SPRT - Its-90 TG3

MSL Technical Guide 3
Measurement Standards Laboratory Propagating uncertainty in the

SPRT sub-ranges of ITS-90
Step 1: Calculate the resistance ratio at the un-

1. Introduction and use of this guide known temperature
Interpolation is the mathematical process of inserting The resistance ratio is defined as
another entry between two known entries in a table or
series of numbers. Perhaps the simplest and most R(T )
W (T ) = (1)
common example of the use of interpolation in metrol- R (0.01 C)
ogy is in the use of calibration certificates where, for
example, we might average the corrections at 30 °C and
where R(T) is the resistance of the SPRT at temperature
40 °C to obtain a value for the correction at 35 °C. The
T, and R(0.01 °C) is the resistance of the SPRT at the
international measurement community uses interpola-
triple point of water. The ratio W(T) is therefore a di-
tion in a more sophisticated form in the establishment of
mensionless quantity that describes the temperature
the International Temperature Scale of 1990 (ITS-90).
dependence of the SPRT. ITS-90 constraints on the W
Although we have a very good understanding of ther-
values at the gallium and mercury points ensure that all
modynamics, it is impracticably difficult to measure tem-
SPRTs have a very similar W(T) relationship.
peratures to high accuracy on the thermodynamic scale.
As a solution the international measurement community, Step 2: Interpolate a value for the reference resis-
through the Conférence Générale des Poids et Mesures tance ratio
(CGPM), has developed an approximation to the ther- While SPRTs have very similar resistance-
modynamic temperature scale based on more practical temperature characteristics, they are not identical and
thermometers, and interpolation between fixed points somehow this must be compensated. This is done by
with defined temperatures. The fixed points are the melt- using each SPRT to interpolate between reference re-
ing-, freezing- and triple-point temperatures of various sistance ratios defined for each fixed point. The SPRT
pure substances. This ‘wire’ scale is a good approxima- interpolations specified in ITS-90 all have the form
tion to the thermodynamic temperature scale, and unlike
the thermodynamic scale, is sufficiently reproducible, Wr = W − ∆W (W ) (2)
low-cost, and accessible, to satisfy almost all our trade,
safety and research needs in thermometry.
The purpose of this document is to provide a guide and transform the measured resistance ratio W to the
reference resistance ratio Wr. The form of the deviation
to the propagation of uncertainty on the Standard Plati-
num Resistance Thermometer (SPRT) sub-ranges of function ∆W is defined by the ITS-90 and is typically a
ITS-90, which cover the temperature range from polynomial in W − 1 and/or log(W) with the coefficients
–260 °C to 962 °C. This guide is intended primarily for of the equation determined from measurements of the
SPRT users for whom the uncertainty in SPRT meas- thermometer resistance at the specified fixed points
urements is a significant factor in their temperature (see Appendix A). Eight different deviation functions are
measurements. defined by ITS-90 for the eleven different SPRT sub-
The guide begins with an overview of the mathe- ranges. The various deviation functions and temperature
matical operations that define the SPRT sub-ranges of ranges are given in Appendix B.
ITS-90. Section 3 then describes, as a simple example, It is worth considering a simple example at this point
the propagation of uncertainty in the water-tin-zinc (0 °C to see how the interpolation works. The interpolation
equation for the water-zinc sub-range is
to 420 °C) sub-range of ITS-90. The example also intro-
duces the general principles of the uncertainty calcula-
tions described in detail in Section 4. Section 5 then W r = W − a(W − 1) − b(W − 1) 2 . (3)
considers several detailed examples illustrating applica-
tion to specific cases. The appendices include a sum- The values of the coefficients a and b are deter-
mary of the sensitivity coefficients for each of the sub- mined by requiring the equation to be satisfied at the tin
ranges. and zinc points and are therefore determined from the
two equations
2. The SPRT sub-ranges of ITS-90
Wr,Sn = WSn − a(W Sn − 1) − b(WSn − 1) 2
The standard platinum resistance thermometer (4a,b)
(SPRT) sub-ranges of the International Temperature Wr,Zn = W Zn − a(W Zn − 1) − b(W Zn − 1) 2
Scale of 1990 (ITS-90) define temperature in terms of
the resistance of SPRTs. The mathematical definition
follows three algebraic steps:
Measurement Standards Laboratory of New Zealand • FAX 64 (0)4 931 3003 • email: msl@irl.cri.nz • www.irl.cri.nz/msl/
where WSn and WZn are the measured values of resis- Wr (W ) = Wr,H2O LH2O (W )
tance ratio at the two fixed points, and Wr,Sn and Wr,Zn (6)
are the two values of reference resistance ratio as- + Wr,Sn LSn (W ) + Wr,Zn LZn (W ) .
signed to the tin and zinc points by ITS-90. Once the
values of a and b have been determined from Equations where the L(W) functions are called Lagrange polyno-
(4a) and (4b), the value of Wr is calculated from W at the mials after the French mathematician who discovered
unknown temperature using Equation (3). Figure 1 them, and are given by
shows a simple graphical illustration of the interpolation
for three different SPRTs. (W − W Sn )(W − W Zn )
LH2O (W ) = ,
(WH2O − W Sn )(WH2O − W Zn )
3
Reference resistance ratio (W r)
Zinc Points (W − WH2O )(W − W Zn )

LSn (W ) = , (7a,b,c)
2.5 (W Sn − WH2O )(W Sn − W Zn )
2 (W − WH2O )(W − W Sn )
Tin Points
L Zn (W ) = .
(W Zn − WH2O )(W Zn − W Sn )
1.5
Note that we have explicitly included the variables

