exclusive eDn e-Book

CHICAGO, IL
from the
Technical Seminar
September 29, 2010

EDN 1
Welcome to the e-book version of
EDN’s 2010 Designing with LEDs seminar
EDN’s fourth Designing with LEDs event, held in Chicago on September 29, 2010 at the Rosemont/Stephens Convention
Center addressed high-brightness (HB) LED design challenges from the viewpoint of the hardware engineer. The event featured
keynotes by lighting industry luminary Cary Eskow as well as a panel discussion among representatives from three leading LED
manufacturers on “The 50,000-hour lumen maintenance myth.” Paper topics included power control, thermal management,
and optics—all of which affect the cost, efficiency, and lifespan of LEDs. Technical papers also discussed the impact of lighting
control and communication on system cost and usability.
In order to broaden the reach of the event and make its content available to those unable to attend, we have published a
transcript of the LED manufacturers’ panel discussion and six of the technical papers in this e-book. We hope you benefit from
these and look forward to seeing many of you at our next LED Workshop!

Best Regards,
Margery Conner
Technical Editor, Power and Components
EDN magazine
margery.conner@ubm.com

Table of Contents

4 15
Panel Discussion: “LEDs and the White Paper: Practically Speaking:
50,000-hour lifetime myth” LED Light Measurement
Mark Hodapp of Philips Lumileds Wolfgang Daehn & Bob Angelo, Gigahertz-Optik

7 20
Panel Discussion: What “LM-80” Is & Is Not White Paper: What Mechanical Engineers
Paul Scheidt, Cree Should Know About LEDs
Richard Zarr, National Semiconductor
9
White Paper: Design Challenges for 21
Solar-powered HBLED Lighting White Paper: Integrated Solar Powered
Heather Robertson, Avnet Lighting Solutions
Luca Difalco, STMicroelectronics
12 25
White Paper: Understanding Efficient Heat Removal
White Paper: On the Useful Lifetime
in HB LED Applications
of LED Lighting Systems
Barry Dagan, Cool Innovations
Geof Potter, Texas Instruments

Designing with LEDs EDN 2

Allied Electronics, is a small order, high service level distributor of electronic components and
electromechanical products with more than 50 sales branches across the United States and in
Canada.

Avnet LightSpeed Whether you are considering a new application or are interested in re-visiting
an existing design, Avnet Electronics Marketing has the LED technology and product experts you
need to get the job done. Our team of engineers can help with all areas of LED technology, thermal
management, power driver stage and secondary optics. From design to delivery – Avnet Electronics
Marketing brings together the world’s foremost LED, high-performance analog and optical/
electromechanical manufacturers, along with best-in-class technical expertise and supply chain
management services.

Bergquist Thermal Clad is an insulated metal substrate circuit board providing complete thermal
management systems for surface mount and High Power LED applications. Available in standard and
custom configurations, Bergquist Thermal Clad solutions provide better thermal management with
lower die temperatures, extended LED lifetimes, and increased light output. The Bergquist Company
designs and manufactures high performance thermal management materials used to dissipate
heat and keep electronic components cool. With some of the best-known brands in the business,
including Sil-Pad, Gap Pad, Gap Fillers, Bond-Ply, and Hi-Flow phase change grease replacement
materials, Bergquist is your total thermal management supplier.

Coilcraft: See how Coilcraft’s LED Design Center makes it easy to pick the perfect inductor for your
LED driver circuit. Start with a specific IC, a driver topology, or inductor specs. In seconds, you’ll get
a list of options with performance data, pricing, even detailed loss calculations.

International Rectifier (NYSE:IRF) is a world leader in power management technology. IR’s analog
and mixed-signal ICs, advanced circuit devices, integrated power systems and components enable
high performance computing and reduce energy waste from motors, the world’s single largest
consumer of electricity. Leading manufacturers of computers, energy efficient appliances, lighting,
automobiles, satellites, aircraft and defense systems rely on IR’s power management benchmarks to
power their next generation products.

Jameco has been supplying electronic components to design engineers, product developers, educators
and hobbyists for over 35 years. Known for its personalized service, Jameco offers a wide range of
name brand and house brand pricing options, a low-price guarantee and the highest quality catalog
in the industry. Get your free catalog and start saving today.

National Semiconductor’s energy-efficient LED drivers provide constant current to arrays of

LEDs, enabling color and brightness matching over a wide temperature range. Through dimming,
thermal management, and fault protection, National’s drivers improve performance in a variety of
applications. Use the new WEBENCH® LED Architect design tool to build the optimal LED driver
system to meet your requirements. Check out National’s LED drivers and design tools today.

Philips Lumileds delivers reliable, illumination-grade power LEDs and outstanding world-wide
service and the design support that customers need to rapidly develop LED lighting solutions. As
an industry leader, Philips Lumileds’ LUXEON LEDs are used widely in retail, entertainment,
outdoor, automotive, and many other applications. Philips Lumileds is a technological pioneer in
LED development, delivering the most lumens per watt, per package and per dollar with the widest
operating range, reliability and lumen maintenance. LUXEON LEDs are enabling lighting solutions
that are more environmentally friendly, help reduce CO2 emissions and reduce the need for power
plant expansion..

Designing with LEDs EDN 3

Panel Discussion Transcript: “LEDs and the
50,000-hour lifetime myth”
Panel Discussion: “LEDs and the 50,000-hour lifetime myth”
Mark Hodapp of Philips Lumileds
Some things you may not know: 10 years ago the myth was the
exception, not the rule. 10 years ago was about the date when [the
industry] started developing what we call high-brightness LEDS.
Before that, LED technology was made with other technologies,
gallium arsenide, gallium arsenide phosphide, the allen gap
technology, and colors were red, yellow, and green And we used
epoxies for most of the LED construction. Then Nichia developed
the GaN LED that lead to the blue and white {LED color]
technologies. [The industry] found out that those materials that
were used with building LEDs up until then really weren’t all that
compatible with the blue and white [product] lines.
Source: Rensselaer Polytechnic Institute, Troy, NY; National
Lighting Product Information Program publication “Lighting
Answers: Light Emitting Diode Lighting Systems”, Volume 7,
Issue 3, May 2003, Figure 13
LRC also found out that these epoxies were really sensitive to
temperature and drive current. When run at a higher temperature,
the LEDs would degrade a lot faster. Drive them at higher drive
current, which also means higher flux levels, and these devices
would degrade a lot faster. So these LEDs really weren’t all they
were cracked up to be. And realize that 10 years ago the market was
pretty much all this type of LED technology [red, yellow, and green].
There were just a few [LED manufacturers] starting to release high-
power LEDs.

Source: Rensselaer Polytechnic Institute, Troy, NY; National

Lighting Product Information Program publication “Lighting
Answers: Light Emitting Diode Lighting Systems”, Volume 7,
Issue 3, May 2003.

Here are some of the studies that were done roughly ten years
ago. This is work that was done by the Lighting and Research
Center [LRC] at Rensselaer. They looked at lumen maintenance
for different types of LEDs. These are the standard 5mm LED
which was the commonest led at the time. LRC found out that
the shorter wavelength devices did not last all that long. The red
devices, gallium arsenide phosphide, really do last a long time:
You can see that at 10,000 hours of operating they’re still brand
new at 85% of their original light output. But they found that the Source: Narendran, N., L. Deng, R.M. Pysar, Y. Gu, and H. Yu,
shorter wavelength devices, particularly the blue and white, weren’t 2004, Performance characteristics of high-power light-emitting
doing that well. They found out that the epoxies weren’t all that diodes. Third International Conference on Solid State Lighting,
compatible with this really short wavelength light. So, at least the Proceedings of SPIE 5187: 267-275.
first white LED technology really didn’t do so well. Instead of lasting
50k hours, we were lucky to get a couple thousand hours.
This is some work that LRC did on comparing two white LEDs –
one was a standard 5mm LED and one was a high-power LED. After
18,000 hours of operation the high-power LED is still running at
80% of its initial light output while the 5mm LED after 10,000
hours was down to 40%. Their difference was in the packaging.
Remember, an LED is not only the LED chip itself and it turns out
the packaging has a huge effect. [The industry]started using silicon
material instead of epoxies and this made all the difference.

Designing with LEDs EDN 4

Here’s a slide of some independent studies with different types of So basically this is what is meant by LM80 testing. At a minimum
LEDs. What you can see from these is that the LED wavelength you’re supposed to test for 6,000 hours, with data reported at
has a huge effect – particularly the shorter wavelengths . Materials 1,000 hour increments. I believe all major LED manufacturers are
tend to be more affected by the shorter wavelengths. When you see complying with LM80 test protocols.
a product that advertises a 50,000 hour LED, you want to question
that and see if there is actually lumen maintenance test data to
back it up. Package construction and material had a huge affect on
lumen maintenance. The drive conditions, particularly with the
temperature of material and the delta flux levels of the light have a
huge affect on lumen maintenance.

These studies were done several years ago. Since then the industry
has developed a protocol for lumen maintenance testing called
the LM80 test. My message here is that you shouldn’t assume just
because it’s an LED it will last for 20,000 hours: You really need to
look at the manufacturer’s test data to prove that’s really the case.
And realize that when this myth was promulgated, 50,000 hours
was the exception, not the rule. Today, I don’t know that I would say
that it’s the rule not the exception, but it really depends on how the
LEDs were manufactured because there’s still a huge industry out With this data I wanted to show a couple of things: This is actual
there building epoxy-based LEDs that probably are not 50,000-hour data for an LED, tested after 15,000 hours. It also shows the DOE
LEDs. (Department of Energy) 6,000 hour limits – those are the little red
diamonds. Limits based on the equation shown on the slide – they
So to go to the present, I’d like to talk briefly about LM80 testing. do a simple exponential extrapolation of lumen maintenance with
Several years ago the LED manufacturers got together in conjunction simple equation of flux as a measure of time using the exponential
with the US government and wrote a spec on how to do lumen equation. If you follow the limits in the spec, and you use this
maintenance testing. The spec says how to do testing in a way that equation what you’ll find is that if the device lumen maintenance
you can compare different manufacturer’s results. The key things exceeds those limits at 6,000 hrs – or actually met those limits at
in the spec are to do the test at different control case temperatures. 6,000 hrs – then with this simple exponential model the lifetime
As a minimum there are three different temperatures: 55°C, 85°C, will be 25,000 or 35,000 hours.
and a third temperature. It turns out that we weren’t able to get
agreement on what that third temperature should be because we Equation: flux(t) = EXP (alpha t) where alpha = LN(EnergyStar
were testing at different temperatures. So 55 and 85 were standards limit)/6,000hr
but not the third which we left it up to the manufacturer to pick.
That’s how those limits were developed. The limits are from the US
When you build the product you’re supposed to thermocouple the government. These are the limits that we manufacturers would like
LEDs in the product and compare the thermocouple measurements to have our products meet or exceed to meet those requirements.
you’re making with these temps. As long as you’re temp
measurement is within the range that that the LEDs were tested by
the manufacturer, you can use that test report to show how you’re
product will behave. So the key thing is if you’re running at a high
temp then the third temperature needs to be similarly high.

The spec also sets limits for the effective air temperature in the
chamber which needs to be approximately the same as the case
temperature because it sets tolerances for how the devices are
measured.

Designing with LEDs EDN 5

When the LM80 spec was finished it talked about how to test the Paper available on reliability: http://apps1.eere.energy.gov/buildings/
parts for 6,000 hours minimum, but it didn’t talk about how to publications/pdfs/ssl/led_luminaire-lifetime-guide.pdf
extrapolate the data from, for example, 6,000 hrs to 25,000 hrs.
“Evaluating the Lifetime Behavior of LED Systems” (a methodology
There is a fairly new committee called the TM21 committee for LED testing) www.philipslumileds.com/pdfs/WP15.pdf
which has been chartered with how to take the 6,000 hr data and
extrapolate it out to 25,000 hr and onward. Our goal is to have the
spec written this year.

Lumen maintenance is important, but it’s not the whole story on

LED reliability. Lumen maintenance takes a really narrow focus,
looking at the LED component only. It doesn’t say that the light as a
whole will last 50,000 hours. The whole system is quite complicated
– you have drivers, you have optics. You have to look at the rest of
the system as well.

You can imagine all the points of the previous slide as like the inks
of this chain. We don’t really know necessarily up front where the
weakest link is.

