A22

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 20

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/23720406

Evaluation of expanded uncertainties in luminous intensity and illuminance


calibrations

Article · December 2008


DOI: 10.1364/AO.47.005829 · Source: PubMed

CITATIONS READS

3 4,870

1 author:

Ferhat Sametoglu
The Scientific & Technological Research Council of Turkey
46 PUBLICATIONS 226 CITATIONS

SEE PROFILE

All content following this page was uploaded by Ferhat Sametoglu on 13 July 2014.

The user has requested enhancement of the downloaded file.


Evaluation of expanded uncertainties in luminous
intensity and illuminance calibrations

Ferhat Sametoglu
TÜBİTAK-Ulusal Metroloji Enstitüsü (UME), Gebze, 41470 Kocaeli, Turkey, (ferhat.sametoglu@ume.tubitak.gov.tr)

Received 1 August 2008; accepted 12 September 2008;


posted 22 September 2008 (Doc. ID 99709); published 24 October 2008

Detector-based calibrating methods and expressions for calculation of photometric uncertainties related
to uncertainties in the calibrations of luminous intensity of a light source, illuminance responsivity of a
photometer head, and calibration factors of an illuminance meter are discussed. These methods permit
luminous intensity calibrations of incandescent light sources, luminous responsivity calibrations of
photometer heads, and calibration factors of illuminance meters to be carried out with relative expanded
uncertainties (with a level of confidence of 95.45%) of 0.4%, 0.4%, and 0.6%, respectively. © 2008 Optical
Society of America
OCIS codes: 120.0120, 120.5240.

1. Introduction value must be shown with a range of its reliability.


In photometry, light sources are calibrated by In other words, in the metrological point of view,
measuring their related photometric quantities, such the statement of the result of a calibration or a mea-
as the luminous intensity (cd), the illuminance (lx), surement is complete only if it contains both the
the luminous flux (lm), and the luminance (cd=m2 ). value attributed to the measurand and the uncer-
Photometry became a significant branch of optical tainty of calibration associated with that value.
metrology in the early 18th century and ever since it The photometric standard that is used by a labora-
has maintained a tradition that the unit of luminous tory (or a company) for giving calibration results
intensity has been regarded as the fundamental should be traceable to a national laboratory and it
photometric quantity among the photometric quanti- is also preferred that the laboratory should be accre-
ties and other derived units. Modern calibrations of dited by an accreditation body according to the
the photometric quantities of a light source require International Organization for Standardization/
using a calibrated photometer of known illuminance International Electrotechnical Commission (ISO/
responsivity (A=lx or V=lx) because the detector- IEC) standard 17025 [1]. This international standard
based technique is preferable due to its repeatability, contains the requirements for sound management
reproducibility, transportation convenience, and (Clause 4) and specifies the requirements for techni-
measurement convenience as compared to the cal competence for the type of calibrations (or tests)
source-based technique. the laboratory undertakes (Clause 4). As specified in
To provide international validity of a photometric Clause 5.46, a calibration laboratory shall have and
calibration, the calibration process should be done shall apply a procedure to estimate the uncertainty
according to international standards. In addition of measurement for all calibrations and types of
to this, calibrated values must be given according calibrations. In essence, uncertainty is a measure
to International System (SI) units, which are defined of the “trustworthiness” of the result of a measure-
ment [2]. The calculation of uncertainty for a photo-
by the meter convention by the General Conference
metric calibration is an effort to set reasonable
on Weights and Measures (CGPM), and the given
bounds for the calibration result according to stan-
dardized rules [3]. The ideal method for evaluating
0003-6935/08/315829-19$15.00/0 and expressing uncertainty in each the photometric
© 2008 Optical Society of America calibration should be capable of readily providing

1 November 2008 / Vol. 47, No. 31 / APPLIED OPTICS 5829


such an interval, in particular, one with a coverage necessary to analyze the calibration procedure more
probability or level of confidence that corresponds precisely and use an improved model equation, as
in a realistic way with that required. To estimate a given below, by taking into account all input quanti-
calibration uncertainty in a photometric quantity it ties that may affect the y, whether significant or not:
is necessary to use the Guide to the Expression of  
Uncertainty in Measurement (GUM) [4]. The GUM x1
y¼f
provides general rules for evaluating and expressing x2 ðx3 þ x4 Þ
uncertainty in most fields of physical measurements
and for use within standardization, calibration, la- × ð1  δx5  δx6  δx7  …  δxN Þ: ð2Þ
boratory accreditation, and measurement services.
The GUM uncertainty framework has been adopted The value of each input quantity xi and its stan-
by many organizations, is widely used, and has been dard uncertainty uðxi Þ , given in Eq. (2), is necessary
implemented in standards and guides on measure- to form the uncertainty budget of a calibration. The
ment uncertainty [5,6]. standard uncertainty associated with an estimate
To evaluate the uncertainty in a photometric cali- must have the same dimension as the estimate.
bration, all sources of uncertainty that influence the The main input quantities given within the first par-
calibration result and relations between each source entheses are generally obtained from statistic com-
should be known. Such relations are generally given putations. Values of other input quantities, which
in the uncertainty budget, which is evaluated using a are presented within the second parentheses of the
procedure depicted in Fig. 1. uncertainty model, are obtained from statistical ana-
A selected calibration method should, first, be in- lysis or can be extracted from other sources (litera-
ternationally acceptable in order to determine any of ture, books, articles, etc.). If there is no adequate
photometric calibration quantity (Y) properly and knowledge about the presented input quantity, the
trustworthy. Present methods published in interna- value for that input quantity is estimated to be zero.
tional, regional, or national standards are generally In general, knowledge about input quantities are
used as preferable methods in a calibration. Never- expressed by appropriate probability density func-
theless, laboratory-developed and validated methods tions in accordance with the Bayesian probability
may also be used if they are appropriate for the in- concept, based on the principle of maximum informa-
tended use [1]. tion theory and the Bayes theorem [7,8]. The stan-
In most cases a measurand Y is not measured di- dard uncertainty of each the input quantity uðxi Þ
rectly, but is determined from other influencing input is evaluated using a Type-A or a Type-B method,
quantities (X i ). The knowledge about the calibration which are defined as the method of evaluating the
process is to be represented by the so-called model uncertainty by the statistical analysis (Type A)
equation, which is a key element of modern uncer- and the method of evaluating the uncertainty by
tainty evaluation [4]. The model is formed on physical means of other than the statistical analysis (Type B).
or empirical grounds and establishes the mathemati- Type-A standard uncertainty is obtained from a
cal interrelation between the measurand Y and the probability density function derived from an ob-
input quantities X 1, X 2 ; …; X N according to the func- served frequency distribution, while Type-B stan-
tional relationship, as given as an example below: dard uncertainty is obtained from an assumed
probability density function based on the degree of
  belief that an event will occur, often called subjective
X1 probability. Standard uncertainty of an input quan-
Y¼f : ð1Þ
X 2 ðX 3 þ X 4 Þ tity uðxi Þ obtained from a statistical analysis has a
normal type of probability distribution and is esti-
Values of some input quantities given in the model mated from the experimental standard deviation
equation are taken from measurements performed sðxi Þ of the arithmetic mean of repeated measure-
during a calibration. These quantities are also used ments. As it is stated in the GUM, the standard un-
as basic quantities that generate the uncertainty certainty of an input quantity is equal to the
model. Estimated value (y) of the output quantity standard deviation of measurements if the standard
(Y) is then obtained by putting measured intrinsic va- deviation is known from n ≥ 10 former measure-
lues (x1 , x2 , x3 , and x4 ) of input quantities (X 1 , X 2 , X 3 , ments (uðxi Þ ¼ sðxi Þ). To show how many measure-
and X 4 ) into the model equation given in Eq. (1). ments (n) are considered for the estimation of such
The reliability of a calibration depends not only on an uncertainty evaluation, it is also necessary to de-
basic input quantities, but also on other important clare the degrees of freedom (vi ¼ n − 1). In uncer-
parameters that may affect calibration. They can tainty analyses other than the statistical analysis,
be classified as the long-term behavior of a reference the methods of rectangular, triangular, or U-shaped
material that is used in calibration, stability of in- probability distributions are used. Therefore, the de-
struments, environmental conditions, alignments, grees of freedom of the standard uncertainty uðxi Þ ob-
etc. These effects may be categorized into random tained from a Type B may be taken to be
and systematic effects. In order to sense the influ- vi → ∞ [4]. The rectangular distribution is a reason-
ences of these factors on the output quantity (y) and able default model in the absence of any other infor-
to estimate the output quantity more reliably, it is mation. But if it is known that the values of the

5830 APPLIED OPTICS / Vol. 47, No. 31 / 1 November 2008


quantity in question are more likely near the center by using “the law of propagation of uncertainty”
of the limits than close to the limits, a triangular or a method, which is known as the “root-sum-of-squares”
normal distribution may be a better model. On the method of combining uncertainty components. How-
other hand, if values close to the limits are more ever, in order to examine the standard uncertainty of
likely than values near the center, a U-shaped distri- an output quantity more reliably, it is important to
bution may be more appropriate. investigate the contribution coefficient uci ðyÞ of each
The standard uncertainty uðyÞ of an output quan- input quantity to the standard uncertainty. Thus, the
tity Y can now be roughly estimated using standard sensitivity coefficient (ci ) for each input quantity
uncertainties uðxi Þ of input quantities X i obtained should be calculated by taking the partial derivative
from Type-A or Type-B evaluations. This can be done of the model function developed for the uncertainty

Fig. 1. Standard procedure necessary to constitute the uncertainty budget.

