Ansi Ies TM-35-19
Ansi Ies TM-35-19
Ansi Ies TM-35-19
TECHNICAL MEMORANDUM:
PROJEC TING LONG-TERM CHROMATICITY
COORDINATE SHIFT OF LED PACKAGES,
ARRAYS, AND MODULES
AN AMERICAN NATIONAL S TA N DARD
Approved by the /ES Standards Committee August 23, 2019 as a Transaction of the Illuminating Engineering Societ y.
All rights reserved. No part of this publ ication may be reproduced in a ny form, in any electronic retrieva l system or
otherwise, without prior written permission of the I ES.
Publ ished by the I l l u m i nating Engineering Society, 1 20 Wal l Street, New York, New York 1 0005.
I ES Standards and G u ides are developed through committee consensus and produced by the I ES Office in New York.
Careful attention is g iven to style and accuracy. If a ny errors a re noted in this document, please forward them to
Brian Liebel, Director of Standards, at sta [email protected] or the above address for verification and correction. The I ES
welcomes and u rges feedback and comments.
ISBN# 978-0-87995-082-8
DISCLAIMER
IES publications are developed through the consensus standards development process approved by the American
National Standards Institute. This process brings together volunteers representing varied viewpoints and interests to
achieve consensus on lighting recommendations. While the IES administers the process and establishes policies and
procedures to promote fairness in the development of consensus, it makes no guaranty or warranty as to the accuracy
or completeness of any information published herein.
The IES disclaims liability for any injury to persons or property or other damages of any nature whatsoever, whether
special, indirect, consequential or compensatory, directly or indirectly resulting from the publication, use of, or reliance
on this document.
In issuing and making this document available, the IES is not undertaking to render professional or other services for or on
behalf of any person or entity. Nor is the IES undertaking to perform any duty owed by any person or entity to someone
else. Anyone using this document should rely on his or her own independent judgment or, as appropriate, seek the
advice of a competent professional in determining the exercise of reasonable care in any given circumstances.
The IES has no power, nor does it undertake, to police or enforce compliance with the contents of this document.
Nor does the IES list, certify, test or inspect produc ts, designs, or installations for compliance with this document. Any
certification or statement of compliance with the requirements of this document shall not be attributable to the IES and
is solely the responsibility of the certifier or maker of the statement.
Approval of an American National Standard requires verification by ANSI that the requirements for due process,
consensus, and other criteria have been met by the standards developer.
Consensus is established when, in the judgment of the ANSI Board of Standards Review, substantial agreement has been
reached by directly and materially affected interests. Substantial agreement means much more than a simple majority,
but not necessarily unanimity. Consensus requires that all views and objections be considered, and that a concerted
effort be made toward their resolution.
The use of American National Standards is completely voluntary; their existence does not in any respect preclude anyone,
whether that person has approved the standards or not, from manufacturing, marketing, purchasing, or using products,
processes, or procedures not conforming to the standards.
The American National Standards Institute does not develop standards and will in no circumstances give an interpretation
to any American National Standard. Moreover, no person shall have the right or authority to issue and interpretation of an
American National Standard in the name of the American National Standards Institute. Requests for interpretations should
be addressed to the secretariat or sponsor whose name appears on the title page of this standard.
CAUTION NOTICE: This American National Standard may be revised at any time. The procedures of the American National
Standards Institute require that action be taken to reaffirm, revise, or withdraw this standard no later than five years from
the date of approval. Purchasers of American National Standards may receive current information on all standards by
calling or writing the American National Standards Institute.
Prepared by the IES Testing Procedures Committee
Members
C. K. Andersen K.C. Fletcher M. Kotrebai E. Radkov
J. N. Hulett
E. Bretschneider J. Hospodarsky P. McCarthy J.E. Walker
P. Elizondo G. McKee
Advisory Members
L. M. Ayers J. Frazer K. M. Liepmann A. W. Serres
T. Kawabata
M. Damle H. Kashaninejad D. P ark A.Thorseth
CONTENTS
3.4 DUT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
3. 5 LED light source . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
3.6 relative chromaticity cSu', cSv' . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
7.0 Report . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Additional Reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
Technical Memorandum: Projecting Long-Term Chromaticity Coordinate Shift of LED Packages, Arrays, and Modules
1.1 Introduction
The I l l u m inating Engineering Society (IES) has defined standard test methods to be used by the man ufacturers of LED
products to ensure that the prod ucts meet the expectations of the lighting com m u nity. One of the most i m portant of
these is ANSI/JES LM-80-15, Approved Method: Measuring Luminous Flux and Color Maintenance of LED Packages, Arrays,
and Modules. LM-80 provides a test and measurement protocol for how these LED light sou rces in order to document
cha nges in their light and chromaticity characteristics with respect to time. By a pplying the methodology of ANSI/JES
TM-21-19, Technical Memorandum: Projecting Long Term Lumen Maintenance of LED Light Sources to a m i n i m u m of 6,000
hours of LM-80-compliant test results, it is possi ble to project the l u men maintenance life (e.g., L10, L80, or L90) of LED
packages, arrays, and modu les.
I n addition to l u m inous flux, LM-80 also add resses another key photometric parameter, the chromaticity coordinates
(often cal led the "color poi nt" ) of the LEDs. The i nitia l chromaticity coordinate req u i rements for white solid
state l i g hting (SSL) l ight sources and l u m i naires are defined i n ANSl/N EMA/ANSLG 08.377, Specifications for the
Chromaticity of Solid State Lighting Products. Proper selection of LEDs used to build SSL products a l lows l u minaire
manufacturers to easily achieve a desired ANSI 08.377-defined chromaticity.
It is important to note that in some applications-such as retail and museum lighting-a characterization of
a ntici pated chromaticity shift of SSL products over their usable l ifetime is very im portant to the user. Further, the
chromaticity coordinates of some light sou rces can change significantly from their initial val ues over time. This
docu ment recommends a method of projecting the shift i n c h romaticity coordi nates of LED packages, a rrays, and
modules using data obtained d u ring LM-80 testing.
1.2 Scope
This document provides recom mendations for projecting long-term chromaticity coordinate stabil ity of LED l ig ht
sources using data obta ined per ANSI/JES LM-80-75, Approved Method: Measuring Luminous Flux and Color Maintenance
of LED Packages, Arrays, and Modules.
This method shall only be used for phosphor converted LEDs with i nitial chromaticity coordi nates that fal l within the
ranges defined for solid state l i g hting products in ANSl/NEMA/ANSLG 08.377-201 7.
This method sha l l not be used to predict chromaticity sh ifts in excess of �u'v' = 0.01 0.
1
ANSl/IES TM-35-1 9
3.0 Definitions
3.4 OUT
Device u nder test; the LED l ig ht source defined i n Section 3.5.
Since the changes in OUT chromaticity coordinates between time i nterva l s is typica lly very small (i.e., Llu'v' < 0.001),
this method shall only be used with LM-80 datasets in which the OUT chromaticity data a re reported to at least 5
decimal places. When the average change in relative chromaticity coordinates between time intervals is much less
than 0.00005, the predicted trend for chromaticity shift may exceed the actual trend and therefore could be i na ccurate.
All calculations sha l l be made using at least 5 significant dig its.
2
Technical Memorandum: Projecting Long-Term Chromaticity Coordinate Shift of LED Packages, Arrays, and Modules
The preferred sample size is a m i n i m u m of 30 u nits. As with the measurement interval, a change in sample size leads
to a change in u ncertainty and therefore results i n a different extrapolation l imit for projection of chromaticity sh ift.
Specifical ly:
• For a sample size of 30 u nits or g reater and a measurement interval of 600 hours or less, the extrapolation limit as
specified in Section 5.2.7 for chromaticity sh ift shal l be 6.0 times the tota l ANSI/I ES LM-80-1 5 test duration. For a
sample size of 30 u nits or g reater and a measurement i nterva l g reater than 600 hours up to a max i m u m of 1 ,000
hours, the extrapolation l im it as specified in Section 5.2.7 for chromaticity sh ift shall be 5.5 times the total ANSI/
IES LM-80-1 5 test d u ration.
• For a sample size of 20 u n its to 29 u nits and a measu rement interval of 600 hours or l ess, the extra polation limit
as specified in Section 5.2.7 for chromaticity shift sha l l be 5.5 times the total ANSl/IES LM-80-1 5 test d u ration.
For a sample size of 20 u n its to 29 u nits and a measu rement interval g reater than 600 hours u p to a max i m u m of
1 ,000 hours, the extra polation l i m it as specified in Section 5.2.7 for chromaticity shift sha l l be 5.0 times the total
ANSl/IES LM-80-1 5 test duration.
• For a sample size of 10 u n its to 19 u nits and a measurement i nterval of 600 hours or less, the extra polation limit
as specified in Section 5.2.7 for chromaticity shift shall be 5.0 times the total ANSl/IES LM-80-1 5 test d u ration.
For a sample size of 10 u nits to 19 u nits and a measu rement interval greater than 600 hours u p to a max i m u m of
1 ,000 hours, the extrapolation limit as specified in Section 5.2 .7 for ch romaticity shift shall be 4.5 times the total
ANSl/IES LM-80-1 5 test duration.
This method for chromaticity shift projection shall not be used for a sample size of fewer than 10 u nits or a
measurement i nterval g reater than 1 ,000 hours.
The O UT sam ples tested per ANSl/IES LM-80-1 5 sha l l be selected to be sufficiently representative of the overa l l
popu lation of the product. A l l data collected for the OUT sam p les tested per ANSl/IES LM-80-1 5 shall b e used for
projection, with the exception of fa iled OUT sam ples as defined in ANSI/I ES LM-80-1 5. Specially built samples to be
used as OUTs shall not be a l l owed. Selectively excluding data based on performance shall not be a l l owed.
5.1 Method
The recom mended method of chromaticity shift projection is to apply a cu rve-fit to the col lected data for differential
chromaticities, and then to use this i nformation to extrapolate the changes i n relative chromaticity to the time point
3
ANSl/IES TM-35-1 9
where the chromaticity sh ift exceeds an acceptable level (for exa mple CS? or /:::; u'v' = 0.007). That time point is the
chromaticity shift life. The same cu rve-fit i nformation for the d ifferential chromaticities can a lso be used to determ ine
the relative chromaticities, and hence chromaticity shift, at future points i n time (e.g., 20,000 hours, 25,000 hours).
Since the differential chromaticity data is calcu lated using averaged data from the OUTs, data for each OUT shall be
complete (i.e., obtained and recorded). Si nce it is a lso u n l i kely that each device u nder test (OUT) will have the same
initial chromaticity coordinates, a change in the n u m ber of data points used in calcu lating the average may result i n a
significant change in the trend. Therefore, if no chromaticity data a re reported for a OUT at a single time interval, then
no data for that OUT shall be used for this method. Similarly, if a OUT fa ils d u ring LM-80 testing, then no data for that
OUT shall be used for this method.
This method is a pplied separately for each set of OUT test data collected at each operational (e.g., d rive cu rrent) and
environmental (e.g., case temperatu re) condition as specified in LM-80-1 5.
