apmp supplementary comparisons of led measurements · apmp supplementary comparisons of led...

APMP Supplementary Comparisons of

LED Measurements

APMP.PR-S3a Averaged LED Intensity

APMP.PR-S3b Total Luminous Flux of LEDs

APMP.PR-S3c Emitted Colour of LEDs

Final Report (July 2012)

Dong-Hoon Lee, Seongchong Park, and Seung-Nam Park

Division of Physical Metrology, Korea Research Institute of Standards and Science (KRISS)

1 Doryong-Dong, Yuseong-Gu, Daejeon 304-340, Rep. Korea

Correspondance to: [email protected]

APMP.PR-S3b Total Luminous Flux of LEDs Final Report

2

Table of Contents

1. Introduction ...................................................................................................................................................... 5

2. Comparison Protocol .................................................................................................................................... 5

3. Arttifact LEDs ................................................................................................................................................... 7

4. Measurement Capabilities of Participants........................................................................................... 9

4.1. KRISS .......................................................................................................................................................... 9

4.2. MIKES ...................................................................................................................................................... 13

4.3. CMS-ITRI ................................................................................................................................................ 21

4.4. PTB ........................................................................................................................................................... 28

4.5. NMIJ ......................................................................................................................................................... 34

4.6. CENAM ................................................................................................................................................... 41

4.7. LNE ........................................................................................................................................................... 48

4.8. METAS ..................................................................................................................................................... 58

4.9. NMC-A*STAR ....................................................................................................................................... 68

4.10. VSL ....................................................................................................................................................... 72

4.11. NIST ..................................................................................................................................................... 79

4.12. VNIIOFI ............................................................................................................................................... 89

4.13. INM ...................................................................................................................................................... 89

5. Reported Results of Participants .......................................................................................................... 96

5.1. KRISS ....................................................................................................................................................... 96

5.2. MIKES ...................................................................................................................................................... 98

5.3. CMS-ITRI ................................................................................................................................................ 99

5.4. PTB ........................................................................................................................................................... 99

5.5. NMIJ ....................................................................................................................................................... 100

5.6. CENAM ................................................................................................................................................. 100

5.7. LNE ......................................................................................................................................................... 101


3

5.8. METAS ................................................................................................................................................... 101

5.9. NMC-A*STAR ..................................................................................................................................... 102

5.10. VSL ..................................................................................................................................................... 102

5.11. NIST ................................................................................................................................................... 103

5.12. VNIIOFI ............................................................................................................................................. 103

5.13. INM .................................................................................................................................................... 104

6. Pre-draft A Process .................................................................................................................................. 104

6.1. Verification of Reported Results ............................................................................................... 105

6.2. Temperature Correction and Artifact Drift ........................................................................... 105

6.3. Review of Relative Data ................................................................................................................ 113

6.4. Review of Uncertainty Budgets ................................................................................................. 114

6.5. Identification of Outliers ............................................................................................................... 114

7. Data Analysis ............................................................................................................................................... 115

7.1. Calculation of Difference to Pilot ............................................................................................. 115

7.2. Calculation of Comparison Reference Value ....................................................................... 116

7.3. Calculation of Degree of Equivalence .................................................................................... 117

7.4. Data Analysis Spreadsheet .......................................................................................................... 117

8. Comparison Results ................................................................................................................................. 118

8.1. Red LEDs .............................................................................................................................................. 118

8.2. Green LEDs ......................................................................................................................................... 120

8.3. Blue LEDs ............................................................................................................................................. 121

8.4. White LEDs.......................................................................................................................................... 123

9. Discussion ..................................................................................................................................................... 126

9.1. Test of Consistency ......................................................................................................................... 126

9.2. Accuracy of Color Correction ..................................................................................................... 126

10. Summary .................................................................................................................................................. 129


4

Acknowledgement ............................................................................................................................................. 129

Appendix A: Technical Protocol ................................................................................................................... 130

Appendix B: Review of Relative Data ........................................................................................................ 131

Appendix C: Comments from Review of Relative Data .................................................................... 132

Appendix D: Comments from Review of Uncertainty Budgets ..................................................... 133

Appendix E: Identification of Outliers ....................................................................................................... 134

Appendix F: Comments and Revision to Draft A Report ................................................................. 135


5

1. Introduction

With the recent growth of the solid state lighting and display industry, the interest and

importance of accurate measurement of light-emitting diodes (LEDs) are increasing.

Photometric measurement of LEDs, however, is influenced by the specific properties of

individual LED such as spectral distribution, spatial emission profile, temperature

dependence, etc. In general, the measurement uncertainty of LEDs is larger than that of

the conventional incandescent lamps, and greater care is required to avoid or correct the

systematic errors related to the LED properties.

The Asia Pacific Metrology Programme (APMP) Technical Committee of Photometry

and Radiometry (TCPR) decided at its meeting in December 2006 to conduct

supplementary comparisons on measurement of LEDs to test the metrological

equivalence among national metrology institutes (NMIs) under the CIPM Mutual

Recognition Arrangement (MRA)1. The participation was not limited to NMIs in APMP, but

also NMIs of other regional metrology organizations (RMOs). The Korea Research

Institute of Standards and Science (KRISS) of Republic Korea is designated as the pilot

laboratory.

Three measurement quantities of LEDs are selected for the comparisons, which are

listed as service categories for Calibration and Measurement Capabilities (CMCs):

averaged LED intensity in condition B defined by International Commission on

Illumination (CIE) 2 , total luminous flux, and emitted color expressed as chromaticity

coordinates (x, y) according to the CIE 1931 standard colorimetric system3. The three

comparisons are registered as APMP.PR-S3a, -S3b, and -S3c, respectively.

In this report, we summarize the results of the comparison S3b on total luminous

flux of LEDs.

2. Comparison Protocol

The organization, the artifact LEDs, and the guidelines for measurement and report of all

the three comparisons (S3a, S3b, S3c) are settled on one technical protocol before the

start of the comparisons. The protocol is drafted by the pilot lab, agreed by the

participants, and approved by the APMP TCPR in January 2008. The protocol is once

revised in November 2008, as the INM of Romania has joined as an additional participant.

1 http://www.bipm.org/en/cipm-mra/ 2 Measurement of LEDs, 2nd edition, CIE Technical Report 127-2007. 3 Colorimetry, 3rd edition, CIE 015:2004.


6

The final version of the technical protocol is included in 오류! 참조 원본을 찾을 수

없습니다. as an electronic file. Table 2-1 shows the final list of participants to the S3b

comparison with the measurement schedules planned and performed. We note that the

NPL of the UK listed on the technical protocol has withdrawn its participation in August

2009.

Table 2-1. List of participants and measurement schedules of APMP.PR-S3b.

NMI country contact person(s) measurement

planned LED set

measurement

performed

results

reported

KRISS

(pilot) Korea

Seongchong Park,

Dong-Hoon Lee -- -- -- --

NMC-

A*STAR

Singapore Yuanjie Liu,

Gan Xu

June ~ Aug.

2008 #8

10 July ~ 28 Aug.

2008

12 Jan.

2009

MIKES Finland (Pasi Manninen),

Tuomas Poikonen,

March ~ May

2008 #1

7 April ~ 13 April

2008

17 June

2008

NIST USA

Cameron Miller,

Yoshi Ohno,

Yuqin Zong

Aug. ~ Oct.

2008 #3

18 Feb. ~ 25 Feb.

2009

31 July

2009

CMS-

ITRI

Chinese

Taipei Cheng-Hsien Chen

March ~ May

2008 #2

26 May 2008 ~ 2

Oct. 2009*

26 Oct

2009

PTB Germany

Matthias

Lindemann,

Robert Maass

April ~ June

2008 #3 May ~ July 2008

18 July

2009

CENAM Mexico

Laura P. González,

Anayansi Estrada,

Eric Rosas

May ~ July

2008 #5

17 July ~ 21 July

2008

08 May

2009

NMIJ Japan Kenji Godo,

(Terubumi Saito)

April ~ June

2008 #4

17 April ~ 22

June 2008

01 Aug.

2008

METAS Switzerland Peter Blattner June ~ Aug.

2008 #7

08 Sept ~ 17 Sept

2008

07 April

2009

LNE France Jimmy Dubard May ~ July

2008 #6

15 June ~ 13 July

2008

15 April

2009

VSL The

Netherlands

(Eric van der Ham),

M. Charl Moolman,

Daniel Bos

July ~ Sept.

2008 #1

13 Oct 2008 ~ 12

Jan 2009

1 Oct

2009

VNIIOFI Russia Tatiana Gorshkova,

Stanislav Shirokov

Sept. ~ Nov.

2008 #5

28 Nov ~ 05 Dec

2008

06 Feb.

2009

INM Romania Mihai Simionescu Nov. ~ Dec.

2008 #7 Dec 2008

30 March

2009

* The CMS-ITRI had the initial measurement in May 2008, but it had to repeat the measurement on the red

LEDs in Oct 2009 due to damages in the initial measurement.

The comparison was performed as a star-type circulation of multiple sets of artifact

LEDs. The round for each participant had the following sequence: (1) first measurement

by the pilot, (2) measurement by the participant, (3) second measurement by the pilot.

The results of the repeated measurement by the pilot are used to evaluate the stability of

the artifact LEDs.


7

3. Arttifact LEDs

Five different types of LEDs are used as comparison artifacts: RED (Nichia model

NSPR518S), GREEN (Nichia model NSPG518S), BLUE (Nichia model NSPB518S), WHITE

(Nichia model NSPW515BS), and DIFFUSER-TYPE GREEN (NSPG518S mounted in a

cylinder-type cap with an opal diffuser). All the bare LEDs had a lamp diameter of 5 mm

and were to be operated at a forward direct current of 20 mA. The detailed information

of the LEDs is included in the technical protocol (Appendix A). Note, however, the

diffuser-type green LEDs are not measured for the comparison S3b.

Each set of artifact LEDs consisted of three pieces of the red (R), green (G), blue (B),

and white (W) LEDs and two pieces of the diffuser-type green (D) LEDs. They were

packaged and identified as shown in Fig. 3-1. The pilot prepared eight sets of artifact

LEDs for the LED comparisons S3a, S3b, and S3c. Each artifact LED is designated in a

form #N-X-M with three codes:

- #N as the artifact set number: N = 1, 2, …, 8

- X as LED color and type code: X = R for red, G for green, B for blue, W for white, D for

diffuser-type green

- M as sample serial number for each type: M = 1, 2, 3

Fig. 3-1. Artifact LED set circulated in the LED comparisons S3a, S3b, and S3c.


8

The artifact LEDs are prepared based on the functional seasoning 4 that records

during the pre-burning the relative change of luminous intensity and spectral distribution

of each individual LED together with its junction voltage under the ambient temperature

periodically varied from 18 °C to 33 °C. From the recorded data, the temporal drift and

the temperature dependence of the optical characteristics of each LED could be

separately determined. Each artifact LEDs has passed a seasoning procedure over 300

hours.

Since the photometric properties of LEDs have a very high dependence upon

temperature, their comparison requires a sensitive control or monitoring of the junction

temperature. As the junction voltage Vj of a LED can be approximated as a linear

function of the junction temperature T in a small interval, say ±10 °C, around a reference

temperature of T0,5 we can model the temperature dependence of its total luminous flux

ΦLED as a third-order polynomial with three coefficients:

2 3

0 0 0

0

1 ( ) ( ) ( ) ( ) ( ) ( )LED

j j j j j j

LED

Ta V T V T b V T V T c V T V T

T

. (3-1)

The coefficients a, b, and c of each artifact LED could be determined by fitting the

function of Eq. (3-1) to the functional seasoning data. With these results, the pilot was

capable to calculate a temperature correction factor for the measurement result of any

artifact LED to the same measurement condition, as long as the junction voltage at the

time of measurement is known. The uncertainty of this correction factor is estimated to

be less than 0.5 % as a relative standard uncertainty from the goodness of fit for the

coefficients.

In the comparison S3b, the measurement condition was specified with an ambient

temperature of 25 °C. In addition, the junction voltage of each LED was to be recorded

to monitor the junction temperature and to apply the aforementioned temperature

correction. In the chapters 오류! 참조 원본을 찾을 수 없습니다.~오류! 참조 원본을

찾을 수 없습니다., we will show and discuss this effect of the temperature correction to

the comparison results.

4 Seongchong Park et al., Metrologia 43, 299 (2006). 5 See, for example, E. F. Schubert, Light-Emitting Diodes (Cambrige University Press, 2003)


9


10

4. Measurement Capabilities of Participants

In this chapter, we summarize the information on measurement capabilities and

uncertainty budgets for total luminous flux of LEDs, which are reported by each

participant.

4.1. KRISS

4.1.1. Measurement setup

Fig. 4-1 shows the measurement setup of total luminous flux in KRISS. This setup is

implemented in a similar way to the NIST absolute integrating sphere method. The

integrating sphere has a diameter of 300 mm. There are 2 photometers: one (photometer

#1) is located outside the sphere for luminous flux measurement of a collimated

reference beam, and the other one (photometer #2) is attached to the sphere surface,

which acts a comparator of the illuminance between the reference beam and an LED. The

photometer #1 has a diameter of 15 mm (P15F0T made by LMT), and the photometer #2

has an aperture of 1 cm2 (P11S0Ts made by LMT).

For spectral mismatch correction, we use a CCD-mounted spectrograph-type

spectroradiometer (CAS140CT-153 made by Instrument Systems), of which the input

optics is composed of an 1.5” integrating sphere and fiber bundle. The aperture area of

the integrating sphere is 1 cm2. It covers 380 nm to 1050 nm, and its spectral bandwidth

(FWHM) is about 3 nm at 633 nm. The photometer #2 can be substituted by the

spectroradiometer input optics. Other geometry is shown in the right-side of Fig. 4-1.

The LED is driven by a source-meter unit (2400 Sourcemeter made by Keithley),

which provides both of current sourcing and voltage measuring function. The LED is

connected to the source-meter unit using 4-wire connection.


11

Fig. 4-1. LED total luminous flux measurement setup in KRISS.

4.1.2. Mounting and alignment

Normally, the LED holder is positioned as the right-side of Fig. 4-1, thus the LED tip is

aimed at 115° from z-axis. For spatial response distribution measurement, we use

another LED holder with an LED beam source, which enables to adjust the aiming angle

over nearly 4 solid angle. Based on the SRDF measurement, the spatial mismatch

correction is performed.

4.1.3. Traceability

The absolute spectral responsivity of photometer #1 and the relative spectral responsivity

of photometer #2 are calibrated using a KRISS working standard photodiode. The scale is

traceable to KRISS cryogenic radiometer. For the spectroradiometer, the relative spectral

responsivity is calibrated using a spectral irradiance standard lamp traceable to NIST

spectral irradiance scale.

4.1.4. Measurement uncertainty

Tables in the following are the detailed uncertainty budgets of total luminous flux

measurement for the LEDs used in this APMP LED comparison. The uncertainty

evaluation is carried out according to Guide to the Expression of Uncertainty in

Measurement (GUM). Expanded uncertainty are evaluated at a confidence level of

approximately 95% with a coverage factor normally k = 2. Table 4-5 is the detailed

uncertainty budget of the junction voltage measurement.

Table 4-1. KRISS uncertainty budget of total luminous flux measurement for red LEDs (R).

baffle

Collimated

QTH Lamp

photometer 1

photometer 2

Linear stage

Integratingsphere

baffle

Collimated

QTH Lamp

photometer 1

photometer 2

Linear stage

Integratingsphere

z

40

35

y

65

x

REF. beam

test LED

z

40

35

y

65

x

REF. beam

test LED


12

Uncertainty Component Standard

uncertaint

y Ty

pe Probability

distribution

Sensitivity

coefficient

Contribut

ion (%)

DoF Correl

ated?

sphere photometer repeatability

(DUT)

0.00 % A t 1 0.00 9 N

current feeding accuracy 0.05 % B rectangular 1 0.05 Y

near field reflection loss 0.50 % B rectangular 1 0.50 Y

external photometer repeatability

(REF)

0.00 % A t 1 0.00 9 N


(REF)

0.00 % A t 1 0.00 9 N

external photometer linearity 0.05 % B rectangular 1 0.05 Y

sphere photometer linearity 0.05 % B rectangular 1 0.05 Y

transfer procedure repeatability 0.01 % A t 1 0.01 9 N

spatial mismatch correction 0.75 % B normal 1 0.75 Y

luminous flux responsivity 0.46 % B normal 1 0.46 Y

stray light 0.20 % B rectangular 1 0.20 Y

color correction 0.24 % B normal 1 0.24 Y

reproducibility 0.33 % A t 1 0.33 >30 N

Combined standard

uncertainty (%)

normal 1.11 >20

Table 4-2. KRISS uncertainty budget of total luminous flux measurement for green LEDs (G).


uncertain

ty Ty

pe Probability

distribution

Sensitivity

coefficient

Contribut

ion (%)

DoF Correl

ated?


(DUT)

0.00 % A t 1 0.00 9 N




(REF)

0.00 % A t 1 0.00 9 N


(REF)

0.00 % A t 1 0.00 9 N









Combined standard

uncertainty (%)

normal 1.09 >20


13

Table 4-3. KRISS uncertainty budget of total luminous flux measurement for blue LEDs (B).


uncertain

ty Ty

pe Probability

distribution

Sensitivity

coefficient

Contribut

ion (%)

DoF Correl

ated?


(DUT)

0.00 % A t 1 0.00 9 N




(REF)

0.00 % A t 1 0.00 9 N


(REF)

0.00 % A t 1 0.00 9 N









Combined standard

uncertainty (%)

normal 1.09 >20

Table 4-4. KRISS uncertainty budget of total luminous flux measurement for white LEDs (W).


uncertain

ty Ty

pe Probability

distribution

Sensitivity

coefficient

Contribut

ion (%)

DoF Correl

ated?


(DUT)

0.00 % A t 1 0.00 9 N




(REF)

0.00 % A t 1 0.00 9 N


(REF)

0.00 % A t 1 0.00 9 N









14


Combined standard

uncertainty (%)

normal 1.08 >20

Table 4-5. KRISS uncertainty budget of junction voltage measurement.


uncertainty

Ty

pe Probability

distribution

Sensitivity

coefficient

Contribut

ion (mV)

DoF Correl

ated?

sourcemeter calibration 0.05 mV B normal 1 0.05 Y

sourcemeter offset 0.10 mV B normal 1 0.10 Y

repeatability 0.04 mV A t 1 0.04 9 N

stray resistance 0.02 mV B rectangular 1 0.02 Y

Combined standard

uncertainty (mV)

t 0.12 >10

4.2. MIKES


The total luminous flux of LEDs was measured using a 30-cm integrating sphere. The

sphere has three ports: a main port for the LED under calibration, a detector port for a

photometer head, and an auxiliary port for an auxiliary LED. An LED holder used for total

luminous flux and a 5-cm precision aperture for the luminous flux responsivity of the

sphere photometer can be attached in the main port. The photometer used was made

by PRC Krochmann and had good cosine response. The auxiliary port was utilized in the

self-absorption measurements of the LEDs and in the transfer calibration of the total flux

mode.

The integrating sphere photometer has been calibrated for the illuminance

responsivity with an external source (luminous intensity standard lamp) when the 5-cm

entrance aperture is mounted in the main port. The illuminance in the center of the

entrance aperture is measured with a reference photometer, and the corresponding

photocurrent is measured with the sphere photometer at the same distance (70 cm) from

the external source. A correction due to illuminance non-uniformity of radiation field at

the aperture plane has been made. The light beam of the LED under calibration hit the

sphere wall at the same angle of incidence as the reference light from the external

source. The obtained illuminance responsivity of the sphere with the 5-cm aperture has

been transferred to the total flux mode by measuring the signal from a white LED in the

auxiliary port with two cases: when the 5-cm aperture and the LED holder have been


15

attached in the main port.

For calculating the spectral mismatch correction factor of the LEDs, the relative

spectral responsivity of the photometer has been calibrated with a reference

spectrometer of MIKES, and relative spectral throughput of the integrating sphere and

spectral power distribution of the LEDs have been measured with a spectroradiometer of

type DM150 from Bentham inc.

The total luminous flux measurements for each LED were made with the

integrating sphere photometer. The self-absorption measurements were made with an

auxiliary 5-mm white LED used in the auxiliary port by measuring the signal of the

photometer with and without the LED under calibration. To calculate the spectral

mismatch correction factor, the relative spectral power distributions were measured by

steps of 1 nm within the wavelength range of 380-780 nm, and the relative spectral

responsivity of the used photometer and the relative throughput of the integrating

sphere were measured by steps of 2 nm and 5 nm within the wavelength range of the

380-780 nm. During the measurements, the ambient temperature was (23.0 ± 1.0) °C and

the relative humidity of air was (31 ± 5) °C.


The LED holder used in the total luminous flux measurements of the LEDs is shown in Fig.

4-2. The LED is located in the center of the integrating sphere. The sensitivity of the

system to the positioning of the LEDs was tested by repeating the LED mounting and

signal measurement. The V(λ)-corrected photometer used for luminous flux signal

measurements and the diffuser of the spectroradiometer for the spectral measurements

were mounted to the detector port one at a time.

Fig. 4-2. LED holder used in the measurements of the total LED luminous flux in MIKES.

4.2.3. Traceability

The illuminance responsivity of the photometer used is traceable to MIKES’ reference

photometer. The reference photometer includes a precision aperture, a V(λ) filter, and a


16

silicon trap detector. The absolute transmittance of the V(λ) filter used in the reference

photometer is traceable to the national standard of the regular transmittance [Calibration

certificate T-R 479]. The spectral responsivity of the trap detector is traceable to a

cryogenic electrical substitution radiometer at SP in Sweden [Calibration certificate

MTeP501362-025] and modeling the spectral shape [Calibration certificate INT-028]. The

determinations of the areas of the precision apertures are traceable to the realization of

the meter at MIKES [Calibration certificate M-07L193]. The spectral irradiance responsivity

of the spectroradiometer is traceable to the national standard of spectral irradiance

[Calibration certificate T-R 506]. The calibrations of the current-to-voltage converter

Vinculum SP042 and digital voltmeter HP 3458A are traceable to the national standards

of electricity [Calibration certificates INT-033, INT-032].


Uncertainty components for the total luminous flux and junction voltage of the LEDs

have been presented in Tables below. The sensitivity coefficients of the uncertainty

components have been calculated as the ratio between the relative standard uncertainty

of the component and the standard deviation of the probability distribution of the

component. The uncertainty components due to wavelength errors and relative spectral

responsivity are based on Monte Carlo simulations.

Table 4-6. MIKES uncertainty budget of total luminous flux measurement for red LEDs (R).


uncertainty

(%)

Ty

pe

Probability

distribution

Sensitivity

coefficient

Contrib

ution

(%)

Deg. of

freedo

m

Correl

ated?

Setup-related

Repeatability 0.41 A normal 1 0.41 11 X

Near-field absorption 1.00 B rectangular 1 1.00 ∞ O

Self-absorption correction

factor

0.02 A normal 1 0.02 5 X

Non-uniformity of sphere

wall

0.20 B rectangular 1 0.20 ∞ O

Photocurrent measurement

(flux signal)

0.03 B rectangular 1 0.03 ∞ X

Current feeding B rectangular 3 –

5 %/mA

0.03 ∞ O

Integrating sphere

calibration


17

Illuminance responsivity of

photometer

0.20 B normal 1 0.20 ∞ O


(illuminance)

0.01 A normal 1 0.01 19 X

Drift of the external source 0.01 B rectangular 1 0.01 ∞ O

Long-term stability of

photometer


Distance setting of sphere-

photometer

B rectangular 0.5 %/mm 0.06 ∞ X

Aperture diameter B rectangular 0.006

%/μm

0.04 ∞ O

Reflection from aperture

land

B rectangular 1 0.05 ∞ O

Illuminance non-uniformity

correction

0.02 A normal 1 0.02 8 X

Calibration transfer factor 0.20 B rectangular 1 0.20 ∞ O

Repeatability of calibration 0.04 A normal 1 0.04 9 X

Spectral mismatch

correction

Wavelength error in LED

spectrum

B normal 0.05 –

0.2 %/nm

0.02 ∞ O

Wavelength error in

photometer response

B normal 0.5 –

4.7 %/nm

0.19 ∞ O

Relative spectral

responsivity of photometer


Throughput of integrating

sphere


Measurement geometry of

relative spectral response of

photometer


Combined standard

uncertainty (%)

-- -- normal -- 1.32 ∞ --

Table 4-7. MIKES uncertainty budget of total luminous flux measurement for green LEDs (G).


uncertainty

(%)

Ty

pe

Probability

distribution

Sensitivity

coefficient

Contrib

ution

(%)

Deg. of

freedo

m

Correl

ated?

Setup-related




18


factor

0.02 A normal 1 0.02 5 X


wall



(flux signal)



5 %/mA

0.02 ∞ O

Integrating sphere

calibration


photometer

0.20 B normal 1 0.20 ∞ O


(illuminance)

0.01 A normal 1 0.01 19 X



photometer



photometer



%/μm

0.04 ∞ O


land



correction

0.02 A normal 1 0.02 8 X



Spectral mismatch

correction


spectrum

B normal 0.05 –

0.2 %/nm

0.03 ∞ O

Wavelength error in

photometer response

B normal 0.5 –

4.7 %/nm

0.15 ∞ O

Relative spectral




sphere




photometer


Combined standard

uncertainty (%)

-- -- normal -- 1.24 ∞ --

Table 4-8. MIKES uncertainty budget of total luminous flux measurement for blue LEDs (B).


19


uncertainty

(%)

Ty

pe

Probability

distribution

Sensitivity

coefficient

Contrib

ution

(%)

Deg. of

freedo

m

Correl

ated?

Setup-related




factor

0.02 A normal 1 0.02 5 X


wall



(flux signal)



5 %/mA

0.02 ∞ O

Integrating sphere

calibration


photometer

0.20 B normal 1 0.20 ∞ O


(illuminance)

0.01 A normal 1 0.01 19 X



photometer



photometer



%/μm

0.04 ∞ O


land



correction

0.02 A normal 1 0.02 8 X



Spectral mismatch

correction


spectrum

B normal 0.05 –

0.2 %/nm

0.02 ∞ O

Wavelength error in

photometer response

B normal 0.5 –

4.7 %/nm

0.28 ∞ O

Relative spectral




20


sphere




photometer


Combined standard

uncertainty (%)

-- -- normal -- 2.80 ∞ --

Table 4-9. MIKES uncertainty budget of total luminous flux measurement for white LEDs (W).


uncertainty

(%)

Ty

pe

Probability

distribution

Sensitivity

coefficient

Contrib

ution

(%)

Deg. of

freedo

m

Correl

ated?

Setup-related




factor

0.02 A normal 1 0.02 5 X


wall



(flux signal)



5 %/mA

0.03 ∞ O

Integrating sphere

calibration


photometer

0.20 B normal 1 0.20 ∞ O


(illuminance)

0.01 A normal 1 0.01 19 X



photometer



photometer



%/μm

0.04 ∞ O


land



correction

0.02 A normal 1 0.02 8 X



21


Spectral mismatch

correction


spectrum

B normal 0.05 –

0.2 %/nm

< 0.01 ∞ O

Wavelength error in

photometer response

B normal 0.5 –

4.7 %/nm

0.03 ∞ O

Relative spectral




sphere




photometer


Combined standard

uncertainty (%)

-- -- normal -- 1.89 ∞ --

Table 4-10. MIKES uncertainty budget of junction voltage measurement for red LEDs (R).


uncertainty

Ty

pe

Probability

distribution

Sensitivity

coefficient

Contrib

ution

(mV)

Deg. of

freedo

m

Correl

ated?

Calibration of voltmeter B normal 1 0.02 ∞ O

Junction position

dependence

B rectangular 1 0.03 ∞ X

Stability of junction voltage A normal 1 0.01 –

0.02

19 X

Combined standard

uncertainty (mV)

-- -- normal -- 0.04 –

0.045

∞ --

Table 4-11. MIKES uncertainty budget of junction voltage measurement for green LEDs (G).


uncertainty

Ty

pe

Probability

distribution

Sensitivity

coefficient

Contrib

ution

(mV)

Deg. of

freedo

m

Correl

ated?


Junction position

dependence



0.04

19 X

Combined standard

uncertainty (mV)

-- -- normal -- 0.13 ∞ --

Table 4-12. MIKES uncertainty budget of junction voltage measurement for blue LEDs (B).


22


uncertainty

Ty

pe

Probability

distribution

Sensitivity

coefficient

Contrib

ution

(mV)

Deg. of

freedo

m

Correl

ated?


Junction position

dependence



0.06

19 X

Combined standard

uncertainty (mV)

-- -- normal -- 0.11 –

0.12

∞ --

Table 4-13. MIKES uncertainty budget of junction voltage measurement for white LEDs (W).


uncertainty

Ty

pe

Probability

distribution

Sensitivity

coefficient

Contrib

ution

(mV)

Deg. of

freedo

m

Correl

ated?


Junction position

dependence



0.04

19 X

Combined standard

uncertainty (mV)

-- -- normal -- 0.21 ∞ --

4.3. CMS-ITRI


As Fig. 4-3, the test LED is located within the integrating sphere centre. The integrating

sphere diameter is 1500 mm, include one auxiliary lamp for calculating absorption effect

and a optical detector for measuring optical signal. By substitute method, comparing the

output signal from the LED to that from the standard lamp in the integrating sphere.

Using the DC multiple standard resistor, two voltage meter and DC power supply that

give the LED current and monitor the current and voltage of the junction of LED. The

detector is the V(λ) optical detector connect the optical current meter for getting the

optical signal.


23

Fig. 4-3. Total Luminous Flux of LEDs measurement system in CMS-ITRI.


Fig. 4-4 is the vertical view of LED alignment. The LED at the centre of integrating sphere

and the beam direction is at the uniform area of the sphere that is flat spatial response

of distribution area. The LED is mounting by a holder that has two pins connect and has

two wires at the end of holder for power current connecting.

Fig. 4-4. The vertical view of LED alignment in CMS-ITRI.

4.3.3. Traceability

The traceability of LED total luminous flux is trace to the standard total luminous flux

lamp by total luminous flux measurement system. The standard total luminous flux lamp

is trace to the standard reference lamp then trace to NIST.

Baffle

Baffle

Detector Auxiliary

lamp

LED

(Vertical view)

LED

holder

Detector

(100 mm2 circular

aperture) LED

Alignment CCD

Alignment CCD

100 mm


24

Fig. 4-5. Traceability of measurement system in CMS-ITRI.


Uncertainty budget of total luminous flux measurement:

1. Repeatability of standard lamp:

The repeatability of standard lamp is record the optical current by using current meter

several times a day and measure several days. Calculate the standard deviation of all the

data.

2. Repeatability of test LED:

The repeatability of test LED is record the optical current by using current meter several

times a day and measure several days. Calculate the standard deviation of all the data.

3. Current ratio repeatability of standard lamp and LED:

Due to the different measurement condition between standard lamp and LED, such as

alignment angle, environment condition, and the small deviation of lamp, to consider the

optical signal ratio of repeatability of standard lamp and LED.

4. LED spatial light distribution:

Because of the geometrical structure in the integrating sphere, cause the non-uniform

distribution in the integrating sphere. Consider the deviation of LED alignment angle in

the relative uniform area, to calculate the deviation of LED.

5. Self-absorption factor:

Standard total

luminous flux lamp

Standard

reference lamp

Test LED

Total luminous flux

measurement

system

NIST

Total luminous flux

measurement

system


25

The self-absorption factor is when turn on the auxiliary lamp to measure the optical

signal of standard lamp and LED lamp, then to calculate the both of two ratio.

6. Spectral mismatch correction:

Because of the correction of spectrometer which the wavelength shifts affect the spectral

correction factor (SCF). Consider the wavelength shifts cause the error of SCF.

7. Calibration of standard lamp:

The uncertainty of calibration of standard lamp is drive from the relative expand

uncertainty calibrated by National measurement laboratory (NML) in Taiwan.

Uncertainty budget of junction voltage measurement:

1. Repeatability of test LED:

The repeatability of test LED is record the junction voltage by using voltage meter several

times a day and measure several days when measuring the LED averaged intensity.

Calculate the standard deviation of all the data.

2. Resolution of voltmeter:

To consider the drift when measure the junction voltage that is the maximum digit of

voltage meter.

3. Long-term drift of voltmeter:

Long-term drift of voltmeter is the drift of the traceability since the past. Calculate the

maximum deviation of the uncertainty drift.

4. Voltmeter calibration:

The uncertainty of voltmeter is drive from the relative expand uncertainty calibrated by

National measurement laboratory (NML) in Taiwan.

Table 4-14. CMS-ITRI uncertainty budget of total luminous flux measurement for red LEDs

(R).


uncertainty

Ty

pe

Probability

distribution

Sensitivity

coefficient

Contrib

ution

(%)

Deg. of

freedo

m

Correl

ated?