Designing with LEDs EDN 6

Panel Discussion Transcript: What LM-80 Is -
and Is Not
by Paul Scheidt, Product Marketing Manager at Cree
Introduction
You can’t go to an LED or lighting conference without hearing
LM-80 come up so let’s go to “LM-80 School” and address what
LM-80 is and is not.
AN LM-80 report is really cheap if you want to go buy one – it’s only
$25.
What LM-80 Is
LM-80 covers the measurement of lumen maintenance for
LED packages, arrays and modules and measures their lumen
maintenance over time. Where people get tripped up is that they
think it has something to do with extrapolation methods, or has
something to do with LED lifetime. What is it is an approved
method for doing the long term lumen maintenance testing
of an LED but it does not provide any guidance or make any
recommendation about prediction methods. In fact, we have to do
6,000 hrs of testing , and there is a note which many people don’t
know about that recommends 10,000 hrs of testing in order to do
extrapolation .
y “LM-80, covers the measurement of lumen maintenance of
inorganic LED-based packages, arrays and modules”
y “This approved method does not provide guidance or make
any recommendation regarding predictive estimations or
extrapolation for lumen maintenance beyond the limits of the
lumen maintenance determined from actual measurements.”
y “..the unit shall be driven for at least 6,000 hours … 10,000
hours are preferred for the purposes of improved predictive
modeling.”
What LM-80 Reports Look Like
Here’s what the report looks like – it’s a bunch of numbers.
What LM-80 Is NOT
y An extrapolation of any kind
y A lifetime number
- IES TM-21 (scheduled for Q1 2011) will provide an
industry-standard way to derive L70 lifetime from
LM-80 data
• BUT only for each particular data set
Designing with LeDs
y A report card
- There is no “pass” or “fail”, LM-80 is simply data
- Criteria must be applied to the data
y Specific to anyone’s real LED design
-LM-80 requires that Ta=Tc, which is usually NOT the
case for real designs
It’s not a report card. It’s information overload and generally not
what peope are looking for – its just a manner of collecting data. It’s
also not a lifetime number. Like I said, LM-80 is not a report card,
there is no passing or failing: LM-80 simply is data. There must be
criteria applied to the LM-80 data in order to figure out whether you
have passed or failed anything.
Last point: It’s not really specific to anyone’s real LED design.
The LM-80 requires that the ambient temperature and the case
temperature of the LED are held within 5 degrees of each other,
and this is not actually the case for most real LED designs. So when
this question comes, “Do you have LM-80?” it usually doesn’t mean,
please give me the complete LM-80 report, because I don’t know
how to interpret it.
What’s the Real Question?
I want to get down to two more basic questions that are underneath
this “Do you have LM-80?” type of question.
First: Is your combination of thermal management & LED
choice good enough to “pass” a certain criteria set?
Which criteria?
y ENERGY STAR SSL: DOE pre-LM-80, DOE LM-80,
EPA (D&R)
y ENERGY STAR ILL: no LM-80 option
y Design Lights Consortium (DLC)
Is your combination of thermal management in your LED system
and the LED that you’ve chosen to use in the system, are those two
things together good enough to pass a certain criteria set? And this
is a question asked of the system designers of the world -- not the
LED manufacturer -- as to whether they made the right choice in
their LED system design.
And the question there is: Which criteria are you trying to pass?
There have been several instances of the Energy Star SSL [solid-
state lighting]. There’s now the Energy Star “integrated LED lamp
program” which doesn’t have an LM-80 option – you have to test the
bulbs. And then there’s the Design Lights Consortium, which was
put together to help validate designs that are not covered by Energy
Star programs.
Second: Is your lifetime rating full of crap? Is your luminaire/
lamp lifetime rating full of crap?
y All LED manufacturers have their own models
(opinions)
y TM-21 will provide a consensus opinion for certain
data sets
(con’t)
EDN 7
 y Can’t prove anything unless you always want to use 6 References
year old LEDs
ENERGY STAR SSL, Category A Applications
All the LED manufacturers have their own models for how to • Residential
– Kitchen under-cabinet 24 lm/W
project LED lifetime. They all boil down to the manufacturer’s – Portable desk lights 29 lm/W CRI>75;
opinions of how long the LEDs are going to last. TM 21 will provide – Recessed , pendant downlights 35 lm/W 2700K
3000K
a consensus opinion for certain data sets, and again those data – Ceiling-mounted luminaires 30 lm/W
3500K
– Cove lighting 45 lm/W
sets are provided under LM-80 conditions. But the note I want to – Surface-mounted directional lights 35 lm/W
leave you with is that you can’t prove anything unless you’re always – Outdoor porch lights 24 lm/W

going to use 6 year old LEDs – unless you’re always willing to wait – Outdoor step lights 20 lm/W
– Outdoor pathway lights 25 lm/W
the whole 50,000 hours --which is about 6 years --for that complete – Outdoor decorative lights 35 lm/W
testing to be done. So again you have to ask yourself: How much • Non-residential

data do you need in order to feel good about the lifetime ratings – Recessed downlights
– Under-cabinet
35 lm/W
29 lm/W
that are out there?. – Portable desk lights 29 lm/W CRI>75;
– Wall-wash luminaires 40 lm/W 2700K
y All LED manufacturers have their own models – Bollards 35 lm/W -
5000K
(opinions)
y TM-21 will provide a consensus opinion for certain http://www.energystar.gov/ia/partners/product_specs/program_reqs/SSL_prog_req_V1.1.pdf

data sets 8

y Can’t prove anything unless you always want to use 6

[1] ENERGY STAR SSL, Category A Applications
year old LEDs
y

Semiconductor Reliability Testing ESILL Lumen Maintenance Requirements

The Minimum Requirements

• Products may not be qualified on LED component testing alone (i.e. via LM-80)
• Full ENERGY STAR approval requires the following:

Testing Requirement Bulb Type Requirements

10 Lamp Test Decorative Ta ! 25°C

(15,000 hrs L70 lifetime)
(5 base-up, 5 base-down) Average of 10 lamps > 86.7% LF at 6,000 hours (250 days)
LM-79 Report from Non-Standard, • LED Lamp Power < 10W: Ta ! 25°C
Omnidirectional, • LED Lamp Power ! 10W: Ta ! 45°C
Accredited Testing Lab –
Directional
Temperature Stabilized (25,000 hrs L70 lifetime) Average of 10 lamps > 91.8% LF at 6,000 hours (250 days)

If you’ve recently flown in an airplane, driven in a

car, or talked on a cell phone, you’ve trusted your [2] ESILL Lumen Maintenance Requirements
life on this body of scientific work and testing…

[3] ESILL Lumen Maintenance Options

Fortunately we have the model of semiconductor reliability to follow.
First, a BIG caveat – electron-based testing is different than photon- ESILL Lumen Maintenance Options 9

based testing. But we are learning. The more data we get, the better Lumen Maintenance Options
we get at predicting lifetime with early data. But we do have decades 1. Claims of L70 lifetime > 25,000 hours (up to 50,000 hours)
of knowledge in how accelerated lifetime testing works. Maximum L70 Lifetime Initial Approval Final Approval
30,000 hours ! 93.1% LF @ 6000 hours (250 days) ! 91.2% LF @ 7,500 hours (312 days)
y Reliability test methods and acceptance criteria for 35,000 hours ! 94.1% LF @ 6000 hours (250 days) ! 91.5% LF @ 8,750 hours (365 days)

semiconductor components have been standardized (JEDEC, 40,000 hours ! 94.8% LF @ 6000 hours (250 days) ! 91.5% LF @ 10,000 hours (417 days)

EIAJ, others…) and practiced for decades 45,000 hours ! 95.4% LF @ 6000 hours (250 days) (not specified)

50,000 hours ! 95.8% LF @ 6000 hours (250 days) ! 91.8% LF @ 12,500 hours (521 days)

Think: processors, regulators, microcontrollers, etc.. 2. Early qualification with LM-80 data

10

Designing with LeDs EDN 8

Design Challenges for
Solar-Powered HBLED Lighting
by Heather Robertson, Avnet
Solar powered high brightness LED (HBLED) lighting systems are a
smart energy choice, and one which will continue to see increasing
adoption. Powered by a clean energy source—the sun, they also take
advantage of an extremely efficient form of lighting in HBLEDs. In
this whitepaper, we will discuss the major components of a solar
powered HBLED lighting system, as well as the design challenges in
implementing such a system.
Figure 1 Solar Powered HBLED Lighting System
In typical HBLED lighting applications, power conversion is an
important factor—and even more so in solar power applications. Not
only must power be maximized from the solar panel, but it must also
be conditioned for maximum energy storage in the battery array, and
converted for use powering the HBLEDs.
Fundamentals of Driving HBLEDs
As a diode, LEDs exhibit a steep voltage-current (V-I) curve, which
means that even a small change in voltage will result in a large
change in current and brightness. It follows then, that the most
convenient method to regulate their output is to control current
and not voltage. When multiple HBLEDs are used in a current-
controlled configuration, they are often wired in series to ensure
uniformity of brightness.
Based upon application, there are a large number of high efficiency,
highly integrated HBLED driver solutions available. Common power
conversion topologies for HBLED constant current drivers are:
boost, buck, sepic, and flyback.
HBLED drivers typically look like a well behaved load on the
battery—in most applications, there will be no need to support a
high current surge during start up due to the fact that LEDs are not
a capacitive load, and most integrated HBLED driver circuits have
soft startup protection.
Estimating Current Load
When designing a system, the size of batteries and solar panels must
be established. To do this, an estimate of current consumption is
required, usually measured in Amp-hours. Current consumption
in a design consists of standby power and active power. In this case,
standby power would refer to times when the HBLEDs are off,
such as in the daytime. In standby mode, current use would be
minimized by design, with only a small amount of sense/control
circuitry “awake”, or shut off completely with only leakage currents
Designing with LeDs
remaining. Active power refers to power consumption when the
system is operating, in this case when the HBLEDs are on. Designs
may have only one active/standby mode, or multiple active/standby
modes. Active modes usually have a well understood duty cycle-- a
percentage of time in a 24hr day in which the system is active.
IL=(IST*24)+(IAM*FDutyCycle*24) (1)
IL=Current load (Amp Hours)
IST=Standby current (Amps)
I AM =Active mode current (Amps)
FDutyCycle=Duty Cycle Factor (Percentage)
Sizing & Choosing Solar Panels
There are a number of technologies available for solar panels, and
each has its own unique characteristics. Table 1 is an overview of
technologies and their notable features. For the majority of solar
lighting applications, crystalline silicon panels are used because of
their general availability and high efficiencies. Most HBLED lighting
systems require panels less than 100W. More so than any of the
other technologies, crystalline silicon panels are available in the
form factors and smaller sizes suitable for industrial applications—
ranging from 5W to 100W. In an example of an interesting fit for
a different technology, flexible amorphous silicon has been used
to wrap a light pole with solar materials, forming the solar panel
around the light pole itself.
See Margery Connor’s article on off-grid lighting on the EDN
website for more information.
Table 1 Solar Panel Technology Summary
To size the solar panels, a number of factors need to be considered.
Optimum performance for solar panels is achieved (assuming no
active tracking) when the panels are pointed south, with a tilt angle
approximately equal to the latitude of the installation + 20 degrees.
Further optimization can be achieved by adjusting the tilt angle of
the panels 2 or 4 times per year to track seasonal changes in the path
of the sun. If the solar panels cannot be placed with optimum tilt
angle, or facing south, they will not be able to generate maximum
power. As an example, some systems require panels to be placed
parallel to the ground, with no tilt angle at all, or, they may be placed
subject to significant shading. An orientation/shading multiplication
factor for these situations must be figured into the panel sizing
equation. In addition, it is prudent to factor degradation into the size
of the panels as well. Over a 20 year period, solar panels may lose as
much as 20% of their rated output.
It is important to understand the amount and power of sunlight that
is received in a given geographic location in order to plan a solar
powered lighting system. Insolation is the term used to describe
the amount of solar irradiation received on a given surface area in a
given time, usually measured in KW-hrs/m2 per day.
EDN 9
The National Renewable Energy Lab (NREL) has a variety of solar it is important to include a design factor to account for the impact
insolation data available. When sizing solar panels for a system, of ambient temperature, as lead acid battery capacity is reduced
another fact that must be considered is the number of days that the in cold temperatures. Table 3 shows design factors for lead acid
system must be able to run without any sun at all; this is known as batteries.
solar autonomy days. Estimates for general and conservative systems
are shown in Figure 2 below, based on the worst case winter solar
insolation for the regions where the system will be installed.

Table 3 Lead Acid Battery Temperature Design Factor

A simple equation for battery capacity is shown below:

Figure 2 Solar Autonomy Days
Wpanel (2)
Wpanel=Panel Size
ILoad=Current Load
Forientation =Orientation/Shading Factor/Degradation
V battery=Nominal Battery Voltage
Winter Peak Insolation=Number of Winter Sun Hours (NREL data typically
used)
Sizing & Choosing Batteries
Another aspect of a system is energy storage. While sealed lead acid
(SLA) batteries are the most common battery type used in off-grid
solar powered HBLED lighting applications, other options may be
feasible, as shown in Table 2 below. SLA batteries can be charged
below freezing, which is desirable in solar powered applications.
While other technologies may be able to operate below freezing,
charging may be an issue at low temperature.
(3)
Cbattery=Battery Capacity
IL=Current Load
Tsolar=Number of Solar Autonomy Days
F Temperature=Temperature Design Factor
FDoD=Max Battery Discharge %
This equation is a good estimate for lead acid batteries where the
rated discharge rate is similar to the expected discharge rate. If the
discharge rate is higher or lower than the specified discharge rate of
the battery, this estimate must be adjusted accordingly. Peukert’s Law
may be used to calculate effects due to discharge rate. Lithium ion
batteries generally do not have a significant sensitivity to discharge
rate.
Maximum Power Point Tracking
Solar panels have a characteristic I-V curve which varies depending
upon irradiance and temperature. As can be seen in Figure 3 below,
there is a point on the IV curve where the panel will be generating
maximum power. In many solar applications, the system is designed
to operate the panel at this point, generating maximum power. This
is called MPPT (maximum power point tracking). Many MPPT
algorithms exist, but the goal of all of them is the same—to operate
the panel as close as possible to the characteristic peak power point
of the power curve.

