61

FUNJ3AMENTAL.S OF THERMOMETRY

PART IV

MEASURING THE RESISTANCE OF

STANDARD PLATINUM RESISTANCE

THERMOMETERS

by Henry E. Sostmann

ABSTRACT

The Standard Platinum Resistance Thermometer, acronym SPRT, is the

stipulated interpolation instrument for realising the International Tem-
perature Scale of 1990 between the defining fixed points, over the range
from 13.8 K (the triple point of eka-hydrogen) to 961.78-C (the freezing
point of silver). The transfer function of the SPRT is electrical resis-
tance as a function of temperature; therefore an accurate measurement
of resistance, referred to a fixed resistance base, is an essential compo-
nent of the measurement chain. This article discusses fixed resistors,
and then both traditional and modern means for making the resistance
measurement.

1: FIXED RESISTORS

The ohm is now maintained, in most National Laboratories, as a quantum-

Hall (QHE) effect device; the Von Klitsing ohm. In the apparatus which
allows realisation of the QHE ohm, semiconductor devices of standard
Hall-bar geometry are placed in a large applied magnetic field at a
temperature near 1 K. For an applied current through the device, there
are regions where the Hall voltage remains constant, as a function of
the fundamental constants h/e2 as the field or the gate voltage are
varied. There are a number of such plateaus; the conventional value of
the effective resistance of the first plateau is 25 813.807 Q (1). This
value is believed consistent with the SI ohm to within an assigned lo
uncertainty of 0.2 ppm.

(The use of the QHE ohm in nations which are members of the Treaty of
the Meter became effective on January 1, 1990, replacing at NIST the
NBS ohm of 1948. The relationship is:

1 Q(NBS-48) = 0.99999831(NIST-90)
 62

Thus the ohm is now realised as a universal constant independent of the

variables of experiment. This is much more satisfactory than the earlier
definition, where the ohm is a quantity derived from the fundamental SI
units as Q = kg*m2/(A2*s2), or the still earlier definition in terms of the
electrical properties of a column of mercury.

Prior to the adoption of the QHE ohm as a standard representing a

fundamental constant, the ohm was maintained in most National services
as the mean resistance of a bank of 1 Q resistors (usually lo), which
were, or were generally similar to, the 1 Q resistor developed by Thomas
at the NBS (2). Because of the cost and complexity of maintaining the
QHE ohm, it is highly probable that most laboratories which are not
National, and many National Laboratories in smaller nations, as well as
scientific and industrial organisations in the private sector, will continue
to rely on the Thomas-type 1 Q resistor to maintain the local standard
of the ohm, and it is to these that the fixed resistors which are the
reference basis for platinum resistance thermometry will be referred.

The primary requirement for a resistance standard is permanence of

value with respect to time and use. (I do not include accurate knowl-
edge of the resistance, since every standard resistor must be accompa-
nied by a valid and up-to-date calibration certificate). (In the unusual
case where all measurements will be made using the same fixed resistor
and bridge, and there is no need to report a calibration that is to be
transferrable to another set of fixed resistor and bridge; that is, where
results in terms of bridge units for a specific bridge are all that is de-
sired, stability may be the only paramount requirement). Secondary and
highly desirable characteristics are (a) a low and stable temperature co-
efficient (b) a low thermal e.m.f with respect to copper in the external
circuit (c) a design which permits dissipation of the 12R heating due to
the passage of the measuring current (d) low or zero reactance, which
can also be characterised as a fast time constant in response to an im-
pressed input. (d) is of particular interest when the resistor is used as
the standard resistor of modern automatic resistance bridges or com-
parators, most of which operate at low ac frequencies.

As a generality, all resistors which can be considered as reference stan-

dards use bulk wire as the resistance element. An exception to this is
the resistor design of Vishay, which comprises a fine grid of metal laid
as a thick film on a glassy substrate, in such a pattern that the ad-
justment of resistance to a precise value can be made by physically
cutting certain conducting lines. The success of this scheme in resistors
intended to be stable standards is not yet proven; there arise questions
of, for example, strain developed because of the differing characteristics
 63

of the deposited metal and the substrate. These resistors, however, have
proven to be very valuable as highly precise circuit elements.

Various alloys are available for the bulk wire resistance element. Per-
haps the most important of these is manganin, an alloy developed by
Edward Weston in 1889. The composition is 84% copper, 12% manganese
and 4X nickel. The resistivity at 20’C is about 49.3 #J/cm (290 Q per
circular-mil-foot). The temperature coefficient of resistance is about
KLOOOO15 Q/Q/-C over the normal laboratory range of 15’ to 35-C,
although it varies from lot to lot for reasons which are not well
understood or predictable. It can be adjusted somewhat by heat-
treatment in fabrication. For precise work, the temperature of the
resistor must be noted and a correction made. The calibration certificate
of the specific resistor should include the measured coefficients of the
correction equation, which assumes standardisation at 25-C.

