Selection Guide for Pressure Measuring Instruments

Selection Guide S 1833 X 05

Published by Evaluation International, April 2005

Index classification 2.1

SELECTION GUIDE FOR TEMPERATURE INSTRUMENTATION

INTERNATIONAL INSTRUMENT USERS’ ASSOCIATIONS EVALUATION INTERNATIONAL-WIB-EXERA

CIRCULATION

This report has been prepared for EI, WIB & EXERA members. Because of the general interest shown in it, and because it contains only published information freely provided for the purpose by equipment manufacturers, or objective information derived by independent, impartial research, it is made available to non-member organisations.

The normal comprehensive evaluation reports produced by EI-WIB-EXERA remain confidential to members of these organisations.

About EI, WIB and EXERA EWE is the acronym for three international instrument users' associations which collaborate in the sponsoring, planning and organisation of instrument evaluation programmes. They have the long term objective of encouraging improvements in the design, construction, performance and reliability of instrumentation and related equipment.

The evaluation of the selected instruments is undertaken by approved, independent laboratories with respect to the manufacturers' performance specifications and to relevant international and national standards.

Each evaluation report describes the assessment of the instrument concerned and the results of the testing. No approval or certification is intended or given. It is left to the reader to determine whether the instrument is suitable for its intended application. All reports are circulated throughout the entire membership of EWE.

Evaluation International - International Instrument Users' Association The Pool House, South Hill, Chislehurst, Kent, England BR7 5EH

International Instrument Users' Association WIB Prinsessegracht 26, 2514 AP, The Hague, The Netherlands

EXERA Association des Exploitants d'Equipments de Mesure, de Regulation et d'Automatisme, 9 Rue de Rocroy, 75010, Paris, France.

International Instrument Users' Associations

EWE Membership List

January 2005

Acetex Chimie E Heineken Technical Services W Aeroport de Paris E Infraserve/Hoechst W Agence de L'Eau Artois Picardie E Institut National de l'Environment E Industriel at des Risques Air Liquide E Institut National de Recherche & Securite E Akzo Nobel Engineering W Institut de Regulation et Automation E ASM Brescia E Italcimenti/CTG E Atofina Italie E Jacobs Engineering BV W Aventis Pasteur E KEMA Nederland BV W BNFL plc EI Laborelec W BP-Amoco EI Lubrizol France E BP France E MEMC E British Energy plc EI Nancie E CETIAT E Nantes Metropole – Direction de l’Eau E CETIM E Nederlands Meetinstituut W Chiyoda Corporation W Nestec Ltd W COGEMA E Petro SA W DGA/DCE E Polimeri Europa E DOW Benelux W RATP E DSM Technopartners BV W Renault SA E Du Pont de Nemours BV W Rhoditech E EADS LV E Saudi Arabian Oil Company EI Electricite de France (EDF) E Severn Trent Water EI ENEL E Shell France E EniACQUA E Shell Global Solutions International W Environment Agency EI Solvay BV Benelux W ExxonMobil USA W SNCF E Federelettrica E SIP Standardiserad Instrumentprovning EI Fluor Daniel Consultants BV W Suez Environnement E Gaz de France E Technip E GEMCEA E Total France E/W Generale des Eaux E UKAEA EI Health & Safety Executive EI Universite De Genes E

CONTENTS

1. INTRODUCTION

1.1 Aims of the study 1.2 Scope 1.3 Limits of the study 1.4 Summary of contents

2. OVERVIEW OF THE CONCEPT OF TEMPERATURE, TERMS & DEFINITIONS

2.1 What is temperature? Definition of 2.2 Thermal equilibrium and the Zeroth Law of Thermodynamics 2.3 What is a thermometer? 2.4 Summary of the five main Temperature Scales 2.5 The International System of Units (SI) 2.6 Definition of the SI Unit of Thermodynamic Temperature - Kelvin 2.7 The Multiplying Prefixes of the SI 2.8 Temperature Conversions 2.9 The Spectrum of Temperature 2.10 The Evolution of Thermometry and Temperature Scales 2.11 Practical Realisation of the Definition of Thermodynamic Temperature & The International Temperature Scale of 1990 (ITS-90)

3. CATEGORIES OF TEMPERATURE INSTRUMENTATION

3.1 Liquid-in-glass thermometers 3.1.1 Introduction 3.1.2 Principles and materials of construction 3.1.3 Common defects 3.1.4 Calibration of liquid-in-glass thermometers 3.1.5 Precautions in use 3.2 Dial-type expansion thermometers 3.2.1 Bulbs, wells, and capillaries 3.2.2 Liquid-Filled Systems 3.2.3 Vapour Systems 3.2.4 Gas-Filled Systems 3.2.5 Mercury-Filled Systems 3.2.6 Ambient temperature compensation 3.2.7 Effects of bulb elevation 3.2.8 Barometric errors 3.3 Resistance temperature detectors (RTDs) 3.3.1 Introduction and Overview 3.3.2 Materials and construction 3.3.3 Sources of error 3.3.4 Measurement of RTD resistance 3.4 Thermistors 3.4.1 Overview 3.4.2 Sensor types 3.4.3 Self-heating effect 3.4.4 Measuring bridges 3.4.5 Thermistors combined with resistors 3.4.6 Specialised applications 3.4.7 Calibration and testing 3.5 Thermocouples 3.5.1 Overview 3.5.2 Thermocouple materials 3.5.3 Hardware and fabrication 3.5.4 Emf measurement

