The SC475A is a versatile, constant on-time synchronous
buck, pseudo-fi xed-frequency, PWM controller intended for
notebook computers and other battery operated portable
devices. The SC475A contains all the features needed to
provide cost-effective control of system elements needing
voltage slewing. An integrated switch provides two resistorprogrammable DC output voltages controlled by the G0
input.
The output voltage is adjustable from 0.75V to 5V.
Additional features include cycle-by-cycle current limit,
voltage soft-start, under-voltage protection, programmable
over-current protection, soft shutdown, automatic power
save and non-overlapping gate drive. The SC475A
provides an enable input and a power good output which is
automatically blanked during output voltage transitions.
The constant on-time topology provides fast dynamic
response. The excellent transient response means that
SC475A based solutions require less output capacitance
than competing fi xed-frequency converters. Switching
frequency is constant until a step in load or line voltage
occurs, at which time the pulse density and frequency
will increase or decrease to counter the change in
output voltage. After the transient event, the controller
frequency returns to steady state operation. At light loads,
the automatic power save mode reduces the SC475A
frequency for improved effi ciency.
V
Programmable 0.75V to 5.25V with Integrated
OUT
Transition Support
V
Soft Shutoff at Output
Current Sense Using Low-side R
Range 3V to 25V
BAT
or Resistor
DS(ON)
Sensing
Adjustable Cycle-by-Cycle Valley Current Limit
325kHz Fixed-Frequency
Constant On-Time for Fast Dynamic Response and
Reduced Output Capacitance
Automatic Smart Power Save
Internal Soft-Start
Over-Voltage/Under-Voltage Fault Protection
Power Good Output with Transition Blanking
1μA Typical Shutdown Current
500μA Typical Operating Current
Tiny 3×3mm, 16 Pin MLP Package
Low External Part Count
Industrial Temperature Range
0.85% Internal Reference
1A/3A Non-Overlapping Gate Drive with
†
SmartDriver™ Technology
High Effi ciency > 90%
Device is Fully WEEE and RoHS Compliant
Exceeding the specifi cations below may result in permanent damage to the device or device malfunction. Operation outside of the parameters specifi ed in the
Electrical Characteristics section is not implied. Exposure to Absolute Maximum rated conditions for extended periods of time may affect device reliability.
ParameterSymbolMinMaxUnits
DH, BST to GND (DC)
DH, BST to GND (transient - 100nsec max)
DL to GND (DC)
DL to GND (transient - 100nsec max)
Smart Power Save ThresholdFB with Respect to Nominal+8%
Over-Temperature Shutdown
Logic Inputs/Outputs
Logic Input High VoltageEN, G01.2V
Logic Input Low VoltageEN, G00.4V
EN Input Bias CurrentEN = 5V-1+1μA
G0 Input Bias CurrentG0 = 5V5010μA
Power Good Output
Low Voltage
Gate Drivers
FB with Respect to Nominal+20+17+23%
FB Forced 50mV Above
Over-Voltage Fault Threshold
FB with Respect to Nominal;
Valid for 32 cycles after G0
Transition
(1)
Latching, >10°C Hysteresis160°C
R
= 10kΩ to VCC0.4V
PWRGD
5μs
+50%
Shoot-Through
Protection Delay
DL Pull-Down Resistance0.81.6Ω
DL Sink CurrentVDL = 2.5V3.1A
DL Pull-Up Resistance24Ω
DL Source CurrentVDL = 2.5V1.3A
DH Pull-Down ResistanceBST - LX = 5V24Ω
DH Pull-Up Resistance
DH Sink/Source CurrentV
Notes:
1) Guaranteed by design.
2) Semtech’s SmartDriver™ FET drive fi rst pulls DH high with a pull-up resistance of 10Ω (typical) until LX = 1.5V (typical). At this point,
an additional pull-up device is activated, reducing the resistance to 2Ω (typical). This negates the need for an external gate or boost resistor.
1) Available in tape and reel packaging only. A reel contains 3000
devices.
2) Available in lead-free packaging only. This product is fully WEEE,
RoHS and J-TD-020B compliant. This component and all homogenous subcomponents are RoHS compliant.
(1)
MLPQ-16 3X3
(2)
Marking Information
475A
yyww
xxxx
Marking for the 3x3mm MLPQ 16 Lead Package
nnn = Part Number (Example: 475A)
yyww = Date Code (Example: E652)
xxxx = Semtech Lot No. (Example: E901)
2BSTBoost capacitor connection for high-side gate drive.
3VCC5V power input for the internal circuits and gate drive outputs.
4DLGate drive output for the low-side external MOSFET.
SC475A
5GND
6RTN
7NCNot connected internally — leave unconnected or connect to GND.
8D0Drain of the internal MOSFET which is controlled by G0.
9FB
10VOUTOutput voltage sense point for determining On-Time.
11PGOOD
12ENEnable input - connect EN to RTN to disable the SC475A.
13ILIM
14G0
15NCNot connected internally — leave unconnected or connect to GND.