Q = R25[I + a(tl - t-25) + R(tl - t25) 2l

For properly selected manganin, a generally

has a value smaller than 10
x 10 -6, and 9 between -0.3 x 10m6 and -0.8
x 10w6. Fig. 1 shows a typi-
cal curve of resistance versus temperature for good manganin. The ,most
desirable situation is that the peak of the curve be located very close
to the laboratory ambient temperature, so that any effects of 12R heat-
ing have minimum influence on the resistance. A further precaution is to
maintain standard resistors in thermostated oil or air baths. Oil baths,
in particular, alSO assist in dissipating any self-heating. The
thermoelectric power versus copper is low; 2 to 3 H/C’.

In constructing manganin resistors, it is particularly important that

where the manganin joins internal copper lead wires, the joint be made
by welding, or brazing with suitable materials. The components of soft-
solders can, with time, migrate into the manganin wire structure and
cause irreversible alloy (and consequently resistance) changes.

fit&fJq
I

Ei;
. m
10 20 30 40 50 10 20 30 40 50
‘C ‘C
FIG. 1 FIG. 2

MANGANIN: Typical change in EVANOHM: Typical change in

resistance with temperature resistance with temperature
for selected wire
 64

Another resistance wire alloy in common use is known by the trade

names Evanohm and Karma. Its major advantage is a high specific re-
sistance, 800 Q/cmf, which permits all resistors to be made of wire of
larger cross section than manganin, and consequently of better mechani-
cal stability in situations of normal shock, vibration, etc. incidental to
use. The composition of the alloy is about 75% nickel, 20% chromium plus
a few per cent each of aluminum and copper. The temperature coefficient
is about ?0.00002 Q/Q/‘C but there is no peak such as that of manganin.
Fig. 2 shows a typical curve for Evanohm.

I cannot comment from personal experience on the resistance stability of

Evanohm relative to that of manganin. I can mention that manganin is a
solid solution of its constituents, while Evanohm is an intermetallic; that
is, some of the elements of the alloy remain as discrete crystals. From
experience long ago in using potentiometer wires which contained alu-
minum, I can report that as surface crystals it is subject to eventual
oxidation, which in potentiometers caused local spots of high contact rs-
sistance, and in standard resistors may be a source of calibration drift.
Nevertheless the Australian laboratory CSIRO has made 1 Q Thomas-
derivative resistors of Evanohm with remarkably small changes in resis-
tance over a number of years. The composition of Evanohm suggests
that all joints to copper be made by welding, since it is difficult to wet
chromium and aluminum with soft solders, and the migration problem is
probably equivalent to that with manganin.

The basic design of the

Thomas 1 Q resistor is shown
in Fig. 3. Manganin wire of
heavy gage is wound on a
temporary mandrel in a bifilar
fashion with spaced turns.
The wire is bare, so that full
heat treatment (at 550-C in an
inert atmosphere) may be ap-
plied, as would not be per-
missible were the wire enam-
elled or served with a textile
fiber serving. After heat
treatment, the helix is slipped
onto a silk-insulated metal
cylinder and sealed into a dry
FIGURE 3 double-walled container. The
wire gradually reaches equi-
A Thomas 1 Q Resistor librium with the slight amount
(After Leeds and Northup) of contained dry air.
 65

Such a resistor is capable of stability of 1 part per million over several

years. The construction is not optimal for heat transfer, and so any im-
pressed current must be carefully limited. There is some slight effect on
resistance from changes in barometric pressure, and the calibration cer-
tificate will furnish information about this.

While the 1 Q Thomas-type resistor will be regarded by most users as

the reference base for the ohm, it is not a very useful value for most
resistance thermometry, where it is more desirable to have a reference
resistor whose value is in the vicinity of, for example, the resistance of
the thermometer at 0-C. For a 25.5 Q thermometer this might be 25 Q; for
an 0.25 Q thermometer, 0.25 Q, with 1 Q another choice. Also, one would
not wish to use the laboratory’s standards base as a working standard.
Most laboratories will maintain the 1 Q standard and possibly an equiva-
lent 10 kQ standard built to the general Thomas design, and use these
to build up and build down, by ratiometric methods, to the working
standards of more convenient value.

- A design of working standard

which has proven itself
through the years is that of
Rosa, also known as the NBS-
type resistor, shown schemati-
cally in Fig. 4. The resistance
wire, which is manganin for at
least the values of 10 KQ and
lower, is insulated with either
enamel or a special textile
fiber in which cotton and silk
are mixed (this mixture said
to provide minimal strain to
the manganin it covers when
the fiber is subjected to
changes in ambient humidity).
The wire is doubled at its
midpoint and wound in bifilar
fashion onto an insulated
brass bobbin (a material
chosen because the thermal
FIGURE 4 expansion coefficient of brass
closely matches that of the
A ROSA, CR NBS TYPE RESIS-
wire, and because it aids in
TOR. The space within the
housing and around the ther- heat dissipation).
mometer well is filled with
neutral oil.
 66

The ends of the winding are brazed to a pair of short heavy copper
leads, the free ends of which are connected to massive terminals.
Externally, four connections are proved, a potential connection and a
current connection to each end of the resistance winding, so that the
resistor may be connected four-terminal.

