You are on page 1of 20

This publication contains policy, guidance and requirements

applicable to NAMAS testing laboratories

NIS81
EDITION 1 • MAY 1994

The Treatment of Uncertainty


in EMC Measurements

Contents
Section Page

1 Introduction 2
2 Concept of uncertainty 2
3 Steps in establishing an uncertainty budget 3
4 Compliance with specification 8
5 References 10
Appendix I Summary of equations used 11
Appendix II Examples of uncertainty budgets 12
1 Radiated emissions 12
2 Conducted emissions 14
3 Radiated immunity 15
Appendix III Calculation of hp 17
Appendix IV Calculation of uncertainty in logarithmic or
linear quantities 18
© Crown Copyright 1994

NAMAS Executive, National Physical Laboratory, Teddlngton, Middlesex; TW11 OLW j England
Tel: 081';;9437140 Fax: 081-943 7134 Telex: 262344 N~L G

EDITION 1 • MAY 1994 PAGE 1 OF 18


NIS 81 • TREATMENT OF UNCERTAINTY IN EMC MEASUREMENTS

1 Introduction
1. 1 The general requirements for the estimation and reporting ofuncertainty are
given in the NAMAS Accreditation Standard, M10. General gUidance on the
estimation and reporting of uncertainties in testing is given in NAMAS
publication, NISBO. This publication provides specific gUidance on the
application of the principles set out in NISBO for laboratories seeking or
holding NAMAS accreditation for EMC testing.

1. 2 It is recommended that laboratories follow the methods for estimating


uncertainty described in NISBO and in this publication except where the
method is defined in the test specification.

1.3 It should be noted that the small differences in uncertainty estimates


obtained using the methods given in other, similar, documents [3], [4], [5]
and [6] are not considered to be significant.

lA This publication does not attempt to define what the uncertainty


contributions are, or what they should be, since these are dependent on the
equipment used and the method of test. However, examples of uncertainty
budgets are given in Appendix 11 for some common EMC measurements and
have been made as realistic as possible.

2 Concept of uncertainty
2.1 When a measurement is made the result will be different from the true or
theoretically correct value. This difference is the result of an error in the
measured value and it should be the aim of the measurement process to
minimise this error. In practice the extent to which this can be achieved may
be limited and a statement of uncertainty is used to reflect the
quality/ accuracy of the measured result as compared with the true value.
A statement of uncertainty is incomplete without an accompanying
statement of the confidence that can be placed in the value of the
uncertainty.

2.2 Uncertainties arise from random effects and from imperfect correction for
systematic effects. The recommendations ofthe InteITlational Committee for
Weights and Measures (CIPM)[l], which will be followed by NAMAS, are that
uncertainty components be grouped into two categories, based on their
method of evaluation. These categories are referred to as Type A and Type B.

2.3 Type A evaluation is by calculation from a series of repeated observations


and therefore includes random effects. The statistically-estimated standard
deviation is sometimes called a Type A standard uncertainty for convenience.
Type B evaluation is by means other than Type A. For example, by
judgement based on data in calibration certificates, previous measurement
data, experience with the behaviour of the instruments, manufacturers'
specifications and all other relevant information. This category includes
uncertainties arising from systematic effects. The components evaluated by

PAGE 2 OF 18 EDITION 1 • MAY 1994


NIS 81 • TREATMENT OF UNCERTAINTY IN EMC MEASUREMENTS

Type A and Type B methods are combined together to produce an overall


value of uncertainty.

2.4 NIS8l. in line with other gUidance documents on uncertainty, recommends


that the reported uncertainty is calculated from the root sum of squares of
the standard deviations of the individual components multiplied by a
coverage factor, k, of 2, which approximates to a level of confidence of 95%.
If a higher level of confidence is required then k = 3 (CL of 99.7%) can be
used.

3 Steps in establishing an uncertainty budget


3.1 Decide on the range of measurement to which the budget will apply.

An uncertainty budget is a list of the probable sources of error with an


estimation of their uncertainty limits and probability distribution. It is likely
that some uncertainty contributions will not be the same for the complete
range of the measurement and a decision has to be made about the
breakdown that will be most appropriate. A single budget covering the
complete range may mean that a larger uncertainty is assigned than is
strictly necessary. However, this may be preferable in some cases where it
is not necessary to over complicate the calculation and reporting process.
Priority should be given to calculating the uncertainty in the region of the
test specification limit, or limits.