1
Water Triple Points W r,H2O and WH2O in Equations (6) and (7) to highlight
0.5 the form of the equations, although both variables are
0.5 1 1.5 2 2.5 3 equal to 1 by definition. Figure 2 plots the three L(W)
Measured resistance ratio (W ) functions identified in Equation (7). Each of the functions
takes the value 1.0 exactly at the fixed point for which it
Figure 1: SPRT interpolation in ITS-90 for three different is named and is zero at the other two fixed points. If you
SPRTs. For the water-tin-zinc sub-range, the water, tin and have not seen these functions before it is worth spend-
zinc fixed-point measurements define a quadratic relationship ing a few minutes successively substituting W = 1,
between W and Wr. Note that any horizontal line on the graph
is an isotherm. Also the large variations in the values of WSn W = W Sn , and W = W Zn into Equation (5) to see how
and WZn shown in the graph are chosen for illustrative purpose the interpolation works.
only and are not realistic.
1.5
Step 3: Calculate the temperature
Once the Wr value has been calculated, the tem- L H2O(W ) L Sn(W ) L Zn(W )
perature is determined using the Wr(T90) function de- 1
fined by the ITS-90. This function is defined in two seg-

L (W ) value
ments, one for temperatures below 0 °C and one for

0.5
temperatures above 0 °C. The definitions are rather
complex polynomials. However they are exact one-to-
one relationships and introduce no uncertainty into 0
ITS-90.
3. Propagation of uncertainty – a spe- -0.5

1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6
cific case W value
Let us continue with the water-tin-zinc interpolation

equation. If the values of a and b determined from equa- Figure 2: The three Lagrange polynomials for the water-tin-
tions (4a) and (4b) are substituted back into Equation (3) zinc sub-range.
then Equation (3) can be rearranged as
Wr (W ) =
(W − WSn ) (W − W Zn ) Figure 3 shows another graph, this time of two inter-
polations. In the first, the curve passes through all three
(1 − WSn ) (1 − W Zn ) marked points. In the second a small error has been
+ Wr,Sn
(W − 1) (W − W Zn ) (5)
introduced into the measured zinc-point value, WZn. The
new curve passes through the new point but continues
(WSn − 1) (WSn − W Zn ) to pass through the water- and tin-point values. Of par-
+ Wr,Zn
(W − 1) (W − WSn ) . ticular note is the difference between the two curves,
which has the same shape as the LZn(W) function of
(W Zn − 1) (W Zn − WSn ) Figure 2. In a similar fashion, any errors in the determi-
nation of the resistance ratio at the tin point also cause
Equation (5) is written in the form of a sum of three the interpolation to change by an amount proportional to
functions, each associated with one of the fixed points LSn(W). Figure 3 illustrates the fact that the Lagrange
and multiplied by the reference resistance ratio for the polynomials (Equations (7a-c)) tell us how errors at the
respective fixed point: fixed points propagate to other temperatures. This is the
rationale for choosing the particular form of Equation (5).
3 weighting of the fixed-point measurement terms by the
L(W) functions, except that there is apparently no term
∆W Zn for the LH2O ( W ) function.
Interpolated W r Values
2.5
Equation (9) assumes that we have used three sepa-
rate measurements of the water-triple-point resistance.
2 A number of different cases arise depending on whether
the measurements of the triple-point resistance are dif-
ferent; some of these are treated in Section 5. In the
1.5 simplest case, all three triple-point resistance values are
the same, hence
1
1
1 2 3
dWr = {dR +
Measured W values RH2O
Figure 3: The influence of an error in one of the measured [WSn LSn (W ) + W Zn LZn (W ) − W ]dRH O 2
(11)
fixed-point values. The solid line shows an interpolation be-
tween the measured W values and the Wr values. The dotted − LSn (W )dR Sn − LZn (W )dR Zn } .
line shows the effect of a small error in the measured WZn
value. The difference in the two interpolations has the same The term in the square brackets can be simplified
shape as the LZn(W) function in Figure 2.
because the Lagrange polynomials satisfy the relation
Mathematically the propagation of errors in the inter-
polation can be expressed as LH2O (W ) + W Sn LSn (W ) + W Zn L Zn (W ) = W (12)
dW r = dW − LSn (W )dW Sn − LZn (W )dW Zn (8) so that
where the dW symbols (differentials) represent small 1

errors in the respective quantities. Note there is no term dWr = [dR − LH2O dRH2O
RH2O (13)
for errors in the water-triple-point measurements; this is
because the W value for the triple point is 1.0 by defini- − LSn dR Sn − LZn dR Zn ] ,
tion. However each of the W values in Equation (8) is
affected by errors in the triple-point-resistance meas-
urements, which we must somehow unravel. To simplify which is the error propagation equation for the simplest
the analysis we need to calculate how the errors in the case. In this case the LH2O (W ) function reappears with
W values depend on the errors in the respective resis- the same pattern as the L(W) functions for the other
tance measurements. The required relation for the resis- fixed points. Assumptions made about the water-triple-
tance ratio at the tin point is point values and the correlations between them have a
significant impact on the final form of the uncertainty
dW Sn =
1
RH2O
[
dR Sn − W Sn dR H2O, Sn ] (9)
equations. The uncertainty propagation law correspond-
ing to Equation (13), without correlations, can be written
by substituting uncertainties for the differentials and
adding all terms in quadrature:
where RSn is the resistance of the SPRT at the tin-point
and RH2O,Sn is the value of the triple-point resistance 2
 1 
used to calculate WSn. A similar relationship applies to
the resistance ratio at the zinc point and the resistance
U 2
Wr =
 RH O  R
(
 U 2 + L2H OU R2
2 H2O
(14)
ratio at the unknown temperature, but not for the triple
 2 
point of water (Equation (9) is zero in that case). If + L2SnU R2Sn + L2Zn U R2Zn .)
Equation (9) is duplicated for each of the measured W
values, the total error in Wr due to errors in the resis-
This gives us the uncertainty in the Wr value; the uncer-
tance values is
tainty at any temperature is scaled according to the
slope of the T90(Wr) function
dWr =
1
RH2O
[(dR − WdR H2O )
2
 dT  2
UT290 =  90  UW .
 (15)
− LSn (W )(dR Sn − WSn dRH2O,Sn ) (10)  dWr 
r
− LZn (W )(dR Zn − W Zn dRH2O,Zn ) . ]

