Are you sure?
This action might not be possible to undo. Are you sure you want to continue?
EA 4/02
Publication
Reference
Expression of the
Uncertainty of
Measurement in
Calibration
PURPOSE
The purpose of this document is to harmonise evaluation of uncertainty of measurement
within EA, to set up, in addition to the general requirements of EALR1, the specific
demands in reporting uncertainty of measurement on calibration certificates issued by
accredited laboratories and to assist accreditation bodies with a coherent assignment of
best measurement capability to calibration laboratories accredited by them. As the rules
laid down in this document are in compliance with the recommendations of the Guide
to the Expression of Uncertainty in Measurement, published by seven international
organisations concerned with standardisation and metrology, the implementation of
EA4/02 will also foster the global acceptance of European results of measurement.
December 1999 rev00 page 1 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
Authorship
This document has been drafted by EAL Task Force for revision of WECC Doc. 19
1990 on behalf of the EAL Committee 2 (Calibration and Testing Activities). It
comprises a thorough revision of WECC Doc. 191990 which it replaces.
Official language
The text may be translated into other languages as required. The English language
version remains the definitive version.
Copyright
The copyright of this text is held by EA. The text may not be copied for resale.
Further information
For further information about this publication, contact your National member of EA.
You will find the list of members at the website: www.europeanaccreditation.org
December 1999 rev00 Page 2 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
CONTENTS
1 INTRODUCTION 4
2 OUTLINE AND DEFINITIONS 5
3 EVALUATION OF UNCERTAINTY OF MEASUREMENT OF INPUT
ESTIMATES 6
4 CALCULATION OF THE STANDARD UNCERTAINTY OF THE
OUTPUT ESTIMATE 9
5 EXPANDED UNCERTAINTY OF MEASUREMENT 12
6 STATEMENT OF UNCERTAINTY OF MEASUREMENT IN
CALIBRATION CERTIFICATES 13
7 STEPBYSTEP PROCEDURE FOR CALCULATING THE
UNCERTAINTY OF MEASUREMENT 13
8 REFERENCES 14
APPENDIX A 15
APPENDIX B 18
APPENDIX C 20
APPENDIX D 21
APPENDIX E 24
SUPPLEMENT 1 27
SUPPLEMENT 2 51
December 1999 rev00 Page 3 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
1 INTRODUCTION
1.1 This document sets down the principles of and the requirements on the evaluation of
the uncertainty of measurement in calibration and the statement of this uncertainty in
calibration certificates. The treatment is kept on a general level to suit all fields of
calibration. The method outlined may have to be supplemented by more specific
advice for different fields, to make the information more readily applicable. In
developing such supplementary guidelines the general principles stated in this
document should be followed to ensure harmonisation between the different fields.
1.2 The treatment in this document is in accordance with the Guide to the Expression of
Uncertainty in Measurement, first published in 1993 in the name of BIPM, IEC,
IFCC, ISO, IUPAC, IUPAP and OIML [ref. 1]. But whereas [ref. 1] establishes
general rules for evaluating and expressing uncertainty in measurement that can be
followed in most fields of physical measurements, this document concentrates on the
method most suitable for the measurements in calibration laboratories and describes
an unambiguous and harmonised way of evaluating and stating the uncertainty of
measurement. It comprises the following subjects:
• definitions basic to the document;
• methods for evaluating the uncertainty of measurement of input quantities;
• relationship between the uncertainty of measurement of the output quantity and
the uncertainty of measurement of the input quantities;
• expanded uncertainty of measurement of the output quantity;
• statement of the uncertainty of measurement;
• a step by step procedure for calculating the uncertainty of measurement.
Worked out examples showing the application of the method outlined here to
specific measurement problems in different fields will be given in supplements.
Evaluation of uncertainty of measurement is also addressed in several of the EAL
documents which provide guidance on calibration methods, some of these documents
containing specific worked out examples.
1.3 Within EAL the best measurement capability (always referring to a particular
quantity, viz. the measurand) is defined as the smallest uncertainty of measurement
that a laboratory can achieve within its scope of accreditation, when performing more
or less routine calibrations of nearly ideal measurement standards intended to define,
realize, conserve or reproduce a unit of that quantity or one or more of its values, or
when performing more or less routine calibrations of nearly ideal measuring
instruments designed for the measurement of that quantity. The assessment of best
measurement capability of accredited calibration laboratories has to be based on the
method described in this document but shall normally be supported or confirmed by
experimental evidence. To assist accreditation bodies with the assessment of the best
measurement capability some further explanations are given in Annex A.
December 1999 rev00 Page 4 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
2 OUTLINE AND DEFINITIONS
Note: Terms of special relevance to the context of the main text are written in bold
when they appear for the first time in this document. Appendix B contains a
glossary of these terms together with references
2.1 The statement of the result of a measurement is complete only if it contains both the
value attributed to the measurand and the uncertainty of measurement associated
with that value. In this document all quantities which are not exactly known are
treated as random variables, including the influence quantities which may affect the
measured value.
2.2 The uncertainty of measurement is a parameter, associated with the result of a
measurement, that characterises the dispersion of the values that could reasonably be
attributed to the measurand [ref. 2]. In this document the shorthand term uncertainty
is used for uncertainty of measurement if there is no risk of misunderstanding. For
typical sources of uncertainty in a measurement see the list given in Annex C.
2.3 The measurands are the particular quantities subject to measurement. In calibration
one usually deals with only one measurand or output quantity Y that depends upon
a number of input quantities X
i
(i = 1, 2 ,…, N) according to the functional
relationship
Y = f(X
1
, X
2
, …, X
N
) (2.1)
The model function f represents the procedure of the measurement and the method of
evaluation. It describes how values of the output quantity Y are obtained from values
of the input quantities X
i
. In most cases it will be an analytical expression, but it may
also be a group of such expressions which include corrections and correction factors
for systematic effects, thereby leading to a more complicated relationship that is not
written down as one function explicitly. Further, f may be determined
experimentally, or exist only as a computer algorithm that must be evaluated
numerically, or it may be a combination of all of these.
2.4 The set of input quantities X
i
may be grouped into two categories according to the
way in which the value of the quantity and its associated uncertainty have been
determined:
(a) quantities whose estimate and associated uncertainty are directly determined in
the current measurement. These values may be obtained, for example, from a
single observation, repeated observations, or judgement based on experience.
They may involve the determination of corrections to instrument readings as
well as corrections for influence quantities, such as ambient temperature,
barometric pressure or humidity;
(b) quantities whose estimate and associated uncertainty are brought into the
measurement from external sources, such as quantities associated with calibrated
measurement standards, certified reference materials or reference data obtained
from handbooks.
December 1999 rev00 Page 5 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
2.5 An estimate of the measurand Y, the output estimate denoted by y, is obtained from
equation (2.1) using input estimates x
i
for the values of the input quantities X
i
y f x x x
N
= ( , ,.., )
1 2
(2.2)
It is understood that the input values are best estimates that have been corrected for
all effects significant for the model. If not, the necessary corrections have been
introduced as separate input quantities.
2.6 For a random variable the variance of its distribution or the positive square root of
the variance, called standard deviation, is used as a measure of the dispersion of
values. The standard uncertainty of measurement associated with the output
estimate or measurement result y, denoted by u(y), is the standard deviation of the
measurand Y. It is to be determined from the estimates x
i
of the input quantities X
i
and their associated standard uncertainties u(x
i
). The standard uncertainty associated
with an estimate has the same dimension as the estimate. In some cases the relative
standard uncertainty of measurement may be appropriate which is the standard
uncertainty of measurement associated with an estimate divided by the modulus of
that estimate and is therefore dimensionless. This concept cannot be used if the
estimate equals zero.
3 EVALUATION OF UNCERTAINTY OF MEASUREMENT OF INPUT
ESTIMATES
3.1 General considerations
3.1.1 The uncertainty of measurement associated with the input estimates is evaluated
according to either a 'Type A' or a 'Type B' method of evaluation. The Type A
evaluation of standard uncertainty is the method of evaluating the uncertainty by
the statistical analysis of a series of observations. In this case the standard
uncertainty is the experimental standard deviation of the mean that follows from an
averaging procedure or an appropriate regression analysis. The Type B evaluation
of standard uncertainty is the method of evaluating the uncertainty by means other
than the statistical analysis of a series of observations. In this case the evaluation of
the standard uncertainty is based on some other scientific knowledge.
Note: There are occasions, seldom met in calibration, when all possible values of a
quantity lie on one side of a single limit value. A well known case is the so
called cosine error. For the treatment of such special cases, see ref. 1.
3.2 Type A evaluation of standard uncertainty
3.2.1 The Type A evaluation of standard uncertainty can be applied when several
independent observations have been made for one of the input quantities under the
same conditions of measurement. If there is sufficient resolution in the measurement
process there will be an observable scatter or spread in the values obtained.
December 1999 rev00 Page 6 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
3.2.2 Assume that the repeatedly measured input quantity X
i
is the quantity Q. With n
statistically independent observations (n > 1), the estimate of the quantity Q is q , the
arithmetic mean or the average of the individual observed values q
j
(j = 1, 2, …, n)
q
n
q
j
j
n
=
=
∑
1
1
(3.1)
The uncertainty of measurement associated with the estimate q is evaluated
according to one of the following methods:
(a) An estimate of the variance of the underlying probability distribution is the
experimental variance s²(q) of values q
j
that is given by
s q
n
q q
j
j
n
2
1
2
1
1
( ) ( ) =
−
−
=
∑ (3.2)
Its (positive) square root is termed experimental standard deviation. The best
estimate of the variance of the arithmetic mean q is the experimental variance
of the mean given by
s q
s q
n
2
2
( )
( )
= (3.3)
Its (positive) square root is termed experimental standard deviation of the
mean. The standard uncertainty u q associated with the input estimate ( ) q is the
experimental standard deviation of the mean
u q s q ( ) ( ) = (3.4)
Warning: Generally, when the number n of repeated measurements is low (n <
10), the reliability of a Type A evaluation of standard uncertainty, as expressed
by equation (3.4), has to be considered. If the number of observations cannot be
increased, other means of evaluating the standard uncertainty given in the text
have to be considered.
(b) For a measurement that is wellcharacterised and under statistical control a
combined or pooled estimate of variance s may be available that characterises
the dispersion better than the estimated standard deviation obtained from a
limited number of observations. If in such a case the value of the input quantity
Q is determined as the arithmetic mean
p
2
q of a small number n of independent
observations, the variance of the mean may be estimated by
s q
s
n
2
2
( )
p
= (3.5)
The standard uncertainty is deduced from this value by equation (3.4).
December 1999 rev00 Page 7 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
3.3 Type B evaluation of standard uncertainty
3.3.1 The Type B evaluation of standard uncertainty is the evaluation of the uncertainty
associated with an estimate x
i
of an input quantity X
i
by means other than the
statistical analysis of a series of observations. The standard uncertainty u(x
i
) is
evaluated by scientific judgement based on all available information on the possible
variability of X
i
. Values belonging to this category may be derived from
• previous measurement data;
• experience with or general knowledge of the behaviour and properties of
relevant materials and instruments;
• manufacturer’s specifications;
• data provided in calibration and other certificates;
• uncertainties assigned to reference data taken from handbooks.
3.3.2 The proper use of the available information for a Type B evaluation of standard
uncertainty of measurement calls for insight based on experience and general
knowledge. It is a skill that can be learned with practice. A wellbased Type B
evaluation of standard uncertainty can be as reliable as a Type A evaluation of
standard uncertainty, especially in a measurement situation where a Type A
evaluation is based only on a comparatively small number of statistically
independent observations. The following cases must be discerned:
(a) When only a single value is known for the quantity X
i
, e.g. a single measured
value, a resultant value of a previous measurement, a reference value from the
literature, or a correction value, this value will be used for x
i
. The standard
uncertainty u(x
i
) associated with x
i
is to be adopted where it is given. Otherwise
it has to be calculated from unequivocal uncertainty data. If data of this kind are
not available, the uncertainty has to be evaluated on the basis of experience.
(b) When a probability distribution can be assumed for the quantity X
i
, based on
theory or experience, then the appropriate expectation or expected value and the
square root of the variance of this distribution have to be taken as the estimate x
i
and the associated standard uncertainty u(x
i
), respectively.
(c) If only upper and lower limits a
+
and a
–
can be estimated for the value of the
quantity X
i
(e.g. manufacturer’s specifications of a measuring instrument, a
temperature range, a rounding or truncation error resulting from automated data
reduction), a probability distribution with constant probability density between
these limits (rectangular probability distribution) has to be assumed for the possible
variability of the input quantity X
i
. According to case (b) above this leads to
x a a
i
= +
+ −
1
2
( ) (3.6)
for the estimated value and
u x a a
i
2
1
12
( ) ( ) = −
+ −
2
(3.7)
December 1999 rev00 Page 8 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
for the square of the standard uncertainty. If the difference between the limiting
values is denoted by 2a, equation (3.7) yields
u x a
i
2
1
3
( ) =
2
2
(3.8)
The rectangular distribution is a reasonable description in probability terms of
one’s inadequate knowledge about the input quantity X
i
in the absence of any
other information than its limits of variability. But if it is known that values of
the quantity in question near the centre of the variability interval are more likely
than values close to the limits, a triangular or normal distribution may be a better
model. On the other hand, if values close to the limits are more likely than
values near the centre, a Ushaped distribution may be more appropriate.
4 CALCULATION OF THE STANDARD UNCERTAINTY OF THE OUTPUT
ESTIMATE
4.1 For uncorrelated input quantities the square of the standard uncertainty associated
with the output estimate y is given by
(4.1) u y u y
i
i
N
2
1
( ) ( ) =
=
∑
Note: There are cases, seldom occurring in calibration, where the model function is
strongly nonlinear or some of the sensitivity coefficients [see equation (4.2) and
(4.3)] vanish and higher order terms have to be included into equation (4.1). For a
treatment of such special cases see ref. 1.
The quantity u
i
(y) (i = 1, 2, …, N) is the contribution to the standard uncertainty
associated with the output estimate y resulting from the standard uncertainty
associated with the input estimate x
i
u
i
(y) = c
i
u(x
i
) (4.2)
where c
i
is the sensitivity coefficient associated with the input estimate x
i
, i.e. the
partial derivative of the model function f with respect to X
i
, evaluated at the input
estimates x
i
,
c
f
x
f
X
i
i i
X x X x
N N
= =
= =
∂
∂
∂
∂
1 1
..
(4.3)
4.2 The sensitivity coefficient c
i
describes the extent to which the output estimate y is
influenced by variations of the input estimate x
i
. It can be evaluated from the model
function f by equation (4.3) or by using numerical methods, i.e. by calculating the
change in the output estimate y due to a change in the input estimate x
i
of +u(x
i
) and 
u(x
i
) and taking as the value of c
i
the resulting difference in y divided by 2u(x
i
).
Sometimes it may be more appropriate to find the change in the output estimate y
from an experiment by repeating the measurement at e.g. x
i
± u(x
i
).
December 1999 rev00 Page 9 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
4.3 Whereas u(x
i
) is always positive, the contribution u
i
(y) according to equation (4.2) is
either positive or negative, depending on the sign of the sensitivity coefficient c
i
. The
sign of u
i
(y) has to be taken into account in the case of correlated input quantities, see
equation (D4) of Annex D.
4.4 If the model function f is a sum or difference of the input quantities X
i
f X X X p X
N
i
N
( , , , )
1 2
1
K =
=
∑
i i
x
i
i
i
2
(4.4)
the output estimate according to equation (2.2) is given by the corresponding sum or
difference of the input estimates
y p
i i
i
N
=
=
∑
1
(4.5)
whereas the sensitivity coefficients equal p
i
and equation (4.1) converts to
u y p u x
i
i
N
2 2 2
1
( ) ( ) =
=
∑
(4.6)
4.5 If the model function f is a product or quotient of the input quantities X
i
f X X X c X
N
p
i
N
i
( , , , )
1 2
1
K =
=
∏
(4.7)
the output estimate again is the corresponding product or quotient of the input
estimates
y c x
i
p
i
N
i
=
=
∏
1
(4.8)
The sensitivity coefficients equal p
i
y/x
i
in this case and an expression analogous to
equation (4.6) is obtained from equation (4.1), if relative standard uncertainties
w(y) = u(y)/y and w(x
i
) = u(x
i
)/x
i
are used,
w y p w x
i
i
N
2 2
1
( ) ( ) =
=
∑
(4.9)
4.6 If two input quantities X
i
and X
k
are correlated to some degree, i.e. if they are
mutually dependent in one way or another, their covariance also has to be
considered as a contribution to the uncertainty. See Annex D for how this has to be
done. The ability to take into account the effect of correlations depends on the
knowledge of the measurement process and on the judgement of mutual dependency
of the input quantities. In general, it should be kept in mind that neglecting
correlations between input quantities can lead to an incorrect evaluation of the
standard uncertainty of the measurand.
December 1999 rev00 Page 10 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
4.7 The covariance associated with the estimates of two input quantities X
i
and X
k
may
be taken to be zero or treated as insignificant if
(a) the input quantities X
i
and X
k
are independent, for example, because they have
been repeatedly but not simultaneously observed in different independent
experiments or because they represent resultant quantities of different
evaluations that have been made independently, or if
(b) either of the input quantities X
i
and X
k
can be treated as constant, or if
(c) investigation gives no information indicating the presence of correlation between
the input quantities X
i
and X
k
.
Sometimes correlations can be eliminated by a proper choice of the model function.
4.8 The uncertainty analysis for a measurement — sometimes called the uncertainty
budget of the measurement — should include a list of all sources of uncertainty
together with the associated standard uncertainties of measurement and the methods
of evaluating them. For repeated measurements the number n of observations also
has to be stated. For the sake of clarity, it is recommended to present the data
relevant to this analysis in the form of a table. In this table all quantities should be
referenced by a physical symbol X
i
or a short identifier. For each of them at least the
estimate x
i
, the associated standard uncertainty of measurement u(x
i
), the sensitivity
coefficient c
i
and the different uncertainty contributions u
i
(y) should be specified.
The dimension of each of the quantities should also be stated with the numerical
values given in the table.
4.9 A formal example of such an arrangement is given as Table 4.1 applicable for the
case of uncorrelated input quantities. The standard uncertainty associated with the
measurement result u(y) given in the bottom right corner of the table is the root sum
square of all the uncertainty contributions in the outer right column. The grey part of
the table is not filled in.
Table 4.1: Schematic of an ordered arrangement of the quantities,
estimates, standard uncertainties, sensitivity coefficients
and uncertainty contributions used in the uncertainty
analysis of a measurement.
Quantity
X
i
Estimate
x
i
Standard
uncertainty
u(x
i
)
Sensitivity
coefficient
c
i
Contribution to the
standard uncertainty
u
i
(y)
X
1
x
1
u(x
1
) c
1
u
1
(y)
X
2
x
2
u(x
2
) c
2
u
2
(y)
: : : : :
X
N
x
N
u(x
N
) c
N
u
N
(y)
Y y u(y)
December 1999 rev00 Page 11 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
5 EXPANDED UNCERTAINTY OF MEASUREMENT
5.1 Within EAL it has been decided that calibration laboratories accredited by members
of the EAL shall state an expanded uncertainty of measurement U, obtained by
multiplying the standard uncertainty u(y) of the output estimate y by a coverage
factor k,
U = ku(y) (5.1)
In cases where a normal (Gaussian) distribution can be attributed to the measurand
and the standard uncertainty associated with the output estimate has sufficient
reliability, the standard coverage factor k = 2 shall be used. The assigned expanded
uncertainty corresponds to a coverage probability of approximately 95%. These
conditions are fulfilled in the majority of cases encountered in calibration work.
5.2 The assumption of a normal distribution cannot always be easily confirmed
experimentally. However, in the cases where several (i.e. N ≥ 3) uncertainty
components, derived from wellbehaved probability distributions of independent
quantities, e.g. normal distributions or rectangular distributions, contribute to the
standard uncertainty associated with the output estimate by comparable amounts, the
conditions of the Central Limit Theorem are met and it can be assumed to a high
degree of approximation that the distribution of the output quantity is normal.
5.3 The reliability of the standard uncertainty assigned to the output estimate is
determined by its effective degrees of freedom (see Annex E). However, the
reliability criterion is always met if none of the uncertainty contributions is obtained
from a Type A evaluation based on less than ten repeated observations.
5.4 If one of these conditions (normality or sufficient reliability) is not fulfilled, the
standard coverage factor k = 2 can yield an expanded uncertainty corresponding to a
coverage probability of less than 95%. In these cases, in order to ensure that a value
of the expanded uncertainty is quoted corresponding to the same coverage
probability as in the normal case, other procedures have to be followed. The use of
approximately the same coverage probability is essential whenever two results of
measurement of the same quantity have to be compared, e.g. when evaluating the
results of an interlaboratory comparison or assessing compliance with a
specification.
5.5 Even if a normal distribution can be assumed, it may still occur that the standard
uncertainty associated with the output estimate is of insufficient reliability. If, in this
case, it is not expedient to increase the number n of repeated measurements or to use
a Type B evaluation instead of the Type A evaluation of poor reliability, the method
given in Annex E should be used.
5.6 For the remaining cases, i.e. all cases where the assumption of a normal distribution
cannot be justified, information on the actual probability distribution of the output
estimate must be used to obtain a value of the coverage factor k that corresponds to a
coverage probability of approximately 95%.
December 1999 rev00 Page 12 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
6 STATEMENT OF UNCERTAINTY OF MEASUREMENT IN CALIBRATION
CERTIFICATES
6.1 In calibration certificates the complete result of the measurement consisting of the
estimate y of the measurand and the associated expanded uncertainty U shall be
given in the form (y ± U). To this an explanatory note must be added which in the
general case should have the following content:
The reported expanded uncertainty of measurement is stated as the standard
uncertainty of measurement multiplied by the coverage factor k = 2, which for a
normal distribution corresponds to a coverage probability of approximately
95%. The standard uncertainty of measurement has been determined in
accordance with EAL Publication EALR2.
6.2 However, in cases where the procedure of Annex E has been followed, the additional
note should read as follows:
The reported expanded uncertainty of measurement is stated as the standard
uncertainty of measurement multiplied by the coverage factor k = XX, which for
a tdistribution with ν
eff
= YY effective degrees of freedom corresponds to a
coverage probability of approximately 95%. The standard uncertainty of
measurement has been determined in accordance with EAL Publication EALR2.
6.3 The numerical value of the uncertainty of measurement should be given to at most
two significant figures. The numerical value of the measurement result should in the
final statement normally be rounded to the least significant figure in the value of the
expanded uncertainty assigned to the measurement result. For the process of
rounding, the usual rules for rounding of numbers have to be used (for further details
on rounding see ISO 310:1992, Annex B). However, if the rounding brings the
numerical value of the uncertainty of measurement down by more than 5%, the
rounded up value should be used.
7 STEPBYSTEP PROCEDURE FOR CALCULATING THE UNCERTAINTY
OF MEASUREMENT
7.1 The following is a guide to the use of this document in practice (cf. worked examples
in Annex F and in separate supplementary documents):
(a) Express in mathematical terms the dependence of the measurand (output
quantity) Y on the input quantities X
i
according to equation (2.1). In the case of a
direct comparison of two standards the equation may be very simple, e.g.
Y = X
1
+X
2
.
(b) Identify and apply all significant corrections.
(c) List all sources of uncertainty in the form of an uncertainty analysis in
accordance with Section 4.
December 1999 rev00 Page 13 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
(d) Calculate the standard uncertainty u q for repeatedly measured quantities in
accordance with subsection 3.2.
( )
(e) For single values, e.g. resultant values of previous measurements, correction
values or values from the literature, adopt the standard uncertainty where it is
given or can be calculated according to paragraph 3.3.2 (a). Pay attention to the
uncertainty representation used. If no data are available from which the standard
uncertainty can be derived, state a value of u(x
i
) on the basis of scientific
experience.
(f) For input quantities for which the probability distribution is known or can be
assumed, calculate the expectation and the standard uncertainty u(x
i
) according
to paragraph 3.3.2 (b). If only upper and lower limits are given or can be
estimated, calculate the standard uncertainty u(x
i
) in accordance with
paragraph 3.3.2 (c).
(g) Calculate for each input quantity X
i
the contribution u
i
(y) to the uncertainty
associated with the output estimate resulting from the input estimate x
i
according
to equations (4.2) and (4.3) and sum their squares as described in equation (4.1)
to obtain the square of the standard uncertainty u(y) of the measurand. If input
quantities are known to be correlated, apply the procedure given in Annex D.
(h) Calculate the expanded uncertainty U by multiplying the standard uncertainty
u(y) associated with the output estimate by a coverage factor k chosen in
accordance with Section 5.
(i) Report the result of the measurement comprising the estimate y of the
measurand, the associated expanded uncertainty U and the coverage factor k in
the calibration certificate in accordance with Section 6.
8 REFERENCES
[1] Guide to the Expression of Uncertainty in Measurement, first edition, 1993,
corrected and reprinted 1995, International Organization for Standardization
(Geneva, Switzerland).
[2] International Vocabulary of Basic and General Terms in Metrology, second
edition, 1993, International Organization for Standardization (Geneva,
Switzerland).
[3] International Standard ISO 35341, Statistics  Vocabulary and symbols  Part I:
Probability and General Statistical Terms, first edition, 1993, International
Organization for Standardization (Geneva, Switzerland).
December 1999 rev00 Page 14 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
APPENDIX A
Comments on the assessment of best measurement capability
A1 Best measurement capability (see Section 1 of the main text) is one of the
parameters that is used to define the scope of an accredited calibration
laboratory, the others being physical quantity, calibration method or type of
instrument to be calibrated and measurement range. Best measurement
capability is normally stated in the accreditation schedule or in other
documentation that supports either the decision on accreditation or the
accreditation certificate which in many cases is issued as evidence of
accreditation. Occasionally it is stated both in the accreditation schedule and
in the supporting documents. Best measurement capability is one of the
essential pieces of information to be found in directories of accredited
laboratories that are regularly issued by accreditation bodies and is used by
potential customers to accredited laboratories to judge the suitability of a
laboratory to carry out particular calibration work at the laboratory or on
site.
A2 To make it possible to compare the capabilities of different calibration
laboratories, in particular laboratories accredited by different accreditation
bodies, the statement of best measurement capability needs to be
harmonised. To facilitate this, some explanations are given below to the
term best measurement capability, based on its definition as reported in the
main text.
A3 With 'more or less routine calibrations' is meant that the laboratory shall be
able to achieve the stated capability in the normal work that it performs
under its accreditation. Obviously there are instances where the laboratory
would be able to do better as a result of extensive investigations and
additional precautions but these cases are not covered by the definition of
best measurement capability, unless it is the outspoken policy of the
laboratory to perform such scientific investigations (in which case these
become the 'more or less routine' type calibrations of the laboratory).
A4 Inclusion of the qualifier 'nearly ideal' in the definition means that best
measurement capability should not be dependent on the characteristics of
the device to be calibrated. Inherent in the concept of being nearly ideal is
thus that there should be no significant contribution to the uncertainty of
measurement attributable to physical effects that can be ascribed to
imperfections of the device to be calibrated. However, it should be
understood that such a device should be available. If it is established that, in
a particular case, even the most 'ideal' available device contributes to the
uncertainty of measurement, this contribution shall be included in the
determination of the best measurement capability and a statement should be
made that the best measurement capability refers to calibration of that type
of device.
A5 The definition of best measurement capability implies that within its
accreditation a laboratory is not entitled to claim a smaller uncertainty of
measurement than the best measurement capability. This means that the
laboratory shall be required to state a larger uncertainty than that
corresponding to the best measurement capability whenever it is established
December 1999 rev00 Page 15 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
that the actual calibration process adds significantly to the uncertainty of
measurement. Typically the equipment under calibration may give a
contribution. Obviously the actual uncertainty of measurement can never be
smaller than the best measurement capability. When stating the actual
uncertainty, the laboratory shall be asked to apply the principles of the
present document.
A6 It should be pointed out that according to the definition of best measurement
capability the concept is applicable only to results for which the laboratory
claims its status as accredited laboratory. Thus, strictly speaking the term is
of an administrative character and does not necessarily need to reflect the
real technical capability of the laboratory. It should be possible for a
laboratory to apply for accreditation with a larger uncertainty of
measurement than its technical capability if the laboratory has internal
reasons for doing so. Such internal reasons usually involve cases where the
real capability has to be held in confidence to external customers, e.g. when
doing research and development work or when providing service to special
customers. The policy of the accreditation body should be to grant
accreditation on any applied level if the laboratory is capable of carrying out
calibrations on that level. (This consideration refers not only to the best
measurement capability but to all parameters that define the scope of a
calibration laboratory.)
A7 Assessment of best measurement capability is the task of the accreditation
body. The estimation of the uncertainty of measurement that defines the best
measurement capability should follow the procedure laid down in the
present document, with the exception of the case covered in the previous
subsection. The best measurement capability shall be stated to the same
level as required for calibration certificates, i.e. in the form of an expanded
uncertainty of measurement, normally with coverage factor k = 2. (Only in
those exceptional cases where the existence of a normal distribution cannot
be assumed or the assessment is based on limited data, the best
measurement capability has to be stated to a coverage probability of
approximately 95%. See Section 5 of the main text.)
A8 All components contributing significantly to the uncertainty of measurement
shall be taken into account when evaluating the best measurement
capability. The evaluation of the contributions that are known to vary with
time or with any other physical quantity can be based on limits of possible
variations assumed to occur under normal working conditions. For instance,
if the used working standard is known to drift, the contribution caused by
the drift between subsequent calibrations of the standard has to be taken into
account when estimating the uncertainty contribution of the working
standard.
A9 In some fields the uncertainty of measurement may depend on some
additional parameter, e.g. frequency of applied voltage when calibrating
standard resistors. Such additional parameters shall be stated together with
the physical quantity in question and the best measurement capability
specified for the additional parameters. Often this can be done by giving the
best measurement capability as a function of these parameters.
December 1999 rev00 Page 16 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
A10 The best measurement capability should normally be stated numerically.
Where the best measurement capability is a function of the quantity to
which it refers (or any other parameter) it should be given in analytical form
but in this case it may be illustrative to support the statement by a diagram.
It should always be unequivocally clear whether the best measurement
capability is given in absolute or relative terms. (Usually the inclusion of the
relevant unit gives the necessary explanation but in case of dimensionless
quantities a separate statement is needed.)
A11 Although the assessment should be based on the procedures of this
document, in the main text there is the requirement that the assessment
normally shall be 'supported or confirmed by experimental evidence'. The
meaning of this requirement is that the accreditation body should not rely on
an evaluation of the uncertainty of measurement only. Interlaboratory
comparisons that substantiate the evaluation have to be carried out under the
supervision of the accreditation body or on its behalf.
December 1999 rev00 Page 17 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
APPENDIX B
Glossary of some relevant terms
B1 arithmetic mean ([ref. 3] term 2.26)
The sum of values divided by the number of values
B2 best measurement capability (Section 1)
The smallest uncertainty of measurement that a laboratory can achieve
within its scope of accreditation, when performing more or less routine
calibrations of nearly ideal measurement standards intended to define,