3.2 Type A evaluation of uncertainty components.

3.2.1 Random effects result in errors that vary in an unpredictable way while the
measurement is being made or is repeated under the same conditions. The
uncertainty associated with these contributions can be evaluated by
statistical techniques from repeated measurements. An estimate of the
standard deviation, s(qJ, of a series of n readings, qk' is obtained from:

(1)
1 ~ -2
( Lt (qk-q)
--1)
n k-)

where q is the mean value of n measurements.


3.2.2 The random component of uncertainty can be reduced by making repeat
measurements in the process of testing the equipment under test (EUT).
This yields the standard deviation of the mean, s( ij). given by:

s(q) (2)

EDITION 1 • MAY 1994 PAGE 3 OF 18


NIS 81 • TREATMENT OF UNCERTAINTY IN EMC MEASUREMENTS

3.2.3 Practical considerations will normally mean that the number of repeat
readings will be very small and will often be limited to only a single reading.
It is satisfactory to use a predetermination of S(qk) for the measurement
system, based on a larger number of repeats, provided the system, method,
configuration and conditions etc. are truly representative of the test.
However, such a predetermination will not include the contributions of the
particular EUT. The value of n to be used to obtain s( q) under these
circumstances is the number of measurements made in the process of
testing and not the number of measurements made in the predetermination.
Repeat measurements should be undertaken when the measured result is
close to the specification limit.

3.2.5 A value for the random contributions of the measurement system is in any
case an essential part of the uncertainty assessment and a type A evaluation
should be made on the 'typical' processes and configuration involved in the
test. For example, in the case of open site measurements, the type A
evaluation could include reconnecting the antenna and receiver and
adjusting the antenna height to maximise the receiver reading.

3.2.6 The standard uncertainty, U(xt) , of an estimate XI of an input quantity q,


based on a type A evaluation is therefore:
(3)

3.3 Type B evaluation: list all the other signtficant contributions to


uncertainty with an estimation of their limit value.

3.3.1 Contributions to uncertainty arising from systematic effects are those that
remain constant while the measurement is being made but can change if the
measurement conditions, method or equipment is altered. If there is any
doubt about whether a contribution is significant it should be included in
the uncertainty budget in order to demonstrate that it has been considered.

3.3.2 Normally, all corrections that can be applied to the measured result should
be applied. However, in some cases it may be impractical or unnecessary to
correct for all known errors. For example. the calibration certificate for an
EMC receiver may give actual measured input results at specific readings.
with an associated uncertainty. It is possible to correct subsequent readings
by using this calibration to achieve the lowest possible uncertainty. However.
it is more practical to use indicated values with no corrections applied, in
which case the manufacturer's specified uncertainty should be used,
provided it has been confirmed by an accredited calibration or, where this
is not obtainable, a route acceptable to NAMAS.

3.3.3 The individual uncertainty contributions should be in terms of the variation


in the quantity being measured, rather than the influence quantity, and all
in the same units. Most EMC measurements are derived from readings using
logarithmic scales (eg dBllV) , corrections for the gain or loss of system
components are in dB, specification limits are generally given in dB and

PAGE 4 OF 18 EDITION 1. MAY 1994


NIS 81 • TREATMENT OF UNCERTAINTY IN EMC MEASUREMENTS

instrument specification limits are normally in dB. In these cases it is


recommended that the uncertainty calculations are made in terms of dBs.
In some cases, for example, where the addition of signals is the dominant
contribution it may be more correct to calculate the uncertainty in terms of
absolute values, eg Vfm. The use of dB, percentages or absolute values is
discussed in Appendix IV

3.3.4 It is relatively straightforward to assign a value to the uncertainty


contribution when there is already evidence on which to base the value,
such as a calibration certificate or manufacturer's specification. In other
cases there may be little or no data available and an estimation has to be
made based on experience or on other relevant published material. In such
cases it is generally safer to overestimate the size of a contribution until
more substantial evidence is available.