Repeatability of standard

lamp

0.002 A t 1 0.002 87 X

Repeatability of test LED 0.040 A t 1 0.040 87 O

Current ratio repeatability

of standard lamp and LED

0.156 A t 1 0.156 2 O

LED spatial light

distribution

0.664 B rectangular 1 0.664 200 X


26

Self-absorption factor 0.123 A t 1 0.123 89 O

Spectral mismatch

correction

0.090 B rectangular 1 0.090 200 O

Calibration of standard

lamp

0.920 B normal 1 0.920 5000 O

Combined standard

uncertainty (%)

-- -- normal -- 1.16 1264 --

Table 4-15. CMS-ITRI uncertainty budget of total luminous flux measurement for green LEDs

(G).


uncertainty T

yp

e Probability

distribution

Sensitivity

coefficient

Contrib

ution

(%)

Deg. of

freedo

m

Correl

ated?


lamp

0.003 A t 1 0.003 87 X




0.228 A t 1 0.228 2 O

LED spatial light

distribution

0.664 B rectangular 1 0.664 200 X


Spectral mismatch

correction

0.271 B rectangular 1 0.271 200 O


lamp

0.920 B normal 1 0.920 5000 O

Combined standard

uncertainty (%)

-- -- normal -- 1.19 807 --

Table 4-16. CMS-ITRI uncertainty budget of total luminous flux measurement for blue LEDs

(B).


uncertainty

Ty

pe

Probability

distribution

Sensitivity

coefficient

Contrib

ution

(%)

Deg. of

freedo

m

Correl

ated?


lamp

0.003 A t 1 0.003 87 X




0.222 A t 1 0.222 2 O

LED spatial light

distribution

0.664 B rectangular 1 0.664 200 X


27


Spectral mismatch

correction

0.156 B rectangular 1 0.156 200 O


lamp

0.920 B normal 1 0.920 5000 O

Combined standard

uncertainty (%)

-- -- normal -- 1.17 794 --

Table 4-17. CMS-ITRI uncertainty budget of total luminous flux measurement for white LEDs

(W).


uncertainty T

yp

e

Probability

distribution

Sensitivity

coefficient

Contrib

ution

(%)

Deg. of

freedo

m

Correl

ated?


lamp

0.003 A t 1 0.003 87 X




0.252 A t 1 0.252 2 O

LED spatial light

distribution

0.664 B rectangular 1 0.664 200 X


Spectral mismatch

correction

0.032 B rectangular 1 0.032 200 O


lamp

0.920 B normal 1 0.920 5000 O

Combined standard

uncertainty (%)

-- -- normal -- 1.16 586 --

Table 4-18. CMS-ITRI uncertainty budget of junction voltage measurement for red LEDs (R).


uncertainty

(%)

Ty

pe

Probability

distribution

Sensitivity

coefficient

Contrib

ution

(%)

Deg. of

freedo

m

Correl

ated?

Repeatability 0.020 A t 1 0.020 200 X

Resolution of voltmeter 0.003 B rectangular 1 0.003 200 O

Long-term drift of

voltmeter

0.026 B rectangular 1 0.026 200 O

Voltmeter calibration 0.001 B normal 1 0.001 5000 O

Combined standard

uncertainty (%)

-- -- normal -- 0.04 402 --


28

Table 4-19. CMS-ITRI uncertainty budget of junction voltage measurement for green LEDs

(G).


uncertainty

(%)

Ty

pe

Probability

distribution

Sensitivity

coefficient

Contrib

ution

(%)

Deg. of

freedo

m

Correl

ated?



Long-term drift of

voltmeter

0.026 B rectangular 1 0.026 200 O


Combined standard

uncertainty (%)

-- -- normal -- 0.07 261 --

Table 4-20. CMS-ITRI uncertainty budget of junction voltage measurement for blue LEDs (B).


uncertainty

(%)

Ty

pe

Probability

distribution

Sensitivity

coefficient

Contrib

ution

(%)

Deg. of

freedo

m

Correl

ated?



Long-term drift of

voltmeter

0.026 B rectangular 1 0.026 200 O


Combined standard

uncertainty (%)

-- -- normal -- 0.06 294 --

Table 4-21. CMS-ITRI uncertainty budget of junction voltage measurement for white LEDs

(W).


uncertainty

(%)

Ty

pe

Probability

distribution

Sensitivity

coefficient

Contrib

ution

(%)

Deg. of

freedo

m

Correl

ated?




29

Long-term drift of

voltmeter

0.026 B rectangular 1 0.026 200 O


Combined standard

uncertainty (%)

-- -- normal -- 0.15 213 --

4.4. PTB


Fig. 4-6 below shows the measurement setup in principle. To enable the measurement of

all the desired quantities, a special mechanism is needed. This allows the following

functionality: the alignment of the LED transfer standard to the optical axis of the system,

the rotation of the LED transfer standard around its horizontal axis φ and rotation

around its vertical axis θ. Furthermore, it allows the variation of the distance r between

the selected detector and the LED transfer standard. Opposite the LED transfer standard,

a rotating wheel is used for a quick detector selection. Additionally, there is a laser and a

CCD camera mounted to enable the easy alignment of the LED transfer standard. Due to

the rotation of φ angle, the interconnection between the power supply and the LED

under test prohibits an endless rotation.

Thus, in the case of luminous flux measurements after a little more than one

rotation, a stop is needed. The next movement will then be the turn back and so on.

The goniophotometer measured the zonal photocurrent (which is proportional to

the measured averaged illuminance) as a function of the angle θ where θ = 0 represents

the optical axis of the goniophotometer, which is also the mechanical axis of the LED

package in the direction of emittance. See Fig. 4-7 below.


30

Fig. 4-6. Measurement setup for total luminous flux in PTB.

Fig. 4-7. Geometry of the gonio-photometric measurement of LED total luminous flux in PTB.


Fig. 4-8 below shows the holder which was used to hold, align and operate each LED. A

high reflecting cone directly behind the installed LED allows for the indirect measurement

of the backward directed partial luminous flux of the LEDs, which also contributes to the

total luminous flux.


31

Fig. 4-8. Pictures of the LED holder used in the measurement of total luminous flux in PTB.

4.4.3. Traceability

The primary standards for the measured quantities are traceable to national standards.


The uncertainties are determined from up to 30 individual contributions originated in the

operation and alignment of an LED in thermal conditions influenced by the holder and

the environment. The specific properties of the measurement devices and their effects

are considered in detail. The estimated uncertainties of the contributions are maximum

for standard LED calibrations at PTB. They are listed and sorted in uncertainty budgets.

The components are treated as uncorrelated.

The next statement shows the formula to determinate luminous flux:


32

Table 4-22. PTB uncertainty budget of total luminous flux measurement for red LEDs (R).


uncertainty

Ty

pe

Probability

distribution

Sensitivity

coefficient

Contribut

ion (%)

Deg.

of

freedo

m

Correl

ated?

LED nominal current 0 A 22.8679 0

Exponent LED current

correction

0.36 B normal 9.98E-6 5.42E-4 13

LED current reading 2.0E-6 A A normal -22.8683 -6.88E-3 10

Correction factor for

spectral mismatch as

function of θ

0 B normal 0.665126 0 20

Exponent LED voltage

correction

1.6 B normal 1.0678E-3 0.25 13

LED nominal voltage for

25 °C

7.3E-4 V A normal 1.89755 0.21 9

LED voltage reading 6.0E-4 V A normal -1.9006 -0.17 10


straylight

0.00050 B normal 0.665219 0.050 10

LED backward emission 0.0010 B normal 0.664462 0.10 10

Straylight correction of

spectrometer

5.0E-5 B normal 0.665126 0.0050 50

Bandbass correction of

spectrometer

0.00011 B normal 0.665126 0.011 50

Distance 0.00050 m B rectangular 4.20966 0.32 10

Photometric sensitivity of

photometer

8.9E-11 A/lx B normal -2.4003E7 -0.32 10

Spectral mismatch

correction factor

0.0078 B normal 0.648397 0.76 20

Integrated photocurrent,

solid angle weighted

2.3E-10 A B normal 2.34488E7 0.82 90


33

Combined standard

uncertainty (%)

-- -- normal -- 1.27 105 --

Table 4-23. PTB uncertainty budget of total luminous flux measurement for green LEDs (G).


uncertainty

Ty

pe

Probability

distribution

Sensitivity

coefficient

Contribut

ion (%)

Deg.

of

freedo

m

Correl

ated?



correction

0.13 B normal 2.5724E-5 1.2E-4 13

LED current reading 2.0E-6 A A normal -72.1751 -0.0051 10



function of θ

0 B normal 2.85823 0 20


correction

0.45 B normal 6.5639E-3 0.10 13


25 °C

0.0026 V A normal 1.32354 0.12 9

LED voltage reading 0.0011 V A normal -1.32658 -0.052 10


straylight

0.00050 B normal 2.85863 0.050 10



spectrometer

3.0E-5 B normal 2.85823 0.003 50


spectrometer

0.00010 B normal 2.85863 0.010 50



photometer

8.9E-11 A/lx B normal -1.0314E8 -0.32 10

Spectral mismatch

correction factor

0.0035 B normal 2.87028 0.35 20



1.2E-9 A B normal 2.26512E7 0.95 90

Combined standard

uncertainty (%)

-- -

-

normal -- 1.12 135 --

Table 4-24. PTB uncertainty budget of total luminous flux measurement for blue LEDs (B).


uncertainty

Ty

pe

Probability

distribution

Sensitivity

coefficient

Contribut

ion (%)

Deg.

of

free

dom

Correl

ated?


34



correction

0.028 B normal 7.75322E-6 2.8E-5 13




function of θ

0.00020 B normal 0.77 0.020 20


correction

0.10 B normal 0.0016 0.022 13


25 °C

0.0017 V A normal 0.109 0.024 9

LED voltage reading 8.0E-4 V A normal -0.109743 -0.011 10


straylight

0.00050 B normal 0.775426 0.050 10



spectrometer

0.0010 B normal 0.775318 0.10 50


spectrometer

0.0010 B normal 0.775318 0.10 50



photometer

8.9E-11

A/lx

B normal -2.79797E7 -0.32 10

Spectral mismatch

correction factor

0.0071 B normal 0.873302 0.80 50



3.1E-10 A B normal 2.0155E7 0.82 90

Combined standard

uncertainty (%)

-- -- normal -- 1.24 157 -

Table 4-25. PTB uncertainty budget of total luminous flux measurement for white LEDs (W).


uncertainty

Ty

pe

Probability

distribution

Sensitivity

coefficient

Contribut

ion (%)

Deg.

of

freedo

m

Correl

ated?



correction

0.21 B normal 1.6824E-5 2.1E-4 13




function of θ

0.00020 B normal 1.68311 0.020 20


35


correction

0.61 B normal 0.0026366 0.095 13


25 °C

0.0025 V A normal 1.34013 0.20 9

LED voltage reading 0.0011 V A normal -1.34223 -0.09 10


straylight

0.00050 B normal 1.68267 0.050 10



spectrometer

1.0E-5 B normal 1.68244 0.001 50


spectrometer

4.0E-5 B normal 1.68244 0.0040 50



photometer

8.9E-11 A/lx B normal -6.0715E7 -0.32 10

Spectral mismatch

correction factor

0.0023 B normal 1.69072 0.23 50



6.8E-10 A B normal 2.2639E7 0.92 90

Combined standard

uncertainty (%)

-- -- normal -- 1.1 134 --

Table 4-26. PTB uncertainty budget of junction voltage measurement of blue LED (example).


uncertainty

Ty

pe

Probability

distribution

Sensitivity

coefficient

Contrib

ution

(mV)

Deg. of

freedo

m

Correl

ated?

Calibration of voltmeter 0.00005 B rectangular 3.44 0.17 10

Junction position

dependence

0.00052 V B rectangular -1 -0.52 10

Reproducibility 0.00058 V A normal 1 0.58 10

Combined standard

uncertainty (mV)

-- -- normal -- 0.80 21 --

4.5. NMIJ


The measurement of LED luminous flux at NMIJ is based on the goniophotometric

method. The measurement distance is 1.15m. "f1' value" of a photometer for LED

luminous flux (LED-photometer) is 2.4. The Photometer and the LED mount socket were


36

installed on the automatic-move stage.

Fig. 4-9. Calibration facility for LED luminous intensity and total luminous flux in NMIJ.


a) The laser system and the telescope with CCD camera are used for LED alignment.

b) LED holder is mounted to the gonio-stage. (see Fig. 4-10)

c) Fig. 4-11 shows picture of the LED holder. (Pin socket is used to mount LED)

Fig. 4-10. LED mount socket mounted to the gonio-stage in NMIJ.

Fig. 4-11. LED mount socket in NMIJ.


37

4.5.3. Traceability

a) Illuminance responsivity of the LED photometer ⇒ luminous intensity standard at

NMIJ.

b) Relative spectral responsivity of the LED photometer ⇒ spectral responsivity

standard at NMIJ.

c) Relative spectral distribution of the test LED ⇒ spectral irradiance standard at NMIJ.


Table 4-27. NMIJ uncertainty budget of total luminous flux measurement for red LEDs (R).


uncertainty

Ty

pe

Probability

distribution

Sensitivity

coefficient

Contrib

ution

(%)

Deg. of

freedo

m

Correl

ated?

Calibration of illuminance

responsivity

B gaussian 1 0.32 1510 O

Temperature dependence of

illuminance responsivity 1.2 °C B rectangular 0.08 %/°C 0.09 ∞ O

Linearity of illuminance

responsivity


Reference plane of

photometer

0.62 mm B rectangular 0.17 %/mm 0.11 ∞ O

Distance alignment 0.21 mm B rectangular 0.17 %/mm 0.04 ∞ X

Current feeding accuracy B rectangular 1 < 0.01 ∞ O

DMM accuracy B rectangular 1 < 0.01 ∞ O

Axis alignment 0.29 mm B rectangular 0.62 %/mm 0.18 ∞ X

Optical center in LED B rectangular 1 0.61 ∞ X

Angle accuracy B rectangular 1 0.05 ∞ O

Repeatability of LED

lighting (including noise

and drift)

A t 1 0.13 6 X

Stray light B rectangular 1 0.10 ∞ O

measurement angle step

and angular resolution


Spectral mismatch correction factor

Spectral responsivity

calibration (including

A

+

gaussian 1 0.11 ∞ X


38

repeatability) B

Spectral irradiance


repeatability)

A

+

B

gaussian 1 < 0.01 ∞ X

Wavelength uncertainty of

relative spectral

responsivity

Random

0.1 nm,

systematic

0.1 nm

A

+

B

gaussian

(random

factor),

rectangular

(systematic

factor)

-- 0.19 ∞ X


LED spectral distribution

Random

0.1 nm,

systematic

0.1 nm

A

+

B

gaussian

(random

factor),

rectangular

(systematic

factor)

-- 0.02 ∞ X

Effect of slit function width B rectangular 1 0.04 ∞ X

Angular dependence of



Combined standard

uncertainty (%)

-- -- normal -- 1.2 >>

20000

--

Table 4-28. NMIJ uncertainty budget of total luminous flux measurement for green LEDs (G).


uncertainty

Ty

pe

Probability

distribution

Sensitivity

coefficient

Contrib

ution

(%)

Deg. of

freedo

m

Correl

ated?


responsivity





responsivity


Reference plane of

photometer









39



and drift)

A t 1 0.09 6 X








repeatability)

A

+

B


Spectral irradiance


repeatability)

A

+

B



relative spectral

responsivity

Random

0.1 nm,

systematic

0.1 nm

A

+

B

gaussian

(random

factor),

rectangular

(systematic

factor)

-- 0.2 ∞ X



Random

0.1 nm,

systematic

0.1 nm

A

+

B

gaussian

(random

factor),

rectangular

(systematic

factor)

-- 0.02 ∞ X





Combined standard

uncertainty (%)

-- -- normal -- 0.86 >>

20000

--

Table 4-29. NMIJ uncertainty budget of total luminous flux measurement for blue LEDs (B).


uncertainty

Ty

pe

Probability

distribution

Sensitivity

coefficient

Contrib

ution

(%)

Deg. of

freedo

m

Correl

ated?


responsivity





responsivity


Reference plane of

photometer





40







and drift)

A t 1 0.06 6 X








repeatability)

A

+

B


Spectral irradiance


repeatability)

A

+

B



relative spectral

responsivity

Random

0.1 nm,

systematic

0.1 nm

A

+

B

gaussian

(random

factor),

rectangular

(systematic

factor)

-- 0.31 ∞ X



Random

0.1 nm,

systematic

0.1 nm

A

+

B

gaussian

(random

factor),

rectangular

(systematic

factor)

-- < 0.01 ∞ X





Combined standard

uncertainty (%)k=1

-- -- normal -- 0.94 >>

20000

--

Table 4-30. NMIJ uncertainty budget of total luminous flux measurement for white LEDs (W).


uncertainty

Ty

pe

Probability

distribution

Sensitivity

coefficient

Contrib

ution

(%)

Deg. of

freedo

m

Correl

ated?


responsivity





41


responsivity


Reference plane of

photometer










and drift)

A t 1 0.14 6 X








repeatability)

A

+

B


Spectral irradiance


repeatability)

A

+

B



relative spectral

responsivity

Random

0.1 nm,

systematic

0.1 nm

A

+

B

gaussian

(random

factor),

rectangular

(systematic

factor)

-- 0.04 ∞ X



Random

0.1 nm,

systematic

0.1 nm

A

+

B

gaussian

(random

factor),

rectangular

(systematic

factor)r

-- < 0.01 ∞ X





Combined standard

uncertainty (%)k=1

-- -- normal -- 0.71 >>

20000

--

Table 4-31. NMIJ uncertainty budget of junction voltage measurement.


42


uncertainty

Ty

pe

Probability

distribution

Sensitivity

coefficient

Contrib

ution

(V)

Deg. of

freedo

m

Correl

ated?

Calibration of DMM B gaussian 1 0.0001 ∞ O

Repeatability (including

effect of temperature

difference)

A gaussian 1 0.0001

~

0.0033

4 X

Junction position B rectangular 1 0.0003 ∞ X

Combined standard

uncertainty (V) k=1

-- -- normal -- 0.0003

~

0.0033

20 --

4.6. CENAM


The measurement system used for Total Luminous Flux is conformed by a set of standard

incandescent lamps and a 1 m diameter luminous integrating sphere. The integrating

sphere includes a photometric detector coupled to the exit port of a satellite sphere, an

auxiliary lamp, a pair of baffles to avoid the direct incidence of light into the photometric

detector, and a lamp holder. The measurement system is completed with the electronic

instrumentation commonly used to measure photocurrents and other electric operating

parameters of the lamps. The measurement system used is shown in Fig. 4-12 and Fig.

4-13.

Fig. 4-12. Schematic diagram of the total luminous flux measurement setup in CENAM.


43

Fig. 4-13. 1 m diameter integrating sphere at CENAM.


In order to mount the LEDs artefacts inside the integrating sphere, an LED holder was

adapted to the lamp holder as shown in Fig. 4-14. No alignment was provided to the

LEDs.

Fig. 4-14. LED holders for integrating sphere in CENAM.

4.6.3. Traceability

The total luminous flux was measured by using a photometric detector and set of

standard lamps calibrated for this quantity by NIST. Fig. 4-15 shows the traceability chart

for the Total Luminous Flux measurements performed at CENAM, where the expanded

uncertainty presented correspond to a coverage factor of k = 2.


44

Fig. 4-15. Traceability chart for the total luminous flux measurements performed at CENAM.


The total luminous flux of the LED led is determined by using Eq. (4.1):

, (4.1)

where iled is the photocurrent of the photometer head when measuring the LED’s, led is

the LED self-absorption correction, ccf*(Sled) is the LED spectral mismatch correction

factor, ccf*(Sp) is the standard lamp spectral mismatch correction factor, p is the value of

the standard lamps total luminous flux, and T is the system transfer function given by

Eq. (4.2):

, (4.2)

where p is the standard lamps self-absorption correction and ip is the photocurrent of

the photometer head when measuring standard lamps.

The spectral mismatch correction factor used for the standard lamps and the

white LED’s is given by Eq. (4.3):

, (4.3)

where SA(λ) is the relative spectral power distribution of the CIE Illuminant A, Si(λ) is the

relative spectral power distribution of the source when located inside the integrating

Total Luminous Flux

0,5 lm - 5 000 lm

LED’S

U = 11%

volt

[V]

Voltage [V]

Multimeter

M-3457-883

M-3458-334

U ≤ 13 µV/V

CNM-PNE-5

Electric DC

Voltage

ohm

[]

Resistance []

Shunt Resistor

Res-61173

0,0999965

U ≤ 1,7µΩ/Ω

[V]

Multimeters

M-3457-883

M-3457-885

U = 15µV

/Ω

r

M-3457-881

CNM-PNE-3

Electric

Resistance

ampere

[A]

Electrical DC

current [A]

Multimeter

M-3458-334

U ≤ 13 µA/A

CNM-PNE-13

Electric DC

Current

lumen

[lm]

Total Luminous Flux

NIST

Integrating Sphere

CNM-PNF-15

Total Luminous

Flux

Total Luminous Flux

[lm]

Lamps

P486, P487

U = 0.5 %

SI units

External

Services


45

sphere, V(λ) is the spectral luminous efficiency function and Rs(λ) is the relative spectral

responsivity function of the sphere system, that can be obtained by measuring the

relative spectral responsivity of the photometer head, Srel-df (, and the relative spectral

throughput of the integrating sphere Ts(λ) as in Eq. (4.4):

), (4.4)

The relative spectral throughput Ts(λ) of the sphere was obtained using a

spectrorradiometer and calculating the ratios of the spectral irradiance on the detector

port of the sphere to the spectral irradiance of the same lamp or LED measured outside

the integrating sphere, as shown in Eq. (4.5):

, (4.5)

For the red, green and blue LEDs, the spectral mismatch correction factor used is given

by Eq. (4.6):

, (4.6)

where SA(λ) is the relative spectral power distribution of the CIE Illuminant A, Srel-df () is

the relative spectral responsivity of the photometer head and SLED is the LED relative

spectral power distribution, which was simulated from the measured FWHM and peak

wavelength6.

Thus, the uncertainty estimation of the spectral irradiance was done by

considering the input and influence quantities presented in Fig. 4-16.

6 Richard Y., Kathleen M.., Carolyn J., Quantifying photometric spectral mismatch uncertainties in LED measurements, Proceedings of the 2nd Expert Symposium on LED Measurement, CIE, Genève, (2001).


46

Fig. 4-16. Total luminous flux uncertainty components in CENAM.

Table 4-32. CENAM uncertainty budget of total luminous flux measurement for red LEDs (R).


uncertainty

(%)

Ty

pe

Probability

distribution

Sensitivity

coefficient

Contrib

ution

(%)

Deg. of

freedom Correl

ated?

Total

Luminous

Flux

T

Reading repeatibility

Multimeter resolution

Multimeter error

p

i p


Reading repeatibility

Multimeter error

Total luminous flux reference value

ccf*

standard lamps and

white LED’s

S rel

S lamp

Photometer head relative spectral responsivity

Spectroradiometer error

Spectroradiometer repeatibility in the sphere

Spectroradiometer repeatibility out the sphere

i led


Multimeter repeatibility

Multimeter error

led



Multimeter error

ccf*

red, green and blue

LED’s

S rel

S lamp Spectroradiometer error

Photometer head relative spectral responsivity

Spectroradiometer repeatibility in the sphere

Current

feeding

accuracy

R Resistance value

V Resistance



Multimeter error

Voltage

junction

due to position

position vled FLT

V LED



Multimeter error


47

Luminous flux reference

value

0.25 B normal 1 0.25 200 X

System transfer function 0.11 B normal 1 0.11 200 X

Standard lamps spectral

mismatch correction

2.22 B normal 1 2.22 200 X

LED self-absorption

correction

0.06 B normal 1 0.06 200 O

LED readings repeatability 3.87 A normal 1 3.87 14 O

LEDs spectral mismatch

correction

2.65 B normal 1 2.65 200 O

Junction voltage 0.012 A normal 1 0.012 14 X

Current feeding accuracy 0.17 A normal 1 0.17 14 X

Combined standard

uncertainty (%)

-- -- normal -- 5.20 45 --

Table 4-33. CENAM uncertainty budget of total luminous flux measurement for green LEDs

(G).


uncertainty

(%) Ty

pe

Probability

distribution

Sensitivity

coefficient

Contrib

ution

(%)

Deg. of

freedom Correl

ated?


value

0.25 B normal 1 0.25 200 X



mismatch correction

2.44 B normal 1 2.44 200 X

LED self-absorption

correction

0.06 B normal 1 0.06 200 O



correction

2.93 B normal 1 2.93 200 O



Combined standard

uncertainty (%)

-- -- normal -- 4.33 290 --

Table 4-34. CENAM uncertainty budget of total luminous flux measurement for blue LEDs (B).


48


uncertainty

(%)

Ty

pe

Probability

distribution

Sensitivity

coefficient

Contrib

ution

(%)

Deg. of

freedom Correl

ated?


value

0.25 B normal 1 0.25 200 X



mismatch correction

2.22 B normal 1 2.22 200 X

LED self-absorption

correction

0.07 B normal 1 0.07 200 O



correction

2.79 B normal 1 2.79 200 O



Combined standard

uncertainty (%)

-- -- normal -- 4.92 61 --

Table 4-35. CENAM uncertainty budget of total luminous flux measurement for white LEDs

(W).


uncertainty

(%)

Ty

pe

Probability

distribution

Sensitivity

coefficient

Contrib

ution

(%)

Deg. of

freedom Correl

ated?


value

0.25 B normal 1 0.25 200 X



mismatch correction

2.22 B normal 1 2.22 200 X

LED self-absorption

correction

0.06 B normal 1 0.07 200 O



correction

2.62 B normal 1 2.79 200 O



Combined standard

uncertainty (%)

-- -- normal -- 4.55 86 --

Table 4-36. CENAM uncertainty budget of junction voltage measurement.


49


uncertainty

(%)

Ty

pe

Probability

distribution

Sensitivity

coefficient

Contrib

ution

(%)

Deg. of

freedom

Cor

rela

ted?

Readings repeatability 0.01604 A normal 1 0.01604 14 O

Multimeter resolution 0.00001 B rectangular 1 0.00001 200 X

Multimeter error 0.00055 B normal 1 0.00055 200 X

Combined standard

uncertainty (%)

-- -- normal -- 0.016 14 --

4.7. LNE


LNE has developed a measurement set-up to measure photometric and colorimetric

characteristics of LEDs. This set-up is based on a goniophotometer designed to meet the

requirements of the CIE127 standards for averaged intensity and total flux measurements.

It is optimised for high power white LEDs measurements and was adapted for the LEDs

in the framework of the APMP-S3 supplementary comparison. The schematic of the

goniophotometer is shown on Fig. 4-17. It is 2 m long and 1.8 m high.


50

Fig. 4-17. Goniophotometer for LEDs flux measurements in LNE.

The set-up is made of the following parts :

- Optical rails to set the main frame

- A multi-axis LED mount which allow the accurate alignment of the LED along the

horizontal optical axis and with respect to the photometric center of the

goniophotometer. This device is mounted onto a horizontal axis motorised

rotation stage that rotates the LED around the optical axis. A detailed schematic

of the LED mount is shown on figure 2.

- A vertical axis motorised rotation stage on which the multi-axis LED mount is

placed.

A camera placed above the LED allows us to adjust the position of the LED with

respect to the photometric center. The photometer is mounted on an optical rail. The

Photometer

Spectrocolorimeter

LED mount

Stepping

motor driver

Camera


51

distance between the photometer and the LED can be adjusted to meet the

requirements of the measurement conditions. During the measurements the photometer

is kept steady. Laser beam is used to define the optical axis of the goniophotometer.

Fig. 4-18. LED mount in LNE.

Total flux is determined from intensity measurements in any directions I(,)

and integration over 4 steradian according to the following equation:

0

2

0

sin, ddI

Intensity measurement is performed with a photometer, manufacturer LMT, type

P11S00, including a 11,3 mm diameter (1 cm²) sensitive area, with a very fine V()

correction (f’1 1%). Due to the geometry and size of the components of the bench the

angles in is limited to 140°. To take into account backlight emission of the LED, a 5 mm

diameter white paper is put at the back of the LED. The reflectance factor of the white

paper is 0.8. The distance between the LED and the photometer is 350 mm. The

photometric center is aligned onto the LED chip. The angular resolution due to the size

of the sensitive area of the photometer is 2°. The angular measurement step is 5° in

and 1° in .

The instruments used to perform the measurements are listed in Table 4-37.

Table 4-37. Instruments used on the LED photometric bench in LNE.

Instrument Manufacturer Type Function

V() photometer LMT P11S00 Illuminance

measurement

Picoammeter Keithley 486 Photometer current

measurement


52

LED power supply Agilent 3436A Stabilised LED power

supply

Shunt resistor AOIP 1000 / 228RE6 LED current

measurement

Multimeter Hewlett-Packard 3457A LED junction voltage

measurement


Alignment of the LED is performed using a luminancemeter, manufacturer LMT, type

L1009 with reflex viewing.

Fig. 4-19. LED holders in LNE.

4.7.3. Traceability

Photometer

The photometer is calibrated in illuminance at LNE using a set of three standard lamps

calibrated in luminous intensity at LNE-INM. The standards lamps are calibrated using

primary realisation of the candela through filter radiometer.

Electrical Instruments

All electrical instruments with critical impact on the measurements are calibrated by the

LNE electrical department which is COFRAC (Comité Français d’Accréditation) accredited.

COFRAC is the French accreditation body.

Length

The distance between the LED and the photometer is measured using a meter calibrated

by the LNE length department which is COFRAC accredited.


53


Flux measurement

Reading repeatability

This uncertainty is estimated from the standard deviation of 5 measurements performed

in the same operating conditions. The uncertainty associated to each colours are the

following:

- Red: 0.25 %

- Green: 0.10 %

- Blue: 0.10 %

- White: 0.20 %

This uncertainty includes also the uncertainty due to horizontal, vertical and

angular alignment of the LED.

Component due to distance between the LED and the photometer

The distance between the LED and the reference plane of the photometer is known with

an uncertainty of 100 µm. The associated contribution to the intensity measurement is

evaluated by measuring the changes in the photometer signal when the distance is

changed by 5 mm. The result is shown in the following table for the different LED

colours.

LED type Relative uncertainty due to distance

LED-photometer

(%)

Red 0.03

Green 0.03

Blue 0.03

White 0.03

Component due to current feeding accuracy.

The current is measured through a 1000 resistor using a voltmeter. The resistor is

calibrated with an uncertainty of 1. 10-5. The voltmeter is calibrated with an uncertainty

of 1. 10-5. Therefore the current is measured with an uncertainty of 1.4 10-5. The current

is adjusted with an offset of 0.001 mA which corresponds to a relative error of 5. 10-5 .

The intensity is not corrected for this offset which is included in the uncertainty of the

current. The overall uncertainty on the current feeding is obtained from the uncertainty

due to the current measurement and the current offset, that is 5.2 10-5. The

corresponding uncertainty of the LED intensity measurement is determined from the

manufacturer’s data sheets. The results are summarized in the following table:


54

LED type Relative uncertainty due to

current feeding

(%)

Red 0.0052

Green 0.0042

Blue 0.0031

White 0.0042

Diffuser 0.0042

Component due to stray light in the optical bench

Stray light in the optical bench is evaluated by placing a mask on the optical path of the

beam at a distance of about 100 mm from the LED. The size of the mask is 10 mm. For

all types of LED the relative contribution of the stray light to the photometer signal is <

0.01 %.

Component due to ambient temperature

The measurements are performed at 23 °C 1 °C. The measurement uncertainty due to

the uncertainty on the ambient temperature is determined from the manufacturer’s data

sheets. The results are summarized in the following table:

LED type Uncertainty due to ambient temperature

(%)

Red 0.5

Green 0.25

Blue 0.25

White 0.2

Diffuser 0.25

Component due to angular resolution and computation

Flux measurement is performed with a step angle of 5° in and 1° in . The uncertainty

due to the angular resolution is evaluated by comparing results of the measurement

performed with a 2° and 5° step in . The results show an uncertainty of 0.15%.

Component due to backward emission

Contribution of the backward emission of the LED is measured by placing a white

diffused paper at the back of the LED. The reflectance factor of the white paper is 0.8

with an uncertainty of 0.05. Assuming that backward emission represents 4% of the

forward emitted light the uncertainty due to the use of the white paper is 0.2%.

Component due to the calibration of the photometer


55

The photometer is calibrated with a relative uncertainty of 0.6%.

Component due to linearity of the photometer

The photometer is calibrated in linearity. The uncertainty associated to the photometer

linearity varies from 0.02 % to 0.1 %. Therefore the uncertainty on the flux measurement

is 0.1 %.

Component due to spectral mismatch correction

The photometer is calibrated in relative spectral response. The LED flux measurement

results are corrected for the spectral mismatch of the photometer. The uncertainty on the

relative spectral response of the photometer is used to determine the uncertainty on the

spectral mismatch correction. This uncertainty is calculated by taking the average of the

uncertainty of the relative spectral response weighted by the spectral distribution of the

LED. Works using Monte Carlo techniques are underway to take into account correlation

in determining uncertainty on spectral mismatch correction. The actual uncertainties are

the following:

- Red: 0.5 %

- Green: 0.4 %

- Blue: 1 %

- White: 0.2 %

Table 4-38. LNE uncertainty budget of total luminous flux measurement for red LEDs (R).

Uncertainty

Component

Standard

uncertainty Ty

pe Probability

distribution

Sensitivity

coefficient

Contribution

(%)

Deg. of

freedom

Correlated?

Reading

repeatability

0.25 A t 1 0.25 4 X

Distance

setting


Current

feeding

accuracy

0.0052 B rectangular 1 0.0052 ∞ X

Stray light 0.01 B rectangular 1 0.01 ∞ O

Ambiant

temperature


Angular

measurement

step and

computation


Backward

emission



56

Calibration

of

photometer

0.6 B normal 1 0.6 ∞ O

Non-

linearity


Spectral

mismatch

correction

0.5 B normal 1 0.5 ∞ X

Combined

standard

uncertainty

(%)

-- -- normal -- 1.0 ∞ --

Table 4-39. LNE uncertainty budget of total luminous flux measurement for green LEDs (G).

Uncertainty

Component

Standard

uncertainty Ty

pe Probability

distribution

Sensitivity

coefficient

Contribution

(%)

Deg. of

freedom

Correlated?

Reading

repeatability

0.1 A t 1 0.1 4 X

Distance

setting


Current

feeding

accuracy

0.0052 B rectangular 0.8 0.00416 ∞ X


Ambiant

temperature


Angular

measurement

step and

computation


Backward

emission


Calibration

of

photometer

0.6 B normal 1 0.6 ∞ O

Non-

linearity


Spectral

mismatch

correction

0.4 B normal 1 0.4 ∞ X

Combined

standard

uncertainty

(%)

-- -- normal -- 0.82 ∞ --

Table 4-40. LNE uncertainty budget of total luminous flux measurement for blue LEDs (B).

Uncertainty

Component

Standard

uncertainty Ty

pe Probability

distribution

Sensitivity

coefficient

Contribution

(%)

Deg. of

freedom

Correlated?