realise, conserve or reproduce a unit of that quantity or one or more of its
values, or when performing more or less routine calibrations of nearly ideal
measuring instruments designed for the measurement of that quantity.
B3 correlation ([ref. 3] term 1.13)
The relationship between two or several random variables within a distribution
of two or more random variables
B4 correlation coefficient (from [ref. 1] Section C.3.6)
The measure of the relative mutual dependence of two random variables,
equal to the ratio of their covariance to the positive square root of the
product of their variances
B5 covariance (from [ref. 1] Section C.3.4)
The measure of the mutual dependence of two random variables, equal to
the expectation of the product of the deviations of two random variables
from their respective expectations
B6 coverage factor ([ref. 1] term 2.3.6)
A numerical factor used as a multiplier of the standard uncertainty of
measurement in order to obtain an expanded uncertainty of measurement
B7 coverage probability (from [ref. 1] term 2.3.5, NOTE 1)
The fraction, usually large, of the distribution of values that as a result of a
measurement could reasonably be attributed to the measurand
B8 experimental standard deviation ([ref. 2] term 3.8)
The positive square root of the experimental variance.
B9 expanded uncertainty ([ref. 1] term 2.3.5)
A quantity defining an interval about the result of a measurement that may
be expected to encompass a large fraction of the distribution of values that
could reasonably be attributed to the measurand.
B10 experimental variance (from [ref. 1] Section 4.2.2)
The quantity characterising the dispersion of the results of a series of n
observations of the same measurand given by equation (3.2) in the text.
B11 input estimate (from [ref. 1] Section 4.1.4)
The estimate of an input quantity used in the evaluation of the result of a
measurement.
B12 input quantity (from [ref. 1] Section 4.1.2)
A quantity on which the measurand depends, taken into account in the
process of evaluating the result of a measurement.
December 1999 rev00 Page 18 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
B13 measurand ([ref. 2] term 2.6)
The particular quantity subject to measurement.
B14 output estimate (from [ref. 1] Section 4.1.4)
The result of a measurement calculated from the input estimates by the
model function.
B15 output quantity (from [ref. 1] Section 4.1.2)
The quantity that represents the measurand in the evaluation of a
measurement.
B16 pooled estimate of variance (from [ref. 1] Section 4.2.4)
An estimate of the experimental variance obtained from long series of
observations of the same measurand in wellcharacterised measurements
under statistical control.
B17 probability distribution ([ref. 3] term 1.3)
A function giving the probability that a random variable takes any given value
or belongs to a given set of values
B18 random variable ([ref. 3] term 1.2)
A variable that may take any of the values of a specified set of values and
with which is associated a probability distribution.
B19 relative standard uncertainty of measurement (from [ref. 1]
Section 5.1.6)
The standard uncertainty of a quantity divided by the estimate of that
quantity.
B20 sensitivity coefficient associated with an input estimate (from [ref. 1]
Section 5.1.3)
The differential change in the output estimate generated by a differential
change in an input estimate divided by the change in that input estimate.
B21 standard deviation (from [ref. 3] term 1.23)
The positive square root of the variance of a random variable.
B22 standard uncertainty of measurement ([ref. 1] term 2.3.1)
The uncertainty of measurement expressed as the standard deviation
B23 Type A evaluation method ([ref. 1] term 2.3.2)
The method of evaluation of uncertainty of measurement by the statistical
analysis of series of observations
B24 Type B evaluation method ([ref. 1] term 2.3.3)
The method of evaluation of uncertainty of measurement by means other
than the statistical analysis of series of observations.
B25 uncertainty of measurement ([ref. 2] term 3.9)
A parameter, associated with the result of a measurement, that characterises
the dispersion of the values that could reasonably be attributed to the
measurand.
B26 variance (from [ref. 3] term 1.22)
The expectation of the square of the deviation of a random variable about its
expectation.
December 1999 rev00 Page 19 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
APPENDIX C
Sources of uncertainty of measurement
C1 The uncertainty of the result of a measurement reflects the lack of complete
knowledge of the value of the measurand. Complete knowledge requires an
infinite amount of information. Phenomena that contribute to the uncertainty
and thus to the fact that the result of a measurement cannot be characterised
by a unique value, are called sources of uncertainty. In practice, there are
many possible sources of uncertainty in a measurement [ref. 1], including:
(a) incomplete definition of the measurand;
(b) imperfect realisation of the definition of the measurand;
(c) nonrepresentative sampling — the sample measured may not represent
the defined measurand;
(d) inadequately known effects of environmental conditions or imperfect
measurements of these;
(e) personal bias in reading analogue instruments;
(f) finite instrument resolution or discrimination threshold;
(g) inexact values of measurement standards and reference materials;
(h) inexact values of constants and other parameters obtained from external
sources and used in the datareduction algorithm;
(i) approximations and assumptions incorporated in the measurement
method and procedure;
(j) variations in repeated observations of the measurand under apparently
identical conditions.
C2 These sources are not necessarily independent. Some of the sources (a) to (i)
may contribute to (j).
December 1999 rev00 Page 20 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
APPENDIX D
Correlated input quantities
D1 If two input quantities X
i
and X
k
are known to be correlated to some extent
— i.e. if they are dependent on each other in one way or another — the
covariance associated with the two estimates x
i
and x
k
u x x u x u x r x x i k
i k i k i k
( , ) ( ) ( ) ( , ) ( ) = ≠ (D.1)
has to be considered as an additional contribution to the uncertainty. The
degree of correlation is characterised by the correlation coefficient r(x
i
, x
k
)
(where i ≠ k and r ≤1).
D2 In the case of n independent pairs of simultaneously repeated observations
of two quantities P and Q the covariance associated with the arithmetic
means p and q is given by
s p q
n n
p p q q
j j
j
n
( , )
( )
( )( =
−
− −
=
∑
1
1
1
)
k
(D.2)
and by substitution r can be calculated from equation (D.1).
D3 For influence quantities any degree of correlation has to be based on
experience. When there is correlation, equation (4.1) has to be replaced by
u y c u x c c u x x
i i i k i
k i
N
i
N
i
N
2 2 2
1 1
1
1
2 ( ) ( ) ( , ) = +
= + =
−
=
∑ ∑ ∑ (D.3)
where c
i
and c
k
are the sensitivity coefficients defined by equation (4.3) or
u y u y u y u y r x x
i
i
N
i k i
k i
N
i
N
2 2
1 1 1
1
2 ( ) ( ) ( ) ( ) ( , ) = +
= = + =
k
−
∑ ∑ ∑ (D.4)
with the contributions u
i
(y) to the standard uncertainty of the output estimate
y resulting from the standard uncertainty of the input estimate x
i
given by
equation (4.2). It should be noted that the second summation of terms in
equation (D.3) or (D.4) may become negative in sign.
D4 In practice, input quantities are often correlated because the same physical
reference standard, measuring instrument, reference datum, or even
measurement method having a significant uncertainty is used in the
evaluation of their values. Without loss of generality, suppose that two input
quantities X
1
and X
2
estimated by x
1
and x
2
depend on the set of independent
variables Q
l
(l = 1,2,…,L)
X g Q Q Q
X g Q Q Q
L
L
1 1 1 2
2 2 1 2
=
=
( , , .., )
( , , .., )
(D.5)
although some of these variables may not necessarily appear in both
functions. The estimates x
1
and x
2
of the input quantities will be correlated
to some extent, even if the estimates q
l
(l = 1,2,…,L) are uncorrelated. In
that case the covariance u(x
1
,x
2
) associated with the estimates x
1
and x
2
is
given by
December 1999 rev00 Page 21 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
(D.6) u x x c c u q
l l l
l
L
( , ) ( )
1 2 1 2
2
1
=
=
∑
where c and c are the sensitivity coefficients derived from the functions
g
l 1 l 2
1
and g
2
in analogy to equation (4.3). Because only those terms contribute
to the sum for which the sensitivity coefficients do not vanish, the
covariance is zero if no variable is common to functions g
1
and g
2
. The
correlation coefficient r(x
1
,x
2
) associated with the estimates x
1
and x
2
is
determined from equation (D.6) together with equation (D.1).
D5 The following example demonstrates correlations which exist between
values attributed to two artefact standards that are calibrated against the
same reference standard.
Measurement Problem
The two standards X
1
and X
2
are compared with the reference standard Q
S
by means of a measuring system capable of determining a difference z in
their values with an associated standard uncertainty u(z). The value q
S
of the
reference standard is known with standard uncertainty u(q
S
).
Mathematical Model
The estimates x
1
and x
2
depend on the value q
S
of the reference standard and
the observed differences z
1
and z
2
according to the relations
x q z
x q z
1
2 2
= −
= −
S
S
1
)
(D.7)
Standard uncertainties and covariances
The estimates z
1
, z
2
and q
S
are supposed to be uncorrelated because they
have been determined in different measurements. The standard uncertainties
are calculated from equation (4.4) and the covariance associated with the
estimates x
1
and x
2
is calculated from equation (D.6), assuming that u(z
1
) =
u(z
2
) = u(z),
u x u q u z
u x u q u z
u x x u q
2
1
2 2
2
2
2 2
1 2
2
( ) ( ) ( )
( ) ( ) (
( , ) ( )
= +
= +
=
S
S
S
(D.8)
The correlation coefficient deduced from these results is
r x x
u q
u q u z
( , )
( )
( ) ( )
1 2
2
2 2
=
+
S
S
(D.9)
Its value ranges from 0 to +1 depending on the ratio of the standard
uncertainties u(q
S
) and u(z).
D6 The case described by equation (D.5) is an occasion where the inclusion of
correlation in the evaluation of the standard uncertainty of the measurand
can be avoided by a proper choice of the model function. Introducing
directly the independent variables Q by replacing the original variables X
l
1
and X
2
in the model function f in accordance with the transformation
equations (D.5) gives a new model function that does not contain the
correlated variables X
1
and X
2
any longer.
December 1999 rev00 Page 22 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
D7 There are cases however, where correlation between two input quantities X
1
and X
2
cannot be avoided, e.g. using the same measuring instrument or the
same reference standard when determining the input estimates x
1
and x
2
but
where transformation equations to new independent variables are not
available. If furthermore the degree of correlation is not exactly known it
may be useful to assess the maximum influence this correlation can have by
an upper bound estimate of the standard uncertainty of the measurand which
in the case that other correlations have not to be taken into account takes the
form
( )
u y u y u y u y
r
2
1 2
2
2
( ) ( ) ( ) ( ) ≤ + + (D.10)
with u
r
(y) being the contribution to the standard uncertainty of all the
remaining input quantities assumed to be uncorrelated.
Note: Equation (D.10) is easily generalised to cases of one or several
groups with two or more correlated input quantities. In this case a
respective worst case sum has to be introduced into equation (D.10)
for each group of correlated quantities.
December 1999 rev00 Page 23 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
APPENDIX E
Coverage factors derived from effective degrees of freedom.
E1 To estimate the value of a coverage factor k corresponding to a specified
coverage probability requires that the reliability of the standard uncertainty
u(y) of the output estimate is taken into account. That means taking into
account how well u(y) estimates the standard deviation associated with the
result of the measurement. For an estimate of the standard deviation of a
normal distribution, the degrees of freedom of this estimate, which depends
on the size of the sample on which it is based, is a measure of the reliability.
Similarly, a suitable measure of the reliability of the standard uncertainty
associated with an output estimate is its effective degrees of freedom ν
y
eff
,
which is approximated by an appropriate combination of the effective
degrees of freedom of its different uncertainty contributions u
i
(y).
E2 The procedure for calculating an appropriate coverage factor k when the
conditions of the Central Limit Theorem are met comprises the following
three steps:
(a) Obtain the standard uncertainty associated with the output estimate
according to the step by step procedure given in Section 7.
(b) Estimate the effective degrees of freedom ν
eff
of the standard
uncertainty u(y) associated with the output estimate y from the Welch
Satterthwaite formula
ν
ν
eff
( )
( )
=
=
∑
u y
u y
i
i i
N
4
4
1
, (E.1)
where the u
i
(y) (i=1,2,…,N), defined in equation (4.2), are the
contributions to the standard uncertainty associated with the output
estimate y resulting from the standard uncertainty associated with the
input estimate x
i
which are assumed to be mutually statistically
independent, and ν
i
is the effective degrees of freedom of the standard
uncertainty contribution u
i
(y).
For a standard uncertainty u(q) obtained from a Type A evaluation as
discussed in subsection 3.1, the degrees of freedom are given by ν
i
= n
1. It is more problematic to associate degrees of freedom with a
standard uncertainty u(x
i
) obtained from a Type B evaluation. However,
it is common practice to carry out such evaluations in a manner that
ensures that any underestimation is avoided. If, for example, lower and
upper limits a
–
and a
+
are set, they are usually chosen in such a way that
the probability of the quantity in question lying outside these limits is in
fact extremely small. Under the assumption that this practice is
followed, the degrees of freedom of the standard uncertainty u(x
i
)
obtained from a Type B may be taken to be ν
i
→ ∞.
December 1999 rev00 Page 24 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
(c) Obtain the coverage factor k from the table of values given as Table E.1
of this annex. This table is based on a tdistribution evaluated for a
coverage probability of 95,45%. If ν
eff
is not an integer, which will
usually be the case, truncate ν
eff
to the next lower integer.
Table E.1: Coverage factors k for different effective degrees of
freedom ν
eff
.
ν
eff
1 2 3 4 5 6 7 8 10 20 50 ∞
k 13,97 4,53 3,31 2,87 2,65 2,52 2,43 2,37 2,28 2,13 2,05 2,00
December 1999 rev00 Page 25 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
SUPPLEMENT 1
Examples
December 1999 rev00 Page 26 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
CONTENTS
S1 INTRODUCTION 28
S2 CALBRATION OF A WEIGHT OF NOMINAL VALUE 10 KG 29
S3 CALIBRATION OF NOMINAL 10 kΩ STANDARD RESISTOR 31
S4 CALIBRATION OF A GAUGE BLOCK OF NOMINAL LENGTH
50 MM 34
S5 CALIBRATION OF A TYPE N THERMOCOUPLE AT 1000°C 37
S6 CALIBRATION OF A POWER SENSOR AT A FREQUENCY
OF 19 GHz 42
S7 CALIBRATION OF A COAXIAL STEP ATTENUATOR AT A
SETTING OF 30 dB (INCREMENTAL LOSS) 46
December 1999 rev00 Page 27 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
S1 INTRODUCTION
S1.1 The following examples are chosen to demonstrate the method of evaluating the
uncertainty of measurement. More typical and representative examples based on
appropriate models have to be developed by special working groups in the different
areas. Nevertheless, the examples presented here provide a general guidance on how
to proceed.
S1.2 The examples are based on drafts prepared by EAL Expert Groups. These drafts have
been simplified and harmonised to make them transparent to laboratory staff in all
fields of calibration. It is thus hoped that this set of examples will contribute to a
better understanding of the details of setting up the model of evaluation and to the
harmonisation of the process of evaluating the uncertainty of measurement,
independent of the field of calibration.
S1.3 The contributions and values given in the examples are not intended to imply
mandatory or preferred requirements. Laboratories should determine the uncertainty
contributions on the basis of the model function they use in the evaluation of the
particular calibration they perform and report the evaluated uncertainty of
measurement on the calibration certificate they issue. In all the examples given, the
conditions stated in section 5 for the use of the standard coverage factor k = 2 are
fulfilled.
S1.4 The presentation of the examples follows, in accordance with the stepbystep
procedure of section 7 of EALR2, a common scheme containing:
• a short descriptive title,
• a general description of the process of measurement,
• the model of evaluation with a list of symbols used,
• an extended listing of input data with short descriptions of how they have been
obtained,
• the list of observations and the evaluation of the statistical parameters,
• an uncertainty budget in table form,
• the expanded uncertainty of measurement,
• the reported complete result of measurement.
S1.5 This first supplement to EALR2 is intended to be followed by others containing
further workedout examples on the evaluation of uncertainty of measurement in
connection with the calibration of instruments. Examples may also be found in EAL
Guidance Documents dealing with the calibration of specific types of measurement
instruments.
December 1999 rev00 Page 28 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
S2 CALBRATION OF A WEIGHT OF NOMINAL VALUE 10 KG
S2.1 The calibration of a weight of nominal value 10 kg of OIML class Ml is carried out
by comparison to a reference standard (OIML class F2) of the same nominal value
using a mass comparator whose performance characteristics have previously been
determined.
S2.2 The unknown conventional mass m
X
is obtained from:
m
X
= m
S
+ δd
D
+ δm + δm
C
+ δB (S2.1)
where:
m
S
 conventional mass of the standard,
δm
D
 drift of value of the standard since its last calibration,
δm  observed difference in mass between the unknown mass and the standard,
δm
C
 correction for eccentricity and magnetic effects,
δB  correction for air buoyancy.
S2.3 Reference standard (m
S
): The calibration certificate for the reference standard gives
a value of 10 000,005 g with an associated expanded uncertainty of 45 mg (coverage
factor k = 2).
S2.4 Drift of the value of the standard (δm
D
): The drift of the value of the reference
standard is estimated from previous calibrations to be zero within ±15 mg.
S2.5 Comparator (δm, δm
C
): A previous evaluation of the repeatability of the mass
difference between two weights of the same nominal value gives a pooled estimate of
standard deviation of 25 mg. No correction is applied for the comparator, whereas
variations due to eccentricity and magnetic effects are estimated to have rectangular
limits of ±10 mg.
S2.6 Air buoyancy (δB): No correction is made for the effects of air buoyancy, the limits
of deviation are estimated to be ±1×10
6
of the nominal value.
S2.7 Correlation: None of the input quantities are considered to be correlated to any
significant extent.
December 1999 rev00 Page 29 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
S2.8 Measurements: Three observations of the difference in mass between the unknown
mass and the standard are obtained using the substitution method and the substitution
scheme ABBA ABBA ABBA:
no conventional mass reading observed difference
1 standard +0,010 g
unknown +0,020 g
unknown +0,025 g
standard +0,015 g +0,01 g
2 standard +0,025 g
unknown +0,050 g
unknown +0,055 g
standard +0,020 g +0,03 g
3 standard +0,025 g
unknown +0,045 g
unknown +0,040 g
standard +0,020 g +0,02 g
arithmetic mean: δ = 0,020 g m
pooled estimate of standard deviation: s
p
(δm) = 25 mg
(obtained from prior evaluation)
standard uncertainty: u m s m ( ) ( ) , δ δ = = =
25
3
14 4
mg
mg
S2.9 Uncertainty budget (m
X
):
quantity
X
i
estimate
x
i
standard
uncertainty
u(x
i
)
probability
distribution
sensitivity
coefficient
c
i
uncertainty
contribution
u
i
(y)
m
S
10 000,005 g 22,5 mg normal 1,0 22,5 mg
δm
D
0,000 g 8,95 mg rectangular 1,0 8,95 mg
δm 0,020 g 14,4 mg normal 1,0 14,4 mg
δm
C
0,000 g 5,77 mg rectangular 1,0 5,77 mg
δB 0,000 g 5,77 mg rectangular 1,0 5,77 mg
m
X
10 000,025 g 29,3 mg
S2.10 Expanded uncertainty
U = k × u(m
X
) = 2 × 29,3 mg ≅ 59 mg
December 1999 rev00 Page 30 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
S2.11 Reported result
The measured mass of the nominal 10 kg weight is 10,000 025 kg ±59 mg.
The reported expanded uncertainty of measurement is stated as the standard
uncertainty of measurement multiplied by the coverage factor k = 2, which for a
normal distribution corresponds to a coverage probability of approximately 95 %.
S3 CALIBRATION OF NOMINAL 10 KΩ STANDARD RESISTOR
S3.1 The resistance of a fourterminal standard resistor is determined by direct
substitution using a longscale digital multimeter (7½ digit DMM) on its resistance
range, and a calibrated fourterminal standard resistor of the same nominal value as
the item to be calibrated as reference standard. The resistors are immersed in a well
stirred oil bath operating at a temperature of 23 °C monitored by a centrally placed
mercuryinglass thermometer. The resistors are allowed to stabilise before the
measurement. The fourterminal connectors of each resistor are connected in turn to
the terminals of the DMM. It is determined that the measuring current on the 10 kΩ
range of the DMM of 100 µA is sufficiently low not to cause any appreciable self
heating of the resistors. The measuring procedure used also ensures that the effects of
external leakage resistances on the result of measurement can be considered to be
insignificant.
S3.2 The resistance R
X
of the unknown resistor is obtained from the relationship:
R R R R r r R
T X S D S C
( ) = + +
T X
− δ δ δ (S3.1)
where:
R
S
 resistance of the reference,
δR
D
 drift of the resistance of the reference since its last calibration,
δR
TS
 temperature related resistance variation of the reference,
r = R
iX
/R
iS
 ratio of the indicated resistance (index i means ‘indicated’) for the
unknown and reference resistors,
rC  correction factor for parasitic voltages and instrument resolution
δR
TX
 temperaturerelated resistance variation of the unknown resistor.
S3.3 Reference standard (R
S
): The calibration certificate for the reference standard gives
a resistance value of 10 000,053 Ω ±5 mΩ (coverage factor k = 2) at the specified
reference temperature of 23 °C.
S3.4 Drift of the value of the standard (δR
D
): The drift of the resistance of the reference
resistor since its last calibration is estimated from its calibration history to be
+20 mΩ with deviations within ±10 mΩ.
S3.5 Temperature corrections (δR
TS
, δR
TX
): The temperature of the oil bath is monitored
using a calibrated thermometer to be 23,00 °C. Taking into account the metrological
characteristics of the thermometer used and of gradients of temperature within the oil
bath, the temperature of the resistors is estimated to coincide with the monitored
temperature within ±0,055 K. Thus the known value 5×10
6
K
1
of the temperature
coefficient (TC) of the reference resistor gives limits ±2,75 mΩ for the deviation
December 1999 rev00 Page 31 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
from its resistance value according to calibration, due to a possible deviation from
the operating temperature. From the manufacturer’s literature, the TC of the
unknown resistor is estimated not to exceed 10×10
6
K
1
, thus the resistance variation
of the unknown resistor due to a temperature variation is estimated to be within
±5,5 mΩ.
S3.6 Resistance measurements (r
C
): Since the same DMM is used to observe both R
iX
and R
iS
the uncertainty contributions are correlated but the effect is to reduce the
uncertainty and it is only necessary to consider the relative difference in the
resistance readings due to systematic effects such as parasitic voltages and
instrument resolution (see the mathematical note in paragraph S3.12), which are
estimated to have limits of ±0,5×10
6
for each reading. The distribution resulting for
the ratio r
C
is triangular with expectation 1,000 000 0 and limits ±1,0×10
6
.
S3.7 Correlation: None of the input quantities are considered to be correlated to any
significant extent.
S3.8 Measurements(r): Five observations are made to record the ratio r:
No. observed ratio
1 1,000 010 4
2 1,000 010 7
3 1,000 010 6
4 1,000 010 3
5 1,000 010 5
arithmetic mean: r = 1 000 010 5 ,
experimental standard deviation: s(r) = 0,158 × 10
6
standard uncertainty: u r s r ( ) ( ) = =
×
= ×
−
−
0158 10
5
0 0707 10
6
6
,
,
S3.9 Uncertainty budget (R
X
):
quantity
X
i
estimate
x
i
standard
uncertainty
u(x
i
)
probability
distribution
sensitivity
coefficient
c
i
uncertainty
contribution
u
i
(y)
R
S
10 000,053 Ω 2,5 mΩ normal 1,0 2,5 mΩ
δR
D
0,020 Ω 5,8 mΩ rectangular 1,0 5,8 mΩ
δR
TS
0,000 Ω 1,6 mΩ rectangular 1,0 1,6 mΩ
δR
TX
0,000 Ω 3,2 mΩ rectangular 1,0 3,2 mΩ
r
C
1,000 000 0 0,41×10
6
triangular 10 000 Ω 4,1 mΩ
r 1,000 010 5 0,07×10
6
normal 10 000 Ω 0,7 mΩ
R
X
10 000,178 Ω 8,33 mΩ
December 1999 rev00 Page 32 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
S3.10 Expanded uncertainty:
U k u R = × = × ≅ ( )
X
2 8 33 17 , m m Ω Ω
S3.11 Reported result: The measured value of the nominal 10 kΩ resistor, at a measuring
temperature of 23,00 °C and a measuring current of 100 µA, is
(10 000,178 ±0,017) Ω.
The reported expanded uncertainty of measurement is stated as the standard
uncertainty of measurement multiplied by the coverage factor k = 2, which for a
normal distribution corresponds to a coverage probability of approximately 95 %.
S3.12 Mathematical note on the standard uncertainty of measurement of the ratio of
indicated resistance values: The unknown and the reference resistors have nearly
the same resistance. Within the usual linear approximation in the deviations, the
values causing the DMM indications R
iX
and R
iS
are given by
R R
R
R
R R
R
R
X i X
X
S i S
S
( )
( )
'
'
'
'
= +
= +
1
1
δ
δ
(S3.2)
with R being the nominal value of the resistors and δ R
X
' and δ R
S
' the unknown
deviations. The resistance ratio deduced from these expressions is
R
R
rr
X
S
C
'
'
= (S3.3)
with the ratio of the indicated resistance for the unknown and the reference resistor
r
R
R
=
i X
iS
(S3.4)
and the correction factor (linear approximation in the deviations)
r
R R
R
C
X
= +
−
1
δ δ '
S
'
)
(S3.5)
Because of the fact that the difference of the deviations enters into equation (S3.5),
correlated contributions of systematic effects resulting from the internal scale of the
DMM do not influence the result. The standard uncertainty of the correction factor is
determined only by uncorrelated deviations resulting from the parasitic effects and
the resolution of the DMM. Assuming that u R u R u R ( ) ( ) (
X S
δ δ δ ' ' ' = = , it is given
by the expression
u r
u R
R
2
2
2
2 ( )
(
C
=
δ ' )
(S3.6)
December 1999 rev00 Page 33 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
S4 CALIBRATION OF A GAUGE BLOCK OF NOMINAL LENGTH 50 MM
S4.1 The calibration of the grade 0 gauge block (ISO 3650) of 50 mm nominal length is
carried out by comparison using a comparator and a calibrated gauge block of the
same nominal length and the same material as reference standard. The difference in
central length is determined in vertical position of the two gauge blocks using two
length indicators contacting the upper and lower measuring faces. The actual length
of the gauge block to be calibrated is related to the actual length of the
reference standard by the equation
l
X
' l
S
'
l l
X S
' ' = + l δ (S4.1)
with δl being the measured length difference. l and l are the lengths of the gauge
blocks under measurement conditions, in particular at a temperature which, on
account of the uncertainty in the measurement of laboratory temperature, may not be
identical with the reference temperature for length measurements.
X
'
S
'
S4.2 The length l
X
of the unknown gauge block at the reference temperature is obtained
from the relationship:
l l l l l L t t l
X S D C
= + + + −
V
× + × − δ δ δ δ δ ∆ δ ( ) α α (S4.2)
where:
l
S
 length of the reference gauge block at the reference
temperature t
0
= 20 °C according to its calibration
certificate;
δl
D
 change of the length of the reference gauge block since its
last calibration due to drift;
δl  observed difference in length between the unknown and the
reference gauge block;
δl
C
 correction for nonlinearity and offset of the comparator;
L  nominal length of the gauge blocks considered;
α α α = + ( )
X S
/ 2  average of the thermal expansion coefficients of the
unknown and reference gauge blocks;
δt = (t
X
 t
S
)  temperature difference between the unknown and reference
gauge blocks;
δα = (α
X
– α
S
)  difference in the thermal expansion coefficients between the
unknown and the reference gauge blocks;
∆t t t t = + − ( )
0 X S
/ 2  deviation of the average temperature of the unknown and
the reference gauge blocks from the reference temperature;
δl
V
 correction for noncentral contacting of the measuring faces
of the unknown gauge block.
S4.3 Reference standard (l
S
): The length of the reference gauge block together with the
associated expanded uncertainty of measurement is given in the calibration
certificate of a set of gauge blocks as 50,000 02 mm ±30 nm (coverage factor k = 2).
December 1999 rev00 Page 34 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
S4.4 Drift of the standard (δl
D
): The temporal drift of the length of the reference gauge
block is estimated from previous calibrations to be zero with limits ±30 nm. General
experience with gauge blocks of this type suggests that zero drift is most probable
and that a triangular probability distribution can be assumed.
S4.5 Comparator (δl
C
): The comparator has been verified to meet the specifications
stated in EALG21. From this, it can be ascertained that for length differences D up
to ±10 µm corrections to the indicated length difference are within the limits
±(30 nm +0,02·D). Taking into account the tolerances of the grade 0 gauge block to
be calibrated and the grade K reference gauge block, the maximum length difference
will be within ±1 µm leading to limits of ±32 nm for nonlinearity and offset
corrections of the comparator used.
S4.6 Temperature corrections (α , δt, δα, ∆t ): Before calibration, care is taken to
ensure that the gauge blocks assume ambient temperature of the measuring room.
The remaining difference in temperature between the standard and the gauge block to
be calibrated is estimated to be within ±0,05 K. Based on the calibration certificate of
the reference gauge block and the manufacturer’s data for the gauge block to be
calibrated the linear thermal expansion coefficient of the steel gauge blocks is
assumed to be within the interval (11,5±1,0)×10
6
°C
1
. Combining the two
rectangular distributions the difference in linear thermal expansion coefficient is
triangularly distributed within the limits ±2×10
6
°C
1
. The deviation of the mean
temperature of measurement from the reference temperature t
0
= 20 °C is estimated
to be within ±0,5 °C. The best estimates of the difference in linear expansion
coefficients and the deviations of the mean temperature from the reference
temperature are zero. Therefore second order terms have to be taken into account in
the evaluation of their uncertainty contribution resulting in the product of standard
uncertainties associated with the factors of the product term δ ∆ α × t in equation
(S4.2) (see the mathematical note in paragraph S4.13, eq. (S4.5)). The final standard
uncertainty is u t ( ) δ ∆ α × = ×
−
0 236 10
6
, .
S4.7 Variation in length (δl
V
): For gauge blocks of grade 0 the variation in length
determined from measurements at the centre and the four corners has to be within
±0,12 µm (ISO 3650). Assuming that this variation occurs on the measuring faces
along the short edge of length 9 mm and that the central length is measured inside a
circle of radius 0,5 mm, the correction due to central misalignment of the contacting
point is estimated to be within ±6,7 nm.
S4.8 Correlation: None of the input quantities are considered to be correlated to any
significant extent.
December 1999 rev00 Page 35 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
S4.9 Measurements (δl): The following observations are made for the difference between
the unknown gauge block and the reference standard, the comparator being reset
using the reference standard before each reading.
obs.
no.
obs. value
1 100 nm
2 90 nm
3 80 nm
4 90 nm
5 100 nm
arithmetic mean: δl = −94 nm
pooled estimate of standard deviation: s l
p
( ) δ = 12 nm
(obtained from prior evaluation)
standard uncertainty: u l s l ( ) ( ) δ δ = = =
12
5
5 37
nm
nm ,
The pooled estimate of the standard deviation has been taken from the tests made to
confirm compliance of the comparator used with the requirements of EALG21.
S4.10 Uncertainty budget (δl
X
):
quantity
X
i
estimate
x
i
standard
uncertainty
u(x
i
)
probability
distribution
sensitivity
coefficient
c
i
uncertainty
contribution
u
i
(y)
l
S
50,000 020 mm 15 nm normal 1,0 15,0 nm
δl
D
0 mm 17,3 nm triangular 1,0 17,3 nm
δl 0,000 094 mm 5,37 nm normal 1,0 5,37 nm
δl
C
0 mm 18,5 nm rectangular 1,0 18,5 nm
δt 0 °C 0,0289 °C rectangular 575 nm°C
1
16,6 nm
δ ∆ α × t 0 0,236×10
6
special 50 mm 11,8 nm
δl
V
0 mm 3,87 nm rectangular 1,0 3,87 nm
l
X
49,999 926 mm 36,4 nm
S4.11 Expanded uncertainty
U k u l = × = × ≅ ( )
X
2 36 4 73 , nm nm
S4.12 Reported result
The measured value of the nominal 50 mm gauge block is 49,999 926 mm ±73 nm.
The reported expanded uncertainty of measurement is stated as the standard
uncertainty of measurement multiplied by the coverage factor k = 2, which for a
normal distribution corresponds to a coverage probability of approximately 95 %.
December 1999 rev00 Page 36 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
S4.13 Mathematical note on the standard uncertainty of measurement of the product
of two quantities with zero expectation: If a product of two quantities is
considered, the usual method of evaluation of uncertainty contributions based on the
linearisation of the model function has to be modified if one or both of the
expectations of the factors in the product are zero. If the factors in the product are
statistically independent with nonzero expectations, the square of the relative
standard uncertainty of measurement (relative variance) associated with the product
can be expressed without any linearisation by the squares of the relative standard
uncertainties associated with the estimates of the factors:
w x x w x w x w x w x
2
1 2
2
1
2
2
2
1
2
2
( ) ( ) ( ) ( ) ( ) × = + + × (S4.2)
Using the definition of the relative standard uncertainty of measurement this
expression is easily transformed into the general relation
u x x x u x x u x u x u x
2
1 2 2
2 2
1 1
2 2
2
2
1
2
2
( ) ( ) ( ) ( ) ( × = + + × )
)
(S4.3)
If the standard uncertainties u(x
1
) and u(x
2
) associated with the expectations x
1
and x
2
are much smaller than the moduli of the respective expectation values the third term
on the right side may be neglected. The resulting equation represents the case
described by the usual method based on the linearisation of the model function.
If, however, one of the moduli of the expectation values, for example x
2
, is much
smaller than the standard uncertainty u(x
2
) associated with this expectation or even
zero, the product term involving this expectation may be neglected on the right side
of equation (S4.3), but not the third term. The resulting equation is
u x x x u x u x u x
2
1 2 1
2 2
2
2
1
2
2
( ) ( ) ( ) ( ) × ≅ + × (S4.4)
If both moduli of the expectation values are much smaller than their associated
standard uncertainties or even zero, only the third term in equation (S4.3) gives a
significant contribution:
u x x u x u x
2
1 2
2
1
2
2
( ) ( ) ( × ≅ × (S4.5)
S5 CALIBRATION OF A TYPE N THERMOCOUPLE AT 1000°C
S5.1 A type N thermocouple is calibrated by comparison with two reference
thermocouples of type R in a horizontal furnace at a temperature of 1000 °C. The
emfs generated by the thermocouples are measured using a digital voltmeter through
a selector/reversing switch. All thermocouples have their reference junctions at 0 °C.
The thermocouple to be calibrated is connected to the reference point using
compensating cables. Temperature values are give in the t
90
temperature scale.
S5.2 The temperature t
X
of the hot junction of the thermocouple to be calibrated is
t t V V V V
t
C
t t
t V C V C V C V
C
C
t t
X S iS iS1 iS2 R
S
S
D F
S iS S iS1 S iS2 S R
S
S
S D
( )
( )
= + + + − + +
≅ + × + × + × − + +
δ δ δ
t
F
δ
δ δ
δ δ δ δ δ
0
0
0
0
δ
(S5.1)
December 1999 rev00 Page 37 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
S5.3 The voltage V
X
across the thermocouple wires with the cold junction at 0 °C during
calibration is
V t V t
t
C
t
C
V V V V V
t
C
t
C
X X X
X
X
X
iX iX1 iX2 R LX
X
X
X
( ) ( ) ≅ + −
= + + + + + −
∆ δ
δ δ δ δ
∆ δ
0
0
0
0
(S5.2)
where:
t
S
(V)  temperature of the reference thermometer in terms of voltage
with cold junction at 0 °C. The function is given in the
calibration certificate;
V
iS
, V
iX
 indication of the voltmeter;
δV
iS1
, δV
iX1
 voltage corrections obtained from the calibration of the
voltmeter;
δV
iS2
, δV
iX2
 voltage corrections due to the limited resolution of the
voltmeter;
δV
R
 voltage correction due to contact effects of the reversing switch;
δt
0S
, δt
0X
 temperature corrections due to the deviation of the reference
temperatures from 0 °C;
C
S
, C
X
 sensitivities of the thermocouples for voltage at the measuring
temperature of 1000 °C;
C
S0
, C
X0
 sensitivities of the thermocouples for voltage at the reference
temperature of 0 °C;
δt
D
 change of the values of the reference thermometers since their
last calibration due to drift;
δt
F
 temperature correction due to nonuniformity of the temperature
of the furnace;
t  temperature at which the thermocouple is to be calibrated