1 November 2008 / Vol. 47, No. 31 / APPLIED OPTICS 5831


analysis [Eq. (2)] with respect to each input quantity calibration uncertainty by evaluation of the uncer-
(ci ¼ ∂f =∂xi ) and the contribution coefficient is then tainty budget are discussed here.
calculated by multiplying the sensitivity coefficient Calibration is described as the process of establish-
of each input quantity to its standard uncertainty ing the relationship between values of quantities in-
(uci ðyÞ ¼ ci × uðxi Þ). This process is true when the dicated by a measuring instrument or system and
input variables X 1 ; …; X N are mutually uncorre- the values provided by measurement standards [4].
lated. If two input quantities X i and X k are corre- Luminous intensity calibration of a light source is
lated to some degree (mutually dependent), then carried out at UME on a 6 m long optical bench sys-
the degree of correlation is characterized by the cor- tem by using a known standard calibration proce-
relation coefficient rðxi ; xk Þ and the covariance asso- dure [11,13,14]. The base of the bench consists of
ciated with the two estimates xi and xk have to be three, 2 m long marbles. The bench system contains
considered as an additional contribution to the un- (i) an enclosed cabinet with a special lamp alignment
certainty [4]. As a result, the standard uncertainty stage having 6° of freedom for geometric alignment of
associated with the measurement result uc ðyÞ is ob- a lamp under test (LUT), (ii) a rack containing digital
tained from the root sum square of all the uncer- electronic equipment (a voltmeter, a power supply,
tainty contributions. etc.) and a computer for operating and controlling
The reliability of the standard uncertainty as- the electrical parameters of the LUT, (iii) a 6 m long
signed to the output estimate should be declared. rail system with a movable carriage having an align-
In metrology, the reliability of the standard uncer- ment stage that is mounted on the marbles for glid-
tainty of a calibration is represented by an expanded ing the working standard photometer from the light
uncertainty (U ¼ kuC ðyÞ), which is determined from source up to 6 m, (iv) a set of matte-black painted
the product of the coverage factor (k) for the valid baffles between the lamp and the photometer for re-
probability distribution and the standard uncer- ducing the stray light, (v) a working standard photo-
tainty uc ðyÞ. It also defines a coverage interval meter head (WSPH) with a photocurrent meter, and
y − U ≤ Y ≤ y þ U (or Y ¼ y  U) that contains ex- (vi) a laser diode and side-viewing transits for align-
actly the specified proportion of probability distribu- ment of the LUT and electronic equipment, such as
tion (the shortest interval between two quantities of the WSPH, the photometer head under test (PUT), or
distribution for the output quantity values) or level of luxmeters (Fig. 2).
confidence p [9]. The coverage factor is determined by The calibration procedure was started by mount-
the effective degrees of freedom (veff ) of the standard ing the LUT to the lamp holder in the base-down
uncertainty, which depend on the size of the sample position and its orientation and position were accom-
on which it is based (vi ), using the Welch– plished by aligning the lamp so the lamp posts were
Satterthwaite formula [4,6,10]. Tabulated values held vertically and the plane formed by the axes of
that show relations among the coverage factor k, the posts was perpendicular to the optical axis of
the effective degree of freedom veff , and the level of the photometer. Then, the LUT was operated at a
confidence p are given in Annex G of the GUM docu- specified color temperature (2856 K) and electrical
ment [4]. If the coverage factor is k ¼ 2, expanded un- polarity (Fig. 3). After that, the WSPH was glided
certainty will be U ¼ 2uC ðyi Þ, which defines an to a distance of 4 m and the photocurrent generated
interval having a level of confidence of p ¼ ∼95% at the output of the WSPH was measured. Finally,
and p ¼ 99% for U ¼ 3uC ðyi Þðk ¼ 3Þ. the average photocurrent of repeated measurements
The scope of this paper is as follows. In 2003, a new (Pc ), the illuminance responsivity of the WSPH (sv ),
candela was realized at Ulusal Metroloji Enstitüsü and the lamp-to-photometer distance (d) were used
(UME) using a set of filter radiometers with calibra- to determine the luminous intensity (I x ðTÞ) of the
tions of their responsivities based on the cryogenic LUT according to the inverse-square law of illumina-
radiometer, and the old UME standard lamps-based tion by using the following equation:
unit was replaced by the new detector-based unit
[11,12]. Therefore, the photometric calibration proce-
dures have been revised to utilize the detector-based Pc
methods. Descriptions of the detector-based lumi- I T ðTÞ ¼ ccf · d2 ; ð3Þ
sv
nous intensity and illuminance calibration proce-
dures, evaluations of calibration uncertainties, and
established uncertainty budgets are presented in
Sections 2 and 3. where ccf is the color-correction factor of the WSPH.
Measurements were generally repeated six times by
closing and operating the lamp again and the lumi-
2. Uncertainty Evaluation in the Luminous Intensity nous intensity value of the LUT was calculated by
Calibration taking their average.
Calibration of an Osram Wi41/G lamp was the most Calibration equations, known also as physical
popular calibration given by the laboratory on the lu- models, are used in a wide variety of applications
minous intensity calibration. Therefore, the lumi- in measurement science and they are usually helpful
nous intensity calibration of this lamp, operated at a in determining the uncertainty model of calibrations
color temperature of 2856 K, and estimation of the [15]. A univariate (scalar) uncertainty model is

5832 APPLIED OPTICS / Vol. 47, No. 31 / 1 November 2008


Fig. 2. Luminous intensity and the illuminance calibration setups. WSPH and PUT are the working standard photometer head and
photometer under test, respectively, and WL and LUT are the working lamp and lamp under test, respectively.

 m
developed by taking into account all input quantities ðcDVM þ cDVMD Þ · U L i

that may affect the luminous intensity value, I x ðTÞ ¼


U L ðJÞ · Rð1 þ RD þ αRS ΔTÞ
whether significant or not. The uncertainty model  
ðPc þ cPCM þ cPCMD Þ
developed on the basis of the calibration equation × ðd þ dp þ αSR ΔTÞ2 ×
for estimating the uncertainty in the luminous inten- ðsV þ svD Þ
sity calibration of a light source is × ccf × scf × M R × ð1 − Laε − Laφ − Paθ − PN Þ: ð4Þ

Fig. 3. (a) Schematic drawing of the lamp holder and connections for the measurement of electrical parameters, (b) assembled lamp
holder with a lamp. PS is the power supply, U L is the lamp voltage, R is the shunt resistor, and U L ðJÞ is the voltage drop across the
shunt resistor.

1 November 2008 / Vol. 47, No. 31 / APPLIED OPTICS 5833


The first parentheses show estimated uncertainty nance responsivity (sV ) and short-term drift (sVD )
sources in the current stability of the LUT. It of the WSPH. The last parentheses show other fac-
includes uncertainties in (i) calibration factor tors (supplementary input quantities) that affect
(cDVM ) of the digital voltmeter (DVM) that is used the luminous intensity (output quantity). It includes
for measuring the LUT current and its short-term uncertainties in (i) vertical (Laε ) and horizontal (Laφ )
drift (cDVMD ) between recalibrations; (ii) the voltage alignments of the lamp filament, (ii) angular align-
drop (U L ) across the reference shunt resistor (R) that ment of the WSPH (Paθ ), and (iii) the nonlinearity
is used for monitoring the LUT current; (iii) the op- (PN ) of the WSPH. There are also three correction
erating current of the LUT (U L ðJÞ) assigned for a col- factors in the uncertainty model that are used for col-
or temperature of 2856 K; (iv) the resistance value or correction (ccf ), stray-light correction (scf ) and
(R), short-term drift (RD ), and the temperature coef- measurement repeatability (M R ).
ficient (αRS ΔT) of the reference shunt resistor; and Table 1 gives the general outline of uncertainty
(v) the exponent for changes of lamp current affecting sources in the luminous intensity calibration of an
the luminous intensity (mi ). The second parentheses aged 180 W tungsten filament lamp of type Osram
show the estimated uncertainty components in the Wi41/G performed at UME, which make up our gen-
distance measurement between the LUT and the eric uncertainty budget. They are divided into two
WSPH. They are (i) the measured photometric dis- main groups: those related to the main factors of
tance (d) between the center plane of the lamp fila- the modeling equation and those caused by the other
ment and the outermost surface of the WSPH, (ii) the estimated factors.
offset distance (dp ) between the outermost surface of Description of both the estimated value and its
the WSPH and the center of its aperture, and (iii) the standard uncertainty for each input quantity are gi-
thermal expansion coefficient (αSR ΔT) of the stain- ven in Subsections 2.A–2.E and evaluation of the ex-
less steel ruler that is mounted on the rail and used panded uncertainty is given in Subsection 2.F.
for the photometric distance measurement. The third
parentheses show uncertainty sources in the illumi- A. Current-Induced Uncertainty Sources
nance measurement of the lamp at the WSPH A specially designed lamp holder is used for operat-
surface. It includes uncertainties in (i) the photocur- ing the LUT at a fixed polarity (Fig. 3). The holder
rent reading (Pc ), (ii) the calibration factor (cPCM ) of has an E27 lamp socket and four separate contacts,
the digital photocurrent meter (PCM) that is used for two for the power supply and two for voltage mea-
measuring the photocurrent and its short-term drift surements. The lamp was operated with a dc power
(cPCMD ) between recalibrations, and (iii) the illumi- supply (under dc constant-current control) and at

Table 1. Uncertainty Budget Established at UME for the Luminous Intensity Calibration of an Osram Wi41/G Light Source

ui ðyÞ hðy; xi Þ
Xi xi uðxi Þ pdf ci (cd) (%)
cDVM 0.999992 0.000004 NTD 1762:65 cd 0.00705 0.02
cDVMD 0.000001 0.0000004 RTD 1762:65 cd 0.00071 0.0002
UL 0:584720 V 0:0000036 V NTD 3014:50 cd=V 0.01085 0.04
U L ðJÞ 5:8472 A 0:0006 A NTD −301:45 cd=A −0:18087 10.3
R 0:099994 Ω 0:0000020 Ω NTD −17627:44 cd=Ω −0:03525 0.4
RD −0:0000010 0.0000004 RTD −1762:64 cd −0:00071 0.0002
αRS ΔT 0.0 0.0000023 RTD −1762:64 cd −0:00407 0.01
mi 6.24 0.14 RTD −0:00662 cd 0.00096 0.0003
d 4:000 m 0:00018 m RTD 141:16 cd=m 0.02541 0.2
dP 0:00212 m 0:000021 m RTD 141:16 cd=m 0.00296 0.003
αSR ΔT 0.0 0.0000078 RTD 141:16 cd 0.00110 0.0004
Pc 221:12 nA 0:024 nA NTD 1:27 cd=nA 0.03007 0.3
cPCM 1.00033 0.00010 RTD 1:27 cd 0.00013 0.00001
PmD −0:000030 0.000012 NTD 1:27 cd 0.00002 0.0000001
sv 12:592 nA=lx 0:019 nA=lx NTD −22:44 cd · lx=nA −0:42391 56.3
svD −0:0006 0.0024 RTD −22:44 cd −0:05386 0.9
ccf 1.0000 0.0005 RTD 282:47 cd 0.14124 6.3
scf 0.9992 0.0004 NTD 282:70 cd 0.11308 4.0
MR 0.9998 0.00006 NTD 282:54 cd 0.01783 0.1
Laε 0:0° 0:00064° RTD −282:47 cd=° −0:18103 10.3
Laφ 0:0° 0:00064° RTD −282:47 cd=° −0:18103 10.3
Pθ 0:0° 0:00012° RTD −282:47 cd=° −0:03262 0.3
PN 0.0 0.00012 RTD −282:47 cd −0:03390 0.4
I x ðTÞ 281:20 cd combined uncertainty, k ¼ 1 0:56 cd
effective degree of freedom, veff >100
expanded uncertainty, k ¼ 2 1:13 cd
expanded uncertainty, k ¼ 2 0.4%