5.2 Procedures
5.2.1 Relative Chromaticities. The relative chromaticity coordinate data of all sam ples within the same dataset
defined in Section 5.1 are calcu lated:
5.2. 2 Average Relative Chromaticities. The calculated relative chromaticity coord inates with in the same dataset
defined in Section 5.1 are averaged for each test condition and measurement interval.
5.2.3 Differential Chromaticities. The differential chromati cities are ca lcu lated using the averaged relative
chromaticity coord inates with in the same dataset as defined in Section 5.1 for each test condition and each
measurement i nterva l:
5.2.4 Data Used for Curve-Fit. Al l differential chromaticity coordinate data from 2,000 hours forward shal l be used
for cu rve-fits.
5.2.5 Curve-Fit. A li near least-squares cu rve-fit shall be performed for the d ifferential chromaticities as a fu nction of
time using (see Annex D for calcu lation exa m ples):
8u'*(t) = a ·t + b,,·,
,,
where:
= operating time, hours
8u'*(t), 8v '*(t) =differential chromaticity at time t
a,,·, a,,· = slope derived by the least-sq uares fit
4
Technica l Memorandum: Projecting Long-Term Chromaticity Coordinate Shift of LED Packages, Arrays, and Modules
5.2.6 Projecting Future Behaviors. Relative ch romaticity shift shall be projected using a numerical method, as
described here.
Numerical Projection:
Using the information at a reference time and relative chromaticity coord i nates, it is possible to calculate the relative
chromaticity coordinates as a fu nction of time. The reference time and coordinates shall be those for the last data
point used in the calculations (i.e., Bu' and 8v ' at the final measurement interval). Si nce the differential chromaticities
give the rate of change of the relative chromaticity coordi nates as a function of time, they may be used to calcu late
the change in relative ch romaticity coordinates for a given time i nterva l using:
and:
8v'(t + 6.t) = 8v'(t) + 8v'*(t)6.t = 8v'(t) + (aA + b.,)6.t ,
where 6.t shall be the average measu rement interval of the dataset used for the ca lculations.
When performing nu merica l projections, the time interval used shall be the average of the ANSl/I ES LM-80-1 5 data
measurement i nterva l for that condition.
To determ ine CSn, the above formulas shall be used in a n iterative manner u ntil the chromaticity shift threshold is
exceeded. The calculated time sha l l be obta ined using l i near i nterpolation between the two nearest extrapolated
time i nterva ls.
Whenever a CSn value is exceeded experimenta l ly in the course of ANSl/IES LM-80-1 5 testi ng, the reported value shall
be obtained by l i near interpolation between the two nearest test poi nts. The result takes precedence over a ny value
projected by the form u las above.
5.2.7 Adjustment of Results. For a sample size of 30 u nits or m o re and a measurement interval of 600 hours or fewer,
chromaticity sh ift shall not be projected beyond 6.0 times the total test d u ration of the measured data. For a sample
size of 20 to 29 u n its and a measurement i nterval of 600 hours or fewer, chromaticity shift shall not be projected
beyond 5.5 times the total test duration of the measured data. For a sample size of 1 0 to 1 9 units and a measu rement
interva l of 600 hours or fewer, chromaticity sh ift shall not be projected beyond 5.0 time the total test d u ration of the
measured data.
For a sample size of 30 u nits or more and a measurement interval of more than 600 hours and fewer than or equal to
1 ,000 hours, chromaticity sh ift sha l l not be projected beyond 5.5 times the total test d u ration of the measured data.
For a sample size of 20 to 29 u nits and a measurement interva l of more than 600 hours and fewer than or equal to
1 ,000 hours, chromaticity shift shall not be projected beyond 5.0 times the total test d u ration of the measured data.
For a sample size of 10 to 19 un its and a measurement i nterva l of more than 600 hours and fewer than or equal to
1 ,000 hours, chromaticity shift shall not be p rojected beyond 4.5 times the total test d u ration of the measu red data.
For a sample size of less than 10 u nits or a measurement i nterva l of g reater than 1 ,000 hours chromaticity sh ift sha l l
not b e projected.
5
ANSl/IES TM-35-19
Measurement Interval
Test Units >600 to s; 1 ,000
s;600 hours
hours
:2'.30 5.5 6.0
20-29 5.0 5.5
1 0- 1 9 4.5 5.0
Data collected within 48 hours of each measurement point a re acceptable for use i n product chromaticity shift
projections. The exact time val ue(s) sha l l be used in the ca l c u l ations for differentia l chromaticity parameters. The
average measurement time i nterval shall be used for extrapo latio n calcu lations.
5.2.8 Notation for Projected Chromaticity Shift Life. The ch romaticity sh ift life projected i n this method shall be
expressed using CSn(Ok) , where n is 1 ,000 x chromaticity shift measu red using u 'v ' coordi nates, and D is the total
d u ration of the test i n hours divided by 1 ,000 and rounded to the nearest 0.5. Examples:
If the calculated CSn value is red uced by the extrapolation l imits described in Section 5.2.7, the color shift l ife value
shall be expressed with a symbol ">". For exa m p l e:
If the CSn va lue is reached experimental ly in the course of LM-80-1 5 testing, the ch romaticity shift life sha l l be
expressed with the D value equal to the CSn value in hours and rou nded to the nearest 1 00. For exa m p le:
When a CSn va lue is reached experimenta l ly in the course of LM-80-1 5 testing or in extra polated resu lts, the reported
value sha l l be obtained by l inear i nterpolation between the two nearest data poi nts.
No temperature or cu rrent i nterpolation shall be used for projecting chromaticity sta bility. CSn va l ues sha l l be
calculated using test conditions described in ANSl/IES LM-80-1 5 that eq ual or exceed the case temperature and d rive
cu rrent of the LED.
6
Technica l Memorandum: Projecting Long-Term Chromaticity Coordinate Shift of LED Packages, Arrays, and Modules
7.0 Report
The report of chromaticity sh ift l ife projection shall include the i nformation shown in Table 7-1 and Figure 7-1 . The
calculated and reported CS4 and CS? val ues shall be rou nded to the nearest 1 00 hours. Differential chromaticity
coefficients (au·, b,,·, av·, bv.) sha l l be rounded to 4 significant dig its.
A vector g raphic showing the relative chromaticity shift lifetimes for CS4 and CS? shall also be included in the report.
Vector g raphic information for chromaticity shift lifetimes that exceed the a l l owable extrapolation l i mits of Section
5.2.7 shall not be plotted.
7
ANSl/IES TM-35-1 9
AnnexA-
Comments On Color Maintenance Behavior
Previous to the deve lopment of this Technical Memorandum, chromaticity shift specifications for LEDs have been
based on not exceeding a specified va lue over a certai n test i nterva l . This decision was based on the practical issues
associated with a lack of a predictive model. However, numerous exam ples exist i l l u strating that a n LED with essentia l ly
constant chromaticity coordinate shift over a period of several thousand hours suddenly begins to shift in a consistent
(l inearly i ncreasing) manner and soon exh ibits unaccepta ble deg rees of change in chromaticity coord inates.
This is shown by the LED data in Figure A-1, in which t.u'v' is approximately equal to 0.0005 from a bout 1 ,000 hours
u ntil about 6,000 hours, after which it increases l i nearly at a rate of a bout 0.0007 per thousand hours. The obvious
trend can be projected to reach t.u'v' = 0.007 at a bout 1 5,500 hours.
0.007
,.,
0
0.006
y
u
8E-07x- 0.0046 o'
'
=
,)
0.005 R2 = 0.9951
0.004 "'
::::s
.]> ' cf
(),o
'
<1
0.003
,
0.002 ,... ,o
)
,<) -
�
o-
0.001
ooo )000 >Ooo ) <?,'
0.000
0 2,000 4,000 6,000 8,000 10,000 12,000 14,000
Time (hrs)
Figure A-1 . An example of LED color shift behavior, i l l ustrating an essentially l i near i ncrease
i n .iu'v' after a period of about 6,000 hours.
More important, it is possible for at least some LEDs to exh i bit extreme shifts in chromaticity coordi nates while
exh ibiting excellent l u men mai ntenance. Plotting the data from the LED shown i n Figure A-1 against the lumen
maintenance data reveals the potentia l severity of the issue (see Figure A-2).
The cha l lenge of developing a model for predicting the chromaticity coordinate shift of LEDs is related to the
complicated ra nge of behaviors exhibited by LED products. Chromaticity sh ift behavior is usually shown as a plot of
t.u'v' versus time, where t.u'v' is calcu lated as the difference in c h romaticity coordi nates between a g iven time and
time zero.
8
Technica l Memorandum: Projecting Long-Term Chromaticity Coordinate Shift of LED Packages, Arrays, and Modules
105%
� '
Cj
100%
�<> >
=
6
<> <>
�� v <>
95% � V' v
<> <> <> <> <>
-
�
.5 90%
�
=
�
85%
§ 80%
�
75%
70%
0.000 0.001 0.002 0.003 0.004 0.005 0.006 0.007
�u'v'
Figure A-2. Lumen maintenance and chromaticity shift behavior of the LED from Figure
A1 . This LED is projected to reach a chromaticity coordinate shift of du'v ' = 0.007 while
mainta i n i n g lumen mai ntenance greater than 90%.
0.007 I I
0.006 I
0.005
I I
0.004
::s
;> Emergence
I
<]
0.003 _.. ...
F ecovery . ..... ,,.
0.001 I
I
-nOC
00011
I
c;> OO ()000 >OQo >oo
:>
I
I I o o:o 0 )
0.000
0 2,000 4,000 6,000 8,000 10,000 12,000 14,000
Time (hrs}
Figure A-3. An example of LED color mai ntenance behavior illustrating incubation, recovery,
and emergence.
9
ANSl/IES TM-35-1 9
LEDs may not exhi bit a l l three phases of behavior. Most LEDs w i l l exhibit incubation, a lthough u nder extreme operating
conditions (e.g., drive cu rrent and case temperature near the maximum a l lowable) some LEDs may reach emergence
a l most immediately. Other LEDs may exhibit incu bation followed by emergence (see Figure A-4). H owever, once
a n LED enters emergence, the chromaticity sh ift contin ues to i ncrease in a n approximately l i near fashion to until it
exceeds Llu'v' = 0.007 - 0.01 0.
0.007 �
0
0
0.006
0
0.005 0
0.004 no
_o
)- lncubat on )
<I ,... 0
:I
I'
,;d 0 ()
0.003 "'
\On n Cm--
� ,....
..-i:;te-R-W
01
0.002 I
n
I I
I I
I I
0.001 I I
0
I I
I I
I I
0.000
0 2,000 4,000 6,000 8,000 10,000 12,000 14,000
Time (hrs)
Figure A-4. An example of LED chromaticity coordinate shift behavior showing
incubation followed by emergence.
0.001
T -....,
\
� A
/.
......
��
.....
.. ...
.....
0.000
.. : �
I
�
-> -0.001
Liu'v' = 0.0018
c.o
J
,,,' <
, /t
•>
-0.002
-0.003 J
-0.002 -0.001 0.000 0.001 0.002 0.003 0.004
ou'
Figure A-5. Chromaticity shift data for the LED of Figure A4. During incubation, the
chromaticity coordinates move approximately along a circle of constant radius.