(calibration point);
∆t = t  t
X
 deviation of the temperature of the calibration point from the
temperature of the furnace;
δV
LX
 voltage correction due to the compensating cables.
S5.4 The reported result is the output emf of the thermocouple at the temperature of its hot
junction. Because the measurement process consists of two steps — determination of
the temperature of the furnace and determination of emf of the thermocouple to be
calibrated — the evaluation of the uncertainty of measurement is split in two parts.
S5.5 Reference standards (t
S
(V)): The reference thermocouples are supplied with
calibration certificates that relate the temperature at their hot junction with their cold
junction at 0 °C to the voltage across their wires. The expanded uncertainty of
measurement at 1000 °C is U = 0,3 °C (coverage factor k = 2).
December 1999 rev00 Page 38 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
S5.6 Calibration of the voltmeter (δV
iS1
, δV
iX1
): The voltmeter has been calibrated.
Corrections to the measured voltages are made to all results. The calibration
certificate gives a constant expanded uncertainty of measurement for voltages
smaller than 50 mV of U = 2,0 µV (coverage factor k = 2).
S5.7 Resolution of the voltmeter (δV
iS2
, δV
iX2
): A 4½ digit microvoltmeter has been used
in its 10 mV range resulting in resolution limits of ±0,5 µV at each indication.
S5.8 Parasitic voltages (δV
R
): Residual parasitic offset voltages due to the switch
contacts have been estimated to be zero within ±2 µV.
S5.9 Reference temperatures (δt
0S
, δt
0X
): The temperature of the reference point of each
thermocouple is known to be 0 °C within ±0,1 °C.
S5.10 Voltage sensitivities (C
S
, C
X
, C
S0
, C
X0
): The voltage sensitivities of the
thermocouples have been taken from reference tables:
1000 °C 0 °C
reference thermocouple C
S
= 0,077 °C/µV C
S0
= 0,189 °C/µV
unknown thermocouple C
X
= 0,026 °C/µV C
S0
= 0,039 °C/µV
S5.11 Drift of the reference standard (δt
D
): From previous calibrations the drift of the
reference standards are estimated to be zero within the limits ±0,3 °C.
S5.12 Temperature gradients (δt
F
): The temperature gradients inside the furnace have
been measured. At 1000 °C, deviations from nonuniformity of temperature in the
region of measurement are within ±1 °C.
S5.13 Compensating cables (δV
LX
): The compensating cables have been investigated in
the range 0 °C to 40 °C. From this, the voltage differences between the cables and
the thermocouple wires are estimated to be within ±5 µV.
S5.14 Measurements (V
iS
, t
S
(V
iS
), V
iX
): The indications of the voltmeter are recorded in
the following operational procedure which gives four readings for every
thermocouple and reduces the effects of temperature drift in the thermal source and
of parasitic thermal voltages in the measuring circuit:
1
st
cycle:
1
st
standard, unknown thermocouple, 2
nd
standard,
2
nd
standard, unknown thermocouple, 1
st
standard.
Reversion of polarity.
2
nd
cycle:
1
st
standard, unknown thermocouple, 2
nd
standard,
2
nd
standard, unknown thermocouple, 1
st
standard.
December 1999 rev00 Page 39 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
S5.15 The procedure requires that the difference between the two reference standards must
not exceed ±0,3 °C. If the difference is not within these limits the observations have
to be repeated and/or the reasons for such a large difference have to be investigated.
Thermocouple 1
st
reference Unknown 2
nd
reference
Indicated voltage, corrected +10500 µV +36245 µV +10503 µV
+10503 µV +36248 µV +10503 µV
10503 µV 36248 µV 10505 µV
10504 µV 36251 µV 10505 µV
Mean voltage 10502,5 µV 36248 µV 10504 µV
Temperature of the hot junction 1000,4 °C 1000,6 °C
Temperature of the furnace 1000,5 °C
S5.16 From the four readings on each thermocouple given in the table above, the mean
value of the voltages of each thermocouple is deduced. The voltage values of the
reference thermocouples are converted into temperature values by means of the
temperaturevoltage relations stated in their calibration certificates. The observed
temperature values are highly correlated (correlation factor nearly one). Therefore,
by taking their mean value, they are combined to one observation only, which is the
temperature of the furnace at the location of the thermocouple to be calibrated. In a
similar way, one observation of the voltage of the thermocouple to be calibrated has
been extracted. In order to evaluate the uncertainty of measurement associated with
these observations, a series of ten measurements has been previously undertaken at
the same temperature of operation. It gave a pooled estimate of standard deviation
for the temperature of the furnace and the voltage of the thermocouple to be
calibrated.
The respective standard uncertainties of measurement of the observed quantities are:
pooled estimate of standard deviation: s
p
(t
S
) = 0,10 °C
standard uncertainty: u(t
S
) =
s t
p S
( )
1
= 0,10 °C
pooled estimate of standard deviation: s
p
(V
iX
) = 1,6 µV
standard uncertainty: u(V
iX
) =
s V
= 1,6 µV
p iX
(
1
)
December 1999 rev00 Page 40 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
S5.17 Uncertainty budget (temperature t
X
of the furnace):
quantity
X
i
estimate
x
i
standard
uncertainty
u(x
i
)
probability
distribution
sensitivity
coefficient
c
i
uncertainty
contribution
u
i
(y)
t
S
1000,5 °C 0,10 °C normal 1,0 0,10 °C
δV
iS1
0 µV 1,00 µV normal 0,077 °C/µV 0,077 °C
δV
iS2
0 µV 0,29 µV rectangular 0,077 °C/µV 0,022 °C
δV
R
0 µV 1,15 µV rectangular 0,077 °C/µV 0,089 °C
δt
0S
0 °C 0,058 °C rectangular 0,407 0,024 °C
δt
S
0 °C 0,15 °C normal 1,0 0,15 °C
δt
D
0 °C 0,173 °C rectangular 1,0 0,173 °C
δt
F
0 °C 0,577 °C rectangular 1,0 0,577 °C
t
X
1000,5 °C 0,641 °C
S5.18 Uncertainty budget (emf V
X
of the thermocouple to be calibrated):
The standard uncertainty of measurement associated with the temperature deviation
of the calibration point from the temperature of the furnace is the standard
uncertainty of measurement associated with the temperature of the furnace because
the temperature point is a defined value (exactly known).
quantity
X
i
estimate
x
i
standard
uncertainty
u(x
i
)
probability
distribution
sensitivity
coefficient
c
i
uncertainty
contribution
u
i
(y)
V
iX
36 248 µV 1,60 µV normal 1,0 1,60 µV
δ V
iX1
0 µV 1,00 µV normal 1,0 1,00 µV
δ V
iX2
0 µV 0,29 µV rectangular 1,0 0,29 µV
δV
R
0 µV 1,15 µV rectangular 1,0 1,15 µV
δV
LX
0 µV 2,9 µV rectangular 1,0 2,9 µV
∆t 0,5 °C 0,641 °C normal 38,5 µV/°C 24,5 µV
δt
0X
0 °C 0,058 °C rectangular 25,6 µV/°C 1,48 µV
V
X
36 229 µV 25,0 µV
December 1999 rev00 Page 41 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
S5.19 Expanded uncertainties
The expanded uncertainty associated with the measurement of the temperature of the
furnace is
U = k × u(t
X
) = 2 × 0,641 °C ≅ 1,3 °C
The expanded uncertainty associated with the emf value of the thermocouple to be
calibrated is
U = k × u(V
X
) = 2 × 25,0 µV ≅ 50 µV
S5.20 Reported result
The type N thermocouple shows, at the temperature of 1000,0 °C with its cold
junction at a temperature of 0 °C, an emf of 36 230 µV ±50 µV.
The reported expanded uncertainty of measurement is stated as the standard
uncertainty of measurement multiplied by the coverage factor k = 2, which for a
normal distribution corresponds to a coverage probability of approximately 95 %.
S6 CALIBRATION OF A POWER SENSOR AT A FREQUENCY OF 19 GHZ
S6.1 The measurement involves the calibration of an unknown power sensor with respect
to a calibrated power sensor used as a reference by substitution on a stable transfer
standard of known small reflection coefficient. The measurement is made in terms of
calibration factor, which is defined as the ratio of incident power at the reference
frequency of 50 MHz to the incident power at the calibration frequency under the
condition that both incident powers give equal power sensor response. At each
frequency, one determines the (indicated) ratio of the power for the sensor to be
calibrated, respectively the reference sensor and the internal sensor that forms part of
the transfer standard, using a dual power meter with ratio facility.
S6.2 Schematic of the measuring system
Γ
G
Γ
S
Γ
X
or
Transfer standard
Power meter
1.0000 B/A
G
S
X
December 1999 rev00 Page 42 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
S6.3 The quantity K , termed ‘calibration factor’ by some manufacturers, is defined as:
K
P
P
P
P
= =
+
+
Ir
Ic
r Ar
c A
( )
( )
1
1
2
2
Γ
Γ
c
(S6.1)
for the equal power meter indication
where:
P
r
 incident power at the reference frequency (50 MHz),
P
c
 incident power at the calibration frequency,
Γ
r
 voltage reflection coefficient of the sensor at the reference frequency
Γ
c
 voltage reflection coefficient of the sensor at the calibration frequency
P
Ar
 power absorbed by the sensor at the reference frequency
P
Ac
 power absorbed by the sensor at the calibration frequency
S6.4 The calibration factor of the unknown sensor is obtained from the relationship
K K K
M M
M M
p p p
r
X S D
= + ( )
Sr Xc
Sc X
Cr Cc
δ (S6.2)
where:
K
S
 calibration factor of the reference power sensor;
δK
D
 change of the calibration factor of the reference power sensor since
its last calibration due to drift;
M
Sr
 mismatch factor of reference sensor at the reference frequency;
M
Sc
 mismatch factor of standard sensor at the calibration frequency;
M
Xr
 mismatch factor of sensor to be calibrated at the reference
frequency;
M
Xc
 mismatch factor of sensor to be calibrated at the calibration
frequency;
p
Cr
 correction of the observed ratio for nonlinearity and limited
resolution of the power meter at power ratio level of the reference
frequency;
p
Cc
 correction of the observed ratio for nonlinearity and limited
resolution of the power meter at power ratio level of the calibration
frequency;
p
p p
p p
=
Sr Xc
Sc Xr
 observed ratio of power ratios derived from:
p
Sr
 indicated power ratio for the reference sensor at the reference
frequency;
p
Sc
 indicated power ratio for the reference sensor at the calibration
frequency;
p
Xr
 indicated power ratio for the sensor to be calibrated at the reference
frequency;
p
Xc
 indicated power ratio for the sensor to be calibrated at the
calibration frequency.
December 1999 rev00 Page 43 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
S6.5 Reference sensor (K
S
): The reference sensor was calibrated six months before the
calibration of the unknown power sensor. The value of the calibration factor, given in
the calibration certificate, is (95,7±1,1) % (coverage factor k = 2), which may also be
expressed as 0,957±0,011.
S6.6 Drift of the standard (δK
D
): The drift of the calibration factor of the reference
standard is estimated from annual calibrations to be 0,002 per year with deviations
within ±0,004. From these values, the drift of the reference sensor, which was
calibrated half a year ago, is estimated to equal 0,001 with deviations within ±0,002.
S6.7 Linearity and resolution of the power meter (p
Cr
, p
Cc
): The expanded uncertainty
of 0,002 (coverage factor k = 2) is assigned to the power meter readings at the power
ratio level of the reference frequency and of 0,0002 (coverage factor k = 2) at the
power ratio level of calibration frequency due to nonlinearity of the power meter
used. These values have been obtained from previous measurements. Since the same
power meter has been used to observe both p
S
and p
X
, the uncertainty contributions
at the reference as well at the calibration frequency are correlated. Because power
ratios at both frequencies are considered, the effect of the correlations is to reduce the
uncertainty. Thus, only the relative difference in the readings due to systematic
effects should be taken into account (see the mathematical note in paragraph S3.12),
resulting in a standard uncertainty of 0,00142 associated with the correction factor
p
Cr
and 0,000142 with the correction factor p
Cc
.
The expanded uncertainty of measurement stated for the readings of the power meter contains
linearity and resolution effects. The linearity effects are correlated whereas the resolution effects are
uncorrelated. As shown in S3.12, building the power ratio cancels the influence of correlations and
gives a reduced standard uncertainty of measurement to be associated with the ratio. In the
calculations above, however, the separated correlated and uncorrelated contributions are not known
and the values given are upper bounds for the standard uncertainty of measurement associated with
ratios. The uncertainty budget finally shows that the contributions arising from these ratios are
insignificant, i.e. the approximations are justified.
S6.8 Mismatch factors (M
Sr
, M
Sc
, M
Xr
M
Xc
): As the transfer standard system is not
perfectly matched and the phase of the reflection coefficients of the transfer standard,
the unknown and the standard power sensors are not known, there will be an
uncertainty due to mismatch for each sensor at the reference frequency and at the
calibration frequency. The corresponding limits of deviation have to be calculated for
the reference and the calibration frequencies from the relationship:
M
S,X G S,X
= ± 1 2Γ Γ (S6.3)
where the magnitudes of the reflection coefficients of the transfer standard, the
reference sensor and the sensor to be calibrated are:
50 MHz 18 GHz
Γ
G
0,02 0,07
Γ
S
0,02 0,10
Γ
X
0,02 0,12
December 1999 rev00 Page 44 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
The probability distribution of the individual contributions is Ushaped. This is taken
into account by replacing the factor 1/3 for a rectangular distribution by 1/2 in
calculating the variance from the square of the halfwidth determined from the limits.
The standard uncertainty due to mismatch is therefore obtained from:
u M ( )
S,X
G S
=
2
2
Γ Γ
(S6.4)
Note: The values of the reflection coefficients are the results of measurements which
are themselves subject to uncertainty. This is accounted for by adding the square root
of the sum of the uncertainty of measurement squared and the measured value
squared.
S6.9 Correlation: None of the input quantities are considered to be correlated to any
significant extent.
S6.10 Measurements (p): Three separate readings are made which involve disconnection
and reconnection of both the reference sensor and the sensor to be calibrated on the
transfer standard to take connector repeatability into account. The power meter
readings used to calculate the observed power ratio p are as follows:
obs. no p
Sr
p
Sc
p
Xr
p
Xc
p
1 1,0001 0,9924 1,0001 0,9698 0,9772
2 1,0000 0,9942 1,0000 0,9615 0,9671
3 0,9999 0,9953 1,0001 0,9792 0,9836
arithmetic mean: p = 0 976 , 0
s
experimental standard deviation: p ( ) = 0,0083
standard uncertainty: u p s p ( ) ( ) = = =
0 0083
3
0 0048
,
,
S6.11 Uncertainty budget (K
X
):
quantity
X
i
estimate
x
i
standard
uncertainty
u(x
i
)
probability
distribution
sensitivity
coefficient
c
i
uncertainty
contribution
u
i
(y)
K
S
0,957 0,0055 normal 0,976 0,00537
δK
D
0,001 0,0012 rectangular 0,976 0,00113
M
Sr
1,000 0,0006 Ushaped 0,933 0,00053
M
Sc
1,000 0,0099 Ushaped 0,933 0,00924
M
Xr
1,000 0,0006 Ushaped 0,933 0,00053
M
Xc
1,000 0,0119 Ushaped 0,933 0,01110
p
Cr
1,000 0,0014 normal 0,933 0,00132
p
Cc
1,000 0,0001 normal 0,933 0,00013
p 0,976 0,0048 normal 0,956 0,00459
K
X
0,933 0,01623
December 1999 rev00 Page 45 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
S6.12 Expanded uncertainty:
U = k × u(K
X
) = 2 × 0,01623 ≅ 0,032
S6.13 Reported result:
The calibration factor of the power sensor at 18 GHz is 0,933 ±0,032, which may
also be expressed as (93,3±3,2) %.
The reported expanded uncertainty of measurement is stated as the standard
uncertainty of measurement multiplied by the coverage factor k = 2, which for a
normal distribution corresponds to a coverage probability of approximately 95 %.
S7 CALIBRATION OF A COAXIAL STEP ATTENUATOR AT A SETTING OF
30 DB (INCREMENTAL LOSS)
S7.1 The measurement involves the calibration of a coaxial step attenuator at 10 GHz
using an attenuation measuring system containing a calibrated step attenuator which
acts as the attenuation reference. The method of measurement involves the
determination of the attenuation between matched source and matched load. In this
case the unknown attenuator can be switched between settings of 0 dB and 30 dB and
it is this change (called incremental loss) that is determined in the calibration
process. The attenuation measuring system has a digital readout and an analogue null
detector which is used to indicate the balance condition.
S7.2 Schematic of the measuring system
Γ
L
Γ
G
s
22a
s
11a
s
11b
s
22b
Step attenuator
RF attenuation measuring system
G
dB 0 0 0 3
30.052 dB
S7.3 The attenuation L of the attenuator to be calibrated is obtained from the relation:
X
L
X
= L
S
+ δL
S
+ δL
D
+ δL
M
+ δL
K
+ δL
ib
 δL
ia
+ δL
0b
 δL
0a
(S7.1)
where:
L
S
= L
ib
– L
ia
 attenuation difference of reference attenuator derived from:
L
ia
 indicated attenuation with the attenuator to be calibrated, set at
0 dB;
L
ib
 indicated attenuation with the attenuator to be calibrated, set at
30 dB;
δL
S
 correction obtained from the calibration of the reference
attenuator;
δL
D
 change of the attenuation of the reference attenuator since its
last calibration due to drift;
δL
M
 correction due to mismatch loss;
December 1999 rev00 Page 46 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
δL
K
 correction for leakage signals between input and output of the
attenuator to be calibrated due to imperfect isolation;
δL
ia
, δL
ib
 corrections due to the limited resolution of the reference detector
at 0 dB and 30 dB settings;
δL
0a
, δL
0b
 corrections due to the limited resolution of the null detector at
0 dB and 30 dB settings.
S7.4 Reference attenuator (δL
S
): The calibration certificate for the reference attenuator
gives a value of attenuation for the 30,000 dB setting at 10 GHz of 30,003 dB with
an associated expanded uncertainty of 0,005 dB (coverage factor k = 2). The
correction of +0,003 dB with the associated expanded uncertainty of 0,005 dB
(coverage factor k = 2) is considered to be valid for attenuation settings of the
reference attenuator that differ not more than ±0,1 dB from the calibrated setting of
30,000 dB.
S7.5 Drift of the reference (δL
D
): The drift of the attenuation of the reference attenuator
is estimated from its calibration history to be zero with limits ±0,002 dB.
S7.6 Mismatch loss (δL
M
): The reflection coefficients of the source and the load at the
insertion point of the attenuator to be calibrated have been optimised by impedance
matching to as low magnitudes as possible. Their magnitudes and the magnitudes of
the scattering coefficients of the attenuator to be calibrated have been measured but
their phase remains unknown. Without any phase information, a correction for
mismatch error cannot be made, but the standard uncertainty (in dB) due to the
incomplete knowledge of the match is estimated from the relationship [1]:
u L = s s s s s s ( ) ( ) ( ) ( ) δ
M S 11a 11b L 22a 22b S L 21a 21b
8 686
2
2 2 2 2 2 2 2 2 4 4 ,
Γ Γ Γ Γ + + + + × +
(S7.2)
with the source and load reflection coefficients
Γ
L
= 0,03 and Γ
S
= 0,03
and the scattering coefficients of the attenuator to be calibrated at 10 GHz
0 dB 30 dB
s
11
0,05 0,09
s
22
0,01 0,01
s
21
0,95 0,031
as u(δL
M
) = 0,02 dB.
Note: The values of scattering and reflection coefficients are the results of
measurements which are themselves not exactly known. This is accounted
for by adding the square root of the sum of uncertainty of measurement
squared and the measured value squared.
December 1999 rev00 Page 47 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
S7.7 Leakage correction (δL
K
): Leakage signals through the attenuator to be calibrated
have been estimated from the measurements at 0 dB setting to be at least 100 dB
below the measurement signal. The correction for leakage signals is estimated from
these findings to be within ±0,003 dB at the 30 dB setting.
S7.8 Resolution of the reference attenuator setting (δL
ia
, δL
ib
): The digital readout of
the reference attenuator has a resolution of 0,001 dB from which the correction for
resolution is estimated to be within ±0,0005 dB.
S7.9 Resolution of the null detector (δL
0a
, δL
0b
): The detector resolution was
determined from a previous evaluation to have a standard deviation of 0,002 dB at
each reading with assumed normal probability distribution.
S7.10 Correlation: None of the input quantities are considered to be correlated to any
significant extent.
S7.11 Measurements (L
S
): Four observations are made of the incremental loss of the
attenuator to be calibrated between settings of 0 dB and 30 dB:
obs. no. obs. values at
0 dB setting 30 dB setting
1 0,000 dB 30,033 dB
2 0,000 dB 30,058 dB
3 0,000 dB 30,018 dB
4 0,000 dB 30,052 dB
arithmetic mean: L
S
= 30 040 , dB
experimental standard deviation: s(L
S
) = 0,018 dB
standard uncertainty: u L s L ( ) ( )
S S
= = =
0 018
4
0 009
,
,
dB
dB
S7.12 Uncertainty budget (L
X
):
quantity
X
i
estimate
x
i
standard
uncertainty
u(x
i
)
probability
distribution
sensitivity
coefficient
c
i
uncertainty
contribution
u
i
(y)
L
S
30,040 dB 0,0090 dB normal 1,0 0,0090 dB
δL
S
0,003 dB 0,0025 dB rectangular 1,0 0,0025 dB
δL
D
0 dB 0,0011 dB Ushaped 1,0 0,0011 dB
δL
M
0 dB 0,0200 dB Ushaped 1,0 0,0200 dB
δL
K
0 dB 0,0017 dB Ushaped 1,0 0,0017 dB
δL
ia
0 dB 0,0003 dB Ushaped 1,0 0,0003 dB
δL
ib
0 dB 0,0003 dB rectangular 1,0 0,0019 dB
δL
0a
0 dB 0,0020 dB rectangular 1,0 0,0020 dB
δL
0b
0 dB 0,0020 dB normal 1,0 0,0020 dB
L
X
30,043 dB 0,0224 dB
December 1999 rev00 Page 48 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
S7.13 Expanded uncertainty:
U = k × u(L
X
) = 2 × 0,0224 dB ≅ 0,045 dB
S7.14 Reported result:
The measured value of the step attenuator for a setting of 30 dB at 10 GHz is
(30,043 ±0,045) dB.
The reported expanded uncertainty of measurement is stated as the standard
uncertainty of measurement multiplied by the coverage factor k = 2, which for a
normal distribution corresponds to a coverage probability of approximately 95 %.
S7.15 Reference
[1] Harris, I. A. ; Warner, F. L. : Reexamination of mismatch uncertainty when
measuring microwave power and attenuation. In: IEE Proc., Vol. 128, Pt.
H, No. 1, Febr. 1981
December 1999 rev00 Page 49 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
SUPPLEMENT 2
Examples
December 1999 rev00 Page 50 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
CONTENTS
S8 INTRODUCTION 52
S9 CALBRATION OF A HANDHELD DIGITAL MULTIMETER
AT 100 V DC 55
S10 CALIBRATION OF A VERNIER CALLIPER 59
S11 CALIBRATION OF A TEMPERATURE BLOCK CALIBRATOR
AT A TEMPERATURE OF 180°C 64
S12 CALIBRATION OF A HOUSEHOLD WATER METER 68
S13 CALIBRATION OF A RING GAUGE WITH A NOMINAL
DIAMETER OF 90 MM 72
December 1999 rev00 Page 51 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
S8 INTRODUCTION
S8.1 The following examples are chosen to demonstrate further the method of evaluating
the uncertainty of measurement. They supplement the examples presented in
Supplement 1 to EALR2 (Edition 1, November 1997). The present collection of
examples focuses on situations where there are one or two dominant terms in the
uncertainty propagation or where the number of repeated measurements is small.
S8.2 The examples are chosen to illustrate situations encountered in practice. It should be
emphasised, however, that in practical applications there is no need to go through the
mathematical derivations presented in these examples, in particular in the
mathematical notes appended to some of the examples. Rather, the user is
encouraged to employ the results of the theoretical presentations after having made
himself acquainted with the conditions that have to be fulfilled. For instance, if it is
ascertained, in a given situation, that the result of measurement has a rectangular
distribution (as would be the case if there were only one term, rectangularly
distributed, that needed to be considered in the propagation), one can immediately
draw the conclusion that the coverage factor to be used to arrive at a coverage
probability of 95 % is k = 1,65 (see S9.14).
S8.3 One general conclusion that may be drawn from the uncertainty propagation is that in
the case of only one dominant contribution the type of distribution of this
contribution applies for the result of measurement as well. However, to evaluate the
uncertainty of the result of measurement, the applicable sensitivity coefficient has to
be employed, as usual.
S8.4 It should be added that the situation where there is only one or a few dominant terms
to the uncertainty of measurement is often met in connection with less complicated
measuring instruments, where the dominant term often is due to the limited
resolution of the instrument. Thus it may appear a paradox that the treatment of
uncertainty of measurement for less complicated instruments, as shown by the
examples of this Supplement, is more complicated than the treatment of the more
straightforward examples in Supplement 1. However, it should be kept in mind that
the mathematical derivations, which may be felt as complications, are inserted for
pedagogical reasons at places where they are needed instead of presenting them in
the main document.
S8.5 The examples are based on drafts prepared by EA Expert Groups. These drafts have
been simplified and harmonised to make them transparent to the laboratory staff in
all fields of calibration. It is thus hoped that this set of examples, like the preceding
set published as Supplement 1 to EALR2, will contribute to a better understanding
of the details of setting up the model of evaluation and to the harmonisation of the
process of evaluating the uncertainty of measurement, independent of the field of
calibration.
December 1999 rev00 Page 52 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
S8.6 The contributions and values given in the examples are not intended to imply
mandatory or preferred requirements. Laboratories should determine the uncertainty
contributions on the basis of the model function they use in the evaluation of the
particular calibration they perform and report the evaluated uncertainty of
measurement on the calibration certificate they issue.
S8.7 The presentation of the examples follows the common scheme presented and
implemented in the first supplement to EALR2. For details the reader is referred to
clause S1.4 of that document.
S8.8 The uncertainty analysis of the examples is intended to represent the fundamentals of
the specific measurement process and the method of evaluating the measurement
result and the associated uncertainty. To keep the analysis transparent, also for those
who are not experts in the relevant metrological field, a uniform method for the
choice of the symbols of quantities has been followed, focused more on the physical
background than on the current practice in different fields.
S8.9 There are several recurrent quantities involved in all cases. One of them is the mea
surand, i.e. the quantity to be measured, another is the quantity presented by the
working standard, which realises the local unit; with this quantity the measurand is
compared. Besides these two quantities there are several others, in all cases, which
take the role of additional local quantities or corrections.
S8.10 Corrections describe the imperfect equality between a measurand and the result of a
measurement. Some of the corrections are given by complete results of measurement,
i.e. a measured value and its associated measurement uncertainty. For others the
distribution of values is inferred from more or less complete knowledge of their
nature. In most cases this will lead to an estimation of the limits for the unknown
deviations.
S8.11 In certain cases the quantity presented by a working standard is characterised by the
nominal value of the standard. Thus nominal values, which generally speaking
characterise or identify calibration artefacts, often enter the uncertainty analysis.
S8.12 To distinguish in the mathematical models of evaluation between these concepts, the
examples have been designed to follow the notational rules given below. It is
evident, however, that it is not possible to follow such rules strictly, because the
practice concerning the use of symbols is different in different metrological fields.
S8.13 The notation applied here distinguishes between main values, nominal values,
correction values and values of limits:
Main values are measured or observed values that contribute an essential part to the
value of a measurand. They are represented by lowercase letters in italics; they will
be preceded by an uppercase Greek delta if the quantity represents a difference.
December 1999 rev00 Page 53 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
EXAMPLE:
t
i X
 temperature indicated by a thermometer X to be calibrated. (index i means
indicated),
l ∆  observed difference in the displacement of a measuring spindle.
Nominal values are assigned values of the realisation of a quantity by a standard or a
measuring instrument. They are approximate values that give the main part of the
realised value. They are represented by uppercase letters in italics.
EXAMPLE:
L  nominal length of a gauge block to be calibrated.
Correction values give small deviations from the main values that are known or have
to be estimated. In most cases they are additive. They are represented by the symbol
chosen for the quantity under consideration, preceded by a lowercase Greek delta.
EXAMPLE:
δ m
D
 possible deviation because of the drift of the value of a reference weight
since its last calibration
δ m
C
 correction for eccentricity of load and magnetic effects in the calibration of
a weight.
Values of limits are fixed, estimated values of possible variations of the unknown
values of a quantity. They are represented by the symbol chosen for the quantity
under consideration, preceded by a uppercase Greek delta.
EXAMPLE:
X
α ∆  estimated halfwidth of the interval of possible deviations of a linear thermal
resistivity coefficient given in a manufacturer’s specification for a resistor to
be calibrated.
The differentiation between different quantities of the same kind is effected by
indices as shown in the examples. The internationally accepted notational rules for
physical quantities have been followed: indices representing physical quantities are
given in italics whereas indices that symbolise artefacts, instruments and so on are
written in upright letters.
S8.14 Defined reference values are represented by a quantity symbol with the index zero.
EXAMPLE:
p
0
 reference pressure, e.g. of 1000 mbar.
December 1999 rev00 Page 54 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
S8.15 Ratios of quantities of the same kind (dimensionless ratios) are represented by lower
case letters in italics.
EXAMPLE:
r R R =
i X i N
/  ratio of indicated resistance of an unknown resistor and a reference
resistor (index i means indicated).
S8.16 If several indices are used, the sequence of indices is chosen in such a way that the
index representing the most general concept is leftmost and the one representing the
most specific concept is rightmost.
EXAMPLE:
V
i 1
,V  voltage indicated by voltmeter '1' and voltmeter '2', respectively
i 2
S8.17 The examples in this second supplement to EALR2 are intended to be followed by
others, illustrating different aspects encountered in connection with the calibration of
measuring instruments. Examples may also be found in EAL and EA Guidance
Documents
1
dealing with the calibration of specific types of measuring instruments.
S9 CALBRATION OF A HANDHELD DIGITAL MULTIMETER AT 100 V DC
S9.1 As part of a general calibration, a handheld digital multimeter (DMM) is calibrated
at an input of 100 V DC using a multifunction calibrator as a working standard. The
following measuring procedure is used:
(1) The calibrator’s output terminals are connected to the input terminals of the DMM
using suitable measuring wires.
(2) The calibrator is set to its 100V setting and, after a suitable stabilising period, the
DMM reading is noted.
(3) The error of indication of the DMM is calculated using the DMM readings and the
calibrator settings.
S9.2 It must be noted that the error of indication of the DMM which is obtained using this
measuring procedure includes the effect of offset as well as deviations from linearity.
S9.3 The error of indication of the DMM to be calibrated is obtained from E
X
S X i S X i X
V V V V E δ δ − + − = (S9.1)
where
1
EALG26, Calibration of pressure balances
EALG31, Calibration of thermocouples
EALG32, Measurement and generation of small ac voltages with inductive voltage dividers
EA10/10, EA Guidelines on the Determination of Pitch Diameter of Parallel Thread gauges by Mechanical
Probing
December 1999 rev00 Page 55 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
V
i X
 voltage, indicated by the DMM (index i means indication),
V
S
 voltage generated by the calibrator,
X i
V δ  correction of the indicated voltage due to the finite resolution
of the DMM,
S
V δ  correction of the calibrator voltage due to
(1) drift since its last calibration,
(2) deviations resulting from the combined effect of
offset, nonlinearity and differences in gain,
(3) deviations in the ambient temperature,
(4) deviations in mains power,
(5) loading effects resulting from the finite input
resistance of the DMM to be calibrated.
S9.4 Because of the limited resolution of the indication of the DMM, no scatter in the
indicated values is observed.
S9.5 DMM readings (V )
X i
The DMM indicates the voltage 100,1 V at the calibrator setting 100 V. The DMM
reading is assumed to be exact (see S9.4).
S9.6 Working standard (V )
S
The calibration certificate for the multifunction calibrator states that the voltage
generated is the value indicated by the calibrator setting and that the associated
expanded relative uncertainty of measurement is W=0,000 02 (coverage factor
), resulting in an expanded uncertainty of measurement associated with the
100 V setting of U=0,002 V (coverage factor k
k = 2
= 2).
S9.7 Resolution of DMM to be calibrated (
X i
V δ )
The least significant digit of the DMM display corresponds to 0,1 V. Each DMM
reading has a correction due to the finite resolution of the display which is estimated
to be 0,0 V with limits of +0.05 V (i.e. one half of the magnitude of the least
significant digit).
S9.8 Other corrections (
S
V δ )
Because of the fact that individual figures are not available the uncertainty of
measurement associated with the different sources is derived from the accuracy
specification given by the manufacturer of the calibrator. These specifications state
that the voltage generated by the calibrator coincides with the calibrator setting
within ±(0,000 1×V
S
+1 mV)
2
under the measuring conditions
(1) the ambient temperature is within the range 18 °C to 23 °C
(2) the mains voltage powering the calibrator is in the range 210 V to 250 V,
(3) the resistive load at the calibrator’s terminals is greater than 100 kΩ,
2
A widely used method of presenting accuracy specification of measuring instruments in data sheets or manuals
consists in giving the specification limits in terms of ‘settings’. For the calibrator, the statement would be
±(0,01% of setting + 1 mV). Even if this method is considered to be equivalent to the expression given above it
is not used here because it may be misleading in many cases and because it does not represent an equation of
physical quantities in the internationally accepted symbolic nomenclature.
December 1999 rev00 Page 56 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
(4) the calibrator has been calibrated within the last year.
Since these conditions of measurement are fulfilled and the calibration history of the
calibrator shows that the manufacturer’s specification may be relied upon, the
correction to be applied to the voltage generated by the calibrator is assumed to be
0,0 V within ±0,011 V.
S9.9 Correlation
None of the input quantities are considered to be correlated to any significant extent.
S9.10 Uncertainty budget ( )
X
E
quantity
i
X
estimate
i
x
standard
uncertainty
) (
i
x u
probability
distribution
sensitivity
coefficient
i
c
uncertainty
contribution
) ( y u
i
V
i X
100,1 V    
V
S
100,0 V 0,001 V normal 1,0 0,001 V
X i
V δ
0,0 V 0,029 V rectangular 1,0 0,029 V
S
V δ
0,0 V 0,0064 V rectangular 1,0 0,0064 V
E
X
0,1 V 0,030 V
S9.11 Expanded uncertainty
The standard uncertainty of measurement associated with the result is clearly
dominated by the effect of the finite resolution of the DMM. The final distribution is
not normal but essentially rectangular. Therefore, the method of effective degrees of
freedom described in Annex E of EALR2 is not applicable. The coverage factor
appropriate for a rectangular distribution is calculated from the relation given in eq.
(S9.8) in the mathematical note S9.14.
V 0,05 V 030 , 0 65 , 1 ) ( ≅ ⋅ = ⋅ =
X
E u k U
S9.12 Reported result
The measured error of indication of the handheld digital voltmeter at 100 V is
(0,10 ±0,05) V.
The reported expanded uncertainty of measurement is stated as the standard
uncertainty of measurement multiplied by the coverage factor which has
been derived from the assumed rectangular probability distribution for a coverage
probability of 95%.
65 , 1 = k
S9.13 Additional remark
The method used for calculating the coverage factor is clearly related to the fact that
the measurement uncertainty associated with the result is dominated by the effect of
the finite resolution of the DMM. This will be true for the calibration of all low
resolution indicating instruments provided the finite resolution is the only dominant
source in the uncertainty budget.
December 1999 rev00 Page 57 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
S9.14 Mathematical note
If the situation of measurement is such that one of the uncertainty contributions in
the budget can be identified as a dominant term, for instance the term with index 1,
the standard uncertainty to be associated with the measurement result can be
written as
y
) ( ) ( ) (
2
1
y u y u y u
2
R
+ = . (S9.2)
Here is
∑
=
=
N
i
i
y u y u
2
2
) ( ) (
R
(S9.3)
denotes the total uncertainty contribution of the nondominant terms. As long as the
ratio of the total uncertainty contribution u of the nondominant terms to the
uncertainty contribution u of the dominant term is not larger than 0,3, eq. (S9.2)
may be approximated by
y
R
( )
y
1
( )
(
(
¸
(
¸