5834 APPLIED OPTICS / Vol. 47, No. 31 / 1 November 2008


fixed polarity. The lamp voltage was measured di- [uðU L ÞÞ ¼ 0:0000036 V] assuming a normal type of
rectly between two contacts, whereas the lamp distribution (NTD), where the degree of freedom of
current was measured as the voltage across a cali- this estimate is equal to vV ¼ 19. The DVM used
brated standard shunt resistor [11]. for monitoring the lamp current has been calibrated
A high-precision PTN55 125-20 series dc power by the Voltage Standards Laboratory of UME. The
supply, manufactured by a Heinzinger Electronics calibration has been performed by comparing read-
GmbH, is used for operating the LUT at constant- ings from the DVM with the reference calibrator
current mode. It provides dc up to 20 A and dc voltage (Fluke 5720 A). The calibration certificate stated
up to 125 V. The LUT has a flat filament with an in- that the calibration factor of the DVM for the voltage
ert gas inside. The properties of incandescent lamps level of 1 V is found to be cDVM ¼ 0:999992 with a re-
are not very sensitive to thermodynamic conditions, lative expanded uncertainty of 0.000008 (k ¼ 2).
such as ambient temperature, air pressure, speed of Thus, the standard uncertainty assuming a NTD
moving air, and humidity [16]. The lamp current is (vi → ∞) and calculated as uðcDVM Þ ¼ ð0:999992×
automatically controlled by a computer feedback sys- 0:000008=2Þ ¼ 0:000004. The DVM is recalibrated
tem to keep the current variation within 0:002%. annually, utilizing the same reference calibrator-
The operating current required to achieve the color based calibration setup. The drift of the calibration
temperature of the lamp at 2856 K has been deter- factor of the DVM since its last calibration was esti-
mined as 5:8472 A  0:0006 A (coverage factor k ¼ mated from its calibration history to be þ0:000001
1) by repeated measurements of the spectral irradi- with a standard deviation of sðcDVMD Þ ¼ 0:0000007.
ance of the lamp [11,17]. The lamp current is ramped Thus, the standard uncertainty of uðcDVMD Þ assum-
p a RTD and estimated as uðcDVMD Þ ¼ sðcDVMD Þ
up slowly (within 1 min) to the specified value and ing
allowed to stabilize (typical warm-up time is 15 min 3 ¼ 0:0000004.
for incandescent lamps). The operating current of the A precision four-terminal standard dc current
LUT is monitored via the voltage drop across a cali- shunt operating in air, Guildline Instruments
brated dc current shunt (0:1 Ω) using a calibrated 9230/15, was used to monitor the current of the
Hewlett–Packard 3456A DVM. A digital multimeter LUT. The terminations of the shunt are specially de-
(Datron Wavetek 1271) is also used for monitoring signed for giving low thermal emfs. The temperature
the LUT voltage. Data acquisition and measure- coefficient of the reference resistor is stated by the
ments are controlled by a computer program via manufacturer as 4 × 10−6 =°C. The environmental
an IEEE-488 bus. A control program written in Lab- conditions of the laboratory are stable and controlled
View 6.0 is used to monitor the lamp current, voltage, over time (the temperature of the laboratory, for ex-
and standard deviation in the current setting during ample, is 23 °C  1 °C). Thus, the resistance varia-
the whole calibration period. As soon as the measure- tion of the reference shunt resistor due to a
ment is finished, the electrical parameters of the temperature variation (1 °C) is estimated to be
LUT are stored and the lamp current is ramped down within 0:000004 Ω using the known R ¼ R0 þ αðT −
and is, finally, turned off. T 0 Þ formula, which causes the standard uncertainty
The voltage drop across the reference shunt resis- of uðαΔTÞ ¼ 0:0000023 assuming a RTD. The resis-
tor during the calibration process was obtained from tance value of the calibrated dc current shunt is R ¼
the arithmetic mean of 20 readings (v ¼ 19) as 0:099994 Ω at the specified lamp current. The resis-
0:584720 V (Table 2). Each reading was obtained tor has been calibrated with a relative expanded un-
by subtracting the offset voltage (U L0 ) of the DVM certainty of 0:000040 Ω (k ¼ 2) by the Impedance
from the reading (U L ¼ U LR − U L0 ). Thus, the rela- Standards Laboratory of UME at the specified refer-
tive standard deviation of repeated measurements ence temperature of 23 °C. The calibration has been
(n ¼ 20) is used as the standard uncertainty performed by direct substitution using an 8½-digit
precision digital multimeter (HP 3458 A) and a cali-
Table 2. Lamp Voltage, Photocurrent, and Stray Light Measurement brated four-terminal standard dc current shunt
Results (Guildline 9211 A) of the same nominal value. Thus,
Lamp Stray the standard uncertainty in the resistance value was
Measurement Voltage Photocurrent Light estimated as uðRÞ ¼ ð0:099994 × 0:000040=2Þ ¼
№: n UL Pc scf 0:0000020 Ω. The current shunt is recalibrated an-
(V) (nA) nually utilizing the same reference dc current
1 0.584718 221.16 0.9987
shunt-based calibration setup. The drift of the resis-
2 0.584712 221.10 0.9991 tance value of the shunt since its last calibration was
3 0.584723 221.09 0.9989 estimated from its calibration history to be
4 0.584720 221.15 0.9992 −0:0000010 with a standard deviation of sðRD Þ ¼
… … … … 0:0000007. Thus, the standard uncertainty of uðRD Þ
18 0.584719 221.13 0.9993 assuming
p a RTD and estimated as uðRD Þ ¼ sðRD Þ
19 0.584720 221.09 0.9996 3 ¼ 0:0000004.
20 0.584717 221.11 0.9990 The spectral distribution and brightness of a tung-
Mean 0.584720 221.12 0.9992
sten or a tungsten halogen lamp are functions of
sðU L Þ 0.0000036 0.024 0.0004
the filament temperature and, hence, the electrical

1 November 2008 / Vol. 47, No. 31 / APPLIED OPTICS 5835


current flowing through the filament. It is known measurement stage, which is equal to the distance
that current changes affect short-wavelength (0.2% from the outermost surface of the WSPH and its
at 400 nm) radiation more than long-wavelength aperture plane (acceptance are of the WSPH), has
(0.1% at 800 nm) radiation [18]. An approximate re- been measured at the Dimensional Standards
lation for determining the change in the luminous in- Laboratory of UME using a Coordinate Measuring
tensity of a lamp due to a current change in the lamp Machine and found to be 2:12 mm  0:02 mm
is ðI=I R Þ ≈ ðJ=J R Þmi, where I and J are the changed (k ¼ 2), which causes a standard uncertainty of
luminous intensity and current and I R and J R are the uðdp Þ ¼ 0:000021 m assuming a NTD.
rated values of luminous intensity and current, re-
spectively [19,20]. The exponent in the relation is C. Photometer-Induced Uncertainty Sources
found to be 6:24  0:25 assuming an interval of The luminous intensity value of the LUT was
RTD, which form the standard uncertainty of measured by using a thermostabilized commercial
uðmi Þ ¼ 0:14. This means that an increase in current WSPH, manufactured by PRC Krochmann GmbH
of 1 mA in a lamp operating at 5:8572 A would give (TH15BA). The WSPH consists of a well-defined in-
ð5:8582=5:8572Þ6:24 ¼ 1:0011 or a change of about put aperture having a nominal area of 0:5 cm2 that
0.11% in luminous intensity. defines the solid angle and the distance measure-
ment plane, the VðλÞ matching filter, and a Si photo-
B. Distance-Induced Uncertainty Sources diode. All these components are assembled in a
In photometry, the area of the limiting aperture of temperature-controlled cylindrical housing. Inside
the photometer head and the lamp-to-photometer the housing, the temperatures of the Si photodiode
distance define the solid angle over which light is col- and the filter are kept constant at 35 °C during op-
lected. On the other hand, the intensity of light or eration to eliminate the influence of the surrounding
other linear waves radiating from a point source (en- temperature. The temperature regulation occurs
ergy per unit of area perpendicular to the source) is within the photometer head with special control elec-
inversely proportional to the square of the distance tronics. The manufacturer’s specifications state
from the source, which is known as the inverse- that the temperature coefficient of the WSPH is
square law. The departure from the inverse-square f 6 < 0:2%.
law at short distances has influence on the accuracy The illuminance responsivity (sV ) of the WSPH is
of calibration. Therefore, the photometric distance sv ¼ 12:592 nA=lx. The photometer head has been
for the luminous intensity calibration should be 10 calibrated using the new detector-based UME illumi-
times more than the dimensions of the lamp. The dis- nance scale with an expanded uncertainty of 0.30%
tance in the calibration was aligned exactly to 4 m by (k ¼ 2) [21]. Thus, the standard uncertainty in the
measuring it in two stages: (i) the distance between illuminance responsivity value was estimated as
the filament plane of the lamp and the outermost uðsv Þ ¼ 0:019 nA=lx. The WSPH is recalibrated an-
surface of the photometer (d) and (ii) the distance off- nually by comparing it with the absolutely calibrated
set of the photometer (dp ). The first stage is mea- filter radiometers constructed for the realization of
sured mechanically by using a calibrated stainless the candela [22]. The drift of the illuminance respon-
steel ruler attached to the bench. The overall length sivity value of the WSPH since its last calibration
of the ruler is 5 m and it has been calibrated at sev- was estimated from its calibration history to be
eral points by comparing it with a reference bench −0:006 nA=lx (∼0:05%) with a standard deviation
system having a calibrated line scale of 6 m, which of sðsvD Þ ¼ 0:0042. Thus, the standard uncertainty
has been developed by the Dimensional Standards of the drift value p assuming a RTD and estimated
Laboratory of UME. The calibration certificate sta- as uðsvD Þ ¼ sðsvD Þ= 3 ¼ 0:0024 nA=lx.
ted that the ruler was calibrated with an expanded
p The photocurrent generated at the output of the
uncertainty formula of Uðk ¼ 2Þ ¼ 0:052 þ WSPH was measured with a PRC-321 model digital
ð0:005 × LÞ mm. This means that the uncertainty
2
photocurrent-meter (PCM). A mechanical shutter
in the length scale is 4:000 m  0:054 mm (k ¼ 2). was used in front of the WSPH for taking real sig-
The worst case was considered for uncertainty ana- nals. The measured photocurrent was the average
lysis (∼0:3 mm) because it was assumed that there of 20 readings (v ¼ 19) taken after the stabilization
may be a deviation in the thickness of the line show- of the LUT current (Table 2). Each reading was per-
ing the distance on the ruler. Thus, standard uncer- formed with the shutter opened (Pso ) and closed
tainty for the distance measurement was assuming p a (dark signal, Psc ). In order to obtain the real photo-
RTD and estimated as uðdÞ ¼ 0:0003= 3 ¼ current value (Pc ), the dark signal was subtracted
0:00017 m. The thermal expansion coefficient from the reading (Pc ¼ Pso − Psc ). The standard un-
(αSR ΔT) of the stainless steel ruler is stated by the certainty of the mean photocurrent reading of
manufacturer as 17 × 10−6 =°C. Thus, the length 221:12 nA was obtained as uðPc Þ ¼ 0:024 nA, which
variation of the stainless ruler due to a temperature is the standard deviation of 20 readings.
variation within the laboratory is estimated to be The PCM has been calibrated by Impedance Stan-
within ∂L=∂T ¼ 0:000014 m, which causes the neg- dards Laboratory of UME. A Decat resistor (Quad
ligible standard uncertainty of uðαSR ΔTÞ ¼ Tech 1433 W), a DVM (HP 3458 A), a voltage calibra-
0:0000078 assuming a RTD. The second distance tor (Fluke 5500 A), and reference resistors (Guildline

