FUNDAMENTALS
OF
THERMOMETRY
PART
INDUSTRIAL
PLATINUM
V
<USE)
RESISTANCE
THERMOMETERS
by
COMMON
TEMPER.xXTURE
Henry
ERRORS
by
INDUSTRIAL
PLATINUM
E. Sostmann,
and
IN
INDCJSTRIAL
MEASUREMENT
John
RESISTANCE
P. Tavener
THERMOMETERS
ABSTRACT
The interpolation
Scale
of 1990,
standard
over
the
stipulated
temperature
in the
range
International
from
-189’
Temperto +962-C,
ature
is a long-stem
Standard
Platinum
Resistance
Thermometer
(SPRT);
or
rather,
a pair
of SPRTs,
since
no one thermometer
can encompass
the
entire
range
[ll. Some characteristics
required
of an SPRT
make
it unsuitable
and
unfit
for
use outside
the standardising
or calibration
laboratory;
therefore
a variety
of industrial-grade
platinum
resistance
thermometers,
standards,
of high
and in
quality,
have
been developed
applications
such
as process
for
use
monitoring
as
field
working
and control.
DISCUSSION
The
constrains
that
ITS(90)
121, are
the gallium
point
and
and
specific
the
water
triple
on
the
SPRT,
stipulated
in the
text
of
the
it be strain-free,
that
the
ratio
of resistances
at
the water
triple
point,
or the mercury
triple
point
point,
Wga=
whg
and,
in
(962-C),
sistances
addition,
that
the
Rga/Rwt
= Rhg/Rwt
be
2 1.118 07
2 0.844235
if the
thermometer
ratio
of the silver
is
point
Eq.
Eq.
to
be used
to
and the water
1
2
the silver
point
triple
point
re-
be
wag
= Rag/Rtp
2 4.2844
Eq.
3
6
(These
IPTS(68),
constraints
that
the
replace
ratio
of
W = RlOO/R,
Eq.
the
requirement
the resistances
Z .0032950
at
of the
previous
100’ and at O’C
Scale,
be
the
)
4
These
requirements
attempt
to specify
the
tance
element
der
to conform
are
a well-meant
and
strain-freedom
purity
is not difficult
to think
of a better
formulation).
the
platinum
wire
must
to this
requirement,
and
it must
be mounted
on
some
support
in
rectly,
that
if an SPRT is put
that
it can be heard
to touch
Ironically,
it may be strained
from
a Laboratory
garding
the quality
e.g.,
unfortunately
of the
platinum
indirect
resis-
(it
ideally
pure,
which
(contradictorily)
provides
only
straint.
Thus,
while
the
well-designed
instrument
of extraordinary
sensitivity
and
can
be easily
knocked
out
of
erature;
but
be
a
In oralmost
fashion
a minimum
of support,
and no conand
constructed
SPRT
[31 is an
and repeatability,
it is delicate,
calibration.
It has
been
said,
cor-
down
onto
a surface
with
enough
force
so
it, it may be strained
out of calibration.
due to shipping
shock
as it is returned
has
been
newly
calibrated!
Precautions
reof SPRTs
are provided
throughout
the lit-
where
it
assurance
[41.
In addition,
SPRTs
are
physically
the platinum
element
inside
a sheath
forgiving
material,
in that
its elastic
fragile.
of fused
limit
and
The best
of them
quartz.
Quarts
is
its breaking
point
contain
an unare the
same,
but its use
is justified
by its physical
tolerance
of high
temperatures
(its
softening
point
is
about
15OO’C),
its
available
purity,
its
transparency,
which
allows
the
platinum
element
to be viewed,
its
impermeability
to gasses
(except
under
special
circumstances)
and its obvious
indication
that
it and the construction
it contains
are fragile
and
Metallic
sheaths,
on the other
hand,
can
must
be treated
with
respect.
emit vapors
which
are poisonous
to the
platinum
element,
and provide
a
One
manufacturer
provides
a quarts
spurious
sense
of
robustness.
sheath
inside
a metal
sheath,
with
the net effect
that
the protective
inside
ment
quarts
exposed
user
that
the
“INDUSTRIAL
this
I have
heading
mometers
is
other
fields
can be broken
to the
metallic
system
GRADE”
provided
to indicate
is
by
a slight
bending,
with
no
vapors,
and
visible
the platinum
eleindication
to the
corrupt.
RESISTANCE
quotation
customary
by no means
limited
of science,
technology,
THERMOMETERS
marks
around
terminology.
to
industry,
and broad
the word
The use
“Industrial”
of such
but
extends
into
common
application.
in
ther-
numerous
7
Industrial
resistance
thermometers
where
the compromise
is struck
between
stability)
that
it be as free from
strain
and the requirement
that
it be rugged
for which
it is intended.
(IRTs)
are compromise
devices,
the requirement
(for resistance
as may be compatible
with
use,
and durable
in the environment
It should
be mentioned,
briefly,
that
not all industrial
thermometers
are made of platinum
wire.
Other
constructions
resistance
are:
(a) Nickel,
copper,
and alloy
wire.
Nickel
and copper
wire
were
once very
popular
as thermometer
materials.
Both
have
temperature
coefficients
of resistance
higher
than
that
of platinum;
approximately
6.9 x low3 and 6.5 x low3 Q/Q/-C,
respectively,
as opposed
to 3.9 x 10m3
Q/Q/‘C
for
platinum.
This
higher
change
of
resistance
with
temperature
reduced
the burden
on earlier
signal
amplifiers;
with
modern
solid
state
electronics,
the advantage
is
negligible.
Nickel
is highly
non-linear
and passes
through
a sharp
change
in coefficient
in the vicinity
of 370-C
where
its magnetic
characteristics
change.
Copper
has a very
low specific
resistance,
a disadvantage
for thermometry.
Both
materials,
since
they
are
base metals,
are more
susceptible
to contamination
and oxidation
than
platinum.
Both
cost less than
platinum,
but
the fraction
of
the cost of an IPRT
which
represents
platinum
wire is small.
Both
have
faded
from
popularity
as resistance
thermometer
materials
(although
copper
is still
employed
in some
in-slot
temperature
monitors
in the protective
circuits
for electric
motors).