Power ground. This is the return point for the DL driver output, and the
reference point for the ILIM and zero cross circuits.
Return or analog ground for the FB input and FB resistor divider . Connect to
GND directly at the IC. All feedback components should connect to this ground.
Feedback sense point. The FB threshold is 0.75V; the resistor divider ratio
between VOUT and FB sets the output voltage. This ratio can be modifi ed
using the G0 input to switch a resistor in or out at D0.
Open-drain power good indicator - a high impedance indicates power is good.
An external pull-up resistor is required.
Current limit sense point — to program the current limit connect a
resistor from ILIM to LX or to a current sense resistor.
Control input for the D0 MOSFET. A logic low energizes the D0 MOSFET,
pulling D0 to ground.
16DHGate drive output for the high-side external MOSFET.
The SC475A is a synchronous power supply controller
which simplifi es the task of designing a dual-level power
supply suitable for controlling video chip sets and other
dual-voltage circuits. The SC475A provides an input (G0)
which controls an internal pull-down transistor used to
select from two adjustable output voltages.
Battery and +5V Bias Supplies
The SC475A requires an external +5V bias supply in addition to the VBAT supply. If stand-alone capability is required,
the +5V bias supply can be generated with an external
linear regulator.
The PWM control method is a constant-on-time, pseudofi xed-frequency PWM controller, see Figure 1. The ripple
voltage seen across the output capacitor’s ESR provides
the PWM ramp signal, eliminating the need for a current
sense resistor. The on-time is determined by an internal
one-shot whose period is proportional to output voltage,
and inversely proportional to input voltage. A separate oneshot sets the minimum off-time (typically 350ns).
TON
V
LX
C
IN
L
V
ESR
+
FB
C
OUT
FB Threshold
0.75V
R1
R2
V
FB
OUT
Q1
Q2
VBAT
V
LX
On-Time One-Shot (T
ON
)
The internal on-time one-shot comparator has two inputs.
One input senses output voltage via the VOUT pin, while
the other input samples VBAT via the LX pin and creates
a proportional current which charges an internal capacitor. The TON time is the time required for this capacitor to
charge from zero volts to VOUT, thereby making TON directly
proportional to output voltage and inversely proportional
to input voltage. This implementation results in a fairly
constant switching frequency without the need of a clock
generator. The internal frequency is optimized for 325kHz.
The general equation for the on-time is:
TON (nsec) = 2560 • (V
OUT/VBAT
) + 35
Immediately after the DH on-time, the DL output drives high
to energize the low-side MOSFET. DL has a minimum high
time of typically 350nsec, after which DL will continue to
stay high until one of the following occur:
FB drops to the 0.75V reference
•
The Zero Cross detector trips if power save is active
•
The Negative Current Limit detector trips
•
The Zero Cross detector monitors the voltage across the
low-side MOSFET and trips when it reaches zero. If this occurs on eight consecutive cycles, then DL will subsequently
shut off when the Zero Cross detector trips. See the PSAVE
Operation section. Both MOSFETS will then stay off until
FB drops to 0.75V, which will begin the next DH on-time.
This is normal operation at light load.
The Negative Current Limit detector trips when the drain
voltage at the low-side MOSFET reaches typically +80mV,
indicating a large negative current is being drawn through
the inductor from VOUT. When this occurs, DL drives low.
Both MOSFETS will then stay off until FB drops to 0.75V,
which will begin the next DH on-time. Tripping the Negative
Current detector is rare.
If DL drives low because FB has dropped to the 0.75V refer-
Figure 1
ence, then another DH on-time is started: this is normal
operation at heavy load. If DL drives low because of the
Zero Cross detector, then both DH and DL will remain low
until FB drops to the 0.75V reference, at which point the
next DH on-time will begin. This is normal operation at
light load.
The typical operating frequency is 325kHz. It is possible to
raise the frequency by placing a resistor divider between
the output and the VOUT pin, see Figure 2. This reduces
the voltage at the VOUT pin which is used to generate the
on-time according to the previous equation. Note that
this places a small minimum load on the output. The new
frequency is approximated by the following equation:
FREQ (kHz) = 325 ∙ (1 + R1/R2)
V
LX
L
ESR
C
OUT
+
R2
Figure 2
It is also possible to lower the frequency using a resistive
divider to the 5V bias supply, see Figure 3. This raises the
voltage at the VOUT pin which will increase the on-time.
Note that this results in a small leakage path from the 5V
supply to the output voltage. The resistor values should be
fairly large (>50kOhm) large to prevent the output voltage
from drifting up during shutdown conditions. Note that
the feedback resistors act as a dummy load to limit how
far the output can rise.
The new operating frequency is approximated by the
equation:
FREQ (kHz) = 325 ∙ ((R1 + R2) / (R1 + R2 ∙ V
Power Output
V
R1
pin 10
(VOUT)
1nF
CC/VOUT
OUT
))
SC475A
VCC
V
LX
L
ESR
C
OUT
R1R2
+
Figure 3
VOUT Voltage Selection
Output voltage is regulated by comparing VOUT as seen
through a resistor divider to the internal 0.75V reference,
see Figure 1. With D0 in the open state, the output voltage
is at the lowest value and is set by the equation:
V
= 0.75 • (1 + R1/R2)
OUT
Voltage Transition Control
The SC475A provides a G0 control input to allow selecting
between two output voltages. The output voltage is regulated by comparing the FB pin (connected to VOUT via an
external resistor divider) to the internal 0.75V reference.