5836 APPLIED OPTICS / Vol. 47, No. 31 / 1 November 2008


9330=10 M and 9330=100 k) had been used in the depend on the spectral power distributions of several
calibration. Desired currents from the calibrator types of light sources, including tungsten-filament
and reference resistor combination had been applied incandescent, fluorescent, high-pressure sodium,
to the input of the PCM and the calibration has been and metal-halide light sources, and white/colored
performed by taking current readout from the PCM light-emitting diodes have been investigated [23].
display. The calibration certificate stated that the The ccf of the WSPH is equal to unity for the Osram
calibration factor of the PCM for the current level Wi41/G light source operated at 2856 K. The uncer-
from 100 nA to 500 nA is found to be cPCM ¼ tainty of the ccf has been calculated as uðccf Þ ¼
1:00033 with a relative expanded uncertainty of 0:0005 using the square-root of the sum of the
0.02% (k ¼ 2). Thus, the standard uncertainty squares of the two relative spectral measurements
assuming a NTD (vi → ∞) and it is calculated on the detector and the lamp, as well (relative spec-
as uðcPCM Þ ¼ ð1:00033 × 0:00020=2Þ ¼ 0:00010. The tral responsivity of the WSPH and the spectral dis-
PCM is recalibrated annually utilizing the same tribution of the LUT) [23].
calibration procedure. The drift of the calibration fac- The second correction was the scf (unwanted light),
tor of the PCM since its last calibration was esti- which is more problematic when a detector-based
mated from its calibration history to be −0:000030 method is used. The stray light is commonly mani-
with a standard deviation of sðcPMCD Þ ¼ 0:000021. fested as random scatter or directional reflection of
Thus, the standard uncertainty of the drift value the light from its surroundings. The stray light is
p
was estimated as uðcPMCD Þ ¼ sðcPMCD Þ= 3 ¼ generally obstructed by using a set of baffles between
0:000012 assuming a RTD. the light source and the photometer (Fig. 4).
The baffles were placed so that when the lamp is
D. Correction-Induced Uncertainty Sources viewed with naked eyes in any bench position and
There are three corrections that were used within photometer circumference, no stray light can be seen.
the uncertainty modeling. The first correction was At the front surface of the WSPH, a screen with a cir-
the ccf, which is a function of the spectral power dis- cular beveled aperture of diameter 25 mm (a little bit
tribution of a light source. It is known that an error more than the field of view of the WSPH) was used to
occurs when the photometer measures a light source mask stray rays and to determine the scf. The
having a spectral power distribution different from amount of stray light has been calculated as
the calibrated source (CIE Illuminant A). The varia- 0.9992 by taking the mean ratio of photometer sig-
tions of the illuminance responsivities of the WSPH nals at lamp-off and lamp-on conditions (Table 2).

Fig. 4. (a) Stray light measurement setup showing alignment procedures for the lamp filament and photometer head. (b) Perpendicularity
alignment of the WSPH.

1 November 2008 / Vol. 47, No. 31 / APPLIED OPTICS 5837


Thus, the experimental standard deviation of re- perpendicular to the laser beam, where the second
peated measurements (n ¼ 20, vV ¼ 19) was used one was set up on the optical axis and used for the
as the standard uncertainty of stray light measure- vertical alignment. Each transit has a precision
ments (uðscf Þ ¼ 0:0004) assuming a NTD. cross-axis telescope having a glass reticule with a
The luminous intensity measurements were re- filar/bifilar pattern. Perpendicularity alignments of
peated six times by reoperating the LUT each time the filament at both axes were performed using ob-
to guarantee the measurement reliability. It is served images of the filament from the transits. The
known that the repeatability is the closeness of manufacturer’s specifications stated that each the
the agreement between the results of successive transit has a resolution of 4:0 arc sec( 0:001°). This
measurements of the same measurand carried out means that the alignments of the filament of the
under the same conditions of measurement. Mea- LUT were estimated to be 0° with limits 0:001°.
surement results are shown in Table 3. Thus, the standard uncertainty of each alignment
As can be seen from Table 3, there is good agree- was estimatedp assuming a RTD as uðLaφ Þ ¼
ment between the results (≤0:03%). The maximum uðLaε Þ ¼ 0:001= 3 ¼ 0:00064°.
luminous intensity within repeated measurements The perpendicularity alignment of the photometer
was 281:71 cd, which was used for normalizing mea- head was performed before alignment of the lamp to
surement results to unity. The repeatability was then the optical axis. The photometer head was fixed on a
calculated by taking the arithmetic mean of normal- movable carriage. The perpendicular alignment of
ized luminous intensities (M R ¼ 0:99978). The ex- the photometer head was performed by using a laser
perimental standard deviation of the normalized diode having a 1 mm output aperture that was also
results obtained under repeatable conditions (n ¼ 6) used for determining of the optical axis. The surface
was calculated as sðM R Þ ¼ 0:00015. Thus, the stan- of the WSPH was closed with a lid having an align-
dard uncertainty of the repeatability measurements ment mirror and then the laser beam was reflected
was estimated from the pooled estimate
p of standard
p back from the mirror. The maximum error in the fall-
deviation as (uðM R Þ ¼ sðM R Þ= n ¼ 0:00015= 6 ¼ ing of the laser beam onto the output point of the
0:00006). laser beam will inevitably/unavoidably be 0:5 mm
E. Other Uncertainty Sources because the aperture used in front of the laser diode
has a diameter of 1 mm, causing back-reflection
Alignment of a lamp or a detector is also a critical [Fig. 4(b)]. The maximum error in adjustment of
point in optical measurements because their center the perpendicularity of the photometer head and
points must be at the same optical axis and their the laser diode aperture takes place Pθ ¼
planes must be perpendicular to each other. Both 2 × arctanð5000=0:5Þ ¼ 0:0002 due to the fact that
the LUT and the WSPH are mounted on positioning
the distance between them is 5 m. Thus, the standard
stages that allow adjustments of the rotations, hori-
uncertainty of the photometer head alignment
zontal and vertical positions, and height. The laser
was estimated
p assuming a RTD as uðPθ Þ ¼
diode was used behind the LUT to define the optical
0:0002= 3 ¼ 0:00012°. The same type of side-
axis and the center of the lamp filament, which had
viewing transit was also used to check the alignment
been aligned parallel to the optical axis [22]. After
of the WSPH.
that, the alignment of the LUT filament was started.
A detector is said to be linear if its output signal is
The LUT was mounted to the holder on the photo-
directly proportional to the incident power. The lin-
metric bench in the base-down position. The horizon-
tal (Laφ ) and vertical (Laε ) alignments of the filament earity of the WSPH has been measured using a flux-
were performed using a side-viewing precision tran- addition technique [21]. Measurements had been
sit (Brunson, 76-RH190), shown in Fig. 4(a). The first performed between the photocurrents of 2 nA and
transit was set up and aligned along a horizontal axis 2 mA by applying different neutral density filter com-
binations to the beam ways. To reduce the effects of
random noise, the measurements were repeated at
Table 3. Repeatability Results of the Luminous Intensity Calibrations least ten times at each photocurrent level. Measure-
Luminous ment results indicated that the WSPH is linear over
Intensity an output current range of 10−9 A to 10−4 A, with a
I x ðTÞ standard uncertainty of 0.012% (k ¼ 1), which corre-
Measurement Luminous Intensity sponds to an illuminance range of 8 × 10−2 lx to
№: n (cd) Normalized Ixnor ðTÞ 8 × 103 lx. Therefore, the WSPH can be used to mea-
1 281.22 0.9997 sure the luminous intensity of a light source between
2 281.25 0.9998 1:3 cd and 1:3 × 105 cd at 4 m.
3 281.29 0.9999
4 281.20 0.9996 F. Evaluation of the Expanded Uncertainty
5 281.31 1.0000 The next step for forming the uncertainty budget,
6
P
281.22 0.9997 after determining the standard uncertainties uðxi Þ,
Mean value M R ¼ n1 ni¼1 I xnori ðTÞ 0.9998 is to calculate the sensitivity coefficient (ci ) of each
Standard
pP deviation sðM R Þ ffi¼
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 0.00015
n ðI
input quantity (X i ) by taking the partial derivative
k¼1 xnori ðTÞ − M R Þ
1
n−1 of the model function f [Eq. (4)] with respect to the