57

Reading

repeatability

0.1 A t 1 0.1 4 X

Distance

setting


Current

feeding

accuracy

0.0052 B rectangular 0.6 0.00312 ∞ X


Ambiant

temperature


Angular

measurement

step and

computation


Backward

emission


Calibration

of

photometer

0.6 B normal 1 0.6 ∞ O

Non-

linearity


Spectral

mismatch

correction

1 B normal 1 1 ∞ X

Combined

standard

uncertainty

(%)

-- -- normal -- 1.2 ∞ --

Table 4-41. LNE uncertainty budget of total luminous flux measurement for white LEDs (W).

Uncertainty

Component

Standard

uncertainty Ty

pe Probability

distribution

Sensitivity

coefficient

Contribution

(%)

Deg. of

freedom

Correlated?

Reading

repeatability

0.2 A t 1 0.2 4 X

Distance

setting


Current

feeding

accuracy

0.0052 B rectangular 0.8 0.00416 ∞ X


Ambiant

temperature


Angular

measurement

step and

computation


Backward

emission


Calibration

of

0.6 B normal 1 0.6 ∞ O


58

photometer

Non-

linearity


Spectral

mismatch

correction

0.2 B normal 1 0.2 ∞ X

Combined

standard

uncertainty

(%)

-- -- normal -- 0.75 ∞ --

Junction Voltage

Repeatability

This uncertainty is estimated from the standard deviation of 20 measurements performed

in the same operating conditions. For all type of LED the uncertainty is 0.02%.

Component due to the calibration of the voltmeter

The voltmeter used for the junction voltage measurement is calibrated with an

uncertainty of 0.001 %.

Component due to position of junction voltage measurement point.

The leads of the LED are made of iron for the red LED and of copper for the green,

blue and white LED. The 4-wires device used to measure the junction voltage is located

20 mm away from the LED chip. Taking into account the geometry of the leads (40 mm

long and 0.25 mm² area) and the conductivity of the material used for the leads we

determine the voltage drop due to the leads. The results are summarized in the following

table.

LED type Relative voltage drop @ 20 mA

(%)

Red 0.008

Green 0.0008

Blue 0.0008

White 0.0008

Diffuser 0.0008

Table 4-42. LNE uncertainty budget of junction voltage measurement of red LEDs.

Uncertainty

Component

Standard

uncertainty Ty

pe Probability

distribution

Sensitivity

coefficient

Contribution

(%)

Deg. of

freedom

Correlated?

Repeatability* 0.04 A normal 1 0.04 29 X


59

Calibration of

voltmeter

0.001 B normal 1 0.001 ∞ O

Junction

position

dependence*


Combined

standard

uncertainty

(%)

-- -- normal -- 0.041 ∞ --

4.8. METAS


The measurements were performed in two steps. First the DUT-LED is used for

calibrating the luminous flux sensitivity of the integrating sphere. For this purpose the

LED is placed at 100 mm in front of a 100 mm2 aperture. A LED of same colour is used

inside the sphere in order to minimize self absorption effects. In the second step the LED

is placed inside the sphere and the flux of the DUT-LED is measured. The main

components of the system are listed in the following diagram.

Fig. 4-20. Schematic setup for LED total luminous flux in METAS.


The LED was mounted inside the integrating sphere in a way that the absorption of light

emitted on the back side of the LED is as small as possible. The output of the LED is

100 mm

aperture 100 mm2

1-m integrating sphere, Czibula & Grundmann GmbH, BaSO4, ρ>0.98

cos-corrected Photometer LMT

baffle

Keithley Sourcemeter 2400, 4wires

Keithley Multimeter 2010


60

oriented in the same direction than the light beam generated during the sphere

calibration process. No mapping of the whole integrating sphere was made, but some

uniformity tests around the measurement and calibration direction were made.

Fig. 4-21. LED mount in the integrating sphere in METAS.

4.8.3. Traceability

All primary quantities (i.e. illuminance, length, current, voltage etc) and secondary

quantities (temperature, humidity, etc) are traceable to national standards realized at

METAS. The detailed view of the traceability of the primary quantities is shown in the

following diagram.

Averaged LED intensity

METAS Electricity Section

ULED, ILED

Reference Integrating sphere METAS

Length Section

Distance, Aperture

Luminous flux of LED

METAS Electricity Section

ULED, ILED, IPhoto

APMP-PR.S3a

(METAS)


61


The uncertainty budgets are based on the recommendation of CIE TC2-43

“Determination of measurement uncertainty in photometry”, Draft 9, 2008, and thus

following the GUM.

For simplicity only the uncertainty budget for a green LED is illustrated explicitly

in the following. The estimated input quantities of the other LED’s are listed in their

description.

Model for total luminous flux:

22

11111

2

1

0

01

/21

)(/21

aSSSUU

aSSSSSUSDMCPCM

SU

UCS

cc

mmCS

Tdd

ThddfG

d

AI

VV

VV

Description of terms:

1CS output quantity: luminous flux of the LED at certified conditions.

mV = 0.269054 V, DVM signal photometer, the DUT-LED is installed inside the

integrating sphere, 10n independent readings, the SDM is taken as standard

MU mVu 0.000004 V and is significantly larger than the resolution; Type A

with DOF 9v , no correlation.

0mV = 0.000054 V, DVM dark signal photometer, the DUT-LED is installed inside the

integrating sphere, 10n independent readings, the SDM is taken as standard

MU m0Vu 0.000001 V and is significantly larger than the resolution; Type A

with DOF 9v , no correlation.

cV = 0.299314 V, DVM signal photometer, the DUT-LED is installed outside the

integrating sphere (calibration), a dummy LED of same color is inside, 10n

independent readings, the SDM is taken as standard MU cVu 0.000011 V and

is significantly larger than the resolution; Type A with DOF 9v , no correlation.

0cV = 0.000241 V, DVM dark signal photometer, the DUT-LED is installed outside

the integrating sphere (calibration), a dummy LED of same color is inside,

10n independent readings, the SDM is taken as standard MU c0Vu

0.000007 V and is significantly larger than the resolution; Type A with DOF

9v , no correlation.

1CSI = 2.7951 cd, luminous intensity of the LED used for calibrating the sphere (c.f.

METAS report on APMP PR-S3.a); The standard MU Type B )( 1CSIu = 0.0224

with DOF v , no correlation.

UA = 100 mm2, limiting entrance aperture used in front of the integrating sphere for its


62

calibration. The standard MU Type B )( uAu = 0.01 mm2 with DOF v , no

correlation.

SUd = 0.100 m, distance between tip of the LED and the limiting entrance aperture of

the integrating sphere, interval ±0.00020 m with RPD, converted into standard

measurement uncertainty (MU) )( SUdu = (0.00020/ 3 = 0.000115) m; Type B

with degree of freedom (DOF) v , no correlation.

PCMG = , gain factor of the photometer when switching calibration to the

measurement certified with absolute standard MU PCMGu = 0.2 ; Type B with

DOF v , no correlation.

DMCf = 1.00 (spatial) distribution mis-match correction factor of the integrating when

switching from calibration to the measurement. No mapping of the whole

integrating sphere was made. But some uniformity tests around the measurement

and calibration direction were made. As a high reflectance integrating sphere is

used and the DUT is illuminating similar part of the integrating sphere the absolute

standard MU is estimated to DMCfu = 0.003 ; Type B with DOF v , no

correlation.

SUS1 dd = (0 ± 0.2) mm/100 mm, distance alignment of LED tip within interval with

RPD, converted into standard MU SUP ddu = 0.2/(100* 3 ) = 0.0012;

Type B with DOF v , no correlation.

)( 11 SSh = 0.0, angular misalignment of the LED within interval 1S 2° with RPD

converted into standard MU 2022

11 ghu SS )( = 0.0025;

Type B with DOF v , no correlation. )log(cos/).log( .5050 g = 9.0, is

determined from the FMHW 50. (datasheet of the green LED). For the other

LED’s the values are g (red) = 6.9, g (blue) = 9.0, g (white) = 3.2, g (diffuse)

= 1.0 . The uncertainty on g is neglected.

S1 = -0.0019 -1K , relative temperature coefficient of the green LED (based on the

datasheet) used during calibration procedure, with standard MU

S1u = (0.0002/2 = 0.0001) -1K ; Type B with DOF v , no correlation. For

other LED’s the temperature coefficient is estimated as: S1 (red) = (-0.0074 ±

0.0005) -1K , S1 (blue) = (0.00175 ± 0.00020)

-1K , S1 (white) = (0.0016

± 0.0005) -1K

aS1T = 0.0°C, above nominal ambient temperature near LED (outside the integrating

sphere, i.e. during the calibration procedure), with standard MU aS1Tu = (0.5/

3 =0.28) °C; Type B with DOF 1000v , no correlation.

SUU dd = (0 ± 0.2) mm/100 mm, distance alignment of integrating sphere aperture

within interval with RPD, converted into standard MU )( SUU ddu = 0.2/(100*

3 ) = 0.0012; Type B with DOF v , no correlation.


63

S2 = S1 relative temperature coefficient of the green LED (based on the datasheet)

used inside the integrating sphere (i.e. during the measurement procedure).

aS2T = 0.0°C, above nominal ambient temperature near LED inside the sphere

(measurement procedure), with standard MU aS2Tu = (0.5/ 3 =0.28)°C;

Type B with DOF 1000v , no correlation.

The following quantities were ignored:

- The influence of the ambient temperature uncertainty on the photometer as a temperature

stabilized photometer was used.

- ageing of the DUT as no relevant information was available.

- variation of the output intensity as a change of electrical current (c.f. luminous intensity

report).

- Calibration factor and its uncertainty of the DVM (c.f. luminous intensity report).

- straylight effects (not estimated) during calibration.

- angular and directional misalignment of the integrating sphere during calibration.

- influence of the directional change of spectral distribution of LED (the sphere is calibrated

with an LED in CIE averaged intensity condition (100mm), therefore not all directions are

included during calibration process).

Sensitivity coefficients:

m

CS

m

CS

VVc 11

1

9.344 lm/V

m

CS

m

CS

VVc 1

0

12

-9.344 lm/V

C

CS

C

CS

VVc 11

3

-8.3993 lm/V

C

CS

C

CS

VVc 1

0

14

-8.3993 lm/V

1

1

1

15

CS

CS

CS

CS

IIc

0.8994 lm/cd

U

CS

U

CS

AAc 11

6

25140 lm/m

2

SU

CS

SU

CS

ddc 11

7 2

50.28 lm/m

PCM

CS

PCM

CS

GGc 11

8

0.0251 lm

DMC

CS

DMC

CS

ffc 11

9

2.5140 lm

1

110 2 CS

CS

ddc

)( SUS1

5.0281

lm

1

11

111 CS

SS

CS

hc

)( 2.5140 lm

aS

S1

Tc CSCS

1

112

0.000 lm K

S1

aS

1

113 CS

CS

Tc

= -0.004777 lm K

-1

1

114 2 CS

CS

ddc

SUU

-


64

5.0281 lm

aS2

S2

Tc CSCS

1

115

0.000 lm K S2

aS2

1

116 CS

CS

Tc

=0.004777 lm K

-1

Table 4-43. METAS uncertainty budget of total luminous flux measurement for red LEDs (R).


uncertainty

Ty

pe

Probability

distribution

Sensitivity

coefficient

Contribut

ion (%)

Deg. of

freedo

m

Correl

ated?

Mean value photosignal

mV

4E-6 V A t 1.1342 lm/V 0.006 9 X

Mean value dark

photosignal 0mV

1E-6 V A t -1.1342

lm/V

<0.001 9 X


calibration CV

1.1E-6 V A t -0.1341

lm/V

-0.002 9 X

Mean value dark

photosignal calibration

0CV

7E-6 V A t 0.1341 lm/V <0.001 9 X

Intensity of calib. LED at

normal current 1CSI

0.0224 cd B normal 0.1190 lm/cd 0.69 ∞ O

Limiting entrance

aperture UA

1.0E-8 m2 B normal 822 lm/ m

2 0.01 ∞ O

Distance LED to

integrating sphere SUd

0.000115

m

B rectangular -1.64 lm/m -0.23 ∞ X

Gain switching factor of

the photometer PCMG

0.2 B normal 0.00082 lm 0.20 ∞ O

(Spatial) distribution

mismatch correction

factor DMCf

0.003 B normal 0.0822 lm 0.30 ∞ X

Relative distance

variation of LED SUS1 dd

0.0012 B rectangular 0.1643 lm 0.24 ∞ X

Angular misalignment of

LED )( 11 SSh


Temperature coefficient

of LED S1 and S2

0.0001 B normal 0 0 ∞ X

Temperature above

nominal temp., calibration

aS1T

0.28 K B rectangular -0.000608

lm/K

-0.21 ∞ X

Temperature above


aS2T

0.28 K B rectangular 0.000608

lm/K

0.21 ∞ X

Distance alignment of

integrating sphere

aperture SUU dd

0.0012 B rectangular -0.1643 lm -0.24 ∞ X


65

Combined standard

uncertainty (%)

-- -- normal -- 0.95 > 1000 --

Table 4-44. METAS uncertainty budget of total luminous flux measurement for green LEDs

(G).


uncertainty

Ty

pe

Probability

distribution

Sensitivity

coefficient

Contribut

ion (%)

Deg. of

freedo

m

Correl

ated?


mV

4E-6 V A t 9.3440 lm/V <0.001 9 X

Mean value dark

photosignal 0mV

1E-6 V A t -9.3440

lm/V

<0.001 9 X


calibration CV

1.1E-6 V A t -8.3993

lm/V

-0.004 9 X

Mean value dark


0CV

7E-6 V A t 8.3993 lm/V 0.002 9 X


normal current 1CSI

0.0224 cd B normal 0.8994 lm/cd 0.80 ∞ O

Limiting entrance

aperture UA

1.0E-8 m2 B normal 25140 lm/

m2

0.01 ∞ O

Distance LED to


0.000115

m



the photometer PCMG

0.2 B normal 0.0251 lm 0.20 ∞ O


mismatch correction

factor DMCf

0.003 B normal 2.5140 lm 0.30 ∞ X

Relative distance




LED )( 11 SSh



of LED S1 and S2

0.0001 B normal 0 0 ∞ X

Temperature above


aS1T


lm/K

-0.05 ∞ X

Temperature above


aS2T


lm/K

0.05 ∞ X


integrating sphere

aperture SUU dd


Combined standard

uncertainty (%)

-- -- normal -- 1.00 > 1000 --


66

Table 4-45. METAS uncertainty budget of total luminous flux measurement for blue LEDs (B).


uncertainty

Ty

pe

Probability

distribution

Sensitivity

coefficient

Contribut

ion (%)

Deg. of

freedo

m

Correl

ated?


mV

4E-6 V A t 1.2876 lm/V 0.005 9 X

Mean value dark

photosignal 0mV

1E-6 V A t -1.2876

lm/V

<0.001 9 X


calibration CV

1.1E-6 V A t -0.1588

lm/V

-0.002 9 X

Mean value dark


0CV

7E-6 V A t 0.1588 lm/V <0.001 9 X


normal current 1CSI

0.0224 cd B normal 0.1229 lm/cd 1.61 ∞ O

Limiting entrance

aperture UA


2 0.01 ∞ O

Distance LED to


0.000115

m



the photometer PCMG

0.2 B normal 0.00109 lm 0.20 ∞ O


mismatch correction

factor DMCf

0.003 B normal 0.1094 lm 0.30 ∞ X

Relative distance




LED )( 11 SSh



of LED S1 and

S2

0.0001 B normal 0 0 ∞ X

Temperature above


aS1T


lm/K

-0.05 ∞ X

Temperature above


aS2T


lm/K

0.05 ∞ X


integrating sphere

aperture SUU dd


Combined standard

uncertainty (%)

-- -- normal -- 1.72 > 1000 --

Table 4-46. METAS uncertainty budget of total luminous flux measurement for white LEDs

(W).


67


uncertainty

Ty

pe

Probability

distribution

Sensitivity

coefficient

Contribut

ion (%)

Deg. of

freedo

m

Correl

ated?


mV

4E-6 V A t 4.5301 lm/V 0.005 9 X

Mean value dark

photosignal 0mV

1E-6 V A t -4.5301

lm/V

<0.001 9 X


calibration CV

1.1E-6 V A t -2.3580

lm/V

-0.007 9 X

Mean value dark


0CV

7E-6 V A t 2.3580 lm/V 0.005 9 X


normal current 1CSI

0.0224 cd B normal 0.5188 lm/cd 0.66 ∞ O

Limiting entrance

aperture UA


2 0.01 ∞ O

Distance LED to


0.000115

m



the photometer PCMG

0.2 B normal 0.00356 lm 0.20 ∞ O


mismatch correction

factor DMCf

0.003 B normal 0.3560 lm 0.30 ∞ X

Relative distance




LED )( 11 SSh



of LED S1 and

S2

0.0001 B normal 0 0 ∞ X

Temperature above


aS1T


lm/K

-0.04 ∞ X

Temperature above


aS2T


lm/K

0.04 ∞ X


integrating sphere

aperture SUU dd


Combined standard

uncertainty (%)

-- -- normal -- 1.73 > 1000 --

Model for junction voltage:

L0L1aL0aLrelL,CLL 1 UUTTccU

Description of terms:


68

LU output quantity: junction voltage of the LED at certified conditions.

Lc = 1.0000, DVM calibration factor with absolute standard MU )( Lcu = 1E-5;


Cc = 1.000, non-equivalence of the contact. We have tried different connectors. A

spread in junction voltages have been observed even with 4 wires connections. The

estimated absolute standard MU )( Ccu = 0.0020; Type B with DOF v , no

correlation.

relL, = 0.000015, relative temp. coefficient according standard MU )( relL,u = 5E-6;


aLT = 22.6 °C, ambient temperature with ±0.5°C RPD, converted into standard MU

)( aLTu = (0.5/ 3 = 0.29)°C; Type B with DOF v , no correlation.

aL0T = 23.0 °C, nominal ambient temperature, no uncertainty

L1U = 1.94058 V, measured voltage (DVM), with standard MU of L1Uu = 0.00011

V, 361 readings, Type A with DOF 360v , no correlation.

L0U = 0.00002 V, measured zero voltage (DVM), with standard MU of L1Uu =

0.00011 V, 361 readings , Type A with DOF 360v , no correlation.

Table 4-47. METAS uncertainty budget of junction voltage measurement.


uncertainty

Ty

pe

Probability

distribution

Sensitivity

coefficient

Contrib

ution

(%)

Deg. of

freedo

m

Correl

ated?

DVM mean value voltage

L1U

0.00011 V A t 0.99999

V/V

0.0055 360 X

DVM calibration factor

Lc

1.0E-5 B normal 1.94055 V 0.0010 ∞ O

Relative temperature

coefficienct relL,

5.0E-6 K-1

B normal -0.77622

VK

-0.0002 ∞ X

Ambient temperature aLT 0.29 °C B rectangular 0.00003

V/°C

0.0004 ∞ X

Offset voltage L0U 0.00011 V A t -0.99999

V/V

-0.0055 360 X

Non-equivalence of contact

Cc

0.0021 B rectangular 1.94055 V 0.21 ∞ X

Combined standard

uncertainty (V)

-- -- Normal -- 0.21 >1000 --


69

4.9. NMC-A*STAR


The measurement setup of the total luminous flux of LED is shown from Fig. 4-22 to Fig.

4-24. The LED is mounted at the centre of a 1-meter integrating sphere. The LED light in

the sphere is fed to a spectroradiometer (Model OL770 made by Optronic Laboratories,

see report for S3a) through an optical fibre as shown in Fig. 2. A baffle and an opal

glass diffuser are mounted in front of the tip of the optical fibre to avoid the direct

illumination from the LED.

Fig. 4-22. LED total luminous flux measurement setup in A*STAR.

Fig. 4-23. Relative spectral responsivity calibration;

Fig. 4-24. Absolute luminous flux calibration in A*STAR.


70


The LED holder has a flexible arm which allows the LED to be pointed to any direction of

the sphere to access the correction factor of the spatial non-uniformity of the integrating

sphere for each type of the LED.

4.9.3. Traceability

The relative spectral responsivity of the sphere spectroradiometer is calibrated by a

spectral irradiance standard lamp traceable to NMC’s spectral irradiance scale as shown

in Fig. 4-23 similar to Yoshi Ohno’s method. The stray light error of the spectroradiometer

is corrected using cut-on filters. The absolute luminous flux responsivity of the sphere

spectroradiometer is calibrated using a luminous flux standard lamp traceable to NMC’s

total luminous flux scale as shown in Fig. 4-24.

A 50 W tungsten halogen auxiliary lamp is used for substitution error

compensation affected by lamp holder, calibration lamps, test LED and any other items

used inside the sphere or at its opening port. The absorption corrections were carried

out over the whole wavelength range of 380 nm to 780 nm in 1 nm interval for both the

sphere calibration and the LED measurement.


Tables in the following are the detailed uncertainty budgets of total luminous flux

measurement for the LEDs used in this APMP LED comparison.

The uncertainty evaluation is carried out according to Guide to the Expression of

Uncertainty in Measurement (GUM). The artefact-dependent uncertainties shown in the

table with * adopt the largest uncertainty values registered among the same type of LEDs

measured. Expanded uncertainty are evaluated at a confidence level of approximately 95%

with a coverage factor normally k = 2.

Table 4-48. A*STAR uncertainty budget of total luminous flux measurement for red LEDs (R).


uncertainty

Ty

pe

Probability

distribution

Sensitivity

coefficient

Contrib

ution

(%)

Deg. of

freedo

m

Correl

ated?

Calibration of flux standard

lamp

B normal 1 0.450 ∞ Yes

Drift of flux standard lamp B rectangular 1 0.289 ∞ Yes


71

Sphere-radiometer transfer

measurement (non-

linearity)*

B rectangular 1 0.405 ∞ No

Sphere spatial uniformity B rectangular 1 0.289 ∞ Yes

Calibration of current

feeding

0.0058 % B rectangular 0.8 0.005 ∞ Yes

LED holder absorption B rectangular 1 0.116 ∞ Yes

Wavelength scale of

spectroradiometer*

0.2 nm B rectangular 2.17 %/nm 0.434 ∞ No

stray light correction of

spectroradiometer (20 % of

correction)*


Reproducibility A t 1 0.173 2 No

Combined standard

uncertainty (%)

-- -- normal -- 0.90 1444 --

Table 4-49. A*STAR uncertainty budget of total luminous flux measurement for green LEDs

(G).


uncertainty

Ty

pe

Probability

distribution

Sensitivity

coefficient

Contrib

ution

(%)

Deg. of

freedo

m

Correl

ated?


lamp




measurement (non-

linearity)*




feeding



Wavelength scale of

spectroradiometer*




correction)*



Combined standard

uncertainty (%)

-- -- normal -- 0.78 4148 --

Table 4-50. A*STAR uncertainty budget of total luminous flux measurement for blue LEDs (B).


72


uncertainty

Ty

pe

Probability

distribution

Sensitivity

coefficient

Contrib

ution

(%)

Deg. of

freedo

m

Correl

ated?


lamp




measurement (non-

linearity)*




feeding



Wavelength scale of

spectroradiometer*




correction)*



Combined standard

uncertainty (%)

-- -- normal -- 0.87 395 --

Table 4-51. A*STAR uncertainty budget of total luminous flux measurement for white LEDs

(W).


uncertainty

Ty

pe

Probability

distribution

Sensitivity

coefficient

Contrib

ution

(%)

Deg. of

freedo

m

Correl

ated?


lamp




measurement (non-

linearity)*




feeding



Wavelength scale of

spectroradiometer*



73



correction)*



Combined standard

uncertainty (%)

-- -- normal -- 0.81 295 --

Table 4-52 is the detailed uncertainty budget of the junction voltage measurement,

representatively presented for the red LEDs. The artefact-dependent uncertainties shown

in the table with * adopt the largest uncertainty values registered among the same type

of LEDs measured.

Table 4-52. A*STAR uncertainty budget of junction voltage measurement.


uncertainty

Ty

pe

Probability

distribution

Sensitivity

coefficient

Contribut

ion (V)

Deg.

of

freedo

m

Correl

ated?

Calibration of DVM B normal 1 9.50E-5 ∞ Yes

Position of junction (0.05

Ω)

B rectangular 1 5.78E-4 ∞ No

Drift of junction voltage B rectangular 1 1.73E-4 ∞ No

Reproducibility* A t 1 4.66E-4 5 No

Combined standard

uncertainty (V)

-- -- normal -- 7.7E-4 37 --

4.10. VSL


The quantity for average LED intensity and total luminous flux of LEDs (as defined by the

key-comparison protocol) are measured with a goniometer facility specifically designed

and build for small single LED light sources. The facility is based on the method where

the light source is turned and the detector stands still. Therefore the facility consists out

of a detector platform and a turn-able light source unit. The light source unit includes

two rotation stages, a LED mounting unit and one linear translation stage. The linear

translation stage is applied to be able to change the distance between the turn-able

light source unit and the detector platform. The two rotation stages are perpendicular

mounted to each other so that the LED can be rotated exactly in the midpoint of each


74

stage.

The detector platform consists out of an illuminance meter with a circular

aperture with a surface of 100mm2 and an array-spectroradiometer (SRM). The SRM is

used to correct for colour mismatch introduced by the detector and the individual LED. In

order to reduce stray light a baffle was places between the detector platform and the

turn-able light source unit. The aperture of the baffle was large compare to the diameter

of the detector and the LED to be measured.

Fig. 4-25. Schematic drawing of LED goniometer facility at VSL.


The LED is fixed in a holder, which is mounted into a mounting unit. The mounting unit

is mounted on the turn-able light source unit consisting out of the two rotation stages.

The LED holder is shown in the following figure.

Fig. 4-26. VSL LED holder.

The LED holder clamps the two LED pins with two parallel copper plates. The

copper plates are connected to the current source which provides the LED with operating


75

current. The mounting unit allows one to translate the LED in both vertical as well as

horizontal direction, and also to tilt the LED. This alignment unit is in turn mounted to

the two rotation stages. The layout of the alignment system of the LED facility together

with the mounted holder is shown in the following figure.

Fig. 4-27. Turn-able light source unit of the LED goniometer facility at VSL.

In Fig. 4-27, one sees the LED mounted on the mounting unit fixed on a two axis

rotational system. The alignment of the LED with regards to the detector as well as axis

of rotation is done as follows:

1. A high resolution camera is placed perpendicular to the mounted LED.

2. The mounted LED is rotated and visually inspected by using the high resolution camera.

3. If the mounted LED is in the centre of the rotational axis, no movement is detected

with the camera, otherwise translation is observed. The mounted LED is then

iteratively adjusted until no translation of the mounted LED is visible with the camera.

This is iteratively repeated also for the polar rotation. When varying the polar angle the

alignment criteria was that the location of the LED tip remained constant.

4. The mounted LED and illuminance detector are then optically aligned with the double

alignment laser.

The nominal distance between LED and detector is brought to 100 mm by making

use of an electronic translation stage where the LED alignment axes are mounted on, as

well as a calibrated gauge block of nominal length 100 mm. The gauge block is placed

against the detector reference surface and the LED is translated precisely until contact is

made with the gauge block. This translation distance is recorded. The gauge block is

then removed and the LED is translated back to the correct position. The distance is then

100 mm between detector and LED. The following figure illustrates this graphically.


76

Fig. 4-28. Schematic drawing of the detector versus LED distance determination at VSL.

4.10.3. Traceability

The total luminous flux of a LED measurement at VSL has as the traceability route as

shown in Fig. 4-29.

Fig. 4-29. Traceability of LED total luminous flux measurement at VSL.

The spectral responsivity scale is derived from an Absolute Cryogenic Radiometer

(ACR) by using a double monochromator facility 7 . The same facility is used for the

determination of the illuminance responsivity by using a scanning beam method and the

relative spectral irradiance responsivity of the illuminance meter 8 . Knowing the

illuminance responsivity of an illuminance meter and using a calibrated gauge block one

can determine the luminous intensity of a LED. The gauge block is calibrated and

traceable to the national standard for length. Each measurement within the traceability

chain is conducted by using digital multimeters for measurement of detector current, LED

current and LED voltage. These measurements are traceable to the national standard for

current and voltage by the use of calibrated meters.

7 Comparison of monochromator-based and laser-based cryogenic radiometry, Metrologia 1998, 35, 431-435. 8 Novel calibration method for filter radiometers, Metrologia 1999, 36, 179-182.

Cryogenic radiometer VSL Spectral responsivity scale

(A/W)

ACR facility VSL Illuminance responsivity

(A/lx)

LED Goniomter facility VSL Average luminous intensity and total

luminous flux (cd) and or (lm)

Electrical department for the traceability to the national standard of current and voltage (A) and (V)

Length department for the traceability to the national standard of length (m)


77


After the LED and detector are aligned, the following steps are performed to measure

the total luminous flux of each of the twelve LEDs respectively:

1. The LED is brought to an operating current of nominal 20 mA.

2. The whole setup is enclosed by a thermal insulation box and allowed to stabilize for at

least 20 minutes.

3. The measurement of the illuminance at different angles are performed by varying the

polar angle of the LED from 0° to 125° in 5° increments, repeating this for an azimuthal

rotations from 0° to 360° in 5° increments, thereby effectively scanning a partial sphere

of 2.78π around the LED tip.

4. The stray light was measured by blocking light only on the optical axis and repeating

step 3. The light was blocked by using a strip with an effective area just greater then

the surface of the LEDs so no direct light from the LED was seen by the detector.

5. The dark signal was measured, by closing the baffle situated in front of the detector

completely and repeating step 3.

6. The illuminance of the LED at each goniometer position is calculated as described in the

report for S3a. The responsivity of the detector is corrected for the spectral mismatch

by using the spectral irradiance measurement conducted with the spectroradiometer

at polar position 0, 0.

7. Finally the total luminous flux of the LED is calculated using model equation below.

Model equation for the total luminous intensity:

)sin())(,()(sin),(),(

1

2

2

2

2

nm

i

v ErddErdAE

A is the surface area

E

(ε,η)

is the measured illuminance at a certain position

r is the radius of the sphere

ε is the polar angle

η is the azimuthal angle

δ is step size for or azimuth or polar axis

m the amount of steps along the polar angle

n the amount of steps along the azimuth angle

The comparison protocol states that the participant describes the total uncertainty

in detail for the LEDs of each color. As the total uncertainty of each LED is depending on

individual components the uncertainty from one LED to one other is different. Knowing


78

this we chose to present a detailed uncertainty budget of that LED that has the lowest

uncertainty, instead of determining the average total uncertainty of the LEDs with the

same color. This was done since no information is given how to determine the average

uncertainty of a group of LEDs. The detailed uncertainty budgets are summarized in the

tables below.

Table 4-53. VSL uncertainty budget of total luminous flux measurement for red LEDs (R).


uncertainty

(%) Ty

pe Probability

distribution

Sensitivity

coefficient

Contri-

bution

(%)

Deg. of

freedom

Correlated

Spectral mismatch

correction

B normal 1 0.21 ∞ X

Reproducibility B rectangular 1 0.45 ∞ X

Current feeding of LED B normal 1 0.01 ∞ O

Near-field absorption of

backward emission


Stray light A normal 1 0.28 9 O

Missing emitted flux B rectangular 1 1.17 ∞ X

Alignment of LED A normal 1 0.10 28 X

Distance between LED and

detector


Responsivity of detector B normal 1 0.15 ∞ O

Detector readout A normal 1 0.03 9 O

Combined standard

uncertainty (%)

-- -- normal -- 1.46 ∞ --

Table 4-54. VSL uncertainty budget of total luminous flux measurement for green LEDs (G).


uncertainty

(%) Ty

pe Probability

distribution

Sensitivity

coefficient

Contri-

bution

(%)

Deg. of

freedom

Correlated

Spectral mismatch

correction





backward emission






detector




Combined standard

uncertainty (%)

-- -- normal -- 1.20 ∞ --


79

Table 4-55. VSL uncertainty budget of total luminous flux measurement for blue LEDs (B).


uncertainty

(%) Ty

pe Probability

distribution

Sensitivity

coefficient

Contri-

bution

(%)

Deg. of

freedom

Correlated

Spectral mismatch

correction





backward emission






detector




Combined standard

uncertainty (%)

-- -- normal -- 1.30 ∞ --

Table 4-56. VSL uncertainty budget of total luminous flux measurement for white LEDs (W).


uncertainty

(%) Ty

pe Probability

distribution

Sensitivity

coefficient

Contri-

bution

(%)

Deg. of

freedom

Correlated

Spectral mismatch

correction





backward emission*



Missing emitted flux** B rectangular 1 1.41 ∞ X



detector




Combined standard

uncertainty (%)

-- -- normal -- 1.58 ∞ --

* The LED mount used in the measurements is black to absorb backwards emission i.e. rather than choosing

a highly reflective mount to include the backwards emission in the illuminance measurement. This means

that the backwards emission is filtered out from the goniometric illuminance measurements alternatively.

However some backwards emission may reflect of the black mount and contribute to the forward

illuminance measurement. As this leads to an uncertainty we have measured the flux emitted directly at the

backside of the LED to estimate the flux reflecting from the mount with taken into account the reflection

coefficient of the mount surface and the effective area illuminated by the reflected light. This then is taken

as the uncertainty for the near-field absorption of backwards emission.

** Do to the structure of the goniometer facility it is not possible to measure the total polar plane from 0° to

180°. Therefore the illuminance measured at polar angle 125° is extrapolated till 180°. The model for

extrapolation is based on the knowledge from a measurement directly behind the LED itself performed

outside the goniometer facility at a distance of 100 mm. The associated uncertainty for the extrapolation is


80

based on the estimated flux within the missing cone from 125° to 180°.

Table 4-57 is the detailed uncertainty budget of the junction voltage

measurement.

Table 4-57. VSL uncertainty budget of junction voltage measurement.


uncertainty

Ty

pe

Probability

distribution

Sensitivity

coefficient

Contribut

ion (%)

Deg.

of

freedo

m

Correl

ated?