When the chromaticity coordinates cease tracking the circular arc, the LED enters
the emergence phase.
10
Technica l Memorandum: Projecting Long-Term Chromaticity Coordinate Shift of LED Packages, Arrays, and Modules
The time at which emergence begins varies with both d rive cu rrent and operating tem peratu re. Similarly, the rate at
which t:.u'v' increases d u ring emergence also changes with both d rive cu rrent and operating temperatu re. The va lue
of t:.u'v' d uring incubation will vary slig htly with both drive cu rrent and operating temperature. Figure A-6 provides
an example of this behavior.
0.007
o
-t-���--+-����1--���+-���-H
105 C, 500 mA
D
0.006
105 C, 700 mA
6. 105 C, 1,000 mA
0.005
Xb.
xb.
:> 0.004 -----+----.....,...--+< 0 55 C l,500mA -
-----+----...-....--��--+< X
<I
::::J
85 C, 1,500mA
0.003
0.002
0.001
0.000
0 3,000 6,000 9,000 12,000 15,000 18,000
Time (hrs)
Figure A-6. An example of chromaticity sh ift behavior for an LED at different operating
conditions.
The similarity of measured t:.u'v' va l ues during incu bation is one of the factors that has com pl icated the deve lopment
of models capable of predicting l ong-term chromaticity shift behavior. After emergence, predicting long-term
chromaticity shift behavior is relatively straig htforward and s i m ple, but as i l lustrated in Figure A-6, this may req u i re
testing for periods in excess of 1 2,000 hours. This is better than testing u ntil end of l ife, but not significantly so.
The time req u ired for an LED to reach emergence and the rate of i ncrease in t:. u 'v' is related to both the d rive cu rrent
and the operating temperature, but i nadeq uate data was available to develop a reliable interpolation method for
either operating temperature or d rive cu rrent.
Annex B-
Types of Chromaticity Shift Behavior
Chromaticity shift in LED devices is a complex process that often i nvolves m u ltiple processes occurring at different
operational times d u ring the LED l ifetime. I nitia l ly, many LEDs tend to show a rapid shift in chromaticity slig htly toward
either the blue or g reen direction. This shift typical corresponds to t:.u'v' of less than 0.003 and occurs within about
the first 1 ,000 hours of operation. The initial chromaticity shift may then be fol lowed by a prol onged chromaticity
shift that can be much larger in magnitude. This prolonged, or terminal, shift principally occurs in directions that
rou g h ly correlate to one of the major col ors-red, g reen, blue, or yellow (see Figure B-1)-and can be in a completely
d ifferent direction from the initial shift. Since the chromaticity coord inates of a n LED device are a bala nce of emissions
over the visible spectrum, chromaticity sh ifts can be caused by either a relative i ncrease or decrease in emissions over
11
ANSl/IES TM-35-1 9
a particular wavelength ra nge. For exa m ple, a blue shift can be caused by either an i ncrease in b l u e emissions from
the LED or a relative decrease in yellow emissions from the phosphor. Likewise, a relative increase in g reen emissions
may produce a g reen shift, while a relative decrease i n g reen em ission may produce a red shift.
0.70
0.60
0.50
0.40
->
0.30
0.20
0.10
u'
Figure B-1. Chromaticity shift directions for LED devices, shown using the
CIE 1 976 color space. A representative initial chromaticity point is given
by the white ci rcle, and the chromaticity points of the blue and yellow
emitters are also shown.
The time req u i red for the i m pact of the terminal chromaticity shift to be manifested in the device is correlated to the
incubation period (see Section A.1). Since the terminal sh ift is responsible for the chromaticity shift behavior of the
LED device over the majority of its lifetime, u nderstanding this period of prolonged chromaticity shift is especially
im portant in b u i l d i ng projections of l ifetime chromaticity shift. Developing such projections a lso req u i res projecting
the length of the incubation time for the term inal chromaticity s h ift processes.
By exa mining the chromaticity behavior of LEDs and LED devices, five chromaticity shift modes (CSMs), roug h ly
corresponding to the principal color directions, have been identified in the l iterature. These CSMs a re listed in Table B-1
and provide g u ida nce on the potential terminal chromaticity sh ift behaviors that can occur in LED packages.
12
Technica l Memorandum: Projecting Long-Term Chromaticity Coordinate Shift of LED Packages, Arrays, and Modules
Chromaticity
Direction of Terminal Chromaticity Shift Potential Causes
Shift Mode
CSM-1 Blue Annealing processes i n LED package d u ring operation
CSM-2 Green Oxidation of phosphors
CSM-3 Yellow Delami nation or interior cracks in the binder
Starts in the yellow d i rection, then turns Second (blue) shift is likely caused by photo-oxidation
CSM-4
toward the blue d i rection of molding resin
CSM-5 Red Reduction i n q u a ntum efficiency of green phosphors
Annex C-
Mathematical Model Development
The TM-35 Working G roup (WG) conducted analyses on more than 30 sets of IES LM-80-08 a nd ANSl/IES LM-80-1 5
test data, collected from a nu mber of sources. These datasets i n c luded m u ltiple d rive cu rrent and case temperature
conditions. Wel l over 1 00 i nd ividual d rive current and case temperature datasets were analyzed. The duration
conditions for these datasets ranged from 6,000 hours to 20,000 hours. The LEDs in col lective of datasets represented
a broad spectrum of package formats and technologies, including PLCC, COB, hig h-power cera mic, and EMC
(sometimes cal led QFN) packages.
The TM-35 WG investigated a nu mber of d ifferent mathematical m odels. One early proposed model was based
on testing LEDs u ntil the trend in emergence had been firmly estab l ished. After emergence, a linear fit was used
to p roject t:.u'v' forward in time. The difficulty with this model was that it req u i red potentia l l y excessive testing, as
emergence could take 1 5,000 hours or more to determ ine, depending on the LED package construction, materials,
d rive cu rrent, and operating temperature. (See Figure C-1.)
0.009
0.008
0.007
0.006
0.005
>
<]
:::J
0.004
0.003
0.002
0.001
0
0 4,000 8,000 12,000 16,000 20,000
Time ( h )
13
ANSl/IES TM-35-1 9
Another model used a step fu nction and assumed a power law behavior for !:,.u'v' after i ncubation. This model takes
the form:
where:
()is the step function,
dis the incubation period
c is the average value of !:,. u 'v' d uring incubation
a and n are constants that model the behavior of !:,. u 'v' after emergence
One com plication with the model is that 3 of the parameters (d, a, n) can only be determ ined after emergence, which
equates to a req u i rement to test u ntil emergence. With enough data points after emergence it is possible this model
could predict behavior after emergence to large va l ues of t,.u'v', but with lim ited data points the model has only an
a pproximate a b i l ity to model behavior.
The model was tested with a dataset as shown in Figure C-2. The data from 5,000 to 8,000 hours was used to ca lculate
the parameters by an optim ization scheme minim ized the difference between the data and the model.
0
0.007 0
I
o Data
0
0
• Data Po ints for Fit
0.006
o Model oo
0
Oo
0 0
0
0.005
0
0
f-
0
0
0
0.004
:> 0
<I
:J
0
f-- 0 0
0 0
0.003
0
•o
0.002 - to
0oi
•
0.001 ,_.g�<> oo�<> oo�� •
0 I I I
0 2,000 4,000 6,000 8,000 10,000 12,000 14,000 16,000
Time (h)
Figure C-2. Evaluation of ..:iu' v
'
(t) = B(t-d)"a·t11 + c model.
After emergence, modeling the chromaticity shift versus time reduces to approximately a l i near equation in time. Many
models were capable of matching this behavior. Therefore, the ability to predict emergence was chosen as a metric for
comparing different models. A benchmark LED dataset was chosen to evaluate several different models. The dataset
extended to 1 2,000 hours and exhibited emergence at just after 6,000 hours. The dataset was chosen based on the
proposition that any model incapable of predicting emergence just before it occurred would be unable to accu rately
predict emergence at some point in the future. A plot of !:,.u'v' versus time for the test dataset is shown in Figure C-3.
14
Technica l Memorandum: Projecting Long-Term Chromaticity Coordinate Shift of LED Packages, Arrays, and Modules
0.007
0.006
/
rr
0.005 n/
O' ,
.Y
->
0.004
-:I (I
p"
<] 0.003 ,,u
�,( e
.()
0.002
n ,... ':>"'�
0.001 e-- - - - - - --- -o-o--;.
0.000
0 2,000 4,000 6,000 8,000 10,000 12,000 14,000
Time (hours)
incubation (short black dashes), <iu'v' = 0.000998 ± 0.0001 7. During emergence (long gray
Figure C-3. <iu'v' versus time plot of benchmark dataset used to evaluate models. During
Most of the models eval uated assumed a l inear behavior of chromaticity coordi nates with respect to time or a type of
exponential decay. This method uses a d ifferential chromaticity a n a lysis (DCA) model that is based on the assumption
that differential chromaticity exh i bits a linea r behavior with respect to time. Some models that were eval uated
assumed an exponential behavior of tristimulus val ues with respect to time. Others req uired prior knowledge of
the chromaticity shift behavior after emergence. Of a l l the models eva l uated, o n ly DCA and derivatives were able to
accu rately predict the i m m i nent emergence behavior in the test dataset with no assu m ptions related to the specific
package or package format.
In hindsight, chromaticity coord inates may exh ibit many different behaviors with respect to time (see Figure C-4).
They may show a monotonic increase, a monotonic decrease. These increases or decreases may be constant or slowly
change rate over time. Other behaviors a re possi ble including a n i ncrease fol l owed by a decrease, a decrease fol l owed
by an increase. Chromaticity coordinates may also remain essentia l l y constant over long periods of time.
All of these behaviors may be a pproximated with quadratic fu n ctions. The first derivative of a quadratic fu nction is
a l inear function. Si nce DCA assumes a l inear first derivative, it is mathematica lly equivalent to a quadratic model for
each chromaticity coordinate with respect to time. Since predicting the rate of chromaticity sh ift is dependent on the
rate of change of chromaticity coord inates with respect to time, DCA only calculates the parameters associated with
that rate of cha nge.
The differential aspect of the model provides some robustness against noise in the data and is mathematica lly less
com plex, as it req u i res i nverting a 2x2 matrix for each chromaticity coordinate as opposed to i nverting a 3x3 matrix
for each chromaticity coordinate.
Additional ly, since only the shift relative to the chromaticity at time = 0 is considered, DCA uses relative chromaticity
coordinates - coordinates that a re referenced to the values at time = 0. This is a form of normal ization and a l l ows
averaging of data between different LEDs. Relative chromaticity coordinates a lso all the behavior of different LEDs to
be show on the same plot without loss of deta il.
15
ANSl/IES TM-35-1 9
0.0020 0.0000
0
0.0015
� -0 .000 5
0
·� :B..
.;:;
0
., 0
�
-5
..
0.0010 0 0
-5
E
-0.0010 0
0
e
�
0.0005
·!