A four-terminal resistor
Pl P2 schematic is shown in Fig. 5.
The value of the four-terminal
resistor is the potential dif-
ference between the potential
terminals divided by the cur-
R3 R4 rent through the current ter-
minals, so that external cir-
cuitry is not a part of the
measured resistance. The po-
tential and current terminals
FIGURE 5 are interchangeable. For both
Thomas and NBS-type stan-
A FOUR-TERMINAL RESISTOR. RS is the dard resistors, the potential
resistance to be measured. Rl, R2, R3, terminals are usually binding
R4 are external resistances; e.g., lead posts and the current ter-
resistances. Cl, C2 are current
minals are connected into the
terminals, PI P2 potential terminals.
For an SPRT these are conventionally circuit using massive mer-
labeled c, C, t, T respectively. cury-wetted amalgam contacts.

The finished winding assembly must be stabilized by heat treatment, and

this process cannot be carried out to completion; the temperature at
which the wire would be rendered ideally strain-free is higher than the
insulation will tolerate. Thus it is of great importance to avoid ever im-
pressing sufficient current on the finished resistor to heat the winding
appreciably, or additional annealing may occur, with a permanent shift
in value. During the production anneal: the value may decrease perma-
nently by as much as 1% or 2%, and this must be allowed for in the
winding length. The final adjustment of resistance is made coarsely by
removing wire, and fine adjustment by reducing the wire cross-section,
locally, by abrasion.

Rosa resistors are filled with a bland neutral oil, which assists in
dissipating and transferring heat. They are equipped with a central
well, into which a thermometer is placed to measure, with very close ap-
proximation, the coil temperature. A typical manufacturers’ specification
for stability is 20 ppm per year, which is, in my experience, highly con-
servative for a unit which is handled carefully.
 67

Safe operating currents are stipulated by the manufacturer. These val-

ues should be posted on labels, and well-known by operating personnel.
The use of an air or oil bath will increase the safe operating current
limit, as well as provide temperature stability.

As far as I know, the principle supplier in the US of Rosa-type resis-

tors, Leeds and Northrup, has never qualified or made a statement about
the ac properties of these or the Thomas 1 Q resistors of their manu-
facture.

NIST does not at this time offer a calibration of standard resistors

which includes a comparison of dc with low-frequency ac characteristics,
although, according to a telephone conversation (November 1991) with
Norman Belecki of the Electricity Section, it plans to do so “in about a
year”. The National Physical Laboratory (NPL) of England has been of-
fering this service since 1987. Resistors submitted by Isotech to NPL,
which are used in conjunction with ac bridges for thermometry, are re-
ported including a ratio of Rf/Rdc, where f = 75 Hz. Fig. (6) reproduces
a typical calibration report.

2: STANDARD RESISTORS USED WITH ALTERNATING CURRENT

Many modern devices for measuring the resistance of an SPRT operate

in a semi-automatic or automatic fashion, based on principles which re-
quire that the signal received from the bridge be ac. Usually it has
been possible to arrange circuitry allowing the frequency to be very
low, with an upper bound of 400 Hz and a lower as low as 15 Hz. (I
except an experimental bridge of Cutkosky which operated at l/2 Hz,
but which was never commercially available.)

The reactances of reference resistors are a function of their inductances

(from the area included within the winding loop where the wire is dou-
bled) and distributed capacitance. In general the inductance is small but
the capacitive reactance is not, although it may be reduced by such de-
vices as dividing the coil into sections. Reactance may include effects
of the mandrels on which the winding is supported. The time constant of
a resistor may be expressed as

T = (L - CR2)/R

and may also be expressed as the time required for the current to
reach l/s of its final value after a fixed voltage is impressed. Obviously,
for a-c bridge service, the time constant must be short enough to allow
the full value to be achieved within 1 ppm or 0.1 ppm.
 68
NATIONAL PHYSICAL LABORATORY
Tcddington Middlesex TWll OLW England

Certificate of Calibration
STANDARD RESISTOR

No 248712 10 OHM H TINSLEY AND CO LTD

FOR : Isothermal Technology Limited

Pine Grove
Southport
Merseyside PR9 9AG

REFERFXE: Order No. gOPOO65/10 dated 11 May 1990

BASIS OF TEST; NPL Measurements Services - Direct Current and LOW

Frequency Electrical Measurements (1987). Section 3.2

PREVIOUS
CERTIFICATE; NOIF2

MFASUREMFNTS :
The resistor wa+ immersed in an oil-bath controlled at II temperature Of
20.000 t 0.005'c for at least 2 hours prior to and during the meesurement
and was measured in a 4-terminal configuration. The power dissipated in
the resistor was less thnn 1 mW.

Resist@Jxe Uncertainty
1 2
Confidence Level Arithmetic Sum
fit least yjx of Contributions
9.999 917 ohm * 0.25 ppm * 0.43 ppln

The uncertainty of the q eesurement is quoted in two ways :

(1) AC "A Confidence Level of at least 95%” - where the individual
contributions have been combined in quadrature where appropriate.

(2) As expressed on certificates priot to October 1% - where the

individual contributions have been added arithmetically, "Arithmetic Sun
of Contributions".