Calibration of DVM B normal 1 1.2E-5 ∞ O

Junction position

dependence


Reproducibility* A t 1 0.0001 9 X

Combined standard

uncertainty (%)

-- -- normal -- 0.081 ∞ --

4.11. NIST


The test LEDs were measured for total luminous flux with 4π geometry in the NIST 2.5 m

detector-based absolute integrating sphere (with 98 % reflectance barium sulfate coating)

with the scale realized in 2009. The schematic of the measurement setup is shown in Fig.

4-30. The reference standard of the 2.5 m absolute sphere system is the luminous flux of

the external source introduced into the sphere through a Ø50 mm precision aperture.

The illuminance of the external source at the precision aperture plane is measured by

two standard photometers to calculate the luminous flux entering into the sphere. For a

measurement of total luminous flux, the test LED and the external source illuminated

directly, in turn, a different part of the sphere wall on the equator. The error arising from

the spatial mismatch in comparison to an isotropic light source inside the sphere was

analyzed and corrected for both the LED and the external source. The details of the

measurement facility and procedures are described in Reference9.

9 Ohno Y. and Zong Y., Detector-Based Integrating Sphere Photometry, in Proc. of 24th Session of the CIE, Vol. 1, Part 1, 155-160. (1999) / Miller C. C., and Ohno Y., Luminous Flux Calibration of LEDs at NIST, in Proc. of 2nd CIE Expert Symposium on LED Measurement, 45-48. (2001)


81

Fig. 4-30. Illustration of the setup for measurement of total luminous flux of the test LEDs in NIST.

The test LED was operated on DC power at a constant current of 20 mA using a

four-wire connection. The wiring diagram for this measurement is shown in Fig. 4-31. The

operating current of the LED was measured with an 8.5 digit multimeter. The test LED

was measured after it was powered on for 10 minutes. The output signal of the sphere

photometer was simultaneously recorded with the LED current, LED voltage, sphere

ambient temperature, room temperature, and room humidity. Corrections were applied

for the dark reading, the self-absorption (automatically corrected), the spectral mismatch,

the spatial mismatch, and the sphere fluorescence (see next paragraph), to calculate the

total luminous flux of a test LED. Each LED was measured for a total of two lightings to

check its reproducibility. The mean value of the two measurements was reported, and the

variation was included in the uncertainty budget of the measurement.

Fig. 4-31. Wiring diagram for measurement of a test LED in NIST.

After the measurement of total luminous flux, each LED was measured in the

same 2.5 m integrating sphere for relative total spectral radiant flux using a CCD-array

spectroradiometer in order to correct the spectral mismatch error and sphere


82

fluorescence error. The measurement was based on the NIST spectral irradiance scale10

as described in Reference11. The sphere-spectroradiometer system, shown in Fig. 4-32,

was calibrated for total spectral radiant flux responsivity against two standard spectral

irradiance FEL lamps aligned in turn at 0.5 m away from the Ø50 mm precision aperture.

The two standard FEL lamps were calibrated in the direction of its optical axis for

absolute spectral irradiance at 0.5 m in the NIST Facility for Automated

Spectroradiometric Calibrations (FASCAL). The CCD-array spectroradiometer has a

bandpass of approximately 3 nm (FWHM) and the spectral range from 200 nm to 800

nm. A heat-absorbing optical filter (Schott KG-5) was inserted between the opal glass

diffuser and the optical fiber bundle to prevent the unwanted infrared radiation of the

standard spectral irradiance FEL lamp from entering into the spectroradiometer in order

to reduce stray light inside the spectroradiometer. The integrating-time nonlinearity and

signal-level nonlinearity of the spectroradiometer were both corrected. The

spectroradiometer was first characterized for spectral stray light12 and then was used to

measure a set of laser sources to characterize the fluorescence of the 2.5 m sphere

coating. The measured relative total spectral radiant flux of the test LED was corrected

for both spectral stray light of the spectroradiometer and the fluorescence of the 2.5 m

sphere, and was used to correct the spectral mismatch error. The error resulting from the

sphere fluorescence was analyzed and corrected based on the characterization result of

the sphere fluorescence.

10 J. H. Walker, R. D. Saunders, J. K. Jackson, and D. A. McSparron, Spectral Irradiance Calibrations, NBS Special Publication 250-20. (1987) / Yoon H. W., Gibson C. E., and Barnes P. Y., Realization of the National Institute of Standards and Technology detector-based spectral irradiance scale, Appl. Opt. 41, 5879-5890. (2002) 11 Zong Y., Miller C. C., Lykke K. R., and Ohno Y., Measurement of total radiant flux of UV LEDs, Proc. CIE, CIE x026:2004, 107–110. (2004) 12 Zong Y., Brown S. W., Johnson B. C., Lykke K. R., and Ohno Y., Simple spectral stray light correction method for array spectroradiometers, Appl. Opt., 45, 1111-1119. (2006)


83

Fig. 4-32. Schematic of the setup for measurement of relative total spectral radiant flux of LEDs in NIST.


The test LED was mounted horizontally on the lamp post at the center of the NIST 2.5 m

integrating sphere by using a four-wire, C-shaped LED socket/holder for minimizing the

near-field absorption and for including any backward light. Fig. 4-33 is a photograph of a

test LED mounted at the center of the 2.5 m integrating sphere.

Fig. 4-33. Photograph of a test LED mounted at the center of the 2.5 m integrating sphere in

NIST.


The two standard photometers, mounted on the wheel (shown in Fig. 4-30), used to

measure illuminance of the external source were calibrated for spectral irradiance

responsivity in the NIST tuneable-laser-based SIRCUS facility. The calibration was done by

direct comparison of the photometer with two of the NIST trap detectors, which maintain


84

the NIST spectral irradiance scale and are periodically calibrated against the NIST

Reference Cryogenic Radiometer - Primary Optical Watt Radiometer (POWR).


The uncertainty budgets for measurement of total luminous flux of the red, green, blue,

and white LEDs are shown in the tables below, and the uncertainty budget for

measurement of junction voltage of the test LEDs is shown in Table 4-62. The NIST policy

on uncertainty statements is described in Reference13.

13 B. N. Taylor, and C. E. Kuyatt, Guidelines for Evaluating and Expressing the Uncertainty of NIST Measurement Results, NIST Technical Note 1297. (1993)


85

Table 4-58. NIST uncertainty budget of total luminous flux measurement for red LEDs (R).


86

Table 4-59. NIST uncertainty budget of total luminous flux measurement for green LEDs (G).


87


88

Table 4-60. NIST uncertainty budget of total luminous flux measurement for blue LEDs (B).


89

Table 4-61. NIST uncertainty budget of total luminous flux measurement for white LEDs (W).


90

Table 4-62. NIST uncertainty budget of junction voltage measurement (typical).

4.12. VNIIOFI

Not submitted.

4.13. INM


A lumen-meter equipped with a 150 mm dia. integrating sphere provided with a

precision aperture was used (Fig. 4-34). It allowed for comparison of the LED under

calibration with a standard luminous intensity lamp basically using the substitution

method.


91

Fig. 4-34. LED total luminous flux measurement setup in INM Romania.

For electrical measurements, a four wire technique as described in the comparison

protocol was used in order to (almost) simultaneously measure the current fed into the

measured LED and the junction voltage. The LED current was generated by a finely

adjustable voltage supply and a current measurement shunt across which the voltage

was measured with a digital voltmeter. The LEDs junction voltage was measured with a

similar digital voltmeter.

A photometric head provided with a diffusing IR filter and calibrated in terms of

spectral responsivity was attached to the lumen-meter integrating sphere. The

photocurrent generated by the photometric head was fed into Current to Voltage

converter with a transimpedance factor of 1E6 V/A. The output voltage was measured

with a third digital voltmeter.

The measurement of the spectral densities of the emitted flux of the standard

lamp and of the LED under calibration was performed with a CCD spectrometer

providing a (1 ± 0.1) nm bandwidth. The spectrometer input fibre head was provided

with a diffusing IR filter.


The lumen-meter was calibrated in terms of luminous flux responsivity against the

luminous flux produced by a luminous intensity lamp. The calibration of the lumen-meter

as a whole was performed on the INM optical bench using the regular procedure for

photometers calibration (based on the distances inverse squares law).

Subsequently, the LEDs to be calibrated were mounted in the lumen-meter sphere

in such a position as to illuminate almost the same area previously illuminated by the


92

luminous intensity lamp during the lumen-meter calibration. In order to avoid the direct

illumination of the photometer, a shade was mounted in front of the photometer

transducer (Fig. 4-34).


The lumen-meter (including the 150 mm dia. integrating sphere, the photometer head,

the current to voltage converter and the associated multimeter) was calibrated against

a luminous intensity standard traceable to the national reference for luminous intensity

(group of absolute photometers) maintained by INM-RO. The calibration was performed

at several distances so that the lumen-meter photometric linearity could be checked to

be within ±0.5 %.

The lumen-meter transducer (IR filtered photo-diode) spectral responsivity was

characterised against the INM spectral responsivity references traceable to LNE-INM

primary reference (cryogenic radiometer).

The 150 mm dia. sphere wall was coated with multiple layers of BaSO4 (>20

layers). The last 10 layers were sprayed without any binder. A test sample coated in a

similar manner was characterised in terms of spectral reflectance (0/d geometry) against

standards traceable to the INM reference standard (primary reflectance standard based

on the Taylor-Budde method).

The spectral densities of the standard lamp and of the LED under calibration were

measured with a fibre optic input spectrometer. The spectrometer wavelength scale was

calibrated against low pressure spectral Hg, Cd and He lamps traceable to the INM

reference for length measurements (stabilised He-Ne laser). For all wavelengths within

the visible range it was found to be accurate within ±0.3 nm.

The spectrometer irradiance scale was calibrated against an irradiance spectral

density lamp, traceable to the MIKES–TKK reference. The spectrometer photometric

linearity was calibrated and further checked against a set of spectral transmittance filters

(neutral glass of NG type), traceable to the INM reference spectrophotometer.

The length measurements (standard lamp-lumen-meter aperture plane, the

diameter of the lumen-meter sphere aperture) are traceable to the INM-RO national

reference for length (stabilised He-Ne laser).

All voltage measurements were traceable to the national references of Romania

(group of stabilised Zener diodes of Fluke 732 B). The shunt resistance used to generate

the feeding current was calibrated with traceability to the national references (group of

electrical resistors).


93

The temperature was measured with a digital thermometer calibrated with

traceability to the INM maintained SIT90 fixed points.


The total luminous flux )( x expression is:

)1()(2

,

321 lmAd

IRCCCC

ev

xspvx

where: 1C is the lamp-LED illumination non-equivalence factor;

2C is the LED feeding

current factor; 3C is the correction factor for the ambient temperature;

ve

xx

Y

YR with

xY

the output generated by the LED emitted flux and veY the output generated by the

luminous intensity standard lamp; evI ,is the value of the luminous intensity standard

lamp; A is the area of the integrating sphere aperture (1256,6 mm2); d is the standard

lamp-lumen-meter sphere aperture distance;

spC is the spectral correction factor:

)2(2

1

2

1

2

1

2

1

,,,

,,,

VSsS

VSsS

C

erphrxr

xrphrer

sp

where: )(, erS is the relative spectral density of the luminous intensity standard lamp;

)(, xrS is the relative spectral density of the LED under calibration; )(, phrs is the

relative spectral responsivity of the lumen-meter; 21, are the extreme wavelengths of

the visible spectrum; )(V is the relative responsivity of the CIE standard observer.

Tables in the following are the detailed uncertainty budgets of the total luminous

flux measurement for the LEDs used in this APMP LED comparison.

Table 4-63. INM uncertainty budget of total luminous flux measurement for red LEDs (R).


uncertainty

Ty

pe

Probability

distribution

Sensitivity

coefficient

Contrib

ution

(%)

Deg. of

freedo

m

Correl

ated?

Lamp-LED illumination

non-equivalence factor 1C

0.060 B rectangular vx 6.0 ∞ O

LED feeding current factor

2C

0.001 B normal vx 0.1 ∞ X


94

correction factor for the

ambient temperature 3C


Spectral correction factor

spC

0.05 spC B normal spvx C/ 5.0 ∞ X

Output ratio xR 0.010 xR B normal

xvx R/ 1.0 ∞ O

Value of the luminous

intensity standard lamp

evI ,

0.010evI , B normal

vevx I/ 1.0 ∞ O

Integrating sphere aperture

area A

0.005 A B normal Avx / 0.5 ∞ O

Repeatability 0.010vx A normal 1 1.0 ∞ X

Combined standard

uncertainty (%)

-- -- normal -- 6.4 ∞ --

Table 4-64. INM uncertainty budget of total luminous flux measurement for green LEDs (G).


uncertainty

Ty

pe

Probability

distribution

Sensitivity

coefficient

Contrib

ution

(%)

Deg. of

freedo

m

Correl

ated?





2C






spC



xvx R/ 1.0 ∞ O



evI ,

0.010evI , B normal

vevx I/ 1.0 ∞ O


area A


Repeatability 0.010 vx A normal 1.0 1.0 ∞ X

Combined standard

uncertainty (%)

-- -- normal -- 6.0 ∞ --

Table 4-65. INM uncertainty budget of total luminous flux measurement for blue LEDs (B).


uncertainty

Ty

pe

Probability

distribution

Sensitivity

coefficient

Contrib

ution

(%)

Deg. of

freedo

m

Correl

ated?


95





2C






spC



xvx R/ 1.0 ∞ O



evI ,

0.010evI , B normal

vevx I/ 1.0 ∞ O


area A



Combined standard

uncertainty (%)

-- -- normal -- 6.4 ∞ --

Table 4-66. INM uncertainty budget of total luminous flux measurement for white LEDs (W).


uncertainty

Ty

pe

Probability

distribution

Sensitivity

coefficient

Contrib

ution

(%)

Deg. of

freedo

m

Correl

ated?





2C






spC



xvx R/ 1.0 ∞ O



evI ,

0.010evI , B normal

vevx I/ 1.0 ∞ O


area A



Combined standard

uncertainty (%)

-- -- normal -- 6.6 ∞ --


96

The junction voltage expression is:

readj VCCV 21

readV : the mean reading ; 1C : temperature factor and 2C : position factor

Table 4-67 is the detailed uncertainty budget of the junction voltage

measurement.

Table 4-67. INM uncertainty budget of junction voltage measurement.


uncertainty T

yp

e

Probability

distribution

Sensitivity

coefficient

Contribut

ion (%)

Deg.

of

freedo

m

Correl

ated?

Mean reading readV 2E-5 V B normal 1 0.002 ∞ O

Temperature factor 1C 0.0010 B rectangular readV

0.10 ∞ X

Position factor 2C 0.0005 B rectangular readV

0.05 ∞ X

Repeatability 0.0005 jV

A normal 1 0.05 ∞ X

Combined standard

uncertainty (%

-- -- normal -- 0.12 ∞ --


97

5. Reported Results of Participants

In this chapter, the results of the comparison S3b are presented, which are reported by

each participant as the final version, i.e., after the verification in the pre-draft A process.

We note that, throughout this report document, the uncertainty values with a symbol U

indicate the expanded uncertainties for a confidence level of 95 % normally with a

coverage factor of k = 2, while the values with a symbol u indicate the standard

uncertainties.

5.1. KRISS

As the pilot laboratory of the comparison, KRISS measured each LED at most three times:

the first measurement before sending the LEDs for the first round, the second after

receiving the LEDs from the first round, and the third after receiving the LEDs from the

second round. The final control measurement of the first round is also regarded as the

initial control measurement of the second round. Note that the artefact sets #2, #4, #6,

and #8 are circulated only one round. The repeated measurements provide information

on the stability of the artefact LEDs, which will be discussed in Section 6.2.

Table 5-1 sumarizes the measurement results of KRISS of all the artefact LEDs. The

uncertainty values are not explicitly shown in this table but refered to the budgets in

Table 4-1 ~ Table 4-4. The laboratory conditions are kept at a temperature of (25 ± 2)

ºC and a relative humidity of (45 ± 15) %. The burning time of each measurement was

20 minutes in average.

Table 5-1. Measurement results of KRISS.

artifact

set LED

1. measurement 2. measurement 3. measurement

Φ (lm) Vj (V) Φ (lm) Vj (V) Φ (lm) Vj (V)

#1

R-1 0.6757 1.8849 0.6730 1.8848 0.6710 1.8826

R-2 0.6781 1.8888 0.6755 1.8888 0.6750 1.8866

R-3 0.6506 1.9211 0.6481 1.9211 0.6493 1.9191

G-1 3.0107 3.2912 2.9976 3.2911 2.9756 3.3190

G-2 2.8639 3.4307 2.8467 3.4300 2.8258 3.4543

G-3 2.9543 3.3098 2.9450 3.3122 2.9262 3.3381

B-1 0.7512 3.3723 0.7488 3.3731 0.7340 3.3994

B-2 0.7648 3.3744 0.7608 3.3751 0.7389 3.3991

B-3 0.7974 3.3412 0.7967 3.3435 0.7842 3.3671

W-1 1.5951 3.4358 1.6992 3.4371 1.6839 3.4561

W-2 1.5890 3.4568 1.5749 3.4574 1.5525 3.4788

W-3 1.7533 3.4123 1.7413 3.4124 1.7087 3.4320


98

#2

R-1 0.6367 1.8886 0.6399 1.8932

R-2 0.6945 1.8981 0.6950 1.9024

R-3 0.7023 1.9020 0.7047 1.9064

G-1 2.7609 3.4710 2.8661 3.4689

G-2 2.8444 3.3910 2.8291 3.3889

G-3 2.9915 3.3017 2.9757 3.2996

B-1 0.7438 3.4519 0.7321 3.4517

B-2 0.8121 3.3697 0.8045 3.3677

B-3 0.7318 3.3510 0.7259 3.3509

W-1 1.7008 3.3014 1.7061 3.2989

W-2 1.7136 3.4204 1.6983 3.4167

W-3 1.5545 3.4492 1.5425 3.4460

#3

R-1 0.6749 1.8900 0.6675 1.8873 0.6724 1.8906

R-2 0.6887 1.8931 0.6811 1.8906 0.6856 1.8934

R-3 0.6864 1.8975 0.6796 1.8948 0.6845 1.8986

G-1 2.9676 3.5036 2.9421 3.4951 2.9298 3.5025

G-2 2.7133 3.3718 2.6977 3.3652 2.6758 3.3732

G-3 2.6582 3.3323 2.6377 3.3262 2.6189 3.3322

B-1 0.7804 3.4291 0.7744 3.4237 0.7597 3.4286

B-2 0.8262 3.4177 0.8196 3.4108 0.7993 3.4141

B-3 0.6624 3.5096 0.6599 3.5033 0.6454 3.5102

W-1 1.7035 3.4353 1.6824 3.4262 1.6709 3.4321

W-2 1.6709 3.3377 1.6534 3.3291 1.6334 3.3348

W-3 1.7216 3.3060 1.7029 3.2992 1.6805 3.3024

#4

R-1 0.7098 1.8982 0.7029 1.8957

R-2 0.6725 1.8946 0.6654 1.8923

R-3 0.6933 1.8961 0.6876 1.8938

G-1 2.9251 3.5108 2.8934 3.5034

G-2 3.1816 3.2985 3.1474 3.2946

G-3 2.9647 3.3586 2.9395 3.3527

B-1 0.8849 3.4215 0.8801 3.4165

B-2 0.7567 3.4645 0.7523 3.4590

B-3 0.8270 3.4018 0.8229 3.3965

W-1 1.7656 3.4352 1.7270 3.4287

W-2 1.7379 3.3397 1.7039 3.3341

W-3 1.7825 3.4446 1.7491 3.4388

#5

R-1 0.6827 1.9146 0.6868 1.9182 0.6896 1.9194

R-2 0.6829 1.9187 0.6881 1.9226 0.6885 1.9230

R-3 0.6495 1.8824 0.6542 1.8857 0.6537 1.8857

G-1 2.9119 3.2991 2.9171 3.3075 2.9012 3.3151

G-2 2.8201 3.4336 2.7947 3.4434 2.7550 3.4530

G-3 2.8484 3.3686 2.8501 3.3766 2.8421 3.3862

B-1 0.7790 3.4020 0.7768 3.4097 0.7672 3.4168

B-2 0.8820 3.4045 0.8803 3.4130 0.8714 3.4204

B-3 0.8248 3.4180 0.8219 3.4268 0.8128 3.4352

W-1 1.6785 3.3057 1.6842 3.3123 1.6805 3.3206

W-2 1.7536 3.4314 1.7598 3.4387 1.7476 3.4460

W-3 1.7000 3.4379 1.7101 3.4475 1.6980 3.4544

#6 R-1 0.6992 1.9041 0.6927 1.9005

R-2 0.6610 1.8912 0.6530 1.8870


99

R-3 0.6906 1.9016 0.6818 1.8968

G-1 3.0411 3.3048 3.0086 3.2964

G-2 2.9213 3.3042 2.9000 3.2986

G-3 2.8715 3.3227 2.8551 3.3188

B-1 0.8924 3.4183 0.8908 3.4158

B-2 0.7632 3.3799 0.7599 3.3756

B-3 0.7932 3.3856 0.7904 3.3812

W-1 1.7634 3.4041 1.7047 3.4018

W-2 1.7206 3.4038 1.6762 3.4014

W-3 1.7303 3.4208 1.6969 3.4169

#7

R-1 0.6559 1.9181 0.6570 1.9189 0.6619 1.9222

R-2 0.7196 1.9003 0.7234 1.9016 0.7254 1.9039

R-3 0.6466 1.9170 0.6483 1.9178 0.6527 1.9212

G-1 3.0373 3.2876 3.0222 3.2896 3.0194 3.2945

G-2 2.8805 3.3519 2.8689 3.3541 2.8643 3.3587

G-3 3.0100 3.2931 2.9972 3.2953 3.0015 3.2991

B-1 0.8098 3.4509 0.8039 3.4523 0.7630 3.4596

B-2 0.7816 3.3859 0.7763 3.3896 0.7567 3.3942

B-3 0.7966 3.4154 0.7914 3.4159 0.7792 3.4243

W-1 1.6594 3.4605 1.6525 3.4638 1.6534 3.4698

W-2 1.5769 3.3495 1.5720 3.3502 1.5728 3.3579

W-3 1.5905 3.4025 1.5882 3.4036 1.5817 3.4085

#8

R-1 0.6490 1.8866 0.6524 1.8876

R-2 0.6793 1.8904 0.6833 1.8918

R-3 0.7060 1.8953 0.7098 1.8968

G-1 2.8770 3.5206 2.8631 3.5245

G-2 3.0247 3.2848 3.0101 3.2875

G-3 2.8590 3.2859 2.8468 3.2899

B-1 0.8507 3.4371 0.8449 3.4422

B-2 0.7713 3.3549 0.7665 3.3588

B-3 0.7712 3.4550 0.7641 3.4587

W-1 1.6656 3.4149 1.6645 3.4198

W-2 1.3472 3.4155 1.3425 3.4197

W-3 1.5594 3.4507 1.5509 3.4528

5.2. MIKES

MIKES of Finland measured the artifact set #1 in its first round from 07 April 2008 to 14

April 2008. The laboratory conditions are reported as temperature of (21.5 ± 1.0) ºC and

relative humidity of (31 ± 5) %. Table 5-2 shows the reported results of MIKES.

Table 5-2. Measurement results of MIKES.

artifact

set LED Φ (lm) U(Φ) (lm) Vj (V) U(Vj) (V)

burning

time (min)

#1

R-1 0.717 0.019 1.88996 0.00009 90

R-2 0.721 0.019 1.89352 0.00008 30

R-3 0.691 0.018 1.92577 0.00008 30

G-1 3.080 0.077 3.30151 0.00025 90


100

G-2 2.922 0.073 3.44605 0.00026 30

G-3 3.028 0.076 3.31910 0.00026 30

B-1 0.765 0.043 3.38394 0.00024 150

B-2 0.775 0.043 3.38170 0.00023 30

B-3 0.811 0.045 3.35990 0.00022 30

W-1 1.771 0.067 3.44276 0.00042 160

W-2 1.659 0.063 3.46634 0.00042 90

W-3 1.822 0.069 3.41695 0.00042 70

5.3. CMS-ITRI

CMS-ITRI of Chinese Taipei measured the artifact set #2 in its first round from 6 May

2008 to 8 May 2008. The laboratory conditions are reported as temperature of (23.0 ±

1.5) ºC and relative humidity of (45 ± 10) %. During the measurement at CMS-ITRI,

however, all the three red LEDs were damaged so that the red LEDs of the set #2 had to

be completely replaced for the second round. On the agreement of the other

participants, CMS-ITRI repeated the measurement of the new red LEDs of the set #2 in

Sept. ~ Oct. 2009. Table 5-3 shows the reported results of CMS-ITRI.

Table 5-3. Measurement results of CMS-ITRI.

artifact


burning

time (min)

#2

R-1 0.638 0.015 1.896 0.002 35

R-2 0.694 0.016 1.905 0.002 35

R-3 0.703 0.017 1.909 0.001 35

G-1 2.777 0.067 3.516 0.005 35

G-2 2.868 0.069 3.430 0.004 35

G-3 3.008 0.073 3.337 0.005 35

B-1 0.723 0.017 3.456 0.004 35

B-2 0.797 0.019 3.411 0.003 35

B-3 0.711 0.017 3.420 0.003 35

W-1 1.729 0.040 3.341 0.003 35

W-2 1.743 0.041 3.463 0.003 35

W-3 1.584 0.037 3.495 0.011 35

5.4. PTB

PTB of Germany measured the artifact set #3 in its first round from 16 June to 2 July

2008. The laboratory conditions are reported as temperature of (25.0 ± 0.7) ºC and

relative humidity of (50 ± 10) %. Table 5-4 shows the reported results of PTB.

Table 5-4. Measurement results of PTB.

artifact LED Φ (lm) U(Φ) (lm) Vj (V) U(Vj) (V) burning


101

set time (min)

#3

R-1 0.6652 0.0183 1.8945 0.0012 590

R-2 0.6758 0.0186 1.8976 0.0012 446

R-3 0.6757 0.0186 1.9024 0.0012 426

G-1 2.8550 0.0608 3.5230 0.0026 560

G-2 2.6163 0.0557 3.3894 0.0025 278

G-3 2.5650 0.0546 3.3503 0.0025 399

B-1 0.7750 0.0206 3.4455 0.0017 574

B-2 0.8211 0.0218 3.4338 0.0017 213

B-3 0.6630 0.0176 3.5314 0.0017 410

W-1 1.6826 0.0372 3.4509 0.0025 495

W-2 1.6511 0.0365 3.3525 0.0025 495

W-3 1.6996 0.0376 3.3217 0.0024 395

5.5. NMIJ

NMIJ of Japan measured the artifact set #4 in its first round from 15 April 2008 to 22

June 2008. The laboratory conditions are reported as temperature of (23 ± 2) ºC and

relative humidity of (50 ± 30) %. Table 5-5 shows the reported results of NMIJ.

Table 5-5. Measurement results of NMIJ.

artifact


burning

time (min)

#4

R-1 0.701 0.017 1.8995 0.0007 527

R-2 0.663 0.016 1.8957 0.0007 494

R-3 0.685 0.016 1.8973 0.0006 499

G-1 2.9530 0.0508 3.5223 0.0065 374

G-2 3.2061 0.0551 3.3075 0.0053 371

G-3 2.9883 0.0514 3.3682 0.0065 378

B-1 0.9373 0.0176 3.4340 0.0056 370

B-2 0.8020 0.0151 3.4724 0.0064 382

B-3 0.8746 0.0164 3.4087 0.0048 378

W-1 1.7788 0.0254 3.4469 0.0052 655

W-2 1.7441 0.0249 3.3505 0.0044 457

W-3 1.7897 0.0256 3.4562 0.0029 374

5.6. CENAM

CENAM of Mexico measured the artifact set #5 in its first round from 17 July 2008 to 21

July 2008. The laboratory conditions are reported as temperature of (22.7 ± 2.2) ºC and

relative humidity of (47.5 ± 8.0) %. Table 5-6 shows the reported results of CENAM.

Table 5-6. Measurement results of CENAM.

artifact


burning

time (min)


102

#5

R-1 0.5350 0.0534 1.9228 0.0007 26

R-2 0.5367 0.0582 1.9271 0.0007 19

R-3 0.5245 0.0593 1.8898 0.0044 21

G-1 3.3739 0.2837 3.3208 0.0008 22

G-2 3.2135 0.2864 3.4362 0.0010 23

G-3 3.4003 0.2985 3.3913 0.0009 19

B-1 0.9105 0.0951 3.4235 0.0009 21

B-2 0.9486 0.1179 3.4264 0.0009 18

B-3 0.8750 0.0889 3.4403 0.0015 20

W-1 1.7994 0.1850 3.3253 0.0012 25

W-2 1.8595 0.1760 3.4553 0.0010 22

W-3 1.8181 0.1670 3.4633 0.0015 19

5.7. LNE

LNE of France measured the artifact set #6 in its first round from 15 June 2008 to 13 July

2008. The laboratory conditions are reported as temperature of (22 ± 2) ºC and relative

humidity of (50 ± 10) %. Table 5-7 shows the reported results of LNE.

Table 5-7. Measurement results of LNE.

artifact


burning

time (min)

#6

R-1 0.687 0.014 1.90903 0.00057 720

R-2 0.651 0.013 1.89522 0.00057 1080

R-3 0.680 0.014 1.90532 0.00057 720

G-1 2.972 0.053 3.3120 0.0013 900

G-2 2.870 0.052 3.3137 0.0013 1200

G-3 2.814 0.051 3.3359 0.0013 1230

B-1 0.882 0.021 3.4343 0.0014 1440

B-2 0.755 0.018 3.3927 0.0014 720

B-3 0.785 0.019 3.3988 0.0014 1440

W-1 1.717 0.024 3.4182 0.0017 1410

W-2 1.689 0.024 3.4171 0.0017 2550

W-3 1.701 0.024 3.4346 0.0017 2130

5.8. METAS

METAS of Switzerland measured the artifact set #7 in its first round from 30 Sept. 2008

to 8 Oct. 2008. The laboratory conditions are reported as temperature of (25.0 ± 0.5) ºC

and relative humidity of (45 ± 5) %. Table 5-8 shows the reported results of METAS.

Table 5-8. Measurement results of METAS.

artifact


burning

time (min)

#7 R-1 0.5801 0.0114 1.9296 0.0082 290


103

R-2 0.6445 0.0127 1.9130 0.0082 101

R-3 0.5751 0.0115 1.9297 0.0082 103

G-1 2.5280 0.0519 3.2997 0.063 119

G-2 2.4159 0.0497 3.3643 0.063 82

G-3 2.5336 0.0510 3.3066 0.063 162

B-1 0.7186 0.0252 3.4650 0.075 135

B-2 0.7096 0.0246 3.3995 0.075 96

B-3 0.7079 0.0252 3.4310 0.075 104

W-1 1.3201 0.0237 3.4748 0.083 183

W-2 1.2496 0.0222 3.3635 0.083 88

W-3 1.2603 0.0226 3.4162 0.083 100

5.9. NMC-A*STAR

NMC-A*STAR of Singapore measured the artifact set #8 in its first round from 10 July

2008 to 28 August 2008. The laboratory conditions are reported as temperature of (23 ±

2) ºC and relative humidity of (60 ± 10) %. Table 5-9 shows the reported results of

NMC-A*STAR.

Table 5-9. Measurement results of NMC-A*STAR.

artifact


burning

time (min)

#8

R-1 0.661 0.012 1.8918 0.0016 51

R-2 0.692 0.012 1.8965 0.0016 39

R-3 0.719 0.013 1.9022 0.0016 62

G-1 2.832 0.045 3.5329 0.0033 50

G-2 2.987 0.048 3.2946 0.0033 38

G-3 2.834 0.045 3.2958 0.0033 54

B-1 0.872 0.015 3.4500 0.0026 48

B-2 0.795 0.014 3.3650 0.0026 38

B-3 0.800 0.014 3.4684 0.0026 53

W-1 1.670 0.027 3.4273 0.0078 49

W-2 1.347 0.022 3.4273 0.0078 37

W-3 1.564 0.025 3.4617 0.0078 53

5.10. VSL

VSL of the Netherlands measured the artifact set #1 in its second round from 13 October

2008 to 12 January 2009. The laboratory conditions are reported as temperature of (24.0

± 0.5) ºC and relative humidity of (45 ± 10) %. Table 5-10 shows the reported results of

VSL.

Table 5-10. Measurement results of VSL.

artifact LED Φ (lm) U(Φ) (lm) Vj (V) U(Vj) (V) burning


104

set time (min)

#1

R-1 0.683 0.020 1.8870 0.0031 231

R-2 0.676 0.021 1.8913 0.0031 215

R-3 0.652 0.019 1.9239 0.0031 333

G-1 2.978 0.077 3.2986 0.0056 705

G-2 2.872 0.069 3.4396 0.0059 925

G-3 2.855 0.081 3.3170 0.0054 715

B-1 0.769 0.020 3.3805 0.0055 292

B-2 0.790 0.021 3.3806 0.0058 285

B-3 0.809 0.023 3.3491 0.0055 620

W-1 1.737 0.055 3.4444 0.0057 270

W-2 1.575 0.058 3.4646 0.0059 267

W-3 1.767 0.070 3.4193 0.0059 352

5.11. NIST

NIST of the USA measured the artifact set #3 in its second round on 23 April 2009. The

laboratory conditions are reported as temperature of 25 ºC and relative humidity of

17 %. Table 5-11 shows the reported results of NIST.

Table 5-11. Measurement results of NIST.

artifact


burning

time (min)

#3

R-1 0.669 0.010 1.887 0.005 40

R-2 0.683 0.010 1.891 0.005 25

R-3 0.682 0.010 1.895 0.005 25

G-1 2.919 0.022 3.494 0.008 50

G-2 2.658 0.020 3.364 0.008 25

G-3 2.604 0.020 3.323 0.008 40

B-1 0.789 0.019 3.421 0.008 40

B-2 0.831 0.020 3.409 0.008 25

B-3 0.672 0.016 3.500 0.009 25

W-1 1.668 0.012 3.423 0.008 40

W-2 1.629 0.012 3.326 0.008 25

W-3 1.680 0.013 3.295 0.008 25

5.12. VNIIOFI

VNIIOFI of Russia measured the artifact set #5 in its second round from 12 January 2009

to 20 January 2009. The laboratory conditions are reported as temperature of (22.0 ±

0.5) ºC and relative humidity of (62 ± 2) %. Table 5-12 shows the reported results of

VNIIOFI. We note that VNIIOFI reported two sets of results: the one based on the

goniophotometer method, and the other based on the integrating sphere method. We

use the integrating sphere results for the comparison, which had slightly lower


105

uncertainties.

