ACCREDITATION SCHEME FOR LABORATORIES

Guidelines on the Evaluation and

Expression of Measurement Uncertainty
for Electrical Testing Field

Guidance Notes EL 001, May 02

The SAC Accreditation Programme is managed by SPRING Singapore

All rights reserved

Basic Principles on Measurement Uncertainty

1. Evaluation of Uncertainty
The uncertainty of the result of a measurement generally consists of several
components. They can be grouped into two categories according to the
method used to estimate their numerical values:

Type A evaluation
Calculation of uncertainty is by statistical analysis through repetitive
observations.

Type B evaluation
Calculation of uncertainty is by means other than statistical analysis.

2. Modeling the Measurement Process

A measurand Y can be determined from N inputs quantities X1, X2, X3

XN, through a function f:

Y = f(X1, X2, X3 XN)

An estimate of Y, denoted by y, is obtained from x1, x2, x3 xN, the

estimates of the input quantities X1, X2, X3 XN, through the same
function f:

y = f(x1, x2, x3 xN)

The uncertainty associated with the estimate y is obtained by

appropriately combining the estimated standard deviation (or standard
uncertainty) of each of the input estimate xi.

3. Type A Evaluation of Standard Uncertainty

The arithmetic mean for n independent observations:

The standard deviation of the n independent observations:
1 n
q = qk
n k =1

1 n
s(q k ) = (qk q )2
n 1 k =1

The standard deviation of the mean (estimate the spread of the distribution of
the means):
s(qk )
s(q ) =
n
Guidance Notes EL 001, May 02 Page 1 of 15
 For an input estimate xi determined from n repeated observations, the Type A
standard uncertainty u(xi), with degrees of freedom v is given by:

u (xi ) = s(q )

vi = n 1

Note: the degree of freedom should always be given when Type A evaluation
of an uncertainty component is reported.

4. Type B Evaluation of Standard Uncertainty

Covert a quoted uncertainty to a standard uncertainty from the knowledge of

the probability distribution of the uncertainty.

Commonly used probability distributions:

- Normal or Gaussian probability distribution

- Rectangular probability distribution

Degree of freedom is assumed to be infinite

Normal Probability Distribution

A normal distribution can be assumed when an uncertainty is quoted with a given

confidence level. For example, a calibration report states that the uncertainty of a
voltmeter is 0.1 V with a confidence level of 95%. The standard uncertainty of
the voltmeter is given by:

k 0. 1
u (x ) = = = = 0.051 V
k 1.96

(Note: 95 % level of confidence has a coverage factor of 1.96)

Guidance Notes EL 001, May 02 Page 2 of 15

 Rectangular Probability Distribution

When an uncertainty is given by maximum bound within which all values are
equally probable, the rectangular distribution can be assumed. For example, the
accuracy of a voltmeter of a specific range is quoted as 0.2 V. The standard
uncertainty of the voltmeter is given by:

a 0. 2
u (x ) = = = 0.115 V
3 3

5. Combined Standard Uncertainty

The estimate of a measurand Y is given by:

y = f(x1, x2, x3, . xN)

f f f f
y = x1 + x2 + x3 + + x N
x1 x2 x3 x N

It can be shown that the above equation leads to:

2 2 2 2
f 2 f 2 f 2 f 2
uc ( y ) = u (x1 ) + u (x2 ) + u (x3 ) + + u (x N )
2

x1 x2 x3 x N
= c1 u 2 (x1 ) + c2 u 2 (x2 ) + c3 u 2 (x3 ) + + c N u 2 (x N )
2 2 2 2

The combined standard uncertainty:

u c ( y ) = c1 u 2 (x1 ) + c2 u 2 (x2 ) + c3 u 2 (x3 ) + + c N u 2 (x N )

2 2 2 2

where c1, c2, c3.cN are the sensitivity coefficients

Guidance Notes EL 001, May 02 Page 3 of 15

 Each component of the combined standard uncertainty could be calculated using
either Type A or Type B evaluation method.

6. Coverage Factor of Combined Uncertainty

To determine the coverage factor of combined uncertainty, the effective degree

of freedom must be first calculated from the Welch-Satterthwaite formula:

uc (y )
4
eff =
ci u 4 (xi )
N 4

i =1 vi

Based on the calculated veff, obtain the t-factor tp(veff) for the required level of
confidence p from the t-distribution table.

