V25N2 - JULY

July 2015
Volume 25 Number 2
extend remote sensor
Op Amp Combines Femtoamp Bias
Current with 4GHz Gain Bandwidth
Product, Shines New Light on
Photonics Applications
battery life with thermal
Glen Brisebois
I N
T H I S
I S S U E
4-phase supply supports
120A in tiny footprint 10
one driver is all you
need for automotive LED
headlight clusters 17
energy harvesting 24
simplify small solar
systems with hysteretic
controller 27
powering a Dust® mote
from a piezo 29
Einstein published his seminal paper on the photoelectric effect
110 years ago, essentially inventing the discipline of photonics. One
would think that over so many years the science and engineering
surrounding photonics must have fully matured. But not so. Optical
sensors—photodiodes, avalanche photodiodes, and photomultiplier
tubes—continue to achieve astoundingly high dynamic ranges,
enabling electronics to peer ever more
deeply into the photonic world.
Photosensors typically convert photons to electron current
and are followed by a transimpedance function to transform
the current into a voltage. The transimpedance function
may be either a simple resistor or, for higher bandwidth,
the summing node of an op amp, in which case it is called
a transimpedance amplifier (TIA). The traditional enemies
of the TIA are voltage noise, current noise, input capacitance, bias current and finite bandwidth. Enter the new
LTC®6268-10 with 4.25nV/√Hz voltage noise, 0.005pA/√Hz
current noise, a very low 0.45pF of input capacitance,
3fA of bias current and 4GHz of gain-bandwidth.
(continued on page 4)
The LTC6268’s performance meets the demands of the latest photonics applications.
w w w. li n ea r.com
Linear in the News
In this issue...
COVER STORY
Op Amp Combines Femtoamp Bias Current
with 4GHz Gain Bandwidth Product, Shines
New Light on Photonics Applications
Glen Brisebois
THE INDUSTRIAL INTERNET OF THINGS
1
DESIGN FEATURES
4-Phase Power Supply Delivers 120A in Tiny Footprint,
Features Ultralow DCR Sensing for High Efficiency
Yingyi Yan, Haoran Wu and Jian Li
10
3mm × 3mm Monolithic DC/DC Boost/Inverting
Converters with 65V Power Switches
Joshua Moore
14
One LED Driver Is All You Need for
Automotive LED Headlight Clusters
Keith Szolusha and Kyle Lawrence
17
DESIGN IDEAS
What’s New with LTspice IV?
Gabino Alonso
22
Extend Remote Sensor Battery Life
with Thermal Energy Harvesting
Dave Salerno
24
Simplify Small Solar Systems with Hysteretic Controller
Mitchell Lee
27
Powering a Dust Mote from a Piezoelectric Transducer
Jim Drew
29
new product briefs
31
back page circuits
32
Everywhere you turn, there is discussion around the emerging Internet of
Things (IoT)—the concept that the Internet of the future will create a vast network of interconnected physical objects or “things” with embedded sensors,
electronics and software. IoT devices are able to exchange data with other connected devices and human operators. According to industry analyst, Stifel, the
IoT market today is already sizable, with an estimated 19.7 billion IoT devices
deployed, and expected to grow to 95.5 billion devices deployed by 2025.
Much of the recent discussion about IoT revolves around “wearable technology”—Google glasses and Apple watches—but these applications are only part of a much larger IoT picture. There remains
significant opportunity in the Industrial Internet of Things; analysts
expect industrial to be one of the fastest growing IoT segments.
One major area of promise is to leverage real-time data gathered via wireless sensor networks (WSNs) to improve efficiency and streamline processes.
Sensors can be placed in a broad range of environments, including buildings,
city streets, bridges and tunnels, industrial plants, moving vehicles, or in remote
locations such as along pipelines and weather stations. These applications
require that WSNs draw extremely low power and yield wired-like reliability, often over a range of network architectures, sizes and data rates. Wireless
mesh networks are increasingly popular because they can easily cover large
areas using low power radios that reliably relay data from node to node.
Linear has developed expertise in this area via its Dust Networks® WSN
product line. Linear’s wireless sensor network products are suited for
a wide range of Industrial IoT applications, including:
•Flow and process monitoring in industrial environments
•Data center energy management
•Fence line security
•Rail car preventive maintenance
•Smart street parking systems
SOFTWARE DEVELOPMENT KIT ACCELERATES INDUSTRIAL IoT
APPLICATION DEVELOPMENT
Linear Technology now offers a software development kit for the Industrial
IoT. Linear Technology’s SmartMesh IP™ wireless sensor networking products
now provide the ability to program applications directly on the embedded ARM
Cortex-M3, running Micrium’s µC/OS-II real-time operating system. Application
2 | July 2015 : LT Journal of Analog Innovation
Linear in the news
synchronous buck converter that delivers
up to 50m A of continuous output current
and operates from an input voltage range
of 2.7V to 20V, ideal for a wide range of
energy harvesting and IoT battery-powered applications including “keep-alive”
sensor and industrial control power.
AWARDS
The LTC2000 16-bit, 2Gsps DAC
received the 2014 Product Award
from 21ic.com in China.
Linear’s Dust Networks wireless
sensor network products fuel
applications in the Industrial Internet
of Things.
CONFERENCES & EVENTS
IEEE Nuclear and Space Radiation Effects
Conference, MarrIott Copley Place, Boston,
development time is accelerated with
a library of reference code and source
code examples. Based on 6LoWPAN,
SmartMesh IP mesh networking products
include a pre-compiled networking stack
that delivers >99.999% network reliability at ultralow power. This is particularly
important for Industrial IoT applications, where wireless sensor networks
may be deployed in harsh and remote
environments, with little to no chance
for maintenance. The On-Chip Software
Development Kit (SDK) provided with
the LTC5800-IPM (system-on-chip) and
LTP™5901/2-IPM (PCB modules) is architected to ensure that developers can stably
run both the pre-compiled SmartMesh IP
networking stack and their applications
simultaneously. For more information,
see www.linear.com/solutions/5457.
ENERGY HARVESTING AND IoT
The proliferation of wireless sensors
supporting the Internet of Things has
increased the demand for small, compact
and efficient power converters tailored to
untethered lower power devices. State-ofthe-art and off-the-shelf energy harvesting (EH) technologies—for example, in
vibration energy harvesting and indoor or
wearable photovoltaic cells—yield power
levels on the order of milliwatts under
typical operating conditions. The operation of harvesting elements over several
years makes them comparable to long-life
primary batteries, both in terms of energy
provision and the cost per energy unit.
Systems incorporating EH are typically
capable of recharging after depletion,
something that systems powered by primary batteries cannot do. Most implementations use an ambient energy source as
the primary power source, supplemented
by a primary battery in case the ambient
energy source goes away or is disrupted.
Massachusetts, July 13–17, Booths 8 & 9—
Linear Technology’s LTC3331 is a complete
regulating EH solution that delivers up
to 50m A of continuous output current
to extend battery life when harvestable
energy is available. It requires no supply
current from the battery when providing
regulated power to the load from harvested energy and only 950n A operating
when powered from the battery under
no-load conditions. The LTC3331 integrates a high voltage EH power supply,
plus a synchronous buck-boost DC/DC
converter powered from a rechargeable
primary cell battery to create a single
non-interruptible output for energy
harvesting applications such as those in
wireless sensor networks. Another device,
the LTC3388-1/-3, is a 20V input-capable
Expo, Suburban Collection Showplace, Novi,
Linear is showcasing its products
for space and harsh environments.
More info at www.nsrec.com
2nd Dust Networks Consortium, Tokyo Conference
Center, Tokyo, Japan, July 21—Software
and device vendors, integrators and
monitoring service providers are
invited to learn about wireless sensor
network system capabilities. More
info at www.dust-consortium.jp/
The Battery Show/Electric & Hybrid Vehicle Tech
Michigan, September 15–17—Presenting Linear’s
battery management system products.
More info at www.thebatteryshow.com/
IoT World Congress, Gran Via Venue, Barcelona,
Spain, September 16–18—Linear will high-
light its Dust Networks wireless sensor network products. More info at
www.IoTsworldcongress.com/en/home
Sensors & Instrumentation for Test, Measurement &
Control, The National Exhibition Centre, Birmingham,
UK, September 30 to October 1—Linear will
showcase products and solutions related
to wireless sensor networks. More info at
www.sensorsandinstrumentation.co.uk/ n
July 2015 : LT Journal of Analog Innovation | 3
The calculated CV + I noise for the LTC6268-10 at 1MHz
is 0.052pA/√Hz, compared to 0.156pA/√Hz for the
OPA657; a factor of three better for the LTC6268-10.
UNDERSTANDING VOLTAGE
NOISE AND CURRENT NOISE
CONTRIBUTIONS IN TIAs
Output noise in TIAs is a result of combined input voltage noise and input current noise. This combined effect is often
specified as a current noise referred to the
input—essentially the output voltage noise
divided by the gain in ohms—but it actually arises from both input noise sources.
In fact, the dominant cause of output noise
is usually input voltage noise (Figure 1).
By virtue of feedback, the minus input is
fixed at virtual ground so the current noise
in passes directly through RF and contributes to total current noise with a factor of
1. Also by virtue of feedback, the voltage
noise en is placed in parallel with the input
capacitance CIN and induces a current
noise of en /Z(CIN). The impedance of a
capacitor is 1/2πfC, so the effective current noise due to input voltage noise and
capacitance is 2πfCIN en. So the total op
amp noise (ignoring RF thermal noise) is
INOISE =
(2
2
fCINen ) + (in )
2
This is sometimes referred to as CV + I
noise and makes an excellent figure of
merit for an op amp, because it incorporates only op amp characteristics,
neglecting external aspects of the circuit such as photosensor capacitance
and RF thermal noise. It is essentially
the best the op amp can do.
Figure 1. The op amp with its noise sources and input capacitance. Total op amp noise
(ignoring RF thermal noise) is INOISE = in + 2πfCINen (added rms-wise).
IPD + INOISE
IPD
Figure 2. CV + I current noise versus frequency for
the LTC6268-10 and OPA657. The LTC6268-10 is
considerably quieter.
1.8
1.6
1.4
1.2
1
OPA657
0.6
LTC6268-10
0.4
0.2
0
1
2
3 4 5 6 7
FREQUENCY (MHz)
8
9
–
VOUT = RF • (IPD + INOISE)
+
The CV + I noise is a useful figure of merit
for comparing op amps, but it does have
a dependency on frequency. An insightful
comparison can be made by initially comparing them at a specific frequency and
then observing the differences in the plots
of CV + I noise versus frequency that inevitably arise. For example, let’s compare
the LTC6268-10 and competitive OPA657
by starting with a calculation at 1MHz.
0.8
RF
in
SAMPLE CALCULATION AND
COMPARISON BETWEEN LTC6268-10
AND COMPETITIVE OPA657
0
4 | July 2015 : LT Journal of Analog Innovation
en
IPD = PHOTODIODE CURRENT
en = OP AMP VOLTAGE NOISE
in = OP AMP CURRENT NOISE
CIN
CIN = OP AMP INPUT CAPACITANCE
INOISE = EFFECTIVE COMBINED CURRENT NOISE
RF = FEEDBACK RESISTANCE, OR TIA GAIN
“CV + I” CURRENT NOISE (pA/√Hz)
(LTC6268, continued from page 1)
10
The LTC6268-10 data sheet gives plots
of current noise versus frequency
showing 0.05pA/√Hz at 1MHz , and of
voltage noise versus frequency showing 4nV/√Hz at 1MHz. Using the input
capacitance of 0.55pF (0.45pF for CCM,
plus 0.1pF for CDM), the total CV
noise at 1MHz can be calculated as
CV NOISE = 2 •1MHz •0.55pF • 4nV
= 0.014pA
Hz
Hz
Summing this rms-wise with the native I
noise of 0.05pA/√Hz, we get 0.052pA/√Hz
of total CV + I noise at 1MHz.
The same calculation for the competitive OPA657 can also be performed. It
specifies 4.8nV/√Hz voltage noise, 5.2pF
input capacitance (4.5pF for CCM plus
0.7pF for CDM), and 1.3fA/√Hz current
noise. Calculating total CV + I noise gives
0.156pA/√Hz at 1MHz for the OPA657,
about three times worse than LTC6268-10.
Figure 2 shows a plot of CV + I noise for
LTC6268-10 and OPA657 versus frequency.
