16
Date: 1/21/05 SP6120 Low Voltage, AnyFET
TM
, Synchronous, Buck Controller © Copyright 2005 Sipex Corporation
lum capacitors are considered. Tantalum capaci-
tors are known for catastrophic failure when ex-
posed to surge current, and input capacitors are
prone to such surge current when power supplies
are connected ‘live’ to low impedance power
sources. Certain tantalum capacitors, such as AVX
TPS series, are surge tested. For generic tantalum
capacitors, use 2:1 voltage derating to protect the
input capacitors from surge fall-out.
MOSFET Selection
The losses associated with MOSFETs can be
divided into conduction and switching losses.
Conduction losses are related to the on resistance
of MOSFETs, and increase with the load current.
Switching losses occur on each on/off transition
when the MOSFETs experience both high current
and voltage. Since the bottom MOSFET switches
current from/to a paralleled diode (either its own
body diode or a Schottky diode), the voltage across
the MOSFET is no more than 1V during switching
transition. As a result, its switching losses are
negligible. The switching losses are difficult to
quantify due to all the variables affecting turn on/
off time. However, the following equation pro-
vides an approximation on the switching losses
associated with the top MOSFET driven by SP6120.
SOUTINrssSH
FIVCP
(max)(max)(max)
12=
where
C
rss
= reverse transfer capacitance of the top
MOSFET
Switching losses need to be taken into account for
high switching frequency, since they are directly
proportional to switching frequency. The conduc-
tion losses associated with top and bottom
MOSFETs are determined by:
DIRP
OUTONDSCH
2
(max))((max)
=
)1(
2
(max ))((max)
DIRP
OUTONDSCL
=
where
P
CH(max)
= conduction losses of the high side
MOSFET
P
CL(max)
= conduction losses of the low side
MOSFET
R
DS(ON)
= drain to source on resistance.
The total power losses of the top MOSFET are the
sum of switching and conduction losses. For syn-
chronous buck converters of efficiency over 90%,
allow no more than 4% power losses for high or
low side MOSFETs. For input voltages of 3.3V
and 5V, conduction losses often dominate switch-
ing losses. Therefore, lowering the R
DS(ON)
of the
MOSFETs always improves efficiency even
though it gives rise to higher switching losses due
to increased C
rss
.
Top and bottom MOSFETs experience unequal
conduction losses if their on time is unequal. For
applications running at large or small duty cycle, it
makes sense to use different top and bottom
MOSFETs. Alternatively, parallel multiple
MOSFETs to conduct large duty factor.
R
DS(ON)
varies greatly with the gate driver voltage.
The MOSFET vendors often specify R
DS(ON)
on
multiple gate to source voltages (V
GS
), as well as
provide typical curve of R
DS(ON)
versus V
GS
. For
5V input, use the R
DS(ON)
specified at 4.5V V
GS
. At
the time of this publication, vendors, such as
Fairchild, Siliconix and International Rectifier,
have started to specify R
DS(ON)
at V
GS
less than 3V.
This has provided necessary data for designs in
which these MOSFETs are driven with 3.3V and
made it possible to use SP6120 in 3.3V only
applications.
Thermal calculation must be conducted to ensure
the MOSFET can handle the maximum load cur-
rent. The junction temperature of the MOSFET,
determined as follows, must stay below the maxi-
mum rating.
JA
MOSFET
AJ
R
P
TT
θ
(max)
(max)(max)
+=
where
T
A(max)
= maximum ambient temperature
PMOSFET(max) = maximum power dissipa-
tion of the MOSFET
R
Θ
JA
= junction to ambient thermal resistance.
R
Θ
JA
of the device depends greatly on the board
layout, as well as device package. Significant
APPLICATIONS INFORMATION: Continued
17
Date: 1/21/05 SP6120 Low Voltage, AnyFET
TM
, Synchronous, Buck Controller © Copyright 2005 Sipex Corporation
thermal improvement can be achieved in the maxi-
mum power dissipation through the proper design
of copper mounting pads on the circuit board. For
example, in a SO-8 package, placing two 0.04
square inches copper pad directly under the pack-
age, without occupying additional board space,
can increase the maximum power from approxi-
mately 1 to 1.2W. For DPAK package, enlarging
the tap mounting pad to 1 square inches reduces
the R
Θ
JA from 96°C/W to 40°C/W.
Schottky Diode Selection
When paralleled with the bottom MOSFET, an
optional Schottky diode can improve efficiency
and reduce noises. Without this Schottky diode,
the body diode of the bottom MOSFET conducts
the current during the non-overlap time when both
MOSFETs are turned off. Unfortunately, the body
diode has high forward voltage and reverse recov-
ery problem. The reverse recovery of the body
diode causes additional switching noises when the
diode turns off. The Schottky diode alleviates
these noises and additionally improves efficiency
thanks to its low forward voltage. The reverse
voltage across the diode is equal to input voltage,
and the diode must be able to handle the peak
current equal to the maximum load current.
The power dissipation of the Schottky diode is
determined by
P
DIODE
= 2V
F