3.5.5 Calibration 3.6 Radiation thermometers or pyrometers 3.6.1 Overview 3.6.2 Optical systems 3.6.3 Selecting a radiation thermometer 3.6.4 Calibration 3.6.5 Precautions necessary in the use of radiation thermometers 3.7 Bimetallic thermometers 3.8 Calibrators and simulators 3.9 Integrated circuit (IC) transistors and diodes 3.10 Quartz crystal thermometry 3.11 Ultrasonic thermometers 3.12 Miscellaneous temperature sensors 3.12.1 Self-measuring devices 3.12.2 Acoustic time domain reflectometry 3.12.3 Carbon resistors 3.12.4 Capacitance cable for detecting hot spots 3.12.5 Fluidic sensors 3.12.6 Johnson noise thermometer 3.12.7 Liquid crystals 3.12.8 Paramagnetic salts 3.12.9 Spectroscopic temperature measurement 3.12.10 Thermography 3.12.11 Colour Indicators, Crayons and Pellets 3.12.12 Pyrometric Cones

4. GENERAL SELECTION CRITERIA

4.1 Temperature range 4.2 What is the smallest temperature change you need to record? (required resolution) 4.3 How well the temperature must be known (required accuracy) 4.4 External environment (atmospheric effects) 4.5 Type of use 4.6 Use of thermowells 4.7 Other factors

5. REFERENCES

APPENDICES

APPENDIX A RELEVANT BRITISH/EUROPEAN STANDARDS APPENDIX B HISTORICAL EVOLUTION OF THE THERMODYNAMIC AND PRACTICAL TEMPERATURE SCALES APPENDIX C THE ROLE OF THE PLATINUM RESISTANCE THERMOMETER IN THE ITS-90 APPENDIX D TOLERANCE CLASSES FOR Pt100 THERMOMETERS (IEC 751: 1983) APPENDIX E Pt100 REFERENCE TABLES (IEC 751: 1983) APPENDIX F TOLERANCE CLASSES FOR THERMOCOUPLES (IEC 584-2: 1982) APPENDIX G THERMOCOUPLE REFERENCE TABLES (IEC 584-1:1995) APPENDIX H DATA TABLES FOR RADIATION THERMOMETERS OR PYROMETERS APPENDIX I EXAMPLE INSTRUMENT SPECIFICATION: LABFACILITY TEMPMASTER 100 DIGITAL THERMOMETER APPENDIX J PARTIAL LIST OF UK MANUFACTURERS AND SUPPLIERS OF TEMPERATURE INSTRUMENTATION

LIST OF TABLES

LIST OF FIGURES

SELECTION GUIDE FOR TEMPERATURE INSTRUMENTATION

Author: S Croft MSc MInstMC

April 2005

1. INTRODUCTION

Sira Test & Certification Ltd (Sira) on behalf of Evaluation International (EI) has compiled this report. It has been compiled from technical and commercial literature in the public domain and from information data sheets available from manufacturers, agents or sales representatives. In addition, it draws upon Sira’s twenty years experience in the realm of independent temperature instrument evaluation and calibration.

1.1 Aims of the study

The aim of this study is to produce a document (for use by the EI membership) to assist a prospective purchaser in selecting an appropriate temperature measuring instrument for his/her particular application. It will identify the main categories of temperature measuring instrument widely available in industry, consider known pros and cons with each, and discuss general selection criteria. An example instrument specification is also included.

1.2 Scope

This document provides advice for those wishing to select and use instruments for measuring temperature. It introduces the main concepts of temperature measurement, together with terms and definitions. The main categories of measuring instrument are discussed as well as principal selection criteria; general parameters associated with each instrument type are identified, together with known pros and cons. An example instrument specification will be given, and a limited selection of market- leading UK manufacturers will be listed.

1.3 Limits of the study

This document does not seek to identify all manufacturers/models of temperature instrumentation as it is estimated that there are many thousands of different models currently available in the UK alone. Instead, this document will steer the user into asking the right questions when selecting an appropriate temperature measuring instrument.

The study will be limited to categories of instrument widely used in industry. While this document provides a general overview of temperature measurement, it is not an in-depth scientific treatment of the subject.

The document primarily covers temperature measurements made within