5838 APPLIED OPTICS / Vol. 47, No. 31 / 1 November 2008


input quantity (Table 1). The ci coefficients are essen- same plane by using a digital height gauge. After this
tially conversion factors that allow one to convert the study, the lamp was mounted to the lamp socket lo-
units of an input quantity into the units of the mea- cated on the photometric bench and then operated.
surand and express mathematically how much f After the lamp had stabilized (∼15 min after turn-
changes given an infinitesimal change in xi . A de- ing on), photocurrents generated at the outputs of
tailed description for calculating the ci coefficient the WSPH and the PUT (PcR and PcT , respectively) at
for each input quantity is described in Appendix A. various photometric distances were measured using
Afterward, the uncertainty contribution for each a substitution principle having a substitution
quantity ui ðyÞ was calculated and then the combined scheme of reference–test–test–reference. Instead of
standard uncertainty uc ðyÞ was calculated, based on the PUT, the lux values (ET ) were taken directly from
the procedure given in Fig. 1, as uc ðI T ðTÞÞ ¼ 0:56 cd the read out of the illuminance meter under test
by taking the square root of the sum of the uncer- (IMUT). The illuminance responsivity of the PUT
tainty contributions. The expanded uncertainty of (svT ðTÞ) and the calibration factor of the IMUT (kðTÞ)
the luminous intensity calibration was estimated as were then calculated using the following equations:
UðI T ðTÞÞ ¼ 1:13 cd with a level of confidence of
95.45% by multiplying the uc ðI T ðTÞÞ by a coverage PcT
factor k ¼ 2 (veff > 100). This estimation shows that svT ðTÞ ¼ svR ðTÞ · · ccf · ucf ; ð5Þ
PcR
the expanded uncertainty in the luminous intensity
calibration within the interval from 1:3 cd to 1:3 ×
PcR
105 cd is 0.4%. E s ðTÞ
kðTÞ ¼ R ¼ vR · ccf · ucf ; ð6Þ
ET ET
3. Uncertainty Evaluations in the Illuminance
Responsivity and Illuminance Calibrations where ucf is the uniformity of the lamp illumination
The same detector-based calibration facility, which is at the measured plane.
shown in Fig. 2, was used for calibrations of a com- Univariate (scalar) uncertainty models [Eqs. (7)
mercial photometer head (LMT P30SCT) and an illu- and (8)] for both the PUT and the IMUT calibrations
minance meter (LMT B360) and to demonstrate were developed by taking into account all input
estimation of calibration uncertainties in illumi- quantities that may affect the calibrations, whether
nance calibrations. An Osram Wi41/G lamp operated they are significant or not:
at a color temperature of 2856 K was used as a work-  m
ing lamp in both calibrations. The detector-based ðcDVM þ cDVMD Þ · U L i
svT ðTÞ ¼
method has a greater advantage in the illuminance U L ðJÞ · Rð1 þ RD þ αRS ΔTÞ
calibrations. With this method, photometers and lux-  
PT
meters are calibrated by direct substitution with the × ðsV þ svD Þ × × ccf × scf
standard photometer by placing them on the same ðPc þ cPCM þ cPCMD Þ
illuminated plane. In such substitution methods, × ucf × M R × ð1 − PθR − PθT − RPa − PNR Þ; ð7Þ
many uncertainty factors, such as distance measure-
ments, lamp alignments, and the departure from the
inverse-square law, are canceled out [24].
To perform accurate illuminance calibrations, the  m
ðcDVM þ cDVMD Þ · U L i
reference plane of a calibrated device must be known kðTÞ ¼
precisely. When the reference plane is not correctly U L ðJÞ · Rð1 þ RD þ αRS ΔTÞ
 
defined, the deviation from the inverse-square law Pc þcPCM þcPCMD
causes the responsivity of the photometer to vary de- sV þsvD
pending on the distance to the source. The reference × × ccf × scf × ucf × M R
ET
plane plays a more important role if calibrations are
performed at shorter distances from the light source × ð1 − PθR − IM φ − RPa − PNR Þ; ð8Þ
[25]. It is not difficult to define the reference plane for
those photometers that have thin, flat diffusers. where PNR is the uncertainty in the nonlinearity of
However, for an illuminance meter equipped with the WSPH, RPa is the uncertainty in the reference
a dome-shaped or mesa-shaped diffuser, it is usually plane determination, and PθR , PθT , and IM φ are un-
difficult to define the correct reference plane and, certainties in the alignments of WSPH, PUT, and
therefore, the plane is usually chosen to be at the IMUT, respectively.
top of the diffuser [24,25]. Tables 4 and 5 give general outlines of uncertainty
Before both types of calibration, both the WSPH sources in the illuminance responsivity calibration of
and the PUT were placed on the linear translation a LMT Lichtmesstechnik GmbH manufactured
stage that acts perpendicular to the optical axis photometer head (P30SCT) and the calibration factor
and perpendicularity alignment of each the device of a LMT B360 type of illuminance meter.
was performed (a detailed procedure for the align- Descriptions of estimated values and their stan-
ments is given in Subsection 2.E). Then. the refer- dard uncertainties in both illuminance calibrations
ence planes of both devices were aligned to the are given in Subsections 3.A–3.D. It should be noted

1 November 2008 / Vol. 47, No. 31 / APPLIED OPTICS 5839


Table 4. Uncertainty Budget Established at UME for the Illuminance Responsivity Calibration of a LMT P30SCT Photometer Head

ui ðyÞ hðy; xi Þ
Xi xi uðxi Þ pdf ci (nA=lx) (%)
cDVM 0.999992 0.000004 NTD 118:95 nA=lx 0.000476 0.02
cDVMD 0.000001 0.0000004 RTD 118:95 nA=lx 0.000048 0.0002
UL 0:584720 V 0:0000036 V NTD 203:42 nA=lx · V 0.000732 0.04
U L ðJÞ 5:8472 A 0:0006 A NTD −20:34 nA=lx · A −0:012205 11.7
R 0:099994 Ω 0:0000020 Ω NTD −1189:52 nA=lx · Ω −0:002379 0.4
RD −0:0000010 0.0000004 RTD −118:94 nA=lx −0:000048 0.0002
αRS ΔT 0.0 0.0000023 RTD −118:94 nA=lx −0:000275 0.01
mi 6.24 0.14 RTD −0:000445 nA=lx 0.000064 0.0003
sv 1:8302 nA=lx 0:0027 nA=lx NTD 10.42 0.028616 64.4
svD −0:0015 0.0006 RTD 10:42 nA=lx 0.006254 3.1
PT 596:74 nA 0:131 nA NTD 0:03 nA=lx · nA 0.004197 1.4
PR 56:16 nA 0:0067 nA NTD −0:33 nA=lx · nA −0:002237 0.4
cPCM 1.00033 0.00010 NTD −0:33 nA=lx −0:000033 0.0001
PmD −0:000030 0.000012 RTD −0:33 nA=lx −0:000004 0.000001
ccf 1.0000 0.0005 RTD 19:06 nA=lx 0.009531 7.1
scf 0.9992 0.0004 NTD 19:08 nA=lx 0.007631 4.6
ucf 0.9992 0.00015 NTD 19:08 nA=lx 0.002862 0.6
MR 0.9997 0.00007 NTD 19:07 nA=lx 0.001430 0.2
PθR 0:0° 0:00012° RTD −19:06 nA=lx · ° −0:002201 0.4
PθT 0:0° 0:00012° RTD −19:06 nA=lx · ° −0:002201 0.4
RPa 0.0 0.00044 NTD −19:06 nA=lx −0:008387 5.5
PNR 0.0 0.00012 RTD −19:06 nA=lx −0:002287 0.4
svT ðTÞ 19:08 nA=lx combined uncertainty, k ¼ 1 0:04 nA=lx
effective degree of freedom, veff >100
expanded uncertainty, k ¼ 2 0:07 nA=lx
expanded uncertainty, k ¼ 2 0.4%

Table 5. Uncertainty Budget Established at UME for the Calibration Factor of a LMT B360 Illuminance Meter

hðy; xi Þ
Xi xi uðxi Þ pdf ci ui ðyÞ (%)
cDVM 0.999992 0.000004 NTD 6.32 0.0000253 0.01
cDVMD 0.000001 0.0000004 RTD 6.32 0.0000025 0.0001
UL 0:584720 V 0:0000036 V NTD 10:81 V−1 0.0000389 0.02
U L ðJÞ 5:8472 A 0:0006 A NTD −1:08 A−1 −0:0006489 5.0
R 0:099994 Ω 0:0000020 Ω NTD −63:24 Ω−1 −0:0001265 0.2
RD −0:0000010 0.0000004 RTD −6:32 −0:0000025 0.0001
αRS ΔT 0.0 0.0000023 RTD −6:32 −0:0000146 0.003
mi 6.24 0.14 RTD 0.000024 0.0000034 0.0001
PR 56:16 nA 0:0067 nA NTD 0:018 nA−1 0.0001189 0.2
cPCM 1.00033 0.00010 NTD 0.018 0.0000018 0.00004
PmD −0:000030 0.000012 RTD 0.018 0.0000002 0.000001
sv 1:8302 nA=lx 0:0027 nA=lx NTD −0:55 lx=nA −0:0015213 27.3
svD −0:0015 0.0006 RTD −0:55 −0:0003325 1.3
ET 30:79 lx 0:067 lx NTD −0:03 lx−1 −0:0022082 57.5
ccf 1.0000 0.0005 RTD 1.01 0.0005067 3.0
scf 0.9992 0.0004 NTD 1.01 0.0001521 0.3
ucf 0.9992 0.00015 NTD 1.01 0.0004057 1.9
MR 0.9994 0.00018 NTD 1.01 0.0001801 0.4
PθR 0:0° 0:00012° RTD −1:01°−1 −0:0001170 0.2
LX θ 0:0° 0:00012° RTD −1:01°−1 −0:0001170 0.2
RPa 0.0 0.00048 NTD −1:01 −0:0004864 2.8
PNR 0.0 0.00012 RTD −1:01 −0:0001216 0.2
kðTÞ 0.9838 combined uncertainty, k ¼ 1 0:0029
effective degree of freedom. veff >55
expanded uncertainty, k ¼ 2 0:0058
expanded uncertainty, k ¼ 2 0.6%