The reason the LTC6268-10 outperforms
the OPA657 is its lower voltage noise and
its much lower input capacitance. And
design features
A powerful method to reduce feedback capacitance is to shield the E field paths that
give rise to the capacitance. In this particular case, the method is to place a ground trace
between the resistor pads. Such a ground trace shields the output field from getting to
the summing node end of the resistor, effectively shunting the field to ground instead.
because the LTC6268-10 has lower voltage noise, it continues to outperform
the OPA657 as the sensor capacitance is
added and increased. Furthermore, the
LTC6268-10 features a rail-to-rail output
and can operate on a single 5V supply,
burning half the power of OPA657.
RISE TIME = 88ns
PARASITIC
FEEDBACK C
+2.5
402k
K
CASE
PD
A
+2.5
–
IPD
GAIN BANDWIDTH, AND ACHIEVING
HIGH BANDWIDTH AT HIGH
IMPEDANCE
LASER DRIVE
(2mA/DIV)
OUTPUT
(500MV/DIV)
VOUT
LTC6268-10
+
–2.5
PD: OSI FCI-125G-006
Another advantage of the LTC6268-10 is
its serious 4GHz gain bandwidth product.
In fact, you’ll find that the LTC6268-10 is
able to find and use tiny parasitic capacitances that other op amps miss. Normally,
high value resistors begin to reduce their
net impedance at high frequency due to
their end-to-end capacitance. The key
to exploiting the 4GHz gain bandwidth
of the LTC6268-10 with higher gain TIAs
200ns/DIV
Figure 4. Time domain response of 402kΩ TIA
without extra effort to reduce feedback capacitance.
Rise time Is 88ns and BW is 4MHz.
Figure 3. LTC6268-10 and low capacitance
photodiode in a 402kΩ TIA
is to minimize the feedback capacitance
around the main feedback resistor. Though
minimized, the LTC6268-10 can use the
tiny residue feedback capacitance to
compensate the feedback loop, extending resistor bandwidth to several MHz.
Following is a design example at 402k.
BOTTOM
RESISTIVE
ELEMENT
ENDCAP
IPD
Figure 5. A normal layout (a) and a fieldshunting layout (b). Circuit board in (c)
shows actual layout with extra shunting
at R9, less at R12. Simply adding a
ground trace under the feedback
resistor does much to shunt field away
from the feedback side, dumping it
to ground. Note that the dielectric
constant of FR4 and ceramic is
typically 5, so most of the capacitance
is in the solids and not through the
air. Such field shunting techniques
reduced feedback capacitance from
approximately 100fF in Figure 4 to
11.6fF in Figure 6. Note also that the
feedback trace is exposed in upper (c)
but entirely shielded in lower (c).
A
G
K
–
FR4
LTC6268-10
+
+2.5
MUCH
LIGHTER
DOSE
(a)
RESISTIVE
ELEMENT
ENDCAP
IPD
A
K
+2.5
VOUT
E FIELD ⇒ C
CERAMIC R SUBSTRATE E
G
TOP
(c)
CERAMIC R SUBSTRATE E
–
FR4
LTC6268-10
+
HEAVY DOSE
OF SHUNTING
CF ≈ 11.6fF
EXTRA GND
TRACE UNDER
RESISTOR
VOUT
TAKE E FIELD TO GND,
MUCH LOWER C
(b)
July 2015 : LT Journal of Analog Innovation | 5
Bandwidth and rise time went from 4MHz (88ns) to 34MHz (10.3ns), a factor
of 8. The ground trace used for LTC6268-10 was much wider than that used
in the case of the LTC6268, extending under the entire resistor dielectric.
RISE TIME = 10.3ns
LASER DRIVE
(2mA/DIV)
OUTPUT
(500MV/DIV)
20ns/DIV
Figure 6. LTC6268-10 in a 402kΩ TIA with extra
layout effort to reduce feedback capacitance
achieves 10.3ns total system rise time, or 34MHz
total system bandwidth. This is an 8x increase in
bandwidth, due to a well placed bit of ground trace.
Good layout practices are essential to
achieving best results from a TIA circuit. The following two examples show
drastically different results from an
LTC6268-10 in a 402k TIA (Figure 3). The
first example is with an 0805 resistor in
a basic circuit layout. In a simple layout, without expending a lot of effort to
Figure 7. A photograph
and x-ray of a Hamamatsu
photomultiplier tube. The
electronics components
visible at right are the
encapsulated high voltage
supply. (Do not x-ray your
PMT unless it is already
unusable.)
6 | July 2015 : LT Journal of Analog Innovation
reduce feedback capacitance, the rise time
achieved is about 88ns (Figure 4), implying a bandwidth of 4MHz (BW = 0.35/tR).
In this case, the bandwidth of the TIA is
limited not by the GBW of the LTC6268-10,
but rather by the fact that the feedback
capacitance is reducing the actual feedback impedance (the TIA gain itself) of the
TIA. Basically, it’s a resistor bandwidth
limitation. The impedance of the 402k is
reduced by its own parasitic capacitance at
high frequency. From the 4MHz bandwidth
and the 402k low frequency gain, we can
estimate the total feedback capacitance as
CF =
1
= 0.1pF
2 • 4MHz • 402k
That’s fairly low, but it can be reduced
further, maybe much further.
With some extra layout techniques to
reduce feedback capacitance, the bandwidth can be increased. Note that we
are increasing the effective “bandwidth”
of the 402k resistance. A very powerful
method to reduce feedback capacitance is
to shield the E field paths that give rise to
the capacitance. In this case, the method
is to place a ground trace between the
resistor pads. Such a ground trace shields
the output field from getting to the summing node end of the resistor, effectively
shunting the field to ground instead.
The trace increases the output load
capacitance very slightly. See Figure 5a
and 5b for a pictorial representation,
and Figure 5c for an example layout.
Figure 6 shows the dramatic increase in
bandwidth simply by careful attention
to low capacitance methods around the
feedback resistance. Bandwidth and rise
time went from 4MHz (88ns) to 34MHz
(10.3ns), a factor of 8. The ground shield
trace used for the LTC6268-10 was much
wider than that used in the high speed case
of the LTC6268 (see LTC6268 data sheet),
extending under the entire resistor dielectric. Assuming all the bandwidth limit is
due to feedback capacitance (which isn’t
fair), we can calculate an upper limit of
design features
Photomultiplier tubes yield photonics gains above one million, meriting their
considerably high cost. Given the high inherent gain, the TIA gain can be reduced,
and bandwidth extended to the point that single photon events can be isolated.
When using the LTC6268-10 at low gain, however, care must be taken to ensure
its gain stability requirement of 10 is met, or there is risk of oscillation.
CF =
1
= 11.6fF
2 •34MHz • 402k
A PHOTOMULTIPLIER TUBE (PMT)
AT LOWER IMPEDANCE
Photomultiplier tubes (photograph and
x-ray shown in Figure 7) yield photonics
gains above one million, meriting their
considerably high cost. Given the high
inherent gain, the TIA gain can be reduced,
and bandwidth extended to the point that
single photon events can be isolated. One
convenient feature of a PMT is self-excitation, drawing energy either from local
cosmic radiation or its own thermionic
electron emission when the plate voltage
is high, producing a random Dirac-deltalike ping of electrons on the output plate.
When using the LTC6268-10 at low gain,
however, care must be taken to ensure
Figure 9. The transmission line was short
compared to a 300MHz assessment,
but long enough to be a problem when
compared to real bandwidth available.
Figure 8. First
attempt at
connecting
LTC6268-10 to PMT
output plate. Note
the 3/4-inch or so
transmission line
created by the PMT
plate pin. That’s far
below 1/4-lambda at
300MHz. What could
possibly go wrong?
See Figure 9.
its gain stability requirement of 10 is
met, or there is risk of oscillation. The
Hamamatsu PMT did not have a specified
output plate capacitance, but the HP4192
impedance analyzer measured it to be
10pF at its maximum test frequency of
Figure 10. Much tighter design on dedicated board.
LTC6268-10 is now much closer to the PMT body
and therefore the PMT output plate capacitance.
Transmission line still exists, but it is hanging midair
and is not “in the way.”
MHz. Given that fact, a feedback capacitance of 1pF should have been adequate
to ensure an apparent noise gain of 11.
13
However, the pins on the PMT were about
3/4 inch long (Figure 8), and with the
LTC6268-10 connected to it in a gain of
Figure 11. Reducing the transmission line
length is key to achieveing good results.
Output pulse half-width is 2.2ns. Exact −3dB
bandwidth is not as relevant as a clean timedomain response.
PMT OUTPUT
PLATE PIN
VOUT
50mV/DIV
VOUT
50mV/DIV
2ns/DIV
2ns/DIV
July 2015 : LT Journal of Analog Innovation | 7
The LTC6268 achieves bias currents two orders of
magnitude lower than any previous Linear Technology
amplifier, which requires accurately measuring femtoamps—
while measuring picoamps is challenging enough.
0.3pF AIR CAPACITANCE
300
AIR WIRED
VCM
301k
5V
18V
+
–
100k
DUT
LTC6268
100k
–
+
−18V
10k
VOUT
LT1055
4.7µF
5V
0.1µF
IBIAS =
(VOUT − VCM)
1TΩ
INPUT BIAS CURRENT (fA)
DUT
(+IN)
8.0
200
4.0
100
+IN
2.0
0
0.0
–IN
–100
–2.0
–4.0
–6.0
–200
–8.0
–300
0.0
10k
6.0
INPUT BIAS CURRENT (fA)
RSENSE
1TΩ
10.0
VS = 5V
RSENSE: OHMITE MINI-MOX MOX1125-23E 10%
1.0
2.0
3.0
4.0
COMMON MODE VOLTAGE (V)
–10.0
5.0
Figure 12. Circuit for measuring femtoamp bias current of LTC6268 (the unity gain stable version of LTC6268-10) and measured results, at various common modes.
a sustained oscillation of 1.05GHz
became apparent alongside the expected
response to a dark-current ping (Figure 9).
Trying a variety of feedback capacitors between 0.2pF and 1pF around the
LTC6268-10 did not help. The conclusion
was that the short transmission line was
changing the appearance of the 10pF plate
at high frequency, and was therefore not
satisfying the gain of 10 requirement.
1.82k ,
With the LTC6268-10 positioned closer
to the PMT body on a new board
(Figure 10), the oscillation was quenched
and the much improved response of
Figure 11 was achieved. Component
feedback capacitance installed was
0.8pF (Murata GJM1555C1HR80).
Another change on the board was that
the feedback resistor was brought to
the topside, eliminating two vias.
MEASURING FEMTOAMPS
The LTC6268 achieves bias currents about
two orders of magnitude lower than
any previous Linear Technology amplifier, which requires accurately measuring
femtoamps—while measuring picoamps
is challenging enough. In production testing, speed is of the essence, so capacitive
switching techniques are employed. In our
tests made on the bench, where speed is
not an issue, a sense resistor was preferred.
Assuming a 1mV op amp offset allowance (actually 0.7mV max), and a
desired resolution of 1fA, the required
sense resistor comes to 1mV/1fA = 1TΩ.
Fortunately, Ohmite makes a 1T resistor, in the long blue MOX1125 package.
In order to measure input bias current
at various input common mode voltrage
levels to the DUT (device under test),
the circuit of Figure 12 was employed.
Circuit board effects were removed by
removing the circuit board. That is,
8 | July 2015 : LT Journal of Analog Innovation
removing the board under the LTC6268
noninverting input and whisker connecting it through air to the 1TΩ resistor. This leaves just the op amp pin, the
resistor and their package materials in
place, hanging midair, as you can see in
Figures 13 (topside) and 14 (bottom side).
Figure 15 shows the time domain response,
settling well in 2.2 seconds. The overshoot
isn’t actually overshoot in the conventional
sense, but rather the charge necessary to
move the total input C, effectively looking
like a short term bias current. The voltage delta of the overshoot is about 190mV,
and extends about 1.25 seconds in width.
The total charge can be estimated by
calculating the area of the triangle created
by the voltage-overshoot in Figure 15:
1
• 190mV • 1.25s
TOTAL CHARGE = 2
= 0.12pC
1T
With Q = CV, and a 200mV step, the
total input C can be calculated as
design features
The LTC6268-10 features extremely low 4.25nV/√Hz voltage
noise, 0.005pA/√Hz current noise, a very low 0.43pF of input
capacitance, 3fA of bias current and 4GHz of gain-bandwidth.
FEEDBACK C FIN
1TΩ SENSE RESISTOR
AIR WIRE TO 1TΩ
SENSE RESISTOR
INTEGRATOR
DUT INPUT PIN
AIR WIRE TO 1TΩ
SENSE RESISTOR
DUT INPUT PIN
Figure 13. Actual board implementation of the femtoamp
measurement board. Note the placement respecting the long blue
resistor. Feedback capacitance to DUT input pin is through-air only.