Figure 4 shows a numerical example of the three
components of the uncertainty, due to uncertainties in
This is the fundamental equation for propagating er-
the water, tin, and zinc points and the total uncertainty.
rors in the water-tin-zinc sub-range of ITS-90. The terms
Note that the calculations should be carried out in terms
in the first parenthesis describe the effect of errors in the
of the uncertainties in resistance; i.e. convert all contrib-
measurement of resistance at the unknown tempera-
uting uncertainties to an equivalent uncertainty in resis-
ture, and the terms in the last two parentheses are due
tance, sum the terms according to Equation (14), then
to errors in resistance measurements made during the
scale to equivalent temperature uncertainty according to
SPRT calibration at the tin and zinc points. Note too the
Equation (15).
Graphs like Figure 4 are very useful for identifying termined by differentiating Equation (16) with respect to
the most significant contributions to uncertainty at any the various measured resistances:
temperature. For example, in Figure 4 we can see that
the uncertainty in the water-triple-point value has little
influence except at temperatures below 40 °C, and the dWr =
1
RH2O
{(
dR − WdRH2O )
uncertainty in the tin-point value dominates in the tem- (18)
perature range from 20 °C to 320 °C. Note that the
( dW
)
N
graph gives only the uncertainty in the calibration of the − ∑ dR i − Wi dRH2O,i Fi (W )} r ,

i =2 dW
SPRT (the last three terms of Equation (14)); the term W =Wi
due to the uncertainty in the measured R value at the

unknown temperature (the first term in Equation (14)) where the first two terms are due to the measurements
must be added to the total in Figure 4. at the unknown temperature, the remaining terms are
due to measurements made during the calibration of the
SPRT at the fixed points, and by design ITS-90
0.8
has dW r / dW ≈ 1 . Note also that the terms involving
0.7
Total uncertainty (mK)
measurements at the triple point of water are now identi-

0.6
fied specifically. Equation (10) is a specific example of
0.5
Total
Equation (18)
0.4 Equation (18) identifies the Fi functions of Equa-
0.3
tion (16) as the sensitivity coefficients for uncertainties
Zn Sn associated with the respective fixed points. For all of the
0.2
sub-ranges between the mercury point (~−38 °C) and
0.1 H2O the aluminium point (~660 °C) the Fi functions are all
0 Lagrange polynomials. The various Fi functions are
0 50 100 150 200 250 300 350 400 450 tabulated in Appendix C. For the simplest and most
o
Temperature ( C) commonly used sub-ranges of ITS-90 the functions are
presented in full algebraic form. For the more complex
Figure 4: The calibration uncertainty in the water-tin-zinc sub- sub-ranges and where practical, a numerical approxima-
range due to uncertainties of 0.1 mK, 0.4 mK and 0.8 mK in the tion is given.
water, tin and zinc points respectively; the uncertainty in the The propagation of uncertainty formula has the gen-
unknown R value is assigned a value of zero. The same triple- eral form
point-resistance value is used to calculate all resistance ratios.
2
M  dz  2 M −1 M 
dz  dz 
u z2 = ∑  dx  u xi + 2 ∑ ∑   ri , j u x u x
4. Propagation of uncertainty – the i =1   i =1 j =i +1 dx i

 dx j
 i j
i 
general case (19)
All of the SPRT interpolation equations of the ITS-90
can be expressed in the form where z = z( x 1, x 2 ,..., x M ) is a function of the M meas-
N
ured variables xi with associated standard uncertainties
Wr (W ) = ∑ Wr ,i Fi (W ,W1,...,W N ) (16) u xi and correlation coefficients ri , j . The problem in
i =1 general is to assign the various variables in Equa-
tion (19) to those in the measurement equation and to
where W is the measured resistance ratio at the un- calculate the various partial derivatives. For the ITS-90
known temperature, W r,i are the reference values as- interpolations Equation (18) is the fundamental equation
signed to the N fixed points, Wi are the values of W for identifying the various sensitivity coefficients for
measured at the fixed points, and Fi are functions of propagating uncertainties. Any propagation-of-
measured resistance ratio only. The first fixed point is uncertainty equation for the ITS-90 equations must be
always the triple point of water so that W1 = Wr,1 = 1 by calculated as follows:
definition.
The Fi functions identified in Equation (16) have the (i) Identify the appropriate interpolation equation and
property that sensitivity coefficients (Appendix B and C).
(ii) Derive, by simplifying Equation (18), the appropriate
differential equation corresponding to the interpola-
 1, for i = j tion equation.
Fi (W j ,W1,...,WN ) =  (17)
0 for i ≠ j . (iii) Depending on assumptions about the measurement
(e.g. use of water triple-point values, proximity of
That is, the functions take the value 1.0 at the fixed unknown temperatures to fixed-point values) sim-
point after which they are named, and are zero at all plify Equation (18) and collect like terms.
other fixed points used in the interpolation. The second (iv) Now write down the uncertainty propagation equa-
order Lagrange polynomials given by Equation (7) and tion by replacing the dR terms by U terms and
shown in Figure 2 are specific examples of functions squaring all collected terms. Include and evaluate
satisfying this property. any correlations between the various UR values.
The effects of small errors in the various measured
resistances dRi on the interpolated value of Wr, is de- There are many possible specific cases of Equa-
tion (18) that can be considered depending on the num-
ber of triple-point measurements made, the number of which is the equation we should have expected. The
different triple-point cells used, the number of different temperature difference between the two cells is
bridges used, and the degree of correlation between the
various uncertainties. All of these factors alter the de-  dT 
gree of correlation between the various contributing un- ∆T = T − TSn =  90 ∆W
certainties and change the total uncertainty.  dW r 
There are two mathematical identities that are occa- (24)
sionally useful when simplifying some of the expres-  dT  R − R Sn 
.
=  90 
sions. The first is  dWr  RH2O 