Table 2 Battery Technology Summary

Lead acid battery performance is impacted significantly by
temperature, depth of discharge, and rate of discharge. These
batteries are often labeled by not only capacity (AH), but also by
rate of discharge. The rate of discharge impacts the capacity of a
battery—the faster the rate of discharge, the lower the capacity. The
converse is also true—the slower the rate of discharge, the higher the
capacity of the battery.
Depth of discharge (DoD) also needs to be carefully considered.
DoD is the amount of energy (expressed as a percent of total
capacity) that will be discharged from the battery. The greater the
DoD, the fewer cycles that the battery will be able to support—there
is a direct relationship between DoD and battery life. 50% of
capacity is the generally recommended limit for deep-cycle lead acid
batteries. Lower DoDs will extend battery life. Figure 3 Solar Panel Voltage/Current and Power/Voltage
Characteristics
In choosing the capacity of a battery, especially lead acid batteries,

Designing with LeDs EDN 10

Implementations of MPPT can be fully analog, or mixed signal,
and often include a microcontroller or state machine. In designing
a system, a cost benefit analysis should be performed to determine
if adding MPPT functionality increases energy capture enough to
offset the cost of implementation.

Charge Control
Charge controllers are used to charge batteries in a safe, efficient
manner. Depending on the application, charge controllers can be
bought off the shelf, or designed for a specific application; often with
the MPPT and charge controller circuits combined. As mentioned
in a previous section one of the most common battery types for
HBLED solar powered lighting applications is lead acid batteries.
Efficient charging of lead acid batteries requires a variety of charging
modes, including bulk charge, absorption, float and equalize.
Each state requires charging with different current and voltage
characteristics, making sensing/feedback and control an important
Figure 4 Cypress PowerPSOC Reference Design
element in the controller.
The Cypress architecture uses a current controlled buck regulator
A common architecture for an off-grid MPPT charge controller
for MPPT and battery charging. The MPPT and battery charging
implementation is the use of a boost, buck, or buck/boost switching
algorithm embedded in the PowerPSoC® uses voltage and current
regulator, and a microcontroller with analog inputs for sensing
feedback from the panel and operates the panel at its peak power
current and voltages from both the solar panel and the batteries, and
by controlling the switches in the buck regulator. The switches in
PWM outputs to control switches in the regulator.
the synchronous buck regulator circuit are also operated in a way to
ensure that the current delivered to the battery is per requirement of
Control and Communication
the current charge state of the battery.
HBLED lighting can be networked, or stand alone. Networked
lighting enables energy saving control and dimming, as well as
communication of environmental activity such as movement,
traffic, etc., as well as battery and fault status. Both wired and
wireless networks are common. Standards based wireless protocols
(such as Zigbee, etc.) and proprietary wireless protocols running
over the ISM bands of 902-928MHz and 2.4GHz are often used.
For wired networking, power line modems (PLM) are often used,
communicating over the grid.
While it may seem contradictory to have a grid connected solar
powered HBLED luminaire, the grid would primarily be used for
networked communication, as well as for an optional power source Figure 5 MPPT/Charge Control Implementation Block Diagram
for battery charging. A potential application using both wired and
The solution features two LED drivers: one floating load buck driver
wireless communication would be a wireless link connecting a
for LED loads up to 8V and 2A where the forward voltage is less
subnet of lights, with each subnet controlled by a node connected to
than the battery voltage and the other is a boost driver for LED
the main control center through PLM.
loads up to 40V and 2A where the forward voltage is more than the
battery voltage.
System Examples
Electronics for solar lighting applications lend themselves well ST Microelectronics is developing a highly integrated solar MPPT
to integration. On the market today are integrated solutions charger/HBLED driver. The fully integrated solution features a
specifically designed for solar lighting, as well as products currently lead acid battery charger with MPPT optimization, and integrated
in development. HBLED drivers. This high level of integration reduces cost, improves
reliability, and simplifies design. Releasing in late 2010, this product
Cypress Semiconductor has a complete solar charger HBLED is ideal for HBLED street lighting applications.
reference board designed around their PowerPSOC® processor.
Developed to be powered by a 12V solar panel, and to charge 12V Additional Resources
lead acid batteries, the reference design includes MPPT optimization,
For additional information on solutions in this whitepaper, as well
a battery charger and both buck and boost HBLED driver circuits.
as industry solar and HBLED related articles, reference designs and
solutions, please visit Avnet’s solar and HBLED web pages:
http://www.em.avnet.com/solar
http://www.em.avnet.com/Lightspeed
Heather Robertson
Technology Director – Solar
Avnet Electronics Marketing

PowerPSoC® is a registered trademark of Cypress Semiconductor

Corporation

Designing with LeDs EDN 11

Understanding Efficient Heat Removal
in HB LED Applications
by Barry Dagan, Cool Innovations
You don’t have to be in the lighting business to know that green
technologies are in very high demand and that at the forefront of
sustainable lighting are LED lights. LEDs have very high luminous
to watt efficiencies, very long lifetime, and are not made of toxic
materials. These factors result in reduced down time, reduced
maintenance and reduced costs. From architectural illumination to
lit street lights/signboards to ordinary home/office lighting, LEDs
are rapidly becoming ubiquitous in our reduced carbon consumption
world.
However, there is one significant limiting factor with LED
lighting. That is heat, or more specifically, temperature. Like all
semiconductors, LEDs are very sensitive to high temperatures. To
get more brightness, LEDs have to run at higher power, meaning
increased heat, and therefore higher temperatures. Running an LED
hot is unfavorable for two reasons. The first is that as temperature
increases, LEDs actually become dimmer and lose lighting efficiency.
Second and perhaps more serious, operating too hot will cut down
the lifetime of an LED, wiping out one of the most beneficial
features of the technology.
Through the wise selection of appropriate heat sink technology, life-
sapping heat can be conducted away from the LED where it is least
desirable to the surrounding air. To understand how to effectively
‘heat sink’ an LED application, it is important to know some of the
physics behind heat transfer.
Understanding Natural Convection
Most High Brightness (HB) LED lighting will be designed with
natural convection cooling, that is no forced air circulation by fan
or other inducement, for a variety of reasons. Fans generate noise,
would typically need to be replaced over the lifetime of the LED and
add to the complexity and cost of the lighting appliance. As a result,
natural convection is preferred in most scenarios.
The motivating force behind natural convection is buoyancy. As the
air in direct contact with the heat sink surface heats up, like any gas,
it expands and becomes less dense. The less dense pocket of hot air
will float upwards like a hot air balloon, drawing cooler denser air
into the heat sink. Assuming the rising hot air can be exhausted and
a source of cooler air is present, this process will cycle as long as the
heat sink remains hot.
These natural convection air currents unfortunately are very weak,
typically one to two orders of magnitude less than any forced
convection. As a result, natural convection can easily be hindered
by friction along the surfaces of the heat sink as well as friction
between the individual air molecules themselves.
Maximizing Natural Convection Performance
What heat sinks do fundamentally is multiply the surface area of a
heat producing device, such that the device can come into contact
with more air simultaneously. The more surface area a heat sink
possesses, the more heat it can transfer to the air. The flip side
though is that the more surface area a heat sink has, the more
friction there will be between the moving air and the heat sink. If
there is too much friction, the air will not move at all, rendering the
heat sink equivalent to a block of solid metal.
Designing with LEDs
For good heat sink design, it is especially important in natural
convection to find the right balance of trying to maximize surface
area but not chocking off the air flow.
Heat sink Geometry Considerations
There are a lot of geometric considerations when selecting a heat
sink. These include choosing between fins or pins, the density of the
fins or pins, the size of the heat sink, the foot print of the heat sink,
the material of the heat sink, etc.
Fin density refers to the amount of fins or pins per square inch
of the heat sink footprint. The optimum fin density should
be determined by the available air flow, in the case of natural
convection, low density is required due to low airflow. Having a low
fin density results in less surface area, but natural convection air
currents will be less encumbered resulting in better performance. If
the fin density is not sufficient for the required cooling, then the use
of a fan or other non-passive forms of cooling may be required.
Another important consideration is whether to choose a continuous
finned heat sink like most extrusions or a pin finned heat sink. In a
finned heat sink, the air can travel along the length of the fins and
to a limited extent up and down the fins perpendicular to the base.
When used in natural convection, a finned heat sink works best
when the fins are aligned parallel to gravity so that the rising air can
travel along the fins. When the base is horizontal, air can only be
drawn in from the two open sides of the heat sink, some air can fall
into the heat sink from above but it will be fighting the rising hot air
currents. This tends to lead to poorer performance for finned heat
sinks when oriented horizontally; they perform worst when the base
is vertical but the fins have been oriented perpendicular to gravity.
With respect to natural convection cooling, forged pin fin heat sinks
are a more advanced heat sink technology. Since the heat sink is
open on all sides, air can be drawn in and exhausted from all sides
of the heat sink except the base. This means that the heat sink will
perform well in any given orientation. Another not very obvious
benefit of pin fins is that since each pin is an individual element, it
is very easy to modify pin finned heat sinks into special shapes by
trimming or removing only specific pins without damaging the rest
of the heat sink. This makes it very easy to add installation features
or fit oddly shaped cavities. Pin finned heat sinks can also be forged
into round or other shape bases.
Traditional pin finned heat sinks are surpassed in performance
only by advanced flared pin finned design heat sinks. A flared pin
fin heat sink differs from a traditional vertical pin fin heat sink in
that all the pins expand outwards in a radial fashion from the base,
resembling a hedgehog.
Flared pin fin designs can significantly improve natural convection
performance by increasing the spacing between each pin row while
maintaining the same amount of total surface area. The additional
space serves to reduce air flow restriction, making it easier for hot
air to escape and fresh cold air to enter the pin matrix.
Another important consideration is the effect of pinned and finned
surfaces with regards to the aerodynamic flow of air. Air travelling
along a straight fin will develop what is called a boundary layer
along the fin wall. A boundary layer is a region of disrupted airflow
EDN 12
where air right against the heat sink surface is slowed by friction
against the heat sink; air adjacent to this slow air region is then also
slowed. At the beginning of the fin, the boundary layer is very thin
but as you travel down the length of the fin, the boundary layer
expands until a theoretical maximum or when it is disrupted by
other aerodynamic effects such as another boundary layer. These
boundary layers insulate the heat sink surfaces further down the fin
length reducing cooling efficiency.
In a pin finned heat sink, these boundary layers are broken
because each pin fin is a distinct element and there is no wall for
the boundary layer to follow. Without a continuous surface, the
boundary layer disperses. With reduced boundary layer effects,
pockets of warm air are less likely to occur, and so with a more even
air temperature, each individual heat sink pin performs better.
Square pins, a hybrid of straight finned heat sinks, do help disrupt
the formation of boundary layers. However, their sharp corners add
to aerodynamic friction comparatively to round pin fins.
Heat sink Materials
Another important consideration when choosing a heat sink is
selecting the right material for the heat sink. Primarily two metals
are used for heat sinking, namely aluminum and copper. Aluminum
is the most commonly used heat sink material for many reasons.
First, aluminum is cheaper than copper, not just in material costs
but also in fabrication. Copper is much more difficult to extrude,
forge and machine than aluminum. Aluminum also weighs less and
naturally forms a thin oxide layer that seals the surface preventing
further oxidization. Copper on the other hand is three times as
heavy and will patina over time.
Copper does have one advantage over aluminum, namely that
thermal conduction inside the metal is nearly twice that of
aluminum. Thermal conduction is the transport of heat along the
metal itself, thus improved thermal conduction means that the
heat will travel faster from the source to the farthest corners of the
heat sink.
This makes copper suitable for applications where the heat source is
very small relative to the heat sink and it is necessary to get the heat
away from a focused area fast. Copper is also suitable when there is
a very high air speed flushing through the heat sink, as the high air
speed will quickly strip heat away from the heat sink surfaces. Heat
sinks operate efficiently when the entire structure is of a uniform
temperature (the heat source will always be at least slightly hotter
than any other part of the heat sink). If the heat is being stripped
away from the ends of the heat sink fins at a high rate, then the
ends will be much cooler than the source, meaning less heat will be
Designing with LeDs
transferred to the air, thereby reducing efficiency.
In most natural convection scenarios, however, the limiting factor
of the performance is not the ability to conduct the heat along
the heat sink itself but the ability to convect the hot air out of the
heat sink. Thusly, while the conduction of heat in copper may be
twice as high as aluminum, in most natural convection situations,
performance between aluminum and copper will be practically
identical.
Heat sinks in Practical Applications
As can be expected, most lighting fixtures are ceiling mounted with
lights aimed downwards. For LEDs, this means a horizontal LED
PCB orientation meaning that the heat sink base is also horizontal.
Less common but also recurring scenarios occur when the light
orientation is unknown or is allowed to move. In this case omni-
directional heat sinks are necessary to ensure continuous, efficient
operation.
To establish how well various heat sinks performed against each
other in possible LED lighting scenarios, three heat sinks of
different geometries were tested and compared. The heat sinks were
all of identical footprint and overall heights, each to be tested at
60W power and each to be tested in three orientations. The three
orientations are vertical heat sink base (parallel with respect to
gravity), horizontal heat sink base (perpendicular to gravity) and at a
45° incline.
The three heat sinks tested were a typical extrusion suited for
natural convection, a forged pin fin heat sink with normal straight
pin fins and a forged flared pin fin heat sink with pin fins that
expand outwards from the heat source. See table for details
SEE TABLE 1
In the next table, we have the results of the thermal tests.
SEE TABLE 2
The test results show that in the most common scenario, downward
aimed lighting, the flared heat sink outperformed the extrusion
by 34% and the straight pin heat sink by 19%. In the vertical
orientation, the extrusion was the strongest performer being 8%
more powerful than the flared pin fin heat sink. The 45° orientation
predictably had a result in between the vertical and horizontal
orientations with the flared pin fin performing best.
From these test results, we can see that the flared pin fin heat sink
is the best fit for horizontal (with respect to gravity) orientation
lighting scenarios. The performance is significantly better than any
of the other heat sinks allowing for very high powered LEDs to
These figures
illustrate the
effect of air flow
around various fin
structures. The blue
areas represent cool
air while the red
areas represent hot,
slowed air.
EDN 13
be cooled without the use of forced air convection. The flared pin Natural Convection Extrusion Heat sink
fin heat sink is also suitable for lights that can be installed in any
orientation as it exhibits very strong performance characteristics in
all orientations.
The natural convection finned heat sink performed best when the
light source was oriented vertically. The finned heat sink is sensitive
to orientation as the fins must be parallel to gravity. Whether
mounted in the horizontal or vertical orientation, the air must be
able to travel up along the fin walls in order to develop natural
convection.
The straight pin finned heat sink is omni-directional and can
handle high power levels. It performs better than the extrusion in
the typical horizontal position. The straight pin finned heat sink is
suitable for LED fixtures with tight space constraints.
Flared Pin Fin Heat sink
Looking Forward
When looking at the expectations of light output from LEDs and
their associated power levels, it is obvious that heat is going to be a Table 1
significant factor in LED designs. Good High Brightness LED design Physical Characteristics of the Tested Heat sinks
will require the incorporation of more efficient and more compact
Extrusion Vertical Pin Flared Pin
thermal solutions.
Length (in) 5.0 5.0 5.0
Natural convection is going to be the dominant method of cooling
Width (in) 5.0 5.0 5.0
LEDs as it requires no power, no maintenance and has no obvious
failure mechanisms. In order to maximize natural convection Height (in) 2.5 2.5 2.5
cooling, it is important to design a system with air flow in mind: Number of fins/ 13 360 360
this means having an unlimited supply of fresh cool air, having an pins
exhaust for hot air and choosing a heat sink with low aerodynamic Pitch (in) 0.38 0.25 0.25
resistance to natural convection airflow yet a high cooling capacity.
Surface area (in2) 332 358 358
The heat sink must also be able to operate in the given orientation(s)
of the application and require minimal space. Weight (lbs) 2.11 1.84 1.84
Power (Watts) 60 60 60
Traditional round based light fixtures will have to be replaced with
more efficient square and rectangular fixtures in order to maximize
heat sink surface area. Rectangular fixtures can also be more cost Table 2
effective when designing for LED applications that don’t have Thermal Performance Results of the Tested Heat sinks
production volumes to justify a dedicated heat sink die. Extrusion Vertical Pin Fin Flared Pin Fin
Further down the road, very high powered LEDs may necessitate Horizontal
0.83(°C/W) 0.77(°C/W) 0.62(°C/W)
position
the development of no-noise, extremely long life air movers. Other
possibilities include developing conduction cooling mechanisms H-V45 Deg.
0.85(°C/W) 1.07(°C/W) 0.78(°C/W)
that use pre-existing structural elements. These may be uncommon position
solutions, but successful high brightness LED lighting will only be Vertical
0.84(°C/W) 1.26(°C/W) 0.88(°C/W)
brought about by experimentation, persistence and innovation. position