(b) Platinum
discussion.
films,
thin
and
(c) Non-metallic
resistance
other
semiconductors.
These
this series.
“INDUSTRIAL”
Thus
an Industrial
PLATINUM
the
RESISTANCE
thick.
These
will
thermometers;
will be discussed
be
e.g.,
in
included
in
this
thermistors
and
a future
issue
of
THERMOMETERS
industrial
resistance
thermometer
Platinum
Resistance
Thermometer
of today
(IPRT).
is likely
The IPRT
has a longer
history
than
its most refined
relative,
SPRT.
Werner
van Siemens
is generally
acknowledged
to have made
first,
proposing
it in his Bakerian
Lecture
of 1871, and advancing
It
came
rapidly
into
use, largely
three-term
interpolation
algorithm.
cause
of its inventor’s
reputation
, and declined
as rapidly,
because
The
Siemens
thermometer
comprised
inherent
problems
of stability.
to be
the
the
a
beof
1
meter
of 0.1 mm (0.004 inch)
diameter
platinum
wire
lain or fire-clay
tube,
the whole
assembly
enclosed
protection.
(Siemens
also experimented
with
sensing
impregnated
with platinum
group
metals).
wound
on a porcein an iron
tube
for
elements
of ceramic
A committee
of the British
Association
for the Advancement
of Science
found
that
the
resistance
of the
Siemens
thermometer
increased
upon
each
heating,
making
it necessary
to calibrate
the
thermometer
each time it was used (and a calibration
is a use). The change
in resistance,
reported
to reach
15% after
repeated
heatings
to 9OO’C, was assigned
to chemical
alteration
in the platinum.
(Two probable
other
reasons for
increase
in resistance
are
(a) gradual
volatilization
of some
platinum,
resulting
in a decrease
in sectional
area of the wire,
and (b)
the growth
of intergranular
boundaries,
affecting
the conduction
mechanism at these points).
About
20 years
later,
Callendar,
and Callendar
and Griffiths,
revived
the platinum
thermometer
for laboratory
use over
moderate
temperature
ranges.
Callendar
found
that
the clay
substrate
was a major
cause
of the
variation
of resistance;
that
the
platinum
wire
“became
brittle
and stuck
to the clay”.
We can guess
now at gross
silica
contamination.
On the other
hand,
a mica strip,
that
the platinum
touched
only at the edges,
appeared
to be “perfect”
insulation
in that it did not
cause contamination.
Callendar
also stipulated
that
all joints
with
platinum
be autogenous
fusion
weldments
without
foreign
material
such as
solders;
that
pressure
joints
(screws
or torsion)
be avoided,
and that
copper
conductors
in the
heated
zone
be eschewed,
because
of the
volatility
of the material;
and we observe
these
strictures
today.
As a generality,
the work
of Callendar
and Griffith6
was confined
to the range
0’ to 550-C.
Over
this
range,
they
found
that
a third-order parabolic
equation
using
three
fixed
points,
ice,
steam,
and
the
boiling
point
of sulfur,
was adequate
to establish
an interpolation
scheme.
Of the famous
Callendar
equation,
which
was the basis for International
temperature
Scales
until
1968, more later.
PLATINUM
AS A THERMALLY
SENSITIVE
MATERIAL
The development
of the platinum
resistance
thermometer
paralleled,
in time,
the development
of platinum
itself
as a workable
and pure
material.
Callendar’s
platinum
was certainly
not equivalent
to pure
platinum
and the SPRT as we think
of it today.
years
Platinum,
of the
as first
refined,
occurs
as a spongy
19th century,
it was beyond
technology
mass.
In the mid
ta force this
mass
9
into
an ingot
by methods
which
retained
its
metal
was
obtained,
essentially,
by hammering.
into
an ingot
which
is then
further
compacted
and
annealings,
ready
to
niques
until
roll
into
for
generally
is
strip,
finally
and
managing
in
151.
OF
INDUSTRIAL
A corollary
of the
in the temperature
the
form
eventually
platinum,
proprietary
A CONFUSION
crease
it
to
from
the
purity,
and
the
Today,
it can
by successive
of
draw
wrought
be melted
swagings
a square-sectioned
into
sponge
wire.
to
bar
Modern
wrought
tech-
wire,
STANDARDS
improvement
coefficient
in
of
metallurgical
techniques
resistance
exhibited
by
The purer
the material,
the higher
the temperature
recall
that
most
useful
coefficient,
a, a regrettable
and its subsequent
replacement
Scales:
is
the
coefficient.
victim
of the
remain
10m3
duction
SPRTs
mum a permitted
use
Industrial
in Europe
certainly
and
control
the
migration
after
the
war
precise
devices
400 process-control
ever
wire,
way
purity
always
turned
than
sensitivity
O’C and
of
approaching
the absence
best,
of
has
a
States.
but
European
Western
process
IPRTs
for
hardly
were
not strain
be stable
under
higher
The
ideal,
process
U.
of
x 10M3
as
to
3.928 x
of pro-
(the
commonplace
This
period
consequence,
memory,
at
wire
that
could
engineers
Hemisphere
thermocouples.
a duPont
than
best
mini-
much
longer
history
I cannot
trace
this
were
century.
platinum,
and,
in
fixed,
before
recent
placed,
at that
time,
in the
and achieved
a coefficients
these
elements
der that
they
ideal,
a is
of strain).
thermometers
in the
20th
wire
stated
is expected
5
IPRTs.
between
3.925
3.928 x 10e3.
thermometry
in the
United
resistance
early
of platinum
100-C,
and
classifying
have
a coefficients
the IPTS-68)
and
of modern
a has
been
probably,
The
order
Cohn
today
on
of the
between
popular
resistance
than
it has
the preparation
standard
for
quite
tained.
and
For platinum
of
Q/P/‘C
(assuming
tory,
but
surement
Eq.
the measure
a straight
line
a useful
inwire.