0
0
.. · -0.0015 0
�
0 0
0.0000 0
0
.. 0
Qi 0
0
0 � -0.0020 0
0
-0.000 5 0
0
0.0010
.,
. 0.0000
0
E
.fi -0.0010 0 0 0 0 0 0 0 0 0 0 0 0 0
..
0 0
�
�
-0.0020
0.0010
·.5
0 0 0
Time (h)
0 0 -0.0030
> 0.0000
0 2,000 4,000 6,000 8,000
..
0
..
-5
0
-5
E
t:
e
-0.0010 -0.0008
g:
e 0 0
·�
..
�
-0.0012
�
�
-0.0020 0 0
0
0
0
0 0 0
--0.0030 -0 .0016
0 0
0 0 0 0
Time (h)
-0.0020
Time (h)
-0.0040
2,000 4,000 6,000 8,000
2,000 4,000 6,000 8,000
Figure C-4. Examples of range of behavior for relative chromaticity versus time.
By itself, the mathematica l ly simpler approach of differential chromaticity might not be considered sufficient
reason to use DCA as a pred iction method. I ndeed, the WG considered a variation of DCA that yielded a n analytical
expression for differential chromaticity. This method involved calcu lating a quadratic fit of the relative chromaticities
and then taking the analytical derivative to yield the same l i near differential chromaticity parameters (au·, bu·, av·, bv.J,
This a pproach lacked some appeal, as it req uired calcu lation of 6 parameters and then discarded 2 of them.
More important, a n in depth review of the datasets revealed that DCA possessed a remarkable robustness agai nst
measurement errors that was not present in a ny of the other models, i ncluding the ana lytical derivative model.
Fol l owing the reasoning of TM-21 -1 1 , which used 6 data points for a l inear cu rve, DCA therefore req u i red a minimum
total test d u ration of 7,000 hours i n order to guara ntee a minimum of 6 data poi nts for curve fitting.
The second reason for beg i nning analysis at the 2,000-hour data point came from more-detailed i nvestigations, using
long-term datasets that extended past 1 2,000 hours. Both CS4 and CS7 were projected using data out to 6,000 hours
and then progressively longer d u rations. This was done starting the analysis at the 1 ,000-hour data point and at the
2,000-hour data poi nt.
16
Technica l Memorandum: Projecting Long-Term Chromaticity Coordinate Shift of LED Packages, Arrays, and Modules
For some datasets, there was no significant difference in the projected times for CS4 and CS?, whether or not the
1 ,000-hour data point was included. In other cases, there was a significant difference in the calcu lated times for CS4
and CS?. An example case is shown in Figure C-5.
4 5,000
40,000 IP
a
35,000
a
n
ll ll
a
30,000
a ,,
ll
ll
- ()
n I
2 5,000
C) 0 0 1) 0 I)
v
Ill 0 C) 0
� 20,000 j1')
0
<i>
:::.. 15,000
,.....
VI
u 10,000
5,000
0
6,000 7,000 8,000 9,000 10,000 11,000 12,000
Analysis End Time (hours)
(0) and
2,000-hour (O) data points for different durations of ana lysis.
Figure C-5. Comparison of calculated value of CS7, starting analysis at the 1,000-hour
While both analyses converged at the same va lue of a pproximately 24,000 hou rs, the analysis that began with the
1 ,000-hour data point started significantly higher and then decreased. In contrast, the analysis that began with the
2,000-hour data point started slightly lower than 24,000 hours and then i ncreased.
This type of trend, of converg ing estimates that slightly increased when beginning the analysis with the 2,000-
hour data poi nt, was noted across the datasets ana lyzed by the WG. The shorter duration datasets tended to g ive
conservative estimates that i m p roved as the total test d u ration was i ncreased.
The contrasting behavior between a portion of the 1 ,000-ho u r analyses and the 2,000-hour ana lyses led to the
adoption of 2,000 hours as the first data point in DCA calcu lations.
It is a lso noted that DCA uses a backwards difference equation when cal c u lating differential chromaticity. This means
that for data with a 1 ,000-hour measurement interval, the initial chromaticity data (i.e., data collected at time = 0
hou rs) wou l d be used in the calcu lations. As noted in Annex B, LEDs commonly exh ibit a chromaticity sh ift with in
the first 1 ,000 hours that is u n related to long-term ch romaticity shift behavior. The mag nitude of this sh ift per unit of
time is often much larger than that exhibited at later periods of time and thus may introduce a significant error in the
DCA calculations.
17
ANSl/IES TM-35-1 9
0.007
0.006
0.005
0.004
::s
:>
--
<I 0.003
-- --
0' -
p-
°' .P
b-
0.002
n,o-0..P
0 -'"'-o- & o. :o.-
0 0 0 0- -Q-&1IT "O-G
0.001
The mean variance between the actual Llu'v' va l ues and the values extrapolated from the 2,000 data point was
0.0000456 over a 1 6,000-hour period ohime. When the behavior is extrapolated from the 2,000 hours data poi nt, the
correlation coefficient over the entire i nterval is r2 > 0.97.
The overa l l trajectory of each OUT in chromaticity coordi nates was highly consistent, however differences in the rate
at which OUTs travel led along this trajectory were noted to vary. Th is was evident when the standard deviation of
the chromaticity coordinates for a set of OUTs was plotted vs Llu'v'. The model thus incl uded slight va riabil ity in the
overa l l trajectory as wel l as different speeds along the trajectory for each OUT. The variab i l ity was adjusted to match
the trends in sta ndard deviation of chromaticity coordi nates vs Llu'v' seen in the datasets.
Measurement uncertainty was introduced for each OUT using data from the NVLAP Round 15 Proficiency Test. This
gave values for the variance for u' and v' were estimated at 0.000200 and 0.000082, respectively. Variances were
assumed to be normal ly d istributed.
Monte Carlo simulations were cond ucted using the model for different behaviors, time intervals and num bers of OUTs.
The nominal trajectory for a g iven run was held constant while the rate of travel along the chromaticity trajectory was
changed in order to estimate the accuracy of the model at different extrapolation l i m its.
Ideal CS? val ues were calcu lated by model i n g a set of OUTs without a ny measurement u ncertai nty. These ideal ized
va l ues were compared to the calculated va l ues based on a set of OUTs to which an uncertainty was a pp lied for each
simulated measurement. By comparing these values, the WG was able to estimate the confidence l i m it at different
extra polation levels.
18
Technica l Memorandum: Projecting Long-Term Chromaticity Coordinate Shift of LED Packages, Arrays, and Modules
Simu lations were run using measurement intervals of 500 hours or 1 ,000 hours and 1 0 DUTs, 20 DUTs and 30 DUTs,
with a total test d u ration that ranged from 7,000 to 1 0,000 hours. Ideal ized CS? val ues ra nge from a bout 1 2,000 hours
to a bout 84,000 hours, which corresponded to extrapolation l i m its that ra nged between a bout 1 .2 and a bout 1 0.5.
Ana lysis of the dataset corresponding to 20 DUTs, total test duration of 7,000 hours and measurement intervals of
500 hours is shown in Figure C-7. As a first comment, it is noted that DCA is u n l i kely to u nder-predict the rate of
chromaticity shift. This shows the conservative nature of the extrapolation. When the duration of testing is extended
past 7000 hours, the accuracy of the prediction is i ncreased.
70,000
90th percentile
-• Median
-+- 10th percentile
60,000
-
.c
-
50,000
"""
"'
u
"Cl
cu
- ·
....
40,000
"'
:::J
u
"'
30,000
u
20,000
10,000
10,000 20,000 30,000 40,000 50,000 60,000 70,000
Idealized CS7 {h)
Figure C-7. Error analysis for 20 DUTs, 7,000-hour test d u ration, 500-hour measurement
intervals.
The second feature to be noted is that the confidence bands approach a horizontal asymptote as CS? increases. These
asym ptotes are correlated to the measurement u n certainty and represent a region where the prediction of the model
is dominated by the measu rement u ncertainty. This provides a clear indication of one method for determ ining an
extrapolation l i m it: the intersection of the 90% one-sided confidence l i mit and the u n certainty boundary. As can be
seen i n Figure C-8, this occurs at a n extra polation l i m it of about 5.5 for 20 DUTs, 7,000-hour test d u ration and 500-
hour measurement i ntervals.
In addition, the mean error of the prediction was considered. The geometric mean of the 90% one-sided error
estimate was plotted as a fu nction of extrapolation l i m it. The extrapolation l i m it for TM-21 was determined based on
the 90% one-sided confidence l i m it. The metric chosen corresponded to a n error estimate of about 1 5%. Applying
a mean error l i m it of 1 5% was found to yield a n extrapolation l i m it that was virtually identical to the intersection of
the 90% one-sided confidence limit and the u ncertainty bou ndary as can be seen in Figure C-9. Both of these criteria
were used to determine the extrapolation l im its of Section 5.2.7.
19
ANSl/IES TM-35-1 9
60,000
-- 90th percentile
-e- Median
-+- 10th percentile
50,000
..c
-
"
-
V)
u
40,000
"tJ
Cl.I
ftl
...
::::s
u
30,000
u
ftj
20,000
10,000
1 2 3 4 5 6 7 8
Extrapolation Limit
Figure C-8. Error analysis for 20 DUTs, 7,000-hour test duration, 500-hour measurement
intervals, plotted as a function of extra polation limit.
25%
0
"
c
20%
(D
0
...
"
15%
LLI
...
0
...
c
ftl
Cl.I 10% u
�
0
0
0
5%
0
0
0
0
oo 0
0%
1 2 3 4 5 6 7 8
Extrapolation Limit
Figure C-9. Mean error estimate versus extrapolation limit for 20 DUTs, 7,000-hour test
duration and 500-hour measurement interval.
20
Technica l Memorandum: Projecting Long-Term Chromaticity Coordinate Shift of LED Packages, Arrays, and Modules
Annex D
Calculation Examples
To assist users of this document in carrying out calcu lations fol lowing the procedu res in Section 5, an exa m ple set of
calculations is presented here. The dataset used is from one of the LM-80 datasets used by the TM-35 working group
(WG) i n the deve lopment of this method. It is recom mended that user use the data and calcu lation resu lts presented
here to make a comparison with his or her own calcu lations to ensure that the calcu lation steps taken are correct.
For this first exa m ple, the operating conditions of the LED were a case temperature of 1 05 °C and d rive cu rrent of 500 mA.
Tables 0-la,b and D-2a,b provide example u ' and v ' data, respectively, for the first exa mple LED.
