Last updated Nov.

10, 2006
Lecture 1
Measurement Systems & Uncertainty
ME56
Instrumentation
& Control Engineering
Asst. Prof. Yong Gyun Kim
College of Engineering
Silliman University
Topics
Part 1 Measurement systems
The functional elements of Measurement sys.
Loading
Requirements of Measurement system

Part 2 The Expression of Measurement
results to comply international
standard
Introduction to uncertainty of measured and
analyzed values

Part 1
Measurement systems
Measurement system elements
Sensor
Sensing physical quantity
Signal processor
Converting physical value to electrical value
Amplifying weak electrical value
True
value of
variable
Signal
processor
Input
Output
Measured
Value of
variable
Display
Record
Transmit
Sensor
Methods of Measurement
Deflection Method
Analog multi-tester
Zero Method or Null Method
Thermocouple
Compensation Method
Substitution Method
Coincidence Method
Differential Method
Assignments
Show Examples
Loading in Measurement
An Act of attempting to make the
measurement has modified
measurand
A problem encountered during measurement
Hot
water
Hot
water
Cold
thermometer
Physical quantity
to be measured
R
R
Ammeter R
A
Voltmeter

R
R
I
I
P = IR
P = (I-I
V
)R
I
V
Requirement of Measurement
Fitness of purpose
Needs to deliver required accuracy
Calibration must be done
Calibration
A process of comparing the output of a
measurement system against standards of
known accuracy
Relationship between the output of a
measurement system and the quantity it senses


Calibration
Tip: Never trust a manufacturers calibration unless
you have to - they have a vested interest

Policy to control quality
Refer. Note Pages 6-7
Establish and maintain a system for calibration
Adequate training
Reviewing the calibration system periodically
Consider errors and uncertainties in measurement
process
Keep Documented procedure of calibration
Calibration records must be kept
Use the measurement equipment traceable back to
National standard
Traceability chain
National Standard
Calibration center standard
In-company Standard
Process Instrument
Primary Standards
1. Mass
2. Length
3. Time
4. Current
5. Temperature
6. Luminous intensity
7. Amount of substance
They are used to define the other quantities.

Supplementary standards
Plane angle : Radian
Solid angle : Steradian
Reliability
Being the probability that it will operate to
an agreed level of performance, for a
specific period, subject to specified
environmental condition.

Reliability = f(t, environment conditions
hot, dusty, humid, corrosion..))
Improved by choosing proper materials.

Failure rate
No. of failures / no. of system observed x time observed
Repeatability
Difference between multiple
measurements of the same quantity
Affected by fluctuation in environment

Ability of a measurement system to give
same value for repeated measurements of
the same value of a variable.

Testing a measurement system
1. Pre-installation testing
Checking calibration and operation of of Each
element or instrument.
2. Cabling and piping testing
Checking shield, continuity, insulation, leak..
3. Pre-commissioning
Checking the completion of installation
Checking full operation order when interconnected.

Assignment
Solve the Problems
Part2
Uncertainty Analysis


Technique
Expression of Measurement results to comply
international standard
Evaluation of Uncertainty
An essential process of measurement
Gives a measure of reliance of measured
values
Accuracy v.s Precision

Not Precise, Not Accurate Precise, Not Accurate
accurate, Not Precise Precise, Accurate
Accuracy vs. Precision
Accuracy
the agreement between a measured quantity and the
true value of that quantity.
Difference between the value indicated by measurement
system and true value.
How close the measurement comes to the true value ?
Every component that appears in the analog signal path
affects system accuracy.

