Calculating calibration

uncertainties in an
automated temperature
calibration system Application Note

Some of the basic calibration uncertainties to

­consider with an automated temperature calibration
system like the Hart 7380 bundles are the calibra-
tion uncertainties of the reference PRT, the drift of
the PRT, uncertainties arising from bath uniformity
and stability, and the uncertainties associated with
the readout used to measure the unit under test and
the reference PRT. This application note describes
the evaluation of those uncertainties. Additional
uncertainties that depend on the performance of
the specific unit under test or the entire class of
instruments for which the procedure applies may
also be evaluated and included on a case-by-case
basis. These include hysteresis, short-term stability,
immersion effects, and self-heating.
The table below presents a sample uncertainty
budget for the 7380-USR. An uncertainty budget is
a tool that is used to help track and evaluate the
uncertainties associated with a calibration. The
uncertainty budget shown below offers one method
of aggregating some of the most important contrib- between calibration points would be a technique
utors to the uncertainty of the 7380-USR system. called linear interpolation. That is, we assume that
Each uncertainty has been given a code B1 through the uncertainty changes linearly between –196 °C
B6 to facilitate discussion below. and –38 °C from ± 0.024 °C to ± 0.011 °C.
B1 Some of the values needed for an uncertainty
budget can be found on the report of calibration for
the model 5615-12. They are the 5615-12’s own
calibration uncertainties shown in the table. 5615 calibration uncertainties
If calibration uncertainties are required at points Expanded
between those expressed on the calibration report, Temperature  Uncertainty (k=2) 
such as at –80 °C, then a rigorous approach would T1 –196 °C U1 0.024 °C
include calculating the propagated uncertain- T2 –38 °C U2 0.011 °C
ties using the law of propagation of uncertainties T3 0 °C U3 0.010 °C
as described in the GUM. A reasonable and less T4 200 °C U4 0.018 °C
strenuous method for establishing the uncertainties T5 420 °C U5 0.029 °C

Specifications and cal report data

Data –80 °C –38 °C 0 °C 100 °C
Non statistical evaluation (type B approach) Code source Type (mK) (mK) (mK) (mK)
Reference PRT calibration (model 5615-12) B1 Cal report Uncertainty 14.0 11.0 10.0 14.0
Manual
Reference PRT drift (model 5615-12) B2 Specification 6.8 8.4 10.0 13.9
(TPW)
Bath uniformity (model 7380) B3 Manual Specification 8.0 10.0 12.0 12.0
Bath stability (model 7380) B4 Manual Specification 6.0 8.0 10.0 10.0
Reference PRT readout (model 1529) B5 Manual Specification 4.0 5.0 6.0 9.0
Unit under test readout (model 1529) B6 Manual Specification 4.0 5.0 6.0 9.0
Total standard uncertainty (comparison
U 10.41 11.18 12.79 15.77
measurements)
Total expanded uncertainty (k=2)
U’ 21 22 26 32
(comparison measurements)

F r o m t h e F l u k e D i g i t a l L i b r a r y @ w w w. f l u k e . c o m / l i b r a r y
 With that assumption, we can just use the equa- Taking –80 °C as an example and pulling the
tion of a line, to solve for the slope of the line to data from the W table below that contains nominal
get m = –8.23E-05 values for the 5615, we obtain the equivalent drift
rate at –80 °C or ΔT–80 = ± 0.007 °C.
U 2 − U1
(1) m= ΔT1 = ±0.010 °C
T 2 − T1 R–80 67.72 Ω SensitivityTPW0.4 Ω/°C
Rearranging the equation, we can solve for the RTPW100ΩSensitivity–80 = 0.41 Ω/°C
uncertainty at any point between the temperatures
T1 and T2: B3 and B4 Other required values for the uncer-
tainty budget are bath stability and uniformity.
(2) U −80 = m(T −80 − T 1) + U1 These can be found in the manufacturers’ speci-
fications.The table is an example. You can rely
By plugging in the numbers from the uncertainty on these numbers for a preliminary uncertainty
table above and the value we calculated for into budget. Better results (and lower uncertainties) can
equation 2, we find that the “propagated” calibra- be obtained by monitoring and evaluating the bath
tion uncertainty at –80 °C is ± 0.014 °C. in use and using this information in the uncertainty
B2 One key specification used in determin- budget. To estimate the stability and uniformity
ing calibration uncertainties is the drift rate. The values between –80 °C and 0 °C, for example,
specification for the model 5615-12 secondary assume the changes are linear between the two
reference PRT reported by the manufacturer is: points. This allows you to use equations
“Drift at 0.010 °C equals ± 0.010 °C per 100 hours 1 and 2 above to determine the stability or
exposure to the maximum temperature of 420 °C.” ­uniformity between specifications at two
To determine the equivalent drift at any other tem- adjacent temperatures.
perature, use equation 3
RT 90 Sensitivity TPW 7380 bath specifications
(3) ∆T 2 = ∆T1 9
× Temperature Stability Uniformity Fluid
RTPW Sensitivity T 90
–80 °C ± 0.006 °C ± 0.008 °C Ethanol
0 °C ± 0.010 °C ± 0.012 °C Ethanol
100 °C ± 0.010 °C ± 0.012 °C Oil 5012

Nominal resistance (Ω) vs. temperature tables (°C)

0 –10 –20 –30 –40 –50 –60 –70 –80 –90
–200 17.08
–100 59.51 55.37 51.22 47.04 42.84 38.61 34.35 30.06 25.74 21.4
0 100 96.01 92.01 87.99 83.97 79.93 75.87 71.81 67.72 63.62