.

\

+ ⋅ ≅
2
1
1
) (
) (
2
1
1 ) ( ) (
y u
y u
y u y u
R
. (S9.4)
The relative error of approximation is smaller than 1 . The maximum relative
change in the standard uncertainty resulting from the factor within the brackets in eq.
(S9.4) is not larger than 5%. This value is within the accepted tolerance for
mathematical rounding of uncertainty values.
3
10
−
×
Under these assumptions the distribution of values that could reasonably be
attributed to the measurand is essentially identical with the distribution resulting
from the known dominant contribution. From this distribution density ϕ( ) y the
coverage probability p may be determined for any value of the expanded
measurement uncertainty U by the integral relation
' . (S9.5)
∫
+
−
=
U y
U y
dy y U p ) ' ( ) ( ϕ
Inverting this relation for a given coverage probability results in the relation between
the expanded measurement uncertainty and the coverage probability U U p = ( ) for
the given distribution density ϕ( ) y . Using this relation, the coverage factor may
finally be expressed as
) (
) (
) (
y u
p U
p k = . (S9.6)
December 1999 rev00 Page 58 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
In the case of the handheld digital voltmeter the dominant uncertainty contribution
resulting from the finite resolution of the indication is u whereas
the total uncertainty contribution of the nondominant terms is .
The relevant ratio is
V 029 , 0 ) ( =
X
X
E
V δ
, 0 ) ( =
X R
E u V 0064
22 , 0 ) ( / ) ( =
X X R
X
E u E u
V δ
. Thus the resulting distribution of
values that can reasonably be attributed as errors of indications is essentially
rectangular. The coverage probability for a rectangular distribution is linearly related
to the expanded measurement uncertainty ( a being the halfwidth of the rectangular
distribution)
a
U
p = . (S9.7)
Solving this relation for the expanded measurement uncertainty U and inserting the
result together with the expression of the standard measurement uncertainty related
to a rectangular distribution as given by eq. (3.8) of EALR2 finally gives the
relation
3 ) ( p p k = . (S9.8)
For a coverage probability p = 95 % applicable in the EA, the relevant coverage
factor is thus k = 1 65 , .
S10 CALIBRATION OF A VERNIER CALLIPER
S10.1 A vernier calliper made of steel is calibrated against grade I gauge blocks of steel
used as working standards. The measurement range of the calliper is 150 mm. The
reading interval of the calliper is 0,05 mm (the main scale interval is 1 mm and the
vernier scale interval 1/20 mm). Several gauge blocks with nominal lengths in the
range 0,5  150 mm are used in the calibration. They are selected in such a way that
the measurement points are spaced at nearly equal distances (e.g. at 0 mm, 50 mm,
100 mm, 150 mm) but give different values on the vernier scale (e.g. 0,0 mm,
0,3 mm, 0,6 mm, 0,9 mm). The example concerns the 150 mm calibration point for
measurement of external dimensions. Before calibration several checks of the
condition of the calliper are made. These include dependence of the result of
measurement on the distance of the measured item from the beam (Abbe error),
quality of the measuring faces of the jaws (flatness, parallelism, squareness), and
function of the locking mechanism.
December 1999 rev00 Page 59 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
S10.2 The error of indication of the calliper at the reference temperature t is
obtained from the relation:
X
E
0
20 = °C
M X i S S X i X
l l t L l l E δ δ α + + ∆ ⋅ ⋅ + − = (S10.1)
where:
X i
l
 indication of the calliper,
l
S
 length of the actual gauge block,
L
S
 nominal length of the actual gauge block,
α  average thermal expansion coefficient of the calliper and the gauge
block,
∆t
 difference in temperature between the calliper and the gauge block,
X i
l δ
 correction due to the finite resolution of the calliper,
M
l δ
 correction due to mechanical effects, such as applied measurement
force, Abbe errors, flatness and parallelism errors of the measurement
faces.
S10.3 Working standards ( l , )
S
L
S
The lengths of the reference gauge blocks used as working standards, together with
their associated expanded uncertainty of measurement, are given in the calibration
certificate. This certificate confirms that the gauge blocks comply with the
requirements for grade I gauge blocks according to ISO 3650, i.e. that the central
length of the gauge blocks coincides within ±0,8 µm with the nominal length. For the
actual lengths of the gauge blocks their nominal lengths are used without correction,
taking the tolerance limits as the upper and lower limits of the interval of variability.
S10.4 Temperature ( , ∆t α )
After an adequate stabilisation time, the temperatures of the calliper and the gauge
block are equal within ±2 °C. The average thermal expansion coefficient is 11,5·10
6
°C
1
. (The uncertainty in the average thermal expansion coefficient and in the
difference of the thermal expansion coefficients has not been taken into account; its
influence is considered negligible for the present case. Cf. EALR2S1, example S4.)
S10.5 Resolution of the calliper (
X i
l δ )
The scale interval of the vernier scale is 0,05 mm. Thus variations due to the finite
resolution are estimated to have rectangular limits of ± 25 µm.
S10.6 Mechanical effects (
M
l δ )
These effects include the applied measurement force, the Abbe error and the play
between the beam and the sliding jaw. Additional effects may be caused by the fact
that the measuring faces of the jaws are not exactly flat, not parallel to each other and
not perpendicular to the beam. To minimise effort, only the range of the total
variation, equal to ±50 µm is considered.
S10.7 Correlation
None of the input quantities are considered to be correlated to any significant extent.
December 1999 rev00 Page 60 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
S10.8 Measurements ( )
X i
l
The measurement is repeated several times without detecting any scatter in the
observations. Thus uncertainty due to limited repeatability does not give a
contribution. The result of measurement for the 150 mm gauge block is 150,10 mm.
S10.9 Uncertainty budget (δ l
X
)
quantity
X
i
estimate
x
i
standard
uncertainty
u x
i
( )
probability
distribution
sensitivity
coefficient
c
i
uncertainty
contribution
u y
i
( )
X i
l 150,10 mm    
l
S
150,00 m 0,46 µm rectangular 1,0 0,46 µm
∆t
0 1,15 K rectangular 1,7 µMk
1
2,0 µm
X i
l δ
0 15 µm rectangular 1,0 15 µm
M
l δ 0 29 µm rectangular 1,0 29 µm
X
E
0,10 mm 33 µm
S10.10 Expanded uncertainty
The uncertainty of measurement associated with the result is clearly dominated by
the combined effect of the measurement force and the finite resolution of the vernier.
The final distribution is not normal but essentially trapezoidal with a ratio β = 0 33 ,
of the halfwidth of the plateau region to the halfwidth of the variability interval.
Therefore the method of effective degrees of freedom described in EALR2, Annex
E is not applicable. The coverage factor k = 1,83 appropriate for this trapezoidal
distribution of values is calculated from eq. (S10.10) of the mathematical note
S10.13. Thus
mm 06 , 0 mm 033 , 0 83 , 1 ) ( ≅ ⋅ = ⋅ =
X
E u k U
S10.11 Reported result
At 150 mm the error of indication of the calliper is (0,10 ± 0,06) mm.
The reported expanded uncertainty of measurement is stated as the standard
uncertainty of measurement multiplied by the coverage factor which has
been derived from the assumed trapezoidal probability distribution for a coverage
probability of 95 %.
k = 183 ,
S10.12 Additional remark
The method used for calculating the coverage factor is clearly related to the fact that
uncertainty of measurement associated with the result is dominated by two
influences: the mechanical effects and the finite resolution of the vernier scale. Thus
the assumption of a normal distribution for the output quantity is not justified and the
conditions of EALR2, paragraph 5.6 apply. In the sense that probabilities and
probability densities in practice may only be determined to within 3 %− 5 %, the
distribution is essentially trapezoidal, obtained by convolution of the two rectangular
distributions associated with the dominant contributions. The halfwidths of the base
and the top of the resulting symmetrical trapezoid are 75 µm and 25 µm,
December 1999 rev00 Page 61 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
respectively. 95 % of the area of the trapezoid is encompassed by an interval ±60 µm
around its symmetry axis, corresponding to k = 183 , .
S10.13 Mathematical note
If the situation of measurement is such that two of the uncertainty contributions in
the budget can be identified as dominant terms, the method presented in S9.14 can be
applied when the two dominant contributions, for instance the terms with indices 1
and 2 , are combined into one dominant term. The standard uncertainty to be
associated with the measurement result y may be written in this case as
u y u y u y ( ) ( ) ( ) = +
0
2
R
2
(S10.2)
where
u y u y u y
0
( ) ( ) ( ) = +
1
2
2
2
(S10.3)
denotes the combined contribution of the two dominant terms and
u y u y
i
i
N
R
( ) ( ) =
=
∑
2
3
(S10.4)
the total uncertainty contribution of the remaining nondominant terms. If the two
dominant contributions arise from rectangular distributions of values with half
widths and , the distribution resulting from convolving them is a symmetrical
trapezoidal distribution
a
1
a
2
0
0,2
0,4
0,6
0,8
1
2 1 0 1 2
normalized deviation [y/a]
p
r
o
b
a
b
i
l
i
t
y
d
e
n
s
i
t
y
[
p
]
Fig. 1: Unified symmetrical trapezoidal probability distribution with the value
β=0,33 of the edge parameter, resulting from the convolution of two
rectangular distributions.
December 1999 rev00 Page 62 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
with halfwidths
and a a a = +
1 2
b a (S10.5) a = −
1 2
of the base and the top, respectively (see example in Fig. 1). The distribution may be
conveniently expressed in the unified form
¦
¦
¹
¦
¦
´
¦
<
≤ ≤ ⋅


.

\

−
−
⋅ <
×
+
=
y a
a y a
a
y
a y
a
y
0
1
1
1
1
) 1 (
1
) ( β
β
β
β
ϕ (S10.6)
with the edge parameter
β = =
−
+
b
a
a a
a a
1 2
1 2
(S10.7)
The square of the standard measurement uncertainty deduced from the trapezoidal
distribution of eq. (S10.6) is
u y
a
2
2
2
6
1 ( ) ( ) = + β . (S10.8)
Using the distribution of eq. (S10.6) the dependence of the coverage factor on the
coverage probability is derived according to the method sketched in S9.14
⋅
¦
¦
¹
¦
¦
´
¦
−
≤ − − −
<
−
+
×
+
=
p
p
p
p
p p
p k
2
) 1 )( 1 ( 1
2 2
) 1 (
6
1
1
) (
2
2
β β
β
β
β
(S10.9)
Fig. 2 shows the dependence of the coverage factor k on the value of the edge
parameter β for a coverage probability of 95 %.
December 1999 rev00 Page 63 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
1,5
1,6
1,7
1,8
1,9
2
0 0,2 0,4 0,6 0,8 1
edge parameter [beta]
c
o
v
e
r
a
g
e
f
a
c
t
o
r
[
k
]
Fig. 2: Dependence of the coverage factor k on the value of the edge parameter β of
a trapezoidal distribution for a coverage probability of 95 %.
The coverage factor for a coverage probability of 95 % appropriate to a trapezoidal
distribution with an edge parameter of 95 , 0 < β is calculated from the relation
6
1
) 1 )( 1 ( 1
2
2
β
β
+
− − −
=
p
k . (S10.10)
S11 CALIBRATION OF A TEMPERATURE BLOCK CALIBRATOR AT A
TEMPERATURE OF 180°C
3
S11.1 As part of a calibration, the temperature that has to be assigned to the calibration
bore of a temperature block calibrator, is measured. This is done when the indication
of the builtin temperature indicator has stabilised at 180,0 °C. The temperature of
the calibration bore is determined by an inserted platinum resistance thermometer,
used as a working standard, by measuring the electrical resistance of the thermometer
by an ac resistance bridge. The temperature , that has to be assigned as the
temperature of the bore when the reading of the builtin temperature indicator is
180,0 °C, is given by:
t
X
V H A R X i D S S X
t t t t t t t t t δ δ δ δ δ δ δ + + + + − + + = (S11.1)
3
A similar example will be found in the EA guideline EA10/xx, Calibration of temperature block calibrators.
It has been included here, in a dimplified form, in order to highligt how a value is assigned to an indication of
an instrument in a calibration process. This process is basic for calibrations in different metrological fields
and, therefore, of general interest. The example further demonstrates that there are two equivalent ways to
tackle this problem: the direct assignment of a value to the indication of the instrument and the association of
a correction to the indication, usually called the error of indication.
December 1999 rev00 Page 64 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
where:
t
S
 temperature of the working standard derived from the ac
resistance measurement,
δ t
S
 temperature correction due to the ac resistance measurement,
δ t
D
 temperature correction due to drift in the value of the
working standard since its last calibration,
δ t
i X
 temperature correction due to the settability limitations of the
block temperature calibrator,
R
t δ  temperature correction due to the radial temperature
difference between the builtin thermometer and the working
standard,
A
t δ  temperature correction due to the axial inhomogeneity of
temperature in the measuring bore,
δ t
H
 temperature correction due to hysteresis in the increasing and
decreasing branches of the measuring cycle,
δ t
V
 temperature variation within the time of measurement.
Temperature corrections due to stem conduction are not considered, since the
platinum resistance thermometer used as working standard has an outer diameter
. Prior investigations have shown that stem conduction effects can be
neglected in this case.
d ≤ 6 mm
S11.2 Working standard ( ) t
S
The calibration certificate of the resistance thermometer used as working standard
gives the relationship between resistance and temperature. The measured resistance
value corresponds to a temperature of 180,1 °C, with an associated expanded
uncertainty of measurement U = 30 mK (coverage factor k = 2 ).
S11.3 Determination of the temperature by resistance measurement(δ t
S
)
The temperature of the resistance thermometer used as working standard is
determined as 180,1 °C. The standard measurement uncertainty associated with the
resistance measurement converted to temperature corresponds to u t ( ) δ
S
= 10 mK.
S11.4 Drift of the temperature of the working standard (δ t
D
)
From general experience with platinum resistance thermometers of the type used as
working standard in the measurement, the change of temperature due to resistance
ageing since the last calibration of the standard is estimated to be within the limits
±40 mK.
S11.5 Settability of the block temperature calibrator (δ t
i X
)
The builtin controlling thermometer of the block temperature calibrator has a scale
interval of 0,1 K. This gives temperature resolution limits of ±50 mK within which
the thermodynamic state of the temperature block can be uniquely set.
Note: If the indication of the builtin temperature indicator is not given in units of temperature the
resolution limits must be converted into equivalent temperature values by multiplying the indication
with the relevant instrument constant.
December 1999 rev00 Page 65 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
S11.6 Radial inhomogeneity of temperature (
R
t δ )
The radial temperature difference between the measuring bore and the builtin
thermometer has been estimated to be within ±100 mK.
S11.7 Axial inhomogeneity of temperature (
A
t δ )
The temperature deviations due to axial inhomogeneity of temperature in the
calibration bore have been estimated from readings for different immersion depths to
be within ±250 mK.
S11.8 Hysteresis effects (δ t
H
)
From readings of the reference thermometer during measurement cycles of
increasing and decreasing temperature, the temperature deviation of the calibration
bore due to hysteresis effect has been estimated to be within ±50 mK.
S11.9 Temperature instability (δ t
V
)
Temperature variations due to temperature instability during the measuring cycle of
30 min are estimated to be within ±30 mK.
S11.10 Correlations
None of the input quantities are considered to be correlated to any significant extent.
S11.11 Repeated observations
Due to the finite resolution of the indication of the builtin thermometer no scatter in
the indicated values has been observed and taken into account.
S11.12 Uncertainty budget ( t )
X
Quantity
X
i
Estimate
x
i
Standard
uncertainty
u x
i
( )
Probability
distribution
Sensitivity
coefficient
c
i
Uncertainty
contribution
u y
i
( )
t
S
180,1 °C 15 mK normal 1,0 15 mK
δ t
S
0,0 °C 10 mK normal 1,0 10 mK
δ t
D
0,0 °C 23 mK rectangular 1,0 23 mK
δ t
i X
0,0 °C 29 mK rectangular 1,0 29 mK
R
t δ 0,0 °C 58 mK rectangular 1,0 58 mK
A
t δ 0,0 °C 144 mK rectangular 1,0 144 mK
δ t
H
0,0 °C 29 mK rectangular 1,0 29 mK
δ t
V
0,0 °C 17 mK rectangular 1,0 17 mK
t
X
180,1 °C 164 mK
December 1999 rev00 Page 66 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
S11.13 Expanded uncertainty
The standard uncertainty of measurement associated with the result is clearly
dominated by the effect of the unknown temperature correction due to the axial
temperature inhomogeneity in the measuring bore and the radial temperature
difference between the builtin thermometer and the working standard. The final
distribution is not normal but essentially trapezoidal. According to S10.13, the
coverage factor corresponding to the edge parameter 43 , 0 = β is k 1. 8 , 1 =
K 3 , 0 mK 164 81 , 1 ) (
X
≅ ⋅ = ⋅ = t u k U
S11.14 Reported result
The temperature of the calibration bore that has to be assigned to an indication of the
builtin controlling thermometer of 180,0 °C is 180,1 °C ± 0,3 °C.
The reported expanded uncertainty of measurement is stated as the standard
uncertainty of measurement multiplied by the coverage factor which has
been derived from the assumed trapezoidal probability distribution for a coverage
probability of 95 %.
81 , 1 = k
S11.15 Mathematical note concerning the model
Some metrologists are confused that the indication of the controlling thermometer
does not appear explicitly in the model function of eq. (S11.1). To fit their needs, the
problem can alternatively be formulated with the error of indication
i X X
t t E − = (S11.2)
of the builtin temperature indicator
V H A R X i D S i S X
t t t t t t t t t E δ δ δ δ δ δ δ + + + + − + + − = (S11.3)
The indicated value t is a nominal value. Its effect is to shift the scale of the
measurand. It does, however, not contribute to the uncertainty of measurement
associated with the error of indication
i
(S11.4) u E u t ( ) (
X
= )
X
The model function of eq. (S11.1) can be regained from eq. (S11.3) using the
definition of the error of indication in eq. (S11.2).
This note shows that there is not necessarily only one unique way to choose the
model of evaluation of measurement. The metrologist keeps it in his hands to choose
the model that suits his habits and his approach to the problem. Model functions that
can be transformed mathematically from one into the other represent the same
measurement process. For cases where a continuous scale of indication is involved,
as in the calibration of the temperature block under consideration, model functions
that are connected by linear scale transformations may serve as equivalent
expressions of the measurement problem.
December 1999 rev00 Page 67 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
S12 CALIBRATION OF A HOUSEHOLD WATER METER
S12.1 The calibration of a water meter involves the determination of the relative error of
indication within the applicable flow range of the meter. The measurement is made
using a test rig that supplies necessary water flow with a pressure of approximately
500 kPa, a value typical for municipal tap water systems. The water is received in an
open collecting tank that has been calibrated and determines the reference volume of
the water. It is empty but wetted at the beginning of the measurement. The collecting
tank has a narrow neck with an attached scale by which the filling level can be
detected. The meter to be calibrated is connected between these tanks. It has a
mechanical counter with pointers. The measurement is done at a flow rate of 2500 l/h
with standing startandstop which means that the flow rate is zero both at the
beginning and the end of the measurement. The indication of the meter is recorded at
the beginning and at the end of the measurement. The level is recorded in the
collecting tank at the end of the measurement. The temperature and pressure of the
water at the meter, and the temperature of the water in the collecting tank, are
recorded as well.
S12.2 The relative error of indication in a single run is defined as e
X
1 −
− + ∆
=
X
X1 i X2 i X i
X
V
V V V
e
δ δ
(S12.1)
with
( )( )( ) ) ( 1 ) ( 1 ) ( 1 ) (
0 S X W S X W S S S i S i X
p p t t t t V V V − − − + − + + = κ α α δ (S12.2)
where:
∆V V V
i X i X 2 i X 1
= −
 difference in meter indications,
V
i X 1
, V
i X 2
 indication of the meter at the beginning of the measurement and
at the end of the measurement,
1 X i
V δ ,
2 X i
V δ  corrections due to the finite resolution of the meter indication,
X
V  volume that passed the meter during the measurement under the
prevailing conditions, i.e. pressure and temperature , at the
inlet of the meter,
p
X
t
X
S i
V  volume indicated at the neck scale of the collecting tank at the
end of the measurement,
S i
V δ  correction of the volume indicated at the neck scale of the
collecting tank due to the finite resolution of the scale,
S
α  cubic thermal expansion coefficient of the material of the
collecting tank,
S
t  temperature of the collecting tank,
0
t  reference temperature at which the collecting tank has been
calibrated,
α
W
 cubic thermal expansion coefficient of water,
X
t  temperature of the water at the inlet of the meter,
κ
W
 compressibility of water,
S
p  pressure in the collecting tank (it is zero if excess pressure is
considered)
X
p  pressure of the water at the inlet of the meter.
December 1999 rev00 Page 68 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
S12.3 Collecting tank (V , )
S i 0
t
The calibration certificate states that the neck scale indicates the volume of 200 l at
the reference temperature C 20 ° =
0
t with an associated relative expanded
measurement uncertainty of 0,1 % ( 2 = k ). The expanded measurement uncertainty
associated with the value is 0,2 l ( 2 = k ).
S12.4 Resolution of the collecting tank scale (
S i
V δ )
The water level of the collecting tank can be determined to within ±1 mm. With the
scale factor of the tank of 0,02 l/mm the maximum deviation of the volume of water
in the collecting tank from the observed indicated value is estimated to be within
±0,02 l.
S12.5 Temperature of the water and the collecting tank (
S
α , t )
S
The temperature of the water in the collecting tank is determined to be 15 °C within
±2 K. The stated limits cover all possible sources of uncertainty, such as calibration
of temperature sensors, resolution in reading and temperature gradients in the tank.
The cubic thermal expansion coefficient of the tank material (steel) is taken from a
material handbook to be a constant equal to in the temperature
interval considered. Since there is no uncertainty statement accompanying this value
it is assumed to be known to within its least significant digit. Unknown deviations
are considered to be within the rounding limits of .
1 6
K 10 51
− −
⋅ =
S
α
6
K 10 5 , 0
−
⋅ ±
1 −
S12.6 Temperature of the water at the meter (
W
α , )
X
t
The temperature of the water at the inlet of the meter is determined to be 16 °C
within ±2 K. The stated limits cover all possible sources of uncertainty, such as
contributions from calibration of sensors, resolution in reading and small temperature
changes during one measurement run. The cubic expansion coefficient of water is
taken from a material handbook to be a constant equal to in the
temperature interval considered. Since there is no uncertainty statement
accompanying this value it is assumed to be known to within its least significant
digit. Unknown deviations are considered to be within the rounding limits of
.
1 3
K 10 15 , 0
− −
⋅ =
W
α
1 6
K 10 5 , 0
− −
⋅ ±
S12.7 Pressure difference of the water between the meter and the tank (κ
W
, , )
S
p
X
p
The excess pressure of the water supplied to the inlet of the meter is 500 kPa with
relative deviations not larger than ±10 %. On its way from the inlet to the collecting
tank, the water expands to excess pressure 0 kPa (atmospheric pressure condition).
The compressibility of water is taken from a material handbook to be a constant
equal to in the temperature interval considered. Since there is
no uncertainty statement accompanying this value, it is assumed to be known to
within its least significant digit. Unknown deviations are considered to be within the
rounding limits of ± .
1 6
kPa 10 46 , 0
− −
⋅ =
W
κ
6
10 005 , 0
−
⋅
1
kPa
−
S12.8 Correlation
None of the input quantities are considered to be correlated to any significant extent.
December 1999 rev00 Page 69 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
S12.9 Uncertainty budget (V )
X
quantity
X
i
estimate
x
i
standard
uncertainty
u x
i
( )
probability
distribution
sensitivity
coefficient
c
i
uncertainty
contribution
u y
i
( )
S i
V
200,02 l 0,10 l normal 1,0 0,10 l
S i
V δ
0,0 l 0,0115 l rectangular 1,0 0,0115 l
S
α 51⋅10
6
K
1
0,29⋅10
6
K
1
rectangular 1000 l⋅K 0,29⋅10
3
l
S
t 15°C 1,15 K rectangular 0,0198 l⋅K
1
0,0228 l
W
α 0,15⋅10
3
K
1
2,9⋅10
6
K
1
rectangular 200 l⋅K 0,58⋅10
3
l
X
t 16°C 1,15 K rectangular 0,0300 l⋅K
1
0,0346 l
κ
W
0,46⋅10
6
kPa
1
2,9⋅10
6
kPa
1
rectangular 100 l⋅kPa 0,29⋅10
3
l
X
p 500 kPa 29 kPa rectangular 9,2⋅10

6
l⋅kPa
1
0,0027 l
S
p 0,0 Pa    
X
V 199,95 l 0,109 l
The standard uncertainty of measurement associated with the result is clearly
dominated by the volume indication at the neck scale of the collecting tank. The final
distribution is not normal but essentially rectangular. This must be kept in mind in
the further processing of the uncertainty evaluation.
S12.10 Indication of the meter ( ∆ ,
X i
V
1 X i
V δ ,
2 X i
V δ )
The water meter to be calibrated has a resolution of 0,2 l resulting in the limits ± 0,1 l
in both readings for the maximum deviations resulting from the meter resolution.
S12.11 Uncertainty budget ( e )
X
quantity
X
i
estimate
x
i
standard
uncertainty
u x
i
( )
probability
distribution
sensitivity
coefficient
c
i
uncertainty
contribution
u y
i
( )
X i
V ∆ 200,0 l  nominal  
1 X i
V δ 0,0 l 0,058 l rectangular 5,0⋅10
3
0,29⋅10
3
l
2 X i
V δ 0,0 l 0,058 l rectangular 5,0⋅10
3
0,29⋅10
3
l
X
V 199,95 l 0,109 l rectangular 5,0⋅10
3
0,55⋅10
3
l
X
e 0,000 3 0,68⋅10
3
December 1999 rev00 Page 70 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
S12.12 Repeatability of the meter
The relative error of indication of the water meter to be calibrated, determined at the
same flow rate of 2500 l/h, shows considerable scatter. For that reason the relative
error of indication is determined three times. The results of these three runs are
treated as independent observations in the model that determines the average
error of indication :
j
e
X
av X
e
X X av X
e e e δ + = (S12.3)
where:
X
e  relative error of indication of a single run,
X
e δ  correction of the relative error of indication obtained in the different
runs due to the lack of repeatability of the meter.
S12.13 Measurements ( )
X
e
No. observed relative
error of indication
1 0,000 3
2 0,000 5
3 0.002 2
arithmetic mean: 001 , 0 =
X
e
experimental standard deviation: 001 , 0 ) ( =
j
e s
X
standard uncertainty: 60 000 , 0
3
001 , 0
= = = ) ( ) (
X X
e s e u
S12.14 Uncertainty budget ( e )
av X
quantity
i
X
estimate
i
x
standard
uncertainty
u x
i
( )
degrees of
freedom
eff
ν
probability
distribution
sensitivity
coefficient
c
i
uncertainty
contribution
u y
i
( )
X
e 0,001 0,60⋅10
3
2 normal 1,0 0,60⋅10
3
X
e δ 0,0 0,68⋅10
3
∞ normal 1,0 0,68⋅10
3
av X
e 0,001 10 0,91⋅10
3
December 1999 rev00 Page 71 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
S12.15 Expanded uncertainty
Because of the small number of effective degrees of freedom of the standard
uncertainty associated with the mean relative error of indication the standard
coverage factor has to be modified according to table E1
3 3
10 2 10 91 , 0 28 , 2 ) (
− −
⋅ ≅ ⋅ ⋅ = ⋅ =
av X
e u k U
S12.16 Reported result
The average relative error of indication of the water meter determined at a flow rate
of 2500 l/h is 0,001 ± 0,002.
The reported expanded uncertainty of measurement is stated as the standard
uncertainty of measurement multiplied by the coverage factor , which for a
tdistribution with
28 , 2 = k
10
eff
= ν effective degrees of freedom corresponds to a coverage
probability of approximately 95 %.
S13 CALIBRATION OF A RING GAUGE WITH A NOMINAL DIAMETER OF 90
MM
S13.1 A steel ring gauge of D
x
= 90 mm nominal inner diameter is calibrated applying the
procedure introduced in EALG29. A length comparator of the Abbe type and a steel
setting ring, whose nominal inner diameter (D
s
= 40 mm) differs significantly from
that of the ring to be calibrated, are employed. In this case the length comparator and
the steel setting ring both take the role of working standards. The rings are gently
clamped sequentially on a 4degrees of freedom table, which includes all position
elements for aligning the test pieces. The rings are contacted at several points
diametrically apart by two Cshaped arms, fixed on the steady and the measuring
spindle, respectively. The Cshaped arms are supplied with spherical contact tips.
The measuring force is generated by a tension weight ensuring a constant force of
nominally 1,5 N over the whole measuring range. The measuring spindle is rigidly
connected with the gauge head of a steel line scale of resolution 0,1 µm. The line
scale of the comparator has been verified periodically to meet the manufacturer’s
specification of maximum permissible error.
The ambient temperature is monitored in order to maintain the environmental
conditions stated by calibration procedure. The temperature in the comparator
working volume is maintained at 20 °C within ±0,5 K. Care is taken to ensure that
the rings and the line scale (ruler) maintain the monitored temperature throughout the
calibration.
S13.2 The diameter d of the ring to be calibrated at the reference temperature t
is obtained from the relationship:
X
C ° = 20
0
A E P T i S X
l l l l l l d d δ δ δ δ δ + + + + + ∆ + = (S13.1)
December 1999 rev00 Page 72 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
where:
S
d  diameter of the reference setting ring at the reference temperature,
l ∆  observed difference in displacement of the measuring spindle when the
contact tips touch the inner surface of the rings at two diametrically apart
points,
i
l δ  correction for the errors of indication of the comparator,
T
l δ  correction due to the temperature effects of the ring to be calibrated, the
reference setting ring and the comparator line scale,
δ l
P
 correction due to coaxial misalignment of the probes with respect to the
measuring line,
E
l δ  correction due to the difference in elastic deformations of the ring to be
calibrated and the reference setting ring,
A
l δ  correction due to the difference of the Abbe errors of the comparator
when the diameters of the ring to be calibrated and the reference setting
ring are measured.
S13.3 Working standard ( ) d
S
The inner diameter of the setting ring used as the working standard together with the
associated expanded uncertainty of measurement is given in the calibration
certificate as 40,0007 mm ± 0,2 µm (coverage factor k = 2).
S13.4 Comparator (
i
l δ )
The corrections for the errors of indication of the line scale (ruler) were determined
by the manufacturer and prestored electronically. Any residuals are within the
manufacturers specification of with being the indicated
length. The specifications are ascertained by periodical verifications. For the actual
length difference unknown residuals are estimated to be within
± (0,375) µm.
) 10 5 , 1 m 3 , 0 (
i
6
l ⋅ ⋅ + µ ±
−
mm
i
l
50 = −
S X
D D
S13.5 Temperature corrections (
T
l δ )
Throughout the measurement care is taken to ensure that the ring to be calibrated, the
setting ring and the comparator scale maintain the monitored temperature. From
previous measurements and general experience with the measurement system it can
be ascertained that the deviations of temperatures of the ring to be calibrated, the
setting ring and the comparator scale from ambient temperature stay within ±0,2 K.
The ambient temperature of the measuring room, however, is estimated to be within
±0,5 K. The knowledge on the measurement, therefore, is best described by the
deviation of the ambient temperature from the reference temperature and the
deviations of the temperatures of the ring to be calibrated, the setting ring and the
comparator scale (ruler) from the ambient temperature. The correction
T
l δ due to
temperature influences is determined from the model:
( )
R R X S X X X S S S
A X X X R S S T
t D D t D t D
t D D l
δ α δ α δ α
α α α α δ
⋅ ⋅ − − ⋅ ⋅ − ⋅ ⋅ +
∆ ⋅ − ⋅ − − ⋅ =
) (
) ( ) (
(S13.2)
December 1999 rev00 Page 73 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
where:
D
X
,
D
S
 nominal diameters of the ring to be calibrated and the reference
setting ring,
α
X
,
S
α ,
R
α  linear thermal expansion coefficients of the ring to be calibrated,
the reference setting ring and the comparator line scale (ruler),
0
t t t − = ∆
A A
 deviations of the ambient temperature of the measuring room from
the reference temperature t
0
20 = °C,
X
t δ ,
S
t δ ,
R
t δ  deviations of the temperature of the ring to be calibrated, the
reference setting ring and the comparator line scale (ruler) from
ambient temperature
Since the expectations of the four temperature differences entering eq. (S13.2) are
zero, the usual linearized version will not include effects of the measurement
uncertainty associated with the values of the three linear thermal expansion
coefficients. As depicted in section S4.13 the nonlinear version has to be used to
determine the standard uncertainty associated with the four product terms:
(S13.3)
( ) δ α α α α
δ α δ
δ α δ
δ α δ
l D D
l D t
l D t
l D D t
TA S S R X X R A
TS S S S
TX X X X
TR S X R R
= ⋅ − − ⋅ − ⋅
= ⋅ ⋅
= ⋅ ⋅
= − ⋅ ⋅
( ) ( )
( )
∆t
Based on the calibration certificate of the setting ring, on the manufacturer’s data for
the ring to be calibrated and the comparator scale, the linear thermal expansion
coefficients are assumed to be within the interval (11,5 ± 1,0) 10
6
°K
1
. Using this
value and the limits of temperature variation stated at the beginning, the standard
uncertainties associated with the four product terms are m 012 , 0 ) ( µ =
TA
l u δ ,
m 053 , 0 ) ( µ =
TS
l u δ , m 12 , 0 ) ( µ =
TX
l u δ and m 066 , 0 ) ( µ =
TR
l u δ . The standard
uncertainty associated with the combined temperature corrections is derived from
these values with the use of the following uncertainty subbudget:
quantity
X
i
estimate
x
i
standard
uncertainty
u x
i
( )
probability
distribution
sensitivity
coefficient
c
i
uncertainty
contribution
u y
i
( )
δ l
TA
0,0 µm 0,012 µm  1,0 0,012 µm
δ l
TS
0,0 µm 0,053 µm  1,0 0,053 µm
δ l
TX
0,0 µm 0,12 µm  1,0 0,12 µm
δ l
TR
0,0 µm 0,066 µm  1,0 0,066 µm
δ l
T
0,0 µm 0,15 µm
December 1999 rev00 Page 74 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
S13.6 Coaxiality correction (δ l
P
)
The deviation from coaxiality of the two spherical probes and the measuring line is
assumed to be within ±20 µm. Using the equations stated in the mathematical note
(S13.13) the correction due to possible noncoaxiality and the associated standard
uncertainty of measurement is given by
) (
1 1
2
2
c u
D D
l δ δ ⋅


.

\

− ⋅ =
S X
P
(S13.4)
) (
1 1
5
16
) (
4
2 2
2
c u
D D
l u δ δ ⋅


.

\

+ ⋅ =
S X
P
(S13.5)
Here c δ is the small distance of the measured cord from the centre of the ring. The
values resulting for the correction and the associated standard measurement
uncertainty are m 004 , 0 µ − ≅
P
l δ and m 0065 , 0 ) ( µ ≅
P
l u δ . As can been seen from
the uncertainty budget (S13.10), these values are two orders of magnitude smaller
than the remaining uncertainty contributions so that their influence need not be taken
into account under the current measurement conditions.
S13.7 Elastic deformation correction (
E
l δ )
The elastic deformation of the ring to be calibrated or the reference setting ring are
not determined during the current measurement. From previous experience, however,
the effects resulting from elastic deformations are estimated to be within ± 0,03 µm.
S13.8 Abbe error correction (
A
l δ )
The actual values of the Abbe errors of the comparator are not determined during the
current measurement. From experience and periodical verification data of the
comparator, however, the effects due to Abbe errors are estimated to be within
± 0,02 µm.
S13.9 Measurements ( ∆l )
The following observations are made of the inner diameter of the unknown and the
setting ring:
No Object Observation Measurand
1 reference
setting ring
0
during this step the
comparator display is
zeroed
diameter in the nominal direction of the
symmetry plane orthogonal to the
cylinder axis
2 ring to be
calibrated
49,99935 mm diameter in the nominal direction of the
symmetry plane orthogonal to the
cylinder axis
3 ring to be
calibrated
49,99911 mm diameter in the symmetry plane
orthogonal to the cylinder axis rotated
around the axis with respect to the
nominal direction by +1 mm on the
circumference
December 1999 rev00 Page 75 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
No Object Observation Measurand
4 ring to be
calibrated
49,99972 mm diameter in the symmetry plane
orthogonal to the cylinder axis rotated
around the axis with respect to the
nominal direction by −1 mm on the
circumference.
5 ring to be
calibrated
49,99954 mm diameter in the nominal direction
translated to the plane parallel to
symmetry plane orthogonal to the
cylinder axis by 1 mm upwards
6 ring to be
calibrated
49,99996 mm diameter in the nominal direction
translated to the plane parallel to
symmetry plane orthogonal to the
cylinder axis by 1 mm downwards
The observations may be divided into two groups: the observation of the diameter of
the setting ring (observation no 1) that is used to set the comparator display to zero
and the observation of the diameter of the ring to be calibrated (observations no 2 to
no 6) that give the difference in diameters:
arithmetic mean: mm 54 999 , 49 = ∆l
standard deviation of a single observation: m 33 , 0 ) ( µ = ∆l s
standard deviation of the mean: m 15 , 0
5
) (
) ( µ =
∆
= ∆
l s
l s
The standard deviation of a single observation m 18 , 0 ) ( µ = ∆l s takes into account
effects due to form deviations of the ring to be calibrated as well as due to the
repeatability of the comparator. To obtain the standard uncertainty of measurement
to be associated with the observed mean difference of the diameters, the uncertainty
resulting from the zeroing of the comparator display must also be taken into account.
This is deduced from the pooled estimate of the standard deviation m 25 , 0 ) 0 ( µ =
p
s
obtained in a prior measurement under the same conditions of measurement. The
resulting standard measurement uncertainty to be associated with the observed
diameter difference is:
m 30 , 0 ) 0 ( ) ( ) (
2 2
µ = + ∆ = ∆
p
s l s l u
December 1999 rev00 Page 76 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
S13.10 Uncertainty budget ( d )
X
quantity
X
i
estimate
x
i
standard
uncertainty
u x
i
( )
probability
distribution
sensitivity
coefficient
c
i
uncertainty
contribution
u y
i
( )
d
S
40,000 7 mm 0,10 µm normal 1,0 0,10 µm
l ∆ 49,999 55 mm 0,30 µm normal 1,0 0,30 µm
i
l δ 0,0 mm 0,22 µm rectangular 1,0 0,22 µm
T
l δ 0,0 mm 0,15 µm normal 1,0 0,15 µm
P
l δ 0,000 004 mm 0,0065 µm rectangular 1,0 0,0065 µm
E
l δ 0,0 mm 0,018 µm rectangular 1,0 0,018 µm
A
l δ 0,0 mm 0,012 µm rectangular 1,0 0,012 µm
X
d 90,000 25 mm 0,433 µm
S13.11 Expanded uncertainty
m 9 , 0 m 433 , 0 2 ) ( µ ≅ µ ⋅ = ⋅ =
X
d u k U
S13.12 Reported result
The diameter of the ring gauge is (90,000 3 ± 0,000 9) mm.
The reported expanded uncertainty of measurement is stated as the standard
uncertainty of measurement multiplied by the coverage factor k = 2 which for a
normal distribution corresponds to a coverage probability of approximately 95%.
S13.13 Mathematical note on noncoaxiality
Since it is not possible to make an exact adjustment of the rings with respect to the
measuring axis of the comparator, the quantity determined in the measurement is a
chord of the respective ring in the proximity of its diameter. The length of this
chord, which is observed in the measurement, is related to the diameter of the ring
by
' d
d

.

