An Approach of LED Lamp System Lifetime Prediction

X. P. Li, L. Chen, M.Chen

Philips Lighting, Shanghai, China

(xiupeng.li@philips.com, li.chen@philisp.com, m.chen@philips.com )

Abstract – Different from the MTTF/MTBF widely used to feasible for traditional lamp since their lifetime is only
describe the reliability of electronic products, lifetime is used thousands hours which could be tested in several months.
for the LED lamp. The lifetime is defined as a time when
50% of lamp’s lumen output is less than 70% of initial value, However, LED lamp’s lifetime is more than 25,000 hours;
either by lumen decay of LED light source or catastrophic the lamp manufacturer cannot put LED lamps into the life
failure of electronic component. In other words, the
test for 25,000 hours or more before the LED lamps
reliability of LED lamp consists of not only the
MTTF/MTBF of electronic components, but also other launch into market. The traditional standard to predict
failure modes in the whole lifetime, ie electronic component LED is obviously not acceptable, and a new reliability
or lumen degradation. However, the currently available methodology has to be developed for prediction of LED
reliability standards for LED lamp only focus on lumen lifetime.
maintenance, while cannot give any method on how to
measure the LED lamps’ lifetime. With reviewing the system The paper will investigate the structure of LED lamp and
structure and failure modes of LED lamp, the paper develop a methodology of reliability demonstration test of
proposes a methodology for the reliability of LED lamp. LED lamp based on the traditional lamp and the new
features of LED Lamp system.
Keywords – Lifetime, B50L70, LED Lamp, Lumen
.
Maintenance, Thermal Shock
II. METHODOLOGY
I. INTRODUCTION To address the reliability of LED lamp, it is necessary to
understand its structure and build a reliability model for it.
A light-emitting diode (LED), widely used as Normally a LED Lamp includes the four subsystems:
indicator lamps in many devices, is increasingly used for
lighting in last few years and increasing dramatically with 1. LED as light source
energy saving requirement in global. Different from 2. Electronic driver which provides power to LED
traditional lighting system, LED requires the profoundly lighting source
lighting technologies, including physics, chemical, photo
3. Mechanical housing used for thermal dissipation,
electronics and imaging etc. Due to rather easy electronic isolation and final installation
technologies involved in downstream products, more and 4. Optical lens or bulb, fulfill the optic
more manufacturer join into the lighting market with requirements, e.g. color over angle, beam angle.
limited knowledge of LED and Lighting application,
resulting in a lot of reliability and quality issues. A lot of The reliability of the whole LED Lamp could be
lamps failed with significant lumen degradation or dead in illustrated in shown in Fig1.
4 or 5 months after installation because of poor product
design in LED, electronic and thermal etc.
Fig.1: LED Lamp System Reliability Diagram
Therefore, the requirement of LED lighting system
Optical
standards is very importance for the LED lamp market, in
which the reliability standard requirement is the most
urgent one.
LED LED Driver
Approved methods for traditional light life testing are Light Lamp
called as measuring and reporting rate lamp life. These Source

methods are published by the IES, e.g. LM-49-01 for

incandescent lamp [1], LM-65-01 for compact fluorescent
Mechani
lamp [2], in which the lifetime is used as the reliability cal
indicator of lamp, called B50L70, which means 50% (B50)
of lamp’s lumen output is less than 70%(L70) of initial
value. The lamp manufacturer normally takes a certain
lamps into account of life test, and records the time of
50% lamp failing L70 as the lifetime of lamp. It is

978-1-4577-0628-8/11/$26.00 ©2011
Authorized licensed use limited to: IEEESimon Bolivar (Colombia). Downloaded
Universidad 110 on March 14,2024 at 03:19:14 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the 2011 IEEE ICQR

The reliability of whole LED Lamp can be described as The manufacturer should provide the raw data in 6000hrs
follows: and predict the time of L70/50, indicated as
RLED  system RLED u RDriver u ROptical u RMechanical (1) RL70/50.However, 6000hrs life test of lamp last almost one
year, which is not applicable in current LED lamp
development cycle.
Each subsystem has multi failure modes, in which
each failure mode has its own failure distribution. A
The thermal shock test is only 5 cycles per IEC/PAS
single subsystem’s reliability can be described as follows:
62612, which does not simulate lamp’s usage profile in
Rsubsystem1 R failuremod e1 u R failuremod e2 u R failuremod e... the whole life cycle and thousands cycles of thermal
n (2) stresses during switch on/off in 15 years lifetime. Rapid-
¦R failur mod ei (i 1....n) cycle stress test also limits stresses to the LED light
i 1 source and electronic components, which could not assess
For example, a failure of LED Lamp could be the reliability of LED lamp.
resulted from 30% lumen decay caused by LED die and
package material or the catastrophic failure caused by In order to evaluate and assess the reliability of LED
LED die crack, breakdown. The reliability of LED light lamp, a systematic reliability approach is required to
source is modeled per equation (2) as below: identify the major failure modes of each subsystem, then
RLED Rlumen _ decay u Rcatastrophic _ failure (3) build a system reliability based on equation(2).