21
ANSl/IES TM-35-1 9
22
Technica l Memorandum: Projecting Long-Term Chromaticity Coordinate Shift of LED Packages, Arrays, and Modules
C I E v' Sample #
Time, h 11 12 13 14 15 16 17 18 19 20
0 0.52072 0.52 1 83 0.5 1 997 0.5 1 901 0.5 1 903 0.5 1 6 1 4 0.52068 0.52286 0.52579 0.52280
504 0.52022 0.52 1 63 0.5 1 966 0.5 1 855 0.5 1 870 0.5 1 573 0.52048 0.52266 0.52566 0.52253
1 ,008 0.52024 0.5 2 1 58 0.5 1 960 0.5 1 843 0.5 1 85 9 0.5 1 575 0.52037 0.52261 0.52552 0.52248
1 ,5 1 2 0.52020 0.5 2 1 54 0.5 1 959 0.5 1 842 0.5 1 862 0.5 1 569 0.52033 0.52264 0.525 5 1 0.52244
2,0 1 6 0.520 1 5 0.521 52 0.5 1 957 0.5 1 838 0.5 1 856 0.5 1 572 0.52028 0.52274 0.52549 0.52253
2,520 0.52009 0.5 2 1 49 0.5 1 954 0.5 1 843 0.5 1 850 0.5 1 563 0.52034 0.52257 0.52549 0.52243
3,024 0.52008 0.52 1 48 0.5 1 953 0.5 1 836 0.5 1 849 0.5 1 566 0.52034 0.52257 0.52544 0.52239
3,528 0.52009 0.5 2 1 49 0.5 1 952 0.5 1 838 0.5 1 847 0.5 1 567 0.52030 0.52254 0.52542 0.52242
4,032 0.520 1 2 0.52 1 48 0.5 1 953 0.5 1 844 0.5 1 853 0.5 1 56 1 0.520 3 1 0.52257 0.52548 0.52244
4,536 0.520 1 2 0.52 1 48 0.5 1 953 0.5 1 845 0.5 1 849 0.5 1 565 0.520 3 1 0.52254 0.52545 0.52244
5,040 0.52005 0.52 1 43 0.5 1 947 0.5 1 835 0.5 1 839 0.5 1 56 1 0.52026 0.52248 0.52537 0.52242
5,544 0.5 1 998 0.52 1 46 0.5 1 938 0.5 1 838 0.5 1 835 0.5 1 558 0.52033 0.52244 0.52538 0.5223 1
6,048 0.520 1 2 0.52 1 49 0.5 1 956 0.5 1 848 0.5 1 85 1 0.5 1 566 0.52036 0.52259 0.52540 0.52242
6,552 0.52006 0.52 1 5 1 0.5 1 954 0.5 1 856 0.5 1 844 0.5 1 567 0.52034 0.52252 0.52543 0.52234
7,056 0.52004 0.52 1 36 0.5 1 948 0.5 1 85 1 0.5 1 836 0.5 1 568 0.52035 0.5225 1 0.52531 0.52228
7,560 0.520 1 0 0.52 1 44 0.5 1 938 0.5 1 854 0.5 1 83 1 0.5 1 562 0.52035 0.52259 0.52 5 3 1 0.52232
8,064 0.520 1 0 0.52 1 46 0.5 1 952 0.5 1 855 0.5 1 833 0.5 1 565 0.52045 0.52262 0.52529 0.52226
8,568 0.520 1 2 0.52 1 47 0.5 1 959 0.5 1 867 0.5 1 845 0.51 563 0.52045 0.52265 0.52535 0.52229
The first step in the calcu lations is calculating the relative chromaticity coordi nates for the each LED at each time
interval. The relative chromaticity coordinates are calcu lated using:
23
ANSl/IES TM-35-1 9
Ta ble D-3a. llu' Relative Chromaticity Coordinates for the First Exa m ple LED.
6u' Sample #
Time, h 2 3 4 5 6 7 8 9 10
0 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000
504 -0.0005 1 -0.00047 -0.00045 -0.00055 -0.00056 -0.00051 -0.00052 -0.00051 -0.00055 -0.00061
1 ,008 -0.00053 -0.00044 -0.00057 -0.00057 -0.00057 -0.00051 -0.00054 -0.00047 -0.00052 -0.00060
1 ,5 1 2 -0.00058 -0.00047 -0.00057 -0.00058 -0.00058 -0.00057 -0.00061 -0.00052 -0.00059 -0.00067
2,0 1 6 -0.00063 -0.00050 -0.00062 -0.00068 -0.00065 -0.00060 -0.00066 -0.00055 -0.00063 -0.00061
2,520 -0.00068 -0.00059 -0.00068 -0.00074 -0.00064 -0.00059 -0.00070 -0.00058 -0.00066 -0.00077
3,024 -0.00064 -0.00053 -0.00066 -0.00073 -0.00061 -0.00060 -0.00069 -0.00058 -0.00062 -0.00070
3,528 -0.00063 -0.00049 -0.00067 -0.00067 -0.00066 -0.00050 -0.00058 -0.00052 -0.00061 -0.00067
4,032 -0.00065 -0.0005 1 -0.00058 -0.00067 -0.00056 -0.0005 1 -0.00061 -0.00051 -0.00055 -0.00064
4,536 -0.00056 -0.00040 -0.00055 -0.00065 -0.00048 -0.00049 -0.00055 -0.00045 -0.00053 -0.00057
5,040 -0.00072 -0.00056 -0.00071 -0.00078 -0.00065 -0.00063 -0.00073 -0.00059 -0.00063 -0.00076
5,544 -0.00072 -0.00060 -0.0007 1 -0.00075 -0.00064 -0.00063 -0.00073 -0.00059 -0.00068 -0.00076
6,048 -0.00065 -0.00058 -0.00063 -0.00075 -0.00062 -0.00064 -0.00074 -0.00059 -0.00065 -0.00071
6,552 -0.00066 -0.00053 -0.00073 -0.0007 1 -0.00065 -0.00063 -0.00071 -0.00054 -0.00065 -0.00075
7,056 -0.00067 -0.00059 -0.00068 -0.00079 -0.00063 -0.00064 -0.00072 -0.00047 -0.00062 -0.00076
7,560 -0.00071 -0.00063 -0.00085 -0.00087 -0.00070 -0.00078 -0.00075 -0.00058 -0.00067 -0.00085
8,064 -0.00078 -0.00068 -0.00079 -0.00096 -0.00077 -0.00082 -0.00080 -0.00062 -0.00075 -0.00085
8,568 -0.00074 -0.00070 -0.00082 -0.00095 -0.00075 -0.00087 -0.00085 -0.00063 -0.00074 -0.00082
Ta ble D-3b. llu' Relative Chromaticity Coordinates for the First Exa m ple LED.
6u' Sample #
Time, h 11 12 13 14 15 16 17 18 19 20
0 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000
504 -0.00060 -0.00058 -0.00058 -0.00062 -0.00063 -0.00058 -0.00046 -0.00060 -0.00065 -0.00069
1 ,008 -0.00064 -0.00063 -0.00058 -0.00069 -0.00065 -0.00065 -0.00052 -0.00059 -0.00068 -0.00074
1 ,5 1 2 -0.00065 -0.00064 -0.00064 -0.00071 -0.00066 -0.00063 -0.00054 -0.00063 -0.00071 -0.00074
2,0 1 6 -0.00074 -0.00067 -0.00067 -0.00071 -0.00074 -0.00064 -0.00058 -0.00052 -0.00078 -0.00069
2,520 -0.00076 -0.00074 -0.00071 -0.00077 -0.00072 -0.00071 -0.00059 -0.00065 -0.00077 -0.00078
3,024 -0.00080 -0.00070 -0.00072 -0.00072 -0.00069 -0.00073 -0.00057 -0.00061 -0.00075 -0.00077
3,528 -0.00075 -0.00066 -0.00064 -0.00074 -0.00067 -0.00065 -0.00052 -0.00063 -0.00069 -0.00074
4,032 -0.00076 -0.00061 -0.00066 -0.00071 -0.00067 -0.00063 -0.00056 -0.00060 -0.00070 -0.0007 1
4,536 -0.00069 -0.00061 -0.00062 -0.00066 -0.00056 -0.00062 -0.00050 -0.00054 -0.00072 -0.00064
5,040 -0.00074 -0.00065 -0.00070 -0.00077 -0.00064 -0.00070 -0.00056 -0.00066 -0.00076 -0.00066
5,544 -0.00087 -0.00075 -0.00076 -0.00082 -0.00073 -0.00076 -0.00065 -0.00069 -0.00079 -0.00083
6,048 -0.00079 -0.00072 -0.0007 1 -0.00077 -0.00067 -0.00074 -0.00059 -0.00065 -0.00076 -0.00079
6,552 -0.00080 -0.00072 -0.00074 -0.00076 -0.00069 -0.00075 -0.00060 -0.00064 -0.00076 -0.00078
7,056 -0.00081 -0.00079 -0.00068 -0.00077 -0.00064 -0.00076 -0.00063 -0.00057 -0.00085 -0.00076
7,560 -0.00084 -0.00083 -0.00086 -0.00087 -0.00080 -0.00091 -0.00066 -0.00056 -0.00089 -0.00084
8,064 -0.00098 -0.00090 -0.00083 -0.00088 -0.00074 -0.00093 -0.00072 -0.00065 -0.00 1 00 -0.00088
8,568 -0.00090 -0.00090 -0.00086 -0.00095 -0.00081 -0.00099 -0.00077 -0.00064 -0.001 04 -0.00088
24
Technica l Memorandum: Projecting Long-Term Chromaticity Coordinate Shift of LED Packages, Arrays, and Modules
Ta ble D-4a. cSv' Relative Chromaticity Coordinates for the First Exa m ple LED.
6u' Sample #
Time, h 1 2 3 4 5 6 7 8 9 10
0 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000
504 -0.00052 -0.00035 -0.00044 -0.00043 -0.00036 -0.00022 -0.00041 -0.0001 9 -0.00022 -0.00037
1 ,008 -0.00054 -0.00033 -0.00046 -0.00057 -0.00038 -0.00031 -0.00057 -0.000 1 7 -0.00030 -0.00040
1 ,5 1 2 -0.00057 -0.00041 -0.00054 -0.00058 -0.00035 -0.00039 -0.00063 -0.000 1 7 -0.00029 -0.00044
2,0 1 6 -0.00066 -0.00043 -0.00056 -0.00058 -0.00043 -0.00036 -0.00070 -0.00026 -0.00029 -0.00038
2,520 -0.00064 -0.00049 -0.00069 -0.00063 -0.00045 -0.00047 -0.00066 -0.00025 -0.00036 -0.00048
3,024 -0.00061 -0.00050 -0.00057 -0.00061 -0.00048 -0.00040 -0.00060 -0.00025 -0.00037 -0.00052
3,528 -0.00063 -0.00045 -0.00062 -0.00062 -0.00047 -0.00042 -0.00066 -0.00028 -0.00037 -0.00052
4,032 -0.00061 -0.00049 -0.00066 -0.00057 -0.00047 -0.00049 -0.00054 -0.00028 -0.00043 -0.00050
4,536 -0.00059 -0.00043 -0.00062 -0.00054 -0.00050 -0.00039 -0.00048 -0.00028 -0.00038 -0.00052
5,040 -0.00066 -0.00055 -0.00066 -0.00062 -0.00054 -0.00052 -0.00070 -0.00036 -0.00051 -0.00061
5,544 -0.00073 -0.00055 -0.00070 -0.00067 -0.00060 -0.00048 -0.00073 -0.00040 -0.00049 -0.00069
6,048 -0.00056 -0.00049 -0.00055 -0.00060 -0.00043 -0.00044 -0.00066 -0.00024 -0.00042 -0.00053
6,552 -0.00067 -0.00049 -0.00058 -0.00053 -0.00048 -0.00038 -0.00072 -0.00031 -0.00040 -0.00062
7,056 -0.00066 -0.00062 -0.00052 -0.00068 -0.00047 -0.00043 -0.00073 -0.00033 -0.00046 -0.00067
7,560 -0.00062 -0.00055 -0.00059 -0.00064 -0.00051 -0.00051 -0.00078 -0.0003 1 -0.00046 -0.00076
8,064 -0.00052 -0.00051 -0.00049 -0.00060 -0.00043 -0.00042 -0.00060 -0.00026 -0.00040 -0.00066
8,568 -0.00043 -0.00052 -0.00038 -0.00057 -0.00036 -0.00038 -0.00065 -0.000 1 9 -0.00041 -0.00056
Ta ble D-4b. cSv' Relative Chromaticity Coordinates for the First Exa mple LED.