\

− ⋅ ≅ ⋅ =
2
) (
2
1
1 ) cos( ' ϕ δ ϕ δ d d d (S13.6)
where ϕ δ is the small angle that complements half of the central angle of the chord
to 2 / π . This angle is related on the other hand to the small distance c δ of the chord
form the centre of the ring by
ϕ δ ϕ δ δ ⋅ ⋅ ≅ ⋅ ⋅ = d d c
2
1
) sin(
2
1
(S13.7)
December 1999 rev00 Page 77 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
so that eq. (S13.6) may be rewritten as
D
c
d d
2
) (
2 '
δ
− ≅ (S13.8)
where the diameter of the ring in the ratio has been replaced by its nominal
diameter since the nominator of the ratio is a small quantity already. The best
estimate of the diameter is obtained by taking the expectation of the last relation to
be
d
D
D
c u
d d
) (
2 '
2
δ
+ = . (S13.9)
Here it has been taken into account that the small distance c δ has zero expectation.
It must also be kept in mind that the meaning of , and d ' d c δ in eq. (S13.8) and eq.
(S13.9) is not identical; whereas in eq. (S13.8) these symbols represent the not
exactly known quantities or random variables, in eq. (S13.9) they stand for the
expectations of these quantities. Since the variance of a random variable equals the
expectation of the square of its deviation from the respective expectation, the square
of the standard measurement uncertainty to be associated with the diameter of the
ring is, according to eq. (S13.8),
2
4
2 2
) (
) 1 ( 4 ) ' ( ) (
D
c u
d u d u
δ
α − ⋅ + = (S13.10)
with
) (
) (
2
2
4
c m
c m
δ
δ
α = (S13.11)
being the ratio of the 4
th
order centred moment to the square of its 2
nd
order centred
moment of the small distance c δ . This ratio depends on the distribution that is
assumed for c δ . It takes the value 5 / 9 = α if c δ is assumed to be rectangularly
distributed so that in this case the standard measurement uncertainty to be associated
with the diameter is expressed by
2
4
2 2
) (
5
16
) ' ( ) (
D
c u
d u d u
δ
⋅ + = (S13.12)
December 1999 rev00 Page 78 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
Authorship This document has been drafted by EAL Task Force for revision of WECC Doc. 191990 on behalf of the EAL Committee 2 (Calibration and Testing Activities). It comprises a thorough revision of WECC Doc. 191990 which it replaces. Official language The text may be translated into other languages as required. The English language version remains the definitive version. Copyright The copyright of this text is held by EA. The text may not be copied for resale. Further information For further information about this publication, contact your National member of EA. You will find the list of members at the website: www.europeanaccreditation.org
December 1999 rev00
Page 2 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
CONTENTS 1 2 3 INTRODUCTION 4 OUTLINE AND DEFINITIONS 5 EVALUATION OF UNCERTAINTY OF MEASUREMENT OF INPUT ESTIMATES 6 4 CALCULATION OF THE STANDARD UNCERTAINTY OF THE OUTPUT ESTIMATE 9 5 EXPANDED UNCERTAINTY OF MEASUREMENT 12 6 STATEMENT OF UNCERTAINTY OF MEASUREMENT IN CALIBRATION CERTIFICATES 13 7 STEPBYSTEP PROCEDURE FOR CALCULATING THE UNCERTAINTY OF MEASUREMENT 13 8 REFERENCES 14 APPENDIX A 15 APPENDIX B 18 APPENDIX C 20 APPENDIX D 21 APPENDIX E 24 SUPPLEMENT 1 27 SUPPLEMENT 2 51
December 1999 rev00
Page 3 of 79
the measurand) is defined as the smallest uncertainty of measurement that a laboratory can achieve within its scope of accreditation. But whereas [ref. 1] establishes general rules for evaluating and expressing uncertainty in measurement that can be followed in most fields of physical measurements. December 1999 rev00 Page 4 of 79 . or when performing more or less routine calibrations of nearly ideal measuring instruments designed for the measurement of that quantity.3 Within EAL the best measurement capability (always referring to a particular quantity. The assessment of best measurement capability of accredited calibration laboratories has to be based on the method described in this document but shall normally be supported or confirmed by experimental evidence. It comprises the following subjects: • definitions basic to the document. • a step by step procedure for calculating the uncertainty of measurement. IUPAC. To assist accreditation bodies with the assessment of the best measurement capability some further explanations are given in Annex A. The method outlined may have to be supplemented by more specific advice for different fields. IEC.EA4/02 • Expression of the Uncertainty of Measurement in Calibration 1 1. when performing more or less routine calibrations of nearly ideal measurement standards intended to define. IFCC. viz. • methods for evaluating the uncertainty of measurement of input quantities.1 INTRODUCTION This document sets down the principles of and the requirements on the evaluation of the uncertainty of measurement in calibration and the statement of this uncertainty in calibration certificates. 1. Evaluation of uncertainty of measurement is also addressed in several of the EAL documents which provide guidance on calibration methods. In developing such supplementary guidelines the general principles stated in this document should be followed to ensure harmonisation between the different fields. The treatment is kept on a general level to suit all fields of calibration. conserve or reproduce a unit of that quantity or one or more of its values. • statement of the uncertainty of measurement. • expanded uncertainty of measurement of the output quantity. • relationship between the uncertainty of measurement of the output quantity and the uncertainty of measurement of the input quantities. realize. IUPAP and OIML [ref. some of these documents containing specific worked out examples. first published in 1993 in the name of BIPM.2 1. ISO. Worked out examples showing the application of the method outlined here to specific measurement problems in different fields will be given in supplements. to make the information more readily applicable. The treatment in this document is in accordance with the Guide to the Expression of Uncertainty in Measurement. this document concentrates on the method most suitable for the measurements in calibration laboratories and describes an unambiguous and harmonised way of evaluating and stating the uncertainty of measurement. 1].
In most cases it will be an analytical expression. Appendix B contains a glossary of these terms together with references 2. X2. certified reference materials or reference data obtained from handbooks. such as quantities associated with calibrated measurement standards. f may be determined experimentally. December 1999 rev00 Page 5 of 79 . N) according to the functional relationship Y = f(X1. (b) quantities whose estimate and associated uncertainty are brought into the measurement from external sources. that characterises the dispersion of the values that could reasonably be attributed to the measurand [ref. barometric pressure or humidity. from a single observation.2 2. Further. or it may be a combination of all of these.1) 2. In this document all quantities which are not exactly known are treated as random variables. including the influence quantities which may affect the measured value.4 The set of input quantities Xi may be grouped into two categories according to the way in which the value of the quantity and its associated uncertainty have been determined: (a) quantities whose estimate and associated uncertainty are directly determined in the current measurement. They may involve the determination of corrections to instrument readings as well as corrections for influence quantities. In calibration one usually deals with only one measurand or output quantity Y that depends upon a number of input quantities Xi (i = 1. but it may also be a group of such expressions which include corrections and correction factors for systematic effects. or exist only as a computer algorithm that must be evaluated numerically. It describes how values of the output quantity Y are obtained from values of the input quantities Xi. such as ambient temperature. The measurands are the particular quantities subject to measurement. for example. 2 .1 The statement of the result of a measurement is complete only if it contains both the value attributed to the measurand and the uncertainty of measurement associated with that value. or judgement based on experience.…. XN ) (2. For typical sources of uncertainty in a measurement see the list given in Annex C. The uncertainty of measurement is a parameter. …. These values may be obtained. In this document the shorthand term uncertainty is used for uncertainty of measurement if there is no risk of misunderstanding. 2]. thereby leading to a more complicated relationship that is not written down as one function explicitly. repeated observations.EA4/02 • Expression of the Uncertainty of Measurement in Calibration 2 OUTLINE AND DEFINITIONS Note: Terms of special relevance to the context of the main text are written in bold when they appear for the first time in this document. associated with the result of a measurement. 2.3 The model function f represents the procedure of the measurement and the method of evaluation.
denoted by u(y).1) using input estimates xi for the values of the input quantities Xi y = f ( x1 .. the necessary corrections have been introduced as separate input quantities. is obtained from equation (2.1 December 1999 rev00 Page 6 of 79 . 1. This concept cannot be used if the estimate equals zero. 3 3. x N ) (2. is used as a measure of the dispersion of values. If not. In some cases the relative standard uncertainty of measurement may be appropriate which is the standard uncertainty of measurement associated with an estimate divided by the modulus of that estimate and is therefore dimensionless..1. The standard uncertainty associated with an estimate has the same dimension as the estimate.EA4/02 • Expression of the Uncertainty of Measurement in Calibration 2. see ref. when all possible values of a quantity lie on one side of a single limit value. EVALUATION OF UNCERTAINTY OF MEASUREMENT OF INPUT ESTIMATES General considerations The uncertainty of measurement associated with the input estimates is evaluated according to either a 'Type A' or a 'Type B' method of evaluation.2. the output estimate denoted by y. For the treatment of such special cases..2) It is understood that the input values are best estimates that have been corrected for all effects significant for the model.2 3. If there is sufficient resolution in the measurement process there will be an observable scatter or spread in the values obtained. The Type B evaluation of standard uncertainty is the method of evaluating the uncertainty by means other than the statistical analysis of a series of observations. It is to be determined from the estimates xi of the input quantities Xi and their associated standard uncertainties u(xi).6 For a random variable the variance of its distribution or the positive square root of the variance. The Type A evaluation of standard uncertainty is the method of evaluating the uncertainty by the statistical analysis of a series of observations.1 3. seldom met in calibration. In this case the standard uncertainty is the experimental standard deviation of the mean that follows from an averaging procedure or an appropriate regression analysis. In this case the evaluation of the standard uncertainty is based on some other scientific knowledge.5 An estimate of the measurand Y. The standard uncertainty of measurement associated with the output estimate or measurement result y. is the standard deviation of the measurand Y. called standard deviation.1 Type A evaluation of standard uncertainty The Type A evaluation of standard uncertainty can be applied when several independent observations have been made for one of the input quantities under the same conditions of measurement. A well known case is the socalled cosine error. 2. x 2 . Note: There are occasions. 3.
5) The standard uncertainty is deduced from this value by equation (3. as expressed by equation (3. the estimate of the quantity Q is q .2) Its (positive) square root is termed experimental standard deviation. If the number of observations cannot be increased.2. the reliability of a Type A evaluation of standard uncertainty. The best estimate of the variance of the arithmetic mean q is the experimental variance of the mean given by s2 (q) s (q ) = n 2 (3. With n statistically independent observations (n > 1). 2. the variance of the mean may be estimated by s (q ) = 2 2 sp n (3.2 Assume that the repeatedly measured input quantity Xi is the quantity Q. If in such a case the value of the input quantity Q is determined as the arithmetic mean q of a small number n of independent observations. n) q= 1 n ∑q n j =1 j (3. December 1999 rev00 Page 7 of 79 .4). (b) For a measurement that is wellcharacterised and under statistical control a 2 combined or pooled estimate of variance sp may be available that characterises the dispersion better than the estimated standard deviation obtained from a limited number of observations.EA4/02 • Expression of the Uncertainty of Measurement in Calibration 3. other means of evaluating the standard uncertainty given in the text have to be considered. ….3) Its (positive) square root is termed experimental standard deviation of the mean. The standard uncertainty u(q ) associated with the input estimate q is the experimental standard deviation of the mean u ( q ) = s( q ) (3. the arithmetic mean or the average of the individual observed values qj (j = 1. when the number n of repeated measurements is low (n < 10). has to be considered.4).4) Warning: Generally.1) The uncertainty of measurement associated with the estimate q is evaluated according to one of the following methods: (a) An estimate of the variance of the underlying probability distribution is the experimental variance s²(q) of values qj that is given by 1 n s (q) = ∑ ( q j − q )2 n − 1 j =1 2 (3.
then the appropriate expectation or expected value and the square root of the variance of this distribution have to be taken as the estimate xi and the associated standard uncertainty u(xi). respectively. If data of this kind are not available. A wellbased Type B evaluation of standard uncertainty can be as reliable as a Type A evaluation of standard uncertainty. or a correction value. The standard uncertainty u(xi) is evaluated by scientific judgement based on all available information on the possible variability of Xi.6) for the estimated value and u 2 ( xi ) = 1 (a + − a − ) 2 12 (3. a resultant value of a previous measurement. Values belonging to this category may be derived from • previous measurement data.1 The Type B evaluation of standard uncertainty is the evaluation of the uncertainty associated with an estimate xi of an input quantity Xi by means other than the statistical analysis of a series of observations.EA4/02 • Expression of the Uncertainty of Measurement in Calibration 3. The following cases must be discerned: (a) When only a single value is known for the quantity Xi.3 Type B evaluation of standard uncertainty 3. the uncertainty has to be evaluated on the basis of experience. According to case (b) above this leads to xi = 1 (a + a − ) 2 + 3.3.7) December 1999 rev00 Page 8 of 79 . a single measured value. Otherwise it has to be calculated from unequivocal uncertainty data. • uncertainties assigned to reference data taken from handbooks. The proper use of the available information for a Type B evaluation of standard uncertainty of measurement calls for insight based on experience and general knowledge. • data provided in calibration and other certificates. manufacturer’s specifications of a measuring instrument.3. The standard uncertainty u(xi) associated with xi is to be adopted where it is given. this value will be used for xi. based on theory or experience.g. a temperature range. • manufacturer’s specifications. a rounding or truncation error resulting from automated data reduction). (b) When a probability distribution can be assumed for the quantity Xi. e. a reference value from the literature.2 (3. especially in a measurement situation where a Type A evaluation is based only on a comparatively small number of statistically independent observations. • experience with or general knowledge of the behaviour and properties of relevant materials and instruments.g. a probability distribution with constant probability density between these limits (rectangular probability distribution) has to be assumed for the possible variability of the input quantity Xi. It is a skill that can be learned with practice. (c) If only upper and lower limits a+ and a– can be estimated for the value of the quantity Xi (e.
3) or by using numerical methods.8) 3 The rectangular distribution is a reasonable description in probability terms of one’s inadequate knowledge about the input quantity Xi in the absence of any other information than its limits of variability.2) where ci is the sensitivity coefficient associated with the input estimate xi.3) X 1 = x1 . if values close to the limits are more likely than values near the centre. The quantity ui(y) (i = 1. a triangular or normal distribution may be a better model. X N = x N The sensitivity coefficient ci describes the extent to which the output estimate y is influenced by variations of the input estimate xi. December 1999 rev00 Page 9 of 79 . N) is the contribution to the standard uncertainty associated with the output estimate y resulting from the standard uncertainty associated with the input estimate xi ui(y) = ciu(xi) (4. xi ± u(xi). where the model function is strongly nonlinear or some of the sensitivity coefficients [see equation (4.e. 4 4. Sometimes it may be more appropriate to find the change in the output estimate y from an experiment by repeating the measurement at e. equation (3. …. But if it is known that values of the quantity in question near the centre of the variability interval are more likely than values close to the limits. seldom occurring in calibration. a Ushaped distribution may be more appropriate. ci = 4.2 ∂f ∂f = ∂ xi ∂ X i (4.1) Note: There are cases.e. the partial derivative of the model function f with respect to Xi. If the difference between the limiting values is denoted by 2a.EA4/02 • Expression of the Uncertainty of Measurement in Calibration for the square of the standard uncertainty. For a treatment of such special cases see ref..2) and (4. i.7) yields 1 u 2 ( xi ) = a 2 (3. It can be evaluated from the model function f by equation (4. On the other hand.1). by calculating the change in the output estimate y due to a change in the input estimate xi of +u(xi) and u(xi) and taking as the value of ci the resulting difference in y divided by 2u(xi).1 CALCULATION OF THE STANDARD UNCERTAINTY OF THE OUTPUT ESTIMATE For uncorrelated input quantities the square of the standard uncertainty associated with the output estimate y is given by u ( y ) = ∑ ui2 ( y ) 2 i =1 N (4. i. 2.g. evaluated at the input estimates xi. 1.3)] vanish and higher order terms have to be included into equation (4.
e.K. X N ) = ∑ p i X i i =1 N 4. December 1999 rev00 Page 10 of 79 .K.5) whereas the sensitivity coefficients equal pi and equation (4. the contribution ui(y) according to equation (4.6 If two input quantities Xi and Xk are correlated to some degree.2) is either positive or negative.8) The sensitivity coefficients equal piy/xi in this case and an expression analogous to equation (4. X N ) = c∏ X ipi i=1 N (4.3 Whereas u(xi) is always positive.6) is obtained from equation (4. w ( y ) = ∑ pi2 w 2 ( xi ) 2 i =1 N (4.4 (4.7) the output estimate again is the corresponding product or quotient of the input estimates y = c∏ xipi i=1 N (4. if relative standard uncertainties w(y) = u(y)/y and w(xi) = u(xi)/xi are used. In general. See Annex D for how this has to be done.1) converts to u ( y ) = ∑ pi2 u 2 ( xi ) 2 i =1 N (4. i.9) 4. X 2 .6) 4.EA4/02 • Expression of the Uncertainty of Measurement in Calibration 4. The sign of ui(y) has to be taken into account in the case of correlated input quantities.2) is given by the corresponding sum or difference of the input estimates y = ∑ pi x i i =1 N (4. depending on the sign of the sensitivity coefficient ci. X 2 .4) the output estimate according to equation (2.1). it should be kept in mind that neglecting correlations between input quantities can lead to an incorrect evaluation of the standard uncertainty of the measurand.5 If the model function f is a product or quotient of the input quantities Xi f ( X1 . The ability to take into account the effect of correlations depends on the knowledge of the measurement process and on the judgement of mutual dependency of the input quantities. see equation (D4) of Annex D. If the model function f is a sum or difference of the input quantities Xi f ( X 1 . their covariance also has to be considered as a contribution to the uncertainty. if they are mutually dependent in one way or another.
1 applicable for the case of uncorrelated input quantities.1: Schematic of an ordered arrangement of the quantities. For repeated measurements the number n of observations also has to be stated. because they have been repeatedly but not simultaneously observed in different independent experiments or because they represent resultant quantities of different evaluations that have been made independently. Sometimes correlations can be eliminated by a proper choice of the model function. it is recommended to present the data relevant to this analysis in the form of a table. For the sake of clarity. standard uncertainties. or if (b) either of the input quantities Xi and Xk can be treated as constant. Table 4. In this table all quantities should be referenced by a physical symbol Xi or a short identifier.9 Quantity Xi X1 X2 : XN Y Estimate xi x1 x2 : xN y Standard uncertainty u(xi) u(x1) u(x2) : u(xN) Sensitivity coefficient ci c1 c2 : cN Contribution to the standard uncertainty ui(y) u1(y) u2(y) : uN(y) u(y) December 1999 rev00 Page 11 of 79 . the associated standard uncertainty of measurement u(xi). or if (c) investigation gives no information indicating the presence of correlation between the input quantities Xi and Xk.EA4/02 • Expression of the Uncertainty of Measurement in Calibration 4. For each of them at least the estimate xi. 4. the sensitivity coefficient ci and the different uncertainty contributions ui(y) should be specified. sensitivity coefficients and uncertainty contributions used in the uncertainty analysis of a measurement. for example. estimates.7 The covariance associated with the estimates of two input quantities Xi and Xk may be taken to be zero or treated as insignificant if (a) the input quantities Xi and Xk are independent.8 The uncertainty analysis for a measurement — sometimes called the uncertainty budget of the measurement — should include a list of all sources of uncertainty together with the associated standard uncertainties of measurement and the methods of evaluating them. The dimension of each of the quantities should also be stated with the numerical values given in the table. A formal example of such an arrangement is given as Table 4. The standard uncertainty associated with the measurement result u(y) given in the bottom right corner of the table is the root sum square of all the uncertainty contributions in the outer right column. The grey part of the table is not filled in. 4.
in order to ensure that a value of the expanded uncertainty is quoted corresponding to the same coverage probability as in the normal case.1) In cases where a normal (Gaussian) distribution can be attributed to the measurand and the standard uncertainty associated with the output estimate has sufficient reliability.6 December 1999 rev00 Page 12 of 79 .e. The reliability of the standard uncertainty assigned to the output estimate is determined by its effective degrees of freedom (see Annex E). obtained by multiplying the standard uncertainty u(y) of the output estimate y by a coverage factor k. The assigned expanded uncertainty corresponds to a coverage probability of approximately 95%.g.1 EXPANDED UNCERTAINTY OF MEASUREMENT Within EAL it has been decided that calibration laboratories accredited by members of the EAL shall state an expanded uncertainty of measurement U. in this case.4 5. in the cases where several (i.e. 5. information on the actual probability distribution of the output estimate must be used to obtain a value of the coverage factor k that corresponds to a coverage probability of approximately 95%. it is not expedient to increase the number n of repeated measurements or to use a Type B evaluation instead of the Type A evaluation of poor reliability. For the remaining cases. If one of these conditions (normality or sufficient reliability) is not fulfilled. e. the reliability criterion is always met if none of the uncertainty contributions is obtained from a Type A evaluation based on less than ten repeated observations. U = ku(y) (5.5 5. all cases where the assumption of a normal distribution cannot be justified. N ≥ 3) uncertainty components. However. If. i. the standard coverage factor k = 2 shall be used.g. e. the standard coverage factor k = 2 can yield an expanded uncertainty corresponding to a coverage probability of less than 95%.2 The assumption of a normal distribution cannot always be easily confirmed experimentally. The use of approximately the same coverage probability is essential whenever two results of measurement of the same quantity have to be compared. the method given in Annex E should be used. normal distributions or rectangular distributions. when evaluating the results of an interlaboratory comparison or assessing compliance with a specification. Even if a normal distribution can be assumed.EA4/02 • Expression of the Uncertainty of Measurement in Calibration 5 5. However.3 5. derived from wellbehaved probability distributions of independent quantities. it may still occur that the standard uncertainty associated with the output estimate is of insufficient reliability. the conditions of the Central Limit Theorem are met and it can be assumed to a high degree of approximation that the distribution of the output quantity is normal. In these cases. These conditions are fulfilled in the majority of cases encountered in calibration work. 5. contribute to the standard uncertainty associated with the output estimate by comparable amounts. other procedures have to be followed.
if the rounding brings the numerical value of the uncertainty of measurement down by more than 5%.1 In calibration certificates the complete result of the measurement consisting of the estimate y of the measurand and the associated expanded uncertainty U shall be given in the form (y ± U). For the process of rounding. in cases where the procedure of Annex E has been followed.3 The numerical value of the uncertainty of measurement should be given to at most two significant figures.EA4/02 • Expression of the Uncertainty of Measurement in Calibration 6 STATEMENT OF UNCERTAINTY OF MEASUREMENT IN CALIBRATION CERTIFICATES 6. However. Y = X1+X2. The standard uncertainty of measurement has been determined in accordance with EAL Publication EALR2. To this an explanatory note must be added which in the general case should have the following content: The reported expanded uncertainty of measurement is stated as the standard uncertainty of measurement multiplied by the coverage factor k = 2. (b) Identify and apply all significant corrections. 6. which for a normal distribution corresponds to a coverage probability of approximately 95%.1). The standard uncertainty of measurement has been determined in accordance with EAL Publication EALR2. In the case of a direct comparison of two standards the equation may be very simple. December 1999 rev00 Page 13 of 79 .g. the usual rules for rounding of numbers have to be used (for further details on rounding see ISO 310:1992. e. the rounded up value should be used.2 However. 6. The numerical value of the measurement result should in the final statement normally be rounded to the least significant figure in the value of the expanded uncertainty assigned to the measurement result.1 STEPBYSTEP PROCEDURE FOR CALCULATING THE UNCERTAINTY OF MEASUREMENT The following is a guide to the use of this document in practice (cf. the additional note should read as follows: The reported expanded uncertainty of measurement is stated as the standard uncertainty of measurement multiplied by the coverage factor k = XX. which for a tdistribution with νeff = YY effective degrees of freedom corresponds to a coverage probability of approximately 95%. 7 7. (c) List all sources of uncertainty in the form of an uncertainty analysis in accordance with Section 4. Annex B). worked examples in Annex F and in separate supplementary documents): (a) Express in mathematical terms the dependence of the measurand (output quantity) Y on the input quantities Xi according to equation (2.
the associated expanded uncertainty U and the coverage factor k in the calibration certificate in accordance with Section 6. [3] International Standard ISO 35341. Switzerland).3) and sum their squares as described in equation (4. International Organization for Standardization (Geneva.EA4/02 • Expression of the Uncertainty of Measurement in Calibration (d) Calculate the standard uncertainty u(q ) for repeatedly measured quantities in accordance with subsection 3. If no data are available from which the standard uncertainty can be derived. 1993. (h) Calculate the expanded uncertainty U by multiplying the standard uncertainty u(y) associated with the output estimate by a coverage factor k chosen in accordance with Section 5.2) and (4. If only upper and lower limits are given or can be estimated. 1993. Statistics . calculate the standard uncertainty u(xi) in accordance with paragraph 3. 1993. corrected and reprinted 1995. (e) For single values. [2] International Vocabulary of Basic and General Terms in Metrology.2 (a). calculate the expectation and the standard uncertainty u(xi) according to paragraph 3.3.Vocabulary and symbols . If input quantities are known to be correlated.2 (c). International Organization for Standardization (Geneva. International Organization for Standardization (Geneva.3.1) to obtain the square of the standard uncertainty u(y) of the measurand. resultant values of previous measurements. first edition. correction values or values from the literature. state a value of u(xi) on the basis of scientific experience. first edition. (i) Report the result of the measurement comprising the estimate y of the measurand.g. e. 8 REFERENCES [1] Guide to the Expression of Uncertainty in Measurement. (g) Calculate for each input quantity Xi the contribution ui(y) to the uncertainty associated with the output estimate resulting from the input estimate xi according to equations (4.3. adopt the standard uncertainty where it is given or can be calculated according to paragraph 3.Part I: Probability and General Statistical Terms. Pay attention to the uncertainty representation used. Switzerland). apply the procedure given in Annex D.2. second edition. December 1999 rev00 Page 14 of 79 .2 (b). Switzerland). (f) For input quantities for which the probability distribution is known or can be assumed.
based on its definition as reported in the main text. this contribution shall be included in the determination of the best measurement capability and a statement should be made that the best measurement capability refers to calibration of that type of device. Inherent in the concept of being nearly ideal is thus that there should be no significant contribution to the uncertainty of measurement attributable to physical effects that can be ascribed to imperfections of the device to be calibrated. The definition of best measurement capability implies that within its accreditation a laboratory is not entitled to claim a smaller uncertainty of measurement than the best measurement capability. the others being physical quantity. Best measurement capability is one of the essential pieces of information to be found in directories of accredited laboratories that are regularly issued by accreditation bodies and is used by potential customers to accredited laboratories to judge the suitability of a laboratory to carry out particular calibration work at the laboratory or on site. Inclusion of the qualifier 'nearly ideal' in the definition means that best measurement capability should not be dependent on the characteristics of the device to be calibrated. the statement of best measurement capability needs to be harmonised. unless it is the outspoken policy of the laboratory to perform such scientific investigations (in which case these become the 'more or less routine' type calibrations of the laboratory). calibration method or type of instrument to be calibrated and measurement range.EA4/02 • Expression of the Uncertainty of Measurement in Calibration APPENDIX A Comments on the assessment of best measurement capability A1 Best measurement capability (see Section 1 of the main text) is one of the parameters that is used to define the scope of an accredited calibration laboratory. To make it possible to compare the capabilities of different calibration laboratories. it should be understood that such a device should be available. in a particular case. If it is established that. However. To facilitate this. in particular laboratories accredited by different accreditation bodies. With 'more or less routine calibrations' is meant that the laboratory shall be able to achieve the stated capability in the normal work that it performs under its accreditation. Best measurement capability is normally stated in the accreditation schedule or in other documentation that supports either the decision on accreditation or the accreditation certificate which in many cases is issued as evidence of accreditation. This means that the laboratory shall be required to state a larger uncertainty than that corresponding to the best measurement capability whenever it is established Page 15 of 79 A2 A3 A4 A5 December 1999 rev00 . even the most 'ideal' available device contributes to the uncertainty of measurement. Obviously there are instances where the laboratory would be able to do better as a result of extensive investigations and additional precautions but these cases are not covered by the definition of best measurement capability. some explanations are given below to the term best measurement capability. Occasionally it is stated both in the accreditation schedule and in the supporting documents.
) All components contributing significantly to the uncertainty of measurement shall be taken into account when evaluating the best measurement capability. A6 It should be pointed out that according to the definition of best measurement capability the concept is applicable only to results for which the laboratory claims its status as accredited laboratory. the best measurement capability has to be stated to a coverage probability of approximately 95%. In some fields the uncertainty of measurement may depend on some additional parameter. the contribution caused by the drift between subsequent calibrations of the standard has to be taken into account when estimating the uncertainty contribution of the working standard.EA4/02 • Expression of the Uncertainty of Measurement in Calibration that the actual calibration process adds significantly to the uncertainty of measurement. e. See Section 5 of the main text. Such additional parameters shall be stated together with the physical quantity in question and the best measurement capability specified for the additional parameters. in the form of an expanded uncertainty of measurement. For instance. The best measurement capability shall be stated to the same level as required for calibration certificates. Thus. with the exception of the case covered in the previous subsection. i. e. When stating the actual uncertainty.) Assessment of best measurement capability is the task of the accreditation body. normally with coverage factor k = 2. The evaluation of the contributions that are known to vary with time or with any other physical quantity can be based on limits of possible variations assumed to occur under normal working conditions.g. (This consideration refers not only to the best measurement capability but to all parameters that define the scope of a calibration laboratory. Obviously the actual uncertainty of measurement can never be smaller than the best measurement capability. frequency of applied voltage when calibrating standard resistors. if the used working standard is known to drift. Often this can be done by giving the best measurement capability as a function of these parameters. The estimation of the uncertainty of measurement that defines the best measurement capability should follow the procedure laid down in the present document. (Only in those exceptional cases where the existence of a normal distribution cannot be assumed or the assessment is based on limited data. the laboratory shall be asked to apply the principles of the present document. A7 A8 A9 December 1999 rev00 Page 16 of 79 . It should be possible for a laboratory to apply for accreditation with a larger uncertainty of measurement than its technical capability if the laboratory has internal reasons for doing so.g.e. Such internal reasons usually involve cases where the real capability has to be held in confidence to external customers. strictly speaking the term is of an administrative character and does not necessarily need to reflect the real technical capability of the laboratory. when doing research and development work or when providing service to special customers. The policy of the accreditation body should be to grant accreditation on any applied level if the laboratory is capable of carrying out calibrations on that level. Typically the equipment under calibration may give a contribution.
) Although the assessment should be based on the procedures of this document. (Usually the inclusion of the relevant unit gives the necessary explanation but in case of dimensionless quantities a separate statement is needed. Where the best measurement capability is a function of the quantity to which it refers (or any other parameter) it should be given in analytical form but in this case it may be illustrative to support the statement by a diagram. The meaning of this requirement is that the accreditation body should not rely on an evaluation of the uncertainty of measurement only.EA4/02 • Expression of the Uncertainty of Measurement in Calibration A10 The best measurement capability should normally be stated numerically. in the main text there is the requirement that the assessment normally shall be 'supported or confirmed by experimental evidence'. It should always be unequivocally clear whether the best measurement capability is given in absolute or relative terms. Interlaboratory comparisons that substantiate the evaluation have to be carried out under the supervision of the accreditation body or on its behalf. A11 December 1999 rev00 Page 17 of 79 .