Images and Tables

Straight Pin Finned Heat sink

Designing with LeDs EDN 14

Practically Speaking:
LED Light Measurement
by Wolfgang Daehn & Bob Angelo, Gigahertz-Optik
Gigahertz-Optik is a world class manufacturer of innovative UV-VIS-
NIR optical radiation and color measurement instrumentation for
specification critical industrial, medical and research applications.
Today’s lighting industries continue to be transformed by new
light source technologies. Traditional incandescent lamps are
being replaced by discharge lamps, physical (LED) and organic
(OLED) Light Emitting Diodes. These sources are characterized
predominantly by narrow wavelength band emission spectra.
New alternative light meter technologies are required that can
offer accurate and traceable light measurements in absolute units
independent of the light source emission spectrum. Spectral
measurement technology is recommended for this new type light
source due to high measurement uncertainties of typical filter
photometers and color meters. Moving grating monochromator
spectroradiometers representing the high end in light measurement Photometry deals with the measurement of visible energy. The
instrumentation are too expensive and complex for regular everyday human eye perceives light with different wavelengths as different
industrial applications. Diode array spectrometers currently being colors, as long as the variation of wavelength is limited to the
promoted as the preferred measurement device to the standard range between 400 and 800 nm. Outside this range, the human
light and color meter have limitations in absolute scale light eye is insensitive to electromagnetic radiation and thus we have no
measurements. Due to these limitations Gigahertz-Optik has perception of ultraviolet (UV, below 400 nm) and infrared (IR,
developed a new highly accurate and cost effective spectrolightmeter above 800 nm) radiation. The sensitivity of the human eye to light
based on its new BTS256P Bi-Technology Sensor that combines of certain intensity varies strongly over the wavelength range
a photodiode and diode array for mutual improvement of each between 380 and 800 nm. Under daylight conditions, the average
technology. The hand-held instrument allows qualification of normal sighted human eye is most sensitive at a wavelength of 555
luminous flux, spectral flux distribution and color data of single nm, resulting in the fact that green light at this wavelength
LEDs already assembled to a printed circuit board. In combination produces the impression of highest “brightness” when compared to
with a larger integrating sphere LED arrays or complete luminaires light at other wavelengths. The spectral sensitivity function of the
can be measured. Although compact in size the instrument offers average human eye under daylight conditions (photopic vision) is
all of the latest recommended measurement features like auxiliary defined by the CIE and DIN spectral luminous efficiency function
lamps for integrating sphere based measurements and integrating V(λ).
time related offset compensation for the diode array. The full
function test device is useful in small or large LED processing
companies, in incoming inspection, production control and final
product qualification. Other beam type light sources such as
endoscopes, cool light fiber bundle sources and narrow beam lensed
sources can also be measured.
Light measurement plays a significant role in science and industrial
research as well as in production and quality control. But it is not
as well known a science as say voltage or current measurement. To
help inform and support readers in classifying their application this
article includes basic information on light and light measurement.

Basics of LED Light & Color Measurement

Light, or the visible part of the electromagnetic radiation spectrum,
is the medium through which human beings a major portion of
environmental. Evolution has optimized the human eye into a highly
sophisticated sensor for electromagnetic radiation. Joint performance
between the human eye and visual cortex, a large part of the human
brain, dwarfs recent technical and scientific developments in image
processing and pattern recognition. In fact a major part of the
information flow from external stimuli to our brain is transferred
visually.
Light measurement when correlated to human vision perception
is called photometry. The goal of photometric measurements
is to quantify human impressions, e.g. brightness, brilliance,
brightness contrast, darkness. CIE and DIN specify light
measurement quantities for the quantification of light sources
and lighting conditions in numbers directly relating to the
perception of the human eye. Photometric light measurement
quantities are distinguished from radiometric quantities by the
index “v” for “visual”.
Light-Meters for photometric measurements must offer a light
spectral sensitivity correlating to that of the human eye, typically the
day-light adapted eye response V(λ). The quality of the V(λ) spectral

Designing with LeDs EDN 15

match to that specified by CIE and DIN is one of the key parameters
for photometer specifications. Spectral mismatch error is the key
source for measurement uncertainty with light sources other than
tungsten lamps.
The most common photopic measurements quantities are:
y Luminous flux Φv is the basic photometric quantity and
describes the total amount of electromagnetic radiation emitted
by a source, spectrally weighted with the human eye’s spectral
luminous efficiency function V(λ). Luminous flux is the
photometric counterpart to radiant power. The unit of luminous
flux is lumen (lm), and at 555 nm, where the human eye has its
maximum sensitivity, a radiant power of 1 W corresponds to a
luminous flux of 683 lm.
y Luminous intensity Iv quantifies the luminous flux emitted by a
source in a certain direction. In detail, the source’s (differential)
luminous flux dΦv emitted in the direction of the (differential)
solid angle element dW is given by
dΦv = Iv ∙ dΩ and thus
The unit of luminous intensity is lumen per steradian (lm / sr),
which is abbreviated with the expression “candela” (cd).
1 cd = 1 lm / sr and also foot-Lambert (1 cd/m² = 0.2919 fL)
Designing with LeDs
y Luminance Lv describes the measurable photometric brightness
of a certain location on a reflecting or emitting surface when
viewed from a certain direction. It describes the luminous flux
emitted or reflected from a certain location on an emitting or
reflecting surface in a particular direction (the CIE definition of
luminance is more general. In detail, the (differential) luminous
flux dΦv emitted by a (differential) surface element dA in the
direction of the (differential) solid angle element dΩ is given by
dΦv = Lv cos(θ) ∙ dA ∙ dΩ with θ denoting the angle between
the direction of the solid angle element dΩ and the normal
of the emitting or reflecting surface element dA. The unit of
luminance is 1 lm m-2 sr-1 = 1 cd m-2
y Illuminance Ev describes the luminous flux per area impinging
upon a certain location of an irradiated surface. In detail, the
(differential) luminous flux dΦv upon the (differential) surface
element dA is given by dΦv = Ev ∙ dA. Generally, the surface
element can be oriented at any angle towards the direction
of the beam. Similar to the respective relation for irradiance,
illuminance Ev upon a surface with arbitrary orientation is
related to illuminance Ev,normal upon a surface perpendicular
to the beam by Ev = Ev,normal cos(J) with J denoting the angle
between the beam and the surface’s normal.
The unit of illuminance is lux (lx) and also foot-candle.
1lx = 0.0929 fc (lm/ft²)
Beside brightness, sensitivity to color sensations are human sensory
perceptions and light measurement technology must express them
in descriptive and comprehensible quantities. In light measurement
applications luminous color of incident light and light sources is
of main interest. According to the tristimulus theory, every color
which can be perceived by the normal sighted human eye can be
described by three numbers which quantify the stimulation of red,
green and blue cones. If two color stimuli result in the same values
for these three numbers, they produce the same color perception
even when their spectral distributions are different. Around 1930,
Wright and Guild performed experiments during which observers
had to combine light at 435.8 nm, 546.1 nm and 700 nm in such a
EDN 16
way that the resulting color perception matched the color perception
produced by monochromatic light at a certain wavelength of the
visible spectrum. Evaluation of these experiments resulted in the
definition of the standardize RGB color matching functions which
have been transformed into the CIE 1931 XYZ color matching
functions.
used if the chromaticity differs more than ∆uv=5x10-2 from the
Planckian radiator.
Color rendering is the effect of a light on the color appearance of
objects. Sources that include light of all spectral colors, e.g. sun
light, effect natural color sensations from illuminated objects. Here
the color rendering is good. Light sources with irregular spectral
color distribution effect unnatural color sensations. Here the color
rendering is poor. If for example the color of the object is not
included in the source spectrum the color rendering is gray.

These color matching functions define the CIE 1931 standard

colorimetric observer and are valid for an observer’s field of view
The Color Rendering Index CRI specifies the quality of the color
of 2°. Practically, this observer can be used for any field of view
rendering of illuminants. The CRI is calculated by comparing the
smaller than 4°. Although the XYZ tristimulus values define
color rendering of a sample source to that of a reference source. For
a three-dimensional color space representing all possible color
example a black body radiator with CCT below 5000K as compared
perceptions, for most applications the representation of color in
to day light source like D65 with CCT higher that 5000K. A
a two-dimensional plane is sufficient. One possibility for a two-
selection of reflective test color samples (TCS), specified by the CIE
dimensional representation is the CIE 1931 (x, y) chromaticity
are used to calculate the CRI of a test lamp. The first eight samples
diagram with its coordinates x and y calculated from a projection of
with relative low saturation are used to calculate the general CRI Ra
the X, Y and Z values.
of a light source. The other seven samples provide supplementary
information. Four are with high saturation the others represent well
known objects.

