Metrologia

PAPER You may also like

- Improving acoustic determinations of the
Realization and validation of the detector-based Boltzmann constant with mass
spectrometer measurements of the molar
absolute integrating sphere method for luminous- mass of argon
Inseok Yang, Laurent Pitre, Michael R
Moldover et al.
flux measurement at KRISS
- Total luminous flux measurement for
flexible surface sources with an integrating
To cite this article: Yong-Wan Kim et al 2012 Metrologia 49 273 sphere photometer
Hsueh-Ling Yu and Wen-Chun Liu

- Evaluation of the blackbody radiation shift

of an Yb optical lattice clock at KRISS
Myoung-Sun Heo, Huidong Kim, Dai-Hyuk
View the article online for updates and enhancements. Yu et al.

This content was downloaded from IP address 54.190.231.70 on 08/12/2022 at 02:36

IOP PUBLISHING METROLOGIA
Metrologia 49 (2012) 273–282 doi:10.1088/0026-1394/49/3/273

Realization and validation of the

detector-based absolute integrating
sphere method for luminous-flux
measurement at KRISS
Yong-Wan Kim1,2 , Dong-Hoon Lee1 , Seung-Nam Park1 , Min-Yong Jeon2
and Seongchong Park1,3
1
Division of Physical Metrology, Korea Research Institute of Standards and Science, 267 Gajeong-ro,
Yuseong-gu, Daejeon 305-340, Republic of Korea
2
Department of Physics, Chungnam National University, 99 Daehak-ro, Yuseong-gu, Daejeon 305-764,
Republic of Korea
E-mail: spark@kriss.re.kr

Received 7 November 2011, in final form 16 January 2012

Published 16 March 2012
Online at stacks.iop.org/Met/49/273

Abstract
We present two detector-based realizations of the absolute integrating sphere method for
luminous-flux measurement. A large-size realization with a sphere diameter of 2 m is applied
to high-flux lamps above 1000 lm, and a small-size realization with a sphere diameter of 0.3 m
to low-flux lamps below 100 lm. The uncertainty of both realizations is analysed with a focus
on the systematic effect in the spatial mismatch correction, and evaluated to be 1.2% and 1.1%
(k = 2) for the large-size and small-size realizations, respectively. Comparison with a
goniophotometer using two different types of lamp artefacts for both cases has verified the
validity of the realization within the comparison uncertainty.
(Some figures may appear in colour only in the online journal)

1. Introduction angular distribution of luminous intensity of a test lamp over

4π solid angle by scanning a reference photometer and reduces
Total luminous flux, or luminous flux in short, expressed in the it to a total luminous-flux value by numerical integration. Since
unit of lumen (lm), is a photometric quantity that is defined it is suitable for absolute measurement, i.e. without need of
as a spatial and spectral integral of radiant power emitted another reference lamp, goniophotometers are used by many
from a light source with a weighting function of the luminous National Metrology Institutes (NMIs) to realize their primary
efficacy of the human eye. The quantity is an important luminous-flux scale and transfer it to a group of standard lamps.
specification to describe the energy efficiency of light sources However, a large and high-cost facility and long measurement
for lighting and display applications. With the emerging time limit the practical application of goniophotometers. On
market for solid-state lighting (SSL) based on light-emitting the other hand, an IS photometer takes advantage of the
diodes (LEDs) as an energy-saving alternative for lighting proportionality of an output signal to luminous flux inside
and display, the importance of accurate measurement of total the sphere. The proportionality enables us to measure total
luminous flux especially for LEDs and LED-based products is luminous flux simply by comparing a test lamp with a reference
rapidly increasing. lamp, which results in practical benefits such as simple setup
The luminous flux of a light source can be measured
and fast measurement. However, to ensure accurate operation
using a goniophotometer or using an integrating sphere (IS)
of an IS photometer, the reference lamp should be matched
photometer [1, 2]. A goniophotometer directly measures the
with a test lamp in aspects of spectral distribution, spatial
3 Author to whom any correspondence should be addressed. distribution, lamp material absorption, size, etc.

0026-1394/12/030273+10$33.00 © 2012 BIPM & IOP Publishing Ltd Printed in the UK & the USA 273
Y-W Kim et al

Figure 1. Schematic setup of the large-size absolute IS photometer at KRISS (top view). The dashed lines depict supporting posts of baffles.