Table 5-12. Measurement results of VNIIOFI.

artifact


burning

time (min)

#5

R-1 0.736 0.013 1.932 0.001 37

R-2 0.732 0.015 1.936 0.001 39

R-3 0.707 0.013 1.897 0.001 36

G-1 2.851 0.06 3.339 0.001 40

G-2 2.712 0.04 3.483 0.001 35

G-3 2.792 0.05 3.411 0.001 37

B-1 0.738 0.016 3.441 0.001 36

B-2 0.839 0.016 3.443 0.001 36

B-3 0.816 0.015 3.458 0.001 37

W-1 1.715 0.04 3.345 0.001 36

W-2 1.789 0.04 3.476 0.001 39

W-3 1.745 0.04 3.487 0.001 37

5.13. INM

INM of Romania measured the artifact set #7 in its second round from 13 December

2008 to 16 December 2008. The laboratory conditions are reported as temperature of

(25.0 ± 0.2) ºC and relative humidity of (30 ± 5) %. Table 5-13 shows the reported

results of INM.

Table 5-13. Measurement results of INM.

artifact


burning

time (min)

#7

R-1 0.63 0.04 1.926 0.006 5

R-2 0.69 0.04 1.906 0.006 5

R-3 0.61 0.04 1.924 0.006 5

G-1 2.71 0.17 3.300 0.010 5

G-2 2.69 0.17 3.367 0.010 5

G-3 2.76 0.17 3.306 0.010 5

B-1 0.70 0.10 3.467 0.010 5

B-2 0.64 0.10 3.400 0.010 5

B-3 0.66 0.10 3.433 0.010 5

W-1 1.50 0.20 3.477 0.010 5

W-2 1.34 0.20 3.367 0.010 5

W-3 1.35 0.20 3.418 0.010 5

6. Pre-draft A Process

After the measurement process is completed, the preparation of the comparison report is


106

conducted according to the CCPR Guidelines. 14 The pre-draft A process consists of

verification of reported results, review of uncertainty budgets, and review of relative data.

In this chapter, we also describe the temperature-corrected results and the identification

of outliers.

6.1. Verification of Reported Results

The verification of reported results started in November 2009 after most of the

participants have submitted their results. The pilot sent to each participant the submitted

result values and the technical report including the uncertainty budgets. The participant

reviewed it to correct any error. After the participant confirmed the final version, no

correction is applied in the results and in the technical reports of the participants.

6.2. Temperature Correction and Artifact Drift

After the results are finalized by the verification, the pilot applied the temperature

correction based on the Eq. (3-1). By using the temperature sensitivity coefficients a, b,

and c of each LED and the measured junction voltages reported by the participants, all

the results could be converted to the values expected at the same junction voltage, i.e.,

at the same reference condition with a temperature of T0. We took the initial control

measurement of the pilot for each round as the reference condition for correction.

The tables below summarize the results before and after the temperature correction

for each measurement round. The relative differences of the participant’s results and of

the pilot’s results by the temperature correction are also calculated in the last two

columns to show the magnitudes of the correction. Note that the uncertainty of the

temperature correction was estimated to be 0.5 % as a relative standard uncertainty (see

Chapter 3), while all the participants claimed the uncertainty of the junction voltage

measurement much lower than this.

Table 6-1. Results of temperature correction for the round to MIKES.

artifact

set LED

1. meas.

of pilot

participant

lab

2. meas.

of pilot temperature corrected relative

difference

ΦL* - ΦL

relative

difference

ΦP2* – ΦP2 ΦP1 (lm) ΦL (lm) ΦP2 (lm) ΦL

* (lm) ΦP2

* (lm)

#1 R-1 0.6757 0.717 0.6730 0.7084 0.6732 -1.21% 0.02%

14 CCPR Key Comparison Working Group, Guidelines for CCPR Comparison Report Preparation, Rev. 2 (Sept. 18, 2009), available at http://www.bipm.org/en/committees/cc/ccpr/publications_cc.html

http://www.bipm.org/en/committees/cc/ccpr/publications_cc.html


107

R-2 0.6781 0.721 0.6755 0.7130 0.6756 -1.12% 0.00%

R-3 0.6506 0.691 0.6481 0.6845 0.6482 -0.95% 0.01%

G-1 3.0107 3.080 2.9976 3.0646 2.9977 -0.50% 0.00%

G-2 2.8639 2.922 2.8467 2.9049 2.8475 -0.59% 0.03%

G-3 2.9543 3.028 2.9450 3.0151 2.9417 -0.43% -0.11%

B-1 0.7512 0.765 0.7488 0.7661 0.7488 0.14% 0.01%

B-2 0.7648 0.775 0.7608 0.7755 0.7608 0.06% 0.00%

B-3 0.7974 0.811 0.7967 0.8094 0.7964 -0.20% -0.04%

W-1 1.5951 1.771 1.6992 1.7636 1.6979 -0.42% -0.08%

W-2 1.5890 1.659 1.5749 1.6500 1.5744 -0.55% -0.03%

W-3 1.7533 1.822 1.7413 1.8172 1.7412 -0.27% 0.00%

Table 6-2. Results of temperature correction for the round to CMS-ITRI.

artifact

set LED

1. meas.

of pilot

participant

lab

2. meas.


difference

ΦL* - ΦL

relative

difference


* (lm) ΦP2

* (lm)

#2

R-1 0.6367 0.638 0.6399 0.6256 0.6319 -1.99% -1.26%

R-2 0.6945 0.694 0.6950 0.6820 0.6875 -1.76% -1.09%

R-3 0.7023 0.703 0.7047 0.6904 0.6966 -1.83% -1.16%

G-1 2.7609 2.777 2.8661 2.7317 2.8688 -1.66% 0.09%

G-2 2.8444 2.868 2.8291 2.8224 2.8321 -1.62% 0.11%

G-3 2.9915 3.008 2.9757 2.9518 2.9798 -1.90% 0.14%

B-1 0.7438 0.723 0.7321 0.7226 0.7322 -0.05% 0.00%

B-2 0.8121 0.797 0.8045 0.7917 0.8052 -0.67% 0.08%

B-3 0.7318 0.711 0.7259 0.7173 0.7259 0.87% 0.00%

W-1 1.7008 1.729 1.7061 1.6800 1.7097 -2.92% 0.21%

W-2 1.7136 1.743 1.6983 1.6928 1.7033 -2.97% 0.29%

W-3 1.5545 1.584 1.5425 1.5312 1.5468 -3.45% 0.28%

Table 6-3. Results of temperature correction for the round to PTB.

artifact

set LED

1. meas.

of pilot

participant

lab

2. meas.


difference

ΦL* - ΦL

relative

difference


* (lm) ΦP2

* (lm)

#3

R-1 0.6749 0.6652 0.6675 0.6568 0.6730 -1.27% 0.82%

R-2 0.6887 0.6758 0.6811 0.6675 0.6862 -1.25% 0.73%

R-3 0.6864 0.6757 0.6796 0.6660 0.6849 -1.45% 0.77%

G-1 2.9676 2.8550 2.9421 2.8327 2.9521 -0.79% 0.34%

G-2 2.7133 2.6163 2.6977 2.5953 2.7053 -0.81% 0.28%

G-3 2.6582 2.5650 2.6377 2.5390 2.6461 -1.03% 0.32%

B-1 0.7804 0.7750 0.7744 0.7741 0.7749 -0.12% 0.05%

B-2 0.8262 0.8211 0.8196 0.8198 0.8204 -0.15% 0.09%

B-3 0.6624 0.6630 0.6599 0.6610 0.6605 -0.31% 0.09%

W-1 1.7035 1.6826 1.6824 1.6624 1.6951 -1.21% 0.75%

W-2 1.6709 1.6511 1.6534 1.6324 1.6644 -1.15% 0.66%


108

W-3 1.7216 1.6996 1.7029 1.6788 1.7124 -1.24% 0.55%

Table 6-4. Results of temperature correction for the round to NMIJ.

artifact

set LED

1. meas.

of pilot

participant

lab

2. meas.


difference

ΦL* - ΦL

relative

difference


* (lm) ΦP2

* (lm)

#4

R-1 0.7098 0.701 0.7229 0.6987 0.7077 -0.33% 0.68%

R-2 0.6725 0.663 0.6746 0.6610 0.6698 -0.30% 0.65%

R-3 0.6933 0.685 0.6979 0.6827 0.6923 -0.34% 0.68%

G-1 2.9251 2.9530 2.6238 2.9413 2.9008 -0.40% 0.26%

G-2 3.1816 3.2061 2.7930 3.1902 3.1540 -0.50% 0.21%

G-3 2.9647 2.9883 2.6636 2.9744 2.9479 -0.47% 0.29%

B-1 0.8849 0.9373 0.9182 0.9378 0.8799 0.05% -0.02%

B-2 0.7567 0.8020 0.7942 0.8024 0.7521 0.05% -0.03%

B-3 0.8270 0.8746 0.8769 0.8753 0.8223 0.08% -0.07%

W-1 1.7656 1.7788 0.6897 1.7640 1.7352 -0.84% 0.47%

W-2 1.7379 1.7441 0.6846 1.7300 1.7111 -0.81% 0.42%

W-3 1.7825 1.7897 0.6991 1.7759 1.7561 -0.77% 0.40%

Table 6-5. Results of temperature correction for the round to CENAM.

artifact

set LED

1. meas.

of pilot

participant

lab

2. meas.


difference

ΦL* - ΦL

relative

difference


* (lm) ΦP2

* (lm)

#5

R-1 0.6827 0.5350 0.6868 0.5251 0.6811 -1.89% -0.83%

R-2 0.6829 0.5367 0.6881 0.5267 0.6821 -1.91% -0.88%

R-3 0.6495 0.5245 0.6542 0.5142 0.6484 -2.01% -0.91%

G-1 2.9119 3.3739 2.9171 3.3360 2.9031 -1.14% -0.48%

G-2 2.8201 3.2135 2.7947 3.2099 2.7832 -0.11% -0.41%

G-3 2.8484 3.4003 2.8501 3.3646 2.8383 -1.06% -0.42%

B-1 0.7790 0.9105 0.7768 0.9101 0.7761 -0.04% -0.09%

B-2 0.8820 0.9486 0.8803 0.9461 0.8785 -0.26% -0.21%

B-3 0.8248 0.8750 0.8219 0.8738 0.8207 -0.14% -0.15%

W-1 1.6785 1.7994 1.6842 1.7734 1.6759 -1.46% -0.50%

W-2 1.7536 1.8595 1.7598 1.8299 1.7510 -1.62% -0.51%

W-3 1.7000 1.8181 1.7101 1.7862 1.6984 -1.79% -0.69%

Table 6-6. Results of temperature correction for the round to LNE.

artifact

set LED

1. meas.

of pilot

participant

lab

2. meas.


difference

ΦL* - ΦL

relative

difference


* (lm) ΦP2

* (lm)

#6 R-1 0.7000 0.745 0.7136 0.7257 0.7076 -2.66% -0.84%


109

R-2 0.6563 0.692 0.6635 0.6740 0.6575 -2.67% -0.91%

R-3 0.7023 0.750 0.7144 0.7298 0.7082 -2.77% -0.88%

G-1 2.8398 2.985 2.8575 2.9632 2.8493 -0.74% -0.29%

G-2 2.7226 2.861 2.7450 2.8443 2.7372 -0.59% -0.29%

G-3 2.4871 2.603 2.4902 2.5850 2.4814 -0.70% -0.35%

B-1 0.9185 0.934 0.9231 0.9337 0.9224 -0.03% -0.08%

B-2 0.8098 0.816 0.8125 0.8158 0.8116 -0.03% -0.11%

B-3 0.8244 0.825 0.8236 0.8251 0.8228 0.01% -0.09%

W-1 0.6933 0.709 0.6739 0.7026 0.6716 -0.91% -0.34%

W-2 0.6828 0.702 0.6709 0.6941 0.6680 -1.13% -0.43%

W-3 0.7091 0.731 0.7016 0.7236 0.6986 -1.02% -0.44%

Table 6-7. Results of temperature correction for the round to METAS.

artifact

set LED

1. meas.

of pilot

participant

lab

2. meas.


difference

ΦL* - ΦL

relative

difference


* (lm) ΦP2

* (lm)

#7

R-1 0.6559 0.5801 0.6570 0.5643 0.6556 -2.79% -0.21%

R-2 0.7196 0.6445 0.7234 0.6243 0.7208 -3.24% -0.36%

R-3 0.6466 0.5751 0.6483 0.5571 0.6469 -3.24% -0.22%

G-1 3.0373 2.5280 3.0222 2.5112 3.0185 -0.67% -0.12%

G-2 2.8805 2.4159 2.8689 2.4020 2.8657 -0.58% -0.11%

G-3 3.0100 2.5336 2.9972 2.5135 2.9930 -0.80% -0.14%

B-1 0.8098 0.7186 0.8039 0.7182 0.8038 -0.06% -0.01%

B-2 0.7816 0.7096 0.7763 0.7091 0.7760 -0.06% -0.05%

B-3 0.7966 0.7079 0.7914 0.7058 0.7913 -0.29% -0.01%

W-1 1.6594 1.3201 1.6525 1.3047 1.6478 -1.18% -0.29%

W-2 1.5769 1.2496 1.5720 1.2348 1.5710 -1.19% -0.06%

W-3 1.5905 1.2603 1.5882 1.2459 1.5866 -1.16% -0.10%

Table 6-8. Results of temperature correction for the round to NMC-A*STAR.

artifact

set LED

1. meas.

of pilot

participant

lab

2. meas.


difference

ΦL* - ΦL

relative

difference


* (lm) ΦP2

* (lm)

#8

R-1 0.6490 0.661 0.6524 0.6510 0.6506 -1.53% -0.28%

R-2 0.6793 0.692 0.6833 0.6792 0.6803 -1.88% -0.43%

R-3 0.7060 0.719 0.7098 0.7050 0.7066 -1.99% -0.45%

G-1 2.8770 2.832 2.8631 2.8190 2.8589 -0.46% -0.15%

G-2 3.0247 2.987 3.0101 2.9670 3.0040 -0.67% -0.20%

G-3 2.8590 2.834 2.8468 2.8147 2.8385 -0.69% -0.29%

B-1 0.8507 0.872 0.8449 0.8701 0.8439 -0.22% -0.12%

B-2 0.7713 0.795 0.7665 0.7924 0.7653 -0.32% -0.16%

B-3 0.7712 0.800 0.7641 0.7991 0.7637 -0.12% -0.05%

W-1 1.6656 1.670 1.6645 1.6512 1.6568 -1.14% -0.47%

W-2 1.3472 1.347 1.3425 1.3340 1.3378 -0.97% -0.35%


110

W-3 1.5594 1.564 1.5509 1.5487 1.5480 -0.99% -0.19%

Table 6-9. Results of temperature correction for the round to VSL.

artifact

set LED

1. meas.

of pilot

participant

lab

2. meas.


difference

ΦL* - ΦL

relative

difference


* (lm) ΦP2

* (lm)

#1

R-1 0.6733 0.683 0.6710 0.6795 0.6746 -0.52% 0.53%

R-2 0.6755 0.676 0.6750 0.6718 0.6787 -0.62% 0.55%

R-3 0.6477 0.652 0.6493 0.6482 0.6520 -0.59% 0.42%

G-1 2.9980 2.978 2.9756 2.9672 2.9373 -0.36% -1.30%

G-2 2.8473 2.872 2.8258 2.8613 2.8003 -0.37% -0.91%

G-3 2.9450 2.855 2.9262 2.8487 2.8932 -0.22% -1.14%

B-1 0.7488 0.769 0.7340 0.7696 0.7369 0.08% 0.39%

B-2 0.7609 0.790 0.7389 0.7900 0.7398 0.01% 0.12%

B-3 0.7969 0.809 0.7842 0.8084 0.7822 -0.08% -0.25%

W-1 1.7017 1.737 1.6839 1.7291 1.6639 -0.45% -1.20%

W-2 1.5761 1.575 1.5525 1.5685 1.5327 -0.42% -1.29%

W-3 1.7439 1.767 1.7087 1.7593 1.6885 -0.44% -1.20%

Table 6-10. Results of temperature correction for the round to NIST.

artifact

set LED

1. meas.

of pilot

participant

lab

2. meas.


difference

ΦL* - ΦL

relative

difference


* (lm) ΦP2

* (lm)

#3

R-1 0.6669 0.669 0.6724 0.6696 0.6657 0.08% -1.01%

R-2 0.6809 0.683 0.6856 0.6821 0.6799 -0.13% -0.84%

R-3 0.6791 0.682 0.6845 0.6816 0.6772 -0.05% -1.08%

G-1 2.9418 2.919 2.9298 2.9203 2.9213 0.05% -0.29%

G-2 2.6986 2.658 2.6758 2.6595 2.6667 0.05% -0.34%

G-3 2.6374 2.604 2.6189 2.6084 2.6107 0.17% -0.31%

B-1 0.7744 0.789 0.7597 0.7893 0.7593 0.04% -0.05%

B-2 0.8204 0.831 0.7993 0.8313 0.7989 0.03% -0.05%

B-3 0.6601 0.672 0.6454 0.6724 0.6448 0.06% -0.10%

W-1 1.6872 1.668 1.6709 1.6725 1.6628 0.27% -0.49%

W-2 1.6575 1.629 1.6334 1.6331 1.6261 0.25% -0.45%

W-3 1.7063 1.680 1.6805 1.6859 1.6761 0.35% -0.26%

Table 6-11. Results of temperature correction for the round to VNIIOFI.

artifact

set LED

1. meas.

of pilot

participant

lab

2. meas.


difference

ΦL* - ΦL

relative

difference


* (lm) ΦP2

* (lm)

#5 R-1 0.6868 0.736 0.6896 0.7132 0.6878 -3.20% -0.26%


111

R-2 0.6881 0.732 0.6885 0.7104 0.6879 -3.04% -0.09%

R-3 0.6542 0.707 0.6537 0.6860 0.6537 -3.06% 0.00%

G-1 2.9171 2.851 2.9012 2.8122 2.8901 -1.38% -0.38%

G-2 2.7947 2.712 2.7550 2.6754 2.7448 -1.37% -0.37%

G-3 2.8501 2.792 2.8421 2.7554 2.8297 -1.33% -0.44%

B-1 0.7768 0.738 0.7672 0.7412 0.7672 0.44% 0.00%

B-2 0.8803 0.839 0.8714 0.8410 0.8709 0.24% -0.07%

B-3 0.8219 0.816 0.8128 0.8187 0.8126 0.33% -0.02%

W-1 1.6842 1.715 1.6805 1.6752 1.6701 -2.37% -0.62%

W-2 1.7598 1.789 1.7476 1.7463 1.7390 -2.44% -0.49%

W-3 1.7101 1.745 1.6980 1.7002 1.6899 -2.64% -0.48%

Table 6-12. Results of temperature correction for the round to INM.

artifact

set LED

1. meas.

of pilot

participant

lab

2. meas.


difference

ΦL* - ΦL

relative

difference


* (lm) ΦP2

* (lm)

#7

R-1 0.6570 0.63 0.6619 0.6193 0.6565 -1.73% -0.81%

R-2 0.7234 0.69 0.7254 0.6822 0.7210 -1.14% -0.61%

R-3 0.6483 0.61 0.6527 0.6003 0.6469 -1.61% -0.91%

G-1 3.0222 2.71 3.0194 2.6948 3.0111 -0.56% -0.28%

G-2 2.8689 2.69 2.8643 2.6744 2.8581 -0.58% -0.22%

G-3 2.9972 2.76 3.0015 2.7429 2.9945 -0.62% -0.23%

B-1 0.8039 0.70 0.7630 0.6998 0.7627 -0.03% -0.04%

B-2 0.7763 0.64 0.7567 0.6399 0.7565 -0.02% -0.03%

B-3 0.7914 0.66 0.7792 0.6580 0.7778 -0.30% -0.18%

W-1 1.6525 1.50 1.6534 1.4842 1.6452 -1.07% -0.50%

W-2 1.5720 1.34 1.5728 1.3213 1.5624 -1.41% -0.67%

W-3 1.5882 1.35 1.5817 1.3340 1.5750 -1.20% -0.43%

Based on the temperature-corrected results of the pilot, the drift of the artifact LEDs

could be analyzed. Each LED is measured by the pilot two or three times depending on

the measurement rounds. The relative changes of the total luminous flux measured by

the pilot for each artifact LED are shown in the following figures, separated to a plot

without temperature correction and to a plot after correction. They show that the effect

of the temperature correction is small because the laboratory condition of the pilot was

little changed during the comparison. The most of the artifact LEDs show a drift smaller

than ±1 % for each round that is comparable to the measurement uncertainty of the

pilot. However, a few LEDs underwent a large drift and should be excluded from the data

analysis. The exclusion of the non-stable artifact LEDs is decided by the participant

through the procedure of review of relative data.


112

Fig. 6-1. Drift of the artefact set #1.




113





114



6.3. Review of Relative Data

The review of relative data started in December 2009. The pilot sent to the participants a

document with the relative data of each participant, which are the data reduced to check

only the stability of the artifact LEDs and the internal consistency of each participant. The

document circulated for the review of relative data is included in Appendix B: Review of

Relative Data as an electronic file. Note that both the uncorrected and temperature-

corrected data are separately presented.

The review comments of the participants are collected by the pilot and their

summary is included in Appendix C: Comments from Review of Relative Data. As a result

of the review of relative data, the data of the following artifact LEDs will be excluded

from the analysis on request of the participants.

- #1-W-1 measured by MIKES (large drift)

- #2-G-1 measured by CMS-ITRI (large drift)

- #4-W-1 measured by NMIJ (large drift)

- #7-B-1 measured by INM (large drift)


115

6.4. Review of Uncertainty Budgets

The review of relative data started in March 2010 and completed in June 2010. The pilot

summarized the technical reports and uncertainty budgets of the participants to one

document and sent it to all the participants. We note that VNIIOFI could not participate

to the review process because their submission of the technical report was abandoned.

The discussion among the participants and the revisions of the budgets are conducted

according to the CCPR Guidelines. The review comments of the participants are collected

by the pilot and their summary is included in Appendix D: Comments from Review of

Uncertainty Budgets. The final version of the uncertainty budgets is summarized in

Chapter 4.

6.5. Identification of Outliers

For the identification of outliers that can significantly skew the reference value of the

comparison, the pilot prepared a document with the relative deviation data of each

participant from the simple mean values of all the participants without disclosing the

participant’s identity and the absolute results. The document sent to the participant in

June 2010 is included in Appendix E: Identification of Outliers. As a result of the

discussion, it was agreed in September 2010 that the data with a relative deviation of

more than ±10 % from the mean are to identify as outliers. As the measurements of

each type (color) of LEDs are taken as each separate comparison, the outlier will be

excluded only from the analysis for the related LED type.


116

7. Data Analysis

The data analysis is performed based on the example in Appendix B of the CCPR

Guidelines.15 The only difference was the sequence of each round: “pilot – participant –

pilot” in the LED comparison, while “participant – pilot – participant” in the example of

the Guidelines. In this chapter, the equations of each analysis step are described. The

complete data of the calculation is included as an electronic file (Excel spreadsheet) at

the end of the chapter. Note that the analysis is repeated for each type of LEDs, and also

for the data without and with the temperature-correction.

7.1. Calculation of Difference to Pilot

For each participant with index i and for each LED with index j, the two measurement

results of the pilot (index P), before (index P1) and after (index P2) the participant, are

averaged by

1 2

, , ,

1

2

P P P

i j i j i j . (7-1)

The relative standard uncertainty of the pilot’s average value Φi,jP is calculated from the

relative standard uncertainty ur,corP of the correlated components (scale uncertainty) and

the relative standard uncertainty ur,ucP of the uncorrelated components (transfer

uncertainty) according to

2

2 2

, , ,21

1( )

2

P P Pk

r i j r cor r uc

k

u u u

. (7-2)

The values of ur,corP and ur,uc

Pk are determined by combing the related components in the

reported uncertainty budgets of the pilot in Table 4-1 ~ Table 4-4. Note that the pilot

reported and applied the upper boundary values for all the uncertainty components in

the budgets so that the relative standard uncertainty of each measurement remained the

same for each LED type.

The relative difference Δi,j between the participant i and the pilot (index P) for each

LED j is then calculated by

,

,

,

1i j

i j P

i j

(7-3)

and its uncertainty by

2 22

, , , , ,( ) P

i j r i j r uc r add i ju u u u . (7-4)

15 CCPR Key Comparison Working Group, Guidelines for CCPR Comparison Report Preparation, Rev. 2 (Sept. 18, 2009), available at http://www.bipm.org/en/committees/cc/ccpr/publications_cc.html

http://www.bipm.org/en/committees/cc/ccpr/publications_cc.html


117

Here, ur,add(Φi,j) denotes the additional uncertainty in the measurement of LED j by the

participant i due to non-ideal characteristics of the artifact LEDs. For the results without

temperature correction, we used the drift of the LED for the corresponding round as the

value of ur,add(Φi,j), which is calculated from the relative difference of the two

measurement results of the pilot. For the results with temperature correction, the relative

standard uncertainty of the correction procedure of 0.5 % is additionally combined to

ur,add(Φi,j). The relative standard uncertainty of the participant ur(Φi,j) is determined from

the reported expanded results in Chapter 5.

Finally, the results of the multiple LEDs for each type are averaged for the participant

i by

,

1

3i i j

j

. (7-5)

Under assumption that the results of multiple LEDs measured by the same participant are

strongly correlated, the uncertainty of the relative differences is calculated simply by

,

1

3i i j

j

u u . (7-6)

For the pilot, we use now the index i = 0 and set Δ0 = 0. According to Eq. (7-4), the

uncertainty u(Δ0) for the pilot is the same as the total relative standard uncertainty

averaged over all the measurements by the pilot. For case of the temperature corrected

results, we added also the uncertainty of the correction to u(Δ0).

7.2. Calculation of Comparison Reference Value

The Reference Value (RV) of the comparison for each LED type is calculated using

weighted mean with cut-off. The cut-off value ucut is calculated by

for ; 0,...,cut r i r i r iu average u u median u i N . (7-7)

Note that the outliers are not included in the calculation of the RV so that the number N

denotes the number of the participants whose comparison results are not identified as

outliers (the pilot not counted as a participant here).

The total relative standard uncertainty ur(Φi) of each participant i, averaged over LEDs

with different j, is adjusted by the cut-off (i = 0, …, N):

,

,

for

otherwise

r adj i r i r i cut

r adj i cut

u u u u

u u

(7-8)

Also, the uncertainty of Δi is adjusted after cut-off by

2 2

,adj i r adj i T iu u u . (7-9)


118

Here, uT(Δi) denotes the transfer uncertainty component in u(Δi), which is separated by

2 2

T i i r iu u u . (7-10)

These uncertainties are used to calculate the weights wi for each participant i given by

2

2

0

adj i

i N

adj i

i

uw

u

. (7-11)

Finally, the RV is determined by

0

N

RV i i

i

w

. (7-12)

The uncertainty of the comparison RV is given by

2

40

2

0

Ni

i adj i

RV N

adj i

i

u

uu

u

. (7-13)

7.3. Calculation of Degree of Equivalence

The unilateral degree of equivalence (DoE) of the participant i is defined by

i i RVD . (7-14)

The DoE is calculated according to Eq. (7-13) also for the participants whose comparison

results are identified as outliers. However, the uncertainty of DoE is different. For the

participants whose results are included in the calculation of the RV, the uncertainty of

DoE is given, as an expanded uncertainty with a coverage factor k = 2, by

2

2 2 2

20

2N

i

i i RV adj i

iadj i

uU k u u u

u

. (7-15)

For the participants whose results are excluded in the calculation of the RV, the

uncertainty of DoE is simplified to

2 2

i i RVU k u u . (7-16)

7.4. Data Analysis Spreadsheet

The Excel-file can be opened by a double-click on the icon below.

DoE_flux_rev.xlsx


119

8. Comparison Results

8.1. Red LEDs

The comparison RV for total luminous flux of red LEDs is calculated to be

0.01651, 0.80% ( 2)RV rU k

for the results without temperature correction, and

0.00083, 0.80% ( 2)RV rU k

for the results after temperature correction. Table 8-1

and Table 8-2 summarize the comparison results for red LEDs without and with

temperature correction, respectively. The last column of each table shows the En criteria

of each participant, which is defined as the absolute ratio of Di and U(Di). Note that the

results of CENAM and METAS are identified as outliers and not included in the

calculation of the RV.

Table 8-1. Comparison results for red LEDs without temperature correction.

participant Δi u(Δi) wi Di U(Di) En

MIKES 0.06423 1.41% 0.08441 0.048 0.027 1.778

CMS-ITRI -0.00078 1.27% 0.10341 -0.017 0.024 0.708

PTB -0.01098 1.77% 0.05333 -0.027 0.034 0.794

NMIJ -0.00810 1.56% 0.06876 -0.025 0.030 0.833

CENAM -0.21039 5.41% N.A. -0.227 0.108 2.102

LNE -0.01035 1.56% 0.06827 -0.027 0.030 0.900

METAS -0.11160 1.10% N.A. -0.128 0.023 5.565

NMC-

A*STAR 0.01576 1.10% 0.13038 -0.001 0.020 0.050

VSL 0.00753 1.55% 0.06978 -0.009 0.030 0.300

NIST -0.00036 1.12% 0.10546 -0.017 0.021 0.810

VNIIOFI 0.07135 1.03% 0.15859 0.055 0.019 2.895

INM -0.05148 3.19% 0.01638 -0.068 0.063 1.079

KRISS 0.00000 1.09% 0.14122 -0.017 0.020 0.850

Table 8-2. Comparison results for red LEDs after temperature correction.


MIKES 0.05266 1.50% 0.07649 0.052 0.029 1.793

CMS-ITRI -0.01332 1.59% 0.06800 -0.014 0.031 0.452

PTB -0.02768 1.54% 0.07237 -0.029 0.030 0.967

NMIJ -0.01458 1.37% 0.09185 -0.015 0.026 0.577

CENAM -0.22202 5.49% N.A. -0.223 0.110 2.027

LNE -0.02725 1.19% 0.12225 -0.028 0.022 1.273

METAS -0.13712 1.18% N.A. -0.138 0.025 5.520

NMC-

A*STAR -0.00028 1.10% 0.12349 -0.001 0.021 0.048

VSL -0.00075 1.68% 0.06090 -0.002 0.033 0.061

NIST 0.00422 0.97% 0.12227 0.003 0.019 0.158


120

VNIIOFI 0.03970 1.14% 0.12521 0.039 0.021 1.857

INM -0.06182 3.23% 0.01655 -0.063 0.064 0.984

KRISS 0.00000 1.20% 0.12061 -0.001 0.022 0.045

The DoEs and its uncertainties for red LEDs are plotted in Fig. 8-1 as dot symbols

and error bars, respectively. The red lines indicate the expanded relative uncertainty of

the comparison RV.

Fig. 8-1. DoE for red LEDs without and with temperature correction.


121

8.2. Green LEDs

The comparison RV for total luminous flux of green LEDs is calculated to be

0.00650, 0.76% ( 2)RV rU k


0.01221, 0.84% ( 2)RV rU k

for the results after temperature correction. Table

8-3 and Table 8-4 summarize the comparison results for green LEDs without and with



results of CENAM and METAS are identified as outliers and not included in the


Table 8-3. Comparison results for green LEDs without temperature correction.


MIKES 0.02507 1.37% 0.07955 0.032 0.026 1.231

CMS-ITRI 0.00960 1.36% 0.08110 0.016 0.026 0.615

PTB -0.03270 1.34% 0.08381 -0.026 0.026 1.000

NMIJ 0.01349 1.36% 0.08074 0.020 0.026 0.769

CENAM 0.16523 4.39% N.A. 0.172 0.088 1.955

LNE -0.01622 1.25% 0.09578 -0.010 0.024 0.417

METAS -0.16056 1.16% N.A. -0.154 0.024 6.417

NMC-

A*STAR -0.00998 0.98% 0.15190 -0.003 0.018 0.167

VSL -0.00596 1.52% 0.06463 0.001 0.029 0.034

NIST -0.00860 0.83% 0.12400 -0.002 0.016 0.125

VNIIOFI -0.02055 1.29% 0.09090 -0.014 0.024 0.583

INM -0.08144 3.15% 0.01518 -0.075 0.062 1.210

KRISS 0.00000 1.07% 0.13241 0.007 0.020 0.350

Table 8-4. Comparison results for green LEDs after temperature correction.


MIKES 0.02004 1.47% 0.08502 0.032 0.028 1.143

CMS-ITRI -0.00846 1.43% 0.09050 0.004 0.027 0.148

PTB -0.04258 1.30% 0.10855 -0.030 0.025 1.200

NMIJ 0.00764 1.30% 0.10952 0.020 0.024 0.833

CENAM 0.15880 4.49% N.A. 0.171 0.090 1.900

LNE -0.02291 1.17% 0.13367 -0.011 0.022 0.500

METAS -0.16574 1.32% N.A. -0.154 0.028 5.500

NMC-

A*STAR -0.01490 1.21% 0.12361 -0.003 0.023 0.130

VSL -0.00365 2.31% 0.03444 0.009 0.045 0.200

NIST -0.00615 1.20% 0.09330 0.006 0.023 0.261

VNIIOFI -0.03176 1.62% 0.07027 -0.020 0.031 0.645

INM -0.08572 3.21% 0.01785 -0.074 0.064 1.156

KRISS 0.00000 1.18% 0.13326 0.012 0.022 0.545


122

The DoEs and its uncertainties for green LEDs are plotted in Fig. 8-2 as dot symbols


the comparison RV.