Color Temperature (CT) is a specification for visible light and

Light Emitting Diodes (LED) are semiconductor device incoherent
used to specify lighting conditions in lighting, photography, film
light sources with high electrical power to light power conversion
recording, publishing, and other applications. The color temperature
efficacy. As with any semiconductor device, operating temperature
of a light source is determined by comparing its chromaticity
effects changes in light output and color performance. This
with that of an ideal black body source. The color temperature
is referred to as a device’s temperature coefficient. Thermal
describes the emission spectrum of a black body sources or sources
management is therefore of primary importance in the successful
which match the color temperature of a black body source. Most
implementation of LEDs. Due to thermal drift LEDs are
artificial light sources such as fluorescent or discharge lamps and
often operated in pulsed mode. High peak intensities can be
LEDs are only nearly-Planckian black body sources. They can be
generated in this mode with reduced average electrical power
judged by their correlated color temperature (CCT). The CCT
and therefore reduced junction temperature. Sorting or grading
can be calculated for any chromaticity coordinate but the result
of individual LEDs by color differences caused by tolerances in
is meaningful only if the light sources are nearly white. The CIE
the semiconductor process is a common practice offered by most
recommends that the correlated color temperature should not be

Designing with LeDs EDN 17

semiconductor manufacturers. But due to differences in LED
manufacturer’s sorting processes and environmental conditions,
the LED lighting industry is forced to do in-house qualification
measurements.
The most common light measurement quantity used in LED testing
is luminous flux measured in lumen. This quantity corresponds to
LED efficacy by correlation of the total light output to the electrical
power. Measurement of the total light output in lm instead of
luminous intensity in cd produces much better reproducibility
because it is independent of spatial light distribution which may be
influenced by temperature, humidity, distance, different viewing
angles, misalignment and other experimental error. In research and
industry the most commonly used measurement devices for
luminous flux are light meters with an integrating sphere. The
integrating sphere acts as a light integrator for spatially emitted light.
The light source may be mounted inside or outside the sphere. The
integration effect is the result of multiple diffuse reflections on the
diffuse reflecting surface of the hollow sphere which results in a
uniform light distribution at the sphere surface. The illuminance
measured at any position on the integrating sphere surface is
therefore an indicator of the total flux generated by a light source
inside or outside of the sphere.
The size of the integrating sphere should be much larger than the
size of the test sample to keep measurement uncertainty low and
independent of the spatial light emission characteristics of the
test sample. If a smaller integrating sphere is used this must be
accounted for in the calibration procedure of the measurement
device.
Designing with LeDs
One large source of measurement uncertainty inherent with
integrating sphere use is the absorption effect. During calibration
of the sphere photometer some of the light irradiated into the
sphere will exit the sphere through the measurement port and be
absorbed in the dark room. But during actual use, the measurement
port of the integrating sphere will be fully or partially covered by
the device under test DUT. So light leaving the sphere through the
measurement port will be reflected back into the sphere adding
erroneously to the DUT light signal. Depending on the spectral
reflectivity and color of the DUT the re-reflected light will vary
in intensity and color and affect an unknown measurement error.
Auxiliary lamps are used to compensate this absorption error
by measuring the signal of the auxiliary lamp with and without
the DUT at the measurement port of the integrating sphere. The
difference in intensity is used as a correction factor for subsequent
measurements of the same kind of DUT.
Along with light intensity and color data, spectral intensity
distribution is another important test property in LED analysis.
Spectral based light meters are used for this type of measurement.
Filter type light meters employing photometric or tristimulus (RGB)
detectors are restricted to comparative or relative measurements,
e.g. LED sorting and binning against gold-standards. However
spectrometers offer different levels of quality levels, especially
diode array type spectrometers which are often limited by intensity
linearity and stray light characteristics.
An alternative method is to mate a photodiode with a diode
array like Gigahertz-Optik’s BTS256P Bi-Technology Sensor. It’s
photodiode with a precise photometric response provides a highly
linear ratio between light input and signal output over a very wide
dynamic range for very accurate luminous flux measurements. The
photodiode offers a fast response time mostly independent from the
light intensity so that the measurement signal of the photodiode can
EDN 18
be used for fast data logging and pulse synchronized measurement
applications.
Resources
y CIE 127 Measurement of LEDs
y Colorimetry and Color Quality of Solid-State Light Sources,
Wendy Davis, Optical Technology Division, NIST
y IES LM-79-08 IES Approved Method for the Electrical and
Photometric Measurements of Solid-State Lighting Products
y IES LM-80-08 Approved Method for Measuring Lumen
Maintenance of LED Light Sources
y ANSI/IESNA RP-16-05 Nomenclature and Definitions for
Illuminating Engineering
y Works in Progress by IESNA, CIE, IEC, ANSI, DOE

Spectral distribution data is provided by a separate diode array

sensor. The spectral data enables the measurement device to
calculate color data e.g. xy and u’v’ color coordinates, color
temperature CT, correlated color temperature CCT, color
rendering index CRI, peak wavelength λpeak and dominant
wavelength. The sensitivity of photodiode array is controlled
by integration time so the lower the light level the longer the
measurement time. This makes diode arrays unsuitable for fast
measurements. Longer integration times effect increases in both real
signal and dark signal. To improve the signal to noise ratio offset
compensation becomes an important rule for diode array sensors.
Best offset compensation is done with a dark signal measurement
using the same integration time as the signal measurement. A
remote controlled shutter is built into the BTS256-LED tester to
support synchronized integration time on-line offset compensation.
Low light detection with CMOS photodiode array technology
can therefore be achieved by employing offset compensation in
combination with very long integration times.
As shown, accurate light and color measurement requires a good
basic understanding of radiometric, photometric and colorimetric
principles and quantities.
With this knowledge you will be better prepared to make decisions
regarding objectives, instrumentation and compromise solutions.
A general plan of action would be to:
1. Determine the goals and purposes for the measurements
2. Estimate the acceptable levels of accuracy required
3. Decide which instrumentation best suits 1 &2 and if it meets
budgetary requirements
4. Juggle all decisions for best fit – trade-offs may be necessary

Designing with LeDs EDN 19

What Mechanical Engineers Should
Know About LEDs
by Richard Zarr, National Semiconductor
Abstract
With more lighting companies embracing LEDs as a move to improve
efficiency, mechanical engineers are being asked to redesign fixtures
to accommodate this technology. However, LEDs have different
requirements from incandescent bulbs and florescent tubes which
mechanical engineers need to take into account. This paper addresses
several of these issues in making the move to LEDs.
Introduction
As lighting fixture manufacturers begin to move away from
incandescent and fluorescent light sources in favor of Light Emitting
Diodes (LEDs), engineers need to adapt their designs to accommodate
the changes solid-state lighting technology requires. Most of the focus
on these transitions has been on the electrical changes to provide the
correct currents to high-brightness LEDs. What is often overlooked
are the mechanical considerations and how they affect the life of the
emitters as well as the physical implementation.
There are several key differences between LEDs and other light
sources. Solid-state lighting is more closely related to fluorescent tubes
since both require more complex electrical drive, are efficient and
emit little IR. However, there are other considerations plus new found
advantages that can be exploited to provide new and exciting lighting
fixture designs with LEDs.
LED Basics
Light Emitting Diodes, commonly referred to as LEDs, create light
through a process called electroluminescence. During this process
electrons give up energy in the form of photons when traversing the
LED diode junction. The materials selected to make the junction
(i.e. gallium, arsenic, phosphorous, indium, etc.) provide very specific
properties. On one side of the junction the materials will have an
abundance of charge (the “n-type” with electrons) and the other will
have a deficit of charge (the “p-type” with electron “holes”). Between
them is a zone where the electrons cannot exist (much like the shells
of an atom). This zone is called the energy “band gap” or forbidden
zone and the materials used in LEDs form band-gaps that have specific
energies that provide photon emission. These are called “direct” band-
gaps.
As positive charge enters the “p” side and negative charge enters the
“n” side, the electrons combine with the “holes” and fall to a lower
energy level emitting a photon. This is analogous to water flowing
over a dam driving hydroelectric generators. The water is losing energy
as it falls which turns the turbines and creates electricity. As the
electron crosses the junction, the wavelength of the photon created
is determined by the “band-gap” energy. By varying the band-gap,
engineers can create LEDs that emit anywhere from the infrared,
through the visible spectrum and all the way to the ultraviolet.
Modern high-brightness LEDs go further and use quantum
containment to increase the efficiency of photon generation within
the LED and some even provide optical wave-guides formed in
the device to allow trapped photons to reach the surface. Some
incorporate secondary phosphors which absorb some or all of the
LED’s emission and through a phenomenon called Stokes Shift are
able to create white light. Anyway you look at it these little devices
would have impressed even Edison in their efficiency in converting
electricity to light. But they have characteristics that fixture designers
must consider when including LEDs in a product.
LED Advantages and Disadvantages
Designing with LEDs
Light Emitting Diodes are current devices—that is, they require a
constant current (not voltage) to operate correctly. Also, because of the
construction of many white LEDs, you cannot simply vary the current
and get smooth dimming. LEDs incorporating phosphors will shift
blue if the current gets too low. There are special drive circuits such as
the LM3445 made by National Semiconductor that will allow fixtures
to be used in retrofit applications with existing TRIAC dimmers—this
is quite difficult to accomplish due to the nature of LEDs. The
LM3445 driver simplifies this by “reading” the TRIAC signal from
standard dimmers and correctly controlling pulses of constant current
to “dim” the LEDs providing a smooth, linear range familiar to
everyone. Figure 1 shows a typical drive circuit.
LEDs for general lighting are very energy efficient since the emission
is tailored to be visible to humans. Unlike incandescent bulbs, general
lighting LEDs emit very little infrared energy. One big advantage of
LEDs, especially in task lighting for kitchens, is they do not warm
food or other items sitting below them—a common problem with
incandescent task lighting. This can prematurely cause food products
to spoil or melt. A fun little test is to place an unwrapped chocolate
bar 18 inches from a 60 watt light bulb… it will melt quickly. This is
not due to the visible light, but the large amount of infrared energy
that is emitted and invisible to humans, but absorbed easily by the
chocolate. An equivalent LED fixture would not cause this to happen.
LEDs do have a dark side. The biggest advantage—the lack of IR
emission—is also a problem. Incandescent light bulbs shed most of
their waste energy through the emission of infrared. LEDs must
conduct their waste heat away through thermal management. The
LEDs may not melt the food below them in a kitchen, but they might
bake the cabinets above them (and anything in them). Also, if the
LEDs are allowed to get hot, their lifespan will decrease. LEDs rarely
die a sudden death (as incandescent filament bulbs do). They will
lose intensity over time. An LED is said to need replacing where the
emission has fallen to either the 70% or 50% point from the original
intensity. These are called the L70 or L50 points. Figure 2 shows a
typical L50 point based on the temperature of the LED.
Many new high-brightness LEDs include thermistors or RTDs to help
protect the device from damage in extreme conditions. The thermal
device can be used to fold back the current or energy being supplied
to the LED which in turn will reduce the temperature of the device.
The LM3424 is an example of a special LED driver designed to
incorporate thermal fold-back to protect the LED. This type of control
can greatly increase the lifespan of LEDs where temperatures my rise
and cause damage such as in emergency lighting, out door lighting, or
high intensity applications.
Conclusions
LEDs have great advantages over other light sources. They are very
efficient in producing light, are tiny point emitters which provide
designers with an array of options for fixture designs and come in
every imaginable color (or colors that can be mixed dynamically). They
also have very long life-spans when driven correctly—on the order of
20,000 to 50,000 hours or longer. The major downside is managing
the waste heat of the current devices. Heat sinks and other passive
methods are best suited for low cost fixtures, and active cooling such as
that produced by Nuventix can be applied to commercial high-intensity
applications. With a bit of thermal management and mechanical tricks,
LED fixtures can be engineered for a long life and novel appearance as
well as saving significant energy.
EDN 20
Integrated Solar Powered
Lighting Solutions
by Luca Difalco, STMicroelectronics
1. Introduction
The need for light has become one of the highest priority problems
to deal with for the electronic world. And when this need links very
stringent geographical and environmental conditions with life saving
situations, the priority immediately becomes an urgency—everyone in
the electronics world is called to duty to find a solution.
In the engineering world, the best way to solve a very complex
problem is to reduce it to a combination of smaller, simpler
problems so as to address each of the resulting ones with a quicker,
state-of-the-art solution.
2. The Problem Identification
In order to comply with this approach, we can split the main
problems into two categories:
yy Need for light
yy Lack of a power network
The need for light issue reflects several potential situations, which
can be grouped into four main areas:
yy Entertainment: Outdoor night time activities such as beach
games, tournaments, and camping, often require bright lighting
sources for safety and accuracy.
yy Research and Monitoring: Special weather monitoring and
geological monitoring systems reside in stations very far
from civilization and definitely as far as possible from any
electromagnetically noisy environments. Roads leading to these
remote locations are typically ‘off the grid’ and dangerous, and
overnight expeditions require adequate lighting for safety and
observation.
yy Standard Life/Educational: in non-developed areas, there can be
a complete lack of light during night disruptions and night time
emergencies; lighting is also a must for schools with both day
and evening classes in remote places where the daylight hours
are used for work and family survival.
yy Life-threatening situations: emergency camps or temporary
movable hospitals are normally situated in non-serviced areas
where bright, stationary light is crucial for medical support and
for even the simplest life saving operations.
The need of a power network in non-developed areas seems obvious:
remote villages need lighting for schools, hospitals, and jobs. And
although the very lack of power networks in natural oases like
parks and paradise landscapes is part of their appeal, there are
unfortunately cases where the lack of power suddenly becomes a
top priority problem for life improvement situations and for more
critical life saving situations as well.
Many of the most unfortunate situations happen in cases where
there is a pre-existence of a power network but due either to natural
disasters (tornados, volcano eruptions, etc.) or to war conditions,
the network is inaccessible or very dangerous due to uncapped
connections and open wires with high voltage floating around. In
these horrible situations, the above-mentioned need for light for
emergency camps and any other life saving activities comes into
play.
Designing with LEDs
3. Solar Powered LED solutions
The needs described above require a set of solutions that combined
together will result in a single system that can address all of the
issues triggered by difficult environments.
Requirements for our solutions:
yy Use the Sun
yy Make it Last
yy Make it Bright
Lighting in rural areas and in developing countries is generally
provided by wax candles or combustion lamps (e.g. kerosene), while
battery-powered flashlights are typically only used as occasional,
portable lights for intermittent use.
Combustible sources are cheaper than any form of electricity. On
the other hand the low efficiency, the poor quality of the light, and
the intrinsic fire risk advocate the use of electronic lighting in off-
grid locations. The unavailability of the power grid implies that the
electrical energy must be produced locally. Among the methods by
which energy can be produced, photovoltaic systems (solar cells) are
by far the most universally applicable.
The general principle is to convert the sunlight, in particular energy
carried by photons, into electrical energy.
The use of photovoltaic systems brings some advantages:
yy Solar energy can be produced locally; hence the solar panels can
be installed everywhere (also in areas of difficult access) without
the need for infrastructures. This also minimizes transmission/
distribution losses.
yy Renewable energy source; it does not impact the environment
with pollution.
yy Facilities can operate with little maintenance or intervention
after initial setup; this contributes to reduce the energy cost.
yy Low voltage generation; this simplifies the downstream
conversion of the power.
However, energy production by photovoltaic systems must take into
account the intrinsic periodicity of the solar source (e.g. during the
night the energy source is absent). Electrical storage into batteries
assures a continuous availability of energy.
Once the electrical energy has been produced and stored, the next
step is generating light in a wise way, in terms of energy saving and
environmental respect.
Towards this end, the use of white LEDs is moving to the forefront
for several reasons:
yy High luminous efficiency (more than 100 lumen/watt), which
implies less wasted power in comparison with other light
sources (e.g. incandescent bulbs).
yy Hazardous materials free (mercury or toxic gases), which makes
LEDs the cleanest light in ecological terms.
yy Low driving voltages, making LEDs particularly suitable for
photovoltaic systems.
EDN 21
Lately, significant steps, especially in terms of power capability,
have been achieved in LEDs technology. The availability of LEDs
up to and above 5W, combined with their high luminous efficiency,
contributes to their diffusion in the lighting market as replacements
of traditional, and often less efficient, light sources.
The solar LED lighting system described in basic outline above is
represented in figure 1.
LED lamp
PV cells
+ -
+ -
+ -
batteries
Figure 1 Solar LED lighting system.
4. PV battery charger systems
The operating principle of photovoltaic (PV) systems is quite
simple: the sunlight strikes the solar panel, manufactured from
semiconductor materials; the energy carried by the photons causes
the generation of electron-hole pairs in the semiconductor that, in
turn, generate a current that flows into a load connected to the
panel.
PV panels are realized through different technologies:
monocrystalline silicon, polycrystalline silicon, and thin film.
However, the first two technologies are the most widely employed
since they offer the highest efficiency (in the range of 20%), where
the efficiency is defined as the ratio between the power produced by
the panel and the luminous power that strikes the panel.
Thus the peak power achievable with a PV panel at STC (Standard
Test Conditions(1)) can go to almost 200W/m2.
Commercially available PV panels are offered in different sizes,
proportionally delivering different power: around 200W for the 1-1.2
m2 modules (typically 72 solar cells) used in home applications, up
to 80W for the modules for street-lighting, from 12W to 24W panels
for solar lanterns.
In a PV panel, all the cells are generally connected in series, and the
typical voltage of each cell, when delivering the maximum power, is
around 0.5V (see Figure 2).
Figure 2. Solar cell V-I characteristic.
This voltage, however, depends on the temperature and the
illumination level.
Designing with LeDs
In the interest of maximizing the power transfer, a dynamic
system called a Maximum Power Point Tracker (MPPT) should be
implemented between the solar panels and the load (battery).
This circuit samples the voltage and the current at the output of
the photovoltaic panel. Then, controlling the downstream DC-DC
converter (which can be a step-down or step-up depending on the
panel size and the load), the output current of the photovoltaic
panel is regulated in order to maximize the power transfer (see
Figure 3).
PV panel DC-DC converter battery
V I
Power PACT
calculation
-
controller
MPPT
+
algorithm PMAX
Figure 3. Block diagram of a photovoltaic system
Without the use of a MPPT dynamic system, the power delivered by
the PV panel would change a lot not only with the illumination and
the temperature, but also with the battery charge level.
This can further reduce the PV panel efficiency by 30% or more.
(1) Testing conditions to measure PV cells nominal output power
that supposes an irradiance of 1kW/m2, solar spectrum of air mass
1.5 and module temperature 25°C.
In order to maintain the battery life as long as possible, a dedicated
charging profile has to be performed. At this power level this usually
involves a microcontroller-like implementation to achieve what is
shown in Figure 4 below.
Figure 4 MPPT and battery charger profile
5. ST LED driving solutions
Now that the energy has been converted from the sun and stored
in a battery, the other big portion of the solution would be to
transform this energy into bright light with low power consumption.
In fact, the efficiency of the conversion of the electric energy into
light is always a key point, but it gains further importance in this
specific application with the target of minimizing the size and cost of
the system (battery and PV
panel included).
With their efficiency in power, LEDs make a huge contribution
to luminance conversion. Among the all lighting sources and
techniques available in the market, LEDs provide the best
performance in terms of light efficiency and pave a perfect path to
the future as shown in Figure 5.
EDN 22

