Professional Documents
Culture Documents
The uncertainty associated with all the laboratory measurements involves a random precision error and
a fixed bias error. The errors can each be broken down further into three components. !" calibration
errors #" data ac$uisition errors and %" data reduction errors& and '" conceptual basis erros.
Recall: Random precision errors are observed in repeated independent measurements, while bias is the
difference between an average set of readings and the true value. An example of a bias error is the
sensor uncertainty as specified by the sensor manufacturer.
The procedure attempts to $uantify the bias and precision limits& and total uncertainty for the total
resistance coefficient (
T
" and residuary resistance coefficient (
R
".
To do this& the bias error associated with these coefficients is determined from:
B
CT
#
=
C
T
C
T
!
B
!
C
T
R
x
B
Rx
C
T
#
since (
T
is a function of )&*&R
x
and density:
C
T
Tm
=
R
x
Tm
+.,!
#
and
B
CR
#
=
C
R
C
T
B
CT
C
R
"
B
"
C
R
C
#
B
C#
#
since (
R
is a function of (
T
& k& (
-
:
C
R
=C
T
!,deg
!" C
#
!,deg
)o& the ma.ority of the procedure is concerned with obtainin/ the uncertainty associated with each of
the variables above. The sources of error are or/ani0ed into the followin/ flowchart& where the
variables of interest are estimated for each individual measurement system under the cate/ories of
calibration& data ac$uisition& data reduction& and conceptual basis.
1ote as well that the total uncertainty for (
T
and (
R
is /iven by the root sum s$uare of the uncertainties
of the total bias and precision limits.
$
CT
#
=
B
CT
%
CT
$
CR
#
=
B
CR
%
CR
#
The uncertainty number is some combination of the bias and the precision errors and has a simple
interpretation: the lar/est error reasonably expected. -or example& the interval
& $
represents a band within which the true value of the measurement is expected to lie& A1)23A)45&
!67,".
-inally& note that the bias limit associated with the temperature conversion is not considered in the
analysis& so 8
(T
!, de/rees
9 8
(T
Tm
#inal note regarding lab data:
'ver the course of testing, it was clear that repeatability was going to be an issue (witnessed by what
appeared to be a substantial drift in the resistance reading over time). This may be due to the model
si*e, or insufficient settling time between runs (although +,-+. minutes was allotted for settling time
during the test). /n order to illustrate typical results and ma"e the uncertainty analysis meaningful to
the students, rather than s"ewed by these precision results, the resistance values of the repeated runs
were estimated.