In 1995, Ohno at National Institute of Standards and 100

Technology (NIST) developed the absolute IS method as an spectral reflectance × 100
alternative to the goniophotometer method for realization of 95
the luminous-flux scale [3, 4]. Successful realizations of the 90
absolute IS method have been demonstrated at several NMIs
large sphere
[4–7]. The validity of the realized methods were tested either 85
small sphere
by comparison with a goniophotometer [6] or by comparison
80
with other institutes [7, 8]. The absolute IS methods reported
so far were realized using spheres with a diameter larger 75
than 1 m to optimize their performance for typical E26-based
incandescent bulbs of several hundreds of lumen, which are 70
350 400 450 500 550 600 650 700 750
conventionally used to maintain and disseminate the total wavelength / nm
luminous-flux scale.
At Korea Research Institute of Standards and Science Figure 2. Spectral reflectivity of sphere coatings for two absolute IS
(KRISS), we recently implemented two IS photometers based photometers, measured using a portable spectrophotometer. The
curves ‘large sphere’ and ‘small sphere’ correspond to the
on the absolute IS method. One is based on a large-size
reflectivity of the 2 m diameter sphere and the 0.3 m diameter
sphere with a diameter of 2 m for E26-based incandescent sphere, respectively.
bulbs of several hundreds of lumen, while the other is based
on a small-size sphere with a diameter of 0.3 m for bi-pin,
have the same traceability, the comparison will reflect the
miniature lamps such as a 5 mm, lamp-typed LED of a few
equivalence of the two realizations of the absolute IS method
or tens of lumen. The large IS was originally designed and
with respect to the goniophotometer method.
used for more than ten years as a basic Ulbricht sphere for
substitution measurement. The small one was designed as an
absolute IS from the beginning, and currently works for LED 2. Experimental realization
measurement [9]. The principle of design and operation for
both IS photometers is similar to that described in [3] except 2.1. Large-size absolute integrating sphere photometer
how to measure the reference flux from the external source; in Figure 1 schematically shows the experimental arrangement
our cases, the reference flux of the external source is measured of the large-size absolute IS photometer realized at KRISS.
under an underfill condition using standard illuminance meters The IS has a diameter of 2 m and its inner wall is coated with
whose luminous responsivity (unit: A lm−1 ) are calibrated BaSO4 , the spectral reflectivity of which is shown in the curve
against the spectral responsivity scale of KRISS. ‘large sphere’ of figure 2. The IS is modified from a basic
In this work, we describe the realization and validation Ulbricht sphere design with only one baffle (B1 ) in front of
of the absolute IS method at KRISS. For validation of the a detection port to an absolute IS photometer by adding an
realization, measurement results of two types of lamp artefacts entrance port for the reference flux with another baffle (B2 ).
by each absolute IS photometer are compared with those by A lamp mounting receptacle is positioned at the centre of
a goniophotometer that is currently declared as the primary the sphere, which is compatible with E26-based lamps. An
method for the KRISS luminous-flux scale. Since the two external source, a 50 W quartz–tungsten–halogen lamp (QTH
absolute IS photometers as well as the goniophotometer lamp) with a collimation lens, is operated at a correlated colour

274 Metrologia, 49 (2012) 273–282

 Absolute integrating sphere method for luminous-flux measurement at KRISS

temperature (CCT) of 2750 K to provide the reference flux into critical correction component because the difference of the
the sphere. The spot area illuminated by the reference flux on area illuminated by the test lamp and by the reference flux
the sphere wall has a diameter of approximately 11 cm. The is significant [4, 5, 7]. The general expression for the SCF
reference flux is measured using a standard photometer with for a lamp with a normalized luminous intensity distribution
a clear window under the underfill condition (beam size < I ∗ (θ , φ) and a normalized SRDF K ∗ (θ, φ) is given by [4]
photometer area). The sampled flux through a detection port on
1
the rear hemisphere is measured by a sphere photometer with kSCF =
2π
π . (4)
a flat diffuser. Each photometer has a temperature-stabilized 0 0 I ∗ (θ, φ)K ∗ (θ, φ) sin θ dθ dφ
V (λ) filter with its error parameter f1
< 2% [10]. Two baffles
B1 (35 cm diameter) and B2 (24 cm diameter) inside the sphere For an absolute IS photometer, the spot size of the reference
shield the detection port (3 cm diameter) and the entrance port flux can be smaller than or comparable to the spatial resolution
(14 cm diameter), respectively, from direct illumination by the of the measured SRDF. In this case, I ∗ (θ, φ) in equation (4)
test lamp. can be replaced by a delta function δ(θ − θe , φ − φe ) that is
The measurement principle of the absolute IS photometer equal to unity at the position (θe , φe ) and zero at other positions.
is as described in [4, 5]. The luminous flux T of the test lamp Here, the position (θe , φe ) corresponds to the location of the
is determined by the measurement equation reference flux spot on the sphere wall. Therefore, the equation
R
for kSCF is simply reduced to
yT
T = · R · fspectral · fspatial · fangle , (1) 1
yR R
kSCF = . (5)
K ∗ (θ e , φe )
where y T denotes the sphere photometer reading for the test The last correction factor fangle in equation (1) considers the
lamp, y R the sphere photometer reading for the introduced difference of angle of incidence between the test lamp flux
reference flux and R the luminous-flux value of the reference (0◦ ) and the reference flux (45◦ ) on the sphere wall. Since the
flux provided by the external source and measured by the sphere coating cannot be a perfect Lambertian, this difference
standard photometer. of incident angle between two sources at the reference flux spot
The three factors fspectral , fspatial and fangle in equation (1) position causes an error: the SRDF K ∗ (θe , φe ) in equation (5)
consider the mismatches between the test lamp and the for the reference flux should be measured at an incidence angle
reference flux, which cause errors due to non-ideality of of approximately 45◦ , while the SRDF K*(θ , φ) in equation (4)
the IS. The correction factor fspectral considers the spectral for the test lamp is usually measured at normal incidence (0◦ ).
mismatch between the test lamp and the reference flux. In To correct this systematic error, an additional factor for the
general, it can be determined from the ratio of the respective incident angle difference correction is required, which can be
colour correction factors (CCFs) against the International defined by the ratio of the sphere responses at the reference
Commission on Illuminance (CIE) standard illuminant A: flux spot position:
T
kCCF K45 (θe , φe )
fspectral = R
. (2) fangle = . (6)
kCCF K0 (θe , φe )
T R Here, the sphere responses K45 and K0 measured at incidence
For determination of kCCF and kCCF , one needs to know the
spectral distribution of the test lamp and the external source, angles of 45◦ and 0◦ , respectively, do not need to have the same
respectively, together with the spectral throughput of the IS normalization as the SRDF. By introducing this correction
and the relative spectral responsivity of the sphere photometer. factor in equation (1), both SCFs in equation (3) can be
In our experiment, we adjusted the lamp current of the external determined based on the same SRDF measured at a normal
source so that its CCT is the same as the test lamp. We incidence without causing a systematic error.
experimentally confirmed that the CCT matching resulted Figure 3 shows the SRDF of the large-size absolute IS
in the spectral correction factor fspectral of unity within its photometer measured with a commercial beam scanner based
uncertainty. on a collimated white LED [11]. The spot diameter of the
The correction factor fspatial considers the spatial scanning beam on the sphere wall was 16 cm, and the angular
mismatch between the test lamp and the reference flux. This scan step was 9◦ both in polar and azimuth angles. The circular
can be determined from the ratio of the respective spatial dark area with a bright ring around the region A in the rear
correction factors (SCFs) against an isotropic point source: hemisphere plot of figure 3(a) corresponds to the effect of the
baffle B1 , which screens the sphere photometer from a direct
T
kSCF illumination of the scanner beam (see also figure 1). Around
fspatial = . (3)
R
kSCF the region B in the front hemisphere plot of figure 3(b), there is
another circular dark area that is much larger than the feature
T R
In order to determine kSCF and kSCF , one needs to know around the region A. This is attributed to a secondary screening
the spatial intensity distribution of the test lamp and the by the baffle B1 from illumination of the first reflection on the
reference flux, respectively, together with the normalized front sphere. Both features around the region A and the region
spatial response distribution function (SRDF) of the IS B commonly appear also in a basic Ulbricht sphere system of
photometer. For the absolute IS method, this is the most one baffle and one opening [12]. The feature in the region