Fig. 8-2. DoE for green LEDs without and with temperature correction.

8.3. Blue LEDs

The comparison RV for total luminous flux of blue LEDs is calculated to be

0.01187, 0.92% ( 2)RV rU k



123

0.01081, 0.96% ( 2)RV rU k

for the results after temperature correction. Table 8-5

and Table 8-6 summarize the comparison results for blue LEDs without and with



results of CENAM, METAS, and INM are identified as outliers and not included in the


Table 8-5. Comparison results for blue LEDs without temperature correction.


MIKES 0.01785 2.68% 0.03056 0.006 0.053 0.113

CMS-ITRI -0.01958 1.64% 0.08120 -0.031 0.031 1.000

PTB -0.00084 1.49% 0.09843 -0.013 0.028 0.464

NMIJ 0.06179 1.09% 0.16375 0.050 0.020 2.500

CENAM 0.10325 5.51% N.A. 0.091 0.111 0.820

LNE -0.00932 1.25% 0.14046 -0.021 0.023 0.913

METAS -0.10229 1.89% N.A. -0.114 0.039 2.923

NMC-

A*STAR 0.03487 1.16% 0.13661 0.023 0.022 1.045

VSL 0.03800 2.56% 0.03344 0.026 0.050 0.520

NIST 0.02806 2.54% 0.03397 0.016 0.050 0.320

VNIIOFI -0.02929 1.50% 0.09556 -0.041 0.028 1.464

INM -0.16231 7.97% N.A. -0.174 0.160 1.088

KRISS 0.00000 1.09% 0.18601 -0.012 0.020 0.600

Table 8-6. Comparison results for blue LEDs after temperature correction.


MIKES 0.01793 2.73% 0.03434 0.007 0.054 0.130

CMS-ITRI -0.01923 1.70% 0.08805 -0.030 0.032 0.938

PTB -0.00317 1.55% 0.10659 -0.014 0.029 0.483

NMIJ 0.06262 1.22% 0.15207 0.052 0.023 2.261

CENAM 0.10250 5.55% N.A. 0.092 0.112 0.821

LNE -0.00997 1.33% 0.14503 -0.021 0.024 0.875

METAS -0.10341 1.97% N.A. -0.114 0.041 2.780

NMC-

A*STAR 0.03316 1.32% 0.12349 0.022 0.025 0.880

VSL 0.03757 2.54% 0.03960 0.027 0.050 0.540

NIST 0.02885 2.65% 0.03641 0.018 0.052 0.346

VNIIOFI -0.02587 1.60% 0.09591 -0.037 0.030 1.233

INM -0.16322 8.02% N.A. -0.174 0.161 1.081

KRISS 0.00000 1.20% 0.17851 -0.011 0.022 0.500

The DoEs and its uncertainties for blue LEDs are plotted in Fig. 8-3 as dot symbols



124

the comparison RV.

Fig. 8-3. DoE for blue LEDs without and with temperature correction.

8.4. White LEDs

The comparison RV for total luminous flux of white LEDs is calculated to be

0.00824, 0.92% ( 2)RV rU k


0.00284, 1.02% ( 2)RV rU k

for the results after temperature correction. Table

8-7 and Table 8-8 summarize the comparison results for white LEDs without and with


125



results of METAS and INM are identified as outliers and not included in the calculation of

the RV.

Table 8-7. Comparison results for white LEDs without temperature correction.


MIKES 0.04573 2.09% 0.04962 0.037 0.041 0.902

CMS-ITRI 0.01989 1.42% 0.10838 0.012 0.027 0.444

PTB -0.00672 1.63% 0.08196 -0.015 0.031 0.484

NMIJ 0.01351 2.09% 0.04879 0.005 0.041 0.122

CENAM 0.06501 4.86% 0.00922 0.057 0.097 0.588

LNE -0.00758 2.75% 0.02840 -0.016 0.054 0.296

METAS -0.20539 1.03% N.A. -0.214 0.023 9.304

NMC-

A*STAR 0.00342 0.98% 0.22825 -0.005 0.017 0.294

VSL 0.01885 2.40% 0.03770 0.011 0.047 0.234

NIST -0.00817 1.42% 0.08668 -0.016 0.027 0.593

VNIIOFI 0.02119 1.35% 0.12009 0.013 0.025 0.520

INM -0.12951 7.19% N.A. -0.138 0.144 0.958

KRISS 0.00000 1.04% 0.20093 -0.008 0.019 0.421

Table 8-8. Comparison results for white LEDs after temperature correction.


MIKES 0.04159 2.17% 0.05627 0.044 0.042 1.048

CMS-ITRI -0.01218 1.47% 0.12262 -0.009 0.027 0.333

PTB -0.02169 1.38% 0.13961 -0.019 0.026 0.731

NMIJ 0.00348 1.80% 0.07847 0.006 0.035 0.171

CENAM 0.05096 4.94% 0.01082 0.054 0.098 0.551

LNE -0.01950 2.58% 0.03883 -0.017 0.051 0.333

METAS -0.21406 1.21% N.A. -0.211 0.026 8.115

NMC-

A*STAR -0.00518 1.23% 0.17513 -0.002 0.022 0.091

VSL 0.02059 3.33% 0.02382 0.023 0.066 0.348

NIST -0.00331 1.86% 0.06638 0.000 0.036 0.000

VNIIOFI -0.00095 1.72% 0.08976 0.002 0.033 0.061

INM -0.13775 7.32% N.A. -0.135 0.147 0.918

KRISS 0.00000 1.15% 0.19830 0.003 0.021 0.143

The DoEs and its uncertainties for white LEDs are plotted in Fig. 8-4 as dot symbols


the comparison RV.


126

Fig. 8-4. DoE for white LEDs without and with temperature correction.


127

9. Discussion

9.1. Test of Consistency

In order to test the consistency of the comparison results, the Birge ratio RB is calculated

by the equation

2

RV

B 20 adj

1

( )

Ni

i i

RN u

, (9-1)

where N is the number of the participants, without counting the pilot, whose results are

used for the calculation of the RV. For this calculation, the data of the outliers are not

used. Note that the consistency is satisfied, if RB ≤ 1.

Table 9-1 shows the calculated Birge ratios of the comparison S3b without and with

temperature correction. The results of the white LEDs show the satisfactory consistency.

For the other LEDs, the values of RB range from 1.4 to 2.4, which indicate that the

uncertainties of the participants are underestimated. Table 9-1 also shows that the

temperature correction has the effect of decreasing the Birge ratios and, hence,

improving the consistency. This verifies that the temperature correction based on the

junction voltage measurement described in Chapter 3 is capable to correct the

systematic errors of the artifact LEDs due to different measurement conditions.

Table 9-1. Birge ratio of the comparison S3b.

LED type Birge ratio

without T correction

Birge ratio after T

correction

Red 2.426 1.944

Green 1.437 1.469

Blue 2.087 1.894

White 0.976 0.938

9.2. Accuracy of Color Correction

The narrow spectral bandwidth of LEDs is another important source of systematic errors

in the photometric measurement of LEDs. If a photometer is used for LED measurements,

correction of spectral mismatch, often referred to as color correction, is essential to

achieve high accuracy, which requires both relative spectral distribution of the test LED

and relative spectral responsivity of the photometer. As we have circulated four different

colors of LEDs (R/G/B/W), analysis of the dependence of the comparison results upon the

LED colors can provide important information on the accuracy of color correction. Table


128

9-2 summarizes the DoEs of each participant for different colors of LEDs, which are

based on the temperature corrected data.

Table 9-2. DoEs for different LED colors (after temperature correction).

participant DoE for red

LEDs

DoE for green

LEDs

DoE for blue

LEDs

DoE for white

LEDs MIKES 0.052 0.032 0.007 0.044

CMS-ITRI -0.014 0.004 -0.030 -0.009

PTB -0.029 -0.030 -0.014 -0.019

NMIJ -0.015 0.020 0.052 0.006

CENAM -0.223 0.171 0.092 0.054

LNE -0.028 -0.011 -0.021 -0.017

METAS -0.138 -0.154 -0.114 -0.211

NMC-A*STAR -0.001 -0.003 0.022 -0.002

VSL -0.002 0.009 0.027 0.023

NIST 0.003 0.006 0.018 0.000

VNIIOFI 0.039 -0.020 -0.037 0.002

INM -0.063 -0.074 -0.174 -0.135

KRISS -0.001 0.012 -0.011 0.003

Fig. 9-1 shows plots of the data in Table 9-2. We classified the participants to three

groups. The first group shown on the top plot in Fig. 9-1 have only a weak (< 2 %)

dependence of DoE on the LED colors. The second group shown on the middle plot in

Fig. 9-1 have a moderate (3 % ~ 8 %) dependence of DoE on the LED colors. Especially,

the results of many participants have a maximum or a minimum for blue LEDs. The last

group shown on the bottom plot in Fig. 9-1 have a large dependence of DoE on the LED

colors or a too large offset. The results of Table 9-2 and Fig. 9-1 can be useful for the

participants to investigate the unknown systematic errors in their color correction.


129

Fig. 9-1. Plots of DoEs for different colors of LEDs (R, G, B, W). The participants are classified to three groups.


130

10. Summary

The measurement of total luminous flux is compared by circulating four different types of

artifact LEDs (red, green, blue, and white) to 13 NMIs (including the pilot). The artifact

LEDs are prepared by the functional seasoning to enable a temperature correction based

on the junction voltage measurement. The comparison reference values and the

unilateral degrees of equivalence (DoEs) of each participant are calculated for each type

of LEDs from the reported measurement results. Table 10-1 shows the summary of the

DoEs and their uncertainties of each participant for each type of LEDs as the comparison

result.

Table 10-1. Summary of the unilateral DoEs and their uncertainties for APMP.PR-S3b (temperature correction applied).

NMI

RED GREEN BLUE WHITE

DoE U of

DoE DoE

U of

DoE DoE

U of

DoE DoE

U of

DoE

MIKES 0.052 0.029 0.032 0.028 0.007 0.054 0.044 0.042

CMS-ITRI -0.014 0.031 0.004 0.027 -0.030 0.032 -0.009 0.027

PTB -0.029 0.030 -0.030 0.025 -0.014 0.029 -0.019 0.026

NMIJ -0.015 0.026 0.020 0.024 0.052 0.023 0.006 0.035

CENAM -0.223 0.110 0.171 0.090 0.092 0.112 0.054 0.098

LNE -0.028 0.022 -0.011 0.022 -0.021 0.024 -0.017 0.051

METAS -0.138 0.025 -0.154 0.028 -0.114 0.041 -0.211 0.026

NMC-

A*STAR -0.001 0.021 -0.003 0.023 0.022 0.025 -0.002 0.022

VSL -0.002 0.033 0.009 0.045 0.027 0.050 0.023 0.066

NIST 0.003 0.019 0.006 0.023 0.018 0.052 0.000 0.036

VNIIOFI 0.039 0.021 -0.020 0.031 -0.037 0.030 0.002 0.033

INM -0.063 0.064 -0.074 0.064 -0.174 0.161 -0.135 0.147

KRISS -0.001 0.022 0.012 0.022 -0.011 0.022 0.003 0.021

Acknowledgement

The pilot work of this comparison is partly supported by the Korean Ministry of

Knowledge and Economy under the project of LED standardization, grant B0010209.


131

Appendix A: Technical Protocol

The pdf-file can be opened by a double-click on the image below.


132

Appendix B: Review of Relative Data



133

Appendix C: Comments from Review of Relative Data



134

Appendix D: Comments from Review of Uncertainty Budgets



135

Appendix E: Identification of Outliers



136

Appendix F: Comments and Revision to Draft A Report

Comments from PTB to Data Analysis

Results on 11 April 2011

Replies by the pilot on 17 June 2011

Enclosed please find copies of your files

with some marked blue cells. We think

there are some small bugs.

I have checked them and corrected. Thank

you!

It is possible to refer this comparison to

KCRV using link laboratories.

In principle yes. But the related KC, e.g. of

luminous flux, was done with a different

artifact so that it cannot be directly

compared to this LED comparison. That is

also the reason why this is a

supplementary comparison. We can try to

do such a linkage as an interesting study,

but not as a part of the comparison report.

The resulting excel graphic looks a little bit

strange. We feel is should look similar like

the following graphic (uDoE should be

plotted around DoE):

I agree and I checked that this is also

common for KCs. I will modify the graphs

as you suggest.

It may be helpful to calculate the Birge

ratio to get information about the

consistency of the comparisons. It is

calculated from the internal and external

This is a good suggestion. I will surely try

to calculate both the Birge ratio and the En

values and include the results in the final

report. This will provide valuable


137

consistency. A value of close to 1 or less

indicates that the results are consistent.

Values greater than 1 are not.

in

extB

u

uR with

n

i

i

n

i

ii

Dun

DuD

u

1

2

1

2

ext

)()1(

)](/[

and

2/1

1

2

in )(

n

i

iDuu .

For luminous flux we found values around

2 for most cases. For luminous intensity

(without diffuse LEDs) we found values

around 1. Please see enclosed jpg files (I

apologize this jpg, but is takes a while to

get nice prints with mathematica). We also

calculated the criteria by

)(2Absin,

i

i

DoEu

DoEE

Values greater than 1 indicate a too small

uncertainty of the participant. So we

suggest to use specific enlargements of the

participant uncertainties in that way that

the Birge ratio is equal or less 1 and

criteria is close to 1. This procedure also

would solve the problem of outliers.

information to the next version of the KC

guidelines which should include a

procedure of consistency check and of a

better outlier selection.

Comments from PTB to Draft A Report

on 19 Oct 2011

Replies by the pilot on 22 Nov 2011

We found some typing errors in the draft

A paper. Enclosed please find our errata

ZIP-file.

I have checked the errors. But all the errors

you found were the corrections of the

uncertainty budgets of PTB. These,

however, cannot be corrected in the draft

A report stage, because they are already


138

reviewed by the participants. This is

communicated via email on 21.10.2011.

PTB has acknowledged this and confirms

that these corrections do not affect the

comparison results. Therefore, the

corrections are not considered in the

revision of the draft A report.

The Plots Fig. 9-1 of S3a and S3b are very

helpful. It would be great to have these

plots for S3c, too.

In case of S3c, the plots such as Fig. 9-1 of

S3a and S3b were not easy because a 2

dimensional plot is required to make

systematic effects visible. I will try to realize

this in the next revision of the S3c report,

but I should also manage the workload.

Based on the results data, however, each

participant can make such analysis to

investigate the systematic effect of his

results.

The appendices should include all

important comments, suggestions and

recommendations of the participants to

simplify future comparisons. For example

our Suggestions PTB.docx of 15.04.2011.

I will make another appendix to record the

comments during the draft A report

procedure.

The tables in chapter 8 should include the

criteria

)(2Absin,

i

i

DoEu

DoEE

that would be helpful to evaluate the

stated uncertainty by each participants.

I will consider this in the revision.

The Birge ratios stated in table 9-1

especially for S3a and S3a are often

significant greater than 1. I think the

I agree that the large Birge ratio means

that the uncertainties of the participants

are underestimated. I wrote this also in the


139

meaning of that is, that some stated

uncertainties are too small. Please, refer

the related En criteria.

Here we have an additional hint for that.

The first diagram from S3a (intensity)

shows a relative flat DoE around 0% of PTB

results. But the second diagram from S3b

(flux) shows relative big differences

between (R,G) and (B,W) LEDs. The

luminous flux values were determined by a

goniophotometer directly after the

luminous intensity measurement with the

same operation state of the LED and in the

same system without new alignment of the

LED. So there is no reason for that

difference. We know from hundreds of

measurements that the integration

capability of the goniophotometer has a

very high reproducibility.

So I think this a hint for an inconsistency

of the data as we know from the Birge

ratios.

report. Your statement will be documented

in the Appendix of the revised report.


140

List of Revisions from A-1 to A-2

Draft A-1 Draft A-2

Page 99, section 5.4, the first line: “June ~

July 2008 (the exact dates not reported)”

Corrected to “from 16 June to 2 July 2008”

based on the result verification records.

Chapter 8, the first paragraph of each

section.

Addition of a sentence “The last column of

each table Table 8-2shows the En criteria of

each participant, which is defined as the

absolute ratio of Di and U(Di).”

Chapter 8, Table 8-1 ~ Table 8-8. Addition of a new column with the

calculated En criteria values.

After Chapter 10 Addition of <Acknowledgment> by the

pilot.

After Appendix E Addition of <Appendix F: Comments and

Revision to Draft A Report>

Statement of METAS to the results of S3b on 24 Jan 2012 (added to Draft B)

Comparison APMP.PR-S3a (averaged luminous intensity of single LEDs) has shown that

the photometric scale of luminous intensity and illuminance of METAS is in agreement

with the world mean value.

For the comparison APMP.PR-S3b (luminous flux of single LEDs) a new facility

was build based on a 1 m integrating sphere. The calibration was done directly traceably

through APMP.PR-S3a by determine the flux responsivity of the sphere using a LED of

known averaged luminous intensity and a precision sphere aperture positioned at 100

mm distance to the LED.

Analysing the situation we found different problems in the set up, mainly the

general reproducibility, sphere uniformity, linearity of the photometer, stability of the

photometer, and absorption problems of the LED holder system. Unfortunately no other

validation of the calibration capability was made prior to the APMP.PR-S3.b comparison.


141

No calibration services for the quantity luminous flux for single LEDs were ever

offered to costumers. METAS has no intention to do so in near future. No CMCs are

affected by the results of the APMP.PR-S3.b comparison. N.B. METAS is offering

measurement and calibration of luminous flux of diverse lamps and luminaires (including

LED luminaires). These measurements are performed on the METAS primary flux scale

realized by a goniophotometer. This quantity is directly traceable to the photometric

scale of METAS through calibrated illuminance meters. This competence has been shown

through the successful participation at the most recent CCPR comparison (CCPR-K4) and

are internationally accepted (CMCs on luminous flux).

No further corrective actions are foreseen in near future (no subsequential

comparison on that quantity).

- End of the Report -

APMP supplementary comparison 1

Technical Protocol on


LED Measurements




Approved in January 2008, Revised in November 2008 due to change of participants list

Contents

1. INTRODUCTION........................................................................................................................................ 2

2. ORGANIZATION........................................................................................................................................ 2

2.1. CONDITION OF PARTICIPATION ............................................................................................................... 2 2.2. LIST OF PARTICIPANTS............................................................................................................................ 3 2.3. FORM OF COMPARISON ........................................................................................................................... 4 2.4. TIMETABLE............................................................................................................................................. 4 2.5. TRANSPORT AND HANDLING OF ARTEFACTS ........................................................................................... 6

3. DESCRIPTION OF ARTEFACTS ............................................................................................................ 7

4. MEASUREMENT INSTRUCTIONS ........................................................................................................ 9

4.1. AVERAGED LED INTENSITY (S3A) ......................................................................................................... 9 4.2. TOTAL LUMINOUS FLUX (S3B).............................................................................................................. 11 4.3. EMITTED COLOUR (S3C) ....................................................................................................................... 12

5. REPORTING OF RESULTS AND UNCERTAINTIES ........................................................................ 12

5.1. AVERAGED LED INTENSITY (S3A) ....................................................................................................... 12 5.2. TOTAL LUMINOUS FLUX (S3B).............................................................................................................. 13 5.3. EMITTED COLOUR (S3C) ....................................................................................................................... 13

6. PREPARATION OF COMPARISON REPORT.................................................................................... 14

APPENDIX 1: INSPECTION REPORT ON RECEIPT OF ARTEFACTS................................................. 15

APPENDIX 2: RESULT REPORT OF AVERAGED LED INTENSITY (S3A).......................................... 16

APPENDIX 3: RESULT REPORT OF TOTAL LUMINOUS FLUX (S3B) ................................................ 17

APPENDIX 4: RESULT REPORT OF EMITTED COLOUR (S3C) ........................................................... 18

Technical protocol on comparison of LED measurements



1. INTRODUCTION

Under the Mutual Recognition Arrangement (MRA), the metrological equivalence of national measurement standards will be determined by a set of key comparisons chosen and organized by the consultative committees of CIPM working closely with regional metrology organizations (RMOs). In addition, RMOs can organize supplementary comparisons which should be carried out in the same procedure as that of key comparisons following the guidelines established by BIPM1.

At its meeting in December 2006, Asia Pacific Metrology Programme (APMP) Technical Committee of Photometry and Radiometry (TCPR) proposed several regional comparisons in the field of optical radiation metrology. One of those, a set of photometric quantities of light-emitting diodes (LEDs) has been agreed to be conducted with Korea Research Institute of Standards and Science (KRISS) of Republic Korea as the pilot institute. It is also decided that APMP TCPR invites the institutes of other RMOs to participate this supplementary comparison.

In March 2007, the first invitation to participate is distributed to the members of Consultative Committee of Photometry and Radiometry (CCPR) of CIPM by the chairperson of APMP TCPR. Based on the responses to this invitation, a provisional list of participants is prepared.

Three measurement quantities of LEDs are selected for the comparison, which are listed as service categories for Calibration and Measurement Capabilities (CMCs): averaged LED intensity defined by International Commission on Illumination (CIE), total luminous flux of LEDs, and emitted colour of LEDs expressed as chromaticity coordinates (x, y) according to the CIE 1931 standard colorimetric system.2

It should be noted that total luminous flux is the measurement quantity for CCPR-K4. The current supplementary comparison of total luminous flux of LEDs is, however, not to be linked to this KC, but can be regarded as a pilot study testing the suitability of LEDs as an alternative artefact for CCPR-K4.

This document is to treat the technical protocol for the comparison of LED measurements, and has been prepared by KRISS and agreed by all the participants on the preliminary list.

2. ORGANIZATION

2.1. CONDITION OF PARTICIPATION

KRISS is acting as the pilot institute in the comparison among the participants.

Three comparisons for three measurement quantities are conducted simultaneously by circulating one artefact set. The participant can decide to take part in only one or two of the three comparisons by selecting the measurement quantities. However, it should be declared with the confirmation of participation and stated in the technical protocol.

All the participants must be able to demonstrate traceability to an independent realization of each quantity, or make clear the route of traceability via another named laboratory.

By their declared intention to participate in this comparison, the laboratories accept the general instructions and the technical procedures written down in this document and commit themselves to follow the procedures strictly.

1 Guidelines for CIPM Key Comparisons, March 1999 (modified in October 2003). Available at http://www.bipm.fr/en/convention/mra/guidelines_kcs/ 2 Measurement of LEDs, CIE Technical Report 127-1997.

http://www.bipm.fr/en/convention/mra/guidelines_kcs/



Once the protocol has been agreed, no change to the protocol may be made without prior agreement of all the participants.

2.2. LIST OF PARTICIPANTS

(Nr.) NMI

country contact person email address post address participating comparisons

(1) KRISS

Rep. Korea Seongchong

Park, Dong-Hoon Lee

[email protected] [email protected]

Division of Physical Metrology Korea Research Institute of Standards

and Science 1 Doryong-Dong, Yuseong-Gu Daejeon 305-340, Rep. Korea

all (S3a, S3b,

S3c)

(2)3 NMC-

A*STAR Singapore

Yuanjie Liu, Gan Xu

[email protected]

[email protected]

Optical Metrology Department National Metrology Centre

1 Science Park Drive Singapore 118221

all

(3) MIKES

Finland Pasi Manninen [email protected]

Metrology Research Institute Helsinki University of Technology

P.O.Box 3000 FI-02015 TKK, Finland

all

(4) NIST

USA Cameron Miller,

Yoshi Ohno Yuqin Zong

[email protected]@nist.gov

[email protected]

Optical Technology Division National Institute of Science and

Technology 100 Bureau Drive, Mailstop 8442

Gaithersburg, MD 20899-8442, USA

all

(5) CMS-ITRI

Chinese Taipei

Cheng-Hsien Chen

[email protected]

CMS/ITRI Rm. 301, Bldg. 16, 321, Sec. 2,

Kuang Fu Rd. Hsinchu, Taiwan 300, R.O.C.

all

(6) PTB

Germany Matthias

Lindemann Robert Maass

[email protected]

[email protected]

Physikalisch-Technische Bundesanstalt

AG 4.15, Goniophotometrie Bundesallee 100,

D-38116 Braunschweig, Germany

all

(7) CENAM

Mexico

Laura P. González, Anayansi Estrada,

Eric Rosas

[email protected]@[email protected]

División de Óptica y Radiometría Centro Nacional de Metrología

km 4,5 Carretera a Los Cués 76241, El Marqués, Querétaro,

México

all

(8)

NMIJ Japan

Kenji Godo, Terubumi Saito

[email protected] [email protected]

Optical Radiation Section Photometry and Radiometry Division

National Institute of Advanced Industrial Science and Technology 1-1-1, Umezono, Tsukuba, Ibaraki,

JAPAN 305-8563

all

(9) METAS

Switzerland Peter Blattner Peter.Blattner@metas

.ch

Federal Office of Metrology Lindenweg 50, 3003 Bern-Wabern

Switzerland all

(10) NPL

UK Paul Miller, Nigel Fox

[email protected]

[email protected]

National Physical Laboratory Hampton Road, Teddington, Middx,

TW11 0LW, UK all

(11) LNE

France Jimmy Dubard [email protected]

Laboratoire National de Métrologie et d’Essais

29, avenue Roger Hennequin 78197 TRAPPES, FRANCE

all

3 Formerly SPRING


(12) NMi VSL

The Netherlands

Eric W.M. van der Ham, M.Charl

Moolman

[email protected]@NMi.nl

NMi Van Swinden Laboratorium B.V. Department Electricity, Radiation and

Length Section Optics Thijsseweg 11, 2629 JA Delft

Zuid-Holland, The Netherlands

all

(13) NMIA

Australia Philip Lukins Philip.Lukins@measu

rement.gov.au

National Measurement Institute of Australia

2 Bradfield Rd Lindfield, NSW 2070, Australia

all

(14) VNIIOFI

Russia Tatiana

Gorshkova [email protected]

All-Russian Research Institute for Optical and Physical Measurements

Ozernaya 46 119361 Moscow, Russia

all

(15) MKEH

Hungary George Andor [email protected]

Magyar Kereskedelmi és Engedélyezési Hivatal (MKEH)

Németvölgyi út 37-39 H-1124 Budapest XII.

Hungary

all

(16) INM

Romania Mihai

Simionescu mihai.simionescu@in

m.ro

Institutul National de Metrologie Sos. Vitan Barzesti nr.11, Sector 4

Bucharest, Romania all

2.3. FORM OF COMPARISON

The comparison is carried out by distributing 8 sets of the artefact standard LEDs prepared and provided by the pilot. Each set of the artefact LEDs contains 14 pieces of LED, consisting of 12 lamp-type, 5-mm diameter LEDs (3 x Red, 3 x Green, 3 x Blue, 3 x White) and 2 specially-designed diffuser-type green LEDs. The specifications, preparation, and characteristics of the standard LEDs are described in Chapter 3.

The comparison runs as a star-type. The pilot sends to each participant one set of the artefact LEDs after preparation and characterisation. The participant measures (1) the averaged LED intensity in the CIE condition B, and/or (2) the total luminous flux, and/or (3) the chromaticity coordinate CIE1931 (x,y) of every artefact LEDs according to the introductions described in Chapter 4. After the measurement, the participant sends the artefact set back to the pilot, who characterises it again to check out a possible drift or change. The measurement results should be reported to the pilot as soon as possible after the measurement is finished according to the guidelines in Chapter 5.

The timetable given below shows an overview on how the comparison is to be preceded. Since the preparation of the artefact LEDs takes much time (over 300 hours) due to seasoning process, the pilot requires at least one month preparing the artefact LEDs ready for delivery. The pilot tries to provide as many artefact sets as possible so that the circulation runs without significant loss of time (multiple star-type circulation).

Each participant has two months for measurement after the receipt of the artefact set. With its confirmation to participate, each participant has confirmed that it is capable of performing the measurements in the time allocated to it. If anything happens so that it can not meet the timetable, the participant must contact the pilot immediately.

2.4. TIMETABLE

Time Activity of pilot Activity of participants

July 2007 ~ January 2008

- Preparation of artefact sets (#1 ~ #8) - Preparation of technical protocol draft

- Review of technical protocol draft



January 2008 - Finalization and approval of technical protocol by APMP TCPR

February 2008

- Control measurement of artefact set #1 and #2

- Delivery of artefact set #1 to MIKES - Delivery of artefact set #2 to CMS-ITRI

March 2008


- Delivery of artefact set #3 to PTB - Delivery of artefact set #4 to NMIJ

- Receipt of artefact set #1 in MIKES, Finland

- Receipt of artefact set #2 in CMS-ITRI, Taiwan

April 2008


- Delivery of artefact set #5 to CENAM- Delivery of artefact set #6 to LNE

- Receipt of artefact set #3 in PTB, Germany

- Receipt of artefact set #4 in NMIJ, Japan

May 2008


- Delivery of artefact set #7 to METAS - Delivery of artefact set #8 to NMC-A*STAR

- Receipt of artefact set #5 in CENAM, Mexico

- Receipt of artefact set #6 in LNE, France

- Return of artefact set #1 and #2 to KRISS (MIKES, CMS-ITRI)

June 2008


- Delivery of artefact set #1 to NMi-VSL

- Delivery of artefact set #2 to NMIA

- Receipt of artefact set #7 in METAS, Switzerland

- Receipt of artefact set #8 in NMC-A*STAR, Singapore

- Return of artefact set #3 and #4 to KRISS (PTB, NMIJ)

July 2008


- Delivery of artefact set #3 to NIST - Delivery of artefact set #4 to NPL

- Receipt of artefact set #1 in NMi-VSL, The Netherlands

- Receipt of artefact set #2 in NMIA, Australia

- Return of artefact set #5 and #6 to KRISS (CENAM, LNE)

August 2008


- Delivery of artefact set #5 to VNIIOFI- Delivery of artefact set #6 to MKEH

- Receipt of artefact set #3 in NIST, USA

- Receipt of artefact set #4 in NPL, UK

- Return of artefact set #7 and #8 to KRISS (METAS, NMC-A*STAR)

September 2008 - Control measurement of artefact set #7 and #8

- Receipt of artefact set #5 in VNIIOFI, Russia

- Receipt of artefact set #6 in MKEH, Hungary

- Return of artefact set #1 and #2 to KRISS (NMi-VSL, NMIA)

October 2008 - Control measurement of artefact set #1 and #2

- Delivery of artefact set #7 to INM

- Return of artefact set #3 and #4 to KRISS (NIST, NPL)



November 2008 - Control measurement of artefact set #3 and #4

- Return of artefact set #5 and #6 to KRISS (VNIIOFI, MKEH)

- Receipt of artefact set #7 in INM, Romania

December 2008


- Control measurement of artefact set #7

- Return of artefact set #7 to KRISS (INM)

January 2009 ~ April 2009

- Pre-Draft A process 1: distribution of uncertainty budget - Pre-Draft A process 2: review of relative data

May 2009 ~ June 2009

- Draft A report: preparation and distribution

July 2009 ~ August 2009

- Draft A report: review and approval by the participants

Sept. 2009 ~ October 2009

- Draft B report: preparation and submission to TCPR (Or Draft A-2 report process, if required)

2.5. TRANSPORT AND HANDLING OF ARTEFACTS

Each set of 14 artefact LEDs is transported in a wooden box (size 90 cm x 90 cm x 80 cm) with conductive foam matting, where the LEDs are pinned down at the specified positions. Packaging of the box should be sufficiently robust to be sent by courier, but precautions must be taken to prevent any damage by mechanical impact, heat, water, and moisture. The artefact set will be accompanied by a suitable customs carnet (where appropriate) or documentation identifying the items uniquely.

Each participating laboratory covers the cost for its own measurements, transportation and any customs charges as well as for any damages that may have occurred within its country.

The artefact LEDs should be visually inspected immediately upon receipt. However, care should be taken to ensure that the LEDs have sufficient time to acclimatise to the laboratory environment thus preventing any condensation, etc. The condition of the artefact LEDs and associated packaging should be noted and communicated via email and fax to the pilot by using the form APPENDIX 1: INSPECTION REPORT ON RECEIPT OF ARTEFACTS.

The artefact LEDs should be handled only by the authorized persons, who are well aware of the cautions stated in the manufacturer’s specification sheets of the artefact LEDs.

LEDs can be damaged by static electricity or surge voltage. Using an anti-static wrist band is strongly recommended. When the LEDs are not installed for measurement, they should always be kept at the specified positions on the conductive foam matting in the package box, which prevents not only electrostatic and mechanical damages but also confusion in identifying each LED.

The LEDs should never be touched with bare hands. Please use an anti-static vinyl glove in handling the LEDs. No cleaning of LEDs should be attempted except using dry air.

The mechanical condition of the LEDs should never be changed by actions such as soldering, cutting, polishing, and bonding.

If an artefact LED is damaged or shows any unusual property during operation, the operation should immediately be terminated and the pilot should be contacted.

After measurement, the artefact LEDs should be repackaged as received. Ensure that the content of the package is complete before shipment.




3. DESCRIPTION OF ARTEFACTS

The artefact LEDs are prepared from the commercially available “raw” LEDs in the following procedure:

1. Seasoning: the raw LEDs are pre-burned for more than 300 hours while the temporal change of their electrical and optical properties are recorded. The temporal drift and the temperature dependence of the optical characteristics of each LED are determined during the seasoning process.

2. Selection: based on the seasoning characteristics, the LEDs with predictable seasoning characteristics are selected as the artefact LEDs for the comparison.

3. Test measurement: the photometric quantities of the artefact LEDs are measured by the pilot before sent to each participant. The measurement by the pilot is repeated when the artefacts are received back from the participant after the measurement. If the measured drift of an artefact is greater than expected from the seasoning, it should be replaced by another seasoned LED of the same type for the next measurement round.