The G0 input controls the gate of an internal MOSFET whose
source is connected to D0. Using G0 the user controls
whether D0 is grounded or open, which then controls the
resistor divider ratio for VOUT. A logic low signal on G0 will
connect D0 to ground.
When the G0 input changes state, this change quickly
causes three actions:
D0 changes state.
1.
The power good PGD output is temporarily latched into its
2.
present state. This prevents chattering or false tripping
while VOUT moves to the new level.
The output over-voltage OVP point is raised to 50% above
3.
nominal, or 1.125V at FB. When going from a higher to
lower voltage, the G0 change causes rapid change of D0,
which in turns cause a rapid change at FB. The temporary
increase in OVP allows the output to slew down to the new
VOUT voltage is regulated through the FB pin via resistors
R1 through R3 as shown in Figure 4.
R1
SC475A
the output capacitor via the inductor. DL will remain on
until FB falls to 0.75V, at which point a normal DH switching cycle begins, see Figure 5. This causes the output to
transition to the new voltage level quickly, typically 10~20
usec. Refer to the Smart Power Save Protection section
for a full description.
R2
V
OUT
+
G0
FB
0.75V
R3
D0
TON
Logic
Control
SC475A
Figure 4
The following table shows the equations for VOUT as a
function of control input G0:
VOUT EquationG0
0.75 • (1 + R1/R2)1
0.75 • (1 + R1/R2 + R1/R3) 0
Note that the RDSON of the internal D0 mosfet is in series with
R3, which adds typically 15 ohms in series.
A rapid downward change in VOUT also occurs for downward
changes less than 8%, provided the load is high enough
such that power-save is not active. In this case, after D0
opens and FB rises above the 0.75V trip point, DL will drive
high and stay high until FB drops to the trip point.
V
OUT
D0
R3
D0
FB
DL
DL
R1
FB
R2
RTN
(Smart Psave threshold)
810mV
750mV
(FB threshold)
Voltage Transitioning
The G0 pin allows VOUT to transition to both higher
and lower values. The two directions have differing
responses.
When doing a down transition, the sudden release of R3
will cause FB to go above the 0.75V threshold. Depending
on the level of VOUT change and the load, the IC responds
in different ways.
For the case where the down transition is less than
8%, and the load is light such that power-save is active,
the Smart Power Save detector will not activate. In this
At light load conditions when power-save is active, and
when the downward change is 8% or greater, the rapid
change of D0 is large enough to cause FB to rise up to the
Smart Power Save threshold (810mV). DL will then drive
case, with FB already above the 0.75V reference there
is no switching activity. DL and DH will remain off, and
the output voltage will slowly fall as the output capacitors
discharge into the load, see Figure 6.
high to turn on the low-side MOSFET and draw current from
Note that at light loads it can take many msec for the
output to fall to the new value. This should have no
adverse effect. Many loads such as graphics chipsets
can have a minimum load of several hundred mA, which
will naturally pull VOUT down to the next level.
V
OUT
R1
R3
D0
R2
D0
RTN
(Smart Psave threshold)
FB
COUT Discharge due to load
VOUT
FB
< 810mV
750mV
(FB threshold)
Initial
VOUT
Final
VOUT
When doing an up transition (from lower to higher VOUT),
the G0 change affects D0 and causes FB to drop below
the 0.75V trip point. This quickly trips the FB comparator
regardless of whether psave is active or not, generating a
DH on-time and a subsequent DL high time. At the end
of the minimum off-time (350nsec), if FB is still below
0.75V then another DH on-time is started. This sequence
continues until the FB pin exceeds 0.75V see Figure 7.
V
OUT
R1
R3
D0
R2
D0
RTN
FB
DH
DL
VOUT
FB
750mV
(FB threshold)
Final
VOUT
Initial
VOUT
Figure 6
For the case in Figure 6 the time needed to reach the fi nal
voltage is found from the following equation, where COUT
is in μF, and LOAD is in Amps:
If the VOUT change is signifi cant, there can be several
consecutive cycles of DH on-time followed by minimum DL
time. This can cause a rapid increase in inductor current:
Time (μsec) = COUT * (V
Note that the above equation only applies to the condition
where the VOUT downward change is less that the 8%
limit for Smart Psave, and also the load is light such that
Psave is active.
INITIAL
– V
FINAL
)/LOAD
typically it only takes a few switching cycles for the inductor current to rise up to the Current Limit. At some point
the FB voltage will rise up to the 0.75V reference and the
DH pulses will cease, but the inductor’s LI
then fl ow into the output cap. This can create a signifi cant
overshoot as shown in Figure 8.