6u' Sample #
Time, h 11 12 13 14 15 16 17 18 19 20
0 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000
504 -0.00050 -0.00020 -0.0003 1 -0.00046 -0.00033 -0.0004 1 -0.00020 -0.00020 -0.000 1 3 -0.00027
1 ,008 -0.00048 -0.00025 -0.00037 -0.00058 -0.00044 -0.00039 -0.00031 -0.00025 -0.00027 -0.00032
1 ,5 1 2 -0.00052 -0.00029 -0.00038 -0.00059 -0.00041 -0.00045 -0.00035 -0.00022 -0.00028 -0.00036
2,0 1 6 -0.00057 -0.00031 -0.00040 -0.00063 -0.00047 -0.00042 -0.00040 -0.000 1 2 -0.00030 -0.00027
2,520 -0.00063 -0.00034 -0.00043 -0.00058 -0.00053 -0.00051 -0.00034 -0.00029 -0.00030 -0.00037
3,024 -0.00064 -0.00035 -0.00044 -0.00065 -0.00054 -0.00048 -0.00034 -0.00029 -0.00035 -0.00041
3,528 -0.00063 -0.00034 -0.00045 -0.00063 -0.00056 -0.00047 -0.00038 -0.00032 -0.00037 -0.00038
4,032 -0.00060 -0.00035 -0.00044 -0.00057 -0.00050 -0.00053 -0.00037 -0.00029 -0.000 3 1 -0.00036
4,536 -0.00060 -0.00035 -0.00044 -0.00056 -0.00054 -0.00049 -0.00037 -0.00032 -0.00034 -0.00036
5,040 -0.00067 -0.00040 -0.00050 -0.00066 -0.00064 -0.00053 -0.00042 -0.00038 -0.00042 -0.00038
5,544 -0.00074 -0.00037 -0.00059 -0.00063 -0.00068 -0.00056 -0.00035 -0.00042 -0.0004 1 -0.00049
6,048 -0.00060 -0.00034 -0.00041 -0.00053 -0.00052 -0.00048 -0.00032 -0.00027 -0.00039 -0.00038
6,552 -0.00066 -0.00032 -0.00043 -0.00045 -0.00059 -0.00047 -0.00034 -0.00034 -0.00036 -0.00046
7,056 -0.00068 -0.00047 -0.00049 -0.00050 -0.00067 -0.00046 -0.00033 -0.00035 -0.00048 -0.00052
7,560 -0.00062 -0.00039 -0.00059 -0.00047 -0.00072 -0.00052 -0.00033 -0.00027 -0.00048 -0.00048
8,064 -0.00062 -0.00037 -0.00045 -0.00046 -0.00070 -0.00049 -0.00023 -0.00024 -0.00050 -0.00054
8,568 -0.00060 -0.00036 -0.00038 -0.00034 -0.00058 -0.00051 -0.00023 -0.00021 -0.00044 -0.0005 1
25
ANSl/IES TM-35-1 9
After calculating the relative chromaticity coordinates for each LED, the average relative chromaticity coordi nates for
each time period a re calculated. This is shown in Table D-5.
Average
I
Time, h
liu' l)y'
0 0.000000 0.000000
504 -0.000562 -0.000326
1 ,008 -0.000584 -0.000384
1 ,5 1 2 -0.0006 1 4 -0.00041 1
2,0 1 6 -0.000644 -0.000427
2,520 -0.000691 -0.000472
3,024 -0.000671 -0.000470
3,528 -0.000635 -0.000479
4,032 -0.000620 -0.000468
4,536 -0.000569 -0.000455
5,040 -0.000680 -0.000537
5,544 -0.000723 -0.000564
6,048 -0.000688 -0.000458
6,552 -0.000690 -0.000480
7,056 -0.000691 -0.000526
7,560 -0.000772 -0.000530
8,064 -0.0008 1 6 -0.000475
8,568 -0.000830 -0.00043 1
This is fol l owed by calcu lating the differential chromaticities starting at a pproximately t = 2,000 hours:
It is i m portant to note that the differential is calcu lated as a backwards d ifference. Th is uses the chromaticity data at
1 ,5 1 2 hours. The calcu lation method thus req u i res information from the time interva l before the approximately 2,000-
hour start time. In this exa m ple, although the relative chromaticities were calcu lated at 504 hours and 1 ,008 hours,
these data poi nts have no i m pact on the further calculations and may be omitted.
26
Technica l Memorandum: Projecting Long-Term Chromaticity Coordinate Shift of LED Packages, Arrays, and Modules
I
Time, h 6u'* 6v'*
0
504
1 ,008 I I
1 ,5 1 2
2,0 1 6 I -5.9524E-08 I -3.1 746E-08
2,520 -9.3254E-08 -8.9286E-08
3,024 I 3.9683 E-08 I 3.9683 E-09
3,528 7.1 429E-08 - 1 .7857E-08
4,032 I 2.9762 E-08 I 2.1 825 E-08
4,536 1 .0 1 1 9E-07 2.5794E-08
5,040 I -2.2024E-07 I - 1 .6270E-07
5,544 -8.53 1 7E-08 -5.3571 E-08
6,048 I 6.9444E-08 I 2.1 032E-07
6,552 -3.9683 E-09 -4.3651 E-08
7,056 I - 1 .9841 E-09 I -9. 1 270E-08
7,560 - 1 .6071 E-07 -7.9365 E-09
8,064 I -8.7302E-08 I 1 .09 1 3 E-07
8,568 -2.7778E-08 8.7302E-08
The differential chromaticities a re then p lotted against time, and a l inear fit to the data is calcu lated. For a set of n
experimental data poi nts (x1, y1), (x2, y2), . . . ,(x", y")' the least-squares straight line fit, where n is the total n u m ber of data
points used for cu rve fitting according to Section 5.2 .5, is:
y = mx + b ,
LY - mL x
b= n
,
2'.Y = y I + Y 2 + . . . + Y11
2
2;x 2 = X 1 2 + X 2 2 + + X11
· · ·
27
ANSl/IES TM-35-1 9
xk = tk ,
Yku' i5 u'*k ,
=
where k 1 , 2, = . n
These calcu lations a re shown g raphically in Figures D-1 and D-2. Althou g h the val ues in the preceding tables are
using rou nded values, the calculations should be carried out using as many significant fig u res as feasible.
1.5 E-07
A
1.0E-07
<> <>
<>
<>
5.0E-08
I I V I
O.OE+OO - V
�
C-0
-5.0E-08 - -
<> <>
6
-1.0E-07
<>
y = -8.5830E- 1 2x + l.4809E-08
-1.5E-07
<>
-2.0E-07
-2.5E-07
Time (hou rs)
Figure D-1 . Plot of ou'* data versus time and results of l i near fit calculations for
example LED.
2.5 E-07
0
2.0E-07
* -
-> o o -
-
5.0E-08 -
C-0 r\ -
- - v
0
O.OE+OO
-5 .0E-08
ID 2,Qbo _ �,rroo _6,o�
'-'
8,000 10, JO0
0 0
-1 .0E-07
0
-1 .5E-07
-2 .0E-07
Time (hou rs)
Figure D-2. Plot of ov'* data versus time and results of l i near fit calculations for
example LED.
28
Technica l Memorandum: Projecting Long-Term Chromaticity Coordinate Shift of LED Packages, Arrays, and Modules
The correlation coefficients for the data plotted in Figure D-1 and Figure D -2 are due to the extremely small
changes i n differential chromaticity. The differential chromaticities are used to calcu late the behavior of the relative
chromaticities which are in turn used to calculate t.u'v'.
When the differential chromaticity parameters are compared against the relative chromaticity val ues, it can be
seen that they do a reasona ble job of representing the actual data. If Figure D-3, the modeled data is based on the
d ifferential chromaticity parameters for Ou'*. While the differential chromaticity fit gave a correlation coefficient of
r2 = 0.0381 , the modeled relative chromaticity versus the actual relative chromaticity (Ou') has a much higher
correlation coefficient of r2 = 0.7922.
-0.0002
-0.0010
0 2,000 4,000 6,000 8,000 10,000
Time (h)
Figure D3. Comparison of the behavior of ou' versus modeled behavior for first
example LED.
A repeat the comparison for the behavior of ov' against the modeled behavior from the parameters calculated for ov'*
is shown in Figure D-4. It is noted that the parameters for ov'* had a correlation coefficient of r2 = 0.1 1 46. The modeled
relative chromaticity versus the actual relative chromaticity (Ov') again has a higher correlation coefficient of r2 = 0.2350.
o Data
0.0000
• Model
-0.0002 1--
� 1-- •
0
0 0
-0.0004
0 0
• • • • •
� � �
0
0
� � � .
-0.0006 -
-0.0008 I I I
0 2,000 4,000 6,000 8,000 10,000
Time (h)
Figure D-4. Comparison of the behavior ov' versus modeled behavior for first
example LED.
29
ANSl/IES TM-35-1 9
Com bining the modeled behavior for both relative chromaticity values a llows a comparison against the chromaticity
shift (Llu'v'). This is shown in Figure D-5. The correlation coefficient over the entire i nterval is r2 = 0.6357.
o Data
0. 0020 �
• Model
0.0015 >-
:>
� 0.0010 >-
0 •
� � � � � .
• 0
� �
� � . �
0. 0005 I-
0. 0000 I
0 2,000 4,000 6,000 8,000 10,000
Time (h)
Figure D-5. Comparison of behavior of chromaticity shift (.iu'v') versus modeled behavior for
first example LED.
An i nitia l report is shown in Table D-7. Chromaticity sh ift l ifetimes for the first exa mple LED a re calcu lated in Annex E.
Ta ble D-7. Initial TM-35 Report Data for First Exa mple LED.
30
Technica l Memorandum: Projecting Long-Term Chromaticity Coordinate Shift of LED Packages, Arrays, and Modules
A second example is provided to further assist users to make a comparison with their own calcu lations. The operating
conditions for the second example LED were a case temperature of 85 °C and a drive current of 1 ,000 mA. Tables
0-Sa,b and D-9a,b provide example u' and v' data, respectively, for the second example LED.