experimental variance (from [ref.3.1.4) The estimate of an input quantity used in the evaluation of the result of a measurement. when performing more or less routine calibrations of nearly ideal measurement standards intended to define. 1] Section 4.5.1. 1] Section 4. 1] term 2. conserve or reproduce a unit of that quantity or one or more of its values. realise. input estimate (from [ref.6) The measure of the relative mutual dependence of two random variables. correlation ([ref.EA4/02 • Expression of the Uncertainty of Measurement in Calibration APPENDIX B Glossary of some relevant terms B1 B2 arithmetic mean ([ref. 1] term 2.2.26) The sum of values divided by the number of values best measurement capability (Section 1) The smallest uncertainty of measurement that a laboratory can achieve within its scope of accreditation. 1] Section C. 1] term 2.2) in the text.3.3.6) A numerical factor used as a multiplier of the standard uncertainty of measurement in order to obtain an expanded uncertainty of measurement coverage probability (from [ref. of the distribution of values that as a result of a measurement could reasonably be attributed to the measurand experimental standard deviation ([ref. equal to the ratio of their covariance to the positive square root of the product of their variances covariance (from [ref.4) The measure of the mutual dependence of two random variables. 2] term 3.3.2) The quantity characterising the dispersion of the results of a series of n observations of the same measurand given by equation (3. expanded uncertainty ([ref.5) A quantity defining an interval about the result of a measurement that may be expected to encompass a large fraction of the distribution of values that could reasonably be attributed to the measurand. 1] Section 4.3. or when performing more or less routine calibrations of nearly ideal measuring instruments designed for the measurement of that quantity. B3 B4 B5 B6 B7 B8 B9 B10 B11 B12 December 1999 rev00 Page 18 of 79 . equal to the expectation of the product of the deviations of two random variables from their respective expectations coverage factor ([ref. NOTE 1) The fraction. 3] term 2. input quantity (from [ref. usually large. 1] Section C. taken into account in the process of evaluating the result of a measurement.8) The positive square root of the experimental variance.2) A quantity on which the measurand depends. 3] term 1.13) The relationship between two or several random variables within a distribution of two or more random variables correlation coefficient (from [ref.
2) A variable that may take any of the values of a specified set of values and with which is associated a probability distribution.9) A parameter.1.1. associated with the result of a measurement.3) A function giving the probability that a random variable takes any given value or belongs to a given set of values random variable ([ref.3) The differential change in the output estimate generated by a differential change in an input estimate divided by the change in that input estimate. that characterises the dispersion of the values that could reasonably be attributed to the measurand. 1] term 2. sensitivity coefficient associated with an input estimate (from [ref. 1] term 2.4) An estimate of the experimental variance obtained from long series of observations of the same measurand in wellcharacterised measurements under statistical control.1. relative standard uncertainty of measurement (from [ref. 3] term 1. 1] Section 5.EA4/02 • Expression of the Uncertainty of Measurement in Calibration B13 B14 measurand ([ref.22) The expectation of the square of the deviation of a random variable about its expectation. probability distribution ([ref. output estimate (from [ref.3.6) The particular quantity subject to measurement.3. 1] Section 4.6) The standard uncertainty of a quantity divided by the estimate of that quantity.2.1) The uncertainty of measurement expressed as the standard deviation Type A evaluation method ([ref. B15 B16 B17 B18 B19 B20 B21 B22 B23 B24 B25 B26 December 1999 rev00 Page 19 of 79 . 3] term 1.2) The quantity that represents the measurand in the evaluation of a measurement. standard deviation (from [ref. 1] Section 5.3. 2] term 2. standard uncertainty of measurement ([ref. pooled estimate of variance (from [ref. variance (from [ref. output quantity (from [ref.23) The positive square root of the variance of a random variable.3) The method of evaluation of uncertainty of measurement by means other than the statistical analysis of series of observations.1. 1] Section 4.2) The method of evaluation of uncertainty of measurement by the statistical analysis of series of observations Type B evaluation method ([ref. 3] term 1. 1] term 2.4) The result of a measurement calculated from the input estimates by the model function. uncertainty of measurement ([ref. 3] term 1. 2] term 3. 1] Section 4.
C2 These sources are not necessarily independent. Some of the sources (a) to (i) may contribute to (j). December 1999 rev00 Page 20 of 79 . In practice. (g) inexact values of measurement standards and reference materials. (d) inadequately known effects of environmental conditions or imperfect measurements of these. 1]. (e) personal bias in reading analogue instruments. (j) variations in repeated observations of the measurand under apparently identical conditions. there are many possible sources of uncertainty in a measurement [ref. are called sources of uncertainty. (i) approximations and assumptions incorporated in the measurement method and procedure. (h) inexact values of constants and other parameters obtained from external sources and used in the datareduction algorithm. including: (a) incomplete definition of the measurand. Phenomena that contribute to the uncertainty and thus to the fact that the result of a measurement cannot be characterised by a unique value.EA4/02 • Expression of the Uncertainty of Measurement in Calibration APPENDIX C Sources of uncertainty of measurement C1 The uncertainty of the result of a measurement reflects the lack of complete knowledge of the value of the measurand. (b) imperfect realisation of the definition of the measurand. (f) finite instrument resolution or discrimination threshold. (c) nonrepresentative sampling — the sample measured may not represent the defined measurand. Complete knowledge requires an infinite amount of information.
or even measurement method having a significant uncertainty is used in the evaluation of their values. It should be noted that the second summation of terms in equation (D. input quantities are often correlated because the same physical reference standard. In that case the covariance u(x1. Q2 . D3 For influence quantities any degree of correlation has to be based on experience.….1) has to be replaced by u 2 ( y ) = ∑ c i2 u 2 ( x i ) + 2∑ i =1 N N −1 i =1 k = i +1 ∑ c c u( x .. The estimates x1 and x2 of the input quantities will be correlated to some extent.2. x i k i N k ) (D. q ) = n 1 ∑ ( p − p )(q j − q ) n( n − 1) j =1 j (D. Q2 . QL ) (D. When there is correlation. reference datum..L) are uncorrelated. x k ) (D.3) where ci and ck are the sensitivity coefficients defined by equation (4. Without loss of generality.4) may become negative in sign. xk) (where i ≠ k and r ≤1).1). The degree of correlation is characterised by the correlation coefficient r(xi.x2) associated with the estimates x1 and x2 is given by December 1999 rev00 Page 21 of 79 . D4 In practice.1) has to be considered as an additional contribution to the uncertainty.3) or u 2 ( y ) = ∑ ui2 ( y ) + 2 ∑ i =1 N N −1 i =1 k =i +1 ∑ u ( y)u i N k ( y ) r ( xi .2) and by substitution r can be calculated from equation (D. x k ) = u( xi ) u( x k ) r ( x i .. D2 In the case of n independent pairs of simultaneously repeated observations of two quantities P and Q the covariance associated with the arithmetic means p and q is given by s( p. QL ) X 2 = g 2 ( Q1 .2.….4) with the contributions ui(y) to the standard uncertainty of the output estimate y resulting from the standard uncertainty of the input estimate xi given by equation (4..3) or (D..2).L) X 1 = g1 ( Q1 . x k ) (i ≠ k ) (D. suppose that two input quantities X1 and X2 estimated by x1 and x2 depend on the set of independent variables Ql (l = 1.e.. if they are dependent on each other in one way or another — the covariance associated with the two estimates xi and xk u ( x i . measuring instrument.EA4/02 • Expression of the Uncertainty of Measurement in Calibration APPENDIX D Correlated input quantities D1 If two input quantities Xi and Xk are known to be correlated to some extent — i. equation (4.5) although some of these variables may not necessarily appear in both functions. even if the estimates ql (l = 1.
The correlation coefficient r(x1. Page 22 of 79 December 1999 rev00 . The standard uncertainties are calculated from equation (4.9) The case described by equation (D. Introducing directly the independent variables Ql by replacing the original variables X1 and X2 in the model function f in accordance with the transformation equations (D. the covariance is zero if no variable is common to functions g1 and g2.1). D5 The following example demonstrates correlations which exist between values attributed to two artefact standards that are calibrated against the same reference standard.x2) associated with the estimates x1 and x2 is determined from equation (D. Mathematical Model The estimates x1 and x2 depend on the value qS of the reference standard and the observed differences z1 and z2 according to the relations x1 = qS − z1 x2 = qS − z2 Standard uncertainties and covariances The estimates z1. Measurement Problem The two standards X1 and X2 are compared with the reference standard QS by means of a measuring system capable of determining a difference z in their values with an associated standard uncertainty u(z).3). x2 ) = 2 u ( qS ) + u2 ( z ) Its value ranges from 0 to +1 depending on the ratio of the standard uncertainties u(qS) and u(z).7) (D.6) together with equation (D.5) is an occasion where the inclusion of correlation in the evaluation of the standard uncertainty of the measurand can be avoided by a proper choice of the model function. Because only those terms contribute to the sum for which the sensitivity coefficients do not vanish.6). assuming that u(z1) = u(z2) = u(z). x2 ) = u2 ( qS ) (D.8) The correlation coefficient deduced from these results is u2 ( qS ) r ( x1 .EA4/02 • Expression of the Uncertainty of Measurement in Calibration L u( x1 .4) and the covariance associated with the estimates x1 and x2 is calculated from equation (D. z2 and qS are supposed to be uncorrelated because they have been determined in different measurements. u2 ( x1 ) = u2 ( qS ) + u2 ( z ) u2 ( x2 ) = u2 ( qS ) + u2 ( z ) u( x1 . The value qS of the reference standard is known with standard uncertainty u(qS).5) gives a new model function that does not contain the correlated variables X1 and X2 any longer.6) where c1l and c2 l are the sensitivity coefficients derived from the functions g1 and g2 in analogy to equation (4. x 2 ) = ∑ c1l c2 l u 2 ( ql ) l =1 (D. D6 (D.
g.10) for each group of correlated quantities. December 1999 rev00 Page 23 of 79 .10) is easily generalised to cases of one or several groups with two or more correlated input quantities. Note: Equation (D.10) with ur(y) being the contribution to the standard uncertainty of all the remaining input quantities assumed to be uncorrelated.EA4/02 • Expression of the Uncertainty of Measurement in Calibration D7 There are cases however. In this case a respective worst case sum has to be introduced into equation (D. where correlation between two input quantities X1 and X2 cannot be avoided. using the same measuring instrument or the same reference standard when determining the input estimates x1 and x2 but where transformation equations to new independent variables are not available. e. If furthermore the degree of correlation is not exactly known it may be useful to assess the maximum influence this correlation can have by an upper bound estimate of the standard uncertainty of the measurand which in the case that other correlations have not to be taken into account takes the form u 2 ( y ) ≤ u1 ( y ) + u2 ( y ) ( ) 2 + ur2 ( y ) (D.
Similarly. (b) Estimate the effective degrees of freedom νeff of the standard uncertainty u(y) associated with the output estimate y from the WelchSatterthwaite formula E2 ν eff = ∑ i=1 u4 ( y ) . they are usually chosen in such a way that the probability of the quantity in question lying outside these limits is in fact extremely small. it is common practice to carry out such evaluations in a manner that ensures that any underestimation is avoided. the degrees of freedom of this estimate. The procedure for calculating an appropriate coverage factor k when the conditions of the Central Limit Theorem are met comprises the following three steps: (a) Obtain the standard uncertainty associated with the output estimate according to the step by step procedure given in Section 7. the degrees of freedom of the standard uncertainty u(xi) obtained from a Type B may be taken to be νi → ∞. a suitable measure of the reliability of the standard uncertainty associated with an output estimate is its effective degrees of freedom νeff . and νi is the effective degrees of freedom of the standard uncertainty contribution ui(y). If. For a standard uncertainty u(q) obtained from a Type A evaluation as discussed in subsection 3. Under the assumption that this practice is followed. E1 To estimate the value of a coverage factor k corresponding to a specified coverage probability requires that the reliability of the standard uncertainty u(y) of the output estimate y is taken into account. for example.2). It is more problematic to associate degrees of freedom with a standard uncertainty u(xi) obtained from a Type B evaluation. is a measure of the reliability.1) νi where the ui(y) (i=1. That means taking into account how well u(y) estimates the standard deviation associated with the result of the measurement. lower and upper limits a– and a+ are set. are the contributions to the standard uncertainty associated with the output estimate y resulting from the standard uncertainty associated with the input estimate xi which are assumed to be mutually statistically independent.…. which is approximated by an appropriate combination of the effective degrees of freedom of its different uncertainty contributions ui(y).1. For an estimate of the standard deviation of a normal distribution. which depends on the size of the sample on which it is based.EA4/02 • Expression of the Uncertainty of Measurement in Calibration APPENDIX E Coverage factors derived from effective degrees of freedom. the degrees of freedom are given by νi = n1. However.N). December 1999 rev00 Page 24 of 79 . defined in equation (4. N ui4 ( y ) (E.2.
1: Coverage factors k for different effective degrees of freedom νeff.EA4/02 • Expression of the Uncertainty of Measurement in Calibration (c) Obtain the coverage factor k from the table of values given as Table E. This table is based on a tdistribution evaluated for a coverage probability of 95.37 10 2.53 3 3.43 8 2.13 50 2. which will usually be the case. νeff k 1 13.28 20 2.1 of this annex.52 7 2.05 ∞ 2.45%.87 5 2.00 December 1999 rev00 Page 25 of 79 . If νeff is not an integer. truncate νeff to the next lower integer.31 4 2.65 6 2. Table E.97 2 4.
EA4/02 • Expression of the Uncertainty of Measurement in Calibration SUPPLEMENT 1 Examples December 1999 rev00 Page 26 of 79 .
EA4/02 • Expression of the Uncertainty of Measurement in Calibration CONTENTS S1 S2 S3 S4 S5 S6 S7 INTRODUCTION CALBRATION OF A WEIGHT OF NOMINAL VALUE 10 KG CALIBRATION OF NOMINAL 10 kΩ STANDARD RESISTOR CALIBRATION OF A GAUGE BLOCK OF NOMINAL LENGTH 50 MM CALIBRATION OF A TYPE N THERMOCOUPLE AT 1000°C CALIBRATION OF A POWER SENSOR AT A FREQUENCY OF 19 GHz CALIBRATION OF A COAXIAL STEP ATTENUATOR AT A SETTING OF 30 dB (INCREMENTAL LOSS) 28 29 31 34 37 42 46 December 1999 rev00 Page 27 of 79 .
5 December 1999 rev00 Page 28 of 79 . The contributions and values given in the examples are not intended to imply mandatory or preferred requirements. independent of the field of calibration. S1. It is thus hoped that this set of examples will contribute to a better understanding of the details of setting up the model of evaluation and to the harmonisation of the process of evaluating the uncertainty of measurement. • the reported complete result of measurement.2 S1.1 INTRODUCTION The following examples are chosen to demonstrate the method of evaluating the uncertainty of measurement. These drafts have been simplified and harmonised to make them transparent to laboratory staff in all fields of calibration. The presentation of the examples follows. • the expanded uncertainty of measurement. Nevertheless.4 S1. • a general description of the process of measurement. • an extended listing of input data with short descriptions of how they have been obtained. Laboratories should determine the uncertainty contributions on the basis of the model function they use in the evaluation of the particular calibration they perform and report the evaluated uncertainty of measurement on the calibration certificate they issue.EA4/02 • Expression of the Uncertainty of Measurement in Calibration S1 S1. the conditions stated in section 5 for the use of the standard coverage factor k = 2 are fulfilled.3 S1. the examples presented here provide a general guidance on how to proceed. • the model of evaluation with a list of symbols used. This first supplement to EALR2 is intended to be followed by others containing further workedout examples on the evaluation of uncertainty of measurement in connection with the calibration of instruments. in accordance with the stepbystep procedure of section 7 of EALR2. More typical and representative examples based on appropriate models have to be developed by special working groups in the different areas. In all the examples given. • the list of observations and the evaluation of the statistical parameters. a common scheme containing: • a short descriptive title. The examples are based on drafts prepared by EAL Expert Groups. • an uncertainty budget in table form. Examples may also be found in EAL Guidance Documents dealing with the calibration of specific types of measurement instruments.
4 S2. S2. δmC): A previous evaluation of the repeatability of the mass difference between two weights of the same nominal value gives a pooled estimate of standard deviation of 25 mg. correction for air buoyancy.3 Reference standard (mS): The calibration certificate for the reference standard gives a value of 10 000. drift of value of the standard since its last calibration.6 S2. the limits of deviation are estimated to be ±1×106 of the nominal value. Comparator (δm. S2.1) S2. The unknown conventional mass mX is obtained from: mX = mS + δdD + δm + δmC + δB where: mS δmD δm δmC δB (S2.EA4/02 • Expression of the Uncertainty of Measurement in Calibration S2 S2. Air buoyancy (δB): No correction is made for the effects of air buoyancy.005 g with an associated expanded uncertainty of 45 mg (coverage factor k = 2). Correlation: None of the input quantities are considered to be correlated to any significant extent. observed difference in mass between the unknown mass and the standard.1 CALBRATION OF A WEIGHT OF NOMINAL VALUE 10 KG The calibration of a weight of nominal value 10 kg of OIML class Ml is carried out by comparison to a reference standard (OIML class F2) of the same nominal value using a mass comparator whose performance characteristics have previously been determined.2 conventional mass of the standard.7 December 1999 rev00 Page 29 of 79 .5 S2. Drift of the value of the standard (δmD): The drift of the value of the reference standard is estimated from previous calibrations to be zero within ±15 mg. No correction is applied for the comparator. correction for eccentricity and magnetic effects. whereas variations due to eccentricity and magnetic effects are estimated to have rectangular limits of ±10 mg.
EA4/02 • Expression of the Uncertainty of Measurement in Calibration S2.0 1.77 mg 5.02 g +0.95 mg 14.020 g 0.055 g +0.9 Uncertainty budget (mX): δm = 0.045 g +0.025 g +0.0 uncertainty contribution ui(y) 22.025 g standard uncertainty u(xi) 22.5 mg 8.025 g +0.4 mg 5.005 g 0.0 1.95 mg 14.4 mg quantity estimate Xi mS δmD δm δmC δB mX xi 10 000.020 g sp(δm) = 25 mg u(δ m) = s(δ m) = 25 mg 3 = 14.050 g +0.77 mg 29.10 Expanded uncertainty U = k × u(mX) = 2 × 29.000 g 10 000.5 mg 8.03 g +0.015 g +0.3 mg S2.77 mg 5.77 mg probability distribution normal rectangular normal rectangular rectangular sensitivity coefficient ci 1.040 g +0.020 g +0.8 Measurements: Three observations of the difference in mass between the unknown mass and the standard are obtained using the substitution method and the substitution scheme ABBA ABBA ABBA: no conventional mass reading observed difference 1 standard unknown unknown standard +0.020 g +0.025 g +0.0 1.0 1.000 g 0.01 g 2 standard unknown unknown standard 3 standard unknown unknown standard arithmetic mean: pooled estimate of standard deviation: (obtained from prior evaluation) standard uncertainty: S2.3 mg ≅ 59 mg December 1999 rev00 Page 30 of 79 .010 g +0.020 g +0.000 g 0.4 mg 5.
The reported expanded uncertainty of measurement is stated as the standard uncertainty of measurement multiplied by the coverage factor k = 2. and a calibrated fourterminal standard resistor of the same nominal value as the item to be calibrated as reference standard.00 °C.EA4/02 • Expression of the Uncertainty of Measurement in Calibration S2. .053 Ω ±5 mΩ (coverage factor k = 2) at the specified reference temperature of 23 °C. It is determined that the measuring current on the 10 kΩ range of the DMM of 100 µA is sufficiently low not to cause any appreciable selfheating of the resistors. Temperature corrections (δRTS. drift of the resistance of the reference since its last calibration.1) rC δRTX S3. δRTX): The temperature of the oil bath is monitored using a calibrated thermometer to be 23. The resistors are immersed in a well stirred oil bath operating at a temperature of 23 °C monitored by a centrally placed mercuryinglass thermometer. temperature related resistance variation of the reference.temperaturerelated resistance variation of the unknown resistor.75 mΩ for the deviation S3.1 CALIBRATION OF NOMINAL 10 KΩ STANDARD RESISTOR The resistance of a fourterminal standard resistor is determined by direct substitution using a longscale digital multimeter (7½ digit DMM) on its resistance range. Thus the known value 5×106 K1 of the temperature coefficient (TC) of the reference resistor gives limits ±2.correction factor for parasitic voltages and instrument resolution .5 December 1999 rev00 Page 31 of 79 . Drift of the value of the standard (δRD): The drift of the resistance of the reference resistor since its last calibration is estimated from its calibration history to be +20 mΩ with deviations within ±10 mΩ.3 resistance of the reference.055 K.2 RX = ( RS + δ RD + δ RT S ) rC r − δ RT X where: RS δRD δRTS r = RiX/RiS  (S3. Taking into account the metrological characteristics of the thermometer used and of gradients of temperature within the oil bath.000 025 kg ±59 mg. The resistance RX of the unknown resistor is obtained from the relationship: S3.11 Reported result The measured mass of the nominal 10 kg weight is 10.4 S3. The fourterminal connectors of each resistor are connected in turn to the terminals of the DMM. ratio of the indicated resistance (index i means ‘indicated’) for the unknown and reference resistors. the temperature of the resistors is estimated to coincide with the monitored temperature within ±0. The resistors are allowed to stabilise before the measurement. The measuring procedure used also ensures that the effects of external leakage resistances on the result of measurement can be considered to be insignificant. which for a normal distribution corresponds to a coverage probability of approximately 95 %. Reference standard (RS): The calibration certificate for the reference standard gives a resistance value of 10 000. S3 S3.
0 1.0 10 000 Ω 10 000 Ω uncertainty contribution ui(y) 2.158 × 10−6 u(r ) = s(r ) = = 0. which are estimated to have limits of ±0.000 000 0 and limits ±1. Measurements(r): Five observations are made to record the ratio r: No.000 000 0 1.0707 × 10−6 5 quantity Xi RS estimate xi standard uncertainty u(xi) 2. S3.41×106 0.000 010 6 1.9 Uncertainty budget (RX): r = 1.6 mΩ 3.1 mΩ 0.053 Ω 0.000 Ω 1.0 1.8 mΩ 1.5×106 for each reading.000 010 3 1.EA4/02 • Expression of the Uncertainty of Measurement in Calibration from its resistance value according to calibration.0×106. observed ratio S3.5 mΩ 5.6 Resistance measurements (rC): Since the same DMM is used to observe both RiX and RiS the uncertainty contributions are correlated but the effect is to reduce the uncertainty and it is only necessary to consider the relative difference in the resistance readings due to systematic effects such as parasitic voltages and instrument resolution (see the mathematical note in paragraph S3.000 010 5 arithmetic mean: experimental standard deviation: standard uncertainty: S3.6 mΩ 3.7 S3. Correlation: None of the input quantities are considered to be correlated to any significant extent.178 Ω δRD δRTS δRTX rC r RX December 1999 rev00 Page 32 of 79 .7 mΩ 8.0 1.2 mΩ 4. thus the resistance variation of the unknown resistor due to a temperature variation is estimated to be within ±5.000 Ω 0.2 mΩ 0. the TC of the unknown resistor is estimated not to exceed 10×106 K1. The distribution resulting for the ratio rC is triangular with expectation 1.000 010 5 10 000.5 mΩ 5.000 010 4 1.158 × 106 0.8 mΩ 1.8 1 2 3 4 5 1.000 010 5 s(r) = 0.12).020 Ω 0. due to a possible deviation from the operating temperature.000 010 7 1. From the manufacturer’s literature.07×106 probability distribution normal rectangular rectangular rectangular triangular normal sensitivity coefficient ci 1.33 mΩ 10 000.5 mΩ.
4) and the correction factor (linear approximation in the deviations) rC = 1 + δ RX' −δ RS' R (S3. Within the usual linear approximation in the deviations. S3.6) December 1999 rev00 Page 33 of 79 .5) Because of the fact that the difference of the deviations enters into equation (S3. the values causing the DMM indications RiX and RiS are given by δ RX' ) R δR ' RS' = RiS (1 + S ) R RX' = RiX (1 + (S3. it is given by the expression u2 (rC ) = 2 u2 (δ R' ) R2 (S3. The resistance ratio deduced from these expressions is RX' = rrC RS' RiX RiS (S3.33 mΩ ≅ 17 mΩ S3.5). at a measuring temperature of 23. The reported expanded uncertainty of measurement is stated as the standard uncertainty of measurement multiplied by the coverage factor k = 2.017) Ω. Assuming that u(δ RX' ) = u(δ RS' ) = u(δ R' ) .3) with the ratio of the indicated resistance for the unknown and the reference resistor r= (S3. is (10 000.10 Expanded uncertainty: U = k × u( RX ) = 2 × 8.11 Reported result: The measured value of the nominal 10 kΩ resistor. which for a normal distribution corresponds to a coverage probability of approximately 95 %.EA4/02 • Expression of the Uncertainty of Measurement in Calibration S3. The standard uncertainty of the correction factor is determined only by uncorrelated deviations resulting from the parasitic effects and the resolution of the DMM.2) with R being the nominal value of the resistors and δ RX' and δ RS' the unknown deviations.178 ±0.00 °C and a measuring current of 100 µA.12 Mathematical note on the standard uncertainty of measurement of the ratio of indicated resistance values: The unknown and the reference resistors have nearly the same resistance. correlated contributions of systematic effects resulting from the internal scale of the DMM do not influence the result.
length of the reference gauge block at the reference temperature t0 = 20 °C according to its calibration certificate.temperature difference between the unknown and reference gauge blocks.correction for noncentral contacting of the measuring faces of the unknown gauge block.average of the thermal expansion coefficients of the unknown and reference gauge blocks.2) . . . S4.1 CALIBRATION OF A GAUGE BLOCK OF NOMINAL LENGTH 50 MM The calibration of the grade 0 gauge block (ISO 3650) of 50 mm nominal length is carried out by comparison using a comparator and a calibrated gauge block of the same nominal length and the same material as reference standard.EA4/02 • Expression of the Uncertainty of Measurement in Calibration S4 S4.nominal length of the gauge blocks considered. on account of the uncertainty in the measurement of laboratory temperature. S4. The actual length lX' of the gauge block to be calibrated is related to the actual length lS' of the reference standard by the equation lX' = lS' + δ l (S4. . .observed difference in length between the unknown and the reference gauge block. .change of the length of the reference gauge block since its last calibration due to drift. . . in particular at a temperature which. may not be identical with the reference temperature for length measurements. December 1999 rev00 Page 34 of 79 .difference in the thermal expansion coefficients between the unknown and the reference gauge blocks. .2 The length lX of the unknown gauge block at the reference temperature is obtained from the relationship: lX = lS + δ lD + δ l + δ lC − L( α × δ t + δ α × ∆ t ) − δ lV where: lS δlD δl δlC L α = (α X + α S ) / 2 δt = (tX . .1) with δl being the measured length difference.tS) δα = (αX – αS) ∆t = (t X + tS ) / 2 − t0 δlV (S4.deviation of the average temperature of the unknown and the reference gauge blocks from the reference temperature.3 Reference standard (lS): The length of the reference gauge block together with the associated expanded uncertainty of measurement is given in the calibration certificate of a set of gauge blocks as 50.000 02 mm ±30 nm (coverage factor k = 2).correction for nonlinearity and offset of the comparator. lX' and lS ' are the lengths of the gauge blocks under measurement conditions. The difference in central length is determined in vertical position of the two gauge blocks using two length indicators contacting the upper and lower measuring faces.
Combining the two rectangular distributions the difference in linear thermal expansion coefficient is triangularly distributed within the limits ±2×106 °C1. the correction due to central misalignment of the contacting point is estimated to be within ±6. Therefore second order terms have to be taken into account in the evaluation of their uncertainty contribution resulting in the product of standard uncertainties associated with the factors of the product term δ α × ∆ t in equation (S4. From this. δt.5 mm. it can be ascertained that for length differences D up to ±10 µm corrections to the indicated length difference are within the limits ±(30 nm +0.EA4/02 • Expression of the Uncertainty of Measurement in Calibration S4.05 K. δα. care is taken to ensure that the gauge blocks assume ambient temperature of the measuring room. The best estimates of the difference in linear expansion coefficients and the deviations of the mean temperature from the reference temperature are zero.4 Drift of the standard (δlD): The temporal drift of the length of the reference gauge block is estimated from previous calibrations to be zero with limits ±30 nm.236 × 10−6 . The remaining difference in temperature between the standard and the gauge block to be calibrated is estimated to be within ±0. Variation in length (δlV): For gauge blocks of grade 0 the variation in length determined from measurements at the centre and the four corners has to be within ±0. The final standard uncertainty is u(δα × ∆ t ) = 0.7 nm. General experience with gauge blocks of this type suggests that zero drift is most probable and that a triangular probability distribution can be assumed. ∆ t ): Before calibration. S4.6 S4. Assuming that this variation occurs on the measuring faces along the short edge of length 9 mm and that the central length is measured inside a circle of radius 0.0)×106 °C1.5 °C.12 µm (ISO 3650).2) (see the mathematical note in paragraph S4.8 December 1999 rev00 Page 35 of 79 .5 S4. the maximum length difference will be within ±1 µm leading to limits of ±32 nm for nonlinearity and offset corrections of the comparator used. Taking into account the tolerances of the grade 0 gauge block to be calibrated and the grade K reference gauge block. (S4.5±1. Temperature corrections ( α . Correlation: None of the input quantities are considered to be correlated to any significant extent.13.7 S4.02·D). eq.5)). Based on the calibration certificate of the reference gauge block and the manufacturer’s data for the gauge block to be calibrated the linear thermal expansion coefficient of the steel gauge blocks is assumed to be within the interval (11. Comparator (δlC): The comparator has been verified to meet the specifications stated in EALG21. The deviation of the mean temperature of measurement from the reference temperature t0 = 20 °C is estimated to be within ±0.
37 nm 5 The pooled estimate of the standard deviation has been taken from the tests made to confirm compliance of the comparator used with the requirements of EALG21.EA4/02 • Expression of the Uncertainty of Measurement in Calibration S4.3 nm 0. which for a normal distribution corresponds to a coverage probability of approximately 95 %. value 1 2 3 4 5 100 nm 90 nm 80 nm 90 nm 100 nm δ l = −94 nm sp (δ l ) = 12 nm u(δ l ) = s(δ l ) = arithmetic mean: pooled estimate of standard deviation: (obtained from prior evaluation) 12 nm = 5. no.236×106 0 mm 3.999 926 mm ±73 nm. The reported expanded uncertainty of measurement is stated as the standard uncertainty of measurement multiplied by the coverage factor k = 2.000 020 mm 15 nm 0 mm 17.11 Expanded uncertainty U = k × u(lX ) = 2 × 36. standard uncertainty: S4.0 normal 1. December 1999 rev00 Page 36 of 79 .5 nm 16.4 nm S4. the comparator being reset using the reference standard before each reading.12 Reported result The measured value of the nominal 50 mm gauge block is 49.0 nm 17.87 nm 49.87 nm 36.37 nm 18. obs.6 nm 11.999 926 mm estimate probability distribution sensitivity coefficient ci normal 1.8 nm 3.0 rectangular 575 nm°C1 special 50 mm rectangular 1.3 nm 5.9 Measurements (δl): The following observations are made for the difference between the unknown gauge block and the reference standard.37 nm 0 mm 18.10 Uncertainty budget (δlX): quantity Xi lS δlD δl δlC δt δα × ∆ t δlV lX standard uncertainty u(xi) xi 50.000 094 mm 5. obs.0 uncertainty contribution ui(y) 15.0289 °C 0 0.0 rectangular 1.5 nm 0 °C 0.0 triangular 1.4 nm ≅ 73 nm S4.
3) If the standard uncertainties u(x1) and u(x2) associated with the expectations x1 and x2 are much smaller than the moduli of the respective expectation values the third term on the right side may be neglected.1) December 1999 rev00 Page 37 of 79 . however. Temperature values are give in the t90 temperature scale.1 CALIBRATION OF A TYPE N THERMOCOUPLE AT 1000°C (S4. If the factors in the product are statistically independent with nonzero expectations. The resulting equation represents the case described by the usual method based on the linearisation of the model function.2 C ≅ tS (ViS ) + CS × δ ViS1 + CS × δ ViS2 + CS × δ VR − S δ t0S + δ tD + δ tF CS0 (S5.5) A type N thermocouple is calibrated by comparison with two reference thermocouples of type R in a horizontal furnace at a temperature of 1000 °C. is much smaller than the standard uncertainty u(x2) associated with this expectation or even zero. the product term involving this expectation may be neglected on the right side of equation (S4.EA4/02 • Expression of the Uncertainty of Measurement in Calibration S4. The resulting equation is 2 u2 ( x1 × x2 ) ≅ x1 u2 ( x2 ) + u 2 ( x1) × u2 ( x2 ) (S4. for example x2.2) (S4. The temperature tX of the hot junction of the thermocouple to be calibrated is t X = tS (ViS + δ ViS1 + δ ViS2 + δ VR − δ t0 S ) + δ tD + δ tF CS0 S5. one of the moduli of the expectation values. the square of the relative standard uncertainty of measurement (relative variance) associated with the product can be expressed without any linearisation by the squares of the relative standard uncertainties associated with the estimates of the factors: w2 ( x1 × x2 ) = w2 ( x1) + w2 ( x2 ) + w2 ( x1) × w2 ( x2 ) Using the definition of the relative standard uncertainty of measurement this expression is easily transformed into the general relation 2 2 u2 ( x1 × x2 ) = x2 u2 ( x1) + x1 u2 ( x2 ) + u2 ( x1) × u2 ( x2 ) (S4.3). All thermocouples have their reference junctions at 0 °C. The emfs generated by the thermocouples are measured using a digital voltmeter through a selector/reversing switch.3) gives a significant contribution: u2 ( x1 × x2 ) ≅ u2 ( x1) × u2 ( x2 ) S5 S5.13 Mathematical note on the standard uncertainty of measurement of the product of two quantities with zero expectation: If a product of two quantities is considered. only the third term in equation (S4. The thermocouple to be calibrated is connected to the reference point using compensating cables.4) If both moduli of the expectation values are much smaller than their associated standard uncertainties or even zero. the usual method of evaluation of uncertainty contributions based on the linearisation of the model function has to be modified if one or both of the expectations of the factors in the product are zero. If. but not the third term.
3 The voltage VX across the thermocouple wires with the cold junction at 0 °C during calibration is VX (t ) ≅ VX (t X ) + ∆ t δ t0 X − CX CX 0 ∆ t δ t0 X = ViX + δ ViX1 + δ ViX2 + δ VR + δ VLX + − CX CX 0 (S5. CX0 δtD δtF t ∆t = t . The expanded uncertainty of measurement at 1000 °C is U = 0. CX CS0.temperature of the reference thermometer in terms of voltage with cold junction at 0 °C. .voltage correction due to contact effects of the reversing switch. .tX δVLX S5.voltage corrections obtained from the calibration of the voltmeter. ViX δViS1.temperature correction due to nonuniformity of the temperature of the furnace. . .change of the values of the reference thermometers since their last calibration due to drift. . . Because the measurement process consists of two steps — determination of the temperature of the furnace and determination of emf of the thermocouple to be calibrated — the evaluation of the uncertainty of measurement is split in two parts. .deviation of the temperature of the calibration point from the temperature of the furnace.EA4/02 • Expression of the Uncertainty of Measurement in Calibration S5.temperature at which the thermocouple is to be calibrated (calibration point). S5.sensitivities of the thermocouples for voltage at the reference temperature of 0 °C.4 . . . Reference standards (tS(V)): The reference thermocouples are supplied with calibration certificates that relate the temperature at their hot junction with their cold junction at 0 °C to the voltage across their wires.temperature corrections due to the deviation of the reference temperatures from 0 °C. . The function is given in the calibration certificate.3 °C (coverage factor k = 2). The reported result is the output emf of the thermocouple at the temperature of its hot junction. .sensitivities of the thermocouples for voltage at the measuring temperature of 1000 °C.voltage correction due to the compensating cables. δViX1 δViS2.indication of the voltmeter. δt0X CS. .voltage corrections due to the limited resolution of the voltmeter.2) where: tS(V) ViS.5 December 1999 rev00 Page 38 of 79 . δViX2 δVR δt0S.