The& uncertainties refer only to the measured value and do not carry any
implication regarding the stability of the instrument.

fieference ES 94.69 Date of

ES 15% calibration 7 June 1990
Checked bJ.il.
O] ::.<&. IO, "irlclnr
Page 1 of 2 Signed
4
 69

NATIONAL PHYSICAL LABORATORY

Continuation of Certificate

~c:Dc STANDARD RESISTOR

No 248712 ‘10 OHM H TINSLEY AND CO LTD

R!z.sULTS:

Uncertainty
1 2
Test frequency Rf/Rdc Confidence Level Arithmetic Sum
IIt &least 95% of Contributions

l.ow 000 f 0.9 pm I 2.4 ppm

0.10 * 0.015 -,!H/Q * 0.02 pH/Q

The resistor wns inductive.

The uncertainty of the measurement is quoted in two ways :

(1) At “A Confidence Level of at least 95%” - where the individual

contributions have been combined in quadrature where appropriate.

(2) As expressed on certificates prior to October 1989 - where the

individual contributions have been added arithmetically. “Arithmetic Scm
of Contributions”.

These uncertainties refer only to the measured value and do not carry any
implication regarding the stability of the instrument.

Reference EtA 534.198 Date of

Es 7557 calibration 18 June 1990
Checked +I 19
Page 2 Of 2
 70

Details of design can be used to reduce residuals; for example splitting

the wound coil into two sections connected in series can reduce the ca-
pacitance by a factor of 4. Many designs other than that of Fig. 4 have
been proposed to eliminate residuals over a desired band of frequencies.
These include flat single-layer windings on card-shaped mandrels, for
example of sheet mica. Produced by, for example, Electra-Scientific In-
dustries of Portland, Oregon, these have formed the basis of a large
number of commercial bridges, decade boxes, and the useful and familiar
Dekapots and Dekastats. Other approaches include windings which pass
through a slotted ceramic core and reverse direction with each turn,
etc. Difficulty in manufacture has prevented these from becoming widely
used. High resistance elements have been produced commercially in
which the winding is a flat woven web, or patch, with a resistance-wire
warp and a textile-fiber woof, and these have been used for many years
in certain Leeds and Northup standard resistors, above the range of
interest for platinum resistance thermometry.

A number of workers in the last several decades have attempted to de-

sign standard resistors to give the same value for low-frequency ac (dc
to several kH). A successful configuration has been described by Wilkins
and Swan (3). This general configuration is available from H. Tinsley &
Co. Ltd. (4). It is a Tinsley resistor which is the subject of the NPL
Calibration Report shown in Fig. (6).

Wilkins identifies a number of factors which determine the frequency

characteristic of a resistor. These include thermoelectric effects result-
ing from the nature of the resistance wire alloy, inductance and capaci-
tance of the resistor resulting from the arrangement of the resistive el-
ement, and eddy current and dielectric losses resulting from the
mounting and housing and from the disposition of current and potential
lead wires. All of these require consideration in design for ac use.

Many resistive alloys have thermoelectric properties versus copper.

When alloy to copper junctions are not at the same temperature, these
effects present themselves as (a) Seebeck effects, in which a voltage is
generated in opposition to the voltage impressed upon the resistor for
the sake of measurement (b) Peltier effects, in which the passage of a
direct current through a junction of dissimilar metals causes heating or
cooling at the junction (c) Thompson effects, in which a voltage is
developed between points along a wire of uniform composition with a
temperature gradient along its length. These effects may combine in
various senses and serve to make a measurement of pure resistance
indeterminate. In these cases, the true value of resistance is measurable
only at the instant that the circuit is energised, and another value may
be measured when the circuit values have stabilised. The indications for
 71

resistor design are that (a) resistance wires with 1OW Siebeck
coefficients should be chosen (b) the physical design should exert every
effort to maintain the alloy-to-copper and the winding itself in an
isothermal condition.

In general, the reactance of low-ohm (up to 500 Q) resistors is induc-

tive, and almost any reasonable value of capacitance is acceptable. It is,
however, desirable to keep the inductance low, since the phase angle
between applied voltage and current is largely due to the ratio L/R.
Low inductance indicates as short as possible a length of wire in the
bifilar helix; a condition which, for a given resistance, requires wire of
small diameter. This is at odds with mechanical stability, which argues
for large diameter wire, and so a compromise is required.

With careful consideration of all the above factors, Wilkins has produced
resistors whose dc and ac resistances are the same within 0.1 ppm to
frequencies as high as 1.6 kHz.

3: AC: POSSIBLE EFFECTS ON SPRTS

An SPRT winding is not mechanically or electrically dissimilar from the

winding of a standard resistor. It is a spaced bifilar winding in which
the current path is in one direction for approximately half of the wind-
ing and reverses for the other half, so that it approximates a non-in-
ductive winding, but may be capacitive. Many precise measurements of
both Leeds and Northrup pattern SPRTs and those manufactured by
Isothermal Technology seem to indicate that these effects are sufficiently
small so that the measurement deviation due to electrical reactance is
not more than 0.1 part per million at low impressed frequencies, com-
pared to d-c measurements.