N
∑ Wi Fi (W ) = W , (20) The uncertainty in the Wr value for the new cell is
given by the propagation of uncertainty equation with
i =1
the correlation term included:
which was used in Equation (20). A second identity,
2
which is useful for calculating the FH2O ( W ) functions for  1 
low temperature interpolations, is
2
UWr =
 RH O  R (
 U 2 + U 2 − 2rU U
R ,Sn R R ,Sn ) (25)
 2 
N
∑ Fi (W ) = 1 . (21) where r is the coefficient characterising the correlation
i =1 between the two uncertainties. Now, a further simplifica-
tion is possible if the uncertainties in the two resistance
In the following section we consider a number of measurements have the same magnitude, which is quite
specific examples of the use of the equations. likely. Then the uncertainty in the temperature difference
is
5. Propagation of uncertainty – more
examples  dT  U R 
(2(1 − r ))1/ 2 ,
U ∆T =  90 
 R (26)
 dW r  H2O 

5.1 The simplest case – a fixed-point compari-
son
which shows the advantage of using the same resis-
Let us suppose that we are determining the tempera- tance bridge and measurement procedures to ensure a
ture of a sealed tin-point cell, and are using an SPRT high correlation. Note that this is the uncertainty in the
calibrated at the water, tin and zinc points. The relevant temperature difference between the cells. The uncer-
interpolation equation is Equation (6): tainty in the temperature of the new cell must include the
uncertainty in the temperature of the reference cell.
W r = LH2O (W ) + W r,Sn LSn (W ) + W r,Zn L Zn (W ) . Equations (24) and (26) can be generalised to any of the
fixed points.
Now we calculate the partial derivative, Equa- 5.2 Another simple case − the same water-
tion (10):
triple-point value used throughout
1 This is a generalisation of the example considered in
dWr = (dR − WdRH O − (dR Sn − WSndRH O,Sn )LSn Section 3 for the water-tin-zinc sub-range. The formulae
RH2O 2 2
for the other interpolation ranges and intermediate cor-
− (dR Zn − W Zn dRH2O )LZn ) . relations of triple-point values follow similarly, as shown
below. When the same water-triple-point value is used
Because we are making measurements very near to to calculate all W values, Equation (18) simplifies to
the tin point, we can assume that W = WSn. Under this
condition, the Lagrange polynomials LSn(WSn) = 1, and 1  N
LZn(WSn) = 0 (see Figure 2 and Equation (17)). Hence

dWr =
RH2O 
dR − ∑ Fi dR i
i =2
(27)
1 N 
dWr = (dR − WSn dRH O + dRH2O  ∑ W i Fi − W  ,
RH2O 2
(22)  i =2 
− dR Sn + WSn dRH2O,Sn . )
and hence, by using the identity of Equation (20) to sim-
plify the term in parentheses, this simplifies to
From this equation we can recognise that by using
the same water-triple-point value throughout the meas-
1  N 
urements the terms in dRH2O cancel, hence dW r = dR − ∑ dR i Fi  . (28)
RH2O  i =1 
1
dW r = (dR − dR Sn ) , (23) Equation (13) for the water-tin-zinc sub-range is a
RH2O
specific case of Equation (28).
If it is assumed that the various uncertainties con- that associated with the water triple point, and the re-
tributing to Equation (28) are uncorrelated then the total producibility of triple-point measurements is high so the
uncertainty in the calculated value of Wr is correlation coefficient is high, perhaps about 0.8.
2 5.4 An SPRT calibrated for a client

 1   2 N 2 2 
UW2 r =  U R + ∑ Fi U R,i  . (29) When a national measurement institute (NMI) cali-
 RH O    brates a SPRT for a client, the value of W calculated by
 2   i =1 
the client is based on a different water-triple-point cell,
Equation (14) is a specific example of this equation. different measurement procedures, and different resis-
The equation provides a good approximation of the total tance bridges to those of the NMI. Accordingly the corre-
uncertainty typically achieved in a national standards lation between the uncertainties in the water-triple-point
laboratory. The assumption that all of the triple-point value used by the client and those used by the NMI dur-
values are the same is a good approximation so long as ing the calibration of the SPRT will be close to zero. If
the same triple-point cell, the same procedures, and the we assume the magnitude of the triple-point uncertain-
same resistance bridge are used for all measurements. ties are the same, and the triple-point measurements
This ensures that the uncertainties in the various triple- carried out by the NMI are all highly correlated, then the
point measurements are highly correlated. total uncertainty will be
5.3 A SPRT calibrated using different bridges, 2

 1  2
fixed points and procedures u 2
Wr =
 RH O 
[
 u R + W 2uH2 O
2
If distinct triple-point values are used to calculate  2  (32)
each W value, and none of the contributing uncertainties
]
N
are correlated then the uncertainty propagation law fol- + ∑ Fi 2u i2 + (F H2O −W )2 uH22O .
lowing directly from Equation (18) is i =2
2 The first triple-point term is due to the uncertainty in