Metrologia, 49 (2012) 273–282 275

Y-W Kim et al

Table 1. Uncertainty budget of luminous flux of a 100 W tungsten

lamp for the large-size absolute IS photometer.
Related Uncertainty Relative standard
quantity component uncertainty × 100
y R and y T Repeatability 0.01
yT Lamp operation 0.06
yT Near-field absorption 0.20
y R /y T Non-linearity 0.03
R Reference flux 0.25
fspectral Spectral mismatch 0.01
fspatial Spatial mismatch, ref. flux 0.34
fspatial Spatial mismatch, test lamp 0.30
fangle Angle mismatch 0.15
Combined uncertainty 0.58
Expanded uncertainty (k = 2) 1.2

makes it possible to realize the absolute IS method even with

such a non-ideal sphere. The angle correction factor fangle of
equation (6) is evaluated by measuring the sphere photometer
response under a miniature flashlight illumination onto (θe , φe )
at different incident angles, which amounts to 0.976. Based on
naked-eye inspection of the sphere surface around (θe , φe ), we
presume that the large deviation of fangle from unity comes
from the large size of grains on the BaSO4 coating which
produces a visible texture.
Table 1 shows the representative uncertainty budget of
luminous flux of a 100 W tungsten lamp for the large-size
absolute IS photometer based on the measurement equation (1).
As all the input quantities in the measurement equations
are simply multiplicative and uncorrelated, the uncertainty
components are calculated as relative standard uncertainties
and combined independently with the sensitivity coefficient
Figure 3. Measured SRDF of the large-size absolute IS photometer of unity. Note that most of the components in table 1
(a) for the rear hemisphere and (b) for the front hemisphere. The are evaluated as type-B and sample-independent. The only
sphere photometer is attached at the centre of the rear hemisphere. sample-dependent component of repeatability in the budget
The features marked with A and B are attributed to the baffle B1 ,
corresponds to the maximum standard uncertainty expected
and the feature marked with C to the baffle B2 .
from the similar lamp types.
The repeatability uncertainties of y R and y T are evaluated
C in figure 3(b) is caused by the baffle B2 . Theoretically, it to be less than 0.01% from the standard deviation of the
is expected that the baffle B2 produces a bright area in the repeated signal readings. The lamp operation uncertainty
region C as it effectively shortens the sphere radius. However, component of y T is propagated from the uncertainty of the lamp
the feature measured in the region C is dark and not uniform current measurement. The near-field absorption component of
because the reflectivity of the newly installed baffle B2 is not the y T is related to the loss of flux of the test lamp at its holding
same as the other region of the sphere wall. We also note that receptacle that is impossible to compensate. We estimated the
the overall non-uniformity of the SRDF in figure 3 is mainly amount of the near-field absorption by varying the receptacle
caused by the reduced reflectivity of the sphere wall through reflectivity, i.e. by mounting receptacles with different grey
its long-time use (see also figure 2). levels. The non-linearity component is related to the ratio of
Based on the measured data of K ∗ (θ, φ) in figure 3, we y T and y R , which have different signal levels of the sphere
calculated kSCFR
= 0.946 by equation (5) for the reference photometer by two orders of magnitude. The non-linearity of
flux and kSCF = 1.002 by equation (4) for a typical E26-
T
the sphere photometer is measured based on the flux-addition
based, frosted, 100 W tungsten test lamp. Note that this kind method using LEDs [13]. The non-linearity of current reading
of test lamp is used as the artefact for the validation study, for the sphere photometer is tested using an electrical current
which will be presented in section 3. As a result, the spatial calibration source.
correction factor for the large-size IS photometer is calculated The measurement uncertainty of the reference flux R
to be fspatial = 1.059 by equation (3). This large correction is evaluated from calibration uncertainty of the standard
factor reveals also the poor uniformity of the IS response. photometer, spatial non-uniformity, ambient stray light
However, we will later verify that the proper correction by influence and reading repeatability. However, the dominant
fspatial and fangle based on the careful measurement of the SRDF component among these was the calibration uncertainty of