(At high temperatures other a-c effects have been noted., Studies of the
best insulations used for the former6 on which SPRT windings are sup-
ported indicate that the degradation of the insulating properties of
quarts and sapphire is marked. I have measured insulation resistance in
a thermometer of 120 MP per square at 960-C, which drops to about 20
MP per square at 11OO’C. 20 MP of random shunt resistance cannot be
ignored, and this effect is a prominent reason why ITS-90 terminates
the SPRT range at the silver freezing point (961’C) instead of the gold
freezing point .(1064’C).) Such a shunt effect would be, of course, as
undesirable with dc as with ac, but much of the effect seems to be due
to a long time-constant polarization, so that resistance rises over a time
which may be measurable in minutes. Thus the recovery of most ~of the
insulation resistance would take place with a direct current impressed,
but might not within the reversal time of an alternating current.
 72

4: DC BRIDGES FOR RESISTANCE THERMOMETRY

Resistance measurements can be made by deflection (magnitude of the

unbalance of a circuit which is balanced at some value) or by balance.
Deflection methods are never used in the precise determination of re-
sistance (except as the deflection of the galvanometer is used to further
quantify the last place on the bridge dials).

An elementary (Wheatstone)
bridge is shown schematically
in Fig. 7, where A and B are
fixed resistors (often of
identical value, in which case
the bridge is said to be
“equal-arm”, a resistance
decade S, and an unknown
resistance to be measured, X.
At balance, when the unknown
resistance is balanced by the
decade resistance (assuming A
= B), the potential drop from
Junction 1 to Junction 2 is
equal to the potential drop
from Junction 1 to Junction 4,
so that i,bA = ixsX, and iabB
= rxsS. These equalities can
FIGURE 7 be expressed as
A WHEATETONE BRIDGE
A/B = X/S

which is the equation of balance for the Wheatstone bridge. At this bal-
ance, no current flows in the galvanometer arm G.

The equalities of the balance condition include the entire bridge circuit,
including any external lead wires which extend to the resistance, X,
which is the subject of the measurement. These external resistances may
be negligible, for example is X is very large in comparison with Rll and
R12, but in an elementary bridge, they are always a component of the
measurement in series with X. In industrial resistance thermometry, lead
resistances can be partially compensated, or almost completely compen-
sated under specific restrictions, (e.g., by inserting a dummy loop of
lead wire in the opposite side of the bridge), but must be eliminated for
precise and standards-quality resistance thermometry. Figs. 8(a) and
8(b) show two methods for making such connections. Obviously, the
 degree of compensation depends upon the other resistances in the
circuit; for example, compensation is more closely. approximated if the
bridge is constructed with equal ratio arms.

Sensor

FIG 8A FIG 8B

Bridge connections for an industrial platinum resistance thermometer, pro-

viding partial lead resistance compensation. In 8A, (a 3-wire thermometer)
Ll is in series with the battery, where it is effectively not a part of the
bridge balance circuit. L2 is in series with RA, and L3 in series with the
sensor. In 8B, (a 4-wire thermometer) Ll and L2 are in series with the
sensor, while a dummy loop (L3 and L4) comprising a length of lead wire
equal in resistance to. Ll t L2 is in series with the bridge balance resistor.
The degree of compensation depends upon the bridge values and the
bridge unbalance.

5: D-C BRIDGES: THE MUELLER BRIDGE

The most commonly used d-c bridge for four-terminal measurements of

standard platinum resistance thermometers, from the time of its design
in 1916 until the advent of more modern inductive bridges, was devel-
oped by Mueller at the NBS (5). McLaren’s seminal work in developing
the metal freezing points as calibration standards was done with an L&N
Mueller G-2 Bridge (Mat calls it a G-2*, the & to recognize a least
decade he added, and my own work on the gallium melt point was done
with a Rubicon version of the G-2. The Mueller bridge represents a
modified Wheatstone bridge with a range up to 81.111 Q, 111.111 Q, or
422.1111 Q (6) full-range.

(Anyone visiting NIST will be interested to have a look into the Museum,
which is rather well concealed behind a door to the right of the main
 74

entrance to the Library, in the Administration Building. Many beautiful

original devices are there on display, and among them is Mueller’s first
bridge, which he made with a sheet of marble as its top panel).

The bridge is always used with equal ratio arms of moderate individual
resistance, e.g., 500 Q or 3000 Q. An small slide wire is provided which
is used to balance these resistances exactly, by interchanging them.
Once the arms are equal, the zero resistance value can be determined
precisely. Commutators are supplied for these adjustments.

Consider the bridge circuitry shown in Fig. 9. (A complete diagram of

the Leeds and Northrup G-3 Mueller Bridge is shown in Fig. 10). The
four lead c, C, t and T of the thermometer are connected as shown, and
are integrated into the bridge circuitry as follows:

In the commutator switch po-

sition of the left-hand circuit,
lead c connects into the bat-
tery circuit, where its resis-
tance is of no consequence. In
the commutator position of the
right-hand circuit, lead t is
in the battery circuit. Poten-
tial leads C and T are in op-
posite lower arms of the
bridge circuit, and are
switched to the alternative
arm by the commutator. Thus
battery leads are exchanged
and connections to the decade
C ct T arm and the fixed arm of the
Therm bridge are exchanged simulta-
Therm
neously. The commutator is a
switch whose contacts are
mercury-wetted, which, if
FIGURE 9 they are clean and in good
condition, will add contact
A schematic Mueller Bridge resistance of, at most, several
microhms.