The “raw” LEDs used in this comparison are manufactured by Nichia Corporation.4 The selected models are listed in the following table with the specifications provided by the manufacturer (pdf-files included).

colour model initial characteristics in specifications

(forward current 20 mA, 25 ºC) specification sheets (file)

RED NSPR518S

forward voltage 2.2 V luminous intensity 1 cd dominant wavelength 625 nm spectral bandwidth 15 nm (FWHM) angular directivity 50º (FWHM)

Adobe Acrobat 7.0 Document

GREEN NSPG518S

forward voltage 3.5 V luminous intensity 2 cd dominant wavelength 525 nm spectral bandwidth 40 nm (FWHM) angular directivity 40º (FWHM)


BLUE NSPB518S

forward voltage 3.6 V luminous intensity 0.6 cd dominant wavelength 470 nm spectral bandwidth 30 nm (FWHM) angular directivity 40º (FWHM)


WHITE NSPW515BS

forward voltage 3.6 V luminous intensity 0.7 cd chromaticity near x = 0.31, y = 0.32 angular directivity 70º (FWHM)


The mechanical dimensions are the same for every raw LED as summarized below. The detailed drawing of the LEDs can be found in the specification sheets.

- lamp diameter: 5 mm (diffusion type, epoxy resin mold)

- lamp base diameter: 5.6 mm (LED’s outer diameter)

- lamp length (length of the lamp part with diameter ≤ 5 mm): 7.3 mm

4 More information on the LEDs available at http://www.nichia.co.jp/

http://www.nichia.co.jp/



- wire length (measured from backside of lamp): 20.3 mm for cathode, 22.3 mm for anode

- wire thickness: 0.5 mm

- wire distance: 2.5 mm

In the seasoning process, the relative luminous intensity and spectral distribution of each LED is recorded together with its junction temperature as a function of time for burning time of longer than 300 hours, while the ambient temperature is periodically varied from 18 ºC to 33 ºC. From the recorded data, the temperature dependence and the slow-varying drift characteristics of the LED’s photometric and colorimetric quantities can be separately determined.5 The pilot keeps and uses the measured data and characteristics of each artefact LED during the seasoning, first, to monitor and compensate the temperature effect of the measurands and, second, to control if the drift of the artefact LEDs occurred during the comparison is within the expected range. Note that the record of the junction voltage with the comparison measurands for each artefact LED is essential for this purpose.

Since the mechanical alignment of a LED is known as one of the most critical components affecting the measurement accuracy of averaged LED intensity, the pilot circulates, in addition to the 12 standard-type artefact LEDs, two samples of a specially-designed diffuser-type LED that shows a spatial emission distribution being not sensitive to the alignment. This diffuser-type artefact LED is constructed by putting a green LED (NSPG518S) into a cylinder-type cap with an opal diffuser, as shown in Fig. 1, and should provide a possibility to analyze the result of the comparison. Note, however, that this diffuser-type artefact LEDs are not used in the measurement of total luminous flux.

Fig. 1 Schematic drawing of a diffuser-type artefact LED.

One artefact set finally contains 14 artefact LEDs, and the pilot prepares and circulates 8 different sets for the 14 participants. Each participant receives and measures one among these artefact sets according to the timetable in Section 2.4. Each artefact set is identified with a serial number (set #1, set #2, etc.) and the 14 LEDs in one set is identified and positioned in a package box as shown in Fig. 2. Note that one artefact LED is uniquely identified in a form #N-X-M with three codes: (1) #N as artefact set number (N = 1, 2, …, 8), (2) X as LED colour and type code (X = R for red, G for green, B for blue, W for white, D for diffuser-type), and (3) M as sample serial number for each type (M = 1, 2, 3). As the individual LED could not be indicated by writing the full identification code on the LED due to the small size, only the sample number M of each LED is marked on the wires according to the colour code as shown in the right-hand part of Fig. 2.

5 Seongchong Park et al., Metrologia 43, 299 (2006); Proc. SPIE 6355, 63550G-1 (2006)

13.5 mm

8.3 mm

diffuser diameter 8.3 mm

[side view] [front view]



Fig. 2 Identification of individual LEDs in the box of one artefact set.

4. MEASUREMENT INSTRUCTIONS

4.1. AVERAGED LED INTENSITY (S3A)

The averaged LED intensity (unit: cd) of each artefact LED is to be measured in the standard condition B defined by CIE, as depicted in Fig. 3. 6 Either an illuminance meter or a spectroradiometer is used as the detector measuring the illuminance Ev for a circular area with size A = 100 mm2 at a distance d = 100 mm from the front tip of the LED. This is also valid for the diffuser-type LEDs with a flat front tip (see Fig. 1).

Fig. 3 Measurement condition for averaged LED intensity (CIE standard condition B).

The LED should be mounted so that the geometric axis of the LED is aligned to coincide with the normal of the reference plane of the detector head at the centre of the aperture area. The geometric axis of a LED is defined as the axis of rotational symmetry of the LED lamp cap,

6 Measurement of LEDs, CIE Technical Report 127-1997.

R-1 R-2 R-3

G-1 G-2 G-3

B-1 B-2 B-3

D-1 D-2

[wire marking]

- black for X-1

- red for X-2

- blue for X-3

(X = R/G/B/W/D)

W-1 W-2 W-3

Detector head

distance d

d = 100 mm ( = 0.01 sr)

circular aperture with size A =100 mm2


which, in general, does not coincide with the optical axis of the light emission, as depicted in Fig. 4. Each participant may use a different method to achieve the target alignment condition with high reproducibility. For instance, one can confirm the target alignment condition by visually inspect the LED from the detector head position to check the rotational symmetry of the cap, as shown in Fig. 5.

optical axis


Fig. 4 Definition of the geometric axis of a LED used for alignment to measure its averaged LED intensity.

Fig. 5 Inspection of alignment for the averaged LED intensity measurement by viewing the LED from the detector head position using a camera.

The LED should be mounted so that the backward emission, i.e. radiation emitted from the LED back surface to the direction of the connection wires, does not contribute to the detector signal. For this purpose, it is recommended to design the LED holder so that the backward emission is effectively scattered out of the measurement axis and blocked by a baffle.

The measurement should be performed by applying a constant forward current of 20 mA at an ambient temperature as close as 25 ºC for every artefact LED. In order to determine the junction temperature of the LED, the junction voltage between the anode and cathode should be measured in a 4-wire connection and recorded simultaneously with the value of averaged LED intensity, as shown in Fig. 6.

geometric axis LED front tip

[side view] [front view]

LED lamp cap

well-aligned slightly tilted


anode

cathode

current source

+

−

+

voltmeter

I = 20 mA

−

Fig. 6 Circuit diagram of the 4-wire connection used to measure the junction voltage of a LED while applying the forward current.

The measurement of averaged LED intensity and junction voltage should be performed after a warming-up time of longer than 5 minutes. The turn-on time and turn-off time of each measurement sequence should be recorded so that the total burning time of each artefact LED can be determined and reported.

The measurement should be repeated and reproduced so that its uncertainty can be evaluated with sufficient confidence.

4.2. TOTAL LUMINOUS FLUX (S3B)

The luminous flux integrated for the whole 4 direction (unit: lm) of each artefact LED is to be measured using either a goniophotometer or an integrating sphere. Note, however, that the two diffuser-type LEDs are excluded for the measurement of total luminous flux.

The LED should be mounted so that the contribution of the backward emission is properly included in the total luminous flux. For this purpose, it is recommended to mount the LED back surface as far as possible from the holder and to minimize the near-field absorption in the holder.

The measurement should be performed by applying a constant forward current of 20 mA at an ambient temperature as close as 25 ºC for every artefact LED. In order to determine the junction temperature of the LED, the junction voltage between the anode and cathode should be measured in a 4-wire connection and recorded simultaneously with the value of total luminous flux, as shown in Fig. 6.

The measurement of total luminous flux and junction voltage should be performed after a warming-up time of more than 5 minutes. The turn-on time and turn-off time of each measurement sequence should be recorded so that the total burning time of each artefact LED can be determined and reported.





4.3. EMITTED COLOUR (S3C)

The chromaticity coordinate CIE1931 (x,y) of the emitted colour of each artefact LED is to be determined by measuring the spectral distribution in the geometric condition of averaged LED intensity as shown in Fig. 3.7

The measurement should be performed by applying a constant forward current of 20 mA at an ambient temperature as close as 25 ºC for every artefact LED. In order to determine the junction temperature of the LED, the junction voltage between the anode and cathode should be measured in a 4-wire connection and recorded simultaneously with chromaticity coordinate, as shown in Fig. 6.

The measurement of chromaticity coordinate and junction voltage should be performed after a warming-up time of more than 5 minutes. The turn-on time and turn-off time of each measurement sequence should be recorded so that the total burning time of each artefact LED can be determined and reported.


5. REPORTING OF RESULTS AND UNCERTAINTIES

5.1. AVERAGED LED INTENSITY (S3A)

The measurement results should be reported to the pilot via email and fax by using the form APPENDIX 2: RESULT REPORT OF AVERAGED LED INTENSITY (S3A) immediately after the measurement is finished.

In addition to the result report, the participant is requested to provide the pilot a technical report containing the information listed in the following. This free-form report should be sent to the pilot via email as a Microsoft Word file within one month after the completion of measurement.

- Measurement setup and instruments used

- Mounting and alignment method, including a picture of the LED holder

- Traceability of measurement

- Detailed uncertainty budget for averaged LED intensity including values, evaluation type, probability distribution, degree of freedom, and sensitivity coefficient of each uncertainty component

- Detailed uncertainty budget for junction voltage including values, evaluation type, probability distribution, degree of freedom, and sensitivity coefficient of each uncertainty component

In the uncertainty budgets of the technical report, the participant should state whether and how an uncertainty component is artefact-dependent.

The pilot requests the participants to explicitly include the following uncertainty components in the uncertainty budgets to analyze the critical contributions:

- Component due to axis alignment. Note that the sensitivity to both angular (tilting) and translational (centring) misalignment should be separately considered.

7 This corresponds to a solid angle of 0.01 sr with a detector aperture size of 100 mm2. In case, however, that the aperture size of the instrument cannot be 100 mm2, the emitted colour should be measured for a solid angle of 0.01 sr at an appropriate distance, and the uncertainty budget should include components due to the different geometric condition.



- Component due to current feeding accuracy.

- Component due to stray light in the optical bench. Note that the backward emission of the LED scattered from the LED holder/mount can also contribute to the stray light.

- Component due to spectral mismatch correction, when a filter-type illuminance meter is used. Note that the spectral quantities used for spectral mismatch correction can be strongly correlated.

- For junction voltage: component due to position of junction.8

5.2. TOTAL LUMINOUS FLUX (S3B)

The measurement results should be reported to the pilot via email and fax by using the form APPENDIX 3: RESULT REPORT OF TOTAL LUMINOUS FLUX (S3B) immediately after the measurement is finished.



- Mounting and alignment method, including a picture of the LED holder


- Detailed uncertainty budget for total luminous flux including values, evaluation type, probability distribution, degree of freedom, and sensitivity coefficient of each uncertainty component



The pilot requests the participants to explicitly include the following uncertainty components in the uncertainty budget to analyze the critical contributions:

- Component due to near-field absorption of backward emission


- Component due to stray light, when a goniophotometer is used.

- Component due to spectral mismatch correction, when a filter-type illuminance meter is used. Note that the spectral quantities used for spectral mismatch correction can be strongly correlated.

- Component due to spatial correction, when an integrating sphere is used.

- For junction voltage: component due to position of junction.

5.3. EMITTED COLOUR (S3C)

The measurement results should be reported to the pilot via email and fax by using the form APPENDIX 4: RESULT REPORT OF EMITTED COLOUR (S3C) immediately after the measurement is finished.

8 That means an uncertainty component due to the different distance from the LED junction to the voltage measurement point.






- Detailed uncertainty budget for chromacitycoordinates including values, evaluation type, probability distribution, degree of freedom, and sensitivity coefficient of each uncertainty component



The pilot requests the participants to explicitly include the following uncertainty components in the uncertainty budget to analyze the critical contributions:

- Component due to axis alignment. Note that the sensitivity to both angular (tilting) and translational (centring) misalignment should be separately considered.


- Component in calculating the chromaticity coordinate from the measured spectral distribution. Note that the spectral quantities used for calculation can be strongly correlated.

- For junction voltage: component due to position of junction.

6. PREPARATION OF COMPARISON REPORT

After the measurement schedule of every participant is completed, the pilot prepares the report of the comparisons according to the guidelines by CCPR.9

Since three comparisons are performed together by using one artefact LED set, three reports are to be separately prepared.

Before starting the Pre-Draft A process, the pilot will re-confirm its reception of the artefact sets, the measurement results, and the technical reports from every participant. If any result or report is missing until this time, the pilot will announce a deadline for re-submission. After this deadline, the pilot proceeds the report preparation only with the data submitted so far.

9 Guidelines for CCPR Comparison Report Preparation, Rev. 1 of March 2006. Available at http://www.bipm.org/utils/en/pdf/ccpr_guidelines.pdf

http://www.bipm.org/utils/en/pdf/ccpr_guidelines.pdf



APPENDIX 1: INSPECTION REPORT ON RECEIPT OF ARTEFACTS

Has the artefact set package been opened during transit? (e.g. by Customs) …… Y / N

If Yes, please give details.

Is there any damage to the package box? …… Y / N

If Yes, please give details.

Are the 14 artefact LEDs inside the package box complete and properly fixed into the conductive

matting? …… Y / N

If No, please give details.

Are there any visible signs of damage to the artefact LEDs? …… Y / N

If Yes, please give details (e.g. scratches or contaminations on the lamp, bending of wires, etc).

Is the LED identification sheet prepared by the pilot found in the package? …… Y / N

Laboratory: ………………………………………………………………………………………………

Date: …………………………………………… Signature: ………………………………..……

APPENDIX 2: RESULT REPORT OF AVERAGED LED INTENSITY (S3A)

Artefact set number:

Measurement dates: from to

Laboratory condition: temperature ( ± ) ºC , relative humidity ( ± ) %

LED

measurement value of

averaged LED intensity (cd)

expanded uncertainty* of averaged LED intensity (cd)


junction voltage (V)

expanded uncertainty* of junction voltage

(V)

total burning time (min)

R-1

R-2

R-3

G-1

G-2

G-3

B-1

B-2

B-3

W-1

W-2

W-3

D-1

D-2

* estimated for a 95 % confidence level (normally with a coverage factor k = 2)

Laboratory: ………………………………………………………………………………………………

Date: …………………………………………… Signature: ………………………………..……

APPENDIX 3: RESULT REPORT OF TOTAL LUMINOUS FLUX (S3B)




LED

measurement value of total luminous flux

(lm)

expanded uncertainty* of total luminous

flux (lm)



expanded uncertainty* of junction voltage

(V)

total burning time (min)

R-1

R-2

R-3

G-1

G-2

G-3

B-1

B-2

B-3

W-1

W-2

W-3


Laboratory: ………………………………………………………………………………………………

Date: …………………………………………… Signature: ………………………………..……

APPENDIX 4: RESULT REPORT OF EMITTED COLOUR (S3C)




measurement value of chromaticity

coordinate

expanded uncertainty* of

chromaticity coordinate LED

x y x y



expanded uncertainty* of junction voltage (V)

total burning

time (min)

R-1

R-2

R-3

G-1

G-2

G-3

B-1

B-2

B-3

W-1

W-2

W-3

D-1

D-2


Laboratory: ………………………………………………………………………………………………

Date: …………………………………………… Signature: ………………………………..……

Summary of Comments in Review of Relative Data

VSL

Mail on Dec 14, 2009

Looking to the data of VSL we see a big instability for some of the LEDs. Can you tell me how

you are going to deal with this and what the effect will be for the KCRV values or final

presentation of the results?

Response of KRISS on Dec 22, 2009

I think the stability for the LEDs used for VSL is not so bad (all below 1 % drift). I propose to

average the LEDs of the same type (three of red, three of green, etc.) and take the instability as an

uncertainty component of the difference from the reference value. (There will be no KCRV and

DoE because these are supplementary comparisons.)

Of course, we will exclude particular LEDs which show bad stability based on the opinion and

agreement of the participant.

Mail on March 17, 2010

Looking to the remarks of the temperature correction data we are wondering if the inconsistence

for some of the data has to do with the measurement of the junction voltage. As I can remember

there was a relative large variation in voltage over the legs of the LEDs. So in some cases

depending on the position of the junction measurement this can affect the correction for

temperature. Of course one needs to take this variation into the uncertainty for the voltage

measurement at the junction but maybe some of the inconsistencies can be explained looking to

the uncertainty for junction measurements versus temperature correction and the variation of

junction voltage over the legs of the LEDs.

Response of KRISS on March 24, 2010

It is true that there is a change of junction voltage when the measurement position of the LED

electrodes changes. We have noticed this at the stage of the artefact preparation, and therefore

arranged that this variation due to the junction position should be checked and reported by each

participant as an uncertainty component of junction voltage.

Because we have all the sensitivity data of photometric quantity to junction voltage for each

artefact LED, we can analyze the inconsistency caused by the inaccurate measurement of junction

voltage. We will surely include this in the result report. From our experience, however, the

uncertainty of photometric quantities propagated from the uncertainty of junction voltage

measurement, including the junction position variation, was much lower than 0.5 %, which is the

principal accuracy limit of the temperature correction method via junction voltage.

METAS

Mail on Dec 15, 2009

I have no special observation.

Mail on Feb 10, 2010

I have no special comments in respect to our relative data except that applying the temperature

correction will increase non-consistency of our data. I’ve done this analysis for all participants (see

enclosed excel-file) and it is interesting to see that only for few laboratories the consistency

increases.

Response of KRISS on Feb 12, 2010

You showed that the consistency decreases after the temperature correction, i.e. the standard

deviation of all the relative data for a participant increases. I think this is reasonable because the

process of temperature correction contains also the uncertainty, which is the limitation of the

theoretical model for temperature correction via junction voltage. We estimate this uncertainty to

be less than 0.5 % (see our publication in Metrologia, 43, 299, 2006). Therefore, we expect that

the application of temperature correction unavoidably causes a slight decrease of the consistency

of the relative data. Based on your calculation, the standard deviations of the relative data lie, for

most of the participants, between 0.5 % and 1 % without temperature correction, but the

(absolute) change of them due to temperature correction remains much below 0.5 %. From this,

we can confirm the accuracy of the temperature correction method.

In addition, we could also see the validity of temperature correction in the change of the absolute

data (not published yet) that the consistency between the pilot and the participants clearly

increases after temperature correction.

MKEH

Mail on Jan 20, 2010

After the overview of the MKEH relative data of the comparisons APMP-S3a (averaged LED

intensity) we have two remarks:

The LED G1, which was strongly different, died after the MKEH measurement. So this diode does

not have remeasured value. It might be damaged before the MKEH measurement. We ask for

remove the data of this diode.

The LED B3, which was different as well, died after the MKEH measurement. So this diode does

not have remeasured value. It might be damaged before the MKEH measurement as well. We ask

for remove the data of this diode.


We accept the data you have sent. (with respect to S3c)

MIKES


Could we remove the W-1 LED from the both comparisons?

NIST


We think that the LED set measured by NIST was not so bad if KRISS' measurement results for R1

and R2 were reliable. So we want to confirm that the differences (shown in your relative data)

between the measurement results of R1 and R2 are acceptable to us.

NMIJ

Mail on Dec 28, 2009 (not delivered in time)

By the way, it is the matter of review of relative data, in order to estimate whether it is drift of

LEDs, I would like to know the information of total burning time of our artifact(set #4) including

measurement burning time in KRISS.

I know the burning time in our measurement, but I don't know it in KRISS.

In addition, I would like to know about the seasoning result of our artifact.

Unless KRISS clarifies these information, it is very difficult to judge against our result of relative

data whether it is a drift of LEDs or some issue.


I would like to request to remove the result of B-1, B-3, W-1 from our APMP.PR-S3a results. In

addition, I would like also to request to remove the result of W-1 from our APMP.PR-S3b results.

Because, I think that the change of those LED result is large.

ASTAR


Thanks for the relative data. We have reviewed the data. The data looks in order and we have not

further comments for the relative data of all three comparisons.

Summary of Comments in Review of Uncertainty Budgets

Part 1. General Comments and Revisions

INM (Romania)

Mail on April 02, 2010

As far as the INM reports are concerned, the uncertainty budgets for Green, Blue, White and

Diffuse LEDs were not included in the APMP PR S 3a and APMP PR S3b reports just because they

are very close to our uncertainty budgets for the Red LEDs so we thought not necessary to repeat

the almost exactly same figures. But do you think this is necessary or should we merely mention

this in the reports? Anyway, in order to comply I`ll revise and sent you our reports today, provided

it`s not already too late.

Here attached are our revised reports for APMP PR S3a and APMP PR S3b comparisons, incliding

the uncertainty budgets for all tipes of LEDs.

Please notice that changes only concerned the spectral correction factors for the various LEDs and

while the combined standard uncertainties were of about 5.5 %, the various spectral correction

factors induced quite small changes (less than +/- 0,5 %) in the combined uncertainties values.

That`s why, initially we only reported the uncertainty budgets for the red LEDs.

Response of KRISS on April 12, 2010

I have properly received your two documents including the uncertainty budgets for all color-types

of LEDs. The formats you sent me are ok.

Because your revision deals only with an addition of information, I see no problem to accept your

revision for the report. I will wait for a while for other revisions or corrections, and distribute the

revised files then.

METAS

Mail on April 15, 2010

Please find enclosed an update of our description of the uncertainty budget of the chromaticity

coordinates. I’m sorry to have it sent after your deadline. In the updated version I stated explicitly

uncertainty budgets for the 4 types of LEDs. It’s just to give more information, no value has been

changed.

I also would like to recall our worries in respect to the correlation of chromaticity coordinates (see

the attached file).

APMP.PR-S3 Correlation of chro

Response of KRISS on April 15, 2010

I have received your files well. I will revise the uncertainty review document for S3c and distribute

it again. (But I will wait for a while to collect the revisions also from other participants.)

I think that your suggestion of reporting the correlation can be discussed open. Do you agree to

forward your document directly to all the participants to ask for their opinions?

A*STAR

Mail on June 21, 2010

We found not error in the three files containing technical information and uncertainty budgets.

However we added a paragraph in section 10.3 (in red colour text) of the “uncertainty

budgets_S3b” to mention the absorption correction in integrating sphere calibration and

measurement. The modified file is attached.

All Participants (open discussion)

Mail from KRISS on May 10, 2010

I have a comment which is sent from METAS to all the participants. Peter agreed to discuss this

issue openly.

This deals with a suggestion that, for the uncertainty budgets of chromaticity coordinates (x, y) for

APMP-S3c, the correlation between u(x) and u(y) should be considered by submitting the

correlation coefficient u(x,y)/u(x)u(y). Please see also the attached letter from Peter.

I personally think that it is meaningful to compare also the correlation coefficients among the

participants. However, it may be difficult at this stage to make the report of the correlation

mandatory because we did not mention this in the technical protocol. What we can do instead is

to encourage the participants to voluntarily report the correlation analysis as far as possible. If we

have many volunteers, we can include this part in the comparison report. If we have only a few

participants reporting the correlation, we can prepare this issue to an extra publication.

I would like to ask first who can submit the results of the correlation coefficients for the

chromaticity coordinates as supplementary to the uncertainty budget report. (METAS surely, and

KRISS can also do it.)

Mail from PTB on May 12, 2010

Correlation (x,y): If needed we can add the correlation of (x,y). Please let us know what is the

decision.

Mail from A*STAR on June 21, 2010

Regarding the issue our response is that we cannot submit the correlation coefficients for the uncertainty of the chromaticity coordinates.

Communication from KRISS on June 21, 2010

Typical values of correlation coefficient r(x,y) = u(x,y)/u(x)u(y) are -0.69 for RED, +0.41 for GREEN, -

0.86 for BLUE, and +0.96 for WHITE. The values do not change much as the artifact set changes.

Part 2. Questions and Answers

KRISS

Question to KRISS on May 10, 2010

-S3a average LED intensity

What are the uncertainty of the axis alignment (angular, translational) and distance: expressed in °

and mm?

-S3c, chromaticity coordinates, red LED

For the red LED the main contribution of the uncertainty is given by the spectral straylight. Has

the data been corrected for straylight? Why the contribution for red is much large then for the

others (red x: 0.00148, blue x: 0.00032) and why x and y are so different (usually there is full

correlation for the chromaticity coordinates for red LEDs).

-S3c, chromaticity coordinates, wavelength

For the other LED’s the main contribution of the uncertainty is given by the wavelength accuracy.

It would be useful to know the absolute uncertainty of the wavelength scale (expressed in nm).

Have there been some spectral correlations taking to account in the analysis?

Answers from KRISS on June 21, 2010 -S3a average LED intensity What are the uncertainty of the axis alignment (angular, translational) and distance: expressed in ° and mm? Response: The standard uncertainty of angular axis alignment, translation axis alignment, and distance setting is 0.82°, 0.41 mm and 0.25 mm, respectively. For translational axis alignment, the uncertainty contribution has been revised such as 0.2 % for red (Other else remain the same). -S3c, chromaticity coordinates, red LED For the red LED the main contribution of the uncertainty is given by the spectral stray light. Has the data been corrected for stray light? Why the contribution for red is much large then for the others (red x: 0.00148, blue x: 0.00032) and why x and y are so different (usually there is full correlation for the chromaticity coordinates for red LEDs). Response: The spectral stray light of spectral data is not corrected. We estimated the spectral stray light as an uncertainty based on the spectrograph response under He-Ne laser illumination. Most of stray light readout is distributed around the laser wavelength except the in-band region, which means that the spectral stray light has a similar spectral distribution with the input illumination. Thus, the contribution of the stray spectrum on chromaticity is more or less proportional to that of the input illumination. While the stray spectrum gives more contribution to x in case of a red LED, the stray spectrum of a green LED and a blue LED give more contribution to y and z, respectively. In our calculation, the correlation coefficient r(x, y) of a red LED turned out to -0.69. -S3c, chromaticity coordinates, wavelength For the other LED’s the main contribution of the uncertainty is given by the wavelength accuracy. It would be useful to know the absolute uncertainty of the wavelength scale (expressed in nm). Have there been some spectral correlations taking to account in the analysis? Response: The standard uncertainty of wavelength scale is (0.45 ~ 0.48) nm depending on wavelength. Of the uncertainty,

0.2 nm is a global wavelength offset, which mainly contributes on the chromaticity uncertainty. The spectral correlations are taken account in.

MIKES

Question to MIKES on May 10, 2010


The uncertainty is by far dominated by the repeatability of the measurement. What is the origin of

this? Were measurement noisy? In the case of the diffuse type LED this contribution is smaller

than for the other type. Is it related to the geometry of the source? Is it really repeatability and

not reproducibility (i.e. were the LED realigned?)?

-S3b, luminous flux

The most important contribution (expect for the blue LED) originates from the near field

absorption (1%, with rectangular distribution!). How this value has been determined?

-S3c, chromaticity coordinates, white LED, angular alignment

The uncertainties of the chromaticity coordinates of the white LED are much higher than the other

coloured LEDs (except to the one with diffuser). The main contribution seems to be originated for

the angular alignment, although the sensitivity coefficient of that quantity seems to be the similar.

What is the origin of this?

-S3c, chromaticity coordinates, green LED

The uncertainty of the green LED with diffuser is dominated by the noise. How this contribution

has been determined as it as of Type B with rectangular probability? Usually noise contributions

are included in the repeatability of the measurement (Type A).

Answers from MIKES on May 31, 2010 > /-S3a average LED intensity / > > The uncertainty is by far dominated by the repeatability of the > measurement. What is the origin of this? Were measurement noisy? In > the case of the diffuse type LED this contribution is smaller than for > the other type. Is it related to the geometry of the source? Is it > really repeatability and not reproducibility (i.e. were the LED > realigned?)? > Answer: The uncertainty of repeatability originates mainly from the alignment accuracy of the measurement setup, i.e. the realignment of the LED before each repeat measurement. For the diffuser type of LEDs, the uncertainty due to the alignment was not found as sensitive as for the other type of LEDs. This could be partly explained by the optical properties of the measured LEDs. The LEDs without diffusing output may have nonuniform structure in the light output. > > /-S3b, luminous flux/ > > The most important contribution (expect for the blue LED) originates

> from the near field absorption (1%, with rectangular distribution!). > How this value has been determined? > Answer: The uncertainty of the near field absorption (type B) was estimated by considering the geometry and materials used in the LED holder and the amount of light emitted backward by the measured LEDs. > /-S3c, chromaticity coordinates, white LED, angular alignment/ > > The uncertainties of the chromaticity coordinates of the white LED are > much higher than the other coloured LEDs (except to the one with > diffuser). The main contribution seems to be originated for the > angular alignment, although the sensitivity coefficient of that > quantity seems to be the similar. What is the origin of this? > Answer: In the case of white LEDs, the spectral output may change as a function of angle of observation due to the phosphor coating. Therefore they are more sensitive to the alignment than the other type of LEDs. > /-S3c, chromaticity coordinates, green LED/ > > The uncertainty of the green LED with diffuser is dominated by the > noise. How this contribution has been determined as it as of Type B > with rectangular probability? Usually noise contributions are included > in the repeatability of the measurement (Type A). > Answer: The uncertainty of the diffuser type of LED was obtained by calculating the color coordinates for the original measurement data and for another data, in which the noise of the low signal values was replaced with extrapolated modelled values of the measured LED spectrum.

CMS-ITRI

Question to CMS-ITRI on May 10, 2010

-S3a average LED intensity, LED spatial light distribution

Why the quantity “LED spatial light distribution” is the same for all type of LEDs even the spatial

distribution is very different for the different LEDs (in particular the one with diffuser to the one

without diffuser)

-S3a average LED intensity, red LED,

The uncertainty of the spectral mismatch correction seems to be exceptionally small for the red

LED in respect to the other colours. What is the f1’ of the photometer?


The uncertainty of the “x” - chromaticity coordinate of the red LED is dominated by two

contributions (repeatability :0.0015 and mechanical alignment: 0.0014). Why the combined

uncertainty is only 0.0014?

-S3c, chromaticity coordinates, mechanical alignment

why the uncertainty contribution due to mechanical alignment is the same for all type of LEDs? Is

there an evidence that a misalignment causes the same amount of shift in colour coordinates?

-S3c, chromaticity coordinates, green LED and green LED with diffusor

why the contribution of the wavelength shift of the green LED with diffusor is much higher than

the green LED without diffuser (more than 20x), the spectral distribution of both type of LEDs

being very similar?

PTB

Question to PTB on May 10, 2010


It would be interesting to know the area of the sensitive surface of the photometer head, and in

the case that it is different to 100mm2 how that results were corrected.

-S3a average LED intensity, Correction for LED angular align,

Why the uncertainty due to the correction for angular alignment of the blue LED (0.57%) is much

larger than for the other LEDs (green: 0.11%) although the spatial distribution of is very similar?

-S3b, luminous flux, Integrated photocurrent, solid angle weighted

The most important contribution of uncertainty is originated from the quantity “Integrated

photocurrent, solid angle weighted”. It would be useful to have further information about this

quantity (i.e. eventl. citation). How it has been determined?


The uncertainty of the chromaticity coordinates of the red LED is mainly given by the spectral

bandpass correction and the straylight correction of the spectrometer. There is however no

information about the amount of correction that has been applied and the spectrometer used for

the measurement(bandpass, wavelength accuracy, level of straylight,…)

-There is no information about the uncertainty contributions (input quantities and their

uncertainties) used in the Monte Carlo simulation.

Answers from PTB on May 12, 2010 Here are the answers of PTB concerning some questions of a participant: -S3a average LED intensity It would be interesting to know the area of the sensitive surface of the photometer head, and in the case that it is different to 100mm2 how that results were corrected. PTB: According to CIE Pub. 127 in all cases (S3a, S3b and S3c) the sensitive area of photometers or spectrometer input optics were 100 mm2. So no corrections for a different sensitive area were applied.

-S3a average LED intensity, Correction for LED angular align, Why the uncertainty due to the correction for angular alignment of the blue LED (0.57%) is much larger than for the other LEDs (green: 0.11%) although the spatial distribution of is very similar? PTB: From goniophotometric luminous flux measurements we know the spatial distribution of all LEDs. Especially the spatial distribution of green and blue LEDs are not similar in the interesting range of approx. 0° < ϑ < 2.5° ! Please, see figures below (on the left: example of green LED, on the right: example of blue LED). We describe the spatial distribution with cos[ϑ]g. In case of the green LED we found g=8.9 and in case of the blue LED we found g=39. Please, compare blue plots.

Now we are able the estimate the uncertainty contribution of angular alignment and translational alignment of the LED for luminous intensity measurements by help of a mathematical simulation. The figure below on the left shows a LED aligned in front of a photometer. The angular and aerial responsivity oft he photometer is simulated by a number of hexagons. For our estimations we used a larger number of smaller hexagons (see figure on the right). Based on the knowledge of uncertainty for angular alignment and translational alignment we are able to calculate the estimated uncertainty contributions.

Total area = 100 mm2

5

10

15

20

2530

3540

4550

5560

6570758085

0.2 0.4 0.6 0.8 1.0

0.2

0.4

0.6

0.8

1.0

LEDG101.evk, redmeasured datablueFit Cosg with g8.92036, dashedCos

5

10

15

20

25

30

3540

4550

5560

6570758085

0.2 0.4 0.6 0.8 1.0

0.2

0.4

0.6

0.8

1.0

LEDB101.evk, redmeasured datablueFit Cosg with g39.0866, dashedCos

-S3b, luminous flux, Integrated photocurrent, solid angle weighted The most important contribution of uncertainty is originated from the quantity “Integrated photocurrent, solid angle weighted”. It would be useful to have further information about this quantity (i.e. eventl. citation). How it has been determined? PTB: The figure below on the left shows the goniophotometric measurement of the LEDs in principle. The averaged zonal illuminance is derived from the measured averaged zonal photocurrent ( )ϑj . The figure on the right shows it as a function of the angle ϑ .