Figure 5 Evolution of Lighting and LED positioning
However, the method of driving LEDs should also be optimized to
improve the overall efficiency.
Because they are current controlled devices, the main requirement
to drive LEDs is to control the current, which, in turn, determines
the brightness of the light emitted.
And any LED driving solution must consider the application
conditions:
Taking into account less than ideal illumination conditions, the
actual power delivered by the panel can be halved; thus considering
on average ten hours of sunlight, the energy produced by the panel
during the day can be estimated as around 200Wh.
This implies that the energy produced by the panel would be enough
to supply the lamp for about 16 hours, fully covering nightly lighting
needs in any season.
When the application requires a boost conversion, a solution could
be provided, for example, by the LED7707. This is an appropriate
solution for all the typical input voltages (6V, 12V and 24V
batteries), whenever the output voltage (fixed by the number of LEDs
in series) is higher than the input voltage.
Each one of the integrated six current generators can provide from
20mA up to 85mA, and the current can be simply adjusted by a
resistor (see Figure 7).
In solar LED lanterns, a typical battery voltage is 6V (ranging from
5.5V up to 7V).
VIN = 6V
VOUT
Slope Compensation OVP selection

yy Input voltage range—according to the use, different battery types Switching Frequency selection
+5V
can be chosen: from 6V batteries in the case of solar lanterns, to
12V batteries for home lighting applications, up to 24V batteries

AVCC

VIN

LX

OVSEL

SWF
SLOPE
LDO5 ROW1

mainly dedicated to streetlight solutions. Internal MOS OCP

BILIM ROW2
Rows current selection
RILIM ROW3

yy Number of LEDs and how they are connected—LEDs can be SS ROW4 3W LED string

connected in series in a single string or arranged in multiple COMP LED7707 ROW5

parallel strings. The number of LEDs in series defines the output SGND ROW6

FAULT

MODE

PGND
SYNC
DIM

EN
voltage of the conversion (typical voltage drop across a white Dimming

LED is around 3.5V), whereas the number of parallel strings Fault Faults Management Selection

indicates the total current to provide.

Enable Sync Output

yy LED current—high brightness LEDs are supplied by currents of Figure 7. 3W LED string driven by LED7707
hundreds of milliamperes (up to more than 1A).
Typically, a 15W solar panel is appropriate for solar LED lantern
Depending on the battery voltage and the number of LEDs solutions. Under the same hypothesis of sunlight duration and
connected in series in one string, a buck or a boost conversion can illumination conditions mentioned in the previous section, the
be the most suitable solution. energy produced by a 15W panel is 75Wh.
STMicroelectronics offers, among a wide range of products for LED Therefore, the energy produced by the 15W panel is by far
driving, both a buck converter (L6902D, [2]) and a boost converter sufficient, assuring more than 20 hours of autonomy.
(LED7707, [3]) dedicated to LED driving. Both devices feature
techniques to control the current and a high efficiency conversion. 6. ST Solutions and further integration
The below block schematic (see Figure 8) shows how the ST system
The L6902D is a step-down switching regulator mainly used for
solution approaches the all above requirements in one single system
home and street lighting.
from the MPPT algorithm to the battery charger related profile, the
24V
sense
resistor
350mA LED driving, and every kind of system protection towards the Panel,
IN VOUT
the battery and even the lamp, making the entire architecture safe.
L6902 100mV

CS+
CS- 6W LED string
FB

Figure 6. 6W LED string driven by L6902D

Figure 6 shows a 6W LED string driven by L6902D.
An example of the L6902D used as an LED driver is in street
lighting. A 40W solar panel is suitable for this application, and a
12W lamp, considering the high efficiency of LEDs, could be more
than enough for illumination of local roads in rural areas.
In a typical application, a 12W lamp can be realized by two strings Figure 8 Block Schematic of the ST Solar Battery Charger LED
of 6 LEDs (connecting in series two 3W LED modules), each one Street Light Solution
driven by one L6902D.

Designing with LEDs EDN 23

A real implementation of the above block schematic has also been to further improve the digital function integration, delivering an
realized to prove the technology and system performance (see below optimum battery charger profile on top of the already achieved
Figure 9). This reference design uses state of the art microcontroller functions with the previous integration levels.
technology as well as best in class analog drivers and discrete
The resulting device main features and characteristics are listed in
power components—all from the ST portfolio of technologies and
Table 1.
products.
Table 1 Integrated Solar Powered Battery Charger
Solar lanterns Streetlights
PV panel
Peak power 24W 50-100W
Open circuit 17V 34V
@ MPPT 12V 24V
Lead acid battery
Charge level 6V 13.8V
Energy storage 4.5Ah 20Ah
LED
Power 5-10W 20-40W

Below is a list of the resulting IC features:

Figure 9 Real Implementation of the ST Solar Battery Charger
Led Street Light Solution
For this demanding industrial application, the challenges of
innovation and miniaturization that a leading semiconductor
company has to achieve are even more stringent than usual
requirements for the power conversion segment. To remain at the
leading edge of technology and solutions, ST is engaged in an
integration path whose roadmap is seen in Figure 10.
yy Power switch and synchronous rectifier integrated inside IC
yy Operating voltage range 3V to 38V
yy Perturbation and observation method for MPPT
yy Constant current for bulk battery charge and constant voltage
for floating charge
yy Battery status indication and automatic day/night detection via
solar panel
yy Input pin for battery temperature feedback
yy Protection for solar cell short circuit, battery over voltage and
over temperature

Figure 10. Solar Roadmap Integration Path
Since most of the challenges described above were related to the
MPPT and Battery charging profile setting, the first and most
important activity in the integration path has engaged research and
development resources in first integrating the panel needs and then
the battery charging needs, leaving the LED driving as a separate
function. This will give the final system designer the flexibility to
decide on configurations and other system parameters that could not
be made as flexible in an integrated design.
The first and most simple step taken on the integration path
has been towards efficiency improvement right out of the panel,
making the widely used Bypass schottky diode, an intelligent one
integrating best-in-class analog IC technology with ultra low Rdson
MOSFETs, to reduce by over one third the losses at that portion of
the conversion.
The next and most important integration level was implemented
when achieving the MPPT function together with a high power
DCDC boost converter into a single silicon delivering over 80W
with 92% efficiency. The last portion of the integration has been
7. Conclusions
The challenge of providing light in rural areas can be achieved by
producing energy locally. Photovoltaic systems make it possible
to exploit a readily available energy source while at the same time
respecting the environment.
The storage of the energy into batteries overcomes the intrinsic
discontinuity of the solar energy.
Light emitting diodes, ever more present in lighting solutions, seem
the most appropriate choice for energy saving thanks to their high
luminous efficiency.
Moreover, the availability of different LED drivers, with buck or
boost configuration, provides flexibility in lighting systems design
and high efficient power conversion solutions.
ST’s technologies enable a ready-to-go discrete implementation of
all the above features as well as an integration path towards a one
or two chip implementation, achieving miniaturization and the best
efficiency targets.
Bibliography:
[1] “Imaging India. Ideas for the new century”, pages 258, 454, 458
—Nandan Nilekani, Penguin Books India.
[2] Low Voltage LED Driver Using L6920D, L4971 and L6902D—
Application Note AN1941, STMicroelectronics.
[3] 6 rows—85 mA LEDs driver with boost converter for LCD panels
backlight—Application Note AN2810, STMicroelectronics.