The overshoot can be approximated by the following
equation, where ICL is the current limit, V
is the desired
FINAL
setpoint for the fi nal voltage, L is in μH and COUT is in μF.
VMAX = √( ICL
2
* L/COUT + V
FINAL
2
)
This overshoot can be eliminated by using a small RC
circuit to smooth the voltage seen at FB, see Figure 9.
The presence of Rs/Cs will prevent the rapid changes at
D0 from moving FB too quickly. The result is a gradual
change from VOUT
INITIAL
to VOUT
, to prevent the build-
FINAL
up of high inductor current and reducing overshoot. Note
that Cs can be connected to either VOUT or GND. VOUT is
preferred because this results in higher ripple seen at the
FB pin, which improves stability.
Figure 9
Note that Rs/Cs are part of the FB resistor divider and
therefore affect the output voltage sensing. To minimize
the effect of this, select Rs and Cs according to the
following guidelines:
The total of Rs + R3 should be chosen to give the correct
total resistance needed to adjust VOUT. Set Rs and R3 to
the same value.
Cs should be chosen to give a time constant equal to
approximately 18μsec for 325 kHz operation. Note that
Cs is charging through a resistive network composed
of Rs/R1/R2/R3. The effective resistance seen by Cs
is roughly equal to the parallel combination of Rs and
R3. For example, if the values for R1/R2/R3/Rs are
10k/50k/15k/15k, then the effective resistance would
be (10k + 15k) paralleled with 15k, which is 9.4k. To set
a time constant of 18μsec, CS should be approximately
18μsec/9.4k or 2000pF.
Note that the presence of Rs/Cs will affect the effective
resistance at the FB pin, and therefore modifi es the VOUT
setpoints. If Rs is used, the following table shows the
calculated values for VOUT.
VOUT EquationG0
0.75 • (1 + R1/R2)1
0.75 • (1 + R1/R2 + R1/(R3+Rs)) 0
Enable Input
The EN is used to disable or enable the SC475A. When
EN is low (grounded), the SC475A is off and in its lowestpower state. When EN is high the controller is enabled
and switching will begin.
PSAVE Operation
The SC475A provides automatic power save operation at
light loads. The internal Zero-Cross comparator looks for
inductor current (via the voltage across the lower MOSFET)
to fall to zero on eight consecutive switching cycles. Once
observed, the controller enters power save and turns off
the low-side MOSFET on each cycle when the current
crosses zero. To add hysteresis, the on-time is increased
by 25% in power save. The effi ciency improvement at light
loads more than offsets the disadvantage of slightly higher
output ripple. If the inductor current does not cross zero
on any switching cycle, the controller immediately exits
power save. Since the controller counts zero crossings,
the converter can sink current as long as the current does
not cross zero on eight consecutive cycles. This allows the
output voltage to recover quickly in response to negative
load steps, or to voltage transitions from a higher to a
lower voltage where the change exceeds 8%.
Smart Power Save Protection
In some applications, active loads on VOUT can leak
current from a higher voltage and thereby cause VOUT to
slowly rise and reach the OVP threshold, causing a hard
shutdown; the SC475A uses Smart Power Save to prevent
this. When FB exceeds 8% above nominal (810mV), the
IC exits power save (if already active) and DL drives high to
turn on the low-side MOSFET, which starts to draw current
from VOUT via the inductor. When FB drops to the 0.75V
trip point, a normal TON switching cycle begins. This cycles
energy from VOUT back to VBAT and prevents a hard OVP
shutdown, and also minimizes operating power by avoiding
continuous conduction-mode operation. If a light load is
present, the switching continues for 8 consecutive clock
cycles and then the IC will re-enter power save to reduce
operating power.
Current Limit Circuit
Current limiting can be accomplished in two ways. The
RDSON of the lower MOSFET can be used as a current
sensing element, or a sense resistor at the lower MOSFET
source can be used if greater accuracy is needed. RDSON
sensing is more effi cient and less expensive. In both
cases, the R
The R
MOSFET drain (for RDSON sensing) or the high side of the
current-sense resistor. R
source from the ILIM pin which turns on when the lowside MOSFET turns on, after the on-time DH pulse has
completed. If the voltage drop across the sense resistor
or low-side MOSFET exceeds the voltage across the R
resistor, current limit will activate. The high-side MOSFET
is held off until the voltage drop across the sense element
(resistor or MOSFET) falls below the voltage across the
R
ILIM
This current sensing scheme actually regulates the
inductor valley current, (see Figure 10). This means that
if the current limit is set to 10A, the peak current through
the inductor would be 10A plus the peak ripple current,
and the average current through the inductor would be
10A plus 1/2 the peak-to-peak ripple current.
The RDSON sensing circuit is shown in Figure 11 with
RILIM = R1 and RDSON of Q2.