Since the determination of this averaged zonal photocurrent is a complex system which consists of several dc motor drives, a current/voltage converter and a digital voltmeter a correction factor czone

was introduced. The averaged value of czone = 1, but to consider the uncertainty caused by an

unsharp start and stop angle ( EndStart ϕϕ , ) it is necessary and defined as follows :

πϕϕ

2EndStart

zonec −=

Now we can start the MC-simulation: Repeat the following with normal distributed varied KVVEndStart j,,, ϑϕϕ

( ) ( ) ( ) ϑϑϑϑϑπ

ϑ

d1Sin0

zone ⋅+⋅+⋅+⋅= ∫=

KVVV jjcX

and in principle from X you will get the so called “Integrated photocurrent, solid angle weighted”

( )Xj Mean=Φ with ( ) ( )XU j viationStandardde=Φ .

-S3c, chromaticity coordinates, red LED The uncertainty of the chromaticity coordinates of the red LED is mainly given by the spectral bandpass correction and the straylight correction of the spectrometer. There is however no information about the amount of correction that has been applied and the spectrometer used for the measurement(bandpass, wavelength accuracy, level of straylight,…) -There is no information about the uncertainty contributions (input quantities and their uncertainties) used in the Monte Carlo simulation.

0.5 1.0 1.5 2.0 2.5 3.0Radian

2.107

4.107

6.107

8.107

Photocurrent AMeasured averaged zonal photocurrent as function of zone angle

PTB: As you can see in our uncertainty budgets the correction values of bandpass and spectrometer straylight is always 0. That means no correction was applied. But we estimated the uncertainty contributions by help of some MC simulations. The following figure shows an example result of a similar simulation.

Varied input parameters of the simulation were mainly spectrometer response data during measurement the LED and the halogen lamp used for sensitive calibration with an uncertainty of their spectral irradiance expressed as an uncertainty of the distribution temperature of a planckian radiator (approx. 10 K), an estimated straylight correction matrix (similar to the figure below, which is the real strayight correction matrix of the used array spectrometer from knowledge we have today ), an assumed triangle-shaped bandpass (halfwidth approx. 3nm ), the function between channel-no and wavelengths with a wavelength uncertainty of approx. 0.8nm, etc.

NMIJ

Question to NMIJ on May 10, 2010

0.6998 0.7002 0.7004 0.7006x

0.2988

0.2992

0.2994

0.2996

y

500 500

1000 1000

-S3a average LED intensity, illuminance responsivity

It is very unusual to see a rectangular probability function for the uncertainty of the illuminance

responsivity. Usually this value is either taken from a calibration certificate or determined by

another measurement (traceable to the radiometric scale). In both cases the distribution is

typically Gaussian type. Furthermore the uncertainty seems to be rather large (much larger than

declared CMC values in the KCDB with k=2…).

-S3b, luminous flux, Angular resolution, etc.

Why the contribution of the quantity called « angular resolution, etc » is much larger for the red

LED than for the others (red: 0.91%, green: 0.28%) even if the angular distribution of the LEDs are

very similar (the green LED is even narrower than the red)?


It would be useful to report in the uncertainty budget of the chromaticity coordinates of the red

LED one additional digit (in the column “contribution”). The GUM recommands to report

uncertainty with two significant digit.

Answers from NMIJ on June 02, 2010

I am submitting two file (Reply to Question and Revised verification report).

Revised points in verification file are edited the Word files with red characters. New verification

report is revised according to the comment (Uncertainty Component name, Deg. of freedom, add

to new figure etc,).

But, there is no modify of the combined standard uncertainty .

Q1:-S3a average LED intensity, illuminance responsivity

It is very unusual to see a rectangular probability function for the uncertainty of the illuminance

responsivity. Usually this value is either taken from a calibration certificate or determined by

another measurement (traceable to the radiometric scale). In both cases the distribution is

typically Gaussian type. Furthermore the uncertainty seems to be rather large (much larger than

declared CMC values in the KCDB with k=2…).

Re1:

Thank you for good advice. I made a mistake about probability function of illuminance

responsivity. I would like to correct about probability function and freedom of it.

Next, I would like to explain about uncertainty of illuminance responsivity. In order to consider a

near-field effects which CIE 127:2007 (5.4 P17) described, illuminance responsivity of our

photometer for LED measurement is calibrated by luminous intensity standard lamp at far-field

condition, and then it is calibrated by an integrating sphere source(operated at 2856K) at the

distance corresponding to CIE condition B. Our uncertainties of illuminance responsivity include

uncertainty of near-filed effect. Therefore it becomes larger than uncertainty of CMC.

Q1:-S3b, luminous flux, Angular resolution, etc.

Why the contribution of the quantity called ≪ angular resolution, etc ≫is much larger for the

red LED than for the others (red: 0.91%, green: 0.28%) even if the angular distribution of the LEDs

are very similar (the green LED is even narrower than the red)?

Re2:

Firstly, I would like to change the contribution of the quantity's name from "angular resolution,

etc" to "measurement angle step and angular resolution". I send the modified uncertainty budget.

Sorry, my expressions confuse.

Fig1 indicate an angular distribution of red and green LED. The angular distributions of red LED

is not smoother than it of green LED .I think the angular distribution of the red LED is not the

same as others. Red LED have an irregular angular distribution. For this reasons, the uncertainty of

"measurement step and angular resolution" on red LED became larger than green LED in our

budget.

Fig1: angular distribution

Q3:-S3c, chromaticity coordinates, red LED

It would be useful to report in the uncertainty budget of the chromaticity coordinates of the red

LED one additional digit (in the column "contribution”). The GUM recommends reporting

uncertainty with two significant digits.

A3:

Thank you for good advice. I send the modified uncertainty budget. I add one additional digit to

uncertainty values of contribution, but the combined standard uncertainty isn't changed.

CENAM

Question to CENAM on May 10, 2010

-S3a, average LED intensity, Spectral mismatch correction

Why the uncertainty of the spectral mismatch correction is almost constant for all type of LED’s?

Usually the uncertainty is much lower for white LEDs than for blue LEDs?

-S3b, luminous flux, Standard lamps spectral mismatch correction

The quantity “Standard lamps spectral mismatch correction” seems to be rather large. What kind

of standard lamps was used (usually incandescent lamps are used which are not too far from CIE

illuminant A)? What is the f1’ value of the photometer? What is the estimated relative spectral

throughput of the sphere (i.e. how “flat” is the painting)?

-S3c, chromaticity coordinates

What is the quantity “Propagation from spectral distribution measurement”? Why is it constant for

all colours (15.66% , 13.96%) and why the sensitivity coefficient so small 0.00002 (% per %?) and

constant?


uncertainty of chromaticity coordinates are usually reported as absolute value as chromaticity

coordinates are highly non-linear quantities.


in the case of the red LED the absolute uncertainty are as following: ux= 0.006 and uy=0.0006

(hence a factor of 10 between both coordinates). Is there are an explication for this behavior?

Usually the chromaticity of red LED are fully (negative) correlated resulting in similar uncertainties

in x an y?

Answers from CENAM on May 20, 2010

Please find below the answers to the questions done for CENAM. -S3a, average LED intensity, Spectral mismatch correction

Why the uncertainty of the spectral mismatch correction is almost constant for all type of LED’s?

Usually the uncertainty is much lower for white LEDs than for blue LEDs?

RE: Unfortunately the resolution of the spectrorradiometer we used to measure the LEDs spectra was very bad; thus causing this component to be dominant over the other, and making the spectral mismatch uncertainties to look almost constant.

-S3b, luminous flux, Standard lamps spectral mismatch correction

The quantity “Standard lamps spectral mismatch correction” seems to be rather large. What kind

of standard lamps was used (usually incandescent lamps are used which are not too far from CIE

illuminant A)? What is the f1’ value of the photometer? What is the estimated relative spectral

throughput of the sphere (i.e. how “flat” is the painting)?

RE: Unfortunately the resolution of the spectrorradiometer we used to measure the spectra was very bad; thus causing this spectral mismatch corrections to be very large. We used incandescent lamps operated as CIE Standard illuminant A. The f1=13,36. The estimated relative spectral throughput of the sphere is fairly plain.


What is the quantity “Propagation from spectral distribution measurement”? Why is it constant for

all colours (15.66% , 13.96%) and why the sensitivity coefficient so small 0.00002 (% per %?) and

constant?

RE: We call “Propagation from spectral distribution measurement” to the uncertainty component due to the calculation method from the spectral irradiance lectures. This is constant because we used the average value obtained from the standard lamps used. This also produced such a sensitivity coefficient values, and almost constants.




RE: We reported our final results for those values as absolute to the pilot laboratory; however, according to the final report format, we were requested to report those as relative, and we did it as well.


in the case of the red LED the absolute uncertainty are as following: ux= 0.006 and uy=0.0006

(hence a factor of 10 between both coordinates). Is there are an explication for this behavior?

Usually the chromaticity of red LED are fully (negative) correlated resulting in similar uncertainties

in x an y?

RE: We do not find such a values as they are mentioned. We have double-checked the results we send to the pilot laboratory; as well as those the pilot laboratory sent back for revision; and we found they are ok, within the same magnitude order. Would you please let us know where you found those?

LNE

Question to LNE on May 10, 2010




NMC-A*STAR

Question to A*STAR on May 10, 2010

-S3b, luminous flux,

A*STAR has not used an auxiliary lamp for compensating changes of the integration properties of

the sphere resulting in the different configuration between the LED measurement and the sphere

calibration. Has this influence being estimated?




Answers from A*STAR on June 21, 2010

Question for: -S3b, luminous flux, A*STAR has not used an auxiliary lamp for compensating changes of the integration properties of the sphere resulting in the different configuration between the LED measurement and the sphere calibration. Has this influence being estimated? Reply: The one-meter integrating sphere that we used for LED flux measurement do have a tungsten auxiliary lamp. The absorption corrections were carried out over the whole wavelength range of 380 nm to 780 nm in 1 nm interval for both the LED measurement and the sphere calibration. An additional paragraph explaining this is added in section 10.3 of the uncertainty budgets_S3b. Please refer to the revised file attached. (Dong-Hoon, the revised file is actually attached in my last email to you so I didn’t repeat here) Question for: -S3c, chromaticity coordinates uncertainty of chromaticity coordinates are usually reported as absolute value as chromaticity coordinates are highly non-linear quantities. Reply: The uncertainty of chromaticity coordinates that we reported for the -S3c results are indeed in absolute values.

VSL

Question to VSL on May 10, 2010

-S3a, average LED intensity

What is the quantity “Non-uniformity of source”? Is this due to the non-coincidence of the optical

and mechanical axis? In Figure 11-6 of the report a measurement of the illuminance in function of

different (azimuthal) angles is shown. It is written that this is due to the non-coincidence of the

mechanical axis and the optical axis. However we believe that it is to a misalignment of the

photometer in respect to the rotation axis as illustrated below.

-S3b, luminous flux, Near-field absorption of backward emission

The most important contribution to uncertainty is the quantity “Near-field absorption of backward

emission”. Has the flux also being corrected with this quantity, if yes what was the estimated ratio

from the backwards flux to the total flux?

-S3b, luminous flux

The goniophotometrical measurements were done at an angular increment of 5° (polar angle).

Has the uncertainty due to this rather large increment been estimated (The half angle of the

green LED is only 22°)?

Answers from VSL on May 10, 2010 -S3a, average LED intensity What is the quantity “Non-uniformity of source”? Is this due to the non-coincidence of the optical and mechanical axis? In Figure 11-6 of the report a measurement of the illuminance in function of different (azimuthal) angles is shown. It is written that this is due to the non-coincidence of the mechanical axis and the optical axis. However we believe that it is to a misalignment of the photometer in respect to the rotation axis as illustrated below. Answer VSL As the reference axis for alignment is not defined in the protocol, one needs to make a choice which axis is used for alignment (optical or mechanical). From our research we believe that the mechanical axis for alignment is the best choice for comparability of the measurement results. When using the mechanical axis as a reference axis you will need to check what this means with respect to the azimuthal angle direction in respect to the uncertainty. As measurements show (figure 11-6 of the report) one needs to take the non-coincidence of the mechanical and optical axis into account, again: if you are using the mechanical axis as reference. Please notice that when you align on optical axis you will introduce an angular shift between the optical axis of your LED and the rotation axis of your goniometer. This will also introduce an uncertainty. -S3b, luminous flux, Near-field absorption of backward emission The most important contribution to uncertainty is the quantity “Near-field absorption of backward emission”. Has the flux also being corrected with this quantity, if yes what was the estimated ratio from the backwards flux to the total flux?

Answer VSL The flux has not been corrected for the “Near-field absorption of backward emission”.

-S3b, luminous flux The goniophotometrical measurements were done at an angular increment of 5° (polar angle). Has

the uncertainty due to this rather large increment been estimated (The half angle of the green LED is only 22°)?

Answer VSL As the detector size of our photometer is 10 mm^2 one can calculate the smallest step size that is required to have overlap between the measurement points measuring at a distance of 100 mm. With a step size of 5º we still have an overlap from point to point. Next to this we have taken the green LED and measured ones in steps of 5º and ones in steps of 1º. The results showed that there is a small difference in respect to the total uncertainty between a step of 5º compared to a step size of 1º. We have taken the difference into the uncertainty component for the integration method.

NIST

Question to NIST on May 10, 2010


What is the estimated wavelength uncertainty of the spectrometer measurement (expressed in

nm)? Have there been some spectral correlations taking to account in the analysis?

-S3c, chromaticity coordinates, contributions due to alignment of the LED

Minor comment: It is very unusually that Type A has an infinite number of degree of freedom.

Either the contribution has been determined experimentally and then a statistics is used (Type A

with limited number of degrees of freedom) or a model was used (perhaps also based on

experimental results) to describe the specific input quantity (Type B with infinite number of

degrees of freedom).

MKEH

Question to MKEH on May 10, 2010

-S3a, average LED intensity

Several important contributions are missing: temperature, readout of the photometer (Type A).

Why the calibration accuracy has a rectangular distribution, usually it should be Gaussian

distributed.


Why the uncertainty is stated as a minimum value (>0.0004 and >0.0002). The uncertainty analysis

is used to determine the estimates of the output quantity and its uncertainty (for a given

confidence interval). If only a minimum value is stated either the uncertainty budget is incomplete

or the estimation of some of the contributions are believed to be too small (and should therefore

be adapted).

Answers from MKEH on June 1, 2010

1. In the luminous intensity error budget our main source of error comes from the detector

calibration. We do not have cryogenic radiometer we have Si selfcalibration as an absolute

method. In this case the main source of error is not statistical, but the practical uncertainty of the

method (the internal QE is not measured just believed, based on the literature).

Therefore this is a type B error. All the other participants have cryogenic radiometer……

2. In the color uncertainty budget I have left out data. YOU ARE right…

Revised budget:

source of uncertainty

standard uncertainty

probability distribution

sensitivity coefficient

standard uncertainty in

∆x

standard uncertainty in

∆y spectral

irradiance calibration

1,5% rectangular type B

sample dependent

∆x1 <0,002 0,0003 < ∆x1

∆y1 <0,002 0,0001 < ∆y1

wavelength error 0,1 nm rectangular

type B sample

dependent ∆x2 <0,001

0,00005 < ∆x2 ∆y2 <0,001

0,00005 < ∆y2

linearity 0,05% rectangular type B

sample dependent

∆x3 <0,0005 0,00005 < ∆x3

∆y3 <0,0005 0,00005 < ∆y3

stray light 10-15 – 10-13W rectangular type B

sample dependent

∆x4 <0,0014 0,00005 < ∆x4

∆y4 <0,002 0,00005 < ∆y4

dark noise 2*10-15 W rectangular type B

sample dependent

∆x5 <0,002 0,00003< ∆x5

∆y5 <0,003 0,00001 < ∆y5

room temp. dependence 1 K rectangular

type B sample

dependent ∆x6 <0,00005 ∆y6 <0,00005

light source repeatability as measured normal

type A sample

dependent as calculated as calculated

geometry error rectangular type B

sample dependent as calculated as calculated

combined standard

uncertainty 0,0004 < ∆x

∆x < 0,0026 0,0002 < ∆y ∆y < 0,0032


LED Measurements




Identification of Outliers

1. INTRODUCTION

The relative deviations from the mean value are calculated for each participant and for each type of LEDs and distributed in order to identify the obvious outliers, which can significantly skew the Reference Values of the comparison. Each participant should recommend which data should be removed in the calculation of the Reference Values. The name of the participant is not disclosed in this stage.

The relative deviations from the mean value are obtained as follows:

1. The ratios r1(Xi) and r2(Xi) are calculated for each artefact LED (Xi = R-i, G-i, B-i, or W-i with i = 1, 2 or 3):

1 21 2

( ) ( )( ) ; ( )( ) ( )

L i L ii i

P i P i

y X y Xr X r Xy X y X

= = . (1)

Here, yL(Xi), yP1(Xi), yP2(Xi) denote the measurement result of the participant laboratory, of the pilot laboratory before travel, and of the pilot laboratory after travel, respectively, for the artefact LED Xi.

2. The difference of the ratios corresponding to the artefact drift is calculated for each artefact LED Xi:

2 1( ) ( ) ( )i i id X r X r X= − . (2)

3. The mean value of each type of LEDs is calculated for each type of the artefact LEDs:

,

,

,

,

( ) ( ) ,

( ) ( ) ,

( ) ( ) ,

( ) ( ) .

i j i j

i j i j

i j i j

i j i j

m R Mean r R

m G Mean r G

m B Mean r B

m W Mean r W

=

=

=

=

. (3)

Here, the following data are excluded in the calculation of the mean: firstly, the data which are requested to be removed by the participant in the process of review of relative data, secondly, the data with its drift in Eq. (2) larger than 3 % after the temperature correction.

Note that the mean values of the ratios in Eqs. (3) correspond to the relative deviations of the participant’s data with respect to the pilot’s data.

4. The mean values in Eqs. (3) are normalized to the mean value of the measurement data of all the participants for the same type of the artefact LEDs:

[ ]

[ ]

[ ]

[ ]

( )( ) ,( )

( )( ) ,( )

( )( ) ,( )

( )( ) .( )

Lab xLab x

Lab n n

Lab xLab x

Lab n n

Lab xLab x

Lab n n

Lab xLab x

Lab n n

m RM RMean m R

m GM GMean m G

m BM BMean m B

m WM WMean m W

−−

−

−−

−

−−

−

−−

−

=

=

=

=

(4)

5. The deviations of the mean values in Eqs.(4) from 1 are calculated for each participant and for each type of the artefact LED:

( ) ( ) 1,( ) ( ) 1,( ) ( ) 1,( ) ( ) 1.

Lab x Lab x

Lab x Lab x

Lab x Lab x

Lab x Lab x

R M RG M GB M BW M W

− −

− −

− −

− −

∆ = −∆ = −∆ = −∆ = −

(5)

Note that the deviations in Eq. (5) correspond to the relative deviations of each participant from the mean value over all the participants for each type of the artefact LEDs.

In the case of the LED measurement, the quantity to be measured is a function of junction temperature. Therefore, the junction voltage is simultaneously measured and reported with the comparison quantity. Based on the reported junction voltage data and the characteristic parameters of each artefact LED determined by the pilot laboratory in the preparation stage, the measured comparison quantities can be corrected to one junction voltage.

In the following, the relative deviations in Eqs.(5) of all the participants are listed in a table and plotted for visualization. There are two sets of the data: the first set is based on the submitted measurement data without any correction. The second set is based on the data corrected to one junction voltage as a result of the temperature correction.

In the data table, the relative deviations larger than 6 % are indicated as red, which seem to be the obvious outliers. Note that we have considered here only the result data with an artefact drift much smaller than 3 %.

2. WITHOUT CORRECTION

Lab1 Lab2 Lab3 Lab4 Lab5 Lab6 Lab7

R 2.08% 8.64% 2.01% 0.97% 1.26% -19.39% 1.03%

G 1.03% 3.56% 2.00% -2.27% 2.39% 17.73% -0.61%

B 0.17% 1.96% -1.79% 0.09% 6.36% 10.51% -0.76%

W 1.43% 6.07% 3.45% 0.75% 2.82% 8.03% 0.68%

Lab8 Lab9 Lab10 Lab11 Lab12 Lab13

-9.31% 3.69% 2.85% 2.05% 9.37% -3.17%

-15.19% 0.02% 0.43% 0.16% -1.04% -7.20%

-10.07% 3.67% 3.99% 3.00% -2.76% -14.20%

-19.40% 1.78% 3.35% 0.61% 3.58% -11.71%

3. WITH TEMPERATURE CORRECTION

Lab1 Lab2 Lab3 Lab4 Lab5 Lab6 Lab7

R 3.54% 9.00% 2.16% 0.68% 2.03% -19.45% 0.72%

G 1.65% 3.69% 0.79% -2.67% 2.43% 17.80% -0.68%

B 0.35% 2.15% -1.58% 0.03% 6.63% 10.64% -0.65%

W 2.54% 6.81% 1.29% 0.32% 2.90% 7.77% 0.56%

Lab8 Lab9 Lab10 Lab11 Lab12 Lab13

-10.65% 3.51% 3.47% 3.98% 7.65% -3.10%

-15.19% 0.14% 1.29% 1.03% -1.57% -7.06%

-10.03% 3.68% 4.13% 3.26% -2.24% -16.02%

-19.41% 2.01% 4.67% 2.21% 2.45% -11.58%


LED Measurements




Pre-draft A Process

Review of Relative Data

1. INTRODUCTION

The relative data are calculated and distributed for review to check the stability of the artefact LEDs for each participant before and after travel, and the internal consistency of the artefact LEDs measured at each participant lab.

The relative data are obtained as follows:

1. The ratio R1(Xi) and R2(Xi) are calculated for each artefact LED (Xi = R-i, G-i, B-i, W-i, or D-i with i = 1, 2 or 3)

)()()( ;

)()()(

22

11

iP

iLi

iP

iLi Xy

XyXRXyXyXR == . (1)

Here, yL(Xi), yP1(Xi), yP2(Xi) denote the measurement result of the participant laboratory, of the pilot laboratory before travel, and of the pilot laboratory after travel, respectively, for the artefact LED Xi.

2. The ratios in Eq. (1) are normalized to the mean value of the measurement data for the same type (colour) of artefact LEDs:

[ ] [ ]jiji

ii

jiji

ii XRMean

XRXrXRMean

XRXr,

22

,

11 )(

)()( ;)(

)()( == . (2)

We refer these normalized ratios r1(Xi) and r2(Xi) as to the relative data for the artefact LED Xi. Note that the normalization in Eq. (2) removes any relationship of the absolute scale of the participant laboratory and leaves only internal consistency information within the sub-set of the same LED types.

In the case of the LED measurement, the quantity to be measured is a function of junction temperature. Therefore, the junction voltage is simultaneously measured and reported with the comparison quantity. Based on the reported junction voltage data and the characteristic parameters of each artefact LED determined by the pilot laboratory in the preparation stage, the measured comparison quantities can be corrected to one junction voltage. It is expected that this temperature correction via junction voltage can improve the stability and internal consistency of the artefact LEDs.

In the next chapters, the relative data of all the participants are listed and plotted for visualization. There are two sets of the relative data: the first set is based on the submitted measurement data without any correction. The second set is based on the data corrected to one junction voltage as a result of the temperature correction. By comparison of the two relative data, one can check if the temperature correction via junction voltage works properly by improving the stability of the artefact LEDs. The scale of all the plot of relative data is fixed (from 0.96 to 1.04) for a better comparison. Note that the non-correlated uncertainty of the pilot lab is smaller than 0.1 % (k = 1) for all the type of LEDs.

2. MIKES (SET #1)

2.1. WITHOUT CORRECTION

r1 r2

R-1 0.9971 1.0010

R-2 0.9992 1.0029

R-3 0.9980 1.0018

G-1 0.9980 1.0024

G-2 0.9953 1.0013

G-3 0.9999 1.0031

B-1 1.0005 1.0037

B-2 0.9956 1.0009

B-3 0.9992 1.0001

W-1 1.0515 0.9870

W-2 0.9887 0.9976

W-3 0.9842 0.9910

2.2. WITH TEMPERATURE CORRECTION

r1_cor r2_cor

R-1 0.9961 0.9997

R-2 0.9989 1.0026

R-3 0.9995 1.0032

G-1 0.9979 1.0022

G-2 0.9944 1.0001

G-3 1.0005 1.0048

B-1 1.0018 1.0050

B-2 0.9961 1.0014

B-3 0.9972 0.9985

W-1 1.0512 0.9875

W-2 0.9872 0.9964

W-3 0.9854 0.9922

3. CMS-ITRI (SET #2)


r1 r2

R-1 1.0028 0.9979

R-2 1.0000 0.9993

R-3 1.0017 0.9983

G-1 1.0036 0.9668

G-2 1.0061 1.0115

G-3 1.0033 1.0086

B-1 0.9914 1.0072

B-2 1.0010 1.0104

B-3 0.9909 0.9990

W-1 0.9967 0.9937

W-2 0.9973 1.0063

W-3 0.9991 1.0069


r1_cor r2_cor

R-1 0.9957 1.0033

R-2 0.9952 1.0053

R-3 0.9962 1.0043

G-1 1.0049 0.9671

G-2 1.0078 1.0121

G-3 1.0021 1.0061

B-1 0.9905 1.0063

B-2 0.9940 1.0025

B-3 0.9993 1.0074

W-1 0.9999 0.9947

W-2 1.0000 1.0061

W-3 0.9972 1.0021

4. PTB (SET #3)


r1 r2

R-1 0.9966 1.0076

R-2 0.9921 1.0032

R-3 0.9953 1.0052

G-1 0.9946 1.0032

G-2 0.9968 1.0026

G-3 0.9975 1.0053

B-1 0.9939 1.0016

B-2 0.9946 1.0027

B-3 1.0018 1.0055

W-1 0.9944 1.0068

W-2 0.9948 1.0053

W-3 0.9939 1.0048


r1_cor r2_cor

R-1 1.0009 1.0037

R-2 0.9967 1.0005

R-3 0.9980 1.0002

G-1 0.9970 1.0022

G-2 0.9990 1.0020

G-3 0.9976 1.0022

B-1 0.9950 1.0022

B-2 0.9954 1.0025

B-3 1.0010 1.0039

W-1 0.9975 1.0025

W-2 0.9986 1.0025

W-3 0.9968 1.0022

5. NMIJ (SET #4)


r1 r2

R-1 0.9957 1.0054

R-2 0.9939 1.0045

R-3 0.9961 1.0044

G-1 0.9961 1.0070

G-2 0.9943 1.0051

G-3 0.9945 1.0030

B-1 0.9976 1.0030

B-2 0.9982 1.0040

B-3 0.9961 1.0010

W-1 0.9923 1.0145

W-2 0.9883 1.0080

W-3 0.9890 1.0079


r1_cor r2_cor

R-1 0.9990 1.0018

R-2 0.9976 1.0016

R-3 0.9993 1.0008

G-1 0.9979 1.0063

G-2 0.9951 1.0038

G-3 0.9956 1.0013

B-1 0.9973 1.0030

B-2 0.9979 1.0040

B-3 0.9961 1.0017

W-1 0.9942 1.0116

W-2 0.9905 1.0061

W-3 0.9914 1.0063

6. CENAM (SET #5)


r1 r2

R-1 0.9924 0.9865

R-2 0.9952 0.9878

R-3 1.0227 1.0153

G-1 0.9943 0.9926

G-2 0.9779 0.9868

G-3 1.0245 1.0239

B-1 1.0595 1.0624

B-2 0.9749 0.9767

B-3 0.9616 0.9649

W-1 1.0066 1.0032

W-2 0.9957 0.9921

W-3 1.0042 0.9982


r1_cor r2_cor

R-1 0.9885 0.9908

R-2 0.9912 0.9924

R-3 1.0176 1.0194

G-1 0.9886 0.9916

G-2 0.9822 0.9953

G-3 1.0193 1.0230

B-1 1.0598 1.0637

B-2 0.9730 0.9769

B-3 0.9610 0.9657

W-1 1.0054 1.0069

W-2 0.9929 0.9944

W-3 0.9997 1.0007

7. LNE (SET #6)


r1 r2

R-1 0.9928 1.0021

R-2 0.9952 1.0074

R-3 0.9949 1.0077

G-1 0.9934 1.0041

G-2 0.9986 1.0059

G-3 0.9961 1.0018

B-1 0.9977 0.9994

B-2 0.9985 1.0029

B-3 0.9990 1.0025

W-1 0.9809 1.0147

W-2 0.9889 1.0152

W-3 0.9904 1.0099


r1_cor r2_cor

R-1 0.9969 0.9965

R-2 1.0014 1.0021

R-3 1.0019 1.0012

G-1 0.9963 1.0017

G-2 1.0004 1.0046

G-3 0.9967 1.0003

B-1 0.9975 0.9987

B-2 0.9996 1.0028

B-3 0.9994 1.0020

W-1 0.9818 1.0137

W-2 0.9904 1.0147

W-3 0.9915 1.0078

8. METAS (SET #7)


r1 r2

R-1 0.9955 0.9939

R-2 1.0081 1.0028

R-3 1.0012 0.9985

G-1 0.9915 0.9965

G-2 0.9991 1.0032

G-3 1.0027 1.0070

B-1 0.9884 0.9957

B-2 1.0113 1.0182

B-3 0.9900 0.9964

W-1 1.0011 1.0053

W-2 0.9973 1.0004

W-3 0.9972 0.9986


r1_cor r2_cor

R-1 0.9971 0.9975

R-2 1.0053 1.0036

R-3 0.9985 0.9980

G-1 0.9910 0.9972

G-2 0.9995 1.0047

G-3 1.0010 1.0066

B-1 0.9891 0.9966

B-2 1.0119 1.0193

B-3 0.9883 0.9948

W-1 1.0003 1.0074

W-2 0.9964 1.0001

W-3 0.9967 0.9991

9. A*STAR (SET #8)


r1 r2

R-1 1.0027 0.9975

R-2 1.0029 0.9970

R-3 1.0026 0.9973

G-1 0.9943 0.9991

G-2 0.9975 1.0023

G-3 1.0012 1.0056

B-1 0.9905 0.9973

B-2 0.9960 1.0022

B-3 1.0024 1.0117

W-1 0.9992 0.9999

W-2 0.9965 0.9999

W-3 0.9996 1.0050


r1_cor r2_cor

R-1 1.0034 1.0010

R-2 1.0002 0.9986

R-3 0.9988 0.9980

G-1 0.9947 1.0010

G-2 0.9958 1.0026

G-3 0.9994 1.0066

B-1 0.9899 0.9979

B-2 0.9944 1.0022

B-3 1.0028 1.0127

W-1 0.9965 1.0018

W-2 0.9954 1.0024

W-3 0.9983 1.0056

10. VSL (SET #1)


r1 r2

R-1 1.0068 1.0103

R-2 0.9932 0.9940

R-3 0.9991 0.9967

G-1 0.9993 1.0068

G-2 1.0147 1.0224

G-3 0.9753 0.9815

B-1 0.9892 1.0091

B-2 1.0001 1.0298

B-3 0.9779 0.9938

W-1 1.0018 1.0124

W-2 0.9807 0.9957

W-3 0.9944 1.0149


r1_cor r2_cor

R-1 1.0099 1.0080

R-2 0.9953 0.9906

R-3 1.0015 0.9948

G-1 0.9933 1.0138

G-2 1.0085 1.0254

G-3 0.9708 0.9881

B-1 0.9904 1.0064

B-2 1.0006 1.0291

B-3 0.9776 0.9959

W-1 0.9954 1.0181

W-2 0.9749 1.0025

W-3 0.9883 1.0208

11. NIST (SET #3)


r1 r2

R-1 1.0035 0.9952

R-2 1.0035 0.9965

R-3 1.0046 0.9967

G-1 1.0008 1.0049

G-2 0.9935 1.0019

G-3 0.9959 1.0029

B-1 0.9909 1.0101

B-2 0.9852 1.0112

B-3 0.9901 1.0126

W-1 0.9967 1.0064

W-2 0.9909 1.0055

W-3 0.9926 1.0079


r1_cor r2_cor

R-1 0.9998 1.0016

R-2 0.9977 0.9991

R-3 0.9995 1.0024

G-1 0.9988 1.0058

G-2 0.9916 1.0034

G-3 0.9951 1.0053

B-1 0.9905 1.0102

B-2 0.9847 1.0113

B-3 0.9899 1.0134

W-1 0.9945 1.0091

W-2 0.9885 1.0075

W-3 0.9912 1.0091

12. VNIIOFI (SET #5)


r1 r2

R-1 1.0003 0.9962

R-2 0.9929 0.9923

R-3 1.0087 1.0096

G-1 0.9978 1.0033

G-2 0.9908 1.0050

G-3 1.0001 1.0030

B-1 0.9787 0.9909

B-2 0.9818 0.9918

B-3 1.0227 1.0341

W-1 0.9972 0.9994

W-2 0.9955 1.0024

W-3 0.9992 1.0063


r1_cor r2_cor

R-1 0.9987 0.9973

R-2 0.9929 0.9932

R-3 1.0085 1.0093

G-1 0.9956 1.0049

G-2 0.9887 1.0067

G-3 0.9984 1.0056

B-1 0.9795 0.9918

B-2 0.9807 0.9914

B-3 1.0225 1.0342

W-1 0.9956 1.0040

W-2 0.9932 1.0051

W-3 0.9951 1.0070

13. INM (SET #7)


r1 r2

R-1 1.0110 1.0035

R-2 1.0056 1.0028

R-3 0.9919 0.9852

G-1 0.9762 0.9771

G-2 1.0208 1.0224

G-3 1.0025 1.0010

B-1 1.0166 1.0711

B-2 0.9625 0.9874

B-3 0.9736 0.9888

W-1 1.0428 1.0422

W-2 0.9793 0.9788

W-3 0.9765 0.9805


r1_cor r2_cor

R-1 1.0047 1.0054

R-2 1.0052 1.0085

R-3 0.9870 0.9892

G-1 0.9753 0.9789

G-2 1.0196 1.0235

G-3 1.0009 1.0019

B-1 1.0170 1.0719

B-2 0.9630 0.9883

B-3 0.9714 0.9884

W-1 1.0416 1.0462

W-2 0.9748 0.9808

W-3 0.9741 0.9823

apmp supplementary comparisons of led measurements · apmp supplementary comparisons of led...

Documents