Designing with LEDs EDN 24

On the Useful Lifetime of
LED Lighting Systems
by Geof Potter, Power Technologist, Texas Instruments, November 2009
Abstract
Although they are relatively new players in industrial and residential
lighting, high-brightness light-emitting diodes (HBLED) are actively
being characterized for field reliability prediction by many of the Diffuser
major manufacturers of lighting products. That process requires FRONT VIEW
(14 LEDs )

collection of meaningful long- term experience data upon which

credible estimates of expected operating life under actual field Diffuser
conditions can be published. Without credible estimates, field
acceptance of LED luminaires will suffer.
CHASSIS

This paper reviews what is known about the lifetime of LED’s in the ( SIDE VIEW )
LED
field and discusses how to estimate the operating life of LED-based MOUNTED
&
lighting systems by including reliability factors associated with LED CONNECTED

drivers (sometimes called ballasts) that provide controlled voltage or

DBC
current to power the devices. Coordination of LED and driver life substrate
expectancy is necessary for commercial lighting products if they are DRIVER CIRCUIT
to fulfill the “long-life” promise of solid-state lighting. BOARD

Users of LED lighting products are legitimately concerned about

COVER ALUMINUM
useful life, because the price of HBLED based luminaires has not DRIVER OUTPUT
ABOUT 20 WATTS
ELECTROLYTIC
CAPACITORS (e-caps)
yet seen the cost benefits of sustained mass production on the scale DIAGRAM OF HYPOTHETICAL LUMINARIE PRODUCT SHOWING THE ENTIRE ASSEMBY INCLUDING
of incandescent or florescent lighting devices. Reasonably accurate LEDs, DRIVER BOARD, CHASSIS AND DIFFUSER

prediction of time-to-failure is only practical where substantial Fig. 1

information about the design, construction and environment for
actual applications of LED-based products is available. Norms have
not yet been established in the minds of the public by high volume
manufacturers based on consistent field performance. Consequently,
this paper approaches LED lighting longevity by discussing several
key factors that influence useful life in order to identify ranges of
probable lifetimes based on estimation of application conditions and
the general nature of the product design.
It is important to distinguish lifetime from reliability as applied to
luminaires. Lifetime is the amount of useful operating time available
from the vast majority of lighting product units, under prescribed
conditions, exclusive of random failures. Reliability is a term used
to describe how often such products fail randomly (i.e. exclusive of
infant mortality or wear-out conditions). This document aims to
present factors influencing lifetime. Poor reliability resulting from
design or construction of luminaires can adversely affect the product
lifetime by distorting the impact of premature failures and the
apparent onset of wear-out.
HBLED Lifetime
One advantage that LED devices have over incandescent lamps is
their lack of “hard” failures. LEDs display a wear-out mechanism
rather that a propensity for catastrophic failure. In fact, the lighting
industry has standardized on a definition of failure that declares
“end-of-life” to be the point at which LED output (Lumens) has
diminished by 30%. The so-called L70 point, is the point at which
most people can perceive a loss in intensity and is often used as the
measurement standard for failure.
LED output degradation occurs, in part, as a result of reduced
transmission of light through the LED package, that is, a lens and
the encapsulant that reduce internal reflections within the assembly.
There is also a reduction in actual light generation within the die
due to an accumulation of defects in the lattice occurring with use.
In devices that have external phosphor coatings to adjust color, there
is also a loss in phosphorescence that occurs with use.

A hypothetical LED Luminaire

For reference, in Fig. 1, below, is a simple diagram of an assembly
consisting of a light diffuser, a box (chassis), an LED mounting
/ cooling plate and a circuit board containing components of a
constant current driver that provides regulated power to the LED
array. Of course this is a representation of some basic elements of
a luminaire, not a real device, and is intended to emphasize the
point that such a product has more parts to consider from a lifetime
viewpoint than just the LED devices themselves.

Fig. 2 Side-view (cut-away) of mounted HBLED with associated

structure [source: CREE ]
The net effect of both degradation modes is presented as a set of