VIN
200mV/DIV
Figure 14. The bottom side of the
board, showing the DUT input pin
hanging midair.
Q/V = 0.6pF. A rough allocation would
be 0.45pF for the LTC6268 input CDM
and another 0.15pF for the whisker and
resistor lead. Output noise was measured at just under 1mVP–P, consistent
with the objective of resolving 1fA.
VOUT
200mV/DIV
CONCLUSION
The LTC6268-10 significantly reduces
the traditional enemies of TIAs: voltage noise, current noise, input capacitance and bias current. It features
extremely low 4.25nV/√Hz voltage noise,
0.005pA/√Hz current noise, a very low
0.43pF of input capacitance, 3fA of bias
current and 4GHz of gain-bandwidth. n
1s/DIV
Figure 15. Time domain response. Settles in
2.2 seconds with a 200mV change in common
mode voltage. Overshoot is real, as the teraohm
resistor moves the voltage on the 0.6pF total input
capacitance.
July 2015 : LT Journal of Analog Innovation | 9
4-Phase Power Supply Delivers 120A in Tiny Footprint,
Features Ultralow DCR Sensing for High Efficiency
Yingyi Yan, Haoran Wu and Jian Li
The LTC3875 is a feature-rich dual-output synchronous
buck controller that meets the power density demands
of modern high speed, high capacity data processing
systems, telecom systems, industrial equipment and DC
power distribution systems. The LTC3875 delivers high
efficiency with reliable current mode control, ultralow DCR
sensing and strong integrated drivers in a 6mm × 6mm
40-pin QFN. Multiple LTC3875s can be paralleled to provide
higher current, or it can be combined with the LTC3874
to deliver the same performance with a smaller footprint.
The LTC3874 is a small footprint
(4mm × 5mm QFN), dual PolyPhase® current mode synchronous step-down slave
controller (phase extender). It is suitable
for high current, multiphase applications
when paired with a companion master
controller, such as LTC3875. The LTC3874
can use sub-milliohm DC resistance
power inductors to optimize efficiency.
Immediate response to system faults
guarantees reliability of the total solution.
1V V OUT, 120A CONVERTER WITH
PARALLEL LTC3875s
The LTC3875 can be easily configured as
dual-phase, single-output operation for
high current outputs. This design can
be expanded with more converters and
phases in parallel for even higher current.
Figure 1 shows a 4.5V~14V input, singleoutput application schematic using two
LTC3875s. The LTC3875s’ four channels run
with 90° phase shift, reducing input RMS
current ripple and required capacitor size.
Figure 1. A single-output,
4-phase (1.0V/120A) converter
Each phase supports 30A of current with
one top MOSFET and one bottom MOSFET.
The LTC3875 employs a unique current
sensing architecture to enhance its signalto-noise ratio, enabling current mode
control even with a small sense signal
from a very low inductor DCR—1mΩ
or less. As a result, efficiency is high and
jitter is low. Current mode control yields
fast cycle-by-cycle current limit, current
sharing and easy feedback compensation.
The LTC3875 can sense a DCR value as
low as 0.2mΩ with careful PCB layout.
The LTC3875 uses two positive sense
pins SNSD+ and SNSA+ to acquire signals. The filter time constant of the
SNSD+ should match the L/DCR of the
output inductor, while the filter at SNSA+
should have a bandwidth five times
larger than that of SNSD+. Moreover, an
35
EFFICIENCY (%)
90
85
80
75
VIN = 12V
VOUT = 1V
fSW = 400kHz
70
65
0
20
40
60
80
100
VIN = 12V
VOUT = 1.0V
IOUT = 120A
200 LFM airflow
10 | July 2015 : LT Journal of Analog Innovation
Figure 3. Thermal scan of 4-channel regulator
VIN = 12V
VOUT = 1V
30
25
20
15
10
CHANNEL 1
CHANNEL 2
CHANNEL 3
CHANNEL 4
5
0
120
ILOAD (A)
Figure 2. Efficiency of circuit in Figure 1
INDIVIDUAL CHANNEL CURRENT (A)
95
0
20
40
80
60
ILOAD (A)
100
120
Figure 4. DC current sharing is balanced among the
four channels, even at very high current loads
design features
2.2Ω
10µF
×2
4.7µF
D1
VIN
TG2
MT2
0.25µH
DCR = 0.32mΩ
VOUT
INTVCC
BOOST2
0.1µF
L2
PHASMD
SNSD2+
220nF
715Ω
220nF
SNS1–
SNSA2+
SNSA1+
VOSNS2+
ITH1
VOSNS1–
ITH2
VOSNS2–
IFAST
TK/SS1
EXTVCC
TK/SS2
PGOOD
FREQ
220pF
100k
TAVG
13.3k
220nF
220nF 715Ω
100k
3.01k
ILIM
TRSET2
4.02k
330µF
2.5V
×12
VOSNS1+
RUN2
CLKOUT
0.1µF
+
100µF
6.3V
×14
3.57k
SNSD1+
RUN1
2.2nF
ENTEMPB
SNS2–
L1
MB1
BG1
LTC3875
TCOMP1
TCOMP2
VOUT
1V
120A
0.25µH
DCR = 0.32mΩ
0.1µF
SW1
BG2
3.57k
1µF
MT1
TG1
BOOST1
SW2
MB2
10µF
×2
D2
VIN
4.5V TO 14V
20k
TRSET1
1k
MODE/PLLIN
SGND/PGND
GND
10µF
×2
4.7µF
D3
0.25µH
DCR = 0.32mΩ
VOUT
L4
BG1
LTC3875
TCOMP1
TCOMP2
PHASMD
3.57k
SNSD2+
220nF
220nF
SNSA2+
SNSA1+
RUN1
VOSNS1+
RUN2
VOSNS2+
ITH1
VOSNS1–
ITH2
VOSNS2–
0.1µF
IFAST
TK/SS1
EXTVCC
TK/SS2
PGOOD
FREQ
TRSET2
100k
TAVG
L3
MB3
3.57k
SNSD1+
SNS1–
0.25µH
DCR = 0.32mΩ
0.1µF
ENTEMPB
SNS2–
CLKOUT
100pF
SW1
BG2
1µF
MT3
TG1
BOOST1
SW2
MB4
715Ω
INTVCC
BOOST2
0.1µF
10µF
×2
D4
VIN
TG2
MT4
2.2Ω
220nF
220nF 715Ω
3.01k
ILIM
TRSET1
MODE/PLLIN
SGND/PGND
1k
D1–D4: CMDSH-3
L1–L4: WÜRTH 744301025
MTx: BSC050NE2LS
MBx: BSC010NE2LSI
GND
July 2015 : LT Journal of Analog Innovation | 11
Figure 5. A single-output, 4-phase (1.0V/120A) converter featuring LTC3875 and LTC3874
2.2Ω
10µF
×2
4.7µF
D2
VIN
TG2
MT2
0.25µH
DCR = 0.32mΩ
L2
INTVCC
BOOST2
0.1µF
SW1
SNSD2+
220nF
715Ω
220nF
ENTEMPB
SNS1–
SNSA2+
SNSA1+
RUN2
VOSNS2+
ITH1
VOSNS1–
ITH2
VOSNS2–
0.1µF
TK/SS1
EXTVCC
TK/SS2
PGOOD
220pF
100k
220nF 715Ω
100k
3.01k
ILIM
FREQ
TRSET2
4.02k
13.3k
220nF
IFAST
CLKOUT
2.2nF
TAVG
20k
TRSET1
1k
MODE/PLLIN
SGND/PGND
GND
2.2Ω
10µF
×2
D4
4.7µF
0.25µH
DCR = 0.32mΩ
TG0
BOOST0
0.1µF
L3
MB4
715Ω
220nF
SW1
BG1
ISENSE0
ISENSE1+
ISENSE0–
ISENSE1–
LOWDCR
FAULT1
MODE0
ILIM
MODE1
ITH0
PHASMD
SYNC
12 | July 2015 : LT Journal of Analog Innovation
0.25µH
DCR = 0.32mΩ
L4
MB3
715Ω
220nF
EXTVCC
FAULT0
ITH1
120k
0.1µF
LTC3874
+
1µF
MT3
BOOST1
BG0
RUN1
100pF
INTVCC
TG1
SW0
RUN0
2N7002
10µF
×2
D3
VIN
MT4
+
100µF
6.3V
×14
VOSNS1+
RUN1
10nF
VOUT
1V
120A
3.57k
SNSD1+
SNS2–
L1
MB1
BG1
LTC3875
TCOMP1
TCOMP2
PHASMD
0.25µH
DCR = 0.32mΩ
0.1µF
BG2
3.57k
1µF
MT1
TG1
BOOST1
SW2
MB2
10µF
×2
D1
VIN
4.5V TO 14V
FREQ
GND
75k
GND
D1–D4: CMDSH-3
L1–L4: WÜRTH 744301025
MTx: BSC050NE2LS
MBx: BSC010NE2LSI
330µF
2.5V
×12
design features
The LTC3875 delivers an outsized set of features for its small 6mm × 6mm
40-pin QFN. It offers high efficiency with reliable current mode control, ultralow
DCR sensing and strong integrated drivers. Tracking, multichip operation,
and external sync capability fill out its menu of features.
additional temperature compensation circuit can be used to guarantee
the accurate current limit over a wide
temperature range, and DCR variation.
Efficiency can be optimized with an
ultralow DCR inductor. As shown in
Figure 2, the total solution efficiency in
forced continuous mode (CCM) is 87.1%
at 12V input and 1.0V, 120A output.
The hot spot (bottom MOSFET) temperature rise is 58.1°C with 200 LFM
airflow as shown in Figure 3, where the
ambient temperature is about 25°C.
The DC current sharing among the
four channels is shown in Figure 4. The
difference at full load is about 2.0A
(±3.5%) with a 0.32mΩ DCR inductor.
THE LTC3874 SLAVE CONTROLLER
REDUCES SOLUTION SIZE AND
COMPONENT COUNT IN ALTERNATE
1V, 120A CONVERTER
Figure 5 shows an alternative to the
4.5V~14V input, single-output application shown in Figure 1—in this case using
an LTC3875 and an LTC3874. The LTC3874
phase extender acts as a slave controller, but it supports all the programmable
features as well as fault protection.
• ITH pins of the LTC3875 and LTC3874
are connected for current sharing.
•The CLKOUT pin of the LTC3875 is connected to the SYNC pin of the LTC3874
to synchronize switching frequency.
•The MODE pin of the LTC3874 is
connected to PGOOD, which allows
DCM operation during start-up
period for pre-bias load condition.
•The FAULT pin of the LTC3874 is pulled
up to the INTVCC pin and is connected
to the PGOOD pin of LTC3875 via a
TK/SS pin voltage-controlled MOSFET.
When the PGOOD pin is pulled low due
to a fault, the LTC3874 can shut down
both channels for protection purposes.
CONCLUSION
The LTC3875 delivers an outsized set of
features for its small 6mm × 6mm 40-pin
QFN. It offers high efficiency with reliable current mode control, ultralow DCR
sensing and strong integrated drivers.
Tracking, multichip operation, and external sync capability fill out its menu of
features. Furthermore, the slave controller LTC3874 offers a smaller footprint
solution when paired with the LTC3875.
The LTC3875 and LTC3874 are ideal for
high current applications, such as telecom and datacom systems, industrial
and computer systems applications. n
Like the LTC3875, the LTC3874’s current
mode control is accurate even with
sense signals from an inductor DCR
below 1mΩ. Compared to the master
LTC3875, the LTC3874 simplifies pinout
and uses only one set of RC components
for DCR current sensing. The filter time
constant of the RC filter should have a
bandwidth five times larger than that
of the L/DCR of the output inductor.
The total solution efficiency and thermal performance is similar to that of the
two-LTC3875 solution. The DC current
sharing among four channels is accurate. The difference at full load is about
1.6A with a 0.32m Ω DCR inductor.
July 2015 : LT Journal of Analog Innovation | 13
3mm × 3mm Monolithic DC/DC Boost/Inverting Converters
with 65V Power Switches
Joshua Moore
The vast array of power supply rails required by modern electronics has popularized the
use of compact, easy-to-use monolithic DC/DC converters, such as the LT3580 boost/
inverting converter. The LT8580, LT8570, and LT8570-1 build on the success of the LT3580,
increasing the switch voltage to 65V and the input voltage to 40V, while retaining features
and pin compatibility. The LT8580 includes a 65V, 1A power switch, whereas the LT8570 and
LT8570-1 step the switch current limit down to 0.5A and 0.25A, respectively. Various current
options enable application optimization—a monolithic converter sized for specific demands
can be smaller and more efficient than one designed for greater load currents. Optimized
sizing for current limit helps limit input and output current in the event of a short or failure.