Parasitic voltages (δVR): Residual parasitic offset voltages due to the switch contacts have been estimated to be zero within ±2 µV. 1st standard. S5. 2nd cycle: 1st standard. CX0): The voltage sensitivities of the thermocouples have been taken from reference tables: 1000 °C reference thermocouple unknown thermocouple 0 °C CS = 0. unknown thermocouple. The calibration certificate gives a constant expanded uncertainty of measurement for voltages smaller than 50 mV of U = 2.1 °C. unknown thermocouple. S5. S5. Reversion of polarity. tS(ViS).026 °C/µV CS0 = 0. unknown thermocouple.077 °C/µV CX = 0.039 °C/µV S5.3 °C. Resolution of the voltmeter (δViS2. 1st standard.5 µV at each indication. 2nd standard.12 Temperature gradients (δtF): The temperature gradients inside the furnace have been measured. unknown thermocouple.13 Compensating cables (δVLX): The compensating cables have been investigated in the range 0 °C to 40 °C.8 S5.6 Calibration of the voltmeter (δViS1. CS0.0 µV (coverage factor k = 2). Reference temperatures (δt0S. deviations from nonuniformity of temperature in the region of measurement are within ±1 °C. the voltage differences between the cables and the thermocouple wires are estimated to be within ±5 µV. 2nd standard. δViX2): A 4½ digit microvoltmeter has been used in its 10 mV range resulting in resolution limits of ±0.14 Measurements (ViS. δt0X): The temperature of the reference point of each thermocouple is known to be 0 °C within ±0. At 1000 °C.189 °C/µV CS0 = 0. 2nd standard. Corrections to the measured voltages are made to all results.10 Voltage sensitivities (CS.EA4/02 • Expression of the Uncertainty of Measurement in Calibration S5. S5.9 S5. 2nd standard.7 S5. CX. From this. ViX): The indications of the voltmeter are recorded in the following operational procedure which gives four readings for every thermocouple and reduces the effects of temperature drift in the thermal source and of parasitic thermal voltages in the measuring circuit: 1st cycle: 1st standard.11 Drift of the reference standard (δtD): From previous calibrations the drift of the reference standards are estimated to be zero within the limits ±0. December 1999 rev00 Page 39 of 79 . δViX1): The voltmeter has been calibrated.
6 µV s (V ) u(ViX) = p iX = 1. Therefore. Thermocouple Indicated voltage.4 °C S5. The observed temperature values are highly correlated (correlation factor nearly one). by taking their mean value.16 From the four readings on each thermocouple given in the table above. corrected 1st reference Unknown 2nd reference +10500 µV +10503 µV 10503 µV 10504 µV +36245 µV +36248 µV 36248 µV 36251 µV 36248 µV 1000.3 °C.EA4/02 • Expression of the Uncertainty of Measurement in Calibration S5. It gave a pooled estimate of standard deviation for the temperature of the furnace and the voltage of the thermocouple to be calibrated. The voltage values of the reference thermocouples are converted into temperature values by means of the temperaturevoltage relations stated in their calibration certificates.6 °C Mean voltage Temperature of the hot junction Temperature of the furnace 10502.5 °C +10503 µV +10503 µV 10505 µV 10505 µV 10504 µV 1000. they are combined to one observation only.10 °C 1 sp(ViX) = 1. The respective standard uncertainties of measurement of the observed quantities are: pooled estimate of standard deviation: standard uncertainty: pooled estimate of standard deviation: standard uncertainty: s p( t S ) = 0. the mean value of the voltages of each thermocouple is deduced.6 µV 1 December 1999 rev00 Page 40 of 79 .10 °C s (t ) u(tS) = p S = 0. If the difference is not within these limits the observations have to be repeated and/or the reasons for such a large difference have to be investigated. which is the temperature of the furnace at the location of the thermocouple to be calibrated. one observation of the voltage of the thermocouple to be calibrated has been extracted. In order to evaluate the uncertainty of measurement associated with these observations. a series of ten measurements has been previously undertaken at the same temperature of operation.15 The procedure requires that the difference between the two reference standards must not exceed ±0.5 µV 1000. In a similar way.
60 µV 1.48 µV 25.0 1.022 °C 0.058 °C probability distribution normal normal rectangular rectangular rectangular normal rectangular sensitivity coefficient ci 1. quantity estimate standard uncertainty u(xi) 1.5 µV/°C 25.024 °C 0.173 °C 0.077 °C 0.577 °C 0.0 1.15 °C 0.29 µV 1.00 µV 0.0 1.29 µV 1.15 °C 0.077 °C/µV 0.60 µV 1.00 µV 0.29 µV 1.5 °C 0 µV 0 µV 0 µV 0 °C 0 °C 0 °C 0 °C 1000.15 µV 2.077 °C/µV 0.00 µV 0.0 0.0 1.10 °C 1.0 1.089 °C 0.5 µV 1.6 µV/°C uncertainty contribution ui(y) 1.18 Uncertainty budget (emf VX of the thermocouple to be calibrated): The standard uncertainty of measurement associated with the temperature deviation of the calibration point from the temperature of the furnace is the standard uncertainty of measurement associated with the temperature of the furnace because the temperature point is a defined value (exactly known).10 °C 0.577 °C probability distribution normal normal rectangular rectangular rectangular normal rectangular rectangular sensitivity coefficient ci 1.5 °C standard uncertainty u(xi) 0.641 °C 0.9 µV 24.641 °C S5.15 µV 0.17 Uncertainty budget (temperature tX of the furnace): quantity estimate Xi tS δViS1 δViS2 δVR δt0S δtS δtD δtF tX xi 1000.15 µV 2.EA4/02 • Expression of the Uncertainty of Measurement in Calibration S5.0 µV Xi ViX δ ViX1 δ ViX2 δVR δVLX ∆t δt0X VX xi 36 248 µV 0 µV 0 µV 0 µV 0 µV 0.077 °C/µV 0.5 °C 0 °C 36 229 µV December 1999 rev00 Page 41 of 79 .407 1.9 µV 0.0 1.058 °C 0.173 °C 0.0 38.0 uncertainty contribution ui(y) 0.
which for a normal distribution corresponds to a coverage probability of approximately 95 %.EA4/02 • Expression of the Uncertainty of Measurement in Calibration S5.0 µV ≅ 50 µV S5. using a dual power meter with ratio facility.1 CALIBRATION OF A POWER SENSOR AT A FREQUENCY OF 19 GHZ The measurement involves the calibration of an unknown power sensor with respect to a calibrated power sensor used as a reference by substitution on a stable transfer standard of known small reflection coefficient.19 Expanded uncertainties The expanded uncertainty associated with the measurement of the temperature of the furnace is U = k × u(tX) = 2 × 0. at the temperature of 1000. an emf of 36 230 µV ±50 µV.2 G S X ΓG Γ X or Γ S Power meter December 1999 rev00 Page 42 of 79 . Schematic of the measuring system Transfer standard 1. one determines the (indicated) ratio of the power for the sensor to be calibrated.641 °C ≅ 1.20 Reported result The type N thermocouple shows.0000 B/A S6. At each frequency. S6 S6. which is defined as the ratio of incident power at the reference frequency of 50 MHz to the incident power at the calibration frequency under the condition that both incident powers give equal power sensor response. respectively the reference sensor and the internal sensor that forms part of the transfer standard. The measurement is made in terms of calibration factor. The reported expanded uncertainty of measurement is stated as the standard uncertainty of measurement multiplied by the coverage factor k = 2.0 °C with its cold junction at a temperature of 0 °C.3 °C The expanded uncertainty associated with the emf value of the thermocouple to be calibrated is U = k × u(VX) = 2 × 25.
.correction of the observed ratio for nonlinearity and limited resolution of the power meter at power ratio level of the calibration frequency. . .correction of the observed ratio for nonlinearity and limited resolution of the power meter at power ratio level of the reference frequency.mismatch factor of sensor to be calibrated at the calibration frequency.1) for the equal power meter indication where: Pr Pc incident power at the reference frequency (50 MHz). . . .mismatch factor of sensor to be calibrated at the reference frequency.indicated power ratio for the sensor to be calibrated at the reference frequency. pCc p= pSr pSc pXr pXc pSr pXc pSc pXr December 1999 rev00 Page 43 of 79 .change of the calibration factor of the reference power sensor since its last calibration due to drift.indicated power ratio for the reference sensor at the calibration frequency.4 The calibration factor of the unknown sensor is obtained from the relationship M M KX = ( KS + δ KD ) Sr Xc pCr pCc p (S6. . incident power at the calibration frequency.indicated power ratio for the sensor to be calibrated at the calibration frequency.observed ratio of power ratios derived from: .mismatch factor of standard sensor at the calibration frequency.EA4/02 • Expression of the Uncertainty of Measurement in Calibration S6. voltage reflection coefficient of the sensor at the reference frequency voltage reflection coefficient of the sensor at the calibration frequency power absorbed by the sensor at the reference frequency power absorbed by the sensor at the calibration frequency Γr Γc PAr PAc S6.3 The quantity K . is defined as: K= (1 + Γ r ) PAr P Ir = P (1 + Γ c 2 ) PAc Ic 2 (S6. .mismatch factor of reference sensor at the reference frequency. termed ‘calibration factor’ by some manufacturers.calibration factor of the reference power sensor. .2) M Sc M Xr where: KS δKD MSr MSc MXr MXc pCr . .indicated power ratio for the reference sensor at the reference frequency. .
12 (S6.8 Mismatch factors (MSr. the unknown and the standard power sensors are not known. Because power ratios at both frequencies are considered.12. the separated correlated and uncorrelated contributions are not known and the values given are upper bounds for the standard uncertainty of measurement associated with ratios. The value of the calibration factor. S6. resulting in a standard uncertainty of 0. The corresponding limits of deviation have to be calculated for the reference and the calibration frequencies from the relationship: M S.X where the magnitudes of the reflection coefficients of the transfer standard.02 0. Thus.12). MSc. is (95.7±1. the uncertainty contributions at the reference as well at the calibration frequency are correlated. As shown in S3.001 with deviations within ±0.5 Reference sensor (KS): The reference sensor was calibrated six months before the calibration of the unknown power sensor.002.000142 with the correction factor pCc.02 0. i. Since the same power meter has been used to observe both pS and pX. The expanded uncertainty of measurement stated for the readings of the power meter contains linearity and resolution effects.002 per year with deviations within ±0.1) % (coverage factor k = 2). The linearity effects are correlated whereas the resolution effects are uncorrelated.e. however. In the calculations above.011.0002 (coverage factor k = 2) at the power ratio level of calibration frequency due to nonlinearity of the power meter used.7 S6. is estimated to equal 0. pCc): The expanded uncertainty of 0. which may also be expressed as 0. only the relative difference in the readings due to systematic effects should be taken into account (see the mathematical note in paragraph S3.957±0.004.3) ΓG ΓS ΓX December 1999 rev00 Page 44 of 79 .10 0.00142 associated with the correction factor pCr and 0.X = 1 ± 2 Γ G Γ S.07 0. the approximations are justified. MXr MXc): As the transfer standard system is not perfectly matched and the phase of the reflection coefficients of the transfer standard. there will be an uncertainty due to mismatch for each sensor at the reference frequency and at the calibration frequency.02 18 GHz 0.002 (coverage factor k = 2) is assigned to the power meter readings at the power ratio level of the reference frequency and of 0. The uncertainty budget finally shows that the contributions arising from these ratios are insignificant. the effect of the correlations is to reduce the uncertainty. These values have been obtained from previous measurements. building the power ratio cancels the influence of correlations and gives a reduced standard uncertainty of measurement to be associated with the ratio. which was calibrated half a year ago.6 S6. the reference sensor and the sensor to be calibrated are: 50 MHz 0. Linearity and resolution of the power meter (pCr. Drift of the standard (δKD): The drift of the calibration factor of the reference standard is estimated from annual calibrations to be 0.EA4/02 • Expression of the Uncertainty of Measurement in Calibration S6. given in the calibration certificate. From these values. the drift of the reference sensor.
9953 1.11 Uncertainty budget (KX): p = 0.976 0. The standard uncertainty due to mismatch is therefore obtained from: u( M S.933 0.9942 0.956 uncertainty contribution ui(y) 0.00053 0.9999 0.00459 0.9698 0.0048 3 quantity Xi KS δKD MSr MSc MXr MXc pCr pCc p KX estimate xi 0.00053 0.0001 0.0083 u ( p ) = s( p ) = 0.0014 0.0055 0.01110 0.9772 0.933 0.X ) = 2 ΓG ΓS 2 (S6.0001 1.9671 0.10 Measurements (p): Three separate readings are made which involve disconnection and reconnection of both the reference sensor and the sensor to be calibrated on the transfer standard to take connector repeatability into account.00132 0.000 1.0012 0.933 0. This is accounted for by adding the square root of the sum of the uncertainty of measurement squared and the measured value squared. This is taken into account by replacing the factor 1/3 for a rectangular distribution by 1/2 in calculating the variance from the square of the halfwidth determined from the limits.976 0.0006 0.000 0.933 standard uncertainty u(xi) 0.9615 0. no p pSc pXr pXc pSr 1 2 3 1.0001 1.0083 = 0.0006 0.000 1.4) Note: The values of the reflection coefficients are the results of measurements which are themselves subject to uncertainty.0048 probability distribution normal rectangular Ushaped Ushaped Ushaped Ushaped normal normal normal sensitivity coefficient ci 0.976 0 s( p) = 0.00113 0.0000 1.EA4/02 • Expression of the Uncertainty of Measurement in Calibration The probability distribution of the individual contributions is Ushaped.000 1.0000 0.9924 0.000 1.00924 0.933 0.001 1.957 0.976 0.9 Correlation: None of the input quantities are considered to be correlated to any significant extent.0119 0.000 1.0099 0.00537 0.00013 0.933 0. The power meter readings used to calculate the observed power ratio p are as follows: obs.933 0.9792 0.9836 arithmetic mean: experimental standard deviation: standard uncertainty: S6.0001 0. S6.01623 December 1999 rev00 Page 45 of 79 . S6.
13 Reported result: The calibration factor of the power sensor at 18 GHz is 0.indicated attenuation with the attenuator to be calibrated.correction obtained from the calibration of the reference attenuator. The reported expanded uncertainty of measurement is stated as the standard uncertainty of measurement multiplied by the coverage factor k = 2. Schematic of the measuring system Step attenuator S7. set at 0 dB.933 ±0. .EA4/02 • Expression of the Uncertainty of Measurement in Calibration S6.change of the attenuation of the reference attenuator since its last calibration due to drift.1 The measurement involves the calibration of a coaxial step attenuator at 10 GHz using an attenuation measuring system containing a calibrated step attenuator which acts as the attenuation reference. .correction due to mismatch loss. S7 CALIBRATION OF A COAXIAL STEP ATTENUATOR AT A SETTING OF 30 DB (INCREMENTAL LOSS) S7.032. The method of measurement involves the determination of the attenuation between matched source and matched load.032 S6. In this case the unknown attenuator can be switched between settings of 0 dB and 30 dB and it is this change (called incremental loss) that is determined in the calibration process.3 3 0 0 0 dB ΓG ΓL s22a s22b RF attenuation measuring system The attenuation LX of the attenuator to be calibrated is obtained from the relation: LX = LS + δLS + δLD + δLM + δLK + δLib . December 1999 rev00 Page 46 of 79 .indicated attenuation with the attenuator to be calibrated. . which for a normal distribution corresponds to a coverage probability of approximately 95 %.δLia + δL0b . which may also be expressed as (93.2) %.3±3.12 Expanded uncertainty: U = k × u(KX) = 2 × 0.052 dB G s11a s11b S7.δL0a where: LS = Lib – Lia Lia Lib δLS δLD δLM (S7.1) . .attenuation difference of reference attenuator derived from: . set at 30 dB.01623 ≅ 0. The attenuation measuring system has a digital readout and an analogue null detector which is used to indicate the balance condition.2 30.
005 dB (coverage factor k = 2) is considered to be valid for attenuation settings of the reference attenuator that differ not more than ±0. S7.EA4/02 • Expression of the Uncertainty of Measurement in Calibration δLK δLia.95 30 dB 0.686 2 2 2 2 2 2 2 2 4 4 Γ S ( s11a + s11b ) + Γ L ( s22a + s22b ) + Γ S × Γ L ( s21a + s21b ) 2 (S7. Without any phase information.003 dB with the associated expanded uncertainty of 0. . but the standard uncertainty (in dB) due to the incomplete knowledge of the match is estimated from the relationship [1]: 8.2) with the source and load reflection coefficients u(δ LM ) = S7.corrections due to the limited resolution of the null detector at 0 dB and 30 dB settings.002 dB. .6 ΓL = 0. December 1999 rev00 Page 47 of 79 .1 dB from the calibrated setting of 30.003 dB with an associated expanded uncertainty of 0.5 S7.03 and ΓS = 0.09 0. Drift of the reference (δLD): The drift of the attenuation of the reference attenuator is estimated from its calibration history to be zero with limits ±0.02 dB. Mismatch loss (δLM): The reflection coefficients of the source and the load at the insertion point of the attenuator to be calibrated have been optimised by impedance matching to as low magnitudes as possible. This is accounted for by adding the square root of the sum of uncertainty of measurement squared and the measured value squared.000 dB setting at 10 GHz of 30.03 and the scattering coefficients of the attenuator to be calibrated at 10 GHz 0 dB s11 s22 s21 0. a correction for mismatch error cannot be made. Their magnitudes and the magnitudes of the scattering coefficients of the attenuator to be calibrated have been measured but their phase remains unknown.4 Reference attenuator (δLS): The calibration certificate for the reference attenuator gives a value of attenuation for the 30. δL0b .005 dB (coverage factor k = 2).000 dB. δLib δL0a.01 0. Note: The values of scattering and reflection coefficients are the results of measurements which are themselves not exactly known. The correction of +0.031 as u(δLM) = 0.correction for leakage signals between input and output of the attenuator to be calibrated due to imperfect isolation.corrections due to the limited resolution of the reference detector at 0 dB and 30 dB settings.05 0.01 0.
000 dB 0.033 dB 30.0019 dB 0.0005 dB.043 dB standard uncertainty u(xi) 0.002 dB at each reading with assumed normal probability distribution.10 Correlation: None of the input quantities are considered to be correlated to any significant extent.0011 dB 0.052 dB LS = 30.0 1.0224 dB December 1999 rev00 Page 48 of 79 .000 dB 0.040 dB s(LS) = 0. Resolution of the reference attenuator setting (δLia.0200 dB 0. S7.0090 dB 0. δL0b): The detector resolution was determined from a previous evaluation to have a standard deviation of 0.001 dB from which the correction for resolution is estimated to be within ±0.018 dB 0.000 dB 0. Resolution of the null detector (δL0a.0 1.12 Uncertainty budget (LX): quantity Xi LS δLS δLD δLM δLK δLia δLib δL0a δL0b LX estimate xi 30. obs.0 1. S7. δLib): The digital readout of the reference attenuator has a resolution of 0.0020 dB 0.7 Leakage correction (δLK): Leakage signals through the attenuator to be calibrated have been estimated from the measurements at 0 dB setting to be at least 100 dB below the measurement signal.0 uncertainty contribution ui(y) 0.0 1.0011 dB 0.0 1.058 dB 30.003 dB at the 30 dB setting.003 dB 0 dB 0 dB 0 dB 0 dB 0 dB 0 dB 0 dB 30.0020 dB probability distribution normal rectangular Ushaped Ushaped Ushaped Ushaped rectangular rectangular normal sensitivity coefficient ci 1.040 dB 0.0 1. The correction for leakage signals is estimated from these findings to be within ±0.11 Measurements (LS): Four observations are made of the incremental loss of the attenuator to be calibrated between settings of 0 dB and 30 dB: obs.0 1.9 S7.018 dB u( LS ) = s( LS ) = = 0.0003 dB 0.0090 dB 0.000 dB 30 dB setting 30.0017 dB 0.018 dB 30.0025 dB 0.0003 dB 0.0200 dB 0.0 1.0017 dB 0.0003 dB 0.0020 dB 0.0020 dB 0.0025 dB 0. no.009 dB 4 arithmetic mean: experimental standard deviation: standard uncertainty: S7.8 S7. values at 0 dB setting 1 2 3 4 0.EA4/02 • Expression of the Uncertainty of Measurement in Calibration S7.
No. In: IEE Proc. 128. : Reexamination of mismatch uncertainty when measuring microwave power and attenuation. 1. . 1981 December 1999 rev00 Page 49 of 79 . Febr.045 dB S7. F. which for a normal distribution corresponds to a coverage probability of approximately 95 %.043 ±0. A. Warner.13 Expanded uncertainty: U = k × u(LX) = 2 × 0.15 Reference [1] Harris. S7.0224 dB ≅ 0. Vol.14 Reported result: The measured value of the step attenuator for a setting of 30 dB at 10 GHz is (30. L.045) dB. I. The reported expanded uncertainty of measurement is stated as the standard uncertainty of measurement multiplied by the coverage factor k = 2. Pt. H..EA4/02 • Expression of the Uncertainty of Measurement in Calibration S7.
EA4/02 • Expression of the Uncertainty of Measurement in Calibration SUPPLEMENT 2 Examples December 1999 rev00 Page 50 of 79 .
EA4/02 • Expression of the Uncertainty of Measurement in Calibration CONTENTS S8 S9 S10 S11 S12 S13 INTRODUCTION CALBRATION OF A HANDHELD DIGITAL MULTIMETER AT 100 V DC CALIBRATION OF A VERNIER CALLIPER CALIBRATION OF A TEMPERATURE BLOCK CALIBRATOR AT A TEMPERATURE OF 180°C CALIBRATION OF A HOUSEHOLD WATER METER CALIBRATION OF A RING GAUGE WITH A NOMINAL DIAMETER OF 90 MM 52 55 59 64 68 72 December 1999 rev00 Page 51 of 79 .
The examples are based on drafts prepared by EA Expert Groups.3 S8.EA4/02 • Expression of the Uncertainty of Measurement in Calibration S8 S8. the applicable sensitivity coefficient has to be employed. that the result of measurement has a rectangular distribution (as would be the case if there were only one term.2 S8. independent of the field of calibration. are inserted for pedagogical reasons at places where they are needed instead of presenting them in the main document. that in practical applications there is no need to go through the mathematical derivations presented in these examples.14). one can immediately draw the conclusion that the coverage factor to be used to arrive at a coverage probability of 95 % is k = 1. They supplement the examples presented in Supplement 1 to EALR2 (Edition 1. One general conclusion that may be drawn from the uncertainty propagation is that in the case of only one dominant contribution the type of distribution of this contribution applies for the result of measurement as well. These drafts have been simplified and harmonised to make them transparent to the laboratory staff in all fields of calibration. however. which may be felt as complications. S8.1 INTRODUCTION The following examples are chosen to demonstrate further the method of evaluating the uncertainty of measurement. to evaluate the uncertainty of the result of measurement. the user is encouraged to employ the results of the theoretical presentations after having made himself acquainted with the conditions that have to be fulfilled. as usual. For instance. It should be added that the situation where there is only one or a few dominant terms to the uncertainty of measurement is often met in connection with less complicated measuring instruments. that needed to be considered in the propagation). However. The examples are chosen to illustrate situations encountered in practice. Thus it may appear a paradox that the treatment of uncertainty of measurement for less complicated instruments. rectangularly distributed. It should be emphasised. It is thus hoped that this set of examples.4 S8. will contribute to a better understanding of the details of setting up the model of evaluation and to the harmonisation of the process of evaluating the uncertainty of measurement. as shown by the examples of this Supplement. The present collection of examples focuses on situations where there are one or two dominant terms in the uncertainty propagation or where the number of repeated measurements is small. November 1997). is more complicated than the treatment of the more straightforward examples in Supplement 1. it should be kept in mind that the mathematical derivations. if it is ascertained. in a given situation.65 (see S9.5 December 1999 rev00 Page 52 of 79 . where the dominant term often is due to the limited resolution of the instrument. like the preceding set published as Supplement 1 to EALR2. Rather. in particular in the mathematical notes appended to some of the examples. However.
To keep the analysis transparent.13 The notation applied here distinguishes between main values. often enter the uncertainty analysis. which realises the local unit. In most cases this will lead to an estimation of the limits for the unknown deviations. S8. Besides these two quantities there are several others.EA4/02 • Expression of the Uncertainty of Measurement in Calibration S8. The presentation of the examples follows the common scheme presented and implemented in the first supplement to EALR2.e. They are represented by lowercase letters in italics. Some of the corrections are given by complete results of measurement. It is evident. another is the quantity presented by the working standard. with this quantity the measurand is compared.e. which generally speaking characterise or identify calibration artefacts. the examples have been designed to follow the notational rules given below. a measured value and its associated measurement uncertainty. S8. nominal values. i. S8.6 The contributions and values given in the examples are not intended to imply mandatory or preferred requirements. For details the reader is referred to clause S1. Laboratories should determine the uncertainty contributions on the basis of the model function they use in the evaluation of the particular calibration they perform and report the evaluated uncertainty of measurement on the calibration certificate they issue. which take the role of additional local quantities or corrections. also for those who are not experts in the relevant metrological field. S8.10 Corrections describe the imperfect equality between a measurand and the result of a measurement. correction values and values of limits: Main values are measured or observed values that contribute an essential part to the value of a measurand. i. that it is not possible to follow such rules strictly. There are several recurrent quantities involved in all cases. focused more on the physical background than on the current practice in different fields. because the practice concerning the use of symbols is different in different metrological fields.11 In certain cases the quantity presented by a working standard is characterised by the nominal value of the standard. in all cases. the quantity to be measured. The uncertainty analysis of the examples is intended to represent the fundamentals of the specific measurement process and the method of evaluating the measurement result and the associated uncertainty. however.9 S8. a uniform method for the choice of the symbols of quantities has been followed. they will be preceded by an uppercase Greek delta if the quantity represents a difference.4 of that document. For others the distribution of values is inferred from more or less complete knowledge of their nature. Thus nominal values. One of them is the measurand. December 1999 rev00 Page 53 of 79 .7 S8.8 S8.12 To distinguish in the mathematical models of evaluation between these concepts.
They are represented by the symbol chosen for the quantity under consideration. December 1999 rev00 Page 54 of 79 . EXAMPLE: L nominal length of a gauge block to be calibrated.g. The differentiation between different quantities of the same kind is effected by indices as shown in the examples.correction for eccentricity of load and magnetic effects in the calibration of a weight.EA4/02 • Expression of the Uncertainty of Measurement in Calibration EXAMPLE: ti X ∆l  temperature indicated by a thermometer X to be calibrated. Nominal values are assigned values of the realisation of a quantity by a standard or a measuring instrument. Correction values give small deviations from the main values that are known or have to be estimated. They are approximate values that give the main part of the realised value. EXAMPLE: p0 reference pressure.14 Defined reference values are represented by a quantity symbol with the index zero. EXAMPLE: ∆α X  estimated halfwidth of the interval of possible deviations of a linear thermal resistivity coefficient given in a manufacturer’s specification for a resistor to be calibrated. The internationally accepted notational rules for physical quantities have been followed: indices representing physical quantities are given in italics whereas indices that symbolise artefacts. Values of limits are fixed. preceded by a uppercase Greek delta. instruments and so on are written in upright letters. They are represented by uppercase letters in italics. In most cases they are additive. of 1000 mbar. (index i means indicated). preceded by a lowercase Greek delta. They are represented by the symbol chosen for the quantity under consideration.possible deviation because of the drift of the value of a reference weight since its last calibration δ mC . S8. observed difference in the displacement of a measuring spindle. e. EXAMPLE: δ mD . estimated values of possible variations of the unknown values of a quantity.
Calibration of pressure balances EALG31. a handheld digital multimeter (DMM) is calibrated at an input of 100 V DC using a multifunction calibrator as a working standard.16 If several indices are used.1) (1) (2) (3) S9. EXAMPLE: Vi 1 . Vi 2  voltage indicated by voltmeter '1' and voltmeter '2'. the sequence of indices is chosen in such a way that the index representing the most general concept is leftmost and the one representing the most specific concept is rightmost. S9 S9. illustrating different aspects encountered in connection with the calibration of measuring instruments. Examples may also be found in EAL and EA Guidance Documents1 dealing with the calibration of specific types of measuring instruments. EA Guidelines on the Determination of Pitch Diameter of Parallel Thread gauges by Mechanical Probing Page 55 of 79 December 1999 rev00 . It must be noted that the error of indication of the DMM which is obtained using this measuring procedure includes the effect of offset as well as deviations from linearity.15 Ratios of quantities of the same kind (dimensionless ratios) are represented by lowercase letters in italics. S8. the DMM reading is noted. Calibration of thermocouples EALG32. The error of indication E X of the DMM to be calibrated is obtained from E X = Vi X − VS + δ Vi X − δ VS where (S9.1 CALBRATION OF A HANDHELD DIGITAL MULTIMETER AT 100 V DC As part of a general calibration. EXAMPLE: r = Ri X / Ri N  ratio of indicated resistance of an unknown resistor and a reference resistor (index i means indicated).EA4/02 • Expression of the Uncertainty of Measurement in Calibration S8.3 1 EALG26. respectively S8. The following measuring procedure is used: The calibrator’s output terminals are connected to the input terminals of the DMM using suitable measuring wires. after a suitable stabilising period. Measurement and generation of small ac voltages with inductive voltage dividers EA10/10.17 The examples in this second supplement to EALR2 are intended to be followed by others. The calibrator is set to its 100V setting and. The error of indication of the DMM is calculated using the DMM readings and the calibrator settings.2 S9.
DMM readings ( Vi X ) The DMM indicates the voltage 100. the resistive load at the calibrator’s terminals is greater than 100 kΩ. . (3) deviations in the ambient temperature. S9. S9.EA4/02 • Expression of the Uncertainty of Measurement in Calibration Vi X VS δ ViX δ VS .correction of the indicated voltage due to the finite resolution of the DMM. The DMM reading is assumed to be exact (see S9.01% of setting + 1 mV).1 V. These specifications state that the voltage generated by the calibrator coincides with the calibrator setting within ±(0. resulting in an expanded uncertainty of measurement associated with the 100 V setting of U=0.voltage. .e.0 V with limits of +0.8 Other corrections ( δ VS ) Because of the fact that individual figures are not available the uncertainty of measurement associated with the different sources is derived from the accuracy specification given by the manufacturer of the calibrator. Even if this method is considered to be equivalent to the expression given above it is not used here because it may be misleading in many cases and because it does not represent an equation of physical quantities in the internationally accepted symbolic nomenclature. (1) (2) (3) 2 A widely used method of presenting accuracy specification of measuring instruments in data sheets or manuals consists in giving the specification limits in terms of ‘settings’. (5) loading effects resulting from the finite input resistance of the DMM to be calibrated.7 The least significant digit of the DMM display corresponds to 0.4). . (2) deviations resulting from the combined effect of offset. no scatter in the indicated values is observed. nonlinearity and differences in gain. (4) deviations in mains power. December 1999 rev00 Page 56 of 79 . For the calibrator.correction of the calibrator voltage due to (1) drift since its last calibration.5 Because of the limited resolution of the indication of the DMM.1 V at the calibrator setting 100 V. one half of the magnitude of the least significant digit).4 S9.000 1×VS +1 mV)2 under the measuring conditions the ambient temperature is within the range 18 °C to 23 °C the mains voltage powering the calibrator is in the range 210 V to 250 V.6 Working standard ( VS ) The calibration certificate for the multifunction calibrator states that the voltage generated is the value indicated by the calibrator setting and that the associated expanded relative uncertainty of measurement is W=0.voltage generated by the calibrator.000 02 (coverage factor k = 2 ). Resolution of DMM to be calibrated ( δ Vi X ) S9. the statement would be ±(0.002 V (coverage factor k = 2 ). S9. indicated by the DMM (index i means indication). Each DMM reading has a correction due to the finite resolution of the display which is estimated to be 0.05 V (i.
10 Uncertainty budget ( E X ) quantity Xi estimate xi standard uncertainty u ( xi ) probability distribution normal rectangular rectangular sensitivity coefficient ci uncertainty contribution ui ( y ) Vi X VS δ ViX 100.05) V.001 V 0.65 ⋅ 0.0 V 0.05 V S9.13 Additional remark The method used for calculating the coverage factor is clearly related to the fact that the measurement uncertainty associated with the result is dominated by the effect of the finite resolution of the DMM.10 ±0. the correction to be applied to the voltage generated by the calibrator is assumed to be 0.0 1.12 Reported result The measured error of indication of the handheld digital voltmeter at 100 V is (0. Therefore.0064 V 1. This will be true for the calibration of all lowresolution indicating instruments provided the finite resolution is the only dominant source in the uncertainty budget.1 V 0.65 which has been derived from the assumed rectangular probability distribution for a coverage probability of 95%.0 V within ±0.030 V δ VS EX S9.0 1. (S9.030 V ≅ 0. Since these conditions of measurement are fulfilled and the calibration history of the calibrator shows that the manufacturer’s specification may be relied upon. The coverage factor appropriate for a rectangular distribution is calculated from the relation given in eq.14.1 V 100. S9.9 Correlation None of the input quantities are considered to be correlated to any significant extent.0064 V 0. S9.001 V 0. The final distribution is not normal but essentially rectangular.8) in the mathematical note S9.11 Expanded uncertainty The standard uncertainty of measurement associated with the result is clearly dominated by the effect of the finite resolution of the DMM. The reported expanded uncertainty of measurement is stated as the standard uncertainty of measurement multiplied by the coverage factor k = 1.011 V.EA4/02 • Expression of the Uncertainty of Measurement in Calibration (4) the calibrator has been calibrated within the last year. S9.0 V 0. December 1999 rev00 Page 57 of 79 . U = k ⋅ u ( E X ) = 1.0 0. the method of effective degrees of freedom described in Annex E of EALR2 is not applicable.029 V 0.0 V 0.029 V 0.
3) denotes the total uncertainty contribution of the nondominant terms. The maximum relative change in the standard uncertainty resulting from the factor within the brackets in eq.5) Inverting this relation for a given coverage probability results in the relation between the expanded measurement uncertainty and the coverage probability U = U ( p) for the given distribution density ϕ ( y ) . (S9. As long as the ratio of the total uncertainty contribution uR ( y ) of the nondominant terms to the uncertainty contribution u1 ( y ) of the dominant term is not larger than 0. Under these assumptions the distribution of values that could reasonably be attributed to the measurand is essentially identical with the distribution resulting from the known dominant contribution. for instance the term with index 1.EA4/02 • Expression of the Uncertainty of Measurement in Calibration S9. From this distribution density ϕ ( y ) the coverage probability p may be determined for any value of the expanded measurement uncertainty U by the integral relation y +U p(U ) = y −U ∫ ϕ ( y' )dy' . the coverage factor may finally be expressed as k ( p) = U ( p) .14 Mathematical note If the situation of measurement is such that one of the uncertainty contributions in the budget can be identified as a dominant term. This value is within the accepted tolerance for mathematical rounding of uncertainty values. u ( y ) ≅ u1 ( y ) ⋅ 1 + R 2 u1 ( y ) (S9. Using this relation.4) is not larger than 5%.2) Here is uR ( y ) = ∑u i =2 N 2 i ( y) (S9.6) December 1999 rev00 Page 58 of 79 . (S9.4) The relative error of approximation is smaller than 1 × 10 −3 . u( y) (S9. (S9.3. (S9.2) may be approximated by 1 u ( y) 2 . the standard uncertainty to be associated with the measurement result y can be written as 2 u ( y ) = u12 ( y ) + uR ( y ) . eq.
9 mm). a (S9. at 0 mm. (3. parallelism.65 . 0. The reading interval of the calliper is 0. The relevant ratio is uR ( E X ) / uδ VX ( E X ) = 0. December 1999 rev00 Page 59 of 79 . Thus the resulting distribution of values that can reasonably be attributed as errors of indications is essentially rectangular.1 A vernier calliper made of steel is calibrated against grade I gauge blocks of steel used as working standards. They are selected in such a way that the measurement points are spaced at nearly equal distances (e.EA4/02 • Expression of the Uncertainty of Measurement in Calibration In the case of the handheld digital voltmeter the dominant uncertainty contribution resulting from the finite resolution of the indication is uδ VX ( E X ) = 0. squareness). 0. Before calibration several checks of the condition of the calliper are made.g.5 .8) For a coverage probability p = 95 % applicable in the EA.3 mm.029 V whereas the total uncertainty contribution of the nondominant terms is uR ( E X ) = 0. S10 CALIBRATION OF A VERNIER CALLIPER S10. These include dependence of the result of measurement on the distance of the measured item from the beam (Abbe error). The coverage probability for a rectangular distribution is linearly related to the expanded measurement uncertainty ( a being the halfwidth of the rectangular distribution) p= U . 0.0 mm. the relevant coverage factor is thus k = 1. The example concerns the 150 mm calibration point for measurement of external dimensions. Several gauge blocks with nominal lengths in the range 0.g.05 mm (the main scale interval is 1 mm and the vernier scale interval 1/20 mm). The measurement range of the calliper is 150 mm. 150 mm) but give different values on the vernier scale (e.0064 V .8) of EALR2 finally gives the relation k ( p) = p 3 . 0.22 .150 mm are used in the calibration. 100 mm. 50 mm. (S9. quality of the measuring faces of the jaws (flatness.7) Solving this relation for the expanded measurement uncertainty U and inserting the result together with the expression of the standard measurement uncertainty related to a rectangular distribution as given by eq.6 mm. and function of the locking mechanism.