Let
us
IPTS(68)
a = RlOO/RO
a is, therefore,
the slope
of
are
to
attention
In the
textile
fiber
the
of
his-
in meapredates
the European
3.850
x 10s3;
then
be ob-
United
States
to IPRTs
as more
mid-1950s,
I made
plant;
the
largest
S. I made
these
from
the
best
about
3.916
x 10e3. Obviously,
free;
a compromise
conditions
of industrial
was
necessary
use.
in
or-
10
In
the
de
crete,
pure
the
facto
remained
platinum
absence
of an
Il. S. standard,
American
while
standard
for
the European
3.850 x 10W3, The American
de
wire
in a less than
ideal
physical
a, 3.915
standard,
x 10m3 became
cast
in con-
facto
standard
structure,
while
reflected
the Eu-
ropean
coefficient,
once
derived
from
platinum
wire
less
than
ideally
now
reflected
pure
platinum
wire
doped
with
specific
impurities
[61. The situation
of two
(and
there
were
more
than
two)
accepted
coefficients
would
obviously
cause
problems.
For example,
a controller
scaled
for one coefficient
would
indicate
(except
at 0-C)
improperly
if the
sensor were
of the
other
coefficient.
Many
unsuccessful
attempts
at compure,
promise
were
made,
by
many
standards-writing
bodies.
SAMA,
for
example,
promulgated
a standard
in which
the
sensing
element
a was
required
to be 3.923 x 10e3 (a completely
unrealistic
number
for an IPRT)
and was
then
shunted
with
prescribed
shunts
to an effective
coefficient
of either
3.915
x 10s3
or 3.850
x 10m3 (which
incidentally
altered
the
shape
to the
of the characteristic
confused
situation
whose
SPRTs
are
characteristics
vidual
ever,
calibration
are specified
considered
are
constants
not as
curve).
The
was to produce
by their
completely
response
sensors
users
and
of
of
U. S. manufacturers
both
coefficients.
to be individual
instruments,
adequately
described
by indi-
and printed
interpolation
individual
sensors,
but
as
tables.
members
IPRTs,
howof a group
performing
within
certain
limits.
The primary
emphasis
here
is on the
interchangeability
of like
sensors,
so that
field
replacements
may
be
made
without
the
necessity
of
recalibrating
systems
or
processes.
Elements
may be checked
by the
manufacturer
on an individual
or on a
statistical
basis,
and,
often,
the difference
between
thermometers
of several
accuracy
classes
represents
sorting
at inspection.
An individual
calibration
of such
a sensor
is almost
never
done
except
in batch
qualification,
and is never
offered
to the user
except
as an extra-cost
option.
It seems
today
that
the confusion
will
eventually
be resolved,
not
on technical
grounds,
but
by
demands
of the
marketplace
for
harmonisation
of standards
on a global
basis.
A task
group
of the
International
Electrotechnic
Commission
(IEC)
is formulating
a revision
of the
IEC document
on industrial
resistance
thermometers,
and it will
be based
on the
European
a coefficient
only.
The
IEC
standard
will
be adopted
rapidly
and verbatim
into
the European
Community
standards
documentation,
as a DIN, BSA, etc.
National
standard,
and the need
to comply
will
be persuasive
to all manufacturers,
including
those
in the United
States
who wish
to export.
In
the absence
of
1, which
lists
tolerances
tional
regulations
within
a uniform
standard
and
coefficients
IPTS(68).
(Note
at
for
that
this
IPRTs
the
time,
we offer
Table
promulgated
as Natolerances
shown
in
TABLK
1
INTKRNATIONALLY
AGRKKD
UPON
SPKCIPICATIONS
(PRIOR
TOITS-90)
IKCPUB751 OIWL1985 8s 1901:1981 DIN13760 GOST
6651-U JKHIW SAMRU(1966)
1983
-_-----------__ _____-______
.-. - - __________._
__._,
a(0 DBCcj OBnS ID0
5 to 1000
100
100
IO, 46, lo0
100
too
TOLERANCB
AT0
CLASS
A
t/-0.06
CLASS
B
t/-o.!2
t/-0.15
0.03appr
t/-0.075
t/-0.06
special
t/-0.1
q-0.3
0.5 appr
1stahdard
I
t/-0.12
0.00385
0.00391
0.00?85
0.00385
0.003115
0.00391
0.001916
ID.003923
-183TO630
-200ro 600 -200ro 600
,200TO150 ,220l-0 ID50 -200ro a50 -200TO850
-/ 200TO600
k/-0.78-0.5 t/-O.IK-05 t/-l.?K-05 t/-0.7K-05
b/-1.28-05 +/-2.OE-o5 t/-3.OK-05 t/-l.Ix-Ol
I
COK?F
TYPK
I
A
B
C
COEFP
TYPK
II
A
B
3.908028-03 1.90802E-033.96835X-033.908021-033.908028-033.971718~033 .98153K-03
-5.8028-07 .5.802K-07 -5.83~98-075.801958-07-5.8028-07 5.8775847 _I5.85316-01
-1.23758-12 4.27IK-12 -1.35578-12 -1.27358-12-1.27358-12 -3.48138-12 -/ 1.35158-12
1.969688-03
i.8677K-07
4.lllK-12
3.96847K-03
-5.8478-07
-1.35588-12
NOTK:IEC: INTERNATIONAL
KLKCTROTKCRNIC
COHIIISSION;
OIXL: INTKRNATIONAL
ORGANIZATION
FORLKGAL
KKIROLOGY;
BS:
RRIfIsRSTANDARD;
01~= DmscflK mm
FOBR
NORIIONG;
cosr = Gossrmm (ALL-UNION
STANDARD,
USSR);JKWI~A
:
JAPANKSK
STANDARD;
SANA
: SCIKNTIPIC
APPARTDS
WAKKRS
ASSOCIATION
(UNITKD
STATKS)
12
FIGURE 1
DEG C
5.0
4.5
4.0
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0
-200-100
0
100
200
300
400
500
600
700
800
DEG C
INDUSTRIAL RESISTANCE THERMOMETERTOLERANCES SPECIFIED BY
IEC IN 1983
CLASS A
CLASS I3
13
Table 1 are
peratures).
standard
at 0-C. Fig.
1 shows
The
Table
suggests
will
be welcome,
in place
A Table of values
from any of these
lated
the IEC 1983 tolerances
that
almost
any
single
of these
many
at
conflicting
other
teminternational
standards.
of resistance
versus
temperature
may be calcuusing the following
algorithms
(stipulated
in the
standards),
which
closely
resemble
the formulations
of Callendar
(above
0-C) and Callendar-Van
Dusen
(below O’C). (In these
equations,
A, B and
C are equivalent
to
but
not
numerically
the
same as Callendar-Van
Dusen’s
a, 6 and
Above
R(t)
0-C:
= 1 + At t Bt’
Below
R(t)
accord
8).