In addition to the new options, all
devices in the family—LT®3580, LT8580,
LT8570, and LT8570-1—are pin compatible. With a few simple component
changes, the same PCB layout can be
used for a range of applications, allowing fast turnaround design changes and
reuse. The LT8580, LT8570, and LT8570-1
retain the LT3580’s features such as single
resistor feedback, for both positive and
negative output voltages, overtemperature
protection, frequency foldback, and an
external clock input pin. And, like the
LT3580, many features are user adjustable, including oscillator frequency,
soft-start, UVLO and output voltage. All
are available in thermally enhanced 8-pin
3mm × 3mm DFN or 8-pin MSE packages.
65V POWER SWITCH
The LT8580/LT8570/LT8570-1 incorporate
an internal 65V power switch, for applications with high input and output voltages.
Furthermore, VIN is capable of handling
up to 40V. This can greatly simplify
applications. For example, Figure 1 shows
the necessary circuitry to create a 48V
output with the LT3580; the LT8570’s 65V
switch simplifies the circuit in Figure 2.
Table 1. Feature comparison of monolithic pin-compatible boost/inverting DC/DC converters
LT3580
LT8570
LT8570-1
LT8580
Input Range
2.5V to 32V
2.55V to 40V
2.55V to 40V
2.55V to 40V
Max Switch Voltage
42V
65V
65V
65V
Max Switch Current
2A
1A
0.5A
0.25A
Integrated Power Switch
L
L
L
L
Frequency Foldback
L
L
L
L
External Clock Input
L
L
L
L
Overtemperature Protection
L
L
L
L
Positive And Negative Output Voltages
L
L
L
L
Single Resistor Feedback
L
L
L
L
Packages
8-Pin MSE
8-Pin MSE
8-Pin MSE
8-Pin MSE
14 | July 2015 : LT Journal of Analog Innovation
design features
All devices in the family—LT3580, LT8580, LT8570, and
LT8570-1—are pin compatible. With a few simple component
changes, the same PCB layout can be used for a wide variety of
applications, allowing fast turnaround design changes and reuse.
D3
D2
R1
L1
VIN
12V
VOUT
48V
COUT2
D1
C1
L1
VIN
12V
RSHDN
VOUT
48V
RSHDN
VIN
SW
SHDN
FB
VIN
RFB
COUT1
LT3580
CIN
SYNC
GND
SYNC
RC
RT
CF
CC
CSS
The LT8580, LT8570 and LT8570-1 include
a number of configuration options. The
oscillator frequency can be adjusted from
200kHz to 1.5MHz. While lower switching frequencies tend to be more efficient,
higher switching frequencies offer smaller
solution sizes. Also, choice of oscillator
frequency may be useful for avoiding
interference with sensitive RF circuitry.
15µH
FBX
VIN
56.2k
SYNC
6.04k
47pF
0.22µF
FBX
3.3nF
Figure 3. LT8580 configured as 5V input to 12V
output boost converter
130k
1µF
LT8570-1
0.47µF
SS
VOUT
12V
50mA
SW
SHDN
4.7µF
VC
GND
A final configuration option is softstart. By varying the soft-start capacitor, the user can adjust the rate of
increase of the inductor current. The
faster the inductor current increases, the
faster the output rises during start-up.
However, allowing the inductor current
to increase slowly reduces output voltage overshoot and avoids large input
transient currents during start-up.
10k
130k
LT8580
SYNC
CF
CC
CSS
47µH
VIN
5V
SW
SHDN
2.2µF
Another configuration option is undervoltage lockout, which, for most applications, is configurable with just one
resistor from VIN to SHDN. This allows
the parts to be used in situations where
source impedance may be high, the
source may ramp slowly or where it is
desirable that the part not discharge
the source below some threshold.
VOUT
12V
200mA
10k
VIN
RC
SS
GND
Figure 2. LT8570 configured for 48V output
USER CONFIGURABILITY
VIN
5V
COUT
VC
RT
Figure 1. LT3580 configured for 48V output
RFBX
FBX
LT8570
CIN
SS
RT
SW
SHDN
VC
RT
RT
D1
RT
VC
GND
56.2k
6.04k
SS
47pF
0.22µF
3.3nF
Figure 4. LT8570-1 configured as a 5V input to 12V output
boost converter
July 2015 : LT Journal of Analog Innovation | 15
VIN
9V TO 16V
UP TO 40V
TRANSIENT
•
C1
1µF
L1
22µH
D1
L2
22µH
487k
SW
SHDN
COUT
4.7µF
VC
GND
84.5k
CIN
4.7µF
16.2k
SS
0.22µF
BOOST CONVERTER
COUT
4.7µF
VC
GND
13.7k
SS
47pF
0.22µF
10nF
Figure 6. LT8580 configured as 5V–40V In to −15V out
dual inductor inverting converter
LT8580 required 0805 size capacitors, the
LT8570-1 can use 0603 size capacitors.
SEPIC CONVERTERS
The SEPIC topology creates a positive
voltage where the input voltage may
be less than or greater than the output
voltage. Due to a lack of DC path from
input to output, it also offers output
disconnect, so that there is no output
voltage if the converter is shut down.
Output disconnect makes the converter
resistant to damage in case of output
shorts. The application in Figure 5 shows
the LT8580 configured to produce 12V
from an input range of 9V to 16V, and
able to survive 40V transients on VIN .
90
640
80
560
70
480
60
400
50
320
40
240
160
30
EFFICIENCY
POWER LOSS
20
10
0
50
100
150
LOAD CURRENT (mA)
200
POWER LOSS (mW)
The boost topology creates an output
voltage greater than the input voltage.
Since the boost converter is the simplest
topology for LT8580, LT8570, and LT8570-1,
it can clearly illustrate how the different
current limits affect solution size. Figure 3
shows the LT8580 in a 12V out boost converter and Figure 4 shows the LT8570-1
in the same converter. Note that the
only significant circuit changes required
between the two are the inductor, the
input capacitor, and the output capacitor.
Both applications use similar inductors in
the Würth WE-LQS family, but the LT8580
requires an inductor that is 5mm × 5mm,
while LT8570-1 can use an inductor
that is only 3mm × 3mm. This reduces
the inductor footprint from 25mm2
to 9mm2. At the same time, the height
drops from 4mm to 1.5mm. Also, where
182k
L1, L2: COILCRAFT 22µH MSD7342-223
D1: CENTRAL SEMI CMMSH1-60
CIN: 4.7µF, 50V, 1206, X5R
COUT : 4.7µF, 25V, 1206, X7R
C1: 1µF, 100V, 0805, X7S
EFFICIENCY (%)
The FBX pin on LT8580/LT8570/LT8570-1
makes setting output voltage easy for both
inverting and noninverting topologies. In
both cases, only a single resistor from VOUT
to FBX is needed to set the output voltage—the converter topology determines
whether the output is positive or negative.
16 | July 2015 : LT Journal of Analog Innovation
SYNC
113k
Figure 5. LT8580 configured as 9V–16V In to 12V
output SEPIC converter
SIMPLE AND EASY OUTPUT VOLTAGE
CONFIGURATION
FBX
LT8580
1nF
L1, L2: WÜRTH 22µH WE-DD 744877220
D1: DIODES INC. DFLS1100
CIN: 4.7µF, 50V, 1206, X7R
COUT : 4.7µF, 25V, 1206, X7R
C1: 1µF, 100V, 0805, X7S
VOUT
–15V
90mA (VIN = 5V)
210mA (VIN = 12V)
420mA (VIN = 40V)
•
SW
SHDN
RT
22pF
L2
22µH
D1
VIN
LT8580
RT
C1
1µF
L1
22µH
10k
130k
FBX
SYNC
•
VIN
5V TO 40V
•
VIN
CIN
4.7µF
VOUT
12V
240mA
80
0
Figure 7. Efficiency and power loss for Figure 6 with
VIN = 12V
DUAL INDUCTOR INVERTING
CONVERTER
The dual inductor inverting topology
creates a negative voltage from a positive
input voltage, which may be greater than
or less than the magnitude of the output
voltage. This topology, like the SEPIC,
has output disconnect. In addition, this
topology tends to have a quieter output
than the boost or SEPIC, since L2 is in
series with the output. The converter in
Figure 6 shows the LT8580 configured as
a dual inductor inverting converter with
a −15V output, and Figure 7 shows the
efficiency and power loss versus load.
CONCLUSION
The popular LT3580 monolithic boost/
inverting converter has been joined by
the pin-compatible LT8580, LT8570, and
LT8570-1 converters, which add current
options and higher voltages, while
retaining the features of the LT3580.
These new options provide an additional
means of optimizing a power supply
for a given application. Depending on
the intended load, solution size and
part counts can be reduced. By retaining pin compatibility, transition within
the LT3580, LT8580, LT8570 and LT8570-1
family of parts is easy, allowing simple
design changes and PCB reuse. n
design features
One LED Driver Is All You Need for Automotive LED Headlight
Clusters
Keith Szolusha and Kyle Lawrence
Low beam headlights, high beam headlights, daytime running lights and signal lights are
often fashioned together in a single unit or cluster, allowing designers to produce distinctive
automotive front end looks. LED lighting has found its way into these clusters, distinguishing
the high end faces of today’s luxury vehicles; but LEDs offer more than just good looks.
They have a number of technical advantages over competing lighting technologies—notably
improved efficiency, robustness and lifetime. Despite these advantages, automobile
lighting designers are challenged by the cost of replacing traditional lamps with LEDs.
A significant portion of the cost LED
lighting is driven by the costs of the
LEDs themselves, thermal management
assemblies (such as finned metal heat
sinks) and robust LED driver circuits.
Traditionally, each LED beam or light
type would require its own LED driver
PCB. Costs and complexity can be significantly reduced if a single driver is
used to drive multiple LED strings (in
series) within the lighting cluster.
A comprehensive, multi-LED-string driver
must support the high voltages and high
currents required by high power LED
strings. It must also deftly handle the
on/off transitions of some LED strings
while others remain on and unaffected.
In an automotive environment, it should
accommodate wide ranging input and
output voltages, of the battery at the
input and the LED strings at its output.
Automotive environments also demand
that the driver feature low EMI and
open and short-circuit fault protection.
The LT3795 and LT3952 automotive LED
drivers satisfy these requirements when
used in boost and (patent-pending) boostbuck topologies. These LED drivers can
operate in high voltage boost (step-up)
Figure 1. LT3795 70W (70V 1A) automotive boost LED driver drives daytime running lights, low beam, and high beam strings in series with 95% efficiency.
L1
10µH
VIN
7V TO 18V
33µF
499k
10µF
x2
VIN
110k
IVINP
IVINN GATE
D1
4.7µF
M1 ×2
250mΩ
ES1B
M2
OPTIONAL FOR
LED+-TO-GND
SHORT PROTECTION
WITH LONG LEADS
1M
SENSE
1µF
7mΩ
EN/UVLO
499k
GND
IVINCOMP
OVLO
499k
PWM
VREF
35.7k
LT3795
CTRL2
4.7nF
10k
100k
RAMP
ISMON
6.8nF
10nF
10k
TG
SHORTLED
0.1µF
RT
100k
4.7µF
SHORTLED
OPENLED
SS
OPENLED
4.7nF
2k
D1: DIODES PDS5100
L1: COILCRAFT SER2915H-103
M1: VISHAY SiR878ADP
M2,M3,M4: VISHAY Si7113DN
MFAULT: VISHAY Si2328DS
Q1,Q2: ZETEX FMMT493
Q2
HB SIG
10k
M4
INTVCC
LB
23V
10k
4.7nF
VC
31.6k
250kHz
INTVCC
HB
23V
INTVCC
CTRL1
10k
Q1
M3
ISP
ISN
71.5k
SPREAD SPECTRUM
SELECT
10k
FB
100k
ANALOG
DIMMING
16.2k
MFAULT
DRL
23V
LB SIG
OPTIONAL FOR POLLING
DRL SIG
July 2015 : LT Journal of Analog Innovation | 17
The LT3795 and LT3952 automotive LED drivers can operate
in high voltage boost (step-up) and boost-buck (step-up and
step-down) topologies—driving series-stacked LED strings
directly from a wide automotive battery voltage range.
and boost-buck (step-up and step-down)
topologies. They support large stacks of
LED strings, accept a wide battery voltage range and can gracefully transition
the number of ON LEDs in the output.