Since the bridge ratio is 1:1, if the C and T leads are equal in resis-
tance, they will cancel each other, and the net resistance added to the
X arm of the bridge is only the resistance of the thermometer coil. At
the level of precision required of this measurement, leads C and T can-
not be assumed to have equal resistance. A reversal of leads C and T
 FIGURE 10

The complete circuit of the Leeds and Northrup G-3 Mueller Bridge. The
least dial is 10 pQ. Bridge arm resistances of 500 or 3000 P may be chosen.
Note that the x0.1 Q decade is subtractive.

(Courtesy Leeds and Northrup Co., North Wales, Pennsylvania)

 76

would result in a slightly different balance. The correct measured resis-

tance of the thermometer coil alone, which is what is wanted, is as
follows, where RI is the result of the first balance and R2 is the result
of the second balance:

R1+C =RT+I, orRI=RTtT-C

where RT is the resistance of the thermometer coil alone, and

R2 + T = RT t C, or R2 = RT t C - T

The average, then, is

(RI + R2)/2 = RT t T - C t RT + C - T

= ;RT/2 = RT

In the manufacture of most SPRTs care is taken to adjust lead resis-

tances so that the difference in balance between the normal and re-
versed position of the commutator switch requires, usually, adjustment
of only the lowest dial.

Two design characteristics of the Mueller Bridge are elegant enough to

be worth noting. The first is the nature of the decade switches, which
must be constructed so as to be free from unwanted contact resistances,
even though on the lowest dial the increment of resistance (on the L&N
G-3 Bridge) is 1OpQ (0.00001 Q). No physical switch has contact resis-
tances which would not represent an uncertainty of many times that
level (the best switches, properly maintained and lubricated, are not
better than 0.001 Q uncertain), and the friction of switching frequently
generates spurious and transient thermal emfs. Nor is it possible to
make fixed decade resistors of micro-ohm values. The lower decades of
the Mueller bridge, x 0.1 Q and lower, employ Waidner-Wolff decades.

When a shunt is applied to a resistor, the change in resistance 6R is

6R = R - RS/(R t S) = R2/(R t S)

The reduction in resistance from that of the unshunted resistor is

(10 - n)6R;

R2/(R t S) = (10 - n)6R

and the shunt required is

 77

S q [(R2/(10 - n)6RJ - R

Fig. 11 shows a Waidner-Wolff decade, in which a fixed series resistor,

Rs, is shunted by shunts, Rl...RlO, that can be varied in 11 steps from
0 to infinity. Rs2 should be exactly divisible by all the integers below
10. One possible value for Rs2 , used in the Figure, is 50.4. The minimum
shunt resistance is obtained when the switch is set at 0; for the X0.0001
Q decade, Rl = 50.4 - $0.0504.

FIGURE 11

A WAIDNER-WOLFF DECADE OF A MUELLER BRIDGE

Rs

R6 R7 R8 R9 R10 RI1

Resistors Rl through Rll are the same for all decades. They shunt Rs,
adding one increment of resistance per dial step. The values shown are one
of a number of possible sets of value.

For all decades: Specific to the decade shown:

Decade Rl
R2 = 5.6 Q x 0.1 pi; 50.4 - J50.5
R3 = 1.0 Q X 0.01 50.4 - 45.04
R4 = 9.0 Q X 0.001 llo:504 50.4 - JO.504
R5 = 12.0 Q x 0.0001 JO.0504 50.4 - JO.0504
R6 = 16.8 Q
RI = 25.2 Q Note that with all dials set at 0, the value of the
R8 = 42.0 Q decade string is not zero. This non-zero resistance
R9 = 84.0 Q is balanced by an equal and trimmable resistor in
RlO = 252.0 Q the opposing arm of the bridge.
Rll = m

If the ratio R2/6R is made the same for all decades, then identical sets
of fixed coils per step may be used for all decades. The resistances
which must be put in series with the Rl, when Rs2 = 50.4, are shown in
Fig. 11
 78

The effect of this arrangement is to place a large resistance in series

with any sliding switch contact, so that variations in contact resistance
are negligible. The sum of the decades in their zero positions is, how-
ever, not zero but is the sum of the resistances in series with the
sliding contacts, in the example of Fig. 11, 9.167 Q, and a compensating
resistor is inserted into the opposite leg of the bridge to exactly
balance this residual resistance.

A second feature of the Mueller bridge is provision for self-calibration.