VBAT
Q1
R1
Q2
+
C1
L
VOUT
D2
+
C3
BST
DH
LX
ILIM
VDD
DL
GND
SC475A
+5V
D1
C2
Q1
Q2
R4
VBAT
D2
+
C1
L1
Vout
+
C3
BST
DH
LX
ILIM
VDD
DL
GND
SC475A
+5V
D1
C2
R1
Figure 12
For resistor sensing, the current through the lower MOSFET
and the source sense resistor develops a voltage that
opposes the voltage developed across RILIM. When the
voltage developed across the RSENSE resistor reaches
voltage drop across RILIM, an over-current exists and the
high-side MOSFET will not be allowed to turn on.
The following over-current equation can be used for both
RDSON or resistive sensing. For RDSON sensing, the
MOSFET RDSON rating is used for the value of RSENSE.
R
ILx
OC
Aƫ10Valley
ILIM
R
SENSE
Power Good Output
The power good (PGD) output is an open-drain output
which requires a pull-up resistor. When the output
Figure 11
voltage as sensed at FB is -20% from the 0.75V reference
(600mV), PGD is pulled low. It is held low until the output
voltage returns above -20% of nominal. PGD is held low
The resistor sensing circuit is shown in Figure 12 with
RILIM = R1 and RSENSE = R4
during start-up and will not be allowed to transition high
until soft-start is completed when FB reaches 0.75V.
There is a 5μs delay built into the PGD circuit to prevent
false transitions.
PGD also transitions low if the FB pin exceeds +20% of
nominal, which is also the over-voltage shutdown point.
When G0 changes state, PGD is immediately latched
into its present state for 32 clock cycles while VOUT
and FB change to the new level, after which the latch is
disabled.
Output Over-Voltage Protection
In steady state operation, when FB exceeds 20% of
nominal (900mV), DL latches high and the low-side
MOSFET is turned on. DL stays high and the SMPS stays
off until the EN/PSV input is toggled or VCC is recycled.
There is a 5μs delay built into the OVP detector to prevent
false transitions. PGD is also held low after an OVP.
During G0 transitions, the OVP threshold is temporarily
increased to 50% above nominal (1.125V) for 32 clock
cycles. This is for cases where the output voltage is
slewing from a higher to a lower voltage: the change in G0
affects the D0 pin immediately, which in turn affects the
FB voltage immediately. The increase in OVP from 20%
to 50% is to prevent nuisance OVP tripping caused by the
immediate change at FB. It also protects against the case
of output overshoot for a lower to higher VOUT transition.
Note: since the temporary OVP is +50%, it is not possible to
have a VOUT change which causes an immediate FB change
of +50% or more. To transition VOUT under this condition, use
the RC smoothing circuit to slow the FB transition edges and
prevent +50% OVP.
Output Under-Voltage Protection
When FB falls 30% below nominal (525mV) for eight
consecutive clock cycles, the output is shut off; the DL/
DH drives are pulled low to tristate the MOSFETS, and the
SMPS stays off until the Enable input is toggled or VCC is
recycled.
Soft-Start
The soft-start is accomplished by ramping the FB
comparator’s internal reference from zero to 0.75V in
30mV increments. Each 30mV step typically lasts for
eight clock cycles.
During the soft-start period, the Zero Cross Detector is
active to monitor the voltage across the lower MOSFET
while DL is high. If the inductor current reaches zero, the
FB comparator’s internal ramp reference is immediately
overridden to match the voltage at the FB pin. This soon
causes the FB comparator to trip which forces DL to turn
off and the next DH on-time will begin. This prevents the
inductor current from going too negative which would
cause droop in the VOUT startup waveform. The next
30mV step on the internal reference ramp occurs from
the new point at the FB pin. Since any of the internal
30mV steps can be overridden by the FB waveform, the
startup time is dependent upon operating conditions.
This override feature will stop when the FB pin reaches
approximately 600mV, at which point the ramp resume
30mV steps up to 750mV.
At start-up, during the fi rst 32 switching cycles, the over-
current threshold is reduced by 50%, to reduce overshoot
caused by the fi rst set of switching pulses.
MOSFET Gate Drivers
The DH and DL drivers are optimized for moderate, highside, and larger low-side power MOSFETs. An adaptive
dead-time circuit monitors the DL output and prevents the
high-side MOSFET from turning on until DL is fully off, and
conversely, monitors the DH output and prevents the lowside MOSFET from turning on until DH is fully off.
Note: be sure there is low resistance and low inductance
between the DH and DL outputs to the gate of each MOSFET.
POR and UVLO
Under-voltage lockout circuitry (UVLO) inhibits switching
and tristates the DH/DL drivers until VCC rises above
4.4V. An internal power-on reset (POR) occurs when VCC
exceeds 4.4V, which resets the fault latch and the softstart counter, to prepare the PWM for switching. At this
time the SC475A will come out of UVLO and begin the
soft-start cycle.
Prior to designing a switch mode supply, the input
voltage, load current, switching frequency and inductor
ripple current must be specifi ed. For notebook systems
the maximum input voltage (VIN
highest AC adaptor voltage, and the minimum input voltage
(VIN
accounting for voltage drops due to connectors, fuses and
battery selector switches.