They both feature spread spectrum
frequency modulation for reduced EMI
and short and open LED protection.
BOOST LED DRIVER FOR LOW BEAM,
HIGH BEAM, AND DAYTIME RUNNING
LIGHT
The total voltage of a low beam, high
beam and daytime running light headlight
cluster can be about 70V when driven
with 1A LEDs. The 100V+ LT3795 single
channel LED driver can drive 70W of
LEDs directly from a standard 9V–16V
automotive input—all three lights in
the cluster can be driven in series.
The combination driver circuit in Figure 1
shows how the LT3795 single channel LED
driver can be used to power 1A through
the daytime running light, low beam
and high beam headlights in a boost
topology. This allows the low and high
beam lights to be turned on and off—
daytime running lights are always on.
As the low and high beams are turned on
and off, their LED strings are added to and
subtracted from the daytime running light
strings by high current MOSFET switches
M3 and M4. These switches act as shorting-out devices. When the MOSFET is on, it
shorts out its corresponding beam, turning
it off; when the MOSFET is off, the beam
runs with 1A current. This easy-to-implement design is robust and saves significant
space, requiring no extra controllers.
18 | July 2015 : LT Journal of Analog Innovation
Switching an entire 23V beam string of
LEDs (such as low beam) on and off creates a 23V transient on the output. It is
important that on and off transitions are
not instantaneous. In this design, Q1 and
Q2 control the MOSFET on and off transitions to prevent large spikes of LED string
current, which would otherwise result as
energy that is taken up or released by the
output cap. Instantaneously switching
M3 and M4 would drop the LED current
temporarily to zero, causing a visible blink
VLED
20V/DIV
DRL+LB
DRL
in the low beam lights, or it could induce
a high current spike, up to 3A, that would
stress even the most robust LED string.
Figure 2 shows the controlled switching of M3 and M4, transitioning the LED
current and output voltage over ~500µs.
The shorting-out driver for M3 and M4
works at a rate at which the output
capacitor and the converter can handle
slow transients with less than 20% deviation in output current over a very short
DRL+LB
1A
ILED(DRL)
500mA/DIV
1A
1A
1A
0A
200µs/DIV
DRL+LB+HB
46V (DRL+LB)
DRL+LB
1A
1A
0A
0A
200µs/DIV
VLED
20V/DIV
ILED(DRL+LB)
500mA/DIV
1A
1A
ILED(HB)
500mA/DIV
ILED(LB)
500mA/DIV
200µs/DIV
69V (DRL+LB+HB)
ILED(DRL+LB)
500mA/DIV
VLED
20V/DIV
ILED(DRL)
500mA/DIV
0A
ILED(LB)
500mA/DIV
VLED
20V/DIV
DRL
ILED(HB)
500mA/DIV
200µs/DIV
Figure 2. All cluster LED strings are driven in series by one IC channel, but no running string is significantly
affected by turning on (or off) other strings—constant brightness is maintained even as low beam and high
beam strings are turned on and off. Transitions are controlled by slowly switching on or off LED beams with
shorting-out MOSFETs, preventing current spikes on other, unchanged strings.
design features
In the combination driver circuit, the LT3795 single channel LED driver can
be used to power 1A through the daytime running light, low beam and high
beam headlights in a boost topology. This allows the low and high beam
lights to be turned on and off—daytime running lights are always on.
time. There is no perceivable blinking or
flicker in the low beam or other running
lights when a string is added to or subtracted from the always-on running lights.
The LT3795 boost LED driver circuit in
Figure 1 has 91% and 95% efficiency when
only daytime running lights are on, and
when all beams are on, respectively. It has
short-circuit and open LED protection.
With good layout and sufficient copper
area for the discrete power components,
the highest temperature rise component of
this 70W boost driver can be kept under
40°C without additional heat sinks or
airflow. EMI filters, a GATE drive resistor
and spread spectrum frequency modulation can be used for reduced EMI.
BOOST-BUCK LED DRIVER FOR
DAYTIME RUNNING LIGHT AND
SIGNAL LIGHT COMBO
Some vehicles use LED lighting for daytime
running lights and signal lights, but not
for high or low beams. Daytime running
lights are designed in a variety of different
configurations, from long strings of LEDs
with relatively low current to short strings
with high current. An IC that can support
OPTIONAL FOR POLLING
10k
10k
Q3
MFAULT
4.7nF
5V
0V
VIN
8V TO 36V
PWM
LED+ ES1B
10k
10k
10k
Q2
M2
4× WHITE
10µF
M1
2× AMBER
4.7nF
OPTIONAL SHORT
CIRCUIT PROTECTION
LED−-TO-GND
ISP
10µF
25V
DFLS260
365k
4.7uF
499k
98
96
93.1k
94
VIN
92
90
DRL + LB (VLED = 46V)
DRL (VLED = 23V)
ALL ON (VLED = 69V)
82
8
9
10 11 12 13 14 15 16 17 18
VIN (V)
Figure 3. Efficiencies of various light combinations
are between 94% and 96%.
1.2k
2.2nF
ISP
ISP
ISN
ISN
TG
TG
ISMON
OPENLED
VC
OPENLED
SHORTLED
SS
0.1µF
LED−
FB
ISMON
IVINCOMP
86
84
LT3952
CTRL
76.8k
100k
Q1
GND
PWM
DIM
ANALOG DIM
88
SW
20k
OVLO
VREF
499k
TG
250mΩ
1A LED
DFLS260
1µF
VIN IVINP IVINN
EN/UVLO
OPTIONAL FOR
LED+-TO-GND
SHORT PROTECTION
WITH LONG LEADS
ISN
LED–
L2
L1 4.7µH
4.7µH
100
EFFICIENCY (%)
10k
5V
0V
L1: WÜRTH 74437346047
L2: WÜRTH 74437324047
M1, M2, MFAULT: VISHAY Si7309DN
Q1: ZETEX FMMT591A
Q2, Q3: ZETEX FMMT493
80
both step-up and step-down conversion
can power a combination daytime running light and sometimes-on trim light or
amber signal light. Using an IC that can
seamlessly handle transitions of stackedstring voltages in a step-up and step-down
topology allows designers to focus on
light aesthetics and functionality, without
RT SYNC/SPRD INTVCC
40.2k
2MHz
SHORTLED
100k
100k
2.2µF
Figure 4. This 18W (18V, 1A) automotive boost-buck LED driver runs daytime running lights and amber signal
lights at different brightness levels. 2MHz switching frequency keeps EMI above and outside the AM band.
July 2015 : LT Journal of Analog Innovation | 19
In automotive environments, it is important that a failure of one lamp function
not impede operation of other LEDs. The LT3795 and LT3952 include fault
detection and reporting features that enable a system controller to turn on
operational LEDs, even when other strings in the series are faulty.
worrying about the driver. Dimming can
be thrown into the mix with little effort.
The (patent-pending) boost-buck LT3952
LED driver in Figure 4 regulates 1A
through a compact daytime running light
and a series amber signal or trim light. The
2-LED amber light can be blinked or PWMdimmed via the M2 shorting-out MOSFET
without affecting the brightness of the
constantly running daytime running light.
Figure 5. PWM dimming amber signal lights at 10:1
(and up to 20:1) at 120Hz does not affect the LED
string current of the daytime running lights.
The result is a single, compact 1A boostbuck LED driver whose output drives
a visibly steady daytime running light
of 2–4 LEDs, and a blinking signal light
and/or variably dimmed trim light.
without affecting the brightness of
the daytime running light. Similarly, it
can be blinked on and off at 1Hz —say
10%-dimmed “off” (or other) to 100%
“on” to act as a turn signal light.
LED current transients are minimized
by the controlled switching of MOSFET
M2—which turns on to short out the
The new boost-buck LED driver topology allows the input voltage and output voltage ranges to cross over each
other, simplifying design by reducing
the need for pre-regulation.
amber light and turns off to enable the
amber light. Figure 5 shows PWM dimming of the amber light operates at
Figure 6. Similar automotive boost-buck LED driver to Figure 4, but this one uses a switching frequency of
350kHz for improved efficiency.
5V
0V
PWM
LED+ ES1B
10k
10k
10k
ILED
DRL
500mA/DIV
ILED
AMBER
SIGNAL
LIGHT
500mA/DIV
Hz for flicker-free 10:1 dimming
120
M2
Q2
4× WHITE
M1
2× AMBER
4.7nF
LED−
L1B
L1A 22µH
22µH
VIN
8V TO 36V
120Hz PWM AMBER
10µF
ISP
DFLS260
365k
4.7µF
499k
VIN IVINP IVINN
EN/UVLO
93.1k
200µs/DIV
ANALOG DIM
LT3952
DIM
ISP
ISP
ISN
ISN
TG
TG
ISMON
OPENLED
VC
OPENLED
SHORTLED
SS
0.22µF
RT SYNC/SPRD INTVCC
287k
350kHz
SHORTLED
100k
2.2µF
L1: SUMIDA CDRH8D43-220NC (UNCOUPLED)
L2: SUMIDA CDRH6D38-220NC
L1A, L1B: COILCRAFT MSD1278-223ML (COUPLED)
M1, M2: VISHAY Si7309DN
Q1: ZETEX FMMT591A
Q2: ZETEX FMMT493
20 | July 2015 : LT Journal of Analog Innovation
LED−
FB
ISMON
IVINCOMP
6.8nF
100k
Q1
GND
PWM
CTRL
1.2k
SW
20k
OVLO
VREF
100k
ISN
250mΩ
1A LED
10µF
1
1µF
TG
100k
OPTIONAL FOR
LED+-TO-GND
SHORT PROTECTION
WITH LONG LEADS
design features
The converter is short-circuit and open LED protected. An optional low VF diode in
the LED− path provides LED−-to-GND protection in addition to the LED+-to-GND
protection from the TG MOSFET (M1) and LT3952 overcurrent detection. The
boost-buck topology has both low input and low output ripple for very low EMI,
which is reduced even further with spread spectrum frequency modulation.
The converter is short-circuit and open
LED protected. An optional low VF diode
in the LED− path provides LED−-to-GND
protection in addition to the LED+-toGND protection from the TG MOSFET
(M1) and LT3952 overcurrent detection.
The boost-buck topology has both low
input and low output ripple for very low
EMI, which is reduced even further with
spread spectrum frequency modulation.
To improve efficiency, the converter can
be operated at a switching frequency of
350kHz (Figure 6). Efficiencies of the two
options are compared in Figure 7. Note
that the 2MHz solution has the advantages
of a reduced size inductor, and EMI above
and outside of the AM band. At either
350kHz or 2MHz , uncoupled inductors
can be used in place of the single, coupled
inductor in the boost-buck topology.
Figure 7. Comparison of efficiency of 350kHz
boost-buck solution (Figure 6) and 2MHz
solution (Figure 4).
100
EFFICIENCY (%)
95
fSW = 350kHz
90
85
SHORT AND OPEN POLLING
CONCLUSION
In automotive environments, it is
important that a failure of one lamp
function does not impede operation
of other LEDs. The LT3795 and LT3952
include fault detection and reporting
features that enable a system controller
to turn on operational LEDs, even when
other strings in the series are faulty.
Combination automotive LED lights
can be driven from a single-channel
LED driver to save cost and space. High
power and high voltage strings can be
stacked in a boost topology, or various
brightness or lower voltage strings can
be turned on and off in the new, boostbuck topology. Using a single driver for
several strings saves cost and complexity while retaining aesthetic benefits.
Using the fault flags and an additional,
optional diagnostic switch (MFAULT), the
system computer can poll the LED beams
by turning them on and off to determine
which one has an open. The system
controller can run the remaining nonfaulty LED beams while the faulty beam
is shorted out. The faulty string can be
re-polled, and brought online as soon as
it is healthy again. Both the LT3795 and
LT3952 circuits handle short and open
circuits, so shorting and opening strings
poses no potential harm for the circuits.
The LT3795 and LT3952 are powerful and flexible LED driver ICs that
can be used for combination headlight
cluster LED strings. They feature high
voltage, high current, spread spectrum
frequency modulation, and shortcircuit and open LED protection. n
Additional voltage readings and shortcircuit detections can be put in place to
turn off strings that have been shorted, or
to report shorted segments that require
servicing. The LED driver circuits maintain
functionality and reliability even when
one of the LED strings has been damaged.
fSW = 2MHz
80
350kHz, DRL + SIGNAL
2MHz, DRL + SIGNAL
350kHz, DRL
2MHz, DRL
75
70
8
9
10 11 12 13 14 15 16 17 18
VIN (V)
July 2015 : LT Journal of Analog Innovation | 21
What’s New with LTspice IV?