Eq.
6
Eq.
7
0-C:
= 1 t At
Tables
with
t Bt2
calculated
IPTS(68),
continuous
functions,
(Think
of the IPTS(68),
between
two supports,
t C(t
using
since
- 100)t3
these
these
equations
equations
will not
represent
be
identically
in
mathematically
and the IPTS(68)
interpolation
equations
do not.
and also the ITS(SO),
as a long clothesline
hung
in a catenary
curve,
but interrupted
at interme-
diate points
by the fixed
points
which
enter
into
the equation,
the curvature
of a clothesline
supported
by intermediate
props).
ever it is stated
by the designers
of IPRT standards
that
the
“industrially
significant”.
(The
major
difference
ences
are
not
IPTS(68)
was about
15 mK plus
the difference
of the individual
mometer
from
constructions
the nominal
for ITS(SO),
it
curve).
Until
we have
seen
is not possible
to estimate
the
the
giving
Howdifferfrom
ther-
equivalent
divergence
between
the industrial
scale and the ITS(90).
Our prediction
is that
it
will be much larger,
because
of the nature
of the ITS(90)
algorithms.
Perhaps
what
we need
(I appreciate
what
heresy
I speak)
is a quite
distinct
Industrial
Platinum
Resistance
Thermometer
Scale;
an International
Practical
Temperature
Scale.
THE CONSTRUCTION
Most
platinum
OF IPRT
manufacturers
wire
used
for
of
the
SENSING
ELEMENTS
IPRTs
do not stint
on the quality
of the
sensing
element,
but
commonly
purchase
“reference
grade”
wire
when the a coefficient
is to be 3.92 x 10e3 or
higher,
and the best wire when a is to be 3.85 x 10m2, Lead wires
may
Occasionally
other
materials
than
platinum
be commercial
grade
platinum.
are
used
for
lead
wires,
but
platinum
is preferable
because
it
avoids
ac-
14
cidental
mometer.
thermoelectric
junctions
in
the
hot
sane
of
the
finished
ther-
The resistance
at O’C may be lOOQ, ZGOQ, 5GGQ or some other
value.
For the 1OOQ element
a typical
wire diameter
is 0.025 mm (0.001 inch)
for
the realistic
reason
that
this
is the diameter
at which
the sum of the
cost of the platinum
metal
and the cost of drawing
the wire is a minimum. Higher
O’C resistance
elements
may employ
even
smaller
diameter
wire,
0.015
mm (0.0006
inch)
being
a realistic
lower
limit
imposed
by
physical
handleability.
The wire
is usually
supplied
in the hard-drawn
condition
to avoid
stretching
it during
manufacture.
There
have been a number
of schemes
for
the disposition
of the
platinum
wire
upon
the supporting
structure.
All are contrived
to provide
some freedom
for the wire
to expand,
contract,
etc., with
minimum
induction
of mechanical
strain,
while
still
restraining
the wire
so that
the physical
motion
of the wire as an accelerated
mass will not in itself
induce
strain,
or in the extreme
example,
cause
unwanted
turn-to-turn
contact.
A design
I used for many
years
is as follows.
.OOl inch
diameter
wire, coated
with
a film of Isonel
varnish
0.0001
inch
thick,
was wound
onto
a high-purity
alumina
mandrel,
and the ends
spot welded
to platinum
lead
wires
fixed
within
the mandrel.
After
trimming
to resistance
tolerance,
the
system
was dip-coated
with
ceramic,
dried,
and fired
to
anneal
and to vaporise
and drive
off the varnish.
(In later
production,
the winding
was surrounded
by a loose-fitting
ceramic
tube,
and the interspace
filled
with
ceramic
powder
compacted
centrifugally).
This
left
the wire
free
in a helical
tunnel
perhaps
0.0002
inch
larger
than
the
wire diameter.
Since
the thermal
expansion
of platinum
is higher
than
that of the ceramic
mandrel,
it was necessary,
first
to stretch
the wire
by several
cycles
from room temperature
to liquid
nitrogen
temperature,
and then to re-anneal
the wire above its upper
temperature
of use.
FIGURE
Alumina
2
core
Packed
The
“tunnel”
construction
of
an element
15
In an experiment
to determine
the effectiveness
of this
construction,
a large
number
of thermometers
(ca 100) were
made
using
bare
wire,
which
would
be closely
captured
by the coating,
and Isonel-coated
wire,
which
could
be expected
to have the freedom
described
above.
All
other
details
were common
to both sets. The close-captured
thermometers
e3xhibited
an a coefficient
closely
grouped
about
the mean of 3.915 x lo, while
the loosely-retained
thermometers
were all about
3.920b3.
Obviously,
the latter
were freer
from strain.
Fig. 2 indicates
the tunnel
construction.
The most
usual
design,
due to Curtis
and other
workers,
is to
prepare
two fine-diameter
coils of platinum
wire, and place these
within
two bores
of a four-bore
ceramic
insulator,
the other
two bores
being
used to capture
lead
wires.
The coils are welded
together
at one end
and to the lead
wires
at the other
end.
A cement,
ceramic
or glassbased
is introduced
into
the tubes
holding
the coils,
in such
manner
that the cement
contacts
and secures
only
some specific
portion
of each
turn.
Fig. 3 indicates
this construction.
FIGURE 3
Ceramic
tube
Platinum
The
“Z-coil”
coil
construction
of
an element
and
stability
with
mechanical
Stability
with
temperature
change,
shock
and vibration,
are directly
contradictory
requirements,
and one
many
be satisfied
only
at the expense
of the other.
In the construction
described
above,
the manufacturer
has a wide
range
of choice
between
sticking
down only
a very
small
fraction
of each turn
of wire,
for best
thermal
stability,
a very
large
fraction
(or all)
of each
turn
for best
mechanical
stability,
or any desired
compromise
between
these.
These basic designs
have
been in use for many years.