31
ANSl/IES TM-35-1 9
0 0.52 1 7 1 0.52095 0.5 1 752 0.5 1 846 0.52 1 45 0.5 1 646 0.5 1 9 1 0 0.52389 0.52429 0.52085
1 ,000 0.5 2 1 42 0.52052 0.5 1 703 0.5 1 795 0.52 1 1 3 0.5 1 596 0.5 1 867 0.52362 0.52395 0.52038
2,000 0.52 1 41 0.52043 0.5 1 695 0.5 1 783 0.52099 0.5 1 585 0.5 1 863 0.52350 0.52389 0.52039
3,000 0.52 1 25 0.52035 0.5 1 684 0.5 1 774 0.52091 0.5 1 573 0.5 1 85 1 0.52346 0.52382 0.520 3 1
4,000 0.52 1 24 0.52027 0.5 1 673 0.5 1 770 0.52080 0.5 1 566 0.5 1 847 0.52333 0.52377 0.520 1 6
5,000 0.52 1 1 5 0.52020 0.5 1 658 0.5 1 767 0.52075 0.5 1 562 0.5 1 844 0.52322 0.52365 0.5 1 999
6,000 0.52 1 1 4 0.520 1 3 0.5 1 66 1 0.5 1 768 0.52078 0.5 1 542 0.5 1 8 1 8 0.523 1 4 0.52356 0.5 1 992
7,000 0.52 1 02 0.52007 0.5 1 66 1 0.5 1 773 0.52073 0.5 1 537 0.5 1 832 0.52309 0.52354 0.5 1 99 1
The first step in the calcu lations is calcu lating the relative chromaticity coordi nates for t h e each L E D a t each time
interval. The relative chromaticity coordinates are calcu lated using:
Ta ble D-1 0a. llu' Relative Chromaticity Coordinates for the Second Example LED.
liu' Sample #
Time, h 1 2 3 4 5 6 7 8 9 10
0 0.24680 0.24604 0.24549 0.24675 0.24620 0.246 1 0 0.2461 0 0.24607 0.24778 0.2465 1
1 ,000 0.24628 0.24555 0.24489 0.24622 0.24566 0.24562 0.24566 0.24555 0.24723 0.24590
2,000 0.24620 0.24546 0.24485 0.2461 4 0.2456 1 0.24554 0.24562 0.2455 1 0.247 1 7 0.24578
3,000 0.246 1 7 0.24538 0.24480 0.24607 0.24557 0.2455 1 0.24556 0.24544 0.24709 0.24575
4,000 0.246 1 6 0.24548 0.24485 0.24609 0.24555 0.24552 0.24559 0.2455 1 0.247 1 2 0.24581
5,000 0.246 1 2 0.24546 0.24480 0.24604 0.24553 0.24556 0.24557 0.24550 0.247 1 0 0.24575
6,000 0.24623 0.24555 0.24485 0.246 1 7 0.24565 0.24561 0.24567 0.24563 0.24722 0.24586
7,000 0.2461 8 0.24540 0.24487 0.24608 0.2457 1 0.24543 0.24547 0.24560 0.247 1 5 0.24583
Table D-10b. llu' Relative Chromaticity Coordinates for the Second Sample LED.
liu' Sample #
Time, h 11 12 13 14 15 16 17 18 19 20
0 0.24680 0.24604 0.24549 0.24675 0.24620 0.246 1 0 0.2461 0 0.24607 0.24778 0.2465 1
1 ,000 0.24628 0.24555 0.24489 0.24622 0.24566 0.24562 0.24566 0.24555 0.24723 0.24590
2,000 0.24620 0.24546 0.24485 0.2461 4 0.2456 1 0.24554 0.24562 0.2455 1 0.247 1 7 0.24578
3,000 0.246 1 7 0.24538 0.24480 0.24607 0.24557 0.2455 1 0.24556 0.24544 0.24709 0.24575
4,000 0.246 1 6 0.24548 0.24485 0.24609 0.24555 0.24552 0.24559 0.2455 1 0.247 1 2 0.24581
5,000 0.246 1 2 0.24546 0.24480 0.24604 0.24553 0.24556 0.24557 0.24550 0.247 1 0 0.24575
6,000 0.24623 0.24555 0.24485 0.246 1 7 0.24565 0.24561 0.24567 0.24563 0.24722 0.24586
7,000 0.2461 8 0.24540 0.24487 0.24608 0.2457 1 0.24543 0.24547 0.24560 0.247 1 5 0.24583
32
Technica l Memorandum: Projecting Long-Term Chromaticity Coordinate Shift of LED Packages, Arrays, and Modules
Ta ble D-1 1 a . ov' Relative Chromaticity Coordinates for the Second Sample LED.
6v' Sample #
Time, h 1 2 3 4 5 6 7 8 9 10
0 0.52085 0.52 1 5 5 0.5 1 707 0.5 1 780 0.5 2 1 77 0.5 1 793 0.5 1 98 1 0.52282 0.52260 0.52224
1 ,000 0.52026 0.52 1 26 0.5 1 655 0.5 1 7 1 6 0.5 2 1 3 9 0.5 1 748 0.5 1 946 0.52257 0.52224 0.52 1 73
2,000 0.52024 0.52 1 20 0.5 1 649 0.5 1 707 0.521 28 0.5 1 742 0.5 1 935 0.52247 0.522 1 8 0.52 1 66
3,000 0.520 1 6 0.52 1 1 3 0.5 1 638 0.5 1 697 0.52 1 1 6 0.5 1 73 1 0.5 1 934 0.52243 0.52207 0.52 1 6 1
4,000 0.520 1 1 0.521 06 0.5 1 625 0.5 1 693 0.52 1 1 0 0.5 1 722 0.5 1 928 0.52227 0.52201 0.52 1 49
5,000 0.52002 0.5 2 1 06 0.5 1 620 0.5 1 698 0.52097 0.5 1 7 1 5 0.5 1 925 0.522 1 8 0.52 1 94 0.52 1 38
6,000 0.5 1 995 0.52095 0.5 1 623 0.5 1 687 0.52094 0.5 1 700 0.5 1 925 0.522 1 6 0.52 1 87 0.5 2 1 38
7,000 0.52006 0.52086 0.5 1 640 0.5 1 690 0.52 1 03 0.5 1 693 0.5 1 9 1 9 0.52222 0.5 2 1 87 0.5 2 1 47
Ta ble D-1 1 b. ov' Relative Chromaticity Coordinates for Second Exa mple LED.
6v' Sample #
Time, h 11 12 13 14 15 16 17 18 19 20
0 0.52 1 7 1 0.52095 0.5 1 752 0.5 1 846 0.52 1 45 0.5 1 646 0.5 1 91 0 0.52389 0.52429 0.52085
1 ,000 0.52 1 42 0.52052 0.5 1 703 0.5 1 795 0.52 1 1 3 0.5 1 596 0.5 1 867 0.52362 0.52395 0.52038
2,000 0.52 1 41 0.52043 0.5 1 695 0.5 1 783 0.52099 0.5 1 585 0.5 1 863 0.52350 0.52389 0.52039
3,000 0.52 1 25 0.52035 0.5 1 684 0.5 1 774 0.52091 0.5 1 573 0.5 1 85 1 0.52346 0.52382 0.52031
4,000 0.52 1 24 0.52027 0.5 1 673 0.5 1 770 0.52080 0.5 1 566 0.5 1 847 0.52333 0.52377 0.520 1 6
5,000 0.52 1 1 5 0.52020 0.5 1 658 0.5 1 767 0.52075 0.5 1 562 0.5 1 844 0.52322 0.52365 0.5 1 999
6,000 0.52 1 1 4 0.5201 3 0.5 1 66 1 0.5 1 768 0.52078 0.5 1 542 0.5 1 8 1 8 0.523 1 4 0.52356 0.5 1 992
7,000 0.52 1 02 0.52007 0.5 1 66 1 0.5 1 773 0.52073 0.5 1 537 0.5 1 832 0.52309 0.52354 0.5 1 99 1
After calcu lating the relative chromaticity coordinates for each LED, the averaged relative chromaticity coordinates for
each time period a re calcu lated. This is shown in Table D-1 2.
Average
I
Time, h
6u' 6v '
-
0 0.000000 0.000000
1 ,000 -0.0005 1 2 -0.000420
2,000 -0.000591 -0.000495
3,000 -0.000634 -0.000582
4,000 -0.000605 -0.000664
5,000 -0.000623 -0.000736
6,000 -0.000571 -0.000798
7,000 -0.000622 -0.000790
33
ANSl/IES TM-35-19
This is fol l owed by calcu lating the differential chromaticities starting at a pproximately t = 2,000 hours:
Average
I
Time, h
l> u' l)y'
0
1 ,000
2,000 -7.9500E-08 -7.SOOOE-08
3,000 -4.2500E-08 -8.7500E-08
4,000 2.9000E-08 -8. l 500E-08
5,000 - 1 .8000E-08 -7 .2500E-08
6,000 5.1 500E-08 -6.2000E-08
7,000 -5.0500E-08 8.0000E-09
The differential chromaticities a re then p lotted against time, a n d a l inear fit to the data is calculated. For a set of n
experimental data poi nts (x1, y1), (x2, y2), . . . ,(x,,, y,,), the least-squares straight line fit, where n is the total n u m ber of data
points used for cu rve fitting according to Section 5.2 .5, is:
y = mx + b ,
LY - m:Lx
b= n
'
2;y = Y I + Y 2 + " ·+ Yn
2
2;x = X 1 2 + X 2 2 + . . . + Xn 2
34
Technica l Memorandum: Projecting Long-Term Chromaticity Coordinate Shift of LED Packages, Arrays, and Modules
xk = tk '
Yku' = /5 u'*k ,
Ykv' = (j v'*k
where k = 1 , 2, . . ,
I
. n
The results of these calcu lations are shown g raphica l l y in Figures D-6 and D-7.
6.0E-08
0
4.0E-08
0
2.0E-08 ....
. . . · ··
O.OE+OO ··
*
. . . . . . . . · · ·o·_
· ·_ _·
.
(!) 2,000 4,000 . ·· · ·6:��0 8, 00
� -2 . 0E-08 . . .
.
.. . . .
-4.0E-08 1----- .. . . . . . · ·� ____ _
0
.
y=
-6.0E-08
l.08 57E-llx - 6.7190E-08
-8.0E-08 0
-l.OE-07
.
Time (h)
Figure D-6. Plot of ou'* data versus time a n d results of l i near fit calculations for
exa mple LED
2.0E-08
O. OE+OO I
2,000 4,000 6,000
I 0
8,000
-2.0E-08
··
:
!
*
·· · ·· ·· ·
:
-4.0E-08
->
·
c.o
-6.0E-08
___
. .
.
. .. .- - � -- - - - · · ·· · ·
.
....
-8.0E-08 . -·
O y=
o'
. / ,. /
f------
1.4 300E-llx - l.2610E-07
-l.OE-07 .
-l.2E-07 1 1
.
.
Time (h)
Figure D-7 Plot of ov'* data versus time and results of l i near fit calculations for
exa mple LED
35
ANSl/IES TM-35-1 9
An initial report is shown in Table D-14. Chromaticity sh ift l ifeti mes for the second exa m p l e LED are calcu lated i n
Annex E.
Annex E
Extrapolation Methods
Extra polating ch romaticity s hift behavior forward in time is a m atter of n u merical i nteg ration. For this procedu re,
the relative ch romaticity coord i nates for the final time interva l used in the l i near-fit calcu lations a re used as a
reference point.