Gabino Alonso
—Follow @LTspice at www.twitter.com/LTspice
—Like us at facebook.com/LTspice
BLOG BY ENGINEERS, FOR
ENGINEERS
Check out the LTspice® blog
(www.linear.com/solutions/LTspice)
for tech news, insider tips and
interesting points of view.
New Article: “Parallel MOSFETs in Hot
Swap Circuits” by Dan Eddleman
www.linear.com/solutions/5677
While it is often desirable, and sometimes
absolutely critical, to use multiple parallel
MOSFETs in Hot Swap™ circuits, careful
analysis of safe operating area (SOA) is
essential. Each additional parallel MOSFET
added to a circuit improves the voltage
drop, power loss, and accompanying
temperature rise of the application. But,
the parallel MOSFETs do not necessarily
improve the transient power capability of
the circuit. Unless every MOSFET is driven
by an independent control loop, temporary
high power events such as initial turn-on
What is LTspice IV?
LTspice®
IV is a high performance SPICE
simulator, schematic capture and waveform
viewer designed to speed the process of power
supply design. LTspice IV adds enhancements
and models to SPICE, significantly reducing
simulation time compared to typical SPICE
simulators, allowing one to view waveforms for
most switching regulators in minutes compared
to hours for other SPICE simulators.
LTspice IV is available free from Linear
Technology at www.linear.com/LTspice. Included
in the download is a complete working version of
LTspice IV, macro models for Linear Technology’s
power products, over 200 op amp models, as
well as models for resistors, transistors and
MOSFETs.
22 | July 2015 : LT Journal of Analog Innovation
into a load or current limiting into a shortcircuit fault have a tendency to concentrate the power into a single MOSFET.
That being said, it is safe to connect
MOSFETs in parallel to reduce the overall
resistance, using a single control loop as
long as each MOSFET’s SOA is capable of
withstanding the entire transient event.
SELECTED DEMO CIRCUITS
For a complete list of example simulations utilizing Linear devices, please
visit www.linear.com/democircuits.
Linear Regulators
• LT3086: Adjustable voltage
controlled current source
www.linear.com/solutions/4475
Buck Regulators
• LT8610AC: 5V, 3.5A, 2MHz step-down
converter (5.5V–42V to 5V at 3.5A)
www.linear.com/solutions/5721
• LTC3892: High efficiency dual 3.3V/36V
output step-down converter (7.5V–60V
to 3.3V at 5.0A & 36V at 2A)
www.linear.com/solutions/5668
• LTM®4623: Ultrathin 3A buck µModule®
regulator (4V–20V to 1.5V at 3A)
www.linear.com/solutions/5520
Boost Regulators
• LT8580: 1.5MHz , 5V to 12V boost
converter (3.5V–6V to 12V at 200m A)
www.linear.com/solutions/5236
Buck-Boost Regulators
• LTC3111: 15V, 800k Hz wide input voltage
buck-boost regulator (2.5V–15V to 5V at
A) www.linear.com/solutions/4714
1.5
VIN range regulator with
bootstrapped LDO (2.7V–40V to 5V at
1A) www.linear.com/solutions/5084
• LTC3114-1: Wide
SEPIC Converters
• LT8495: 450k Hz , 5V output
SEPIC
converter (3V–60V to 5V at 1A)
www.linear.com/solutions/5727
Multitopology Converters
• LT8471: Dual output buck &
inverting converter (6V–32V to
+5V at 1.4A & −5V at 800m A)
www.linear.com/solutions/4676
Isolated Converters
• LT3798/LT8309: Energy Star compliant
isolated converter (85V–150VAC to 5V at
2.2A) www.linear.com/solutions/5623
Surge Stoppers
• LTC3810: High efficiency switching surge
stopper (36V–75V to 57Vclamp at 5A)
www.linear.com/solutions/5639
Hot Swap Design
• LTC4218: 12V/100A Hot Swap
design using parallel MOSFETs
www.linear.com/solutions/5685
Filter Building Blocks
• LT1568: Multiple examples of
bandpass, lowpass and highpass
filters, and a sine wave converter
www.linear.com/solutions/5740
SELECT MODELS
To search the LTspice library for a particular device model, choose Component
from the Edit menu or press F2. Since
LTspice is often updated with new features and models, it is good practice to
design ideas
update to the current version by choosing
Sync Release from the Tools menu. The
changelog.txt file (see root installation
directory) list provides a revision history of changes made to the program.
Buck Regulators
• LTC3882: Dual output PolyPhase® step-
down DC/DC voltage mode controller
with digital power system management
www.linear.com/LTC3882
LED Drivers
Hot Swap Controllers
• LT3952: 60V LED driver with 4A switch
• LTC4232-1: 5A integrated
Hot Swap controller (PCIe compliant)
www.linear.com/LTC4232-1
current www.linear.com/LT3952
Supercapcitor Chargers
• LTC4234: 20A guaranteed
SOA Hot Swap
controller www.linear.com/LTC4234
• LTC3128: 3A monolithic buck-boost
supercapacitor charger and balancer
with accurate input current limit
www.linear.com/LTC3128
Op Amps
• LTC6268-10/LTC6269-10: Single/dual 500MHz
ultralow bias current FET input op amp
www.linear.com/LTC6268 n
Power User Tip
SIMPLE IDEALIZED DIODE
To use of this idealized model in LTspice, insert a .model statement for a diode (D)
with a unique name and define one or more of the following parameters: Ron, Roff,
Vfwd, Vrev or Rrev.
.model MyIdealDiode D(Ron=1 Roff=1Meg Vfwd=1 Vrev=2)
The idealized diode model in LTspice has three linear regions of conduction: on, off
and reverse breakdown. The forward conduction and reverse breakdown can further
be specified with current limit parameters Ilimit and revIlimit.
.model MyIdealDiode D(Ron=1 Roff=1Meg Vfwd=1
Vrev=2 Ilimit=1 RevIlimit=1)
Furthermore, to smooth the switch between the off and conducting states the
parameters epsilon and revepsilon can also be defined.
4
3
2
IDIODE (A)
LTspice semiconductor diode models are essential for simulations, especially when
you want to see results that include breakdown behavior and recombination current.
However, as complete as the semiconductor diode model is in LTspice, there are times
when you need a simple “idealized diode” model to quickly simulate, for example, an
active load, a current source or a current limiting diode. To assist, LTspice provides a
representation of an idealized diode model.
D1
D2
1
D3
0
−1
−2
−3
−4
−5 −4 −3 −2 −1
0 1
V1 (V)
2
3
4
5
Just for fun, in the circuit example below an idealized diode model is used to simulate
a MOSFET’s RDS(ON) in an otherwise nonsynchronous step-down controller. By using
an idealized diode model instead of the traditional Schottky diode, the conduction
losses of synchronous rectification can be easily compared.
.model MyIdealDiode D(Ron=1 Roff=1Meg Vfwd=1 Vrev=2
Ilimit=1 RevIlimit=1 Epsilon=1 RevEpsilon=1)
A quadratic function is also used between the off and on state such that the idealized
diode IV curve is continuous in value and slope, so that the transition occurs over a
voltage specified by the value of epsilon and revepsilon.
Once you have inserted your .model statement in your schematic you can edit the
diode symbol’s Value in the component attributes (Ctrl + Right Click) to match the
name you specified in your statement. For more information on LTspice diode models,
please refer to the help topics (F1).
Happy simulations!
July 2015 : LT Journal of Analog Innovation | 23
Extend Remote Sensor Battery Life with Thermal Energy
Harvesting
Dave Salerno
Wireless and wired sensor systems are often found in environments rife with ambient energy,
ideal for powering the sensors themselves. For instance, energy harvesting can significantly
extend the lifetime of installed batteries, especially when power requirements are low,
reducing long-term maintenance costs and down time. In spite of these benefits, a number
of adoption roadblocks persist. The most significant is that ambient energy sources are
often intermittent, or insufficient to power the sensor system continuously, where primary
battery power sources are extremely reliable over the course of their rated life. System
designers may be reluctant to upgrade systems to harvest ambient energy, especially when
seamless integration is paramount. The LTC3107 aims to change their minds by making it
easy to seamlessly extend battery life, by adding energy harvesting to existing designs.
With the LTC3107, a point-of-load energy
harvester requires little space, just enough
room for the LTC3107’s 3mm × 3mm
DFN package and a few external components. By generating an output voltage
that tracks that of the existing primary
battery, the LTC3107 can be seamlessly
adopted to bring the cost-savings of
free thermal energy harvesting to new
and existing battery-powered designs.
The LTC3107, along with a small source
of thermal energy, can extend battery life,
in some cases up to the shelf life of the
battery, thereby reducing the recurring
maintenance costs associated with battery replacement. The LTC3107 is designed
to augment the battery, or even supply
the load entirely, depending on the load
conditions and harvested energy available.
A digital output, BAT_OFF, is provided
to indicate whether or not the battery
is being used to power the load at any
given time. This allows the system to
monitor the effectiveness of the harvester,
and the duty cycle of the battery’s usage
24 | July 2015 : LT Journal of Analog Innovation
SENSOR
µP
3.0V
CR3032 BATTERY
MANGANESE DIOXIDE LITHIUM
XMTR
WIRELESS SENSOR
PAVG = 250µW
Figure 1. Simplified diagram of a typical battery-powered wireless sensor system
for maintenance reporting. BAT_OFF
is internally pulled up to VOUT.
Figure 1 shows a typical wireless sensor
application. This system is powered
entirely by a CR3032 3.0V primary lithium
coin cell with a capacity of 500m A-Hr.
The battery will last about eight months
in continuous operation if the average
system power demand is 250µW.
Figure 2 shows the same system, using
the same battery, with the addition
of the LTC3107-based thermal harvester to extend the battery life.
Figure 3 shows the predicted battery
life extension with the addition of
thermal energy harvesting, using a small
(15mm × 15mm) thermoelectric generator (TEG) and a 24mm2 heat sink over
a range of TEG mounting surface temperatures (assuming a 23°C ambient).
In situations where the harvested thermal
power is greater than the average power
required by the load, the battery is never
used to power the load—only 80n A of current is drawn from the battery—resulting
in a battery life approaching the five to ten
year shelf life of a typical primary battery.
Under these conditions, the battery is used
only as a reference voltage for the LTC3107
to provide the output voltage regulation
target. It is important to note that the
design ideas
The LTC3107, along with a small source of thermal
energy, can extend battery life, in some cases up to the
shelf life of the battery, thereby reducing the recurring
maintenance costs associated with battery replacement.
T1
1:100
+
1nF
C1
+
THERMOELECTRIC
GENERATOR
CIN
100µF
4V
VBAT
22µF
6.3V
LTC3107
330pF
C2
BAT_OFF
SENSOR
499k
T1: Würth 74488540070
Coilcraft LPR6235-752SML
SW
VLDO
GND
VOUT
VAUX
VAUX
+
220µF
4V
VSTORE
10µF
+
OPTIONAL
Figure 2. Wireless sensor system with battery and
the LTC3107 thermal harvester
µP
XMTR
WIRELESS SENSOR
PAVG = 250µW
LTC3107 prevents any charge current into
the battery under all operating conditions.
divider formed by the probe and the
pull-up resistor internal to the LTC3107.)
For example, for the system shown
in Figure 2, with the TEG attached to
a harvesting heat source, such as a
warm pipe or piece of machinery just
12°C above ambient temperature, the
LTC3107 can power the 250µW load
entirely with harvested energy, resulting in the elimination of many battery
service replacements over the shelf life
of the battery, as shown in Figure 3.
If the load demand exceeds the harvester’s capability, the battery is used as
needed to maintain the output voltage
and provide the necessary output power
required by the load. In these cases, the
harvester supplies as much of the load
current as possible to minimize current
from the battery, and maximize battery
life. The BAT_OFF signal remains low,
even though some of the load current is
10k
1k
100
TA = 23°C
CUI CP20151 15mm × 15mm TEG
24mm × 24mm × 22mm HEATSINK
250µW AVERAGE LOAD
10
1
25
35
30
TEG SURFACE TEMPERATURE (°C)
supplied by the harvester. The waveforms
for this condition are shown in Figure 5.
Note that under these conditions, VOUT
is regulated by the LTC3107 to about
220mV below the actual battery voltage.