5840 APPLIED OPTICS / Vol. 47, No. 31 / 1 November 2008


that the electrical parameters of the working stan- standard deviations of 20 repeated measurements
dard lamp and their uncertainties were taken from were taken as the standard uncertainty of the read-
Table 1 because the same type of light source was ings [uðET Þ ¼ 0:067 lx].
used in both calibrations. In the scope of repeatability measurements, the il-
luminance measurements for both the PUT and
A. Photometer (Illuminance Meter)-Induced Uncertainty IMUT calibrations were repeated six times. By tak-
Sources ing normalized values for each calibration, the re-
A PRC Krochmann GmbH manufactured cosine- peatability was calculated using an arithmetic
corrected photometer head (TH15 C) having the mean of the results in Table 7.
same configuration as described in Subsection 2.C As is shown in Table 7, the illuminance responsiv-
and a flat diffuser in its entrance hole was used as ity of the PUT varied within 0:05%. Obtained
the WSPH in both illuminance calibrations. The illu- results were normalized to the maximum illumi-
minance responsivity (svR ) of the WSPH is traceable nance responsivity value (1:2885 nA=lx) and the
to the illuminance scale of UME, which is equal to repeatability of the illuminance responsivity calibra-
svR ¼ 1:8302 nA=lx  0:0027 nA=lx (k ¼ 1). The tion was calculated as the mean value of the normal-
WSPH is recalibrated annually by comparing with ized results (M R ¼ 0:9997). The experimental
the absolutely calibrated reference filter radiometers standard deviation of the normalized results (n ¼
[21]. The drift of the illuminance responsivity value 6) was calculated as sðM R Þ ¼ 0:00018. Thus, the
of the WSPH since its last calibration was estimated standard uncertainty of the repeatability measure-
from its calibration history to be −0:0015 nA=lx ments
p was pestimated as (uðM R Þ ¼ sðM R Þ=
(∼0:08%) with a standard deviation of sðsvRD Þ ¼ n ¼ 0:00018= 6 ¼ 0:00007). The calibration factor
0:0011. Thus, the standard uncertainty of the drift of the calibrated illuminance meter varied more,
value assuming p a RTD and estimated as compared with the illuminance responsivity calibra-
uðsvRD Þ ¼ sðsvRD Þ= 3 ¼ 0:0006 nA=lx. tion of the photometer head (0:12%). The repeat-
The PUT (LMT P30SCT) is a thermostatic stabi- ability of the calibration was calculated as the
lized type photometer head and contains a VðλÞ Si mean value of the normalized results (M R ¼
photoelement with a light sensitive surface of 0:9994) with an estimated
p standard uncertainty
30 mm diameter and a flat diffuser having opal glass of uðM R Þ ¼ 0:00044= 6 ¼ 0:00018.
material. The IMUT has a photometer head with a
B. Correction-Induced Uncertainty Sources
cosine correction and a 3½-digit display (6 measuring
ranges in decades). The device is capable of reading Light sources used in illuminance calibrations
the illuminance level from 10−2 lx to 105 lx, as stated should have a uniform illumination at the reference
by the manufacturer’s specifications. plane of measurement because the reference instru-
Photocurrents generated at the outputs of the ment and a device under calibration may have differ-
WSPH and PUT (Table 6) were measured using a ent sensing areas. Therefore, the illumination
PRC-321 and an LMT I-1000 SD photocurrent meter, uniformity of the light source has been measured
respectively. Measured real photocurrents of PR ¼ and used as a correction factor in the calculation.
56:16 nA and PT ¼ 596:74 nA at distance of 3 m from The measurement of the illuminance uniformity of
the lamp were taken from averages of 20 readings. the Osram Wi41/G lamp at a 3 m distance was car-
Therefore, standard uncertainties in both photocur- ried out using a surface scanning technique. The
rent measurements were taken as standard uncer- WSPH was mounted on a dual axis (x–y) translation
tainties of measurements (uðPR Þ ¼ 0:0067 nA and stage. By assuming that the maximum sensing area
uðPT Þ ¼ 0:131 nA). of the IMUT is within the area of 40 mm × 40 mm,
In the IMUT calibration, the device directly read both stages were moved by servo motors over a
the lux level (ET ) and the obtained experimental
Table 7. Repeatability Results of the Illuminance Responsivity and
Calibration Factor Calibrations
Table 6. Photocurrents and Illuminance Measurement Results
Photometer Head Illuminance
Photometer Photometer Head Illuminance
(PUT) Meter
Measurement Head (TH15 C) (P30SCT) Meter (B360)
№: n [PR ðnAÞ] [PT ðnAÞ] [ET ðlxÞ] Measurement normalized normalized
№: n svT ðTÞ svTn ðTÞ kðTÞ kn ðTÞ
1 56.15 596.7 30.8
2 56.17 596.9 30.7 1 1.2880 0.9996 0.9923 0.9992
3 56.16 596.8 30.8 2 1.2882 0.9998 0.9919 0.9988
4 56.17 596.4 30.8 3 1.2885 1.0000 0.9925 0.9994
… … … … 4 1.2880 0.9996 0.9922 0.9991
18 56.16 596.8 30.9 5 1.2878 0.9995 0.9928 0.9997
19 56.16 596.7 30.8 6 1.2881 0.9997 0.9931 1.0000
20 56.15 596.5 30.7 Mean 0.9997 0.9994
Mean 56.159 596.74 30.79 Standard 0.00018 0.00044
sðU L Þ 0.0067 0.131 0.067 deviation

1 November 2008 / Vol. 47, No. 31 / APPLIED OPTICS 5841


40 mm × 40 mm area with step lengths of 1 mm and probe tip was settled. The adjustment process speci-
controlled by a servo motor control unit with fied above was repeated 20 times and the standard de-
repeatabilities of 12 and 17 μm in the x and y direc- viations were estimated for the reference plane
tions, respectively [26]. The illuminance uniformity alignments by using the standard deviations obtained
plot was then obtained by dividing measured illumi- from the readout values of the micrometer. The uncer-
nances at each of the points to the maximum tainty values estimated for calibrations of the test
value (Fig. 5). photometer and the test luxmeter were uðRPa Þ ¼
After that, the correction factor for the nonunifor- 0:00044 and uðRPa Þ ¼ 0:00048.
mity of the illumination was calculated as 0.9992,
D. Evaluations of Expanded Uncertainties in the
which represents the mathematical mean of the nor-
Illuminance Calibrations Performed
malized results. This value means that the nonuni-
formity of the illumination at the reference plane is Before estimating the combined standard uncertain-
∼0:7%. Measurements were repeated eight times ties, the sensitivity coefficient for each input quan-
and the standard uncertainty of uniformity measure- tity and then their uncertainty contributions were
ments was estimated as uðucf Þ ¼ 0:00015, which is calculated (Tables 4 and 5). A detailed description
the root square of obtained experimental standard for calculating the sensitivity coefficient for each in-
deviation of repeated measurements and the nonuni- put quantity is described in Appendix B. As is shown
formity of the WSPH. in Table 4, the combined standard uncertainty for the
illuminance responsivity calibration of the PUT was
C. Other Uncertainty Sources estimated as uc ðsvT ðTÞÞ ¼ 0:04 nA=lx, whereas the
The nonlinearity of the WSPH as well as the perpen- calibration factor of the IMUT was estimated as
dicularity alignments of the WSPH, the PUT, and the uc ðkðTÞÞ ¼ 0:0029. These results cause the ex-
IMUT were performed in the same manner discussed panded uncertainties of UðsvT ðTÞÞ ¼ 0:4% and
in Subsection 2.E. Therefore, the values of these in- UðkðTÞÞ ¼ 0:6% with confidential levels of 95.45%,
put quantities and their standard uncertainties were respectively.
taken from Table 1.
The reference plane is generally taken as outside of 4. Conclusions
a turbid material (generally opal glass) used in front It is necessary to previously characterize the refer-
of the photometer. This plane was aligned mechani- ence photometer head with the lowest standard
cally by using a Mututoyo manufactured with an ab- uncertainty and to define all the uncertainty compo-
solute digimatic height gauge (570-230) located on the nents that may affect the calibration result for the
fine-adjustment carriage to feed the slider finely. The realization of precision photometric calibrations. We
height gauge has a built-in absolute linear encoder presented the detector-based luminous intensity and
that eliminates the necessity of setting the reference illumination calibration procedures performed at
point at every power-on. First of all, the reference UME, estimated calibration uncertainties, and es-
point was determined on the ruler at a distance of tablished uncertainty budgets. By these budgets,
3 m by the fiber probe joined to the gauge and then dominant and less important contributions to each
the reference plane of the WSPH fixed to the holder calibration quantity were detected. This study may
with a digital micrometer was adjusted. After comple- be helpful to a laboratory that wants to be accredited
tion of the steps stated above, provided that the gauge on luminous intensity and illuminance calibration
probe tip remained at the reference point, the WSPH capacities according to the European Norm ISO/
was substituted for the PUT or the IMUT and the re- IEC 17025:2005 standard.
ference planes of the WSPH, PUT, and IMUT were The uncertainty of luminous intensity calibration
aligned to the reference point at which the gauge is derived according to the uncertainty model based
on Eq. (4). The contribution coefficients (hðy; xi Þ),
which are calculated according to [27] and listed in
the last column of Table 1, show that the combined
standard uncertainty in the luminous intensity cali-
bration of a Osram Wi41/G incandescent lamp was
more influenced by the illuminance responsivity of
the WSPH (56.3%), the alignment of the LUT, and
its driving current (10.3%). The components that
may be affected more for other types of light sources
in luminous intensity calibration are the current sta-
bility of light source, the photocurrent readout, all
correction factors, and alignments.
Uncertainties of illuminance responsivity and ca-
libration factor calibrations are derived according
to the uncertainty model based on Eqs. (7) and (8),
Fig. 5. Illuminance uniformity map of an Osram Wi41/G light respectively. The last columns of Tables 4 and 5 show
source used in the illumination calibrations at UME. uncertainty contributions of each input quantity to

5842 APPLIED OPTICS / Vol. 47, No. 31 / 1 November 2008


ðd þ dp þ αSR ΔTÞ ¼ D; ðPc þ cPCM þ cPCMD Þ ¼ P; ð1 − Laε − Laφ − Paθ − PN Þ ¼ A;
 
ðcDVM þ cDVMD Þ · U L ðA1Þ
¼ L:
U L ðJÞ · Rð1 þ RD þ αRS ΔTÞ

Thus, the sensitivity coefficient of each input quantity will be calculated as follows:

1. Sensitivity coefficients for the calibration factor (cDVM ) and the short-term drift (cDVMD ) of the DVM:

∂I x ðTÞ ∂I ðTÞ ccf × D2 × mi × M R × P × A × scf × U L × L−1þmi


ccDVM ¼ ccDVMD ¼ ¼ x ¼ : ðA2Þ
∂cDVM ∂cDVMD U L ðJÞ × R × ð1 þ RD þ αRS ΔTÞ × ðsV þ svD Þ