 1  2
u 2
Wr =  u + W 2uH2 O
 RH O  R 2
[ the client’s measurement of the triple-point resistance,
the second triple-point term is that due to the uncertainty
 2  (30) in the NMI’s measurements of the triple-point resistance.
( ) ]
N
There are three features of the behaviour of Equa-
+ ∑ u i2 + Wi 2u i,2H2O Fi 2 . tion (32) worth noting. Firstly, for temperatures close to
i =2
the water triple point FH2O ≈ W ≈ 1 , so that the influence
In this case, which is applicable when separate and of the NMI’s measurement at temperatures near the
uncorrelated triple-point measurements are made after triple point of water is negligible. Secondly, at tempera-
each fixed point measurement (perhaps to compensate tures well away from the triple point of water FH2O ≈ 0 so
for oxidation effects in the SPRTs), all of the uncertain-
the effect of uncertainties in the triple-point values in-
ties associated with the water-triple-point measurements
creases in proportion to W, with both the NMI’s and the
propagate in proportion to W. This is not a very likely
client’s measurement contributing equally. Thirdly, it
scenario since measurements made by a single labora-
can also be seen that in the event of the high correlation
tory would almost certainly exhibit some correlation,
between the client’s and NMI’s triple-point values the
especially the water-triple-point measurements.
terms in W cancel and yield Equation (30).
Instead consider a more realistic scenario. If we as-
sume that all of the triple-point uncertainties have the
same magnitude and are correlated with a single corre-
lation coefficient r, then the uncertainty propagates ac- References
cording to H Preston-Thomas, “The International Temperature
scale of 1990, Metrologia, 27, 3, 3-10, 1990
2
 1  Supplementary information for the International tem-
 2 N
u 2
= 
u R + ∑ u Ri Fi + rFH2O u RH2O
2 2 2 2
perature scale of 1990, BIPM Paris, 1990
Wr
 RH O   Guide to the Expression of Uncertainty in Measure-
 2  i =2
(31) ment, International Organisation for Standardisation,
 N
  Geneva, 1993
+ (1 − r )W 2 + ∑ Wi 2 Fi 2 u R2H O  .
  2 D R White and P Saunders, “ Propagation of Uncer-
i =2 
tainty on Interpolated scales, with examples from Ther-
mometry. Metrologia, 37, 4, 285-293, 2000
For complete correlation between triple-point uncer- D R White, The Propagation of Uncertainty with non-
tainties (r = 1), Equation (31) simplifies to Equation (29) Lagrangian Interpolation. Metrologia, 38, 3, 63-69, 2001
and for zero correlation (r = 0) it simplifies to Equa-
tion (30). Prepared by D R White, June 2002
For temperature ranges above the triple point of wa-
ter, Equation (29) is a good practical approximation to
Equation (31). This is because the uncertainties associ-
ated with the other fixed points are much greater than
Appendix A – Fixed points of ITS-90 and reference resistance ratios for fixed
points
Temperature
Substance T90/K t90/°°C Wr(T90)
(Note 1, 2)
e-H2 (T) 13.8033 −259.3467 0.001 190 07
3
e-H2(V) or He (G) or ≈17 ≈−256.15
4
He(G)
3
e-H2(V) or He (G) or ≈20.3 ≈−252.85
4
He(G)
Ne (T) 24.5561 −248.5939 0.008 449 74
O2 (T) 54.3584 −218.7916 0.091 718 04
Ar (T) 83.8058 −189.3442 0.215 859 75
Hg (T) 234.3156 −38.8344 0.844 142 11
H2O (T) 273.16 0.01 1.000 000 00
Ga (M) 302.9146 29.7646 1.118 138 89
In (F) 429.7485 156.5985 1.609 801 85
Sn (F) 505.078 231.928 1.892 797 68
Zn (F) 692.677 419.527 2.568 917 30
Al (F) 933.473 660.323 3.376 008 60
Ag (F) 1234.93 961.78 4.286 420 53
Notes:
3
1 All substances except He are of natural isotopic composition: e-H2 is hydrogen at the equilibrium concentration
of the ortho- and para-molecular forms.
2. The symbols have the following meanings: V: vapour-pressure point; T: triple point; G: gas-thermometer point;
M, F: melting point, freezing point (temperature, at a pressure of 101 325 Pa, at which the solid and liquid
phases are in equilibrium).
Appendix B – Deviation functions for the SPRT sub-ranges of ITS-90
Temperature range Deviation function Fixed points

5
a(W − 1) + b(W − 1) 2 + ∑ c i [ln(W )]
13.8033 K to 0.01 °C 2+i e-H2, Ne, O2, Ar, Hg,
i =1
Note (1)
3
24.5561 K to 0.01 °C a(W − 1) + b(W − 1) 2 + ∑ c i [ln(W )]
i
e-H2, Ne, O2, Ar, Hg
i =1
a(W − 1) + b(W − 1) 2 + c [ln(W )]

2
54.3584 K to 0.01 °C O2, Ar, Hg
83.8058 K to 0.01 °C a(W − 1) + b(W − 1) ln(W ) Ar, Hg
–38.8344 °C to 29.7646 °C a(W − 1) + b(W − 1) 2 Hg, Ga
0.01 °C to 29.7646 °C a(W − 1) Ga
0.01 °C to 156.5985 °C a(W − 1) In
0.01 °C to 231.928 °C a(W − 1) + b(W − 1) 2 In, Sn
0.01 °C to 419.527 °C a(W − 1) + b(W − 1) 2 Sn, Zn
0.01 °C to 660.323 °C a(W − 1) + b(W − 1) 2 + c (W − 1) 3 Sn, Zn, Al
0.01 °C to 961.78 °C [
a(W − 1) + b(W − 1) 2 + c (W − 1) 3 + d W − W (660.323 o C) ]
2 Sn, Zn, Al, Ag,
Note (2)
Note (1): For the sub-range 13.8033 K to 0.01 °C, two of the calibration points are points determined by gas ther-
mometer or vapour-pressure thermometer.
Note (2): For the sub-range 0.01 °C to 961.78 °C the coefficients a, b, c are the same as used for the sub-range
0.01 °C to 660.323 °C, and the coefficient d is determined from the silver point.
Appendix C Sensitivity coefficients for the ITS-90 sub-ranges
C.1 Water-Gallium
The ITS-90 equation for these two sub-ranges takes the form
Wr (W ) = W − a(W − 1)
where the a coefficient is determined by calibration of the SPRT at the melting point of gallium. The interpolation is
rewritten
(W − W Ga ) (W − 1)
W r (W ) = + W r,Ga
(1 − W Ga ) (W Ga − 1)
from which the sensitivity coefficients are identified as
(W − W Ga )
FH2O (W ) =
(1 − W Ga )
(W − 1)
FGa (W ) = .
(WGa − 1)
1.2
Propagated uncertainty (mK)
H 2O Ga
1.0
0.8
0.6
0.4
0.2
0.0
0 5 10 15 20 25 30
Temperature (oC)
Figure C1: Sensitivity coefficients for the water-gallium sub-range. The curves show the propagated uncertainty in millikelvin for a
1 mK uncertainty at each of the fixed points
C.2 Water-Indium
The ITS-90 interpolation equation takes the form
Wr (W ) = W − a(W − 1)
where the a coefficient is determined by calibration of the SPRT at the freezing point of indium. The interpolation
equation can be written in the form
(W − WIn ) (W − 1)
W r (W ) = + W r,In
(1 − WIn ) (WIn − 1)
(W − WIn )
FH2O (W ) =
(1 − WIn )
(W − 1)
FIn (W ) =
(WIn − 1)
1.2
H2O
In
1.0
0.8
0.6
0.4
0.2
0.0
0 20 40 60 80 100 120 140 160
o
Temperature ( C)
Figure C2: The sensitivity coefficients for the water-indium sub-range. The curves show the propagated uncertainty in millikelvin for
a 1 mK uncertainty at each of the fixed points.
C.3 Mercury-Gallium
W r (W ) = W − a(W − 1) − b(W − 1) 2
where the a and b coefficients are determined by calibration of the SPRT at the triple point of mercury and the melt-
ing point of gallium. The interpolation equation may be written in the form
(W − 1)(W − W Ga ) (W − WHg )(W − WGa ) (W − WHg )(W − 1)