3.3.5 Most contributions to uncertainty can be adequately represented by a


symmetrical distribution about the nominal or measured result, for instance
the uncertainty attributed to a receiver. However, some contributions are not
symmetrical and these are most simply dealt with by calculating separate
positive and negative values for the total uncertainty. The decision on
whether this is appropriate will depend on the difference between the two
values and the need for rigour in the uncertainty estimation. An example of
an asymmetric uncertainty is the addition of two signals at the same
frequency where the resultant is dependent on their relative phase, as
occurs with multiple reflections in a screened room and mismatch
uncertainty.

3.3.6 The basis of the RSS approach relies upon uncorrelated contributions. The
judicious selection of test equipment and measurement method can ensure
that adverse correlation between individual contributions is avoided or
minimised. If adverse correlation between any contributions is known or
suspected then the most straightforward approach is to sum the standard
uncertainty of these contributions arithmetically. In some situations it is
necessary to use the same items of test equipment for different steps in the
measurement process. For example, in the pre-calibration for radiated
immunity measurements it is essential that the same transmit antenna is
used for the calibration and testing.

3.4 Assign a probability distribution and detennine the standard


uncertainty of each contribution.

The probability distribution of an uncertainty describes the variation in


probability of the true value lying at any particular difference from the
measured or assigned result. The form of the probability distribution will not
necessarily be a regular geometric shape and an assumption has to be
made, based on prior knowledge or theory, that it approximates to one of the
common forms. It is then possible to calculate the standard uncertainty,
u(x;) , for the assigned form from simple equations. The three main
distributions of interest to EMC measurement are normal, rectangular and
U shaped.

EDITION 1 • MAY 1994 PAGE 5 OF 18


NIS 81 • TREATMENT OF UNCERTAINTY IN EMC MEASUREMENTS

3.4. I Normal:

This distribution can be assigned to uncertainties derived from multiple


contributions. For example, when a NAMAS calibration laboratory provides
a total uncertainty for an instrument this will have been calculated at a
minimum level of confidence of 95% and can be assumed to be normal. The
standard uncertainty of a contribution to uncertainty with assumed normal
distribution is found by dividing the uncertainty by the coverage factor, k,
appropriate to the stated level of confidence.

uncertainty
For Normal Distributions: utXI) = ---:---..::..
k
(4)

where k = 2 if the reported level of confidence is 95%. (Strictly speaking for


a level of confidence of 95%, k = 1.96, however, the difference this makes to
the combined uncertainty is not significant)

3.4.2 Rectangular:

This distribution means that there is equal probability ofthe true value lying
anywhere between the prescribed limits. A rectangular distribution should
be assigned where a manufacturer's specification limits are used as the
uncertainty, unless there is a statement of confidence associated with the
specification, in which case a normal distribution can be assumed.

For Rectangular Distributions: u(xt) (5)

where a l is the semi-range limit value of the individual uncertainty


contribution.

3.4.3 V Shaped:

This distribution is applicable to mismatch uncertainty[7). The value of the


limit for the mismatch uncertainty, M, associated with the power transfer at
a junction is obtained from 2010glO(l±lrallrLlldB, or
100(( 1 ± 1 rail r L1)2 - 1) % where r a and r L are the reflection coefficients for
the source and load. As stated in para 3.3.5, mismatch uncertainty is
asymmetric about the measured result, however, the difference this makes
to the total uncertainty is often insignificant and it is acceptable to use the
larger of the two limits ie 2010glO(l - Irail r LIl

M
For V-Shaped Distributions: (6)
.fi
PAGE 6 OF 18 EDITION 1 • MAY 1994
NIS 81 • TREATMENT OF UNCERTAINTY IN EMC MEASUREMENTS

3.5 Detennine the combined standard uncertainty.

The combined uncertainty, Ue(y), is obtained by taking the square root of the
sum of squares of the individual standard uncertainties. If any of the
standard uncertainties are not already in tenns of the measured quantity
then they should be converted using the appropriate functional relationship.
Cl' for example, the uncertainty in the measurement distance on an open
site should be converted to the uncertainty in the received signal strength,
then:

(7)

Any contributions with known or suspected adverse correlation should be


added together, then for m contributions:

(8)

3.6 Detennine the expanded uncertainty.

The expanded uncertainty, U, defines an interval about the measured result


that will encompass the true value with a specified level of confidence. p%.
The expanded uncertainty is obtained by multiplying the combined standard
uncertainty by a coverage factor, k, thus:
U = kuclY) (9)

The level of confidence recommended by NAMAS for EMC testing is 95%


which can be obtained with k = 2. However, if random error in the
measurement process is a significant proportion of the total and S(qk) has
been determine from a relatively small number of repeat measurements then
the value of k will need to be increased in order to maintain the specified
level of confidence. This revis.ed value of k is le" and can be obtained using
the procedure given in Appendix Ill. The need to use le" in place of k can be
determined by applying the following criteria:

if uc(Y} / s( q} < 3 then a value for ~ should be obtained using the


procedure given in Appendix Ill.

EDITION 1 • MAY 1994 PAGE 7 OF 18


NIS 81 • TREATMENT OF UNCERTAINTY IN EMC MEASUREMENTS

3.7 Reporting of result.

The result of the measurement, after all approprtate corrections have been
made, is y and may typically be reported as follows:

The measured result is: y dBJlV :t U dB

for a level of confidence of approximately 95%, (k = 2).

Alternatively an overall uncertainty may be given for results that are plotted
or tabulated.

4 Compliance with specification


4.1 NAMAS requirements state that when a product is tested against a declared
specification then the report must contain a statement indicating whether
the results show compliance or non-compliance with the specification. This
decision can be made by the test laboratory if the value of the test result is
higher or lower than the specification limit by a margin greater than the
estimated uncertainty. A problem arises when the margin between the
measured result and the specification limit is less than the measurement
uncertainty. In these cases the laboratory cannot be sure that compliance,
or non-compliance, has been demonstrated.

4.2 For some products and circumstances it may be appropriate for the user to
make a judgement of compliance based on whether the measured result is
within the specified limits, with no account taken of the uncertainty. This
is sometimes referred to as "shared rtsk" since the end user takes some of
the risk of the product not meeting the specification. The implications of
such a rtsk will vary considerably. It may be acceptable to ignore
measurement uncertainty for non-safety crttical performance, such as the
EMC immunity characteristics of radio and television for example, but when
testing a heart pacemaker or the ADS system on a vehicle the user may
require that the risk of the product not complying is negligible. In which
case the uncertainty must be taken into account.

4.3 EMC testing is carried out on a very wide range of products intended for a
vartety of applications. It is not therefore possible, or appropriate, for
NAMAS to recommend standard rules for judging compliance. If a test
specification lays down the criteria then this should be followed, but this is
a rare occurrence. If there is a recognised agreement, between
regulatory/ certification bodies and manufacturers for instance, then this
can be followed but again this is not common. If neither the test
specification nor regulatory bodies set down rules for compliance then the
gUidance given in this publication should be followed.

4.4 If an agreement, code of practice or specification stipulates that uncertainty


can be ignored when judging compliance then all parties should know what
that uncertainty is. The responsibility for calculating and declaring the
uncertainty rests with the test laboratory.

PAGE 8 OF 18 EDITION 1 • MAY 1994


NIS 81 • TREATMENT OF UNCERTAINTY IN EMC MEASUREMENTS

4.5 In the absence of any specification criteria. guidance. or code of practice


NAMAS accredited EMC test laboratories should advise the client when the
uncertainty involved in the measurement makes a Judgement of compliance
difficult. Examples of appropriate statements are given below:

Case A Case B CaseC Case D

TA

TA 1
upper
limit ---+- A
1
T
A 1
1
The product complies The measured result Is The measured result Is The product does not
below the specification above the specification comply
limit by a margin less limit by a margin less
than the measurement than the measurement
uncertainty: It Is not uncertainty; It Is not
therefore possible to therefore possible to
determine compliance at determine compliance at
a level of confidence of a level of confidence of
95%. However, the 95%. However. the
measured result measured result
Indicates a higher Indicates a higher
probability that the probability that the
product tested complies product tested does not
wtth the specification comply wtth the
limit. specification limit.