Therefore, in order to evaluate the system reliability, it is In Fig 1, a LED lamp consists of 4 subsystem and the
necessary to understand both the reliability of each more than 30 total failure modes of system. Major failure
subsystem in the lifetime and the failure modes of each modes for each subsystem of LED lamp are listed in table
subsystem. 2:

In September 2008, the IES issued Measuring Lumen Table 2: Failure Modes and Failure Distribution
Subsystem Failure Mode Typical Failure rate
Maintenance (MLM) of LED Light source and IES LM- distribution
80-08[3]. IEC also issued a PAS version of 62612[4] LED Lumen Maintenance Lognormal(ȕ>3)
which majorly focus on the lumen maintenance of LED Depreciation
lamp and reliability tests in these standards as below: Color Shift Over Lifetime
Catastrophic Failure, wire- Weibull (ȕ>1)
bond broken, die crack, et.c
Table 1: Reliability Requirement of LED Light Source Electronic Electronic Component fails Exponential (ȕ=1)
and Lamp in current standards Driver in useful life
Test item Standards Description Remark
Lumen IES LM- 6000hrs life test at 3 10 samples Solder Joint Fatigue Normal (ȕ=3.5)
Maintenan 80- different case by Energy Mechanical Plastic Housing Crack Normal (ȕ=3.5)
ce 08/Energy Temperatures : Star Optical Optical coating Normal (ȕ=3.5)
Star 55°C discoloration
85ºC Glass bulb crack
defined by manufacturer
Rapid- Energy Cycle times: 2 minutes 10 samples
Cycle Star on, 2 minutes off. Lamp by Energy
Stress Test cycled once for Star Weibull distribution is used to describe the failure rate
every two hours of over time (also called as Hazard Function) for each failure
required minimum mode, and shape parameter (ȕ) tells the character of
L70 life
Lumen IEC/PAS 6000hrs life test at 45°C Sample size
failure mode as showed in Fig 2 Bath curve.
Maintenan 62612 ambient temperatures 10 Fig.2: Bath Curve –IEC 61649[5]
ce
Rapid- IEC/PAS Cycle times: 30sec
Cycle 62612 on, 30sec off. Lamp
Stress Test cycled once for
every two hours of
required minimum
L70 life
Thermal IEC/PAS -10°C~+50°C 1hr dwell
Shock 62612 5 cycles

Table 1 showed that instead of measuring lamp life

time, the current standards call for how much a LED light
source or lamp’s lumen output decay over lifetime with With understanding the failure rate distribution of each
expecting extremely low catastrophic failure in the whole failure mode, real shape and scale parameters for each
lifetime. The 6000hrs lumen maintenance data give a failure distribution by reliability tests are found out.
good indication of lumen maintenance in LED lifetime. Failure modes are caused by different stresses, which

111 on March 14,2024 at 03:19:14 UTC from IEEE Xplore. Restrictions apply.
Authorized licensed use limited to: Universidad Simon Bolivar (Colombia). Downloaded
 Proceedings of the 2011 IEEE ICQR

could not be covered by only one reliability test. For The LM-80 data from LED manufacturing is available
example, the wear out failure of solder joint fatigue is already, so it is better to use the LM-80 data of LED light
mainly caused by the thermal cycling; the electronic source for the lumen maintenance degradation prediction
component failure in useful life is commonly resulted with exponential degradation model. A L70 prediction is
from the thermal stresses and electronic stresses in normal shown in Fig 3, based on 30pcs samples 6000 hours LM-
usage. 80 test data.