Wr (W ) = Wr,Hg + + Wr,Ga
(WHg − 1)(WHg − W Ga ) (1 − WHg )(1 − W Ga ) (W Ga − WHg )(WGa − 1)
(W − 1)(W − WGa )
FHg =
(WHg − 1)(WHg − W Ga )
(W − WHg )(W − W Ga )
FH2O =
(1 − WHg )(1 − W Ga )
(W − WHg )(W − 1)
FGa =
(WGa − WHg )(W Ga − 1)
1.2
Hg H2O Ga
1.0
0.8
0.6
0.4
0.2
0.0
-0.2
-0.4
-40 -30 -20 -10 0 10 20 30
o
Temperature ( C)
Figure C3: The sensitivity coefficients for the mercury-gallium sub-range. The curves show the propagated uncertainty in millikelvin
for a 1 mK uncertainty at each of the fixed points.
C.4 Water-tin
W r − W − a(W − 1) − b(W − 1) 2
where the a and b coefficients are determined by calibration of the SPRT at the freezing points of indium and tin. The
interpolation equation can then be written in the form
(W − WIn )(W − W Sn ) (W − 1)(W − W Sn ) (W − 1)(W − WIn )

W r (W ) = + W r,In + W r,Sn
(1 − WIn )(1 − W Sn ) (WIn − 1)(WIn − W Sn ) (W Sn − 1)(W Sn − WIn )
(W − WIn )(W − WSn )

FH2O =
(1 − WIn )(1 − W Sn )
(W − 1)(W − W Sn )
FIn =
(WIn − 1)(WIn − WSn )
(W − 1)(W − WIn )
FSn =
(W Sn − 1)(W Sn − WIn )
1.2
H 2O In Sn
1.0
0.8
0.6
0.4
0.2
0.0
-0.2
-0.4
0 50 100 150 200 250
Temperature (oC)
Figure C4: The sensitivity coefficients for the water-tin sub-range. The curves show the propagated uncertainty in millikelvin for a
1 mK uncertainty at each of the fixed points.
C.5 Water-zinc
Wr = W − a(W − 1) − b(W − 1) 2
where the a and b coefficients are determined by calibration of the SPRT at the freezing points of tin and zinc. The
interpolation equation can then be written in the form
(W − WSn )(W − W Zn ) (W − 1)(W − W Zn ) (W − 1)(W − W Sn )

Wr (W ) = + Wr,Sn + Wr,Zn
(1 − WSn )(1 − W Zn ) (WSn − 1)(WSn − W Zn ) (W Zn − 1)(W Zn − WSn )
(W − W Sn )(W − W Zn )
FH2O =
(1 − W Sn )(1 − W Zn )
(W − 1)(W − W Zn )
FSn =
(W Sn − 1)(WSn − W Zn )
(W − 1)(W − WSn )
FZn =
(W Zn − 1)(W Zn − W Sn )
1.2
H2 O Sn Zn
1.0
0.8
0.6
0.4
0.2
0.0
-0.2
-0.4
0 100 200 300 400
o
Temperature ( C)
Figure C5: The sensitivity coefficients for the water-zinc sub-range. The curves show the propagated uncertainty in millikelvin for a
1 mK uncertainty at each of the fixed points.
C.6 Water-aluminium
Wr (W ) = W − a(W − 1) − b(W − 1) 2 − c (W − 1) 3
where the a, b and c coefficients are determined by calibration of the SPRT at the freezing points of tin, zinc and
aluminium. The interpolation equation can then be written in the form
Wr (W ) = FH2O + Wr,Sn FSn + Wr,Zn FZn + Wr,Al FAl
where the sensitivity coefficients are
(W − WSn )(W − W Zn )(W − W Al )

FH2O =
(1 − WSn )(1 − W Zn )(1 − W Al )
(W − 1)(W − W Zn )(W − W Al )
FSn =
(W Sn − 1)(WSn − W Zn )(WSn − W Al )
(W − 1)(W − WSn )(W − W Al )
FZn =
(W Zn − 1)(W Zn − WSn )(W Zn − W Al )
(W − 1)(W − WSn )(W − W Zn )
FAl =
(W Al − 1)(W Al − WSn )(W Al − W Zn )
1.2
H2 O Sn Zn Al
1
0.8
0.6
0.4
0.2
0
-0.2
-0.4
-0.6
0 200 400 600
o
Temperature ( C)
Figure C6: The sensitivity coefficients for the water-aluminium sub-range. The curves show the propagated uncertainty in millikelvin
C.7 Water-silver
Wr (W ) = W − a(W − 1) − b(W − 1) 2 − c (W − 1) 3 − d (W − W Al ) 2
where the a, b and c coefficients are the same as determined from the water-aluminium sub-range and the d coeffi-
cient is determined by calibration of the SPRT at the freezing point of silver. The equation is applicable only over the
temperature range between the aluminium and silver points (660.323 °C to 961.78 °C). The interpolation equation
can then be written in the form
Wr (W ) = FH2O + Wr,Sn FSn + Wr,Zn FZn + Wr,Al FAl + Wr,Ag FAg
where the sensitivity coefficients (applicable only above 660 °C) are
2
 W − W Al 
FH2O = FHAl2O (W ) − FHAl2O (W Ag ) 
 W Ag − W Al 
 