The coverage factor will be:

kp = tp(veff)

7. Expanded Uncertainty

The expanded uncertainty defines an interval about the estimated result y within
which the true value of the measurand Y is confidently believed to lie. It is given
by:

U = kp uc(y)

The measurand Y is reported in the following format:

Y=yU

It means that the true value of measurand Y is confidently believed to fall within
the following range:

yUYy+U

Guidance Notes EL 001, May 02 Page 4 of 15

 Example #1: Resistance Measurement

A milliohm meter is used to measure the resistance of a current shunt resistor. At

the selected range of the meter for the measurement, the calibration certificate
states an uncertainty of 0.2 m at 95 % of confidence level. Effects of room
temperature and humidity on the measurement are found to be negligible.

Measurement record:

Reading 1 2 3 4 5 6 7 8 9 10
R (m) 9.4 9.1 9.4 9.8 9.7 9.4 9.8 9.7 9.4 9.4

1. Measurement Process Model

The measured resistance is given by:

Rx = Rrdg + Rm

where Rrdg: resistance reading recorded by the meter

Rm: meter uncertainty

2. Uncertainty Equation

The combined standard uncertainty is given by:

u c (R ) = c1 u 2 (R rdg ) + c 2 u 2 ( R m )
2 2

Rx Rx
Since c1 = = 1 and c2 = = 1 , the combined standard uncertainty is give
Rrdg (Rm )
by:

uc (R ) = u 2 (Rrdg ) + u 2 (Rm )

Guidance Notes EL 001, May 02 Page 5 of 15

 where

u(Rrdg) is the standard uncertainty due to the repeatability of the meter reading

u(Rm) is the standard uncertainty due to the meter calibration

3. Calculation of Uncertainty Components

Type A evaluation:
The best estimate of the measured resistance is given by the arithmetic mean:

1 10 1
R = Rk = (95.1) = 9.51 m
10 k =1 10
Standard deviation:

s( R) =
1 10
(Rk R )2 = 1
(2.449) = 0.522 m
10 1 k =1 9

Standard uncertainty:
u (Rrdg ) = s (R ) =
s( R) 0.522
= = 0.165 m
n 10

Degree of freedom, = 9

Type B evaluation:
The uncertainty of the calibration is 0.2 m with 95 % of confidence level (k =
1.96).
0.2
u (Rm ) = = 0.102 m
1.96
Degree of freedom, =

Note: The value of 0.2 m is used as a component for Type B evaluation on the
assumption that the drift and stability of the equipment is negligible.

4. Uncertainty Budget Table

Source of Type Uncertainty Probability k ui (m) ci ci ui vi

Uncertainty Value (m) Distribution
Repeatability A 0.165 - - 0.165 1 0.165 9
u(Rrdg)
Meter B 0.200 Normal 1.96 0.102 1 0.102
Calibration
u(Rm)

Guidance Notes EL 001, May 02 Page 6 of 15

5. Combined Standard Uncertainty

uc (R ) = 0.1652 + 0.102 2 = 0.194 m

6. Effective Degrees of Freedom

0.1944
eff = 17
0.1654 0.102 4
+
9

7. Expanded Uncertainty

For veff = 17, the coverage factor of the combined standard uncertainty (kp) is equal
to 2.11 at 95 % level of confidence.

U = kp uc = 2.11 0.194 = 0.409 m

8. Reporting of Result
R = 9.51 0.409 m

The measured resistance of the current shunt resistor is 9.51 m. The expanded
uncertainty is 0.409 m with a coverage factor of 2.11, assuming a normal
distribution at a level of confidence of 95 %.

Guidance Notes EL 001, May 02 Page 7 of 15

 Example #2: Temperature Measurement

A digital thermometer with a Type K thermocouple is used to measure the

temperature inside a temperature chamber. The temperature controller of the
chamber is set at 400oC.

Digital thermometer specification:

Accuracy = 0.6 oC

Thermocouple specifications:

Temperature correction for the thermocouple at 400 oC is 0.5 1.0 oC at 95

% confidence level
Deviation due to immersion = 0.1 oC
Deviation due to drift = 0.2 oC

Measurement record:

S/N 1 2 3 4 5 6 7 8 9 10
T 400.1 400.0 400.1 399.9 399.9 400.0 400.1 400.2 400.0 399.9
o
( C)

1. Measurement Process Model

The measured temperature is given by:

t x = t rdg + t m + t tc + t imm + t drift

where

trdg is the temperature reading recoded by the digital thermometer

tm is the accuracy of digital thermometer
ttc is the temperature correction of the thermocouple
Guidance Notes EL 001, May 02 Page 8 of 15
timm is the deviation due to immersion of the thermocouple
tdrift is the deviation due to drift of the thermocouple

2. Uncertainty Equation

uc (t x ) = u 2 (t rdg ) + u 2 (t m ) + u 2 (ttc ) + u 2 (timm ) + u 2 (t drift )

All the sensitivity coefficients are equal to unity.