This
is not
to say that
there
has been
no improvement
over these
years.
The modern IPRT is superior
for the following
reasons,
among
others:
16
a: The
drilled
without
b:
fine
platinum
sapphire
or
contaminating
Ceramic
reached
a
years
ago.
wire
employed
diamond
dies,
the platinum.
materials,
mandrels,
development
stage
c: Techniques
have
ti be thin
and
long,
lar,
etc.
to include
Examples
been
short
is
now
which
drawn
give
substrata
of
developed
and fat,
purity
and
laserresults
cements,
undreamed
which
very
two or more
electrically
of available
shapes
and
through
repeatable
of
permit
sensing
small,
flat and
separate
sizes
are
have
even
10
elements
rectangu-
matched
shown
in
elements,
Table
2
[71.
FILM
ELEMENTS
Much
work
has
been
done
to develop
platinum
sensors
thinand thick-film
technology,
but the results
have
been
to
appointing.
The hoped
for
results
of lower
cost
and performance
alent
to that
of wrought-wire
sensors
have
simply
not
been
and there
is now reason
to doubt
that
they
can be.
(a)
Film
thermometers
but a substrate
material
not yet
been
identified.
sion strain
gages!
must
be
of identical
Many
film
mounted
intimately
mechanical
elements
are
(b) The
paste
or film,
which
contaminated,
and
therefore
usually
which
is subject
to the
problems
to
expansion
excellent
based
on
date
disequivrealized,
a substrate,
properties
thermal
has
expan-
contains
very
little
metal,
is easily
is sealed
under
a glassy
coating,
of (a) and also
may contaminate
the
platinum.
tirely
(c)
The
different
(d)
sintering
(e)
inhomogeneity
from
wrought
Characteristics
temperature,
The
thin
of the
paste
wire;
resembling,
vary
from
provides
rather,
batch
to
characteristics
enspongy
platinum.
batch,
as
a
function
of
a problem
not
etc.
film
units
are
so
small
that
self-heat
leads,
etc.,
are
is
solved.
to the
(f) mechanical
generation
of
attachments
of
thermal
emfs.
weak
and
susceptible
Thermal Developments
wide range of Detectors
tc. IEC 751- 1983.
International
using wire
produce
conforming
a
17
All dimensions
in mm
Detectors
shown same size
Resistance
tolerance at 0-c
Ceramic
length
Ceramic
diameter
Sensing
length
0.1%
25+0
- 0,s
1,5+0
- 0.03
z-L1
PlOO/2516
P2100/2516
0.1%
25+0
-0.5
1,6+0
- 0.03
22=,
P100/2020
P2100/2020
0.1%
20+0
-0.5
2.0 + 0
- 003
17 2 1
0.1%
15+0
- 0.5
4,5*0
1Zfi
0.1%
15+0
-0.5
3.2 + II
0.03
12fl
0.1%
15+0
2.8 + 0
12fl
i;--~~
P100/7040
P100/5024
,
P2100/5024
.~~~
P100/5015
P2100/5015
___~
PlOO/3045
P2100/3045
PiOO/3038
P2100/3038
P100/2532
P2100/2532
P100/2528
P2100/2528
P100/2524
P2100/2524
-..
PlOO/2515
P2100/2515
P100/1545
P2100/1545
C
.~~
P100/1532
P2100/1532
I
~~__
PlOO/l528
P2100/1528
.___~~
(
0.5
~~
PiGO/
P2100/1524
__~~~
c
P100/1520
P2100/1520
(--'I
- 0.03
_____
--~~
- 0.03
0.1%
15+0
-0.5
2.4 + 0
- 0.03
0.1%
15+0
- 0.5
2.0 + 0
12?1
- 0.03
~.
~~~~~~~~~~~ ___-.~
P100/1516
P2100/1516
0.1%
15+0
- 0.5
,.s+o
P100/1515
P2100/1515
O,l%
*5+0
-0.5
1.5+0
- 0.03
12c1
O.,%
*5+0
- 0.5
1.2 z 5%
12?1
P100/1512
-.
-___-
P100/1509
P100/1012
lZIl
12Z1
0.03
18
Nevertheless
accuracy
great
tion;
e.g.,
HOUSING
or
home
THE
film sensors
seem
adaptable
stability,
and
where
cost
heating
IPRT
and
air
conditioning
to
is
uses
which
a paramount
require
considera-
no
systems.
SENSOR
Almost
all IPRT
sensors
require
some
sort
of enclosure
when
they
are
put
to use,
to protect
the
sensor,
provide
for
the securing
of external
connecting
wires,
close
off
the
spaces
into
which
they
are
inserted,
couple
them
to a process
or an environment,
etc.
The variety
is
limited
only
by the configuration
of the
sensor
chosen,
and the ingenuity
of
the
designer.
For those
who need
only
one or a few
special
configurations,
it is
not excessively
difficult
to buy
sensors
and assemble
thermometers
on a
do-it-one’s-self
basis.
Precautions
include
the
necessity
to
preserve
cleanliness,
to make
impeccable
joinings
of wires,
to pay
attention
to
thermal
coupling
between
the
sensor
and
the
outside
environment,
and
to provide
adequate
electrical
isolation.
For those
who
require
more
than
a few
thermometers,
it is usually
more
effective
to adapt
to the
many
commercial
configurations
available,
or to seek
an assembler
willing
to
make
up special
configurations.
It is possible
to mount
IPRT
sensors
in
tubes,
wells,
drilled
holes
in casings,
machine
screws,
etc.
CALIBRATING
THE
IPRT
Accurate
calibration
compared
to calibrating
fixed
points
into
which
mersion,
and
the
rest
of an
an IPRT.
an SPRT
is
SPRT,
in at least
one sense,
is
One has only
to have
a suitable
fits,
diametrally
and
with
sufficient
simple
set of
im-
straightforward.
The majority
of IPRTs
cannot
be calibrated
in the
usual
sort
of
ITS(90)
fixed
point
cell, because
they
won’t
fit, or they
are not designed
for sufficient
immersion,
or because
they
won’t
tolerate
the temperatures
along
the length
of the lead wires.
bration
as “like”
whatever
The concept
of the
“Like
Standard”
of IPRTs.