Precision
how exactly the result is determined without
reference to what the result means.
The relative precision indicates the uncertainty in a
measurement as a fraction of the result.
Accuracy
Standard Deviation
Repeatability
Difference between multiple measurements of the same
quantity
S.D of values measured repeatedly within limited time under
the same environment, method, measurer
Presented as % of F.S
Reproducibility
A measure of consistency of measured values with changing
methods or place, condition under the repeatability condition
Presented as % of F.S
Uncertainty
(How to estimate an error)
Background
For an unified rule in measurements
A book - "Guide to the Expression of Uncertainty in Measurement, ISO in
1993
Suggested by CIPM with BIPM, IEC, IFCC, ISO, IUPAC, IUPAP, OIML

The result of measurement = the best estimate instead of true
value
Uncertainty is the error you dont know about

Definition
Parameter, associated with the result of measurement, that
characterizes the dispersion of the values that could reasonably be
attributed to the measurand
Quantified value
Could be S.D, times of S.D, .
It is the existing range of the object being measured
Errors vs. Uncertainty
Error = Systematic Error + Accidental Error
= Measured value true value


Systematic Error = average of values measured infinitely
under repeatable condition - true value
Accidental Error = Measured value the average of values
measured infinitely under repeatable condition
Errors vs. Uncertainty
Systematic Error
Due to Instrumental error,
Due to Environmental errors - temperature, wiring, stress
Enable to decrease it through calibration
Ex) Bias error If you always got same values when you
measured the measurand repeatedly No Error?

Accidental Error
Random error or Non-repeatable error due to Uncertain
causes
Possible Causes : wear, environment, thermal deformation
Calculating it using the average value through many trials
Error Distribution : Normal or Gaussian Distribution
For infinite trials Average : Zero
Errors vs. Uncertainty
Then Corrected or compensated value through calibration is
always certain?
Uncertain (error) Because cant trust the calibration
Uncertainty by systematic effect

If the measured values were random whenever you
measured them repeatedly after calibration of systematic
error? Accidental Error
Because the measurer couldnt know or control parameters to affect
the measurement.
Average of accidental errors about the values infinitely measured = 0
But cant do infinite trials of measurement Uncertain exists
Uncertainty by Accidental Effect

Uncertainty : doubts (about results and processes of
measurement
4 Steps calculating the Uncertainty
Set relating formula
Calculate Standard Uncertainty
Calculate Combined Standard Uncertainty
Calculate Expanded Uncertainty

Set the relating formula
Let y as Estimated value of measuring value
Let x as Estimated values of related parameters
y = f (x1, x2, . . ., xn) <1>

Ex) To Measure an Area, a ruler is used
Assume that the Uncertainty of the measurer and calibration of
the ruler
A = X Y <2>
Then the formula following ISO guide must be
A = (x + 1) ( y + 2 ) <3>
Where 1 and 2 are the uncertainty due to the calibration of the
each length
Example
Power dissipated in a resistor
The power dissipated by a resistor in an electrical circuit is being
estimated using the voltage across it measured using a digital
voltmeter. The resistance R has a nominal value of 100. The
voltmeter reads 28.0 volts, with a resolution of 0.1V. Estimate the
uncertainty in the power measurement

1) Uncertainty in primary measurements ?
2) Uncertainty in power ?
1) Uncertainty in primary measurements
Resistance:
A glance through any manufacturers specifications will show you
that most often nominal resistances are only accurate to within
5%. We shall therefore take our primary uncertainty here
as (R)=5.
ii) Voltage:
The uncertainty in the reading of a digital voltmeter is usually half
the resolution (the true voltage could lie anywhere between 27.95
and 28.05). We therefore have (V)=0.05V.
Example

2) Uncertainty in power
The relationship between power, voltage and resistance is simply

Preparing to apply the equation we have,




And so,



Or


Example
about 5%. You will notice that apart from giving us an
uncertainty estimate this analysis also shows that the likely
error in our power measurement is almost entirely due to
the uncertainty in resistance. To improve the accuracy we
should therefore concentrate on reducing the uncertainty of
the resistance measurement, not on improving the
voltmeter. This kind of information can save a lot of time
and money (voltmeters are expensive, resistors are not).
Calculate Standard Uncertainty u(xi)
Uncertainties according to the parameters u(xi)

Two methods to calculate it
uncertainty evaluation following Type A
By S.D of the average through many trials of measurement
uncertainty evaluation following Type B
If there were only one measured value for the parameter, from
the data or materials measured before, the characteristic of the
measurement system, By the specification presented by
manufacturer, calibration data, calculate the equivalent S.D

Uncertainty evaluation following
Type A
Where, X
i,k
; each independent
measured value for parameter i
x
i,
= /X
i
; estimated value of the
related parameter x
i,
= /X
i