276 Metrologia, 49 (2012) 273–282

 Absolute integrating sphere method for luminous-flux measurement at KRISS

the standard photometer, i.e. of its luminous responsivity

in A lm−1 . The relative standard uncertainty of the scale
realization for luminous responsivity at KRISS is 0.25% based
on the spectral responsivity measurement [14].
The uncertainty of the spectral correction factor fspectral L P
Q
is determined from the uncertainties of the CCFs for both the M R
test lamp and the external source according to equation (2).
S
For typical incandescent tungsten lamps, the relative standard
uncertainty of CCF amounts to less than 0.01%, which R′
can be evaluated by propagating the uncertainty of spectral B1 Q′
distribution measurement using a spectroradiometer [15]. In P′
the actual measurements, the CCT of the reference source is
matched to the test lamp by adjusting its current.
The uncertainty of the spatial correction factor fspatial is
determined from the uncertainties of the SCFs for both the
Figure 4. Illustration to explain the influence of the size of a light
reference flux and the test lamp according to equations (3)–(5).
R source mounted in an IS photometer on its SRDF. See text for
For the SCF of the reference flux kSCF , the dominant component details.
is the spatial coordinate mismatch between I *(θ , φ) and K*(θ ,
φ), which corresponds to the spatial coordinate mismatch of and position of the baffle B1 in front of the sphere photometer.
T
two spots, one from the reference flux and the other from To quantify the uncertainty of kSCF by the shadow-mismatch
the beam scanner for the SRDF measurement, at the position effect, we measured the sphere response as a function of the
(θe , φe ) in equation (5). In order to estimate this, we calculated test lamp position within the extent where the detection port
the SCFs with the SRDF shifted by its measurement resolution and the entrance port remain to be screened by the baffles B1
R and B2 , respectively, from the direct illumination of the test
of θ and φ. From the variation of kSCF by shifting the SRDF
∗
within K (θe ±θ, φe ±φ), we evaluated a relative standard lamp (see figure 1). From the relative variation of the sphere
T
uncertainty of 0.34%. The other uncertainty components such response with the lamp position, the uncertainty of kSCF is
as flux stability of the beam scanner for the SRDF measurement estimated to be 0.3%. It is noteworthy that, in contrast with
R T
or reproducibility of the SRDF measurement at one position the kSCF , the variation of kSCF on an angular coordinate shift
were evaluated to be less than 0.05%. by (θ ± θ, φ ± φ) of the SRDF was less than 0.1%.
The uncertainty of the SCF of the test lamp kSCF T
is another Finally, the uncertainty of the angle correction factor fangle
component for fspatial , which should be evaluated with greater in equation (6) is estimated from the reproducibility of its
care. As described in equation (4), kSCF T
is calculated from measurement to be smaller than 0.15%. By combining the
I *(θ , φ) of the test lamp and K*(θ, φ), both measured over full uncertainty components described so far, the measurement
(4π ) solid angle. Note that the intensity distribution I *(θ , φ) uncertainty of the large-size absolute IS photometer is
of the test lamp is measured using the KRISS goniophotometer evaluated for a 100 W tungsten lamp to be 1.2% as a relative
that will be described in the next section. Although both I *(θ , expanded uncertainty at a level of confidence of approximately
φ) and K*(θ, φ) are measured over full solid angle, there exists 95% with a coverage factor of k = 2.
a systematic error due to a size difference between the test
lamp and the beam scanner used for the SRDF measurement. 2.2. Small-size absolute integrating sphere photometer
This is explained based on the illustration in figure 4. The Figure 5 schematically shows the experimental arrangement of
spot S is supposed to be the size of the beam scanner with the small-size absolute IS photometer realized at KRISS. Note
a small light-emitting area. When the sphere is scanned for that the drawing of figure 5 takes a different viewpoint from
SRDF measurement, the spot S and the baffle B1 produce a that of figure 1 to better show the difference of the arrangement.
shadowed zone marked by P and P
around the sphere detector The IS has a diameter of 0.3 m and its inner wall is coated with
in figure 4. The size of such a shadowed zone depends on BaSO4 like the large one described previously, but the spectral
the size of the light-emitting area at the centre: the spots S, reflectivity is higher as shown in the curve ‘small sphere’ of
M and L with different sizes correspond to the zone P–P
, figure 2. The external source is a 50 W tungsten–halogen
Q–Q
and R–R
, respectively. Note that the large spots M and lamp operated at a CCT of 2650 K. The lamp receptacle for
L can correspond to the size of the test lamp. Since SRDF mounting test lamps is designed for LEDs and other small-size
value abruptly changes around the detection port as shown in lamps. The size and the divergence of the output beam from the
figure 3(a) and as discussed in [12], a small mismatch between external source are limited with two apertures so that the beam
I *(θ , φ) and K*(θ , φ) due to different sizes of the zones diameter of the reference flux at the entrance port of the sphere
P–P
, Q–Q
and R–R
shadowed by the detector baffle can is only 12 mm. The spot illuminated by the reference flux on
T
significantly affect the calculation accuracy of the SCF kSCF the sphere wall has a diameter of 30 mm. The luminous flux
for the test lamp. In addition, the mounting position of the test of the reference flux is measured under the underfill condition
lamp cannot perfectly coincide with that of the beam scanner, (beam size < photometer area) with the standard photometer
which also causes a similar shadow-mismatch effect. Note that with an active-area diameter of 15 mm. Similar to the large-
T
the sensitivity of these two effects to kSCF depends on the size size absolute photometer, the sphere photometer has a flat

Metrologia, 49 (2012) 273–282 277

Y-W Kim et al

Figure 5. Schematic setup of the small-size absolute IS photometer

at KRISS (side view). The dashed lines depict posts supporting the
baffles.

diffuser while the standard photometer has a clear window.