2
 W − W Al 
FSn = Al
FSn (W Al
) − FSn (W Ag ) 
 
 W Ag − W Al 
2
 W − W Al 
Al
FZn = FZn Al
(W ) − FZn (W Ag ) 
 
2
 W − W Al 
FAl = FAlAl (W ) − FAlAl (W Ag ) 
 
2
 W − W Al 
FAg = 
 
where the Fi Al are the sensitivity coefficients for the water–aluminium sub-range, as given in Appendix C6.
1.5
H2O Sn Zn Al Ag
1.0
0.5
0.0
-0.5
-1.0
0 200 400 600 800 1000
Temperature (oC)
Figure C7: The sensitivity coefficients for the water-silver sub-range. The curves plot the propagated uncertainty in millikelvin for a
1 mK uncertainty at each of the fixed points. Below 660 °C (the aluminium point) the curves are identical to those in Figure C6.
C.8 Argon-water
Wr = W − a(W − 1) − b(W − 1) ln(W )
where the a, and b coefficients are determined by calibration of the SPRT at the triple points of argon and mercury.
The interpolation can be written in the form
Wr (W ) = FH2O + Wr,Hg FHg + Wr,Ar FAr
where the sensitivity coefficients are
(W − WH2O )(WHg − W Ar ) ln W + (WHg − WH2O )(W Ar − W ) ln WHg − (WHg − W )(W Ar − WH2O ) ln W Ar

FH2O =
(WHg − WH2O )(W Ar − WH2O )(ln WHg − ln W Ar )
(W − WH2O )(ln W − ln W Ar )
FHg =
(WHg − WH2O )(ln WHg − ln W Ar )
(W − WH2O )(ln W − ln WHg )
FAr =
(W Ar − WH2O )(ln W Ar − ln WHg )
The sensitivity coefficient for uncertainties in the triple point of water is most easily calculated as
FH2O = 1 − FHg − FAr .
2.0
1.5
Ar Hg H2 O
1.0
0.5
0.0
-0.5
-1.0
-1.5
50 100 150 200 250 300
Temperature (K)
Figure C8: The sensitivity coefficients for the argon-water sub-range. The curves show the propagated uncertainty in millikelvin for
a 1 mK uncertainty at each of the fixed points.
C.9 Oxygen-water
Wr (W ) = W − a(W − 1) − b(W − 1) 2 − c (ln W )

2
where the a, b and c coefficients are determined by calibration of the SPRT at the triple points of oxygen, argon and
mercury. The interpolation can be written in the form
Wr = FH2 0 + Wr,Hg FHg + Wr,Ar FAr + Wr,O2 FO2
where the sensitivity coefficient for the oxygen point is
(lnW ) 2 (WHg − 1)(W Ar − 1)(W Ar − WHg ) − (lnW Ar ) 2 (WHg − 1)(W − 1)(W − WHg ) + (lnWHg ) 2 (W − 1)(W Ar − 1)(W − W Ar )
FO2 =
(lnWO2 ) 2 (WHg − 1)(W Ar − 1)(W Ar − WHg ) − (lnW Ar ) 2 (WHg − 1)(W O2 − 1)(WO2 − WHg ) + (lnWHg ) 2 (WO2 − 1)(W Ar − 1)(WO2 − W Ar )
The equations for the argon and mercury functions can be found by permuting the indices, and the water function
found from FH2O = 1 − FO2 − FAr − FHg . Alternatively, the following numerical approximations may be used.
FO2 = −1.598425679 + 3.403991489W − 1.805565809W 2 + 0.4032254894 (ln W )

2
FAr = +3.930746333 − 8.465193581W + 4.534447249W 2 − 0.5593488293 (ln W )

2
FHg = −4.044936086 + 16.33613127W − 12.29119519W 2 + 0.4643043265 (ln W )

2
FH2O = +2.712615432 − 11.27492918W + 9.562313751W 2 − 0.3081809866 (ln W )

2
1.5
O2 Ar Hg H2O
1
0.5
-0.5
-1
50 100 150 200 250 300
Temperature (K)
Figure C9: The sensitivity coefficients for the oxygen-water sub-range. The curves show the propagated uncertainty in millikelvin
C.10 Neon-water
Wr = W − a(W − 1) − b(W − 1) 2 − c ln W − d (lnW ) 2 − e(ln W ) 3
where the a, b, c, d, and e coefficients are determined by calibration of the SPRT at the triple points of hydrogen,
neon, oxygen, argon and mercury. The interpolation can be written in the form
Wr = FH2O + Wr,Hg FHg + Wr,Ar FAr + Wr,O2 FO2 + Wr,Ne FNe .
The sensitivity coefficients in this case are too complex to write down in algebraic form. The following numerical ap-
proximations may be used.
FHg = −24.30265740 + 42.99793720W − 18.69528095W 2 − 14.20610018 ln(W ) − 2.778961721 ln 2 (W ) − 0.1788181624 ln 3 (W )

FAr = +33.0297966 8 − 46.64306946W + 13.61327293W 2 + 20.51522598 ln(W ) + 4.195086350 ln 2 (W ) + 0.2785479955 ln 3 (W )
FO2 = −29.45087407 + 39.79347654W − 10.34260259W 2 − 19.77485453 ln(W ) − 4.269113231 ln 2 (W ) − 0.2941747479 ln 3 (W)
FNe = +7.413452875 − 9.508661062W + 2.095208241W 2 + 5.423905534 ln(W ) + 1.384488984 ln 2 (W ) + 0.1102290468 ln 3 (W )

