and the +3.3V power
supply will supply the other 11W. 7W @ 1.5V is a
4.67A current through the phase 1 inductor. 11W @
1.5V is a 7.33A current through the phase 2 inductor.
The ratio of inductor ESR is inversely proportional to
the current ratio between phases.
ESR1 I2
=
ESR 2 I1
The lower ESR value will have the higher inductor
current. If the ESR of the phase 1 inductor is selected
as 10mΩ, then the ESR value of the phase 2 inductor
is calculated as:
VIN 1
Phase 1
PWM
+
Ref
⎛ 4.67 A ⎞
⎜
⎟ × 10 mΩ = 6.4 mΩ
⎝ 7.33A ⎠
L1
EA
EO1
FB1
VOUT
RF2
+
VIN 2
EA
Phase 2
PWM
-
FB2
L2
Depending on the required accuracy of this power
sharing; inductors can be chosen from standard
vendor tables with an ESR ratio close to the required
values. Inductors can also be designed for a given
application so that there is the least amount of
compromise in the inductor’s performance.
+5V @ 7W
EO2
There are four variants of the LoadSHARETM
topology that are explained in detail in the following
pages.
Bi-Phase, LoadSHARE (ESR Method)
The first method is to change the ratio of the inductors
equivalent series resistance, (ESR). As can be seen
in the previous example, if the offset error is zero and
the ESR of the two inductors are identical, then the
two inductor currents will be identical. To change the
ratio of current between the two inductors, the value of
the inductor’s ESR can be changed to allow more
Copyright © 2000
Rev. 0.2a, 2006-03-08
10mΩ
1.5V +
46.7mV
Figure (11) - LX1672 LoadSHARE Block Diagram
+3.3V @ 11W
4.67A
L1
6.4mΩ
L2
1.5V @ 12A
18W
7.33A
Figure (12) - LoadSHARE Using Inductor ESR
Bi-Phase, LoadSHARE (Feedback Divider Method)
Sometimes it is desirable to use the same inductor in
both phases while having a much larger current in one
phase versus the other. A simple resistor divider can
be used on the input side of the Low Pass Filter that is
taken off of the switching side of the inductors. If the
Phase 2 current is to be larger than the current in
Phase 1; the resistor divider is placed in the feedback
path before the Low Pass Filter that is connected to
Microsemi
Integrated Products
11861 Western Avenue, Garden Grove, CA. 92841, 714-898-8121, Fax: 714-893-2570
Page 16
LX1672
PRODUCT DESIGN GUIDE
the Phase 2 inductor. If the Phase 2 current needs to
be less than the current in Phase 1; the resistor
divider is then placed in the feedback path before the
Low Pass Filter that is connected to the Phase 1
inductor.
As in Figure (12), the millivolts of DC offset created
by the resistor divider network in the feedback path,
appears as a voltage generator between the ESR of
the two inductors.
A divider in the feedback path from Phase 2 will cause
the voltage generator to be positive at Phase 2. With
a divider in the feedback path of Phase 1 the voltage
generator becomes positive at Phase 1. The Phase
with the positive side of the voltage generator will
have the larger current.
Systems that operate
continuously above a 30% power level can use this
method, a down side is that the current difference
between the two inductors still flows during a no load
condition.
This produces a low efficiency condition during a no
load or light load state, this method should not be used
if a wide range of output power is required.
The following description and Figure (13) show how to
determine the value of the resistor divider network
required to generate the offset voltage necessary to
produce the different current ratio in the two output
inductors. The power sharing ratio is the same as that
of Figure (12). The Offset Voltage Generator is
symbolic for the DC voltage offset between Phase 1 &
2. This voltage is generated by small changes in the
duty cycle of Phase 2. The output of the LPF is a DC
voltage proportional to the duty cycle on its input. A
small amount of attenuation by a resistor divider
before the LPF of Phase 2 will cause the duty cycle of
Phase 2 to increase to produce the added offset at V2.
The high DC gain of the error amplifier will force LPF2
to always be equal to LPF1. The following calculations
determine the value of the resistor divider necessary to
satisfy this example.