EDITION 1 • MAY 1994 PAGE 9 OF 18


NIS 81 • TREATMENT OF UNCERTAINTY IN EMC MEASUREMENTS

5 References
(1) Guide to the Expression of Uncertainty in Measurement. BIPM, IEC. IFCC,
ISO. IDPAe, IDPAP, OIML. International Organisation for Standardization,
Geneva. Switzerland, ISBN 92-67-10188-9, First Edition, 1993

(2) The Expression of Uncertainty in Testing. NIS80, NAMAS (to be published)


(3) The Expression of Uncertainty and Corifidence in Measurement, NAMAS
Publication NIS3003, Edition 7 May 1991.

(4) Guidelines for the expression of uncertainty of measurement in calibration.


WECC Doe. 19-1990.

(5) Guide to the evaluation and expression oJuncertainties associated with the
results of electrical measurements. Def Stan 0026/lssue 2 Sept 1988.
(6) Uncertainties in the measurement of mobile radio equipment characteristics.
ETSI Technical Report, ETR028, March 1992. .

(7) Hams, LA. and Warner, F.L. Re-examination oJmismatch uncertainty when
measuring power and attenuation. lEE Proc. Vol 128 Pt H No.l February
1981.

PAGE 10 OF 18 EDITION 1 • MAY 1994


NIS 81. TREATMENT OF UNCERTAINTY IN EMC MEASUREMENTS

APPENDIX I

Summary of equations used


1. estimated standard deviation from a sample of n readings:

1 ~ - 2
--1)
( L (qk - q)
n k-I

2. standard deviation of the mean of n readings:

s(q)

3. standard uncertainty resulting from type A evaluation.


u(xt) = s(q)

4. standard uncertainty for contributions with normal probability distribution:


uncertainty
u(xj) = k

5. standard uncertainty for contributions with rectangular probability


distribution:

6. standard uncertainty for contributions with U shaped probability


distribution:
M
fi
7. standard uncertainties in terms of the measured quantity:
u/(y) = ct• u(xt)

8. combined standard uncertainty:


m
Uc(y) E u;(y)
/-1

9. expanded uncertainty:

EDITION 1 • MAY 1994 PAGE 11 OF 18


NIS 81 • TREATMENT OF UNCERTAINTY IN EMC MEASUREMENTS

APPENDIXll

Examples of typical uncertainty budgets


The following examples give the likely uncertainty contributions for the more·
common EMC measurements. The contributions and values are not intended to
imply mandatory requirements. Laboratories should determine the
uncertainty contributions for the tests they are performing. Where the
uncertainty contribution is considered insignificant a '0' has been used.

Example 1

Measurement of vertically polarised field strength between 30 dBrV Im and 60


dBrV Im over the frequency range 30 MHz to 1 GHz on an open area test site at 3m
and lOm

Uncertainty (dB)
Probability
81conlcal Antenna Log periodic
Contribution Distribution
Antenna

3m lOm 3m lOm

Ambient signals - - - -
Antenna factor calibration normal (k = 2) ±1.0 ±1.0 ±1.0 ±1.0

Cable loss calibration normal (k = 2) ±0.5 ±0.5 ±0.5 ±0.5

Receiver specification rectangular ±1.5 ±1.5 ±1.5 ±1.5

Antenna dlrectlvlty rectangular +0.5 0 +3.0 +0.5


-0 -0 -0

Antenna factor variation with height rectangular ±2.0 ±2.0 ±0.5 ±0.5

Antenna phase centrc variation rectangular 0 0 ±1.0 ±0.2

Antenna factor frequency Interpolation rectangular ±0.25 ±0.25 ±0.25 ±0.25

Measurement distance variation rectangular ±0.6 :0.4 ±0.6 ±OA

Site Imperfections rectangular ±2.0 ±2.0 ±2.0 ±2.0

Mismatch
Receiver VRC: r l= 0.2
Antenna VRC: r e= 0.67 (81) 0.3 (Lp) U-shaped +1.1 +1.1 ±0.5 ±O.5
Uncertainty limits 20Log(l±r1r e) -1.25 -1.25