2. Sensitivity coefficient for the voltage drop (U L ):

∂I x ðTÞ ccf × ðcDVM þ cDVMD Þ × D2 × mi × M R × P × A × scf × L−1þmi


cU L ¼ ¼ : ðA3Þ
∂U L U L ðJÞ × R × ð1 þ RD þ αRS ΔTÞ × ðsV þ svD Þ

3. Sensitivity coefficient for the operating current U L ðJÞ of the LUT:

∂I x ðTÞ ccf × ðcDVM þ cDVMD Þ × D2 × mi × M R × P × A × scf × U L × L−1þmi


cU L ðJÞ ¼ ¼− : ðA4Þ
∂U L ðJÞ U L ðJÞ2 × R × ð1 þ RD þ αRS ΔTÞ × ðsV þ svD Þ

4. Sensitivity coefficient for the resistance value (R) of the reference shunt resistor:

∂I x ðTÞ ccf × ðcDVM þ cDVMD Þ × D2 × mi × M R × P × A × scf × U L × L−1þmi


cR ¼ ¼− : ðA5Þ
∂R U L ðJÞ × R2 × ð1 þ RD þ αRS ΔTÞ × ðsV þ svD Þ

5. Sensitivity coefficients for the short-term drift (RD ) and the temperature coefficient (αRS ΔT) of the re-
ference shunt resistor:

∂I x ðTÞ ∂I x ðTÞ ccf × ðcDVM þ cDVMD Þ × D2 × mi × M R × P × A × scf × U L × L−1þmi


cRD ¼ cαRS ΔT ¼ ¼ ¼− : ðA6Þ
∂RD ∂αRS ΔT U L ðJÞ × R × ð1 þ RD þ αRS ΔTÞ2 × ðsV þ svD Þ

combined standard uncertainties of the illuminance 6. Sensitivity coefficients for the exponent (mi ):
responsivity of the PUT and the calibration factor of
the IMUT, respectively. The dominant factors which
∂I x ðTÞ
affect the illuminance responsivity calibration when cmi ¼
the values are considered are the uncertainties com- ∂mi
ing from the illuminance responsivity of the WSPH ccf × D2 × M R × P × A × scf × Lmi × LogðLÞ
(∼64:4%) and the lamp current (∼11:7%). Addition- ¼ :
ðsV þ svD Þ
ally, the most effective factors on the combined uncer-
tainty obtained for the calibration of the IMUT are ðA7Þ
the uncertainties coming from the readout value
via the IMUT (∼57:5%) and the illuminance respon- 7. Sensitivity coefficients for the photometric dis-
sivity of the WSPH (∼27:3%). These uncertainty bud- tance (d), offset distance (dp ), and the thermal expan-
gets introduce the best measurement capability sion coefficient (αSR ΔT) of the ruler:
which our laboratory can carry out.

Appendix A ∂I x ðTÞ ∂I x ðTÞ ∂I x ðTÞ


cd ¼ cdp ¼ cαSR ΔT ¼ ¼ ¼
Determination of the sensitivity coefficients for lumi- ∂d ∂dp ∂αSR ΔT
nous intensity calibration is described here. The 2 × ccf × D × M R × P × A × scf × Lmi
sensitivity coefficient for each input quantity, given ¼ : ðA8Þ
in Table 1, was calculated by taking the partial ðsV þ svD Þ
derivative of the model function f with respect to
the input quantity using model Eq. (4). For the con- 8. Sensitivity coefficients for the photocurrent
venience of the equation, the following abbreviations reading (Pc ), calibration factor (cPCM ), and short-term
may be accepted: drift (cPCMD ) of the PCM:

1 November 2008 / Vol. 47, No. 31 / APPLIED OPTICS 5843


∂I x ðTÞ ∂I x ðTÞ ∂I x ðTÞ 12. Sensitivity coefficient for the measurement
cPc ¼ ccPCM ¼ ccPCMD ¼ ¼ ¼ repeatability (M R ):
∂Pc ∂cPCM ∂cPCMD
ccf × D2 × M R × A × scf × Lmi ∂I x ðTÞ ccf × D2 × P × A × scf × Lmi
¼ : cM R ¼ ¼ : ðA13Þ
ðsV þ svD Þ ∂M R ðsV þ svD Þ
ðA9Þ
13. Sensitivity coefficients for vertical (Laε ) and
9. Sensitivity coefficients for the illuminance re- horizontal (Laφ ) alignments of the lamp filament, an-
sponsivity (sV ) and short-term drift (sVD ) of the gular alignment of the WSPH (Paθ ), and nonlinearity
WSPH: (PN ) of the WSPH:

∂I x ðTÞ ∂I x ðTÞ ∂I x ðTÞ ∂I x ðTÞ ∂I x ðTÞ


csv ¼ csvD ¼ ¼ cLaε ¼ cLaφ ¼ cPθ ¼ cPN ¼ ¼ ¼
∂sv ∂svD ∂Laε ∂Laφ ∂Pθ
ccf × D2 × M R × P × A × scf × Lmi ∂I x ðTÞ ccf × D2 × M R × P × scf × Lmi
¼− : ðA10Þ ¼ ¼− :
ðsV þ svD Þ2 ∂PN ðsV þ svD Þ
ðA14Þ
10. Sensitivity coefficient for the color-correction
factor (ccf ):
Appendix B
∂I ðTÞ D2 × M R × P × A × scf × Lmi Determination of sensitivity coefficients for illumi-
cccf ¼ x ¼ : ðA11Þ
∂ccf ðsV þ svD Þ nance responsivity and illuminance-meter calibra-
tions are described here. The sensitivity coefficient
11. Sensitivity coefficient for the stray-light-cor- for each input quantity, given in Tables 4 and 5,
rection factor (scf ): was calculated by taking the partial derivative of
the model function f with respect to the input quan-
∂I x ðTÞ ccf × D2 × M R × P × A × Lmi tity using model Eqs. (7) and (8). For convenience, the
cscf ¼ ¼ : ðA12Þ following abbreviations may be accepted:
∂scf ðsV þ svD Þ

 
ðcDVM þ cDVMD Þ · U L
ðPR þ cPCM þ cPCMD Þ ¼ P; ð1 − PθR − PθT − PNR Þ ¼ A; ¼ L: ðB1Þ
U L ðJÞ · Rð1 þ RD þ αRS ΔTÞ

Thus, the sensitivity coefficient of each input quantity will be calculated as follows:
1. Sensitivity Coefficients Given in Table 4
1. Sensitivity coefficients for the calibration factor (cDVM ) and short-term drift (cDVMD ) of the DVM:

∂svT ðTÞ ∂svT ðTÞ ccf × mi × M R × PT × A × scf × ðsVR þ svRD Þ × ucf × U L × L−1þmi
ccDVM ¼ ccDVMD ¼ ¼ ¼ : ðB2Þ
∂cDVM ∂cDVMD U L ðJÞ × P × R × ð1 þ RD þ αRS ΔTÞ

2. Sensitivity coefficient for the voltage drop (U L ):

∂svT ðTÞ ccf × ðcDVM þ cDVMD Þ × mi × M R × PT × A × scf × ðsVR þ svRD Þ × ucf × L−1þmi
cU L ¼ ¼ : ðB3Þ
∂U L U L ðJÞ × P × R × ð1 þ RD þ αRS ΔTÞ

3. Sensitivity coefficient for the operating current U L ðJÞ of the LUT:

∂svT ðTÞ ccf × ðcDVM þ cDVMD Þ × mi × M R × PT × A × scf × ðsVR þ svRD Þ × ucf × U L × L−1þmi
cU L ðJÞ ¼ ¼− : ðB4Þ
∂U L ðJÞ U L ðJÞ2 × P × R × ð1 þ RD þ αRS ΔTÞ

4. Sensitivity coefficient for the resistance value (R) of the reference shunt resistor:

∂svT ðTÞ ccf × ðcDVM þ cDVMD Þ × mi × M R × PT × A × scf × ðsVR þ svRD Þ × ucf × U L × L−1þmi
cR ¼ ¼− : ðB5Þ
∂R U L ðJÞ × P × R2 × ð1 þ RD þ αRS ΔTÞ

5844 APPLIED OPTICS / Vol. 47, No. 31 / 1 November 2008


5. Sensitivity coefficients for the short-term drift (RD ) and the temperature coefficient (αRS ΔT) of the re-
ference shunt resistor:

∂svT ðTÞ ∂svT ðTÞ


cRD ¼ cαRS ΔT ¼ ¼
∂RD ∂αRS ΔT
ccf × ðcDVM þ cDVMD Þ × mi × M R × PT × A × scf × ðsVR þ svRD Þ × ucf × U L × L−1þmi
¼− : ðB6Þ
U L ðJÞ × P × R × ð1 þ RD þ αRS ΔTÞ2

6. Sensitivity coefficients for the exponent (mi ):

∂svT ðTÞ ccf × M R × PT × A × scf × ðsVR þ svRD Þ × ucf × Lmi × LogðLÞ


cmi ¼ ¼ : ðB7Þ
∂mi P

7. Sensitivity coefficients for the illuminance re- 11. Sensitivity coefficient for the stray-light-cor-
sponsivity (sVR ) and short-term drift (svRD ) of the rection factor (scf ):
WSPH:

∂svT ðTÞ
∂s ðTÞ ∂svT ðTÞ cscf ¼
csv ¼ csvD ¼ vT ¼ ∂scf
∂svR ∂svRD ccf × M R × A × PT × ucf × ðsVR þ svRD Þ × Lmi
ccf × M R × PT × A × scf × ucf × Lmi ¼ :
¼ : ðB8Þ P
P ðB12Þ

8. Sensitivity coefficient for the photocurrent


reading (PT ) with the PUT: 12. Sensitivity coefficient for the uniformity-cor-
rection factor (ucf ):

∂svT ðTÞ
cPT ¼ ∂svT ðTÞ
∂PT cucf ¼
∂ucf
ccf × M R × A × scf × ucf × ðsvR þ svRD Þ × Lmi
¼ : ccf × M R × A × PT × scf × ðsVR þ svRD Þ × Lmi
P ¼ :
P
ðB9Þ
ðB13Þ

9. Sensitivity coefficients for the photocurrent


reading (PR ) with the WSPH, calibration factor 13. Sensitivity coefficient for the measurement
(cPCM ) and short-term drift (cPCMD ) of the PCM: repeatability (M R ):

∂svT ðTÞ ∂svT ðTÞ ∂svT ðTÞ ccf × M R × PT × A × scf × ucf × ðsVR þ svRD Þ × Lmi
cPR ¼ ccPCM ¼ ccPCMD ¼ ¼ ¼ ¼− :
∂PR ∂cPCM ∂cPCMD P2
ðB10Þ

10. Sensitivity coefficient for the color-correction factor (ccf ):

∂svT ðTÞ ∂svT ðTÞ


cccf ¼ cM R ¼
∂ccf ∂M R
M R × A × PT × scf × ucf × ðsVR þ svRD Þ × Lmi ccf × A × PT × scf × ucf × ðsVR þ svRD Þ × Lmi
¼ : ¼ :
P P
ðB11Þ ðB14Þ