Since there are 11 positions on each decade switch (0 through X = lo)
any resistance which is the maximum resistance of a dial is the resis-
tance of the first step of the next higher dial. The bridge can be self-
calibrated in terms of internal bridge units using no more equipment
than a stable decade resistance box, and to calibrate in terms of abso-
lute ohms, only one standard resistor, preferably 10 Q, is necessary. The
technique is described in manufacturers’ manuals. It consists simply in
first setting a decade dial to 1 and the next lower decade to 0, using
the next lower dials to balance an appropriate input resistance; and
then, without changing the external resistance, setting the decade dial
to 0 and the next lower dial to X, rebalancing with the lower dials, and
noting the difference. Thus each step of the bridge dials can be com-
pared with the next lower dial, and finally the lower dial can be cali-
brated in terms of galvanometer deflection.

6: DIGITAL OHM METERS

It is tempting to think of multi-digit high-resolution ohm meters as

readouts for SPRTs. In general, they are not satisfactory. Most St and
9f digital ohm meters present much more current through the ther-
mometer than the 1 mA level which is the accepted level for calibration,
and none that I know of permit two current levels (say 1 mA and 42 or
1.414 mA) which would allow the desirable extrapolation to zero-power
resistance, to eliminate the effects of 12R heating. (As an aside, this
qualification becomes even more important for a user who measures the
resistance of industrial temperature sensors of low dissipation, such as
small industrial PRTs and thermistors).

7: A PRECISION CURRENT COMPARATOR WITH DC THERMOMETER EXCITA-

TION

All conventional dc bridges, such as the Mueller Bridge, depend upon a

number of decades of precision fixed resistors. Such resistors are sub-
ject to drift in value, due to a number of factors; strain or the relief of
residual strain, impressed current, atmospheric and other contamination.
Recalibration6 are not difficult to do, using equipment generally
 79

available in the laboratory and one stable fixed resistor of known

characteristics, but must be done at appropriate and perhaps frequent
intervals. The resistors themselves have temperature coefficients.
Attempts to eliminate temperature effects by thermostating the bank of
resistors leads to the generation of thermal e.m.fs at the junctures of
resistance alloys and copper circuitry wiring; the tradeoff is the need
to make temperature corrections.

An alternative to decade resistors is decade inductors. Toroidally wound

inductive dividers and transformers provide a means for generating
very accurate ratios of voltage or current equal to an integer ratio of
turns. Since there is no such thing as dc inductance, bridges which use
inductors require ac excitation, which may be at low frequencies. In
theory, at least, the position of a tap on a well-made inductor is fixed
and stable; if the entire inductor is isothermal the tap position should
be effectively unchanged with environmental changes. Any thermals gen-
erated are dc and cancel in ac circuitry.

Based on the work of Kusters et al (7) at the National Research Council

(Canada), Guildline developed its Model 9975 Precision Current
Comparator. This comparator is unique among modern bridges in that,
while the inductors require ac, the signal applied to the galvanometer is
dc. The principle is simple, although the execution is sophisticated, and
is illustrated in Fig. 12. There are three windings on a high-
permeability toroidal core, suitably shielded electrostatically and
magnetically. If ampere turns IINI and 19N9 are equal and opposite (that
is, in a condition of balance) there will be zero flux in the core and
consequently zero voltage induced in the detector winding. However in
practice, the primary and secondary currents of a current transformer
are never exactly equal. A special core design makes it possible, in the
9975, to adjust to zero flux with a small compensating current.

The operation is as follows. Current from a power supply flows through

an adjustable number of turns of the comparator (Nx) and through the
thermometer (Rx). A second current flows through a fixed number of
turns (NE) and through the reference resistor (Rs). When the bridge is
in balance, both the net ampere-turns imposed on the comparator cores,
which are indicated on an ampere-turn balance meter, and the difference
between the voltages across the thermometer and the reference resistor,
as measured by the sensitive light-beam galvanometer, must not change
as the currents through both are reversed. Reversal may be automatic,
at a selected rate, or manual. This bridge is capable of measurements of
1 part in 10 million, and its expected accuracy and stability is 2 parts
in 10 million. It requires manual manipulation of the decade dials; its
balance and rebalance is not automatic.
 r------A.7TR*wING SIGN*L
-------ItENER*TOR I
80

,- j

FIGURE 12

The circuit schematic of the Guildline Model 9975 Direct Current Comparator.

(Courtesy Guildline Instruments Inc., Orlando, Florida and Smith Falls,

Ontario, Canada)
 81

8: A NEW AUTOMATIC DC CURRENT COMPARATOR

A relative newcomer to the field is the Measurements International Model

6010A DC Resistance Thermometer Bridge (8), announced at the annual
meeting of the National Conference of Standards Laboratories, Al-
buquerque, N.M. in 1991. The bridge appears to be a dc current com-
parator rather like the Guildline 9915, except automatically balanced. (It
does not seem to have the capability to operate in the off-balance mode,
needed to make a chart recorder sweep). According to Duane Brown, VP
Operations, the 20 uncertainty of the bridge is better than 0.1 ppm t 1
least significant bit and linearity and resolution are specified as 0.01
ppm. The measurement range is 0.1 Q to 10 kQ full-scale, and provides
ratio ranges from 0 to 10 RJR,. A suitable number of thermometer cur-
rent choices are provided, all of which may be multiplied by 42. I have
not had an opportunity to see or use this bridge.