+5V @ 7W
L1,
Switch
Side
100
62k
Not
Used
Resistor
Divider
Phase 2
Error Amp
1.5V
+46.7mV
4700pF
-
Resistor
Divider
62k
LPF2
Offset
Voltage
Generator
4.67A
Vout
1.5V @ 12A
18W
+
62k
V2
100
TBD
ESR L1
10mΩ
PWM
Input
FB2
L2,
Switch
Side
V1
Phase 1
+
LPF1
RF2
1.5V
+73.3mV
4700pF
ESR L2
10mΩ
Phase 2
7.33A
+3.3V @ 11W
Figure (13) - LoadSHARE Using Feedback Divider Offset
V1
K × 100
Where V1 = 1.5467 ; V2 = 1.5733 and K =
then TBD =
= 5.814 K
1− K
V2
Bi-Phase, LoadSHARE (Proportional Method)
The best topology for generating a current ratio at full
load and proportional between full load and no load is
shown in Figure (14). The DC voltage difference
between LPF1 and VOUT is a voltage that is
proportional to the current flowing in the Phase 1
inductor. This voltage can be amplified and used to
Copyright © 2000
Rev. 0.2a, 2006-03-08
offset the voltage at LPF2 through a large impedance
that will not significantly alter the characteristics of the
low pass filter. At no load there will be no offset
voltage and no offset current between the two phases.
This will give the highest efficiency at no load.
A speed up capacitor can be used between the offset
amplifier output and the negative input of the Phase 2
Microsemi
Integrated Products
11861 Western Avenue, Garden Grove, CA. 92841, 714-898-8121, Fax: 714-893-2570
Page 17
LX1672
PRODUCT DESIGN GUIDE
error amplifier.
This will improve the transient
response of the Phase 2 output current, so that it will
share more equally with phase 1 current during a
transient condition.
of the offset amplifier, and the value of Ra and Rb, will
determine the ratio of currents between the phases at
full load. Two external amplifiers are required or this
method.
The use of a FET input amplifier is required for the
buffer to prevent loading the low pass filter. The gain
L1,
Switch
Side
Offset Amp
LPF1
+
62k
+
Rin
-
-
Vos
Rf
1.5V
+46.7mV
4700pF
Phase 2
Error Amp
+
RF2
L2,
Switch
Side
+5V @ 7W
ESR L1
10mΩ
V1
4.67A
Phase 1
PWM
Input
Offset
Voltage
Generator
-
Vout
1.5V @ 12A
18W
+
FB2
LPF2
62k
62k
Ra
1.5V
+73.3mV
4700pF
1M
Rb
ESR L2
10mΩ
V2
Phase 2
7.33A
+3.3V @ 11W
Figure (14) - LoadSHARing Using Proportional Control
The circuit in Figure (14) sums a current through a
1MΩ resistor (Rb) offsetting the phase 2 error amplifier
to create an imbalance in the L1 and L2 currents.
Although there are many ways to calculate component
values the approach taken here is to pick Ra, Rb, Rin,
Vout, and inductor ESR. A value for the remaining
resistor Rf can then be calculated.
The first decision to be made is the current sharing
ratio, follow the previous examples to understand the
basics of LoadSHARE. The most common reason to
imbalance the currents in the two phases is because
of limitations on the available power from the input
rails for each phase. Use the available input power
and total required output power to determine the
inductor currents for each phase.
3) Calculate the offset voltage Vos at the output of
the offset amplifier
⎛ V 2 − V1⎞
⎟ × (Ra + Rb )
⎝ Ra ⎠
Vos = V 2 − ⎜
4) Calculate the value for Rf
(select a value for Rin typically 5KΩ)
⎛ Vos − Vout ⎞
⎟
⎝ Vout − V 1 ⎠
Rf = Rin ⎜
V 1 = L 1 Current × L 1 ESR + Vout
Due to the high impedances in this circuit layout can
effect the actual current ratio by allowing some of the
switching waveforms to couple into the current
summing path. It may be necessary to make some
adjustment in Rf after the final layout is evaluated.
Also the equation for Rf requires very accurate
numbers for the voltages to insure an accurate result.
V 2 = L 2 Current × L 2 ESR + Vout
Bi-Phase, LoadSHARE (Series Resistor Method)
Select values for Ra and Rb (Ra is typically 62KΩ ;
Rb is typically 1MΩ)
A fourth but less desirable way to produce the ratio
current between the two phases is to add a resistor in
All references are to Figure (14)
Calculate the voltages V1 and V2.