16
) is determined by the
MAX
) is determined by the lowest battery voltage after
MIN
www.semtech.com
POWER MANAGEMENT
NOT RECOMMENDED FOR NEW DESIGN
NOT RECOMMENDED
FOR NEW DESIGN
Applications Information (continued)
SC475A
In general, four parameters are needed to defi ne the
design:
1) Nominal output voltages (V
2) Static or DC output tolerance
3) Transient response
4) Maximum load current (I
There are two values of load current to consider: continuous
load current and peak load current. Continuous load
current is concerned with thermal stresses which drive
the selection of input capacitors, MOSFETs and diodes.
Peak load current determines instantaneous component
stresses and fi ltering requirements such as inductor
saturation, output capacitors and design of the current
limit circuit.
Design example:
VBAT = 10V min, 20V max
•
VOUT1 = 0.9V +/- 4%
•
VOUT2 = 1.15V +/-4%
•
Load = 20A maximum
•
Inductor Selection
Low inductor values result in smaller size but create
higher ripple current. Higher inductor values will reduce
the ripple current but are larger and more costly. Because
wire resistance varies widely for different inductors and
because magnetic core losses vary widely with operating
conditions, it is often diffi cult to choose which inductor
will optimize effi ciency. The general rule is that higher
inductor values have better effi ciency at light loads due
to lower core losses and lower peak currents, but at high
load the smaller inductors are better because of lower
resistance. The inductor selection is generally based on
the ripple current which is typically set between 20% to
50% of the maximum load current. Cost, size, output ripple
and effi ciency all play a part in the selection process.
The switching frequency is optimized for 325kHz.
The equation for on-time is:
TON (nsec) = 2560 • (VOUT/VBAT) + 35
During the DH on-time, voltage across the inductor is (VBAT
- VOUT). To determine the inductance, the ripple current
OUT
OUT
)
)
must be defi ned. Smaller ripple current will give smaller
output ripple and but will lead to larger inductors. The ripple
current will also set the boundary for PSAVE operation.
The switcher will typically enter PSAVE operation when the
load current decreases to 1/2 of the ripple current; (i.e.
if ripple current is 4A then PSAVE operation will typically
start for loads less than 2A. If ripple current is set at 40%
of maximum load current, then PSAVE will commence for
loads less than 20% of maximum current).
The equation for determining inductance is:
L = (VBAT - VOUT) • TON / IRIPPLE
Use the maximum value for VBAT, and for TON use the
value associated with maximum VBAT. For selecting the
inductor, we start with the highest VOUT setting and a
maximum ripple current of 5A.
TON = 182 nsec at 20VBAT, 1.15VOUT
L = (20 - 1.15) • 182 nsec / 5A = 0.69μH
We will select a slightly larger value of 0.7μH, which will
decrease the maximum IRIPPLE to 4.91A.
Note: the inductor must be rated for the maximum DC load
current plus 1/2 of the ripple current.
The minimum ripple current under is also checked .This
occurs when VBAT and VOUT are set to their minimum
values of 10V and 0.9V.
The output capacitors are chosen based on required ESR
and capacitance. The ESR requirement is driven by the
output ripple requirement and the DC tolerance. The
output voltage has a DC value that is equal to the valley
of the output ripple, plus 1/2 of the peak-to-peak ripple.
Change in the ripple voltage will lead to a change in DC
voltage at the output.
The design goal is +/-4% output regulation. The internal
0.75V reference tolerance is 1%, assuming 1% tolerance
for the FB resistor divider, this allows 2% tolerance due to
VOUT ripple. Since this 2% error comes from 1/2 of the
ripple voltage, the allowable ripple is 4%, or 46mV for a
1.15V output.
The maximum ripple current of 4.05A creates a ripple
voltage across the ESR. The maximum ESR value allowed
would create 46mV ripple:
ESRMAX = VRIPPLE/IRIPPLEMAX = 46mV / 4.91A
ESRMAX = 9.4 mΩ
The output capacitance is typically chosen based on
transient requirements. A worst-case load release, from
maximum load to no load at the exact moment when
inductor current is at the peak, defi nes the required
capacitance. If the load release is instantaneous (load
changes from maximum to zero in a very small time), the
output capacitor must absorb all the inductor’s stored
energy. This will cause a peak voltage on the capacitor
according to the equation:
COUT
With a peak voltage VPEAK of 1.230 (80mV rise above
1.15 upon load release), the required capacitance is:
The above requirements (570μF, 9.4mΩ) will be met using
two capacitors, 330μF 6mΩ.
If the load release is relatively slow, the output capacitance
can be reduced. At heavy loads during normal switching,
when the FB pin is above the 0.75V reference, the DL
output is high and the low-side mosfet is on. During this
time, the voltage across the inductor is approximately
-VOUT. This causes a downslope or falling di/dt in the
inductor. If the load di/dt is not much faster than the
di/dt in the inductor, then the inductor current can track
change in load current, and there will be relatively less
overshoot from a load release. The following can used to
calculate the needed capacitance for a given dILOAD/dt.