The va lues for the differential chromaticities are used along with the average time i nterval of the dataset. The first step
is to enter the time and relative chromaticities of the fi nal data point in the test set into a table, as shown in Table E-1 .
Table E-1 . Initial Values in the Chromaticity Shift Extrapolation Table for First Example LED
I I I I I
I I I I I
36
Technica l Memorandum: Projecting Long-Term Chromaticity Coordinate Shift of LED Packages, Arrays, and Modules
Differential chromaticity analysis (DCA; see Annex C) assumes that d ifferential chromaticities are a function of
time. Therefore, using the slopes and intercepts for each differential chromaticity, it is then possible to calcu late the
differential chromaticity for a ny a rbitrary time in the futu re. Thus,
becomes:
8u'*(8,568)
X
=
Similarly,
becomes:
This a l l ows the first row in the chromaticity shift extrapolation ta ble to be completed (see Table E-2).
The first three col u m ns of the second row of the chromaticity sh ift extra polation table may now be popu lated. The
time is incremented by the average flt for the test dataset, and new va l ues of 6u' and 6v ' a re calcu lated:
Ta ble E-2. Completed Fi rst Row for the Chromaticity Shift Extrapolation Table for Fi rst
Example LED
I I I I I
I I I I I
8u '(t + flt) = 8u '(t) + 8u'*(t)
8u'(8,568+504) = 8u'(8,568) + 8u'*(8,568)
8u '(9,072) = -0.000830 - 2.9600 X 10-5 = -0.000860
37
ANSl/IES TM-35-1 9
Simila rly:
Ta ble E-3. Chromaticity Shift Extrapolation Ta ble after Calculations for the First Time
Step for Fi rst Example LED
I I I I I
I I I I I
Repeating this process for the second time step g ives the results provided in Table E-4.
Ta ble E-4. Chromaticity Shift Extra polation Table after Calculations for the Second Time
Step for Fi rst Example LED
The p rojected behavior for the first example LED is shown g raph ical ly i n Figure E-1 . The time scale was tru ncated at
35,000 hours, si nce the CS? is projected to happen at a n earlier time.
38
Technica l Memorandum: Projecting Long-Term Chromaticity Coordinate Shift of LED Packages, Arrays, and Modules
0.008
0.007
T
-· ·-!----·
0.006
.Iv
0.005
<I
J>
0.004 i-----
-·-- · f-·---- /
-i
::I -·---·- ·
-
LM·80 data
7
0.003
o I /
- Extrapolated behavior I
0.002 JV
0.001
---
�
--
�
0
0 5,000 10,000 15,000 20,000 25,000 30,000 35,000
Time (hours)
Figure E-1 . Projected chromaticity shift behavior for the first example LED. LM-80 data points
are incl uded for reference. Horizontal dotted l i nes i n d icate .iu'v' = 0.004 and .iu'v' = 0.007.
As seen in Table E-5, CS4 is bracketed by the entries for 26,208 hours and 26,7 1 2 hours. Linea r behavior is assumed
between these time data poi nts. The slope and intercept a re calcu lated by:
y = mx + b
x
L1u'v'2 - L1u v 1
f2- f1
= 0.0041 31 1 1 - 0.00395354 = 0.00017757 10-1
' '
m= = 3 . 5253
26,208 - 25,704 504
b = L1u'v'i - m t 1 = 0.00395354 - 3.5253 x 10-7 x 25,704 = -0.0051025
0.004 - b 0.004 + 0.00 51025 hours
3.5253 X 10-7
tCS4 = m
=
25 836
'
Ta ble E-5. Chromaticity Sh ift Extra polation Table for Fi rst Example LED Showing Entries
32,256
I ...
-0.004578
I ...
0.004792
I ...
0.006627 1 6
I ...
-1 .3207E-04
I ...
2.0269E-04
32,760 -0.0047 1 0 0.004994 0.00686509 -1 .3425E-04 2.065 1 E-04
33,264 -0.004844 0.005201 0.0071 0757 -1 .3643E-04 2.1 032E-04
33,768 -0.004981 0.00541 1 0.00735459 -1 .3861 E-04 2 . 1 4 1 4E-04
39
ANSl/IES TM-35-1 9
Fol l owing the same proced u re, the time to reach CS? is calcu lated as 33,040 hours, which is rou nded to 33,000 hours.
With these calcu lations, the table portion of the TM-35 Report ca n now be completed (see Table E-6).
The va l ues for the ou' a n ov' can be calculated using the n u merical extra polation and interpolation procedure above
to calcu late the val ues needed for the vector g raphic (see Table E-7 and Figure E-2).
6u ' 6v'
CS4 -0.003085 0.002541
CS? -0.004785 0.0051 09
40
Technica l Memorandum: Projecting Long-Term Chromaticity Coordinate Shift of LED Packages, Arrays, and Modules
..... --.- �·
..... --,,......, .
.... __.
-
CS7
- .... . _ -1.J
·-
'""� - .... v . _.
�'
--....
CS4
.....
v . .... .... ....
� ..
' �
.... --.....
-v. .... vL
8 u'
Figure E-2. Vector graphic for first example LED.
Repeating the process again for the second example LED, a table is created using the data for the final time i nterva l .
This data includes t h e elapsed time, t h e average relative chromaticity va l ues (Ou', ov') a t t h e f i n a l t i m e interva l a n d the
chromaticity shift (�u'v') is shown in Table E-8.
Ta ble E-8. Initial Values i n the Chromaticity Shift Extrapolation Ta ble for the Second
Example LED
I I I I I
I I I I I
As stated previously, d ifferential chromaticity analysis (DCA; see Annex C) assumes that differential chromaticities
are a function of time. This a l l ows the differential chromaticity va l ues to be calcu lated for a ny a rbitrary time using the
sl ope and i ntercept for each d ifferential chromaticity. Thus,
becomes:
41
ANSl/IES TM-35-1 9
8u'*(7,000) =
Similarly,
becomes:
8v '*(7,000)
-2.6000 X 10-5
=
8v '*(7,000) =
This a l l ows the first row in the chromaticity shift extrapolation ta ble to be completed (see Table E-9).
Ta ble E-9. Completed F i rst Row for the Chromaticity Shift Extra polation Table for First
Example LED
I I I I I
I I I I I
The first three col u m ns of the second row of the chromaticity sh ift extra polation table may now be popu lated. The
time is i ncremented by the average fit for the test dataset, and new val ues of ou' and ov' a re calculated:
Similarly:
42
Technica l Memorandum: Projecting Long-Term Chromaticity Coordinate Shift of LED Packages, Arrays, and Modules
Ta ble E-10. Chromaticity Shift Extra polation Table after Calculations for the Fi rst Time
Step for the Second Example LED
I I I I I
I I I I I
Repeating this process for the second time step g ives the resu lts provided in Table E-1 1 .
Ta ble E-1 1 . Chromaticity Shift Extra polation Table after Calculations for the Second Time
Step for the Second Example LED
The p rojected behavior for the second example LED is shown g raphically i n Figure E-3. The time sca l e was tru ncated
at 35,000 hours, si nce this is the extrapolation l i m it for this dataset.
0.008
0.007
0.006
0.005 /
<l
:>
::J 0.004 /
LM-80 data �v
0.003 I
H
o
- Extrapolated behavior I
0.002 /
v
0.001 /
0 0 u
0 ----
� l/"
0
-
Figure E-3. Projected Chromaticity Shift behavior for the second example LED. LM-80 data
points are included for reference. Horizontal dotted l ines ind icate .iu'v' = 0.004 and 0.007.
43
ANSl/IES TM-35-1 9
A seen in Table E-1 2, CS4 is bracketed by the entries for 31 ,000 hours and 32,000 hours. Linear behavior is assumed
between these time data poi nts. The slope and intercept are ca l c u lated by:
y = mx + b
x
Llu'v' - Llu'v' 1 0.00403463 - 0.00362005 0.00041458 = .
m = 2
t - ti 4 1458 10-7
2 32,000 31 ,000
- 1 ,000
Ta ble E-1 2 . Chromaticity Shift Extra polation Table for the Second Example LED showing
Entries i n the Vicinity of CS4 and CS7
37,000
I ...
0.004366
I ...
0.00465 1
I . ..
0.00637850
I ...
3.3452E-04
I ...
4.0300E-04
38,000 0.004700 0.005054 0.00690 135 3.4538E-04 4.1 730E-04
3 9,000 0.005045 0.005471 0.00744220 3.5623 E-04 4.3 1 60E-04
40,000 0.005402 0.005902 0.00800 1 04 3.6709E-04 4.4590E-04
44
Technica l Memorandum: Projecting Long-Term Chromaticity Coordinate Shift of LED Packages, Arrays, and Modules
Following the same procedu re, the time to reach CS? is calculated as 38,1 82 hours, which exceeds the a l l owable
extrapo lation l i m it for this dataset. CS? shall therefore be reported as CS? > 35,000 hou rs.
With these calcu lations, the table portion of the TM-35 Report can now be com pleted (see Table E-13).
Table E-1 3. Full TM-35 Report for the Second Example LED
The va l ues for ou' and ov' can be calculated using the nu merical extrapolation and i nterpolation procedure above to
calcu late the va l ues needed for the vector g raphic (see Table E-14 and Figure E-4). Since CS? exceeds the a llowable
extra polation l im its for this dataset, no vector information shall be calculated for CS?.
0.005
CS4 0.00361 1 0.003739
CS4
CS? N/A N/A
0.004
->
c,o /
0.003
0.002 v
0.001 v
0 I/
0 0.001 0.002 0.003 0.004 0.005 0.006 0.007
ou'
Fig u re E-2. Vector g raphic for first example LED.
45
ANSl/IES TM-35-1 9
ADDITIONAL READ I N G
Bergman R , Sum mary o f NVLAP Round 1 5 Proficiency Test, CORM 201 8 Annual Conference and Business Meeting,
Gaithersburg, M D, J u ly 29 - August 1, 201 8.
Davis J L et al. Chromaticity maintenance in LED devices. In: van O riel W, Fan X, and Zhang GQ, editors. Solid State
Lighting Reliability: Components to Systems, Vol. 2. New York: Springer; 201 7.
Davis J L, Young J, Royer M, CALi PER Report 20.5: Chromaticity Shift Modes of LED PAR38 Lamps Operated in Steady
State Conditions, Pacific Northwest Nationa l Laboratory, 201 6.
Mehr YM, va n Oriel WD, Zhang GQ, Progress in U nderstanding Co lor Maintenance in Solid-State Lighting Systems,
Engineering, Vol. 1 , No.2 pp170-1 78, 201 5.
Royer MP, Tuttle R, Rosenfeld S, Miller NJ, Color Maintenance in Laboratory and Field Applications, Pacific Northwest
National Laboratory, 201 3.
Wag ner M, Herzog HG, Kha nh TQ, Lifetime Calculation of White H P-LEDs from 1 6,000 Hours Aging Data. In Luger
Research e.U., Issue 59 pp34-38, 201 7.
46
Order# ANSl/IES
ISBN# 978-0-87995-082-8