If the load is dynamic, transitioning from
low to high values, then the BAT_OFF
signal may be pulsing high and low,
indicating when the harvester is able
to supply the load and when the battery is needed. This is illustrated in the
waveforms of Figure 6, which occurred
during a momentary load step.
To further extend battery life, the LTC3107
can store excess harvested energy in a
large-valued capacitor on the VSTORE
pin during light load conditions to support VOUT during periods of heavy load.
To facilitate the use of supercapacitors,
which typically have a maximum voltage
rating of 5V, the voltage on VSTORE is
internally clamped to 4.48V maximum.
100k
BATTERY LIFE EXTENSION (%)
The waveforms of Figure 4 show the
battery voltage and the LTC3107 output
voltage. As shown, the output voltage is
regulated about 30mV below the unloaded
battery voltage—seamless and transparent to the system load—providing an
output voltage for which the system is
designed. Under these conditions, the
BAT_OFF output remains high, indicating that the battery is not being used
to power the load. (Note that in these
figures, the resistive loading of the scope
probe has lowered the BAT_OFF high
voltage below VOUT due to the resistor
3.0V
CR3032 BATTERY
MANGANESE DIOXIDE LITHIUM
45
Figure 3. Battery life can be extended by years by
using a thermal harvester
This energy storage feature reduces or
eliminates battery drain during times of
increased load by automatically using
stored energy to maintain VOUT before
July 2015 : LT Journal of Analog Innovation | 25
resorting to battery power. This is illustrated in the waveforms of Figure 7. Here,
the VSTORE voltage, having charged up
during a period of light load, can be seen
dropping during a period of increased load
as it delivers energy to the load. It can
be seen that VOUT does not drop and the
BAT_OFF signal remains high, indicating
that the battery was not used to support
the output, even during the load transient.
In situations where no harvested power
is available and any stored energy is
depleted, the output power is supplied
entirely by the battery, just as it was
without the harvester, and VOUT is regulated 220mV below the battery voltage. In
this case, the harvester circuitry remains
idle, adding only 6µ A load to the battery.
The harvester waveforms in this scenario
would be the same as in Figure 5.
To protect the battery from short circuits
on VOUT, the current from VBATT to VOUT
is limited to 30m A minimum and 100m A
maximum. Therefore, steady-state loads
of at least 30m A can be supported when
running from the battery. If needed,
higher transient loads can be supported
for short durations with the help of
the decoupling capacitor on VOUT.
The steady-state output current produced by the harvester is dependent on
several factors, but is primarily limited by
the temperature differential that can be
impressed across the TEG. Note that this
is not only a function of the TEG mounting surface temperature and the ambient
temperature, but by the thermal resistance
of the heat sink used on the cool side of
26 | July 2015 : LT Journal of Analog Innovation
VBAT
1V/DIV
VOUT
1V/DIV
VBAT
1V/DIV
VOUT
1V/DIV
BAT_OFF
1V/DIV
BAT_OFF
1V/DIV
50µs/DIV
50µs/DIV
Figure 4. Harvester waveforms when PHARVEST >
PLOAD
VBAT
1V/DIV
VOUT
1V/DIV
Figure 5. Harvester waveforms when PHARVEST <
PLOAD
VBAT
VOUT
VSTORE
1V/DIV
BAT_OFF
1V/DIV
BAT_OFF
1V/DIV
500ms/DIV
500ms/DIV
Figure 6. Harvester waveforms when a brief load
transient exceeds PHARVEST
Figure 7. Using the VSTORE feature to support a
momentary increase in load
the TEG. The harvested output current
can range from as little as microamps to
several milliamps steady state. The current
that can be supplied to VOUT from VSTORE
is limited by the differential voltage
between the two pins and the internal path
resistance through the LTC3107 charge
control circuitry, which is about 120Ω
typical. Therefore the VSTORE current is
typically limited to a few milliamps as
well, and is not intended to support large
load transients. These should be handled
by the VOUT decoupling capacitor.
2.2V LDO also gets its power from the
harvester and the battery if necessary.
In addition to the BAT_OFF feature, the
LTC3107 provides a second output voltage, regulated to 2.2V by an internal low
dropout (LDO) regulator that can be
used to power loads up to 10m A. The
SUMMARY
To facilitate the adoption of thermal
energy harvesting into a wide range of
new and existing primary battery powered
applications, the LTC3107 is designed to
work with battery voltages in the range of
2V to 4V. This includes most of the popular
long-life primary batteries used in lower
power applications, such as 3V lithium
coin cell batteries and 3.6V lithium-thionyl
chloride batteries. The LTC3107 provides
the best of both worlds—the reliability
of battery power and the maintenance
cost savings of thermal energy harvesting with minimal design effort. n
design ideas
Simplify Small Solar Systems* with Hysteretic Controller
Mitchell Lee
Battery-based solar power systems in the 10W–100W range often use a switching
regulator to control battery charge. These have the advantage of high efficiency
and facilitate peak power point tracking, but only at the cost of an inductor, circuit
complexity and noise. As a simpler alternative to a switching regulator, linear
control is feasible in applications up to about 20W. While simple and quiet, linear
charge controllers generate heat, which must be shed by means of a heat sink.
The bulk, cost and assembly complexity of a heat sink somewhat nullify a linear
charge controller’s perceived advantages over a switching regulator approach.
A hysteretic controller that simply
connects or disconnects the solar panel
as needed to limit the battery’s state
of charge provides an excellent anodyne, one devoid of inductors, complexity, noise and heat sinking.
Both series and shunt hysteretic switch
topologies are possible. A series configuration opens the connection to the solar
panel when the battery has reached its
maximum charging voltage, then reconnects when the battery voltage falls to
a lower threshold. The chief difficulty
with a series configuration is driving the
high side switch, which requires either a
charge pump for an n-channel implementation or a high voltage, high side gate
drive circuit for a p-channel MOSFET.
The preferable shunt arrangement is
shown in Figure 1. In this case the switch
(S1) turns off when the battery voltage
falls below a certain threshold, allowing
the solar panel current to charge the battery. When the battery voltage exceeds
a second, higher threshold, the switch
turns on to divert solar panel current to
ground. Diode D1 isolates the battery
D1
ON
SOLAR PANEL
OFF
+
–
BATTERY
REF
when S1 shorts the solar panel. The switch
is easily implemented with an n-channel
MOSFET, directly driven by the output
of a ground-referred comparator.
Figure 2 shows a complete shunt charge
controller for a 12V lead-acid battery
using an LTC2965 100V micropower
voltage monitor as the controlling element. While it is not monitoring 100V
in this application, the LTC2965’s 3.5V
to 100V operating range generously
encompasses the normal voltage range
of a 12V battery, with plenty of margin.
The LTC2965 contains a ~78M, 10:1
divider which monitors the battery
voltage at the VIN pin. Thresholds are
generated from a precision 2.412V reference by a separate, external divider,
and compared against the attenuated
Figure 1. Shunt mode hysteretic
switch regulates battery charge
in small solar system
version of VIN . This arrangement
eliminates the need for precise, high
value resistors in the main divider.
Hysteresis is developed by switching the
comparator’s inverting input back and
forth between high and low thresholds
as set at the INH and INL pins. These trip
points determine the voltages at which battery charging commences and terminates.
Other important features include the
LTC2965’s low power operation (40µ A
total supply current including Q1’s gate
drive), built-in 0.5% accurate reference, and hysteretic operation with
independent threshold adjustment.
Operation is as follows. Initially, with
a battery voltage of less than 13.7V the
comparator output is low and Q1 is off,
allowing all available solar panel current
* Pun intended.
July 2015 : LT Journal of Analog Innovation | 27
D1
+
2A FUSE
100µF
25V
LTC2965
1A SOLAR PANEL
BAT
12V, 7.2Ah
VIN
VREF
1M
70M
+
2.412V
VIN/10
INL
OUT
Figure 2. Shunt mode hysteretic regulator.
Trip points are temperature compensated
from 0°C to 50°C
This charging scheme shares certain
attributes of cycle charging and trickle
charging. Initial charging proceeds
until the battery voltage reaches 14.7V,
whereupon the circuit begins pulse
charging to complete the process.
It is important to correctly size the battery
and the solar panel for a specific application. As a general rule, choose a maximum
or “peak” panel current equal to 10× the
load current averaged over a 24-hour
period, and a battery ampere-hour capacity equal to 100× this same averaged
figure. Peak current of a 36-cell panel is
estimated by dividing the panel’s claimed
“marketing” watts by 15. A 15W panel can
be expected to produce ~1A maximum
28 | July 2015 : LT Journal of Analog Innovation
PS
output current under favorable conditions,
but this should be verified by actual measurement of the panel under consideration.
These relationships were derived for
Milpitas, California to give 4 days’ run
time on unassisted battery power, with
the panel oriented for maximum winter
insolation. In the case of Figure 2, the circuit was designed for a continuous 100m A
load (2.4Ah /day), dictating the use of a
1A panel and 10Ah battery. The somewhat smaller battery specified in Figure 2
is undersized for about 3 days’ operating time, deprived of any solar input.
The charging thresholds are temperature
compensated by an NTC thermistor over a
0°C to 50°C range. If operated in a controlled environment, temperature compensation is unnecessary and the thermistor
VREF
LTC2965
INH
1M
1M
200k
Figure 3. ±250mV trim scheme. Add to VREF and INH
pins in Figure 2.
150k
47nF
95.3k
7.78M
to pass through D1 to the battery and
load. As the battery charges, its voltage
rises and upon reaching an upper charging
limit of 14.7V, Q1 turns on, shorting the
solar panel to ground. D1 isolates the battery from the shunt path. With Q1 on, the
battery voltage falls at a rate dependent on
the state of charge and the magnitude of
the load current. When the battery voltage
reaches a lower float limit of 13.7V, Q1
turns off and the panel current is, once
again, applied to the battery and load.
46.4k
INH
6.81k
–
Q1
100mA MAXIMUM
24-HOUR AVERAGE
LOAD CURRENT
RS
GND
NTC
100k
TRIP POINTS
13.7V AND 14.7V
AT 25°C
Q1: NXP BUK7640-100A (30mΩ)
D1: DIODES INC. B130
BAT: PANASONIC LC-P127R2P
NTC: MURATA NCP18WF104J03RB
and 150k resistor can be replaced by a
fixed, 249k unit. For readers who wish to
trim out errors introduced by the 1% resistors, Figure 3 shows a simple scheme for
adjusting the charging threshold ±250mV.
While solar panels are normally directed
to collect maximum total energy per
annum, a standalone system must be
optimized for operation under conditions
of minimum seasonal insolation, with
allowance made for coincidental weather
patterns. The primary concern is solar
panel orientation, which is a science unto
itself. Calculation of a theoretically ideal,
fixed orientation is relatively straightforward; nevertheless a host of non-idealities
including atmospheric scattering, fog,
clouds, shading, horizon angles and other
factors make this science inexact, at best.
An excellent overview of this subject may
be found at www.solarpaneltilt.com. n
design ideas
Powering a Dust Mote from a Piezoelectric Transducer
Jim Drew
Increasing the level of remote monitoring and control of industrial environments—such
as factories, plants and refineries—enables process engineers and managers to see
the overall health of a system or factory, ultimately improving decision making. The
easiest way to increase monitoring and control coverage is to use Dust Networks®
SmartMesh® wireless sensor networks, which enable easy installation in remote
environments. SmartMesh sensors and controllers are often deployed in locations
where electrical power connections are not readily available. For this reason, using
energy harvesting technology as the source for powering these sensors is attractive.
The LTC3330 is a nanopower buck-boost
DC/DC with energy harvesting battery life
extender technology that can be attached
to a piezoelectric transducer to provide
energy to power a Dust Networks mote.
The LTC3330 integrates a high voltage
step-down energy harvesting power supply
plus a buck-boost DC/DC converter powered by a primary cell battery to create a
single output always-on power supply that
sources power for the remote Dust mote.
PIEZO
MIDÉ V25W
AC2
22µF
25V
4.7µF, 6V
1µF, 6V
AC1
VIN
22µH
SW
CAP
VIN2
LTC3330
SWA
22µH
VOUT = 3.6V FOR EH_ON = 1
VOUT = 2.5V FOR EH_ON = 0
SWB
UV3
VOUT
UV2
SCAP
UV1
BAL
180mF
2.5V
INTERFACING THE LTC3330 WITH
THE DUST MOTE
100µF
6V
180mF
2.5V
UV0
BAT
IPK2
22µF
6V
CR2032
IPK0
EHORBAT
EH_ON
OUT1
OUT2
VIN3
1µF
6V
VSUPPLY
TX
NC7SZ58P6X
OUT0
+
PGOOD
PGVOUT
IPK1
GND
NC7SZ58P6X
When vibration energy is available, the
LTC3330 uses this as its source of power
rather than the battery. For short periods when vibration energy isn’t available, the LTC3330 charges and balances a
supercapacitor that can be called on to
support the load. The LTC3330’s combination of energy harvesting and supercapacitor charging/balancing circuitry
can extend the life of the primary cell
battery by several orders of magnitude,
resulting in significantly fewer maintenance calls to replace batteries (with
the savings multiplied by the number
of installed sensors/controllers).