Copyright © 2000
Rev. 0.2a, 2006-03-08
Microsemi
Integrated Products
11861 Western Avenue, Garden Grove, CA. 92841, 714-898-8121, Fax: 714-893-2570
Page 18
LX1672
PRODUCT DESIGN GUIDE
series with one of the inductors. This will reduce the
current in the inductor that has the resistor and
increase the current in the inductor of the opposite
phase. The example of Figure 7 can be used to
determine the current ratio by adding the value of the
series resistor to the ESR value of the inductor. The
added resistance will lower the overall efficiency
.
22.0 OTHER MULTIPHASE METHODS
A different technique using inductor resistance has
been implemented by taking the AC and DC
components of the inductor current and using them to
drive a current sense amplifier. This technique works
well on paper but has the disadvantage of requiring
close tolerance components that must track over
temperature to work properly. With proper component
values it can work well and has active feedback to
control each phase.
23.0 DDR VTT TERMINATION SUPPLY
Digital Control Logic
Internal digital control logic can be used to break the
PWM cycle into a large number of clock cycles at a
frequency much higher than the PWM switching
frequency and then predict the switch points for the
next PWM cycle of each phase. Due to the open loop
nature of this approach it is prone to errors and may
not divide the current equally between phases. There
is no ability to compensate for differences in the
external components.
RDS(ON)
A method using the RDS(ON) of a FET has had success.
It actively measures the current through each leg for
proper balance. A downside to this is that the RDS(ON)
of every FET must be equal for accurate current
sharing. If all parts are from the same lot number and
operating at the same temperature this will work well.
Fortunately the positive temperature coefficient of a
FET helps to balance the current over temperature. As
FET RDS(ON) values become lower the feedback signal
continues to get lower in amplitude and can be difficult
to filter out of the ambient noise. This method does
have feedback to control current but the low input
signal levels may cause problems.
Copyright © 2000
Rev. 0.2a, 2006-03-08
Inductor Feedback
DDR memory systems require a termination voltage
(VTT) in addition to the line driver supply voltage
(VDDQ) and receiver supply voltage (VDD). The
JEDEC standard (EIA/JESD8-9A) for SSTL_2 does
not require that VDD and VDDQ be equal but it is
common practice to use the same voltage for both. A
unique feature of the VTT supply is that it must be
able to source and sink current as the signals on the
bus change states.
The LX1672 can supply both VTT and VDDQ by
using two of the three PWM phases. Since the
currents for VTT and VDD-VDDQ combined are
generally a significant number of amperes a switching
regulator is the only practical topology.
VTT for DDR SDRAM can be generated with the
LX1672 by using the positive input of the phase 2 error
amplifier (RF2) as a reference input. Using VREF,
which is defined as one half of VDDQ, for the phase
2 reference input will insure that VTT equals VREF
and that all voltages will track each other as required
by the JEDEC specification.
See Microsemi Application Note 17 for details on
DDR
termination requirements and
circuit
configurations.
Microsemi
Integrated Products
11861 Western Avenue, Garden Grove, CA. 92841, 714-898-8121, Fax: 714-893-2570
Page 19
LX1672
PRODUCT DESIGN GUIDE
RSET
ISET
+5V
CS2
VIN
CS Comp
IRESET
PWM
+
VTRIP
VS2
ISET
R
Q
S
Q
VC2
CIN
HO2
L
VTT
ESR
EO2
LO2
COUT
PG2
+5V
Error Comp
+
FB2
VCCL
Hiccup
-
+ Amplifier/
Compensation
+5V
VREF
RF2
20k
16V
Ramp
Oscillator
VREF
UVLO
16V
UVLO
VCC
S
F
FAULT
R
S
S
SS1
5.5V
TEMP
SS2
SS/ENABLE
DIS2
CSS
Figure 15 - DDR VTT Supply using Phase 2
Copyright © 2000
Rev. 0.2a, 2006-03-08
Microsemi
Integrated Products
11861 Western Avenue, Garden Grove, CA. 92841, 714-898-8121, Fax: 714-893-2570
Page 20