7 Correlation None of the input quantities are considered to be correlated to any significant extent. correction due to the finite resolution of the calliper. its influence is considered negligible for the present case. Additional effects may be caused by the fact that the measuring faces of the jaws are not exactly flat. S10. Abbe errors. flatness and parallelism errors of the measurement faces. are given in the calibration certificate. December 1999 rev00 Page 60 of 79 . This certificate confirms that the gauge blocks comply with the requirements for grade I gauge blocks according to ISO 3650.) S10. difference in temperature between the calliper and the gauge block.EA4/02 • Expression of the Uncertainty of Measurement in Calibration S10. length of the actual gauge block. average thermal expansion coefficient of the calliper and the gauge block. For the actual lengths of the gauge blocks their nominal lengths are used without correction.3 Working standards ( lS . Thus variations due to the finite resolution are estimated to have rectangular limits of ± 25 µm. not parallel to each other and not perpendicular to the beam.5·106 °C1.05 mm. To minimise effort. example S4. such as applied measurement force. Cf. S10. the temperatures of the calliper and the gauge block are equal within ±2 °C. equal to ±50 µm is considered.e. ∆t δ li X δ lM S10. α ) After an adequate stabilisation time. only the range of the total variation. i. that the central length of the gauge blocks coincides within ±0. taking the tolerance limits as the upper and lower limits of the interval of variability. correction due to mechanical effects.6 Mechanical effects ( δ lM ) These effects include the applied measurement force. (The uncertainty in the average thermal expansion coefficient and in the difference of the thermal expansion coefficients has not been taken into account.8 µm with the nominal length. EALR2S1. The average thermal expansion coefficient is 11. together with their associated expanded uncertainty of measurement. S10. the Abbe error and the play between the beam and the sliding jaw. nominal length of the actual gauge block. LS ) The lengths of the reference gauge blocks used as working standards.1) α li X lS LS  indication of the calliper.4 Temperature ( ∆t .2 The error of indication E X of the calliper at the reference temperature t 0 = 20° C is obtained from the relation: EX = li X − lS + LS ⋅ α ⋅ ∆t + δ li X + δ lM where: (S10.5 Resolution of the calliper ( δ li X ) The scale interval of the vernier scale is 0.
10 ± 0. the distribution is essentially trapezoidal.0 1. The final distribution is not normal but essentially trapezoidal with a ratio β = 0.83 appropriate for this trapezoidal distribution of values is calculated from eq.11 Reported result At 150 mm the error of indication of the calliper is (0.9 Uncertainty budget ( δ l X ) quantity estimate Xi li X lS ∆t δ li X xi 150. The result of measurement for the 150 mm gauge block is 150. The coverage factor k = 1. obtained by convolution of the two rectangular distributions associated with the dominant contributions.13. paragraph 5. In the sense that probabilities and probability densities in practice may only be determined to within 3 %− 5 %.15 K 15 µm 29 µm rectangular rectangular rectangular rectangular sensitivity coefficient ci 1.10 mm.06) mm.0 uncertainty contribution ui ( y ) 0. (S10.10 Expanded uncertainty The uncertainty of measurement associated with the result is clearly dominated by the combined effect of the measurement force and the finite resolution of the vernier.8 Measurements ( li X ) The measurement is repeated several times without detecting any scatter in the observations. The halfwidths of the base and the top of the resulting symmetrical trapezoid are 75 µm and 25 µm. December 1999 rev00 Page 61 of 79 .00 m 0 0 0 0.10) of the mathematical note S10.83 ⋅ 0.83 which has been derived from the assumed trapezoidal probability distribution for a coverage probability of 95 %. S10.0 1.33 of the halfwidth of the plateau region to the halfwidth of the variability interval. Annex E is not applicable.46 µm 2.033 mm ≅ 0.12 Additional remark The method used for calculating the coverage factor is clearly related to the fact that uncertainty of measurement associated with the result is dominated by two influences: the mechanical effects and the finite resolution of the vernier scale.10 mm 150.EA4/02 • Expression of the Uncertainty of Measurement in Calibration S10. Thus the assumption of a normal distribution for the output quantity is not justified and the conditions of EALR2.7 µMk1 1.46 µm 1.6 apply.10 mm standard probability uncertainty distribution u( x i ) 0.06 mm S10. Thus U = k ⋅ u ( E X ) = 1. Therefore the method of effective degrees of freedom described in EALR2. S10.0 µm 15 µm 29 µm 33 µm δ lM EX S10. The reported expanded uncertainty of measurement is stated as the standard uncertainty of measurement multiplied by the coverage factor k = 1. Thus uncertainty due to limited repeatability does not give a contribution.
6 0. for instance the terms with indices 1 and 2 . the method presented in S9. the distribution resulting from convolving them is a symmetrical trapezoidal distribution 1 0. 1: Unified symmetrical trapezoidal probability distribution with the value β=0.3) denotes the combined contribution of the two dominant terms and uR ( y ) = ∑u i =3 N 2 i (y) (S10.83 . S10.13 Mathematical note If the situation of measurement is such that two of the uncertainty contributions in the budget can be identified as dominant terms. 95 % of the area of the trapezoid is encompassed by an interval ±60 µm around its symmetry axis. The standard uncertainty to be associated with the measurement result y may be written in this case as 2 2 u( y ) = u0 ( y ) + uR ( y ) (S10. December 1999 rev00 probability density [p] Page 62 of 79 .8 0.4 0.2 0 2 1 0 1 2 normalized deviation [y/a] Fig. are combined into one dominant term.33 of the edge parameter. If the two dominant contributions arise from rectangular distributions of values with halfwidths a1 and a 2 . resulting from the convolution of two rectangular distributions.14 can be applied when the two dominant contributions.2) where 2 2 u 0 ( y ) = u1 ( y ) + u 2 ( y ) (S10.EA4/02 • Expression of the Uncertainty of Measurement in Calibration respectively. corresponding to k = 1.4) the total uncertainty contribution of the remaining nondominant terms.
respectively (see example in Fig.6) the dependence of the coverage factor on the coverage probability is derived according to the method sketched in S9. (S10.5) of the base and the top.9) Fig.6) is u 2 ( y) = a2 (1 + β 2 ) . (S10.7) The square of the standard measurement uncertainty deduced from the trapezoidal distribution of eq. 1).8) Using the distribution of eq.EA4/02 • Expression of the Uncertainty of Measurement in Calibration with halfwidths a = a1 + a 2 and b = a1 − a 2 (S10.14 p (1 + β ) 2 × 1 − (1 − p )(1 − β 2 ) p <β 2− p ⋅ p β≤ 2− p k ( p) = 1 1+ β 2 6 (S10. December 1999 rev00 Page 63 of 79 .6) with the edge parameter β= b a1 − a 2 = a a1 + a 2 (S10. 2 shows the dependence of the coverage factor k on the value of the edge parameter β for a coverage probability of 95 %. 6 (S10. The distribution may be conveniently expressed in the unified form 1 1 × ϕ ( y) = a (1 + β ) 1 − β 1 y < β ⋅a y 1 − β ⋅ a ≤ y ≤ a a 0 a< y (S10.
95 is calculated from the relation k= 1 − (1 − p)(1 − β 2 ) 1+ β 2 6 . is given by: t X = t S + δ t S + δ tD − δ tiX + δ tR + δ t A + δ tH + δ t V (S11. The example further demonstrates that there are two equivalent ways to tackle this problem: the direct assignment of a value to the indication of the instrument and the association of a correction to the indication.8 1 edge parameter [beta] Fig. used as a working standard. that has to be assigned as the temperature of the bore when the reading of the builtin temperature indicator is 180.10) S11 CALIBRATION OF A TEMPERATURE BLOCK CALIBRATOR AT A TEMPERATURE OF 180°C3 S11. The temperature t X . It has been included here.1 As part of a calibration. the temperature that has to be assigned to the calibration bore of a temperature block calibrator. This process is basic for calibrations in different metrological fields and.EA4/02 • Expression of the Uncertainty of Measurement in Calibration 2 coverage factor [k] 1.5 0 0.0 °C. usually called the error of indication.2 0. of general interest. in order to highligt how a value is assigned to an indication of an instrument in a calibration process.9 1.7 1. is measured. Calibration of temperature block calibrators.0 °C. by measuring the electrical resistance of the thermometer by an ac resistance bridge.6 1. (S10. 2: Dependence of the coverage factor k on the value of the edge parameter β of a trapezoidal distribution for a coverage probability of 95 %. in a dimplified form.8 1.1) 3 A similar example will be found in the EA guideline EA10/xx.4 0. Page 64 of 79 December 1999 rev00 . therefore.6 0. The coverage factor for a coverage probability of 95 % appropriate to a trapezoidal distribution with an edge parameter of β < 0. This is done when the indication of the builtin temperature indicator has stabilised at 180. The temperature of the calibration bore is determined by an inserted platinum resistance thermometer.
temperature correction due to the radial temperature difference between the builtin thermometer and the working standard. with an associated expanded uncertainty of measurement U = 30 mK (coverage factor k = 2 ). S11.temperature correction due to the settability limitations of the block temperature calibrator. The measured resistance value corresponds to a temperature of 180. . S11. December 1999 rev00 Page 65 of 79 . .1 °C.temperature variation within the time of measurement.temperature correction due to drift in the value of the working standard since its last calibration.temperature correction due to the ac resistance measurement.temperature correction due to hysteresis in the increasing and decreasing branches of the measuring cycle. S11.3 Determination of the temperature by resistance measurement( δ t S ) The temperature of the resistance thermometer used as working standard is determined as 180.1 °C. since the platinum resistance thermometer used as working standard has an outer diameter d ≤ 6 mm . .4 Drift of the temperature of the working standard ( δ t D ) From general experience with platinum resistance thermometers of the type used as working standard in the measurement. . the change of temperature due to resistance ageing since the last calibration of the standard is estimated to be within the limits ±40 mK.temperature correction due to the axial inhomogeneity of temperature in the measuring bore. S11. .1 K. This gives temperature resolution limits of ±50 mK within which the thermodynamic state of the temperature block can be uniquely set.temperature of the working standard derived from the ac resistance measurement. Note: If the indication of the builtin temperature indicator is not given in units of temperature the resolution limits must be converted into equivalent temperature values by multiplying the indication with the relevant instrument constant.2 Working standard ( t S ) The calibration certificate of the resistance thermometer used as working standard gives the relationship between resistance and temperature.5 Settability of the block temperature calibrator ( δ ti X ) The builtin controlling thermometer of the block temperature calibrator has a scale interval of 0. Prior investigations have shown that stem conduction effects can be neglected in this case. Temperature corrections due to stem conduction are not considered.EA4/02 • Expression of the Uncertainty of Measurement in Calibration where: tS δ tS δ tD δ ti X δ tR δ tA δ tH δ tV . The standard measurement uncertainty associated with the resistance measurement converted to temperature corresponds to u(δ t S ) = 10 mK . . .
0 1.1 °C Standard uncertainty u( x i ) 15 mK 10 mK 23 mK 29 mK 58 mK 144 mK 29 mK 17 mK Probability distribution normal normal rectangular rectangular rectangular rectangular rectangular rectangular Sensitivity coefficient ci 1.0 °C 0. S11.0 °C 0.0 °C 0.12 Uncertainty budget ( t X ) Quantity Estimate Xi tS δ tS δ tD δ ti X xi 180. S11.10 Correlations None of the input quantities are considered to be correlated to any significant extent.0 °C 0.0 1.8 Hysteresis effects ( δ t H ) From readings of the reference thermometer during measurement cycles of increasing and decreasing temperature.0 °C 0. the temperature deviation of the calibration bore due to hysteresis effect has been estimated to be within ±50 mK.0 1. S11.7 Axial inhomogeneity of temperature ( δ t A ) The temperature deviations due to axial inhomogeneity of temperature in the calibration bore have been estimated from readings for different immersion depths to be within ±250 mK.0 1.0 1.6 Radial inhomogeneity of temperature ( δ t R ) The radial temperature difference between the measuring bore and the builtin thermometer has been estimated to be within ±100 mK. S11.0 1.0 °C 180.0 1.0 Uncertainty contribution ui ( y ) 15 mK 10 mK 23 mK 29 mK 58 mK 144 mK 29 mK 17 mK 164 mK δ tR δ tA δ tH δ tV tX December 1999 rev00 Page 66 of 79 .0 °C 0.11 Repeated observations Due to the finite resolution of the indication of the builtin thermometer no scatter in the indicated values has been observed and taken into account.9 Temperature instability ( δ t V ) Temperature variations due to temperature instability during the measuring cycle of 30 min are estimated to be within ±30 mK.EA4/02 • Expression of the Uncertainty of Measurement in Calibration S11. S11. S11.1 °C 0.
13. (S11.0 °C is 180.3) using the definition of the error of indication in eq.81 ⋅ 164 mK ≅ 0.1 °C ± 0. According to S10. (S11. Model functions that can be transformed mathematically from one into the other represent the same measurement process.14 Reported result The temperature of the calibration bore that has to be assigned to an indication of the builtin controlling thermometer of 180. not contribute to the uncertainty of measurement associated with the error of indication u( E X ) = u( t X ) The model function of eq.3 K S11. The metrologist keeps it in his hands to choose the model that suits his habits and his approach to the problem.3) The indicated value t i is a nominal value. the coverage factor corresponding to the edge parameter β = 0. The final distribution is not normal but essentially trapezoidal.EA4/02 • Expression of the Uncertainty of Measurement in Calibration S11. (S11. (S11. December 1999 rev00 Page 67 of 79 . (S11. the problem can alternatively be formulated with the error of indication E X = t X − ti (S11.3 °C.15 Mathematical note concerning the model Some metrologists are confused that the indication of the controlling thermometer does not appear explicitly in the model function of eq.13 Expanded uncertainty The standard uncertainty of measurement associated with the result is clearly dominated by the effect of the unknown temperature correction due to the axial temperature inhomogeneity in the measuring bore and the radial temperature difference between the builtin thermometer and the working standard.43 is k = 1.81 which has been derived from the assumed trapezoidal probability distribution for a coverage probability of 95 %. To fit their needs. S11.2) of the builtin temperature indicator E X = t S − ti + δ t S + δ tD − δ tiX + δ tR + δ t A + δ tH + δ t V (S11. U = k ⋅ u (tX ) = 1. however.2).1).1) can be regained from eq. Its effect is to shift the scale of the measurand.4) This note shows that there is not necessarily only one unique way to choose the model of evaluation of measurement. model functions that are connected by linear scale transformations may serve as equivalent expressions of the measurement problem. For cases where a continuous scale of indication is involved. It does.8 1. as in the calibration of the temperature block under consideration. The reported expanded uncertainty of measurement is stated as the standard uncertainty of measurement multiplied by the coverage factor k = 1.
indication of the meter at the beginning of the measurement and at the end of the measurement. S12.corrections due to the finite resolution of the meter indication. pressure pX and temperature t X . The collecting tank has a narrow neck with an attached scale by which the filling level can be detected. Vi X 2 δ ViX1 . Vi S δViS αS tS t0 αW κW pS pX tX December 1999 rev00 Page 68 of 79 .correction of the volume indicated at the neck scale of the collecting tank due to the finite resolution of the scale. a value typical for municipal tap water systems.compressibility of water. The temperature and pressure of the water at the meter. The measurement is made using a test rig that supplies necessary water flow with a pressure of approximately 500 kPa.1 The calibration of a water meter involves the determination of the relative error of indication within the applicable flow range of the meter.pressure in the collecting tank (it is zero if excess pressure is considered) . and the temperature of the water in the collecting tank.difference in meter indications. . It is empty but wetted at the beginning of the measurement.temperature of the collecting tank. . . The meter to be calibrated is connected between these tanks.cubic thermal expansion coefficient of the material of the collecting tank. .1) V X = (Vi S + δ Vi S )(1 + α S (t S − t 0 ) )(1 + α W (t X − t S ) )(1 − κ W ( p X − pS ) ) ∆Vi X = Vi X 2 − Vi X 1 (S12. It has a mechanical counter with pointers.e. . . . i.2 The relative error of indication eX in a single run is defined as ∆Vi X + δ Vi X2 − δ Vi X1 eX = −1 VX with (S12.pressure of the water at the inlet of the meter. The water is received in an open collecting tank that has been calibrated and determines the reference volume of the water.reference temperature at which the collecting tank has been calibrated.temperature of the water at the inlet of the meter. . The indication of the meter is recorded at the beginning and at the end of the measurement.volume that passed the meter during the measurement under the prevailing conditions. are recorded as well.EA4/02 • Expression of the Uncertainty of Measurement in Calibration S12 CALIBRATION OF A HOUSEHOLD WATER METER S12. The level is recorded in the collecting tank at the end of the measurement.volume indicated at the neck scale of the collecting tank at the end of the measurement. . δ ViX2 VX . The measurement is done at a flow rate of 2500 l/h with standing startandstop which means that the flow rate is zero both at the beginning and the end of the measurement. .2) where: Vi X 1 . at the inlet of the meter.cubic thermal expansion coefficient of water. . .
46 ⋅ 10 −6 kPa −1 in the temperature interval considered. Since there is no uncertainty statement accompanying this value it is assumed to be known to within its least significant digit.15 ⋅ 10 −3 K −1 in the temperature interval considered. Unknown deviations are considered to be within the rounding limits of ± 0. p S . The stated limits cover all possible sources of uncertainty.4 Resolution of the collecting tank scale ( δViS ) The water level of the collecting tank can be determined to within ±1 mm. The cubic thermal expansion coefficient of the tank material (steel) is taken from a material handbook to be a constant equal to α S = 51 ⋅ 10 −6 K −1 in the temperature interval considered. resolution in reading and temperature gradients in the tank. Unknown deviations are considered to be within the rounding limits of ± 0. The compressibility of water is taken from a material handbook to be a constant equal to κ W = 0. The cubic expansion coefficient of water is taken from a material handbook to be a constant equal to α W = 0.2 l ( k = 2 ).02 l. p X ) The excess pressure of the water supplied to the inlet of the meter is 500 kPa with relative deviations not larger than ±10 %. On its way from the inlet to the collecting tank. the water expands to excess pressure 0 kPa (atmospheric pressure condition). such as calibration of temperature sensors.5 ⋅ 10 −6 K −1 . resolution in reading and small temperature changes during one measurement run.8 Correlation None of the input quantities are considered to be correlated to any significant extent. t 0 ) The calibration certificate states that the neck scale indicates the volume of 200 l at the reference temperature t 0 = 20 °C with an associated relative expanded measurement uncertainty of 0. With the scale factor of the tank of 0.7 Pressure difference of the water between the meter and the tank ( κ W .5 Temperature of the water and the collecting tank ( α S . S12. it is assumed to be known to within its least significant digit. S12.6 Temperature of the water at the meter ( α W . t X ) The temperature of the water at the inlet of the meter is determined to be 16 °C within ±2 K. Since there is no uncertainty statement accompanying this value. The expanded measurement uncertainty associated with the value is 0.EA4/02 • Expression of the Uncertainty of Measurement in Calibration S12. Unknown deviations are considered to be within the rounding limits of ± 0.005 ⋅ 10 −6 kPa −1 . t S ) The temperature of the water in the collecting tank is determined to be 15 °C within ±2 K. S12. December 1999 rev00 Page 69 of 79 . S12.02 l/mm the maximum deviation of the volume of water in the collecting tank from the observed indicated value is estimated to be within ±0. such as contributions from calibration of sensors.5 ⋅ 10 −6 K −1 . Since there is no uncertainty statement accompanying this value it is assumed to be known to within its least significant digit. S12.1 % ( k = 2 ). The stated limits cover all possible sources of uncertainty.3 Collecting tank ( Vi S .
95 l 0.10 Indication of the meter ( ∆Vi X .0300 l⋅K1 rectangular rectangular 100 l⋅kPa 9. δ Vi X2 ) The water meter to be calibrated has a resolution of 0.15 K 29 kPa  1.058 l 0.2 l resulting in the limits ± 0.46⋅106 kPa1 2.29⋅103 l 0.109 l δ Vi S αS tS rectangular 0.9 Uncertainty budget ( V X ) quantity Xi estimate xi standard uncertainty u( xi ) probability distribution normal rectangular rectangular sensitivity coefficient ci uncertainty contribution ui ( y) Vi S 200.02 l 0.0 l 199. δ Vi X1 .0 l 51⋅106 K1 15°C 0.058 l 0.1 l in both readings for the maximum deviations resulting from the meter resolution.0 1000 l⋅K 0.0⋅103 5.0⋅10 3 0.0 1.11 Uncertainty budget ( e X ) quantity Xi estimate xi standard uncertainty u( xi ) probability distribution nominal rectangular rectangular rectangular sensitivity coefficient ci uncertainty contribution ui ( y) ∆Vi X 200.9⋅106 K1 1.58⋅103 l 0. S12.0115 l 0.29⋅103 l 0.0 l 0.29⋅103 l 0.0228 l 0.9⋅106 kPa1 pS VX The standard uncertainty of measurement associated with the result is clearly dominated by the volume indication at the neck scale of the collecting tank.EA4/02 • Expression of the Uncertainty of Measurement in Calibration S12.15⋅103 K1 16°C 500 kPa 0.2⋅106 l⋅kPa1  αW tX κW pX 0. S12.0115 l 0.0346 l 0.0027 l 0.109 l 5.15 K 2.55⋅103 l 0.29⋅106 K1 1.0198 l⋅K1 rectangular 200 l⋅K rectangular 0. This must be kept in mind in the further processing of the uncertainty evaluation.95 l 0.0⋅103 December 1999 rev00 Page 70 of 79 .0 Pa 199.10 l 0. The final distribution is not normal but essentially rectangular.0 l 0.68⋅103 δ ViX1 δ ViX2 VX eX 5.29⋅103 l 0.10 l 0.000 3 0.
3) .60⋅103 0.000 3 0.relative error of indication of a single run.EA4/02 • Expression of the Uncertainty of Measurement in Calibration S12.68⋅103 0. .001 3 = 0.000 5 0.0 0. The results of these three runs are treated as independent observations e X j in the model that determines the average error of indication e X av : e X av = e X + δ e X where: eX δ eX (S12.60⋅103 0.0 0.12 Repeatability of the meter The relative error of indication of the water meter to be calibrated. For that reason the relative error of indication is determined three times.001 0. determined at the same flow rate of 2500 l/h.002 2 e X = 0.001 s( e X j ) = 0.000 60 quantity estimate standard degrees of probability sensitivity uncertainty uncertainty freedom distribution coefficient contribution u( xi ) Xi eX xi ν eff 2 ci ui ( y) 0.91⋅103 δ eX e X av ∞ 10 December 1999 rev00 Page 71 of 79 .68⋅103 normal normal 1.001 0.correction of the relative error of indication obtained in the different runs due to the lack of repeatability of the meter. shows considerable scatter.14 Uncertainty budget ( e X av ) 0.001 u ( e X ) = s( e X ) = 0. observed relative error of indication 1 2 3 arithmetic mean: experimental standard deviation: standard uncertainty: S12.13 Measurements ( e X ) No.0 1. S12.
EA4/02 • Expression of the Uncertainty of Measurement in Calibration S12.28 . Care is taken to ensure that the rings and the line scale (ruler) maintain the monitored temperature throughout the calibration. The reported expanded uncertainty of measurement is stated as the standard uncertainty of measurement multiplied by the coverage factor k = 2.001 ± 0. which for a tdistribution with ν eff = 10 effective degrees of freedom corresponds to a coverage probability of approximately 95 %.16 Reported result The average relative error of indication of the water meter determined at a flow rate of 2500 l/h is 0.2 The diameter d X of the ring to be calibrated at the reference temperature t 0 = 20 °C is obtained from the relationship: d X = d S + ∆l + δ li + δ l T + δ l P + δ l E + δ l A (S13. respectively.1) December 1999 rev00 Page 72 of 79 .002. The temperature in the comparator working volume is maintained at 20 °C within ±0. whose nominal inner diameter (Ds = 40 mm) differs significantly from that of the ring to be calibrated.1 A steel ring gauge of Dx = 90 mm nominal inner diameter is calibrated applying the procedure introduced in EALG29. The Cshaped arms are supplied with spherical contact tips.15 Expanded uncertainty Because of the small number of effective degrees of freedom of the standard uncertainty associated with the mean relative error of indication the standard coverage factor has to be modified according to table E1 U = k ⋅ u ( e X av ) = 2.5 K. which includes all position elements for aligning the test pieces. fixed on the steady and the measuring spindle. are employed.28 ⋅ 0.91 ⋅ 10 −3 ≅ 2 ⋅ 10 −3 S12. The rings are contacted at several points diametrically apart by two Cshaped arms. The measuring force is generated by a tension weight ensuring a constant force of nominally 1.1 µm. S13 CALIBRATION OF A RING GAUGE WITH A NOMINAL DIAMETER OF 90 MM S13.5 N over the whole measuring range. S13. A length comparator of the Abbe type and a steel setting ring. The rings are gently clamped sequentially on a 4degrees of freedom table. In this case the length comparator and the steel setting ring both take the role of working standards. The line scale of the comparator has been verified periodically to meet the manufacturer’s specification of maximum permissible error. The ambient temperature is monitored in order to maintain the environmental conditions stated by calibration procedure. The measuring spindle is rigidly connected with the gauge head of a steel line scale of resolution 0.
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
where:
dS ∆l
δ li δ lT
δ lP δ lE
δ lA
 diameter of the reference setting ring at the reference temperature,  observed difference in displacement of the measuring spindle when the contact tips touch the inner surface of the rings at two diametrically apart points,  correction for the errors of indication of the comparator,  correction due to the temperature effects of the ring to be calibrated, the reference setting ring and the comparator line scale,  correction due to coaxial misalignment of the probes with respect to the measuring line,  correction due to the difference in elastic deformations of the ring to be calibrated and the reference setting ring,  correction due to the difference of the Abbe errors of the comparator when the diameters of the ring to be calibrated and the reference setting ring are measured.
S13.3 Working standard ( d S ) The inner diameter of the setting ring used as the working standard together with the associated expanded uncertainty of measurement is given in the calibration certificate as 40,0007 mm ± 0,2 µm (coverage factor k = 2). S13.4 Comparator ( δ li ) The corrections for the errors of indication of the line scale (ruler) were determined by the manufacturer and prestored electronically. Any residuals are within the manufacturers specification of ± (0,3 µm + 1,5 ⋅ 10 −6 ⋅ li ) with li being the indicated length. The specifications are ascertained by periodical verifications. For the actual length difference DX − DS = 50 mm unknown residuals are estimated to be within ± (0,375) µm. S13.5 Temperature corrections ( δ l T ) Throughout the measurement care is taken to ensure that the ring to be calibrated, the setting ring and the comparator scale maintain the monitored temperature. From previous measurements and general experience with the measurement system it can be ascertained that the deviations of temperatures of the ring to be calibrated, the setting ring and the comparator scale from ambient temperature stay within ±0,2 K. The ambient temperature of the measuring room, however, is estimated to be within ±0,5 K. The knowledge on the measurement, therefore, is best described by the deviation of the ambient temperature from the reference temperature and the deviations of the temperatures of the ring to be calibrated, the setting ring and the comparator scale (ruler) from the ambient temperature. The correction δ l T due to temperature influences is determined from the model:
δ l T = (DS ⋅ (α S − α R ) − DX ⋅ (α X − α X ) ) ⋅ ∆t A + DS ⋅ α S ⋅ δ t S − DX ⋅ α X ⋅ δ t X − ( DS − DX ) ⋅ α R ⋅ δ t R
(S13.2)
December 1999 rev00
Page 73 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
where:
DX , DS
αX ,αS ,αR
∆t A = t A − t 0
δ t X , δ t S , δ tR
 nominal diameters of the ring to be calibrated and the reference setting ring,  linear thermal expansion coefficients of the ring to be calibrated, the reference setting ring and the comparator line scale (ruler),  deviations of the ambient temperature of the measuring room from the reference temperature t 0 = 20° C ,  deviations of the temperature of the ring to be calibrated, the reference setting ring and the comparator line scale (ruler) from ambient temperature
Since the expectations of the four temperature differences entering eq. (S13.2) are zero, the usual linearized version will not include effects of the measurement uncertainty associated with the values of the three linear thermal expansion coefficients. As depicted in section S4.13 the nonlinear version has to be used to determine the standard uncertainty associated with the four product terms:
δ l TA δ l TS δ l TX δ l TR
= ( DS ⋅ (α S − α R ) − DX ⋅ (α X − α R )) ⋅ ∆t A = DS ⋅ α S ⋅ δ t S = DX ⋅ α X ⋅ δ t X = ( DS − DX ) ⋅ α R ⋅ δ tR (S13.3)
Based on the calibration certificate of the setting ring, on the manufacturer’s data for the ring to be calibrated and the comparator scale, the linear thermal expansion coefficients are assumed to be within the interval (11,5 ± 1,0) 106 °K1. Using this value and the limits of temperature variation stated at the beginning, the standard uncertainties associated with the four product terms are u (δ l TA ) = 0,012 µm , u(δ l TS ) = 0,053 µm , u (δ l TX ) = 0,12 µm and u (δ l TR ) = 0,066 µm . The standard uncertainty associated with the combined temperature corrections is derived from these values with the use of the following uncertainty subbudget: quantity
Xi
estimate
xi
standard uncertainty
u( xi )
probability distribution 
sensitivity coefficient
ci
uncertainty contribution
ui ( y)
δ l TA δ l TS δ l TX δ l TR δ lT
0,0 µm 0,0 µm 0,0 µm 0,0 µm 0,0 µm
0,012 µm 0,053 µm 0,12 µm 0,066 µm
1,0 1,0 1,0 1,0
0,012 µm 0,053 µm 0,12 µm 0,066 µm 0,15 µm
December 1999 rev00
Page 74 of 79
EA4/02 • Expression of the Uncertainty of Measurement in Calibration
S13.6 Coaxiality correction ( δ lP ) The deviation from coaxiality of the two spherical probes and the measuring line is assumed to be within ±20 µm. Using the equations stated in the mathematical note (S13.13) the correction due to possible noncoaxiality and the associated standard uncertainty of measurement is given by
δ lP = 2 ⋅
1 1 2 ⋅ u (δ c ) − D X DS 16 1 1 ⋅ 2 + 2 ⋅ u 4 (δ c ) D 5 X DS
(S13.4)
u 2 (δ lP ) =
(S13.5)
Here δ c is the small distance of the measured cord from the centre of the ring. The values resulting for the correction and the associated standard measurement uncertainty are δ lP ≅ −0,004 µm and u (δ l P ) ≅ 0,0065 µm . As can been seen from the uncertainty budget (S13.10), these values are two orders of magnitude smaller than the remaining uncertainty contributions so that their influence need not be taken into account under the current measurement conditions.
S13.7 Elastic deformation correction ( δ l E ) The elastic deformation of the ring to be calibrated or the reference setting ring are not determined during the current measurement. From previous experience, however, the effects resulting from elastic deformations are estimated to be within ± 0,03 µm. S13.8 Abbe error correction ( δ l A ) The actual values of the Abbe errors of the comparator are not determined during the current measurement. From experience and periodical verification data of the comparator, however, the effects due to Abbe errors are estimated to be within ± 0,02 µm. S13.9 Measurements ( ∆l ) The following observations are made of the inner diameter of the unknown and the setting ring: No 1 Object reference setting ring Observation 0 during this step the comparator display is zeroed 49,99935 mm Measurand diameter in the nominal direction of the symmetry plane orthogonal to the cylinder axis
2 3
ring to be calibrated ring to be calibrated
49,99911 mm
diameter in the nominal direction of the symmetry plane orthogonal to the cylinder axis diameter in the symmetry plane orthogonal to the cylinder axis rotated around the axis with respect to the nominal direction by +1 mm on the circumference
December 1999 rev00
Page 75 of 79
33 µm standard deviation of the mean: s ( ∆l ) = s( ∆l ) 5 = 0.EA4/02 • Expression of the Uncertainty of Measurement in Calibration No 4 Object ring to be calibrated Observation 49.99996 mm Measurand diameter in the symmetry plane orthogonal to the cylinder axis rotated around the axis with respect to the nominal direction by −1 mm on the circumference. This is deduced from the pooled estimate of the standard deviation s p (0) = 0.15 µm The standard deviation of a single observation s (∆l ) = 0.99972 mm 5 ring to be calibrated ring to be calibrated 49. The resulting standard measurement uncertainty to be associated with the observed diameter difference is: 2 u( ∆l ) = s 2 ( ∆l ) + s p (0) = 0.30 µm December 1999 rev00 Page 76 of 79 .18 µm takes into account effects due to form deviations of the ring to be calibrated as well as due to the repeatability of the comparator. diameter in the nominal direction translated to the plane parallel to symmetry plane orthogonal to the cylinder axis by 1 mm upwards diameter in the nominal direction translated to the plane parallel to symmetry plane orthogonal to the cylinder axis by 1 mm downwards The observations may be divided into two groups: the observation of the diameter of the setting ring (observation no 1) that is used to set the comparator display to zero and the observation of the diameter of the ring to be calibrated (observations no 2 to no 6) that give the difference in diameters: arithmetic mean: ∆l = 49. the uncertainty resulting from the zeroing of the comparator display must also be taken into account.999 54 mm standard deviation of a single observation: s( ∆l ) = 0.99954 mm 6 49. To obtain the standard uncertainty of measurement to be associated with the observed mean difference of the diameters.25 µm obtained in a prior measurement under the same conditions of measurement.
The reported expanded uncertainty of measurement is stated as the standard uncertainty of measurement multiplied by the coverage factor k = 2 which for a normal distribution corresponds to a coverage probability of approximately 95%.10 µm 0.000 3 ± 0.433 µm δ li δ lT δ lP δ lE δ lA dX 0.012 µm 0.10 Uncertainty budget ( d X ) quantity Xi estimate xi standard probability sensitivity uncertainty distribution coefficient u( xi ) ci uncertainty contribution ui ( y) dS ∆l 40.7) December 1999 rev00 Page 77 of 79 .0 1.12 Reported result The diameter of the ring gauge is (90.999 55 mm 0.000 7 mm 49.018 µm 0.0 mm 0.018 µm 0.0065 µm 0.0 1.9 µm S13. This angle is related on the other hand to the small distance δ c of the chord form the centre of the ring by δc = 1 1 ⋅ d ⋅ sin(δ ϕ ) ≅ ⋅ d ⋅ δ ϕ 2 2 (S13.0 1.0 mm 0.13 Mathematical note on noncoaxiality Since it is not possible to make an exact adjustment of the rings with respect to the measuring axis of the comparator.22 µm 0.30 µm 0. S13.0 mm 0. which is observed in the measurement.0065 µm rectangular 0.0 0.0 1.15 µm 0. is related to the diameter of the ring d by 1 d ' = d ⋅ cos(δ ϕ ) ≅ d ⋅ 1 − (δ ϕ ) 2 2 (S13.0 1.0 1.11 Expanded uncertainty U = k ⋅ u ( d X ) = 2 ⋅ 0.433 µm ≅ 0.000 9) mm.10 µm 0. The length d ' of this chord.012 µm normal normal rectangular normal 1.22 µm 0. the quantity determined in the measurement is a chord of the respective ring in the proximity of its diameter.000 25 mm rectangular rectangular S13.0 mm 90.6) where δ ϕ is the small angle that complements half of the central angle of the chord to π / 2 .30 µm 0.EA4/02 • Expression of the Uncertainty of Measurement in Calibration S13.000 004 mm 0.15 µm 0.
It takes the value α = 9 / 5 if δ c is assumed to be rectangularly distributed so that in this case the standard measurement uncertainty to be associated with the diameter is expressed by 16 u 4 (δ c ) u (d ) = u (d ' ) + ⋅ 5 D2 2 2 (S13. This ratio depends on the distribution that is assumed for δ c . (S13. d ' and δ c in eq. u 2 ( d ) = u 2 ( d ' ) + 4 ⋅ (α − 1) with u 4 (δ c ) D2 (S13. (S13. The best estimate of the diameter is obtained by taking the expectation of the last relation to be u 2 (δ c ) d = d ' +2 .12) December 1999 rev00 Page 78 of 79 .6) may be rewritten as (δ c) 2 d'≅ d − 2 D (S13. (S13. It must also be kept in mind that the meaning of d .9) is not identical. Since the variance of a random variable equals the expectation of the square of its deviation from the respective expectation.8) and eq. (S13. the square of the standard measurement uncertainty to be associated with the diameter of the ring is. (S13.10) α= m4 (δ c) 2 m2 (δ c) (S13. in eq. according to eq.9) they stand for the expectations of these quantities.8). whereas in eq.9) D Here it has been taken into account that the small distance δ c has zero expectation.11) being the ratio of the 4th order centred moment to the square of its 2nd order centred moment of the small distance δ c .8) these symbols represent the notexactly known quantities or random variables. (S13.8) where the diameter d of the ring in the ratio has been replaced by its nominal diameter D since the nominator of the ratio is a small quantity already. (S13.EA4/02 • Expression of the Uncertainty of Measurement in Calibration so that eq.
This action might not be possible to undo. Are you sure you want to continue?
Use one of your book credits to continue reading from where you left off, or restart the preview.