The system reliability test should include stresses for each Fig. 3:Lumen Maintenance per LM-80 data with Weibull
failure modes: plot
ReliaSoft Weibull++7 - www.ReliaSoft.com

x Lumen maintenance failure (L70) caused by led 99.000

Probability - Lognormal
Probability-Lognormal

light source and optic system degradation Data 1

Lognormal-2P

x Electronic and thermal stresses for random

RRX SRMMEDFM
F=30/S=0
Data Points
Probability Line

failure rate of electronic component and LED

light source and,
x Wear-out failure of aluminum electrolytic

Unreliability, F(t)
capacitor, the thermal cycling for solder joint 50.000

fatigue, mechanical housing crack, wire-bond

broken of LED die.
10.000

The reliability of mechanical of LED lamp is considered

5.000

as 100% in the whole lifetime for normal application

because lamp is normally installed in the socket, and the 1.000
6/9/2011
12:27:32 PM
10000.000 100000.000

mechanical stresses from vibration and external shock are P V U 
Time, (t)

rather small and neglected, which is not covered in the

paper. The distribution of L70 is lognormal, mean is 10.7437,
and standard deviation is 0.2319. This information will be
1.0 Lumen maintenance Failure (L70) used for system level reliability modeling.

IES had published a standard IES LM-80-08 in 2008, 2.0 Random failure rate of driver’s electronic
defined the methodology to measuring the lumen components
maintenance of LED light source. It is widely accepted by
LED light source manufacturing and lighting industrial. The failure rates of electronic components are well known
LEDs are tested in three kinds of temperatures (55C, 85C and several standards are already available, the mostly
and 3rd temperature defined by LED manufacturing) for used in the field are MIL-STD-217[7] and Telcordia SR-
6000 hours. Energy Star [6] also requires a 6000hrs life 332[8]. As long as components’ case temperature, current,
test of LED lamps to demonstrate lumen maintenance. voltage and power are provided, each components’ failure
Table 3 shows the lumen maintenance of lamp after rate are calculated and the total failure rate of whole
6000hrs and the prediction of L70 per exponential driver is the sum of failure rate of each component.
degradation model as listed in table 3.
Table 3:The Lumen Maintenance in 6K vs L70 n n
Minimum lumen maintenance at Maximum L70 Life Rdriver ¦R ith component ¦ exp O  i ut

end of 6000 hours (% of initial Claim (hours) i 1 i

lumens; -3% tolerance) n
 ¦ Oi ut
91.80% 25,000 exp i 1
   (4)
93.10% 30,000
If the standard is not appropriate, the testing data or field
94.10% 35,000
failure data also could be used to calculate the failure rate
94.80% 40,000 () of whole driver.
95.40% 45,000
The reliability of LED lamp is estimated as below:
95.80% 50,000
Rrandom _ failure _ LED _ Lamp exp  OT (5)
where, Ȝ is the failure rate in table 4.T is the specified
The requirement of Energy Star only focuses on the operating hour.
lumen maintenance, and it is an average value of 10 During the ALT, power on/off should be considered to
samples. Therefore it is a B50L70 for lumen maintenance. allow units under test really cool down and moisture
Moreover, lamps could not burn for 6000hrs before could be penetrated into product. 30mins off is minimum
product release which normally is only half of year. for cool down, therefore a 2.5hrs On/0.5hrs Off cycles is
recommended so that 20hrs On per day is achieved.

112 on March 14,2024 at 03:19:14 UTC from IEEE Xplore. Restrictions apply.
Authorized licensed use limited to: Universidad Simon Bolivar (Colombia). Downloaded
 Proceedings of the 2011 IEEE ICQR

III. RESULTS
3.0 Wear out failure
In a typical LED lamp system, the reliability distribution
The heat generated from LED and electronic components for each subsystem and failure modes are as follows:
is dissipated from housing passively. Because the small
form factor of LED lamp, it is more difficult for heat 1) L70: mean=10.7437, standard deviation=0.2319
dissipation, and causes the temperature of components, lognormal distribution per Fig 3.
solder joint are higher than normal electronic equipment, 2) Total Failure Rate () is 2863FIT per Telcordia
the wear out of aluminum electrolytic capacitors, fans, SR-332.
and fatigue of solder joint may have shorter lifetime. For 3) 6ROGHU-RLQW)DWLJXHȕ Ș 
example, the LED solder joint temperature is also driven
to 90°C in operating in room temperature, which has a Based on equation (13), the cumulative failure
65°C temperature change in a cycle of switch on/off in distribution and each subsystem failure distribution are
normal room temperature (25°C). It is necessary to obtained as shown in Fig. 4.
understand the solder joint fatigue failure distribution. Fig. 4: Cumulative Failure Rate of LED Lamp
Thermal shock is an effective and efficient to evaluate the
solder joint fatigue, the acceleration model of thermal
shock is Coffin-Mansion equation [9] as follows:
n
N f C0 u 'T (6)
And acceleration factor is described as below:
n
§ N use · § 'Ttest ·
AF ¨¨ ¸¸ ¨¨ ¸¸ (7)
© N test ¹ © 'Tuse ¹