/X
i
; Average of X
i,k
of n trials
Uncertainty evaluation following
Type B
u
2
(x
i
) ; Estimated S.D related to estimated value
(x
i
) of input value X
i
,not many trials

Or

Standard Uncertainty u(x
i
) evaluated from the
data or materials measured before, the
characteristic of the measurement system, the
specification presented by manufacturer and
calibration data
Need to estimate S.D Technique
Uncertainty evaluation following
Type B
Evaluation of B type Uncertainty according to the characteristic of
data distribution through estimating S.
Estimate the S.D using the distributions shown below.

It shows the estimated value of input value Xi and uncertainty
evaluation from probability distribution of Xi based on the
information obtained
Combined Standard Uncertainty u
c
(y)
Could be calculated using the equation <1>, <2>
by propagation law of uncertainty
Assumption : No corelations among the parameters or
individual standard uncertainty << measured values
u
2
c
(y) = (f/x
i
)
2
u
2
(x
i
) <5>
If there were corelations among the parameters
Include covariance terms of inputs

If the individual standard uncertainty < measured values
Consider using propagation law including high order term

Expanded uncertainty, U
If degrees of freedom is bigger
It approaches to normal distribution
k=2 in 95 % confidence interval
k=3 in 99 % confidence interval
U = k u
c
<6>
What must be described in
Measurement Report
Measured data
Method used to calculate uncertainty
Parameters to contribute uncertainty
Its evaluation method
Data analysis method
Correction value and constants used in
analysis and their sources

Noise in Measurement
White Noise
Essential Noise due to Thermal movement of electron
Electrons start moving @ above 0
o
K
Electrostatic Induction Noise
Prevent it by electrostatic shield (Ground, Earth)
Electromagnetic Induction Noise
Affected by change of magnetic flux
Twisting pair wiring
Avoid vibration of wire
Ripple in Power
60Hz noise
Preventing it by line filter or capacitors

Primary measurement
A measurement not derived from any
other
Ex.
A distance from a dial gauge
Height from a manometer
(not Velocity from a pitot tube)
Determining Uncertainty in a
Primary measurement
... ) ( ) ( ) ( ) (
2 2 2
+
(

c
c
+
(

c
c
+
(

c
c
= c
c
R
b
b
R
a
a
R
R o o o o
Experimental result R is determined from one or more
primary measurements a,b,c
R=f(a,b,c)
Uncertainty in R is due to the primary measurements may
be determined by multiplying the uncertainty in that
measurement by the sensitivity of R to that measurement.
Total Uncertainty R
Example1-Calculating Uncertainty
Uncertainty in velocity for a pitot tube
p
0
-p=0.22kPa (by manometer reading)
P
atm
=941.1mBar=94110Pa
T
atm
=17.0C=290.0K


atm
atm
RT
p
p p
U
) ( 2
0

=
s m U / 72 . 19
290 287
94110
220 2
=

=
Example1-Calculating Uncertainty
Step1 : primary uncertainty
Manometer
(p-p
0
)=0.02kPa = 20pa (eyeball)
Barometer
(p
atm
)=0.5mBar=50Pa
Thermometer
(T
atm
)=0.5C

0.21 2 3
Example-Calculating Uncertainty
Step2 : Partial Derivatives

atm
atm
RT
p
p p
U
) ( 2
0

=
0448 . 0
) ( 2
1 2 ) ( 2
2
1
) (
0
2 / 1
0
0
=

=
|
|
.
|

\
|
=
c
c

p p
U
p
RT
p
RT p p
p p
U
atm
atm
atm
atm
0340 . 0
2
1
= =
c
c
atm atm
T
U
T
U
000105 . 0
2
1
= =
c
c
atm atm
p
U
p
U
Example1-Calculating Uncertainty
Step3 : Combine Uncertainty

2 2 2
0
0
) ( ) ( ) (
) (
) (
(

c
c
+
(

c
c
+
(

c
c
=
atm
atm
atm
atm
p
p
U
T
T
U
p p
p p
U
U o o o o
Example2-Calculating Uncertainty
using computer
Reynolds number