Each photometer has a temperature-stabilized V (λ) filter with
f1
< 2%. Two baffles B1 (18 mm diameter) and B2 (35 mm
diameter) shield the sphere photometer port (12 mm diameter)
and the entrance port (38 mm diameter) of the reference flux,
respectively, from direct illumination by the test lamp.
Figure 6 shows the measured SRDF of the small-size IS
photometer. The beam scanner for measuring the SRDF is
made of a green LED (lamp-type with a diameter of 5 mm)
with a 1 mm aperture and is operated by manually rotating two
axes. A motorized beam scanner could not be used due to the
limited space inside the sphere. The beam spot size of the
scanner was 30 mm in diameter, and the SRDF is measured
with a scan step of 5◦ both in polar and azimuth angles. The
uniformity of the small-size IS photometer is superior to that
of the large-size IS photometer, which can be recognized by
the scale difference of figure 6 compared with figure 3. Figure 6. Measured SRDF of the small-size absolute IS photometer
In the rear hemisphere plot of figure 6(a), the dark spot (a) for the rear hemisphere and (b) for the front hemisphere. The
sphere photometer is attached at the centre of the rear hemisphere.
around the region A is due to low reflectivity of the test lamp The features marked with A and D are attributed to the lamp
mounting post. The region B shows the effect of the detection mounting post, the features marked with B and E to the baffle B1 ,
baffle B1 , which seems to be qualitatively different from that and the feature marked with C to the baffle B2 .
of the large-size IS photometer shown in figure 3(a). This is
because the spot size of the scanning beam for the small-size
uncertainty of near-field absorption is reduced in table 2
IS photometer (30 mm in diameter) was larger than the size
because the lamp itself allows little emission in the direction of
of the baffle B1 (18 mm in diameter). A shadow due to the
the lamp receptacle. The correction factor fspatial is calculated
baffle B2 in front of the entrance port for the reference flux can R
to be 1.015 from the SCFs kSCF of 0.984 for the reference
also be identified in the region C of figure 6(a). In the front T
hemisphere plot of figure 6(b), the feature in the region D flux and kSCF of 0.999 for typical test lamps. The uncertainty
R T
is attributed to the secondary screening by the lamp mounting components of fspatial by kSCF and kSCF are evaluated to be
post. Around the region E, there is another circular dark area as 0.14% and 0.4%, respectively, in the same way as for the large-
R
large as the feature around the region B, which is also attributed size IS photometer. Note that the uncertainty of kSCF for the
to the secondary screening of the first reflection on the front small-size IS photometer is smaller than that for the large-
sphere by the baffle B1 . size one due to the better uniformity, while the uncertainty
T
The measurement equation for the small-size IS of kSCF for the small-size IS photometer is larger than that
photometer is the same as the large-size one of equation (1). for the large-size one due to the small size of the baffle B1 .
Table 2 shows the representative uncertainty budget of the The value of fangle is evaluated to be 0.995. As a result,
luminous flux of a 5 W tungsten–halogen lamp for the small- the measurement uncertainty of the small-size absolute IS
size absolute IS photometer. Compared with table 1, the photometer is evaluated for a 5 W tungsten–halogen lamp to be

278 Metrologia, 49 (2012) 273–282

 Absolute integrating sphere method for luminous-flux measurement at KRISS

Table 2. Uncertainty budget of luminous flux of a 5 W tungsten

lamp for the small-size absolute IS photometer. KRISS spectral
responsivity scale
Related Uncertainty Relative standard
quantity component uncertainty × 100
y R and y T Repeatability 0.01 Luminous responsivity
yT Lamp operation 0.04 of illuminance meters
yT Near-field absorption 0.05
y R /y T Non-linearity 0.03
R Reference flux 0.25
fspectral Spectral mismatch 0.01
fspatial Spatial mismatch, ref. flux 0.14
fspatial Spatial mismatch, test lamp 0.40 Small-size
fangle Angle mismatch 0.15 Large-size Gonio-
absolute IS
absolute IS photometer
Combined uncertainty 0.52 photometer
Expanded uncertainty (k = 2) 1.1 photometer (reference)
(for LEDs)

comparison with comparison with

100 W tungsten 5 W tungsten-
lamp (1100 lm) halogen lamp (40 lm)

Figure 8. Traceability and validation scheme of the absolute IS

method realized at KRISS.