It comprises
the creation
as possible
to the thermometer
to
manner
is necessary
to adapt
it to
An
intended
textile
steam.
is a useful
one in the
caliof a reference
thermometer,
be calibrated,
but altered
in
a fixed
point
cell.
example
is shown
in Fig. 4. This
small
sheathed
thermometer
is
to be screwed
into
the hot zone of a spinnerette
for
making
a
fiber.
In
use,
the
tip
projects
into
a stream
of high-velocity
Since
the leads
are external
to the
heated
system,
they
are
Tel-
19
fan
and
insulated.
the leads
The hexagonal
will not tolerate
nut
the
is too large
temperature.
to
fit
a fixed-point
cell,
A “like
standard”
was made by (a) reducing
the diameter
of the
hex nut and the threaded
portion
(b) substituting
glass-insulated
leads
for the Teflon
leads
of the working
thermometers.
It was then
possible
to obtain
calibrations
of the “like
standard”
at fixed
points,
so that
the
“like
standard”
could
serve
as a reference
thermometer
for comparison
calibrations
of the working
thermometers.
FIGURE 4
Teflon
The
depth
diately
the
when
side
ature
test
leads
object
The
The
hex
removed
fiberglas
“like”
leads
standard
The short
length
of the thermometer
meant
that
the
immersion
would
not be sufficient.
The paper
by John
Tavener
which
immefollows
will suggest
the errors
inherent
in the system.
However
“like
standard”
had,
intentionally,
the
identical
deficiencies,
and
the
standard
and a working
thermometer
were
screwed
side-byinto
a copper
block,
and the block
immersed
in a constant-temperbath,
these
deficiencies
cancelled
almost
exactly.
One
might
well ask:
but in use,
did the working
thermometer
accurately
realise
the temperature
of the steam?
In this
real-world
example, no one really
cared.
The relevant
matters
were (a) once the process
had been optimised,
the working
thermometer
maintained
the preset
temperature
(b) any
replacement,
of the working
thermometer
(say in the
event
of its failure)
had the same calibration
and characteristics.
COMMON
ABSTRACT
ERRORS
- JOHN
IN TEMPERATURE
MEASUREMENT
P. TAVENER
It is the first
law of thermometry
that
a thermometer
senses
temperature
except
its own, While it is easy to obtain
steady
readings
temperature
from
a resistance
thermometer,
it is much
more difficult
be sure
that
these
readings
indicate
the temperature
of the object
environment
of interest.
no
of
to
or
20
Of the
almost
infinite
some of the
more
of the magnitude,
number
obvious
are
and corrective
of
ways
discussed
measures
in
which
here,
together
which
may
errors
can
with
an
be applied.
occur,
estimate
INTRODUCTION
If two
systems
in thermal
equilibrium
are each
the same temperature
as a third,
then
they
are also the same
temperature
as each
other.
But
no perfect
system
exists.
The temperature
of an object
is affected
by the thermometer,
and
the thermometer
may be temporarily
or permanently
affected
by the system
into which
the thermometer
is immersed.
In obtaining
a
need to be considered,
system
temperature.
1: Thermal
steady
before
lag;
the
electrical
inertia
2: The
thermometer’s
3: Immersion
(including
the
5: DC errors;
ilar metals
1: Thermal
large
diameter
ter immersion
or well
were
If
caused
error,
environment)
lead
by
of
is
by
the
the
following
as the correct
thermometer’s
thermal
or
capacity
by
heat
transfer
to the thermometer
the
emfs
necessity
caused
by
of
from
the
passing
a
junctions
system
current
between
dissim-
resistance
a small-diameter
pocket
should
absent.
the thermometer
an air gap.
caused
thermal
effects
of
lag.
delay
caused
thermometer
4: Self-heating,
through
the
6: Effects
reading
with
a thermometer,
one can accept
the reading
thermometer
is
inserted
loosely
into
a
or well,
one feels
instinctively
that
the thermomebe deeper
than
if it were
a tight
fit, or the pocket
Not necessarily
true:
the size of an air gap around
almost
irrelevant,
compared
to
the
fact
that
there
is
(I once
had the problem
of designing
a thermometer
for a reactor
bypass
cooling
water
loop,
which
had to show
a 63% response
time of 2
seconds
in water
flowing
turbulently
at 2000
feet
per
minute
at 3000
psig,
so that
the
fast
response
had
to be achieved
in a mechanically
The
sensing
element
was
mounted
in a hole
very
rugged
construction.
drilled
in
a taper
pin,
and
embedded
in
a
beryllium
oxide
paste.
The
21
interior
of the sheath
time was
8 times
longer
driven
tight,
as
However
mometer
and
the thermometer
was
reamed
with
the
a result
of
the
the
combination
the
temperature
itself,
give
heat flow
required
to
sponse
time.
I describe
to hold
the taper
pin loosely
in place
very
slight
air
gap!
causes
a finite
model
in Fig.
5:
I
analogue
model
Fig 5: Thermal
lag.
Ti = the initial
and T are the temperatures
of the
tively,
R, is the
thermal
resistance
system,
achieve
C is the
heat
a 63% response.
Then
of
Fig.
ways
and
= -(Ts
In
by
- Ti)
exp
the
T
thermal
lag
temperature
of the
thermometer,
T,
system
and the thermometer
respecbetween
the
thermometer
and
the
the
temperature
thermometer,
and
measurement
(the
r
is
time
shaded
to
area
multiples
Eq.
(-r/T)
time constant
are commonly
to be no general
consensus
Usually,
however,
a manufacturer
provides
was
determined.
a practical
waiting
for
Te/(Ts
in
in
of expressing
there
appears
this
information.
formation
he
small
error
of
of
c
6) is
Te
Other
ature,
the
capacity
re-
5
-L
An electrical
between
the
therthe
heat
capacity
of
as thermal
lag. The
or cool the thermometer
effect
with
an electrical
FIGURE
The response
when
it was
- HES)
of thermal
resistivity
to be measured,
and
rise
to an effect
known
warm
this
pin.
than
- Ti)
of
the
situation,
a sufficient
= -exp
response
the user
can make
time. Fig. 6 shows
(-s/T)
time.
1
found
in the literon how to present
specifies
the
the
error
factor
how
the
in-
arbitrarily
Eq.