9: AN AC AUTOMATIC OR MANUALLY-BALANCED BRIDGE

Automatic Systems Laboratories has developed a line of bridges which

can be used in a truly automatic-balance mode, or manual balance may
be selected (9). (The latter is most handy to have, for it is often de-
sirable to operate a bridg~e in a condition of unbalance approaching bal-
ance; e.g., when a chart record is being made of a rising temperature).

The ac carrier of the F-17 bridge is fixed at 1-i line frequency; for 50
Hz supplies at 75 Hz, and for 60 Hz supplies at SO Hz. This choice al-
lows maximum rejection of noise from the line and from active circuit
components, while providing maximum detector bandwidth. The more so-
phisticated F-18 operates at * and 1% line frequency (25 Hz and 75 Hz
for 50 Hz supplies and 30 Hz and 90 Hz for 60 Hz supplies) selectable.
This feature allows the estimate of any effects on the measurement due
tn carrier frequency, and extrapolation to dc where necessary.

A block diagram of the F-17 and F-18 bridges is shown in Fig. 13. A
stable ac signal is produced by a carrier generator current source. This
drives current through the standard resistor and the unknown resistor,
which are connected in series. The voltage generated across R, is used
as the reference signal to excite the input windings of a multistage in-
ductive divider. The inductive divider’s secondary output is compared
with the voltage appearing across the unknown resistor Rx, in this case,
the SPRT, by the detector circuitry. The inductive divider acts as a
precision ratio transformer. Its taps are adjusted to balance (that is, to
bring to zero) the output to the detector circuit. At balance, the voltage
from the inductive divider is exactly equal and opposite to that appear-
ing across Rx. The output of the inductive divider is also a precise ra-
 F17 and FIB Schematic

Detector Waveform
Generator
Referknce
Switched
current
Source

Standard I II 111111---l:zy- b - fJ Q”adrat”re

I llllll I ^
Quadrature
WO
Detector
I

and Ratio
Transformer

,
Unknwm
Resistor
Lke-
Amplifie?
S!?hed
Gain
Amplifier
q
Phase
Sensitive
---L ’
Low
Pass
-

Detector Filter

FIGURE 13
 83

tio of the voltage across Rs. Since the current flowing through Rs and
R, is identical, the ratio set on the inductive divider is equal to the
ratio Rx/R,.

The nominal accuracy of the F-17 bridge is lppm, and of the F-18,
O.lppm. The F-18 is optimised specifically for SPRT values of 100, 25.5,
2.5 and 0.25 Q. The F-18 has 7: decades, and reads resistance ratios
over the range 0.000 000 0 and 1.299 999 9. A generous selection of op-
erating sensitivities and of thermometer currents (from 0.1 mA to 50 mA
rms, (F-18) including a (2 multiplier) are supplied.

IMPORTANT REFERENCES

In this text, I have relied, as I have for many years, upon two impor-
tant references on fixed resistors and resistance measurement. These
me:

Frank A. Laws, Electrical Measurements, 2nd Edition, McGraw Hill Pub-

lishing Co., New York, 1938

Forest K. Harris, Electrical Measurements, John Wiley & Sons, New York,
Chapman and Hall, London, 1956

FOOTNOTES

(I) Norman B. Belecki, Ronald F. Dsiuba, Brice F. Field, Barry N. Taylor,

Guidelines for implementing the new representations of the volt and the
ohm effective January 1 1990, NIST Tecnical Note 1263 (1989)

(2) James L. Thomas, A new de&&n 04 p-aeci.&on ms.6ieAoa

a, Bur. Standards Jour. Research, Vol 5 (1930)

(3) F. J. Wilkins, M. J. Swan, PaecLaLon ac/dc az6L&ance

dtandandd, Proc. IEEE, Vol 117, No. 4 (1970)

(4) [H. Tinsley Ltd, Standards House, Croyden, Surrey, CR0 4RR,
England]

(5) E. F. Mueller, Wheetatone bnidgea and some uccwoay

WWZ%d.M &OJL R-ce %he~notnat..ty, Nat. Bur. Stand. Bull., 13,
547 (1916)

(6) Leeds and Northrup Models 8067 (G-l), 8069B (G-2) and 8071B (G-3)
Mueller Bridges, respectively. [Leeds and Northup Company, North Wales,
Pennsylvania USA]

(7) N. L. Kusters, M. P. McMartin and R. J. Berry, Rea-istance

Thamome-tay wL.th #ae di/re& -en=2 compoms?olr, Temperature,
Its Measurement and Control in Science and Industry, Vol 4 p 1477,
In&r. Sot. Am (1972) and A d-&ted -en-t compa-ux-to-t bridge 604
 84

4edatanCe thc%e.m~~~, Guildline Instruments [Guildline Instruments

Inc., 4403 Vineland Road, Suite B-10, Orlando Florida USA 32811; also
Smith Falls, Ontario, Canada]

(8) [Measurements International, PO Box 2359, Prescott, Ontario, Canada

and SW Murray Boulevard, Beaverton Oregon 97006, USA]

(9) Models F-17 and F-18 [Automatic Systems Laboratories Ltd., 28 Blun-
dells Road, Bradville, Milton K‘2YlES MK13 7HF, England]