GND
LINEAR TECHNOLOGY DC9003A-AB
DUST MOTE FOR WIRELESS MESH NETWORKS
Figure 1 shows the LTC3330 with an output
supercapacitor, a Dust mote attached, a
battery installed and EH_ON connected to
OUT2. In this configuration, when EH_ON
is low, VOUT is set to 2.5V and when
EH_ON is high, VOUT is set to 3.6V. A Midé
V25W piezoelectric transducer is mechanically attached to a vibration source, and
its electrical contacts are connected to
the AC1 and AC2 pins of the LTC3330. The
vibration source produces 1gRMS of force
at a 60Hz acceleration, which produces
Figure 1. Dust mote setup with a supercapacitor, a battery and EH_ON connected to OUT2
July 2015 : LT Journal of Analog Innovation | 29
an open circuit voltage of 10.6VPEAK .
Figure 2 shows the input capacitor being
recharged from the V25W piezoelectric
transducer. The input capacitor charges
from 4.48V to 5.92V in 208ms. The power
delivered from the V25W is 648µW.
The 22µ F capacitor is only 18µ F at
the applied voltage of 5.0V, so every
VIN_UVLO_RISING and FALLING event
produces 26µC of charge that can be
transferred to the output minus the
efficiency (90%) of the buck regulator
within the LTC3330. Figure 3 shows the
charging of the output supercapacitor to
3.6V with the Midé V25W transducer. It
takes approximately 3300 seconds for the
output supercapacitor to charge to 3.6V.
In Figure 1, when EH_ON is low, VOUT
is set to 2.5V and when EH_ON is high,
VOUT is set to 3.6V. The first marker in
Figure 4 indicates where the vibration
source is activated; VIN rises above the
VIN_UVLO_RISING threshold. EH_ON
goes high causing VOUT to rise toward 3.6V
(VOUT starts at 2.5V because the battery
has charge). As EH_ON goes high, PGVOUT
goes low, since the new VOUT level of 3.6V
is not yet reached. As the charge on VIN
is transferred to VOUT, VIN discharges and
when VIN reaches its UVLO_FALLING
threshold, EH_ON goes low, causing
the targeted VOUT to again be 2.5V.
EH_ON
5V/DIV
VOUT
2V/DIV
VIN
2V/DIV
500s/DIV
Figure 3. Midé 25W charging output supercapacitor
to 3.6V
30 | July 2015 : LT Journal of Analog Innovation
EH_ON
5V/DIV
VOUT
1V/DIV
PC(IN) =
=
VIN
2V/DIV
(
C IN • VIN12 VIN2 2
)
2• t
(
18µF • 5.92 2 4.48 2
)
2•208ms
= 648µW
50ms/DIV
Figure 2. Midé V25W charging the 18µF input capacitance from 4.48V to 5.92V in 208ms
Given that the output capacitor is very
large and the average load is less than
the input power supplied by the Midé
piezoelectric transducer, the output
voltage increases to the higher set point
of 3.6V over many cycles. During the
transition from the BAT set point of
2.5V to the energy harvester set point
of 3.6V, VOUT is above the 2.5V PGVOUT
threshold, hence, PGVOUT goes high
every time EH_ON goes low. This cycle
repeats until VOUT reaches the PGVOUT
threshold for the VOUT setting of 3.6V.
Figure 5 shows the discharging of VOUT
when the vibration source is removed
and VIN drops below the UVLO_FALLING
threshold causing EH_ON to go low. The
supercapacitor on VOUT will discharge
down to the new target voltage of 2.5V
at which point the buck-boost regulator
will turn on supplying power to the Dust
mote. The discharging of the supercapacitor on VOUT provides an energy source for
short-term loss of the vibration source
and extends the life of the battery.
CONCLUSION
The LTC3330 provides a complete solution for powering a Dust Networks mote
from a vibration source using the Midé
V25W piezoelectric transducer and a
primary cell battery connected to the BAT
pin. The V25W piezoelectric transducer
supports output power requirements
from a vibration source, thus extending the life of the battery. When combined with a supercapacitor attached to
VOUT, the LTC3330 enables even longer
extended battery life, reducing maintenance calls to replace batteries. n
PGVOUT
5V/DIV
PGVOUT
5V/DIV
EH_ON
5V/DIV
EH_ON
5V/DIV
VOUT
1V/DIV
VOUT
1V/DIV
VIN
2V/DIV
VIN
2V/DIV
200s/DIV
Figure 4. Midé 25W charging output supercapacitor
from 2.5V to 3.6V
200s/DIV
Figure 5. Output supercapacitor discharging when
the vibration source is switched off
new product briefs
New Product Briefs
42V, 6A OUTPUT, SYNCHRONOUS
STEP-DOWN REGULATOR WITH
INTEGRATED CURRENT SENSING
The LT8613 is a 6A, 42V input-capable
synchronous step-down switching regulator with integrated current sensing.
Synchronous rectification delivers efficiency as high as 95% while Burst Mode®
operation keeps quiescent current under
3µ A in no­load standby conditions. Its 3.4V
to 42V input voltage range makes it ideal
for automotive and industrial applications.
Its internal high efficiency switches deliver
up to 6A of continuous output current
to voltages as low as 0.97V. An internal
current sense amplifier with monitor and
control pins enable accurate input or
output current regulation and limiting.
The LT8613’s Burst Mode operation offers
ultralow quiescent current, making it well
suited for applications such as automotive
always-on systems. The LT8613’s unique
design maintains a minimum dropout voltage of only 250mV at 3A under all conditions, enabling it to excel in scenarios such
as automotive cold­crank. Furthermore,
a fast minimum on­time of only 40ns
enables 2MHz constant frequency switching from a 16V input to a 1.5V output,
enabling designers to optimize efficiency
while avoiding critical noise­sensitive
frequency bands. The LT8613’s 28-­lead
3mm × 6mm QFN package and high
switching frequency keeps external inductors and capacitors small, providing a
compact, thermally efficient footprint.
NEGATIVE INPUT (−4.5V TO −80V)
SYNCHRONOUS BUCK-­B OOST/
INVERTING DC/DC CONTROLLER
DELIVERS UP TO 20A OUTPUT
CURRENT
The LT8709 is a synchronous PWM
controller for negative-­to­-negative or
negative­-to-positive DC/DC conversion. The LT8709 is unique in solving the
problem of regulating a negative voltage
with respect to system ground, without
the need of complicated level shifting
circuitry. The device’s synchronous
operation means that the output diode is
replaced with a high efficiency p-channel
MOSFET, thereby increasing efficiency,
allowing for higher output currents (up to
20A), and eliminating the heat sink typically required in medium to high power
applications. The LT8709 can be configured
in buck, boost, buck­-boost, and inverting
topologies, making it highly versatile for
a wide range of power supply designs.
The LT8709 operates over a −4.5V to −80V
input voltage range and produces an output voltage from −0.1V to as high as 60V
or from −1.4V to as low as −80V. Its rail­
to-­rail output current monitor and control
enables the device to be configured as a
current source. The LT8709 features innovative EN/FBIN pin circuitry for slowly
varying input signals and an adjustable
undervoltage lockout function. This pin is
also used for input voltage regulation to
avoid collapsing a high impedance supply.
The fixed operating frequency is selectable from 100kHz to 750kHz and can be
synchronized to an external clock. Current
mode control provides excellent line and
load regulation. The LT8709 is configurable
for either forced continuous or pulse-skipping operating modes during light load
conditions. Additional features include
a power good indicator, thermal shutdown and integrated soft­-start circuitry.
60V/4A & 36V/8A BUCK-­B OOST
µMODULE REGULATORS IN 15mm ×
15mm BGA PACKAGE
The LTM8055 and LTM8056 are buck­
boost µModule regulators that seamlessly regulate an output voltage equal
to, greater or less than the input voltage.
Housed in 15mm × 15mm × 4.92mm BGA
package, these devices include the inductor, DC/DC regulator, MOSFETs and the
supporting components. The LTM8055
operates from input voltages ranging
from 5V to 36V, delivering a load current up to 8A. The LTM8056 has a higher
input voltage, up to 60V with 4A load
current capability. The compact solution
provided by these devices free up PCB
board space in systems such as battery
operated devices, industrial control,
avionics and solar ­powered equipment.
For battery charging or precision load
current adjustment, both devices enable
input and output average current limit setting. Moreover, input and output current
can be monitored by measuring voltage
on a signal pin. Switching frequency is
adjustable from 100kHz to 800kHz , and
each device can be synchronized to an
external clock frequency from 200kHz
to 700kHz. With one resistor, the output
voltage can be adjusted from 1.2V to 36V
(LTM8055) and 1.2V to 48V (LTM8056). n
July 2015 : LT Journal of Analog Innovation | 31
highlights from circuits.linear.com
R4, 100k
GAIN OF 11 INSTRUMENTATION AMPLIFIER
The LT6023 is a low power, enhanced slew rate, precision operational amplifier.
The proprietary circuit topology of this amplifier gives excellent slew rate at low
quiescent power dissipation without compromising precision or settling time. In
addition, proprietary input stage circuitry allows the input impedance to remain
high during input voltage steps as large as 5V. The combination of precision
specs along with fast settling makes this part ideal for MUX applications.
www.linear.com/solutions/5787
R3, 10k
R2, 10k
–
–
1/2 LT6023
VOUT
1/2 LT6023
+
VINM
R1, 100k
+
–3dB BW = 6kHz
VINP
R1 TO R4: FOR HIGH DC CMRR USE LT5400-3
ICHARGE
12V
BUS
FB
MAIN
STEP-DOWN DC/DC
0.1V UP TO 5.5V
10F
1960k
IBACKUP
2.2µH
1µF
10F
VCAP SW1
51.1Ω
SW2 SVSYS
220nF
47µF
VSYS
FBVCAP
MODE
CMPIN
RUN
VSYS
3.25V
2A
13.7k
523k
0.1µF
VMID
LTC3110
2.5V
1.8V
1.2V
SYSTEM
DC/DC
REGULATORS
976k
FB
2A BIDIRECTIONAL BUCK-BOOST DC/DC
REGULATOR AND CHARGER/BALANCER
The LTC3110 is a 2A bidirectional buck-boost DC/
DC regulator with capacitor charger and balancer.
Its wide 0.1V to 5.5V capacitor/battery voltage
and 1.8V to 5.25V system backup voltage ranges
make it well suited to a wide variety of backup
applications using supercapacitors or batteries. A
proprietary low noise switching algorithm optimizes
efficiency with capacitor/battery voltages that are
above, below or equal to the system output voltage.
www.linear.com/solutions/5786
221k
1000k
RSEN
PROG
1.50k
SGND
PGND
DIR
CHRG
CAPOK
CMPOUT
1000k
µC
END OF CHRG
CAPLOW
12V BUS
SUPERVISOR
LT3088
IN
50µA
+
–
LT3088 3.3V, 1.6A OUTPUT LINEAR REGULATOR, PARALLEL DEVICES
The LT3088 is an 800mA low dropout linear regulator designed for rugged industrial
applications. A key feature of the IC is the extended safe operating area (SOA). The LT3088
can be paralleled for higher output current or heat spreading. The device withstands
reverse input and reverse output-to-input voltages without reverse current flow.
www.linear.com/solutions/5783
SET
OUT
LT3088
IN
VIN
4.8V TO 36V
10mΩ
50µA
+
–
1µF
SET
33k
OUT
10mΩ
10µF
VOUT
3.3V
1.6A
L, LT, LTC, LTM, Linear Technology, the Linear logo, Burst Mode, Dust, Dust Networks, LTspice, PolyPhase, SmartMesh and µModule are registered trademarks, and LTP, Hot Swap and SmartMesh IP are trademarks of Linear Technology Corporation. All other
trademarks are the property of their respective owners. © 2015 Linear Technology Corporation/Printed in U.S.A./70.5K
Linear Technology Corporation
1630 McCarthy Boulevard, Milpitas, CA 95035
(408) 432-1900
www.linear.com
Cert no. SW-COC-001530