FH = −1.715380724 + 2.160321157W − 0.4449404100W 2 − 1.291145588 ln(W ) − 0.3520630280 ln 2 (W ) − 0.0326933541 5 ln 3 (W)
FH2O = +16.02566264 − 28.80000438W + 13.77434278W 2 + 9.33296879 ln(W ) + 1.820562646 ln 2 (W ) + 0.1169092222 ln 3 (W )
2.0
1.5
H Ne O2 Ar Hg H 2O
1.0
0.5
0.0
-0.5
-1.0
-1.5
0 50 100 150 200 250 300
Temperature (K)
Figure C10: The sensitivity coefficients for the neon-water sub-range. The curves plot the propagated uncertainty in millikelvin for a
1 mK uncertainty at each of the fixed points. Note that the sub-range extends only to the Neon point (24.5561 K) although the hy-
drogen point is below this temperature.
C.11 Hydrogen-water
W r = W − a(W − 1) − b(W − 1) 2 − c (ln W )3 − d (ln W ) 4 − e(ln W )5 − f (ln W ) 6 − g (ln W ) 7
where the a, b, c, d, e, f, and g coefficients are determined by calibration of the SPRT at the triple points of hydrogen,
neon, oxygen, argon and mercury, and two other points determined by either gas thermometer or vapour pressure
thermometer near 17 K and 20.3 K.
The sensitivity coefficients in this case are too complex to write down in algebraic form, and also depend on the tem-
peratures chosen for the two auxiliary measurements. The graph below shows the sensitivity coefficients when the
auxillary points are exactly 17 K and 20.3K
1.5
H V1 V2 Ne O2 Ar Hg H2O
1
0.5
-0.5
-1
-1.5
-2
10 100
Temperature (K)
Figure C11: The sensitivity coefficients for the hydrogen-water sub-range. The curves plot the propagated uncertainty in millikelvin
for a 1 mK uncertainty at each of the fixed points. V1 and V2 are the two auxiliary fixed points measured by gas thermometer or
vapour pressure thermometer.

SPRT - Its-90 TG3

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

SPRT - Its-90 TG3

Uploaded by

Copyright:

Available Formats

MSL Technical Guide 3

Measurement Standards Laboratory Propagating uncertainty in the

Step 1: Calculate the resistance ratio at the un-

Zinc Points (W − WH2O )(W − W Zn )

Note that we have explicitly included the variables

fined by the ITS-90. This function is defined in two seg-

ments, one for temperatures below 0 °C and one for

3. Propagation of uncertainty – a spe- -0.5

cific case W value

Let us continue with the water-tin-zinc interpolation

dW r = dW − LSn (W )dW Sn − LZn (W )dW Zn (8) so that

where the dW symbols (differentials) represent small 1

− LZn (W )(dR Zn − W Zn dRH2O,Zn ) . ]

graph gives only the uncertainty in the calibration of the − ∑ dR i − Wi dRH2O,i Fi (W )} r ,

due to the uncertainty in the measured R value at the

measurements at the triple point of water are now identi-

LZn(WSn) = 0 (see Figure 2 and Equation (17)). Hence

2 5.4 An SPRT calibrated for a client

5.3 A SPRT calibrated using different bridges, 2

2 The first triple-point term is due to the uncertainty in

Temperature range Deviation function Fixed points

a(W − 1) + b(W − 1) 2 + c [ln(W )]

83.8058 K to 0.01 °C a(W − 1) + b(W − 1) ln(W ) Ar, Hg

–38.8344 °C to 29.7646 °C a(W − 1) + b(W − 1) 2 Hg, Ga

0.01 °C to 29.7646 °C a(W − 1) Ga

0.01 °C to 156.5985 °C a(W − 1) In

0.01 °C to 231.928 °C a(W − 1) + b(W − 1) 2 In, Sn

0.01 °C to 419.527 °C a(W − 1) + b(W − 1) 2 Sn, Zn

0.01 °C to 660.323 °C a(W − 1) + b(W − 1) 2 + c (W − 1) 3 Sn, Zn, Al

from which the sensitivity coefficients are identified as

The ITS-90 interpolation equation takes the form

from which the sensitivity coefficients are identified as

The ITS-90 interpolation equation takes the form

(W − 1)(W − W Ga ) (W − WHg )(W − WGa ) (W − WHg )(W − 1)

from which the sensitivity coefficients are identified as

The ITS-90 interpolation equation takes the form

(W − WIn )(W − W Sn ) (W − 1)(W − W Sn ) (W − 1)(W − WIn )

from which the sensitivity coefficients are identified as

(W − WIn )(W − WSn )

The ITS-90 interpolation equation takes the form

(W − WSn )(W − W Zn ) (W − 1)(W − W Zn ) (W − 1)(W − W Sn )

from which the sensitivity coefficients are identified as

The ITS-90 interpolation equation takes the form

Wr (W ) = FH2O + Wr,Sn FSn + Wr,Zn FZn + Wr,Al FAl

where the sensitivity coefficients are

(W − WSn )(W − W Zn )(W − W Al )

The ITS-90 interpolation equation takes the form

Wr (W ) = FH2O + Wr,Sn FSn + Wr,Zn FZn + Wr,Al FAl + Wr,Ag FAg

The ITS-90 interpolation equation takes the form

Wr = W − a(W − 1) − b(W − 1) ln(W )

Wr (W ) = FH2O + Wr,Hg FHg + Wr,Ar FAr

where the sensitivity coefficients are

(W − WH2O )(WHg − W Ar ) ln W + (WHg − WH2O )(W Ar − W ) ln WHg − (WHg − W )(W Ar − WH2O ) ln W Ar

FH2O = 1 − FHg − FAr .

The ITS-90 interpolation equation takes the form

Wr (W ) = W − a(W − 1) − b(W − 1) 2 − c (ln W )

Wr = FH2 0 + Wr,Hg FHg + Wr,Ar FAr + Wr,O2 FO2

where the sensitivity coefficient for the oxygen point is

FO2 = −1.598425679 + 3.403991489W − 1.805565809W 2 + 0.4032254894 (ln W )

FAr = +3.930746333 − 8.465193581W + 4.534447249W 2 − 0.5593488293 (ln W )

FHg = −4.044936086 + 16.33613127W − 12.29119519W 2 + 0.4643043265 (ln W )

FH2O = +2.712615432 − 11.27492918W + 9.562313751W 2 − 0.3081809866 (ln W )