System repeatablllty (previous assessment Std Deviation ±0.5 ±O.5 ±0.5 ±0.5
of s(q,J from 5 repeats. 1 reading on EUT)

Repeatablllty of EUT • - - - -
Combined standard uncertainty u,,(y) normal +2.19 +2.16 +2.52 +1.74
-2.21 -2.20 -1.82 -1.72

Expanded uncerialnly U normal (k = 2) +4.38 +4.32 +5.04 +3.48


-4.42 -4.40 -3.64 -3.44

PAGE 12 OF 18 EDITION 1 • MAY 1994


NIS 81 • TREATMENT OF UNCERTAINTY IN EMC MEASUREMENTS

Calculation for 3m biconical antenna, positive value:

In this example it is probable that uc(y) / S(qk) > 3, unless the repeatability of the
EUT is particularly poor, and a coverage factor of k = 2 will ensure that the level of
confidence will be approximately 95%, therefore:
U = 2 uc(y) = 2 x ±2.19 = ±4.38 dB

Notes concerning example 1

1. 1 Ambient signals have not been considered in this budget since the
uncertainty will be very dependent on relative signal levels and will not affect
all frequencies equally. The effect on a measurement result due to ambient
signals should be assessed at the time measurement is made and, if
necessary, the uncertainty should be increased.

1.2 The antenna and cable will reqUire traceable calibrations for which an
uncertainty would have been estimated using NAMAS recommendations 1•
based on a normal probability distribution with k = 2.

1.3 The receiver uncertainty would probably be obtained from the


manufacturer's specification for which a rectangular distribution would have
to be assumed.

1.4 The antenna factor uncertainty does not take account ofantenna directivity.
Unless a detailed analysis is made of all the variables and their effect on the
received signal, an estimation has to be made of the limit values and a
rectangular distribution assumed. The angle ofincidence with respect to the
antenna bore sight will generally be greater for a 3 m range. Since the
antenna calibration is with respect to bore sight it is reasonable to assume
that the actual signal strength will not be less than the indicated reading,
but could be higher by an unknown amount, resulting in an asymmetric
uncertainty .

1.5 The antenna factor may vary with height and since the height will not always
be the same in use as when the antenna was calibrated an additional
uncertainty is required. A calibration certificate from NPL will normally give
gUidance on the value for this contribution.

1.6 The phase centre for log periodic antennas will vary with frequency and is
not accounted for in calibration.

1.7 The uncertainty in the measurement distance will be relatively small but will
have some effect on the received signal strength. The increase in
measurement distance as the antenna height is increased is an inevitable
consequence of the method reqUired by most radiated emission specification
standards and is therefore not considered to be a contribution to
uncertainty.

EDITION 1 • MAY 1994 PAGE 13 OF 18


NIS 81 • TREATMENT OF UNCERTAINTY IN EMC MEASUREMENTS

1.8 Site imperfections are difficult to quantify but may include the following
contributions:

-unwanted reflections from adjacent objects.


-ground plane imperfections: reflection coefficient, flatness and edge effects.
-losses or reflections from "transparent" cabins for the EUT or site coverings.
-earth currents in antenna cables (mainly effects biconical antennas).

The specified limits for the difference between measured site attenuation and
the theoretical value (±4 dB) need not be included in total since the
measurement of site attenuation includes uncertainty contributions already
allowed for in this budget, such as antenna factor.

1.9 The contribution from repeatability of the EUT needs to be assessed at the
time. It would be reasonable to base this on repeat measurements at one
or two frequencies. There is no need to make an accurate assessment of the
repeatability of the EUT if the results are clearly well within the specification
limit.

Example 2

Measurement of conducted emissions between 30 dBpV and 60 dBpV over the


frequency range 9 kHz to 30 MHz.