1 November 2008 / Vol. 47, No. 31 / APPLIED OPTICS 5845


6. Sensitivity coefficients for the exponent (mi ):

∂kðTÞ ccf × M R × A × P × scf × ucf × Lmi × LogðLÞ


cmi ¼ ¼ : ðB21Þ
∂mi ET × ðsVR þ svRD Þ

7. Sensitivity coefficients for the photocurrent reading (PR ) with the WSPH, calibration factor (cPCM ), and
short-term drift (cPCMD ) of the PCM:

∂kðTÞ ∂kðTÞ ∂kðTÞ ccf × M R × A × scf × ucf × Lmi


cPR ¼ ccPCM ¼ ccPCMD ¼ ¼ ¼ ¼ : ðB22Þ
∂PR ∂cPCM ∂cPCMD ET × ðsVR þ svRD Þ

14. Sensitivity coefficients for angular alignment 8. Sensitivity coefficients for the illuminance re-
of the WSPH (PθR ) and PUT (PθT ), determination of sponsivity (sVR ) and short-term drift (svRD ) of the
the reference plane (RPa ), and nonlinearity (PN ) of WSPH:
the WSPH:

∂svT ðTÞ ∂svT ðTÞ ∂kðTÞ ∂kðTÞ


cPθR ¼ cPθT ¼ cRPa ¼ cPN ¼ ¼ csv ¼ csvD ¼ ¼
∂PθR ∂PθT ∂svR ∂svRD
∂s ðTÞ ∂svT ðTÞ ccf × M R × A × P × scf × ucf × Lmi
¼ vT ¼ ¼− : ðB23Þ
∂RPa ∂PN ET × ðsVR þ svRD Þ2
ccf × M R × PT × scf × ucf × ðsVR þ svRD Þ × Lmi
¼− :
P 9. Sensitivity coefficient for the illuminance
ðB15Þ reading (ET ) with the IMUT:

2. Sensitivity Coefficients Given in Table 5


1. Sensitivity coefficients for the calibration factor (cDVM ) and short-term drift (cDVMD ) of the DVM:

∂kðTÞ ∂kðTÞ ccf × mi × M R × A × P × scf × ucf × U L × L−1þmi


ccDVM ¼ ccDVMD ¼ ¼ ¼ : ðB16Þ
∂cDVM ∂cDVMD ET × U L ðJÞ × P × R × ð1 þ RD þ αRS ΔTÞ × ðsVR þ svRD Þ

2. Sensitivity coefficient for the voltage drop (U L ):

∂kðTÞ ccf × ðcDVM þ cDVMD Þ × mi × M R × A × P × scf × ucf × L−1þmi


cU L ¼ ¼ : ðB17Þ
∂U L ET × U L ðJÞ × R × ð1 þ RD þ αRS ΔTÞ × ðsVR þ svRD Þ

3. Sensitivity coefficient for the operating current U L ðJÞ) of the LUT:

∂kðTÞ ccf × ðcDVM þ cDVMD Þ × mi × M R × A × P × scf × ucf × U L × L−1þmi


cU L ðJÞ ¼ ¼− : ðB18Þ
∂U L ðJÞ ET × U L ðJÞ2 × P × R × ð1 þ RD þ αRS ΔTÞ × ðsVR þ svRD Þ

4. Sensitivity coefficient for the resistance value (R) of the reference shunt resistor:

∂kðTÞ ccf × ðcDVM þ cDVMD Þ × mi × M R × A × P × scf × ucf × U L × L−1þmi


cR ¼ ¼− : ðB19Þ
∂R ET × U L ðJÞ × R2 × ð1 þ RD þ αRS ΔTÞ × ðsVR þ svRD Þ

5. Sensitivity coefficients for the short-term drift (RD ) and the temperature coefficient (αRS ΔT) of the re-
ference shunt resistor:

∂kðTÞ ∂kðTÞ ccf × ðcDVM þ cDVMD Þ × mi × M R × A × scf × ucf × U L × L−1þmi


cRD ¼ cαRS ΔT ¼ ¼ ¼− : ðB20Þ
∂RD ∂αRS ΔT ET × U L ðJÞ × R × ð1 þ RD þ αRS ΔTÞ2 × ðsVR þ svRD Þ

5846 APPLIED OPTICS / Vol. 47, No. 31 / 1 November 2008


∂kðTÞ ccf × M R × A × P × scf × ucf × Lmi 6. “Expression of the uncertainty of measurement in calibra-
cET ¼ ¼− : tion,” EA-4/02 (European Co-operation for Accreditation,
∂ET E2T × ðsvR þ svRD Þ 1999).
ðB24Þ 7. D. S. Sivia, “Data Analysis: A Bayesian Tutorial”
(Oxford, 1996).
10. Sensitivity coefficient for the color-correction 8. W. Wöger, “Zu den modernen Grundlagen der Datanauswer-
tung in der Metrologie,” PTB Mitteilungen : [Amts- und
factor (ccf ):
Mitteilungsblatt der Physikalisch-Technischen Bundesan-
stalt Braunschweig-Berlin] 111, 210–225 (2001).
∂kðTÞ M R × A × P × scf × ucf × Lmi 9. R. Willink, “Coverage intervals and statistical coverage inter-
cccf ¼ ¼ : ðB25Þ
∂ccf ET × ðsVR þ svRD Þ vals,” Metrologia 41, L5–L6 (2004).
10. F. E. Satterhwaite, “An approximate distribution of estimates
11. Sensitivity coefficient for the stray-light- of variance components,” Biometrics Bull. 2, 110–114
(1946).
correction factor (scf ):
11. F. Samedov, M. Durak, and O. Bazkir, “Filter-radiometer
based realization of candela and establishment of photometric
∂kðTÞ ccf × M R × A × P × ucf × Lmi scale at UME,” Opt. Lasers Eng. 43, 1252–1256 (2005).
cscf ¼ ¼ : ðB26Þ 12. F. Sametoglu, “New traceability chains in the photometric and
∂scf ET × ðsVR þ svRD Þ
radiometric measurements at the National Metrology Insti-
tute of Turkey,” Opt. Lasers Eng., 45, 36–42 (2007).
12. Sensitivity coefficient for the uniformity-
13. R. L. Booker and D. A. McSparron, “NBS Measurement Ser-
correction factor (ucf ): vices: Photometric Calibrations,” Special Publication 250-15
(National Bureau of Standards, 1987).
∂kðTÞ ccf × M R × A × P × scf × Lmi 14. “The basis of physical photometry,” CIE Publication 18.2
cucf ¼ ¼ : ðB27Þ
∂ucf ET × ðsVR þ svRD Þ (Comission Internationalle de I’Eclairage, 1983).
15. D. R. White and P. Saunders, “The propagation of uncertainty
with calibration equations,” Meas. Sci. Technol. 18,
13. Sensitivity coefficient for the measurement
2157–2169 (2007).
repeatability (M R ): 16. G. Sauter, D. Lindner, and M. Lindemann, “CCPR key compar-
isons K3a of luminous intensity and K4 of luminous flux with
∂kðTÞ ccf × A × P × scf × ucf × Lmi lamp as transfer standards,” PTB Report (Physikalisch Tech-
cM R ¼ ¼ : ðB28Þ
∂M R ET × ðsVR þ svRD Þ nische Bundesanstalt, 1999).
17. M. Durak and F. Samedov, “Realization of a filter radiometer
based irradiance scale with high accuracy in the region from
14. Sensitivity coefficients for angular alignment
286 nm to 901 nm,” Metrologia 41, 401–406 (2004).
of the WSPH (PθR ) and IMUT (LX θ ), determination of 18. J. H. Walker and A. Thompson, “Improved automated current
the reference plane (RPa ), and nonlinearity (PNR ) of control for standard lamps,” J. Res. Natl. Inst. Stand. Technol.
the WSPH: 99, 255–261 (1994).
19. M. S. Rea, Lighting Handbook: Reference & Application,
∂kðTÞ ∂kðTÞ ∂kðTÞ 8th ed. (Illuminating Engineering Society of North America,
cPθR ¼ cPθT ¼ cRPa ¼ cPN ¼ ¼ ¼ 1993).
∂PθR ∂LX θ ∂RPa
20. J. W. T. Walsh, Photometry (Constable & Co., 1958).
∂kðTÞ ccf × M R × P × scf × ucf × Lmi 21. F. Sametoglu, “Establishment of illuminance scale at UME
¼ ¼− : with an accurately calibrated radiometer,” Opt. Rev. 13,
∂PNR ET × ðsVR þ svRD Þ
326–338 (2006).
ðB29Þ 22. F. Samedov and O. Bazkir, “Realization of photometric base
unit of candela traceable to cryogenic radiometer at UME,”
Eur. Phys. J. Appl. Phys. 30, 205–214 (2005).
23. F. Sametoglu, “Relation between the illuminance responsivity
of a photometer and the spectral power distribution of a
References source,” Opt. Eng. 46, 093607 (2007).
1. “General requirements for the competence of testing and cali- 24. C. DeCusatis, Handbook of Applied Photometry (Optical So-
bration laboratories,” ISO/IEC 17025 (International Organi- ciety of America, 1997).
zation for Standardization, 2005). 25. J. Hovila, M. Mustonen, P. Kärhä, and E. Ikonen, “Determina-
2. K. D. Sommer and B. R. L. Siebert, “Systematic approach to tion of the diffuser reference plane for accurate illuminance
the modeling of measurements for uncertainty evaluation,” responsivity calibrations,” Appl. Opt. 44, 5894–5898
Metrologia 43, S200–S210 (2006). (2005).
3. T. Doiron and J. Stoup, “Uncertainty and dimensional calibra- 26. F. Samedov, “Laser-based optical facility for determination of
tions,” J. Res. Natl. Inst. Stand. Technol. 102, 647–676 (1997). refractive index of liquids,” Opt. Laser Technol. 38, 28–36
4. Guide to the Expression of Uncertainty in Measurement, 2nd (2006).
ed. (International Organization for Standardization, 1995). 27. R. Kessel, R. Kacker, and M. Berglund, “Coefficient of contri-
5. M. G. Cox and P. M. Harris, “Measurement uncertainty and bution to the combined standard uncertainty,” Metrologia 43,
traceability,” Meas. Sci. Technol. 17, 533–540 (2006). S189–195 (2006).

1 November 2008 / Vol. 47, No. 31 / APPLIED OPTICS 5847

View publication stats

You might also like