in azimuth angle φ. The lamp mount module includes a lamp

receptacle, a post, an electrical slip-ring and a rotation motor.
A standard photometer is mounted in the photometer arm with
a series of apertures which limits its field of view (FOV). The
FOV is adjusted according to the size of a test lamp. The stray
light is kept to less than 0.1% by the limited FOV and the light
trap.
Figure 8 shows the traceability chart of the absolute IS
photometers and their validation scheme to the luminous-flux
scale of KRISS. All the three realizations of the luminous-flux
scale at KRISS have the same traceability to the KRISS spectral
responsivity scale through calibration of luminous responsivity
of standard photometers.
Figure 7. Schematic setup diagram of the KRISS goniophotometer As the large-size and small-size absolute IS photometers
used for the primary realization of the luminous-flux scale at KRISS. are to be operated at different flux levels and with different
lamp bases, two types of artefacts are used for comparison: a set
1.1% as a relative expanded uncertainty at a level of confidence of E26-based, frosted, 100 W tungsten lamps with a nominal
of approximately 95% with a coverage factor of k = 2. luminous flux of 1100 lm and a set of bi-pin-based (non-
standard), clear, 5 W tungsten–halogen lamps with a nominal
3. Validation method luminous flux of 40 lm. Photographs of both lamp types are
shown in figure 8. The diameters of the 100 W and 5 W lamps
The realization of the two absolute IS photometers at KRISS are 60 mm and 6 mm, respectively. The 100 W lamps are
has been described so far. To validate them, we have performed normally operated at a direct current of 0.84 A and a voltage
a comparison study with the goniophotometer that is currently of 105 V, resulting in a CCT of 2750 K. We note that the same
declared as the primary realization of the KRISS luminous- type of 100 W lamps is currently used as the standard lamps
flux scale [16]. Figure 7 shows a schematic diagram of the for maintaining and disseminating the luminous-flux scale of
KRISS goniophotometer. The goniophotometer consists of a KRISS. The 5 W miniature-type lamps have a dimension and
2π-rotating photometer arm of 2 m diameter for scanning in mounting electrodes which are compatible with typical lamp-
polar angle θ and a π-rotating lamp mount module for scanning type LEDs of 5 mm diameter. The 5 W lamps are operated at

Metrologia, 49 (2012) 273–282 279

Y-W Kim et al

a direct current of 0.97 A and a voltage of 5.1 V, resulting in a

CCT of 2650 K [17]. As the original lamp had too short bi-pin
terminals, we extended them by spot-welding terminal wires,
and reinforced the lamp by packaging it into a stainless-steel
housing with ceramic filling.
The comparison is conducted in three steps. First, the
100 W lamps and the 5 W lamps are separately measured
with the large-size and the small-size absolute IS photometers,
respectively. Second, both types of test lamps are measured
with the goniophotometer. Finally, the lamps are re-measured
with the respective IS photometers for a drift check. Note that
a single measurement of total luminous flux via a full scan of
polar angles with the KRISS goniophotometer requires at least
30 min of burning time, while the IS photometer measurement
requires only a few minutes. We selected two test lamps of each
type for comparison data analysis based on a criterion that the
drift of total luminous flux during the comparison should be
less than 0.5%. The comparison results are summarized in the
next section.
Before we discuss the comparison results for validation
of the new realizations, we need to first verify the accuracy
of the goniophotometer for the 5 W test lamps. We note that
the previous uncertainty evaluation of the goniophotometer
at KRISS described in the previous report [16] was limited
to only the 100 W standard lamp type. The main issues
are the influence of different flux levels and spatial emission
characteristics of the two lamp types. To cover a range of total
luminous flux from 30 lm to 1200 lm, we extended the dynamic
range of the photometer by calibrating the trans-impedance
amplifier gain of the goniophotometer at two different ranges of
1000 nA and 10 nA over five decades. A numerical correction
algorithm of the non-linear gain is installed in the measurement
software based on the calibrated gain table. Figure 9. Angular intensity distribution of (a) a 100 W tungsten
A more critical source of error or uncertainty, however, lamp and (b) a 5 W tungsten–halogen lamp, which are measured
is the different spatial emission characteristics of the two with the KRISS goniophotometer for validation of the IS
lamp types. Figure 9 shows the angular intensity distributions photometers by comparison. The measured data are plotted in a
of both lamp types measured with the goniophotometer. polar diagram against the tangential angle θ so that each line
corresponds to a data set for one value of φ.
The difference of the emission characteristics is significant
especially around the axis of θ = 0◦ , at which the lamp of the KRISS scale identified in the previous international
mount module is placed. The shadowed cone by the lamp comparison on total luminous flux [16]. In order to correct this
base is much broader for the 5 W lamp (|θ | < 60◦ ) than error, we first elongated the post of the lamp mount module
for the 100 W lamp (|θ | < 20◦ ) due to differences in base,
as long as possible, as shown in figure 7, to minimize the
filament and bulb finishing (clear or frosted) (see photographs
screening cone. In addition, we replaced the aluminium cap of
in figure 8). Since the 100 W lamp emits a considerable light
the lamp receptacle with a polytetrafluoroethylene (PTFE) cap
flux toward the lamp mount module compared with the 5 W
to minimize the absorption loss at the lamp receptacle. After
lamp, it experiences a higher screening loss in the lamp mount
these modifications, the deviation between the IS photometer
module of the goniophotometer than the miniature-type 5 W
and the goniophotometer for different lamp types is reduced
lamp.
below 0.3%, which is considered as the estimate of the
In order to evaluate such a systematic error due to different
remaining uncertainty component of the screening effect by
losses in the lamp mount module, we compared test lamps
the lamp mount after correction.
of different sizes and emission characteristics, including the
two lamps for comparison, in a conventional IS photometer Table 3 shows the uncertainty budget for the luminous-flux
with a diameter of 2 m for substitution measurement. Initially, measurement of the 100 W and 5 W lamps with the KRISS
the deviation of the luminous-flux ratio between different goniophotometer. The budget is based on the measurement
lamps measured with the IS photometer is measured to be equation
as large as 1% compared with the ratio measured with the  CCF 
goniophotometer. Then, we noticed that the screening loss T = fij Eij R 2 = · R2 fij yij , (7)
Sv
by the lamp mount module can be the main systematic error i,j i,j