Probability Uncertainty (±dB)


Contribution Distribution
9 kHz - 150 MHz 150 - 30 MHz

Receiver specification rectangular 1.5 1.5

LlSN coupling specification rectangular 1.5 1.5

Cable and Input attenuator calibration normal (k = 2) 0.3 0.5

Mismatch
Receiver VRC: f l= 0.03
LlSN VRC: f a= 0.8 (9 kHz) 0.2 (30 MHz)
Uncertainty limits 20Log(l±f.fa) U-shaped 0.2 0.05
System repeatablllty (previous assessment of standard dev. 0.2 0.35
S(qk) from 10 repeats. 1 reading on EUTl

Repeatabillty of EUT • - -
Combined standard uncertainty u.,(y) normal 1.26 1.30

Expanded uncertainty U normal (k = 2) 2.5 2.6


Calculation for 9 kHz to 150 kHz ran ge:

2
'" ± + 0.2 + 0.2 2 '" ± 1.26 dB
-y

As with example 1 it is probable that uc(y) / S(qk) > 3 and k =2 will suffice.
therefore:

PAGE 14 OF 18 EDITION 1 • MAY 1994


NIS 81 • TREATMENT OF UNCERTAINTY IN EMC MEASUREMENTS

U =2 x Uc(y) = 2 x ±1.26 = ±2.5 dB

Notes concerning example 2

2.1 It is probable that there will be an attenuator at the receiver input, providing
a low VRC, so even though the LISN output is not close to son, particularly at
low frequencies, the mismatch uncertainty is relatively insignificant.

2.2 Since the two budgets produce almost the same total it would be sensible to
quote a single figure covertng the whole frequency range, say ±2.6dB.

Example 3

Radiated ElectIic Field Immunity Measurements at 3 V/m and 30 MHz to 300 MHz.

Uncertainty (±VIm)
Contribution Distribution
30 MHz - 300 MHz

Field strength monitor calibration (±1.0 dB) normal (k = 2) 0.37

System repeatablllty (previous assessment of s(qJ std. deviation. 0.3


from 5 repeats. 1 reading on EUn

Repeatabtlily of EUT • -
Combined standard uncertainty normal 0.35

Expanded uncertainly normal (k = 2.4) 0.84

Calculation of combined standard uncertainty:

(0.37)2 + 0.32 = 0.35 V/m


2

since uc(y) / S(qk) < 3 Appendix III was used to obtain a value for k p

0.35 4
veJf = -0-.-3-:-4 --0-.-1-8-5....,..4 = 7.4
--:r- --00-
+

where the number of readings to obtain S(qk) was 5, giving vt = 4

from the table in Appendix III le" = 2.4, therefore:


U = 2.4 x 0.35 = 0.84 V/m

EDITION 1 • MAY 1994 PAGE 15 OF 18


NIS 81 • TREATMENT OF UNCERTAINTY IN EMC MEASUREMENTS

Notes concerning example 3

3. 1 The measurement method used for this example is a substitution technique


where the field strength has been pre-calibrated without the EUT present as
proposed in draft versions of a revision of IEC 801-3: 1984. The standard
requires the uniformity of field strength to be between 0 and +6 dB but it is
not considered necessary to include this as a contribution to the uncertainty.

3.2 In this case the field strength monitor reading and the specification limit will
be in terms of V/m, while the monitor calibration uncertainty and the system
repeatability will most probably be given in dBs. It is recommended that the
uncertainty is calculated in terms of V/m if this how the specification limit is
defined, however, if the calculation is made in dBs the difference in this case
is insignificant.
3.3 In order to determine the probability of compliance at a level of confidence of
95% it is necessary to test at a field strength of 3.84 V/m (specification level
+ 0.84 V/m). If a failure is detected at this level but not at the specification
limit this is equivalent to case B in section 4.5, case C occurs when the
product fails at a level of 3.0 - 0.84 = 2.16 V/m and above.

3.4 The system repeatability should be based on repeat measurements of the field
uniformity calibration, including re-positioning the transmit antenna and
resetting the input power level, at a number of field monitor locations and
frequencies.

• A value for the uncertainty attributable to the equipment under test is not
included in these examples. It will need to be considered dUring the testing
and included in the uncertainty of the test if it is a significant contribution.