0 10 20 30 40 50 60 70 80 90
0 100 103.98 107.94 111.9 115.84 119.77 123.68 127.59 131.48 135.36
100 139.23 143.09 146.93 150.77 154.59 158.4 162.2 165.98 169.76 173.52
200 177.27 181.01 184.74 188.46 192.16 195.86 199.54 203.21 206.87 210.51
300 214.15 217.77 221.39 224.99 228.58 232.15 235.72 239.28 242.82 246.35
400 249.87 253.38 256.88 260.36 263.84 267.3 270.75 274.19 277.61 281.03
500 284.43

Sensitivity (Ω/°C) vs. temperature tables (°C)

0 –10 –20 –30 –40 –50 –60 –70 –80 –90
–200 0.43
–100 0.41 0.41 0.42 0.42 0.42 0.42 0.43 0.43 0.43 0.43
0 0.40 0.40 0.40 0.40 0.40 0.40 0.41 0.41 0.41 0.41

0.00 10.00 20.00 30.00 40.00 50.00 60.00 70.00 80.00 90.00
0 0.40 0.40 0.40 0.40 0.39 0.39 0.39 0.39 0.39 0.39
100 0.39 0.39 0.38 0.38 0.38 0.38 0.38 0.38 0.38 0.38
200 0.38 0.37 0.37 0.37 0.37 0.37 0.37 0.37 0.37 0.36
300 0.36 0.36 0.36 0.36 0.36 0.36 0.36 0.36 0.35 0.35
400 0.35 0.35 0.35 0.35 0.35 0.35 0.34 0.34 0.34 0.34
500 0.34 0 0 0 0 0 0 0 0 0

2 Fluke Calibration Calculating calibration uncertainties in an automated temperature calibration system

 1529 reference readout specifications
1529 Resistance Accuracy
T (°C) R (Ω) Specification Conversion Equation Solution

± 0.0005
–200 °C 17.08 0 Ω to 20 Ω:  ± 0.0005 Ω  ∆T = ± 0.0116 °C
Sensitivity T 90

25
20 Ω to 400 Ω:  ± 25 ppm of ± × RT 90
–80 °C 67.72 1000000 ± 0.0041 °C
reading ∆T =
Sensitivity T 90

Uncertainty evaluation
Coverage –80 °C –38 °C 0 °C 100 °C
Type B standard uncertainties Code factor (mK) (mK) (mK) (mK)
Reference PRT calibration (model 5615-12) B1 1 7.0 5.5 5.0 7.0
Reference PRT drift (model 5615-12) B2 1 3.9 4.8 5.8 8.0
Bath uniformity (model 7380) B3 1 4.6 5.8 6.9 6.9
Bath stability (model 7380) B4 1 3.5 4.6 5.8 5.8
Reference thermometer readout (model 1529) B5 1 2.3 2.9 3.5 5.2
Unit under test readout (model 1529) B6 1 2.3 2.9 3.5 5.2
Total standard uncertainty (does not include
U 1 10.4 11.2 12.8 15.8
evaluation of unit under test)
Total expanded uncertainty (k=2)
U’ 2 21 22 26 32
(comparison measurements)

B5 and B6 The 1529 manual contains accu- with a rectangular distribution to a standard
racy specifications for the readout. As with most uncertainty, the value is divided by the square
readouts, these are expressed in engineering units root of three. The individual standard uncertainties
of resistance (or voltage) rather than temperature. are combined through a process called Root Sum
So equations are used to convert the resistance Squares or RSS. This means that each uncertainty
accuracy specification into a temperature accu- is squared before adding all of the squared com-
racy specification. For example, the 1529 accuracy ponents together. The square root of the result is
specifications for PRTs are given in the table, with taken as the total combined standard uncertainty.
the appropriate conversion equation and solutions
for the 5615 at two temperatures (–200 °C and Root Sum Squares
–80 °C).
U = B12 + B22 + B 32 + B 4 2 + B52 + B62 ...
Combining uncertainties
In order to combine all of the uncertainty compo- The total expanded uncertainty is obtained by
nents that have been evaluated, they first must be multiplying the resulting value by a factor of two
converted to standard uncertainties. There are two (k=2) and is show in the table above.
kinds of uncertainties in the table. Some of them
are limits of error, and some of them are based on
a known or assumed probability distribution. The
PRT calibration uncertainties are given as k=2
uncertainties. When uncertainties are stated with a Fluke Calibration. Precision, performance, confidence.™
k=2 coverage factor, it means a normal, or Gauss-
ian, distribution has been assumed. To convert
this kind of uncertainty to a standard uncertainty,
all that is required is to divide it by two. The rest
of the uncertainties are limits of error that were
assigned without probability distributions. With-
out a known probability distribution, a rectangular
distribution is assumed. To convert an uncertainty CalPlus GmbH
Niederlassung ScopeShop
Normannenweg 30 • 20537 Hamburg
Tel.: 040 3039595-0 • Fax: 040 3039595-50
[email protected] • www.calplus.de

©2007-2011 Fluke Corporation. Specifications subject to change without notice.

Printed in U.S.A. 6/2011 3084747BA A-EN-N Pub_ID: 11277-eng Rev 02

Modification of this document is not permitted without written permission

from Fluke Corporation.

3 Fluke Calibration Calculating calibration uncertainties in an automated temperature calibration system