280 Metrologia, 49 (2012) 273–282

 Absolute integrating sphere method for luminous-flux measurement at KRISS

Table 3. Uncertainty budget of luminous flux of a tungsten lamp for −0.25% for 5W-2 and 5W-6, respectively. In contrast to
the KRISS goniophotometer. the large-size one, the small-size absolute IS photometer
Related Uncertainty Relative standard underestimates the total luminous flux by 0.37% on average.
quantity component uncertainty × 100 The uncertainty of the comparison in the last column
 of table 4 is calculated from the combined relative standard
 fij yij Repeatability 0.10
Angle accuracy 0.05 uncertainty of the comparison ratio uc (ratio) based on the
 fij yij
Stray light 0.20 following equation:
 fij yij
Screening by lamp mount 0.30 
 fij yij
fij yij Lamp operation 0.06 uc (ratio) = u2 ()IS + u2 ()GP − 2u2 (Sv ) + u2drift . (9)
Sv Luminous responsivity 0.25
R Distance (sensitivity factor 2) 0.20 Here, u()IS and u()SP denote the standard uncertainty of
CCF Spectral mismatch 0.01 luminous-flux measurement with the absolute IS photometer
Combined uncertainty 0.50 and the goniophotometer, respectively. The terms u(Sv ) and
Expanded uncertainty (k = 2) 1.0 udrift mean the uncertainty of luminous responsivity of standard
photometers and the maximum drift of test lamps, respectively.
Note that, among the various components in the budgets from
where CCF and Sv denote the spectral mismatch correction table 1 to table 3, the uncertainty of luminous responsivity of
factor and the luminous responsivity of the standard standard photometers u(Sv ) or reference flux u(R ) is regarded
photometer, respectively. The reading yij of the photometer as being correlated in the ratio measurement. The calculated
is recorded at each value of coordinates (θi , φj ) with index expanded relative uncertainties of 1.3% and 1.4% (k = 2) for
i, j at a distance of R from the test lamp with the geometric the comparison of the small-size and large-size absolute IS
weighting factor fij defined by photometers, respectively, are both larger than their relative
     differences from the goniophotometer.
 θ θ 

fij = φ cos θi − − cos θi + . (8) The comparison results in table 4 verify the equivalence
2 2  of the realized absolute IS method with the goniophotometer
method within the comparison uncertainty. However, the
The values of each uncertainty component in table 3 are results also indicate that there exists a noticeable systematic
representative both for 100 W and 5 W lamps. We recall difference of approximately 0.5% in the opposite directions
that the uncertainty of luminous responsivity Sv in table 3 for the two absolute IS photometers. Although this difference
is strongly correlated with the uncertainty components of is smaller than the uncorrelated uncertainties of both methods,
flux measurement in tables 1 and 2 because all the standard the origin of the systematic effect needs to be investigated.
photometers are calibrated against the same reference as shown It is noteworthy that this systematic effect is not related to
in the traceability chart in figure 8. the photometric scale realization since all the implementation
shares the same traceability.
4. Comparison results Based on the experience of the comparison, we point out
the effect of the different spatial emission characteristics and
Table 4 summarizes the comparison results of the two dimensions of test lamps on the aspect of the goniophotometer
realized absolute IS photometers with the goniophotometer as one of the most probable sources of systematic error, which
by measuring total luminous flux of two different types of is related to the screening by the lamp mount module. To
tungsten-filament lamps. alleviate this effect, the screening cone by the lamp mount
Two 100 W test lamps designated as 100W-7 and 100W-16 module should be reduced by minimizing the uppermost
are measured by the large-size absolute IS photometer, by the bulky part of the lamp mount module or by elongating the
post of the lamp mount module. On the aspect of the
goniophotometer, and again by the absolute IS photometer.
absolute IS photometer, the spatial mismatch correction is
From the two values for the absolute IS photometers, we
most probable as a source of systematic error because its
calculate the drift of luminous flux during the comparison to
uncertainty dominates in the budget as shown in tables 2 and 3.
be 0.05% and 0.26% for 100W-7 and 100W-16, respectively.
To reduce this systematic error, the origin position of the
From the ratio of the averaged value of the large-size absolute
SRDF measurement should coincide with that of the I ∗ (θ , φ)
IS photometer and that of the goniophotometer, we determine
measurement in the IS. In addition, the SRDF measurement
the relative difference between the two methods to be 0.33% needs to be performed at multiple positions within an extent
and 0.65% for 100W-7 and 100-16, respectively, which means comparable to the size of test lamp. The average of the multiple
that the large-size absolute IS photometer overestimates the measurements of SRDF minimizes the influence of the baffle-
total luminous flux compared with the goniophotometer by shadow mismatch between the test lamp and the beam scanner.
0.49% on average.
For validation of the small-size absolute IS photometer, 5. Summary
two 5 W test lamps designated as 5W-2 and 5W-6 are measured.
The drift of the test lamps during the comparison was as small We realized the absolute IS method for luminous-flux
as −0.12% and −0.11% for 5W-2 and 5W-6, respectively. The measurement in two different experimental setups. The first
relative difference of the small-size absolute IS photometer absolute IS photometer is based on a large-size sphere with
from the goniophotometer is determined to be −0.49% and a diameter of 2 m for high flux levels above 1000 lm, and the

Metrologia, 49 (2012) 273–282 281

Y-W Kim et al

Table 4. Comparison results and their uncertainties.