2
22
FIGURE 6
Ts
//I/L”‘:
Tay
The effect
1
T
2r
of thermometer
response
3r
etc.
time on a measurement
FIGURE 7
100%
10%
1%
0.1%
100 PPM
10 PPM
1 PPM
0
2
4
6
8
Time (units
Temperature
error Te/(Ts - Ti)
ment time in multiples
of r
10
12
14
of T)
plotted
against
measure-
EXAMPLE: Estimate the minimum measuring
time to achieve an accuracy
of better
than 0.1% when measuring
a temperature
near 15O'C with a
thermometer
whose response time is 15 seconds.
The maximum error
O.l'C/(150'
is
- 20')
From Fig, 7, the minimum
= 0.07%
measurement
Eq. 3
time is 7r = 105 sec.
In systems where the temperature
is not constant,
ment errors
become more complex. Consider
the situation
the measurein which the
23
T
To -
Ar
Time
Temperature
in a system
error
with
due to the thermometer
the
system
temperature
tine
constant
constantly
rising
In systems
where
the temperature
is not
ment
errors
become
more
complex.
Consider
the
system
temperature
rises
continually,
as in Fig.
error
components,
an exponential
component
T el
and
= -(To
a constant
- AT - Tl)
exp
constant,
situation
8. Here
the measurein which
the
there
are two
(-r/T)
4
Eq.
5
component
T e2 = -AT
of
Eq.
where
the system
temperature
rise of the system
temperature.
Ts = To t A,
and
A is the
rate
The exponential
component,
T el, can be dealt with
by waiting
for a long
enough
time.
The constant
component
Te2 can be dealt with
only
by using a thermometer
of shorter
time constant.
2: Thermal
capacity.
When a thermometer
is immersed
into a system,
heat
will flow
between
the system
and
the thermometer
until
equilibrium
is
reached.
Unless
the
system
temperature
is under
external
control,
a
permanent
change
in the system
temperature
will result.
The smaller
the
heat capacity
of the thermometer,
the smaller
will be the effect
upon the
system
temperature.
A simple
model of the measurement
process
provides
an estimate
of the resultant
temperature
error:
Let Ct and Cs be the heat capacities
respectively,
T, be the final
system
initial
and final
temperatures
of the
of the thermometer
temperature,
and
thermometer.
Then
and the
Ti and Tf
system
be the
24
T, = -[Ct/(Cs
and
+ Ct)l[T,
- Til
Eq.
6
rearranging,
T&T,
- Ti)
= Ct/(Cs
+ Ct)
Therefore,
to achieve
less
mometer
Ct should
be at
of the system
C,.
EXAMPLE:
used
to
Suppose
measure
than
least
1% error,
100 times
a thermometer
the
temperature
has a heat
capacity
of a cup
of coffee.
perature
error
due to heat capacity.
coffee
is 80-C,
the
initial
temperature
the heat
capacity
of coffee
= the
lOOOJ’C-‘.
From
Eq.
T,
of the therheat
capacity
of 5J’C-l,
Estimate
the
and is
tem-
Assume
that
the temperature
of the
of the thermometer
is 20-C,
and
heat
capacity
of 250 ml of water
=
5,
= [5J’C-1/1000J’C-1][80’C
The thermometer
thermistors
and
than
O.O2J’C-1,
pacities
would
the heat
capacity
smaller
than
the
of
show
used
very
while
- 2O’Cl
in this
Example
fine
thermocouples
mercury-in-glass
as much
as
a 12-C
error;
200J’C-1.
even
in
= 0.3-C
Eq.
8
is
a small
sheathed
IPRT.
Small
can
have
heat
capacities
less
thermometers
may have
heat
ca-
Such
a mercury-in-glass
30 liters
of coffee,
the
error
thermometer
would
be
0.1 ‘C.
In
many
measurements,
mometer,
so that
the
the system
temperature.
it
initial
is
possible
to
temperature
preheat
of
the
or
precool
thermometer
the
is
close
3: Thermometer
immersion
depth.
A definition:
A thermometer
is
ciently
immersed
in a system
when
there
is zero
heat
flow
between
sensor
and
the
external
non-system
environment
through
the
sheath
or other
thermometer
parts
that
extend
from
the
sensor
to
ent temperature.
Heat
flowing
through
replaced
by the system
ation.
A simple
model:
the
thermometer
in
the
forms
from
of
or
to
conduction,
ambient
therto
suffithe
leads,
ambi-
is absorbed
convection
and
or
radi-
25
6T = qR
Eq.
where
8T = temperature
sistance.
This
can be
Fig. 9:
difference,
q = heat
compared
to Ohm’s
Law,
FIGURE
Contact
resistance
9
flow and R = thermal
reand repfesented
shown
in
9
T
Leakage
Resistance
TS
1
System
Temperature
Therm
Temp
Ambient
Temperature
a: The greater
age; therefore
practicable.
the
the
immersion
thermometer
depth,
the greater
the
should
be immersed
b: The greater
likely
to be.
the
immersion
depth,
the
smaller
the
resistance
ta leakas far as may be
contact
resistance
is
c: The contact
resistance
also depends
upon the thermal
conductivity
of
if a fluid
how fast the fluid
is flowing,
whether
the flow is
the system,
laminar
or turbulent,
and whether
there
is cavitation
behind
the thermometer.
A simple
formula
is:
T, = (Ta - T,)
Eq.
where
length
length
stant
K,
exp
(-L/Lo)
10
T, = temperature
error,
T, = system
temperature,
L = immersion
of the
thermometer,
L,
a constant
called
the
“characteristic
of the thermometer”,
Ta = ambient
temperature,
and K, = a conalways
less than
1.
In instances
where
the
conductivity
of the system
is poor,
where
high
precision
is required,
a simple
experiment
will determine
and estimate
the magnitude
of T,. At least
three
measurements
must
made.
at immersion
depths
of L1, L2 and L3 = 8L. Temperatures
of
T2 and T3 are obtained.
Ts
and,
= Tl
rearranging
+ [T2
Eq.