Designing with LEDs EDN 25

life curves vs operating hours. often for each of several ambient
temperatures and drive levels. The curves, published by LED
manufacturers, apply only to the diodes themselves, and so, do
not provide a complete estimate of the useful life of a complete
luminaire product.
Of course, LED life curves are not meant to represent a single
device, but rather, the center point of a bell curve that represents
lifetime in a large LED population. They predict the point in
time by which 50% of the devices have “failed”, or drifted out of
specification, (i.e. the “average” lifetime as identified by the term
L70).
50% Failed
Number of failures
Point Estimate of Failure Rate
Operating Hours
Typical maximum operating
point target
Fig. 3 Example of HBLED Lifetime curve.
(This is where the commonly used 50K hour life estimate comes
from)
In the chart above, the CREE XLamp XR-E average lifetime (the
bell curve peak) is plotted as a function of junction temperature.
It is apparent from the chart that device junction temperature is a
major determining factor in LED longevity. It is also apparent from
the separation of forward current curves that drive level (forward
current) is another important factor. At lower current levels, defects
in the crystal lattice develop more slowly, so the fraction of current
that generates photons is higher compared to the fraction that
represents non-radiation current.
Building reliable luminaires is very much about controlling the
temperature of LED devices and maximizing the longevity of
their drivers. Generally, LED temperature control involves heat
management through passive dispersion techniques. Heat generated
by the diodes must be conducted to a location where it can be
effectively coupled to ambient air. Active coupling, by artificially
increasing airflow past a coupling medium (heat sink) typically by
means of a fan, is more effective than purely passive techniques, but
air movers add considerable cost to the lighting product and, more
importantly, with their own failure rate they are likely to degrade
system reliability more than they improve it. Of course, fans are also
problematic for noise and power reasons.
Designing with LeDs
Most often, LED devices for lighting applications are mounted onto
some sort of thermally conducting structure which is then exposed
to ambient air for heat dispersion. Aluminum panels with insulated
copper conductors (DBC technology) are particularly useful for large
numbers of devices where 100 watts or more is to be dissipated,
although panels for OEM luminaire applications are available with
fewer than 10 LED units mounted.
Average LED lifetime, from the chart above, is typically 50,000
hours provided forward current, ambient temperature and junction
temperature are maintained as specified, meaning temperatures
below 105 ºC in typical environments. Junction temperature limits
are typically set just above 100ºC to maintain compatibility with
substrate materials and packaging as well as to improve device
lifetime.
LED Drivers
Having briefly described factors influencing the life of LED devices,
it is now appropriate to examine a similar aspect of power supplies
(drivers) that power them to gain a more complete picture of overall
lighting product life.
LED drivers that provide controlled current for LED luminaires
share several things in common. Regardless of electronic topology,
drivers use soldered interconnections, solid-state switching devices
and integrated circuits for control and monitoring. There are also
passive components (fuses, resistors, capacitors, connectors, wire,
etc.) involved in driver construction. All of these components
have failure rates that contribute to the overall reliability picture.
In particular, those with significant wear-out mechanisms will
contribute disproportionately, and those without contribute almost
nothing. There are a few elements used in most drivers that display
wear-out mechanisms that are significant relative to the lifetime of
LED’s themselves.
Non-Solid Electrolytic capacitors are infamous (not always
justifiably) as lifetime limiters in many types of electronic apparatus,
and are often cited as the limiting factor in LED driver designs.
There are, however, other sources of failure that can be significant,
sometimes more significant than “e-caps”.
Soldered interconnections can also display early wear-out,
particularly in thermally dynamic situations. Because solder joints
are ubiquitous in electronic subsystems and because they are
manufactured under a wide variety of quality and process controls
ranging from excellent to non-existent, it is both important to deal
with interconnect reliability and difficult to predict it.
Optical Coupling Devices (Optocouplers) often used in current
(or voltage) regulation feedback paths, have a wear-out mechanism
similar to high brightness LED’s. Because they contain an LED
light source they are subject to light intensity degradation over time
resulting in a reduction in Ctr, or current transfer ratio. If sufficient
margins have not been provided in the driver design to allow for
a gain reduction due to Ctr drift, loop gain can be compromised
resulting in operational failure.
Generally, other active and passive components mentioned above
have field failure rates significantly lower than e-caps, solder
joints and opto-couplers. Connectors, for example have long-term
oxidation issues, especially when used in low voltage circuits (which
is not generally the case in LED drivers), some types of resistors can
have very long-term drift characteristics, and so forth.
EDN 26
Non-solid Electrolytic Capacitors
Because they are required to operate at low “line” frequencies
of 50 or 60 Hz, drivers must have stored energy to fill-in gaps
between cycles where the instantaneous input voltage is too low to
operate the converter. Capacitors with relatively large CV product
(capacitance x voltage rating) to provide cycle-by-cycle ride-through
are necessary, resulting in little choice for designers other than to
use aluminum electrolytic capacitors, either on the input side or on
the isolated output side of the driver.
Non-solid e-caps contain a liquid or gel electrolyte that permits
migration of electric current across an internal insulator placed
between electrodes. These devices are carefully sealed to prevent
loss of electrolyte through evaporation, but in smaller sizes, a “dry-
out” phenomenon can occur with time and elevated temperature.
Larger sized e-caps seldom actually “dry-out” but electrochemical
degradation leads to deterioration of important electrical
parameters such as capacitance and electrical leakage, resulting
in gradual failure. Heating due to I2R loss (ripple current2 x ESR)
contributes to internal temperature rise, as does conduction into
the capacitor from a high external ambient, but radiation from a
nearby high temperature object (a transformer, or even other e-caps
with high ripple current) can also contribute to core temperature
rise and should be considered when designing driver layout.
Thermal insulation placed between the hot object and any nearby
electrolytic capacitors will improve longevity by lowering core
temperature.
Below is a graphical representation, provided by VISHAY, showing
principal trends of parameters associated with electrolytic capacitors
at constant temperature. It is helpful because it shows concisely
the relative expected lifetime of several grades of e-cap products:
Standard, Long Life, Extra Long Life and Solid Aluminum
types. Note that the curves are calibrated at relatively high
temperatures and have three scales to show the effect of lower
temperature operation for each capacitor type. Useful life increases
as temperatures are reduced according to the famous Arrhenius
relationship between temperature and the rate of chemical processes.
Designing with LEDs
Life doubles for each 10ºK decrease in process (capacitor core)
temperature.
A key point to note is that the expected life of standard grade e-caps,
operating at temperatures that are not atypical for passively cooled
commercial and consumer products (65ºC), are in the order of 8000
hours (about 11 months ). Of course, many factors affect actual
lifetime (on-off cycles, ripple current, etc.) of LED drivers, but it is
clear from this chart that an LED based lighting product using a
driver that contains standard grade e-caps is much more likely to
be limited in lifetime by those e-caps than by LEDs.
An empirically derived formula, based on extensive measurements
over long periods of time, used by Cornell Dubilier and other
capacitor manufacturers, includes operating DC voltage in addition
to temperature to provide a more precise estimate of e-cap life.
L=Life in hours
Lb=Life at rated voltage and temperature
Mv=4.3 - 3.3(Va/Vr) Va=applied voltage
L=Lb·Mv·2((Tm-Tc)/10) Vr=rated voltage
Tm=Rated temperature
Tc=Operating core temperature
Values for Lb, Tc and Vr are published by capacitor manufacturers
and identify various grades of e-caps such as 85°C, 105°C, etc.
Notice that 10°C increments in core temperature rise will adjust life
by a factor of 2.
Core temperature, one critical factor in capacitor life, is determined
by a combination of internal power loss I2R (from ripple current
and ESR), external ambient temperature and airflow that tends to
lower the effective thermal resistance of the capacitor body. When
e-caps are used in low ripple applications, ambient temperature often
provides a close approximation to core temperature.
Considering the concept of failure rate (random failures occurring
during useful life of the capacitor) could also be used when
estimating e-cap life. Capacitor reliability can be compromised by
stressful operating conditions leading to premature failures not
Fig. 4 Lifetime curves for
various grades of non-solid
electrolytic capacitors
(Source: Vishay)
EDN 27
associated with wear-out. Again, Cornell Dubilier estimates e-cap
failure rates using a formula that considers temperature, voltage and
capacitor physical size (represented by capacitance). This analysis is
useful when a bank of capacitors is used to increase energy storage.
Fundamentally, reliability decreases with increasing energy storage.
To estimate random failure rate:
λ=Failure rate in FIT
λ=4·105·N·Va3·C0.5·2(Tc-Tm)/10/Lb·Vr2 N=Number of capacitors
C=Capacitance in farads
Failure rate, λ, defines the # of units failing per unit time. One FIT
equates to one random failure in 109 unit hours.
For example: assume 2 x 100uf / 450 Vdc 2000 hour standard
85°C e-caps, running in a 70°C environment, with
80% of their rated voltage applied and full rated
ripple current…..
λexample 1= 325 FIT or 3.25 x 10 -7 failures/hour which equates
to roughly one random failure due to e-caps in 350 years.
If larger capacitors were used in the design and their voltage and
current stress was higher:
assume 2 x 1000 uf/450 volt 2000 hour standard
85°C e-caps, running in an 80°C environment at
rated voltage and ripple current…..
λexample 2= 4025 FIT or 4.02 x 10 -6 failures/hour or roughly
one random failure in 28 years.
It appears from these two examples of e-caps are unlikely to cause
driver failure as long as operating conditions remain within the
e-cap ratings. However, the projected lifetime due to wear-out for the
same examples yields significantly different answers:
L example 1=Lb·Mv·2((Tm-Tc)/10)=9320 hours or about 13 months.
Lexample 2=2828 hours or about 4 months.
Example 1 describes a design scenario for low power luminaires that
might be found in the field. Voltage and temperature de-rating has
been applied, but the wrong capacitor type was chosen, Example 2 is
much less reasonable due to design stress. Changing the parameters
of example 1 by stipulating that 8000 hour, 105°C capacitors be
used would improve the life result to 150,000 hours, eliminating
impact by e-caps on the total product life.
LED lighting products present a wide variety of operational
conditions including high density configurations (“Edison” style
bulb replacements, for example) where 60°C operating temperatures
are not atypical. There are also examples of much more benign
situations where a driver assembly is located in a separate
compartment and is not subject to direct heating by the LED array
being powered. While there are many luminaire designs that do
not stress e-caps enough to significantly reduce their lifetime, and
there are e-caps that are much more suitable for high temperature
applications than the one cited in the example above, the message
is that if longevity is important (as in street lights, for example),
careful attention to e-cap operating temperature and voltage ratings
is critical, or the LED driver can substantially reduce the effective
life of the system.
Soldered Interconnections
Another area of potential field failure in typical LED driver
assemblies, for the purposes of this discussion, is a broad
spectrum of failure mechanisms that are labeled simply as
Soldered Interconnection failures. They are especially significant
in assessment of LED driver reliability because LED drivers are
Designing with LEDs
built by a wide range of companies with widely varying quality
levels associated with their assembly and testing procedures. A
user purchasing an LED luminaire would expect the product to
perform normally over a long period of time, especially if that user
is familiar with lifetime numbers published by LED manufacturers
(commonly 50,000 hours). In fact, the luminaire lifetime is much
more likely to be limited by the life of the driver and may actually
be substantially less than 50,000 hours. This is particularly true if
the particular driver circuit design stresses components (e-caps, for
example) significantly or the manufacturing process used to build
it is not equipped to guarantee the physical integrity of its soldered
assemblies.
Traditional failure rate based analysis of drivers accomplished by
counting components and assigning failure rates from established
tables (MIL HBK 217-x) is not an effective solution to the need for
lifetime information, mainly because of the difficulty extrapolating
all the rate information from published conditions to actual
conditions. Also, the published failure rates assume that design
deficiencies are not involved. That is, failures are assumed to be
random and constant, i.e. the flat bottom of the famous “bathtub
curve”.
Another method for predicting expected lifetime of luminaires
(including drivers) is more meaningful, but much more time
consuming and expensive. One can build large numbers of the
products and subject them to accelerated stress testing to derive a
so-called “activation energy” exponent for the entire assembly and
use that in a failure rate calculation. This usually takes 6 months or
longer, depending on the number of samples, and would be useful
only for the tested configuration or close facsimiles.
Possibly the most attractive scheme for predicting the life of LED
systems is to apply a “physics of failure” approach such as that
proposed (and practiced) by CALCE at the University of Maryland.
This approach is applied in three phases and is aimed at answering
the question “how much life is left in this device / product /
assembly / etc. When performed on a new product design, it can
estimate useful lifetime. The process begins with an analysis of
degradation in the assembly resulting from storage and / or use.
Physics-of-failure models are then applied to the degradation modes
to define test conditions that produce similar degradation over a
much shorter time. Accelerated tests are then performed (on a
smaller number of units) and the results are analyzed to determine
if the test item(s) have enough life left to meet expectations. Failure
models apply to all aspects of a product design including electrical
stress. With well designed drivers, electrical stress is managed to the
point where it is secondary to other causes of wear-out. Non-solid
electrolytic capacitor wear-out has already been discussed.
Often, when a physics-of-failure approach is used on electronic
assemblies, it identifies soldered interconnections as a key source
of degradation. Connectors are also frequently identified as failure
prone, but typical LED drivers have few, if any, connectors so for the
purposes of this discussion they are ignored.
Fig. 5, below, shows a graphical depiction of relative instantaneous
failure rates for solder joints compared to typical electronic
components. These curves represent the right hand side of those
familiar bath tub curves that reliability estimators commonly use. It
is assumed that the left hand side, the so-called “infant mortality”
characteristic, has been eliminated by any of several manufacturing
means prior to shipment of the lighting product. Of course, for
any given electronic assembly, the curves will vary considerably
due to environmental conditions, operational duty cycle and the
quality of assembly processes involved. Note that individual solder
joints are thought to have lower failure rates than typical electronic
EDN 28
components during normal life but begin to “wear-out” earlier due
to structural breakdown induced by normal operational fatigue. In
the case of solder joints that are subjected to high thermal fatigue
(repeated heating/cooling cycles), wear-out begins significantly
earlier. Also note that large numbers of solder joints raise the net
failure curve and should be minimized for driver longevity sake.
Failures-in-Time (FITs) are the reciprocal of instantaneous failure
rate and represent the number of failures per 109 unit operating
hours. For example: if there are 100 joints in the driver assembly,
operating under what are here called “High Stress” conditions, Fig.
5 shows that at the point where the LEDs are reaching end-of-life
(50KHrs) the system solder joints are already in wear out mode with
a failure every 2 weeks.
FIT rate=2 x 106 failures in 109 hours (from Fig. 5 below)
or a Mean-Time-To Failure of 500 hours=about 3 weeks!!
The significant point is the range of wear out time involved. Fig. 5
also shows a substantially longer time between failures (38 Years) for
those same joints if the stress regime is more normal.
FIG. 5 Comparative Failure Rates for Solder Joints
Optical Couplers
Opto-couplers used in isolated converters in output feedback
paths have wear-out mechanisms which are time and temperature
dependent. Degradation occurs in the device transfer coefficient
(Ctr), the amount of current flowing in the optical transistor
compared to the LED drive current on the opposite side of the
isolation barrier. Driver designs will consider sufficient margin for
Ctr to degrade before problems with feedback loop gain become
a source of failure. Typical curves for the lifetime of low cost opto
couplers are shown in Fig. 6 below.
A continuous decrease in quantum efficiency of the diode half of
common opto-couplers is the main cause of transfer coefficient
degradation over time. Junction temperature and forward current
density influence diode output degradation rates. There is also
degradation in light transmission through compounds used to
couple diodes and phototransistor devices resulting in further loss of
Ctr as well as transistor detector degradation. The latter are usually
minor factors. It is possible to encounter coupler degradation by
moisture ingress through plastic molding compounds, but that is
considered a design issue and not an inherent lifetime determining
factor in typical lighting systems.
By running opto coupler circuits with low forward current in
the LED half of the device, useful life can be extended enough
to remove these parts as potential constraints on luminaire
driver lifetime.
Designing with LEDs
There have been advancements in material technologies associated
with optical isolators and couplers in recent years that make newer
couplers more stable and reliable, but they also have increased cost
making improved parts less attractive for low-cost lighting products.
Life defined as 50% loss in C tr
4N32 devices
KHrs
50 KHrs
Diode Current (ma)
Fig. 6 Example Life Curve for common low cost Opto Coupler
Other Components in LED Driver assemblies
Semiconductor components found in typical drivers are not subject
to wear-out within the lifetime range of non-solid electrolytic
capacitors, variable quality soldered interconnections and low-cost
opto-couplers. When solid-state devices are operated with reasonable
thermal, voltage and current margins (relative to their specified
working limits) their lifetimes far exceed the devices discussed above.
The same is true of passive components such as small capacitors,
resistors, wound inductors, and miscellaneous mechanical elements.
Published FIT rates for these “other” components in a driver would
be multiplied by the number of pieces per driver and added to yield
a composite picture of their random failure rate, λother.
Typical FIT rates for semiconductors are ~ 5, resistors, small ceramic
capacitors, signal diodes etc. are <1, so that λother would add up to
perhaps 20 FIT for a small single stage LED driver assuming good
design practice has been followed throughout.
Summary and Recommendations
When users buy LED lighting devices, they naturally ask the
question “how long will this product last?” This question is difficult
to answer precisely, so the best response is: “it depends!”
In fact, the lifetime of LED luminaires does depend, substantially,
on several important factors:
1. Operating temperature of the LED’s, themselves
2. Operating temperature of any power supply (driver) that provides
current to the LEDs
3. Types of components used in the driver design
4. Number of components and soldered interconnections in the
driver and LED array
5. Power cycle rate
6. Environmental conditions including temperature, and
temperature variations
EDN 29
Assuming there are no special conditions to consider such as a Clearly this luminaire has the potential to last a very long time from
corrosive atmosphere or unusual moisture situations, the lifetime the standpoint of random failures. Examining the life based on wear-
of a particular luminaire design can be estimated by studying how out we see a limited life impact from the LED driver.
it might fail and undertaking targeted tests to accelerate the likely
types of failure. From results of accelerated testing, an “activation Life estimate for LED array 50,000 hr
energy” coefficient can be determined and mean-time-to failure Life estimate for e-caps 425,000 hr
Life estimate for solder joints 33,000 hr (based on 3000 FIT, 10 %
calculated from a classic formula. failures because solder joints
are in wear-out period)
A more practical approach, for a typical user dealing with a smaller
number of luminaires, would be to consider: For discussion purposes, the above analysis shows that our
1. Is the product going to be used in an elevated ambient hypothetical luminaire will likely come close to meeting the 50,000
temperature or in a confined space where internal heating may (6 year) lifetime generally advertised by HB LED manufacturers.
cause a substantial temperature rise? There may be a small reduction in life resulting from solder joint
wear-out before the 6 year target is reached, but that estimate
2. If the driver contains electrolytic capacitors, are they rated for would vary widely depending on joint quality and environmental
high temperature (105°C). (Low cost capacitors rated at 85°C can conditions. Our driver designer used “safer” e-caps for this
reduce the useful life of the product significantly). application causing them to have virtually no impact on luminaire
3. Do e-caps have reasonable operating voltage de-rating under life.
expected conditions? 80% is considered reasonable, lower is
better !! References
[1] S.G. Parler, Jr. “Deriving Life Multipliers for Aluminum Electrolytic
4. Does the driver contain a large number of components? A Capacitors”
typical 15 watt single output, driver might contain 30 to 40 Cornell Dubilier Corp. IEEE Power Electronics Society
miscellaneous components yielding a failure rate on the order Newsletter, Vol 16. No.1. Feb 2004, pp 11–12
of 20–30. (Control IC’s raise the FIT rate much more than
resistors, etc.) [2] S.G. Parler, Jr. “Reliability of CDE Aluminum Electrolytic
Capacitors” Cornell Dublier Corp. Whitepaper
5. Higher component count means more solder joints to contribute
to FIT rate and reduced wear-out time. [3] Nichicon, “General Description of Aluminum Electrolytic Capacitors”
Nichicon Technical Notes Cat. .8101E
6. Is the luminaire subject to significant thermal excursion such as
natural cooling outdoor operation, frequent power cycling, hot [4] J. Ben Hadj Slama, H. Helali, A.Lahyani, K.Louati, P.Venet and
or cold air blast, etc.? G.Rojat “Study and Modelling of Optocoupers Ageing”
J. Automation & Systems Engineering, 2008
7. Does the regulation feedback path contain an optical coupler? If
so is it rated for the application conditions? (Low cost opto’s can
be a problem.)
Returning to our hypothetical luminaire diagrammed in Fig. 1, we
can make some assumptions about it’s design and application in
order to evaluate it’s potential useful lifetime.
yy Operating environment 50°C.
yy 14 HBLED devices on an aluminum heat spreader, 700 ma.
100°C junction temp.
yy 2 “long life” (105°C rated) electrolytic capacitors operating at
80% of rated voltage
yy 130 PCB solder joints in the driver assembly and LED array
yy 1 opto-coupler in current control feedback path. Low FIT rate as
applied.
yy 65 assorted passive components and an integrated power control
chip.
Applying typical FIT rates for random failures:
LED 14 x 1.1=15.4
e-cap 2 x 3.5 =7
Solder Joints 130 x 0.2=26
Optocoupler 1 x nil =0
Control IC 1 x 5=5
Misc comp 65 x 0.3=20

FIT rate 73.4 failures in 109 unit hours

MTBF >13 million hours
Time to 1% failures ~ 136,000 hours

Designing with LEDs EDN 30