Measurement results/lm Comparison

Artefact Large-size Goniophotometer Small-size absolute Relative difference Comparison

S/N absolute IS (∅ 2 m) (GP) IS (∅ 0.3 m) from GP uncertainty (k = 2)
100W-7 1099.8 1096.4 0.33% 1.4%
1100.3
100W-16 1068.3 1062.7 0.65%
1071.0
5W-2 38.21 38.04 −0.49% 1.3%
38.00
5W-6 38.98 38.90 −0.25%
38.86

second is based on a small-size sphere with a diameter of 0.3 m References

for low flux levels below 100 lm. Both absolute IS photometers
are detector-based by measuring the luminous flux of the [1] 1989 The Measurement of Luminous Flux CIE Publication
external source with a standard photometer calibrated against No 84 (Vienna: International Commission on Illumination)
[2] 1996 The Photometry and Goniophotometry of Luminaires
the spectral responsivity scale of KRISS. The measurement CIE Publication No 121 (Vienna: International
uncertainties of the large-size and small-size absolute IS Commission on Illumination)
photometers are evaluated to be 1.2% and 1.1%, respectively, [3] Ohno Y 1995 New method for realizing a total luminous flux
as the expanded relative uncertainty at a level of confidence scale using an integrating sphere with an external source
of approximately 95% with a coverage factor of k = 2. J. IES 24 106–15
[4] Ohno Y 1996 Realization of NIST 1995 luminous flux scale
The dominant uncertainty components are the calibration using integrating sphere method J. IES 25 13–22
uncertainty of the standard photometer and the uncertainty of [5] Ohno Y 1998 Detector-based luminous-flux calibration using
the spatial mismatch correction. the absolute integrating-sphere method Metrologia
For validation of the realized absolute IS method, we 35 473–8
tested the equivalence of each absolute IS photometer with [6] Rastello M L, Miraldi E and Pisoni P 1996 Luminous-flux
measurements by an absolute integrating sphere Appl. Opt.
a goniophotometer by comparing the measurement results of
35 4385–91
two different types of test lamps. A set of 100 W tungsten [7] Hovila J, Toivanen P and Ikonen E 2004 Realization of the unit
lamps with a nominal luminous flux of 1100 lm is used for of luminous flux at the HUT using the absolute
comparison between the large-size absolute IS photometer and integrating-sphere method Metrologia 41 407–13
the goniophotometer, while a set of 5 W tungsten–halogen [8] Sauter G, Lindner D and Lindemann M 1999 CCPR Key
lamps with a nominal luminous flux of 40 lm is used for comparisons K3a of luminous intensity and K4 of luminous
flux with lamps as transfer standards PTB Report
comparison between the small-size absolute IS photometer PTB-Opt-62
and the goniophotometer. The comparison results verified [9] Park S, Lee D H, Kim Y W and Park S N 2008 Absolute
the equivalence within +0.49% and −0.37% on average for integrating sphere method for total luminous flux of LEDs
the large-size and small-size photometers, respectively, which Proc. 10th Int. Conf. on New Developments and
are smaller than the comparison uncertainties evaluated to be Applications in Optical Radiometry (NEWRAD)
pp 215–6
1.4% and 1.3% (k = 2) for the large-size and small-size [10] 1982 Methods of Characterizing the Performance of
photometers, respectively. The disagreement is assumed to Radiometers and Photometers CIE Publication No 53
mainly originate from the lamp-dependent screening loss for (Vienna: International Commission on Illumination)
the goniophotometer and the shadow-mismatch effect in the [11] Winter S, Lindermann M, Jordan W, Binder U and Anokhin M
spatial mismatch correction of the test lamp for the absolute 2009 Convenient integrating sphere scanner for accurate
integrating photometer. luminous flux measurements Metrologia 46 S248–51
[12] Ohno Y 1994 Integrating sphere simulation: application to
In conclusion, the realization of the KRISS luminous- total flux scale realization Appl. Opt. 33 2637–47
flux scale is successfully extended by the two new absolute [13] Shin D J, Lee D H, Park C W and Park S N 2005 A novel
IS photometers based on the same traceability. The inter- linearity tester for optical detectors using high-brightness
comparison among the absolute IS photometers and the light emitting diodes Metrologia 42 154–8
goniophotometer provides key information to identify the [14] Ikonen E and Hovila J 2004 Final report of
CCPR-K3.b.2-2004: bilateral comparison of illuminance
unknown systematic errors and to increase the accuracy of responsivity scales between the KRISS (Korea) and the
the scale. HUT (Finland) Metrologia 41 (Tech. Suppl.) 02003
[15] Hwang J, Lee D H, Park S, Kim Y W and Park S N 2009
Acknowledgment Measurement uncertainty evaluation for emission color and
luminance of displays Appl. Opt. 48 99–105
[16] Final report of APMP-K4, to be published at
This work is supported partly by the Korean Ministry of http://kcdb.bipm.org
Knowledge and Economy under the project of ‘Development of [17] Model# 01219-U, Welch Allyn Inc., http://www.welchallyn
a high-speed test handler for LED chips and packages,’ grant .com
10037396.

282 Metrologia, 49 (2012) 273–282