- T,I%(T
10,
2 - Tl)
-
03
- Tl)l
Eq.
or
L,
be
T1,
11
26
L,
= [sL)/ti[(Ts
EXAMPLE:
cm give
system
Suppose
measured
- T1)/(Ts
that
measurements
temperatures
of
temperature
and
the
Ts = [119
= 123-C
-
11512/[2(119
Lo
-
3cm]/&z[(123
q
[4cm
4: Self-heating
ement;
in order
be passed
thermometer,
- T2)l
Eq.
at immersion
depths
115’,
119’
and
121°C.
characteristic
-
length
115)
-
-
(121
115)/(123
-
-
error.
A resistance
thermometer
to make a measurement
of its
through
which
it.
This
inevitably
is inevitable:
of
the
of 3, 4 and
5
What
is the
thermometer?
115)1
119)]
= 1.44
cm
Eq.
13
Eq.
14
is a passive
electrical
resistance,
a current
results
in
12
some
heating
elmust
of
P = 12R
the
Eq.
15
This
results
in an elevation
of the
apparent
temperature
of the
thermometer
(which
measures
only
its own
temperature)
and
an elevation
of
the system
temperature.
The problem
is to evaluate
whether
this
heating
is significant
in terms
of the accuracy
required
of the measurement.
The
mometer
in
self-heating
a controlled
effect
is readily
environment
determined
and
making
by placing
measurements
the
least
two
impressed
currents.
From
this
the so-called
“zero
power
sistance”;
that
resistance
which
would
be measured
if it could
be
sured
with
no impressed
power;
and
the actual
resistance
measured
any current
may be estimated
as a difference
from
Ro:
Ro = [Rl
5: Voltage
- i121[(R2
errors.
Most
age across
the
unknown
Therefore
any
extraneous
are a source
of error.
The largest
amplifier
develops
Any
output
voltage
by an
voltage
equivalent
of the
-
Rl)/(i22
- i12)1
resistance-measuring
resistor
voltages
therat at
remeaat
Eq.
systems
with
that
acrosm
which
arise
in
compare
the
16
volt-
a reference
resistor.
the
measurement
path
DC error
is caused
by imperfect
amplifiers.
The ideal
zero
voltage
when
both
inputs
are
at zero
potential.
which
occurs
under
theme conditions
can be replaced
DC input
voltage
Vo,,
referred
amplifier.
Most
manufacturers
of
to as the
amplifiers
input
supply
offset
data
27
sheets
range
giving
a typical
or maximum
from
20 PV to 5 mV depending
amplifier.
The
(e.g.,
temperature,
important
offset
sources
Thermal
voltage
and
Power,
supply
of
emfs
error
are
in
value
for
Vos,
which
upon
the
type
or
its
sensitivity
voltage
and
operational
another
to
important
various
represent
time)
amplifier
type
may
quality
typically
of the
parameters
the
most
circuits.
of
DC
error.
They
are
generated
when
junctions
of leads
of dissimilar
materials
act as thermocouple
junctions.
(Even
joints
in copper
wire
from
two
different
manufacturers
may produce
emfs
as high
as 0.2 PV’C-I.
The emf of a copperplatinum
junction
is typically
6 to 8 PV’Cvl.
the
assembling
a thermometer,
junctions
from
different
types
of wire,
become
contaminated.
The
net
emf can
pairs
of junctions
close
to each
other
and,
if solders
are used
at all, employing
maker
In
Voltaic
metals,
should
perfect,
emfs
arise
from
should
avoid
T,
where
V,
instrument
error
electrochemical
caused
by
DC
activity
between
voltage
errors
is
dissimilar
The
the
given
problem
joints
are
by
Eq.
= Ve/S
is
the
lead
especially
wire
that
may
have
be further
reduced
by
keeping
(i.e.,
at the
same
temperature),
“low
thermal”
solders.
in the
same
way
that
batteries
generate
voltage.
not
occur
if the
same
wire
is used
throughout,
and the’ environment
is clean
and dry.
Temperature
making
input
voltage
error
(mV)
and
S is
the
sensitivity
of
17
the
(mVC-I).
Although
it is possible
to reduce
some
of these
errors
by
good
by exchanging
leads
and
averaging
measurements)
volttechnique
(e.g.,
age errors
limit
the
practical
accuracy
of a DC resistance
thermometer.
AC techniques
are
free
from
voltage
errors
but
may be subject
to resistance
time constant
limitations
(81.
6: Lead
resistance
3-wire
and 4-wire
errors
and their
errors.
systems.
mitigation
Industrial
PRTs
may be connected
The reference
provides
a discussion
[91.
as
2-wire,
of these
28
FOOTNOTES
[l]
John
standards
P. Tavener,
for
ITS-90,
Platinum
Isotech
[21 H. Preston-Thomas,
90), Metrologia
2’7, pp
131 C.
Isotech
H.
Jour
Meyers,
Therm
The
3-10
I
am
indebted
for
mund
Cohn
Corporation,
the preparation
of pure
[61 Dr.
Sigmund
Standard
p 10 et
[7]
TDI
From
and
[8]
See
resistors,
Therm
[91 Ibid.,
Mount
platinum
the
life
H.
E.
Sostmann,
resistance
V2 N2 pp 67-69
Pages
72-73
of TDI,
Hemisphere
bridges,
of
1990
(ITS-
resistance
thermometers,
platinum
resistance
thermometers,
with
Vernon,
wire.
Michael
New
York,
O’Shaughnessy,
for
Sig-
information
about
the
revered
metallurgist
of the
premier
supplier
of thermometer
coefficient.
to purify
Southport,
agent,
Fundamentals
Scale
interpolation
platinum
many
years
(the
world’s
pure
platinum
learning
how
the
catalog
its Western
Temperature
as
seq.
a conversation
Bert
Brenner,
for
Cohn
Corporation
wire)
favored
not spent
my
taminate
it!“.
Thermometers
Vl Nl pp 31-37
International
(1990)
Coiled-Filament
V2 Nl pp 16-24
[4]
Henry
E.
Sostmann,
Isotech
Jour
Therm
V2 Nl
[51
Resistance
Jour
Therm
He said
to me once:
platinum
in order
then
England.
please
see
of
thermometer
For
Page
Thermometry
measurements,
“I
to
have
con-
the addresses
3 of this
issue.
Part
IV,
of
Standard
Isotech
Jour