I
OUT
T
NOL
F
S
where
T
NOL
= non-overlap time between GH and GL.
V
F
= forward voltage of the Schottky diode.
SP6120
®
®
C2
C1
R1
COMP
Figure 18. The RC network connected to the COMP pin
provides a pole and a zero to control loop.
Loop Compensation Design
The goal of loop compensation is to manipulate
loop frequency response such that its gain crosses
over 0db at a slope of –20db/dec. The SP6120
has a transconductance error amplifier and re-
quires the compensation network to be con-
nected between the COMP pin and ground, as
shown in Figure 18.
The first step of compensation design is to pick
the loop crossover frequency. High crossover
frequency is desirable for fast transient response,
but often jeopardize the system stability. Since
the SP6120 is equipped with 3% window com-
parator that takes over the control loop on tran-
sient, the crossover frequency can be selected
primarily to the satisfaction of system stability.
Crossover frequency should be higher than the
ESR zero but less than 1/5 of the switching
frequency. The ESR zero is contributed by the
ESR associated with the output capacitors and
can be determined by:
f
Z(ESR)
=
1
2
π
C
OUT
R
ESR
Crossover frequency of 20kHz is a sound first
try if low ESR tantalum capacitors or poscaps
are used at the output. The next step is to calcu-
late the complex conjugate poles contributed by
the LC output filter,
f
P(LC)
=
1
2
π√
LC
OUT
The open loop gain of the whole system can be
divided into the gain of the error amplifier,
PWM modulator, buck converter, and feedback
resistor divider. In order to crossover at the
selected frequency fco, the gain of the error
amplifier has to compensate for the attenuation
caused by the rest of the loop at this frequency.
In the RC network shown in Figure 18, the
product of R1 and the error amplifier
transconductance determines this gain. There-
fore, R1 can be determined from the following
equation that takes into account the typical error
amplifier transconductance, reference voltage
and PWM ramp built into the SP6120.
R
1
=
1300V
OUT
f
CO
f
Z(ESR)
V
IN
f
P(LC)
2
APPLICATIONS INFORMATION: Continued
18
Date: 1/21/05 SP6120 Low Voltage, AnyFET
TM
, Synchronous, Buck Controller © Copyright 2005 Sipex Corporation
Current Sense
The SP6120 allows sensing current using the
inductor, PCB trace or current-sense resistor.
Inductor-sense utilizes the voltage drop across
the ESR of the inductor, while PCB trace and
current-sense resistor introduce additional re-
sistance in series with the inductor. The resis-
tance of the sense element determines the
overcurrent protection threshold as follows,
I
LIM
=
43mV
R
SEN
R
SEN
= current-sense resistance which can be
implemented as ESR of the inductor, trace or
discrete resistor.
The maximum power dissipation on the current-
sense element is:
SENOUTSEN
RIP
2
(max)
=
For the inductor-sense scheme shown in the
application circuit, R
S
and C
S
are used to repli-
cate the signal across the ESR of the inductor. R
S
and C
S
can be looked at as a low pass filter
whose output represents the DC differential
voltage between the switch node and the output.
At steady state, this voltage happens to be the
output current times the ESR of the inductor. In
addition, if the following relationship is satisfied,
L
= R
S
C
S
ESR
the output of the RsCs filter represents the exact
voltage across the ESR, including the ripple.
Since the SP6120’s hiccup overcurrent protec-
tion scheme is intended to safeguard sustained
overload conditions, the DC portion of the cur-
rent signal is more of interest. Therefore, design-
ing the R
S
C
S
time constant higher than L/ESR
provides reliable current sense against any pre-
mature triggering due to noise or any transient
conditions. Pick Rs between 10k and 100k, and
Cs can be determined by:
C
S
= 2
L1
ESR R
S
Here the time constant of R
S
C
S
is twice the value
of L/ESR.
In Figure 18, R1 and C1 provides a zero f
Z1
which needs to be placed at or below f
P(LC)
. If f
Z1
is made equal to f
P(LC)
for convenience, the
value of C1 can be calculated as
C
1
=
1
2
π
f
P(LC)
R
1
The optional C2 generates a pole f
P1
with R1 to
cut down high frequency noise for reliable op-
eration. This pole should be placed one decade
higher than the crossover frequency to avoid
erosion of phase margin. Therefore, the value of
the C2 can be derived from
C
2
=
1
20
π
f
CO
R
1
Figure 19 illustrates the overall loop frequency
response and frequency of each pole and zero.
To fine-tune the compensation, it is necessary to
physically measure the frequency response us-
ing a network analyzer.
Gain
-20db/dec
-40db/dec
-20db/dec
-20db/dec
-20db/dec
Error Amplifier
Loop
f
f
Z1
f
P(LC)
f
Z(ESR)
f
CO
f
P1
Figure 19. Frequency response of a stable system and its
error amplifier.
APPLICATIONS INFORMATION: Continued

SP6120EY-L/TR

Mfr. #:
Manufacturer:
MaxLinear
Description:
Switching Controllers Low Voltage, AnyFET Synchronous, Cntrllr
Lifecycle:
New from this manufacturer.
Delivery:
DHL FedEx Ups TNT EMS
Payment:
T/T Paypal Visa MoneyGram Western Union

Products related to this Datasheet