3. Calculation of Uncertainty Components

Type A evaluation:
The best estimate of the measured temperature is given by the arithmetic mean:

1 10
T =
10 k =1
Tk = 400.02 o C

Standard deviation:
1 10
s(T ) = (Tk T )2 = 0.103 o C
10 1 k =1

Standard uncertainty:

s(T ) 0.103
u (t rdg ) = s (T ) = = = 0.033 o C
n 10

Degree of freedom, = 9

Type B evaluation:
The accuracy of the digital thermometer = 0.6 oC. Assume rectangular distribution,
the standard uncertainty of the digital thermometer meter:

0.6
u (t dev ) = = 0.346 o C
3
Degree of freedom, =
o
The uncertainty of the temperature correction of the thermocouple = 1.0 C at 95
% confidence level (k =1.96). The standard uncertainty due to temperature
correction:

1.0
u (ttc ) = = 0.510 o C
1.96

Degree of freedom, =
Guidance Notes EL 001, May 02 Page 9 of 15
 o
The uncertainty of the thermocouple due to immersion = 0.1 C. Assume
rectangular distribution, the standard uncertainty due to immersion:

0.1
u (timm ) = = 0.058 o C
3
Degree of freedom, =

The uncertainty of the thermocouple due to drift = 0.2 oC. Assume rectangular
distribution, the standard uncertainty due to drift:

u (t drift ) =
0.2
= 0.115 o C
3

Degree of freedom, =

4. Uncertainty Budget Table

Source of Type Uncertainty Probability k ui (oC) ci ci ui vi

Uncertainty Value (oC) Distribution
Repeatability A 0.033 - - 0.033 1 0.033 9
u(trdg)
Digital B 0.6 Rectangular 1.732 0.346 1 0.346
Thermometer
u(tm)
Temperature B 1.0 Normal 1.96 0.510 1 0.510
correction
u(ttc)
Immersion B 0.1 Rectangular 1.732 0.058 1 0.058
u(timm)
Drift B 0.2 Rectangular 1.732 0.115 1 0.115
u(tdrift)

5. Combined Standard Uncertainty

uc (t x ) = 0.0332 + 0.346 2 + 0.510 2 + 0.0582 + 0.1152 = 0.63 o C

6. Effective degrees of freedom

0.634
eff =
0.0334 0.510 4 0.058 4 0.115 4 0.346 4 0.029 4
+ + + + +
9
= 1,195,498

Guidance Notes EL 001, May 02 Page 10 of 15

7. Expanded Uncertainty

Degree of freedom for the combined standard uncertainty approaches . Therefore,

coverage factor of the combined standard uncertainty (kp) is equal to 1.96 at 95 %
level of confidence.

U = kp uc = 1.96 0.63 = 1.235 oC

8. Reporting of result

The correction at 400 oC is 0.5 oC, hence

T = (400.02 + 0.5) 1.235 oC = 400.52 1.235 oC

The measured temperature of the chamber is 400.52 oC. The expanded uncertainty
o
is 1.235 C with a coverage factor of 1.96, assuming a normal distribution at a
level of confidence of 95 %.

Guidance Notes EL 001, May 02 Page 11 of 15

 Example #3: Current Measurement

A current of 10 A is measured by using a current shunt and a voltmeter.

Current shunt specifications:

The calibration report gives R = 0.010088 at 10 A (23 oC) and expanded

uncertainty = 0.08% at 95 % confidence level
Temperature coefficient between 15 to 30 oC = 60 ppm/K
Uncertainty due to resistance drift is negligible

Digital voltmeter specifications:

o
Under the condition of 15 to 40 C

Range Full scale Uncertainty

(% of reading + number of counts)
200 mV 199.99 mV 0.03 +2

Measurement record:
o
Room temperature = 23 5 C

Reading 1 2 3 4 5 6 7 8 9 10
Voltage 100.6 100.8 100.7 100.6 100.6 100.9 100.6 100.6 100.7 100.6
(mV) 8 3 9 4 3 4 0 8 6 5

1. Measurement Process Model

V
I = f (V , R ) =
R

Guidance Notes EL 001, May 02 Page 12 of 15

2. Uncertainty Equation

2 2 2 2
I 2 I 2 I 2 I 2
uc (I ) = u1 (V ) + u 2 (V ) + u3 (R ) + u 4 (R )
2

V V R R
2
[ 2 2
] 2
[
= c1 u1 (V ) + u 2 (V ) + c2 u3 (R ) + u 4 (R )
2 2
]
The sensitivity coefficients:

I 1 I V
c1 = = and c1 = = 2
V R R R
where

u1(V): standard uncertainty of measured voltage due to repeatability

u2(V): standard uncertainty of measured voltage due to voltmeter resolution
u3(R): standard uncertainty of current shunt calibrated resistance value
u4(R): standard uncertainty of current shunt resistance due to temperature effect

3. Calculation of Uncertainty Components

Type A evaluation:

The best estimate of the measured voltage is given by the arithmetic mean:

1 10 1
V = Vk = (1007.2 ) = 100.72 mV
10 k =1 10

Standard deviation:

s(V ) =
1 10

10 1 k =1 9
( )
(Vk V )2 = 1 1040 104 = 10.75 102 mV

Standard uncertainty:
s(V ) 10.75 10 2
u1 (V ) = s (V ) = = = 3.40 10 2 mV
n 10

Degree of freedom, 1 = 9

Type B evaluation:

The resolution of the voltmeter = 0.03 % of reading + 2 counts

= (0.03/100) 100.72 + 2(0.01)
= 5.02 102 mV

Assuming rectangular distribution, the standard uncertainty due to voltmeter

resolution:

Guidance Notes EL 001, May 02 Page 13 of 15

 5.02 10 2
u 2 (V ) = = 2.90 10 3 mV
3
Degree of freedom, 2 =

The uncertainty of the shunt resistance = 0.08 % 0.010088

=(0.08/100) 0.010088
= 8.07 10-6

Normal distribution with 95 % level of confidence (k = 1.96)

8.07 10 6
u3 (R ) = = 4.12 10 6
1.96
Degree of freedom, 3 =

The uncertainty of the shunt resistance due to temperature effect:

60 10-6 t R = 60 10-6 5 0.010088 = 3.03 10-6

Assuming rectangular distribution,

3.03 10 6
u 4 (R ) = = 1.75 10 6
3
Degree of freedom, 4 =
1 1
c1 = = = 99.128 S
R 0.010088

V 100.72 10 3
c1 = = = 989.70 V/ 2

R2 0.0100882

4. Uncertainty Budget Table

Source of Type Uncertainty Probability k ui ci ci ui vi

Uncertainty Value Distribution (A)
Voltmeter A 3.40 10-2 - - 3.40 99.128 3.37 9
-2 -3
Repeatability mV 10 mV S 10
u1(V)
Voltmeter B 5.02 10-2 Rectangular 1.732 2.90 99.128 2.87
-2 -2
Resolution mV 10 mV S 10
u2(V)
Shunt B 8.07 10-6 Normal 2 4.12 989.7 4.08
2
Resistance 10-6 V/ 10-3
u3(R)
Shunt Temp. B 3.03 10-6 Rectangular 1.732 1.75 989.7 1.73
2
Effect 10-6 V/ 10-3
u4(R)

Guidance Notes EL 001, May 02 Page 14 of 15

5. Combined Standard Uncertainty

uc (I ) = c1 u1 (V ) + c1 u 2 (V ) + c2 u3 (R ) + c2 u 4 (R )
2 2 2 2 2 2 2 2 2

( ) (2
) ( 2
= 3.37 10 3 + 2.87 10 3 + 4.08 10 3 + 1.73 10 3 ) (
2
)2

uc (I ) = 3.92 10 5 = 6.26 10 3 A

6. Effective Degrees of Freedom

eff =
(6.26 10 ) 3 4

(3.37 10 ) + (2.87 10 ) + (4.08 10 ) + (1.73 10 )

3 4 3 4 3 4 3 4

9
= 107

7. Expanded Uncertainty

Since veff = 10& >100, the coverage factor of the combined standard uncertainty (kp)
approaches 1.96 at 95 % level of confidence.
-3
U = kp uc = 1.96 6.26 10 = 0.012 A

8. Reporting of Results

V 100.72 10 3
I= = = 9.984 A
R 0.010088

I = 9.984 0.012 A

The measured current is 9.984 A. The expanded uncertainty is 0.012 A with a

coverage factor of 1.96, assuming a normal distribution at a level of confidence of 95
%.

References:
1. SAC-SINGLAS Technical Guide 1: Guidelines on the Evaluation and
Expression of Measurement Uncertainty, 2nd Edition, March 2001.

2. ISO Guide to the Expression of Uncertainty in Measurement, 1995.

3. NIST Technical Note 1297: Guidelines for Evaluating and Expressing the
Uncertainty of NIST Measurement Results, 1994.

Guidance Notes EL 001, May 02 Page 15 of 15