PAGE 16 OF 18 EDITION 1. MAY 1994


NIS 81. TREATMENT OF UNCERTAINTY IN EMC MEASUREMENTS

APPENDIX "I

Calculation of k p
When random errors in a measurement system are comparable to the systematic
errors the expanded uncertainty calculated using equation (8) may be an
underestimation, unless a large number of repeat readings have been made.

In these circumstances a coverage factor ~ will need to be obtained from the t-


distribution, based on the effective degrees offreedom, velf ' of ucfy) and the required
level of confidence.

The effective degrees of freedom is calculated from:

U:(y)
velf = -:-----..,..---:-------:--
u~(y) ui(y) r.4(y) ~(y)
- - + - - + - _...... + - -
VI v2 v3 vm

The degrees of freedom, Vt' of the standard uncertainties based on type B evaluation
can be assumed to be infmite in most cases. For standard uncertainties obtained
from a type A evaluation V t = n - 1, where n is the number of readings used to
calculate S(qk)'

The value of ~ is obtained from t-distribution tables for the appropriate level of
confidence. The following table gives values of ~ for various degrees of freedom vel[
for a level of confidence of 95%, (actually 95.45%). Values of k p for other levels ot
confidence are. given in reference [1]

Veil

kp
1

13.97
2

4.53
3

3.31
4

2.87 2.65
5

2.52
6

2.43
7 8

2.37
10

2.28
20

2.13
50

2.05
-
2.0

The criteria given in para 6.3 to detennine the need to use the procedure given in
this Appendix is based on the conclusion that if uc(y) / u(CJk ) > 3 and all the other
contributions are assumed to have infinite degrees of freedom, then velf > 81 (3 4 ),
giving a value for ~ of less than 2.05, which can be approximated by k = 2.

EDITION 1 • MAY 1994 PAGE 17 OF 18


NIS 81 • TREATMENT OF UNCERTAINTY IN EMC MEASUREMENTS

APPENDIX IV

Calculation of uncertainty in logarithmic or linear


quantities
A general expression that describes a measurement, y, with its uncertainty, U, in
relative values, based on the product of a series of input quantities, -Xi, and their
uncertainties, u(Xj) in relative values, is given by:
y(l ± U) = x\[l ± u(x t )) • X2[l ± u(X2)).-X3[l ± u(-X3)) ....x N [l ± u(xN )) IV(l)

the uncertainty terms are:


[l ± U] = [l ± u(x\)).[l ± u(x)].[l ± u(-X3)).[l ± u(xN)) IV(2)

the total uncertainty can be approximately expressed as:


U = u(xt ) + u(X2) + u(-X3) .... + u(xN ) IV(3)

which is in a form that can be treated by the RSS approach. However, by taking
logs of equation IV(2) then:
10g[1 ± V] = 10g[1 ± u(xt )] + 10g[1 ± u(X2)] + 10g[1 ± u(-X3)] + 10g[1 ± u(xJ] IV(4)

this is a more exact expression for the total uncertainty than equation IV(3) and is
also amenable to RSS treatment. However, if u(Xil in linear terms is relatively large
then:
log[ 1 + u(xl)) :I' log[l - u(xJ] IV(5)

Whether it is correct to combine uncertainties in linear form, eg %, or logarithmic


form, eg dB, will depend upon whether their probability distributions can be better
described in linear or logarithmic form. If the uncertainties for the major
contributions are supplied in terms of dB it can only be assumed that the
probability distribution that is assigned to them should also be in dBs.

In practice the difference in the calculation of u.,ly) between dBs or % is relatively


small. In example 1 the positive value for the biconical antenna at 3m is +2.19 dB
when calculated in dB. If the contributions in dB are converted to percentage
voltage ratios and the calculated Ue(y) converted back to dB the result is +2.13 dB,
a difference of 0.06 dB. However, the two expanded uncertainties are +4.38 dB
(calculated in dB) and +3.83 dB (55.5%), a diffemece of 0.55 dB.

It is recommended that if the specification limit is given in dB terms, eg dBJ.lV, and


the contributing uncertainties are mostly stated in dB, then the uncertainty
calculations should be made in dBs. If the specification limit is given in absolute
terms, eg V fm then the calculations should be made in absolute units.

PAGE 18 OF 18 EDITION I • MAY 1994

You might also like