Where n is the material property to thermal shock, about

2.66[10] for lead free.
In the normal user conditions (25ºC), if the major
concerns components, e.g. solder joint is around 90ºC. From the cummulaive failure disctribution curve, the
The estimated acceleration factor in a thermal shock test(- B50L70 is around 35,000hours, which is far lower the
40°C~125°C)is average L70=46,000hours.
2.66
§ 125   40 · (8) IV. CONCLUSION
AF ¨ ¸ 11.9
© 90  25 ¹
Due to longer lifetime and different reliability definition
Note: the thermal shock temperature range could be of lifetime, the lifetime of LED lamp is more complex
adjusted by manufacturer base on the user application than general electronic equipment or traditional lamps, In
conditions and the limitation the product component its lifetime, it includes minimum three types of failure
specification. modes, lumen maintenance, randomly failure in useful life
and wear-out before end of lifetime
Then the failure distribution in a thermal shock test could
be translated into the real application condition. To get a lifetime of LED lamp system (B50L70), the
reliability information for three types of failures are
§ § T ·E · needed::
exp¨  ¨
Rwear _ out _ LED _ Lamp ¸ ¸ ˄9˅
¨ ¨ K ¸ ¸
© © ¹ ¹ x A lumen maintenance data from LED light
Where ȕ is the shaped parameter, ȕ=2 is typical; Ș is the source manufacturing per IES LM-80-08
scale parameter from thermal shock test; T is the switch x A randomly failure of electronic driver per
on/off cycles in operating. external standards, or testing data
x A thermal shock data to address the wear out
failure mode in whole lifecycle.
After getting the failure distribution of lumen Reliability prediction of LED lighting system lifetime can
maintenance, random failure in useful life and wear-out be achieved as Equation (13) based on above data.
failure of on/off switch, the reliability of LED lamp
system is as follows:
ACKNOWLEDGMENT
RLED  system RL 70 / 50 u RRandom _ failure _ LED _ lamp u RWear  out
(10)

113 on March 14,2024 at 03:19:14 UTC from IEEE Xplore. Restrictions apply.
Authorized licensed use limited to: Universidad Simon Bolivar (Colombia). Downloaded
 Proceedings of the 2011 IEEE ICQR

Authors would like to thank Philips Lighting to provide

experimental data for analysis, thank Dr. Willem Van
Driel of valuable guidance in paper writing, and also
thank all reliability team members for their support in
testing and data collection.

REFERENCES

[1] LM-49-01 IESNA Approved Method for Life Testing of

Filament Lamps, Illuminating Engineering Society / 01-
Dec-2001
[2] LM-65-01 IESNA Approved Method for Life Testing of
Single-Ended Compact Fluorescent Lamps, Illuminating
Engineering Society / 01-Dec-2001
[3] IES LM-80-08 Approved Method for Measuring Lumen
Maintenance of LED Light Sources LM-80
[4] IEC 62612 PAS Self-Ballast LED-Lamps for general
lighting services – performance requirement
[5] International Standards Weibull Analysis - 61649
[6] ENERGY STAR® Program Requirements for Integral
LED Lamps
[7] MIL-STD-217 Reliability Prediction of Electronic
Equipment
[8] Telcordia SR-332 Reliability Prediction Procedure for
Electronic Equipment Issue 3, January 2011
[9] Wayne B. Nelson “Accelerated Testing: Statistical
Models, Test Plans, and Data Analysis” ISBN-13: 978-
0471522775
[10] “Thermal Cycling and Thermal Shock Failure Rate
Modeling”, R.C. Blish, IEEE, IRPS, 1997 and “An
Acceleration Model For Sn-ag-cu Solder Joint
Reliability Under Various Thermal Cycle Conditions”,
N.Pan et al., HP2005

114 on March 14,2024 at 03:19:14 UTC from IEEE Xplore. Restrictions apply.
Authorized licensed use limited to: Universidad Simon Bolivar (Colombia). Downloaded