Unstable operation shows up in two related but distinctly
different ways: double-pulsing and fast-feedback loop
instability. double-pulsing occurs due to switching noise
seen at the FB input or because the ESR is too low, causing
insuffi cient voltage ramp in the FB signal. This causes
the error amplifi er to trigger prematurely after the 350ns
minimum off-time has expired. Double-pulsing will result
in higher ripple voltage at the output but in most cases is
harmless. In some cases, however, double-pulsing can
indicate the presence of loop instability, which is caused
by insuffi cient ESR.
One simple way to solve this problem is to add some
trace resistance between the VOUT/FB sense point and
the output capacitor in the high current output path. A
side effect of doing this is output voltage droop with load.
Another way to eliminate doubling-pulsing is to add a small
(e.g. 10pF) capacitor across the upper feedback resistor
divider network, as shown in Figure 13. This capacitance
should be left out until confi rmation that double-pulsing
exists. Adding this capacitance will add a zero in the
transfer function and should eliminate the problem. It is
best to leave a spot on the PCB in case it is needed.
R1
R3
FB
TON
0.75V
Logic
G0
Loop instability can cause oscillations at the output as a
response to line or load transients. These oscillations can
trip the over-voltage protection latch or cause the output
voltage to fall below the tolerance limit.
The best way for checking stability is to apply a zero-tofull load transient and observe the output voltage ripple
envelope for overshoot and ringing. Over one cycle of
ringing after the initial step is a sign that the ESR should
be increased.
SC475A ESR Requirements
The constant on-time control used in the SC475A
regulates the valley of the output ripple voltage. This
signal consists of a term generated by the output ESR of
the capacitor and a term based on the increase in voltage
across the capacitor due to charging and discharging
during the switching cycle. The minimum ESR is set to
generate the required ripple voltage for regulation. For
most applications the minimum ESR ripple voltage is
dominated by PCB layout and the properties of SP or
POSCAP type output capacitors. For applications using
ceramic output capacitors, the absolute minimum ESR
must be considered. If the ESR is low enough the ripple
voltage is dominated by the charging of the output
capacitor. This ripple voltage lags the on-time due to the
LC poles and can cause double pulsing if the phase delay
Control
Figure 13
D0
SC475A
C
R2
V
OUT
+
exceeds the off-time of the converter. To prevent double
pulsing, the ripple voltage present at the FB pin should be
10-15mV minimum over the on-time interval.
Dropout Performance
The VOUT adjust range for continuous-conduction
operation is limited by the fi xed 350nsec (typical) Minimum
Off-time One-shot. When working with low input voltages,
the duty-factor limit must be calculated using worst-case
values for on and off times.
The IC duty-factor limitation is given by:
DUTY = TONMIN/(TONMIN + TOFFMAX)
Be sure to include inductor resistance and MOSFET onstate voltage drops when performing worst-case dropout
duty-factor calculations.
SC475A System DC Accuracy (VOUT Controller)
Three factors affect VOUT accuracy: the trip point of the
FB error comparator, the switching frequency variation
with line and load, and the external resistor tolerance.
The error comparator offset is trimmed to trip when the
feedback pin is 0.75V, +/-1%.
The on-time pulse in the SC475A is calculated to give a
pseudo-fi xed frequency of 325kHz. Nevertheless, some
frequency variation with line and load is expected. This
variation changes the output ripple voltage. Because
constant on-time converters regulate to the valley of
the output ripple, ½ of the output ripple appears as a
DC regulation error. For example, If the output ripple is
50mV with VIN = 6 volts, then the measured DC output
will be 25mV above the comparator trip point. If the ripple
increases to 80mV with VIN = 25 volts, then the measured
DC output will be 40mV above the comparator trip. The
best way to minimize this effect is to minimize the output
ripple.
To compensate for valley regulation it is often desirable
to use passive droop. Take the feedback directly from the
output side of the inductor, placing a small amount of trace
resistance between the inductor and output capacitor.
This trace resistance should be optimized so that at full
load the output droops to near the lower regulation limit.
Passive droop minimizes the required output capacitance
because the voltage excursions due to load steps are
reduced.
The use of 1% feedback resistors contributes typically
1% error. If tighter DC accuracy is required use 0.1%
resistors.
The output inductor value may change with current. This
will change the output ripple and thus the DC output
voltage. The output ESR also affects the ripple and thus
the DC output voltage.
Switching Frequency Variations
The switching frequency will vary somewhat due to line
and load conditions. The line variations are a result
of a fi xed offset in the on-time one-shot, as well as
unavoidable delays in the external MOSFET switching. As
VBAT increases, these factors make the actual DH on-time
slightly longer than the idealized on-time. The net effect
is that frequency tends to fall slightly with increasing input
voltage.
The load variations are due to losses in the power train
from IR drop and switching losses. For a conventional
PWM constant-frequency topology, as load increases
the duty cycle also increases slightly to compensate for
IR and switching losses in the MOSFETs and inductor. A
constant on-time topology must also overcome the same
losses by increasing the duty cycle (more time is spent
drawing energy from VBAT as losses increase). Since the
on-time is constant for a given VOUT/VBAT combination,
the way to increase duty cycle is to gradually shorten
the off-time. The net effect is that switching frequency
increases slightly with increasing load.
Layout Guidelines
One or more ground planes are recommended to minimize
the effect of switching noise and copper losses and to
maximize heat removal. The analog ground reference,
RTN, should connect directly to the thermal pad, which
in turn connects to the ground plane through preferably
one large via. There should be a RTN plane or copper
are near the chip; all components that are referenced to
RTN should connect to this plane directly, not through the
ground plane, and located on the chip side of the PCB if
possible.
GND should be a separate plane which is not used for
routing analog traces. The VCC input provides power to
the internal analog circuits and the upper and lower gate
drivers.
The VCC supply decoupling capacitor should be tied
between VCC and GND with short traces. All power GND
connections should connect directly to this plane with
special attention given to avoiding indirect connections
between RTN and GND which will create ground loops.
As mentioned above, the RTN plane must be connected to
the GND plane at the chip near the RTN/GND pins.
The switcher power section should connect directly to the
ground plane(s) using multiple vias as required for current
handling (including the chip power ground connections).
Power components should be placed to minimize loops
and reduce losses. Make all the power connections on
one side of the PCB using wide copper fi lled areas if
possible. Do not use “minimum” land patterns for power
components. Minimize trace lengths and maximize trace
widths between the gate drivers and the gates of the
MOSFETs to reduce parasitic impedances (and MOSFET
switching losses); the low-side MOSFET is most critical.
Maintain a length to width ratio of <20:1 for gate drive
signals. Use multiple vias as required by current handling
requirement (and to reduce parasitic) if routed on more
than one layer.
For an accurate ILIM current sense connection, connect
the ILIM trace to the current sense element (MOSFET or
resistor) directly at the pin of the element, and route that
trace over to the ILIM resistor on another layer if needed.
The layout can be generally considered in two parts; the
analog control section referenced to RTN, and the switcher
power section referenced to GND.
Looking at the control section fi rst, locate all components
referenced to RTN on the schematic and place these
components near the chip and on the same side if
possible. Connect RTN using a wide trace. Very little
current fl ows in the RTN path and therefore large areas
of copper are not needed. Connect the RTN pin directly to
the thermal pad under the device as the only connection
between RTN and GND.
The chip supply decoupling capacitor (VCC/GND) should
be located near to the pins. Since the DL pin is directly
between VCC and GND, and the DL trace must be a wide,
direct trace, the VCC decoupling capacitor is best placed
on the opposite side of the PCB, routed with traces as short
as possible and using at least two vias when connecting
through the PCB.
There are two sensitive, feedback-related pins at the chip:
VOUT and FB. Proper routing is needed to keep noise
away from these signals. All components connected to
FB should be located directly at the chip, and the copper
area of the FB node minimized. The VOUT trace that
feeds into the VOUT pin, which also feeds the FB resistor
divider, must be kept far away from noise sources such
as switching nodes, inductors and gate drives. Route the
VOUT trace in a quiet layer if possible, from the output
capacitor back to the chip.
For the switcher power section, there are a few key
guidelines to follow:
1) There should be a very small input loop between
the input capacitors, MOSFETs, inductor, and output
capacitors. Locate the input decoupling capacitors
directly at the MOSFETs.
2) The phase node should be a large copper pour, but still
compact since this is the noisiest node.
3) The power GND connection between the input
capacitors, low-side MOSFET, and output capacitors
should be as small as is practical, with wide traces or
planes.
4) The impedance of the power GND connection
between the low-side MOSFET and the GND pin should
be minimized. This connection must carry the DL drive
current, which has high peaks at both rising and falling
edges. Use mulitple layers and multiple vias to minimize
impedance, and keep the distance as short as practical.
Finally, connecting the control and switcher power sections
should be accomplished as follows:
1) Route the VOUT/FB feedback traces in a “quiet” layer,
away from noise sources.
2) Route DL, DH and LX (low side FET gate drive, high side
FET gate drive and phase node) to the chip using wide
traces, with multiple vias if using more than one layer.
These connections are to be as short as possible for loop
minimization, with a length to width ratio less than 20:1
to minimize impedance. DL is the most critical gate drive,
with power GND as its return path. LX is the noisiest node
in the circuit, switching between VBAT and ground at high
frequencies, thus should be kept as short as practical. DH
has LX as its return path. DL, DH, LX, and BST are highnoise signals and should be kept well away from sensitive
signals, particularly FB and VOUT.
3) BST is also a noisy node and should be kept as short
as possible. The high-side DH driver is relies on the boost
capacitor to provide the DH drive current, so the boost
capacitor must be placed near the IC and connect to the
BST and LX pins using short, wide traces to minimize
impedance.
4) Connect the GND pin on the chip to the VCC decoupling
capacitor and then drop vias directly to the ground plane.
Locate the current limit resistor RLIM at the chip with a
kelvin connection to the drain of the lower MOSFET at the
phase node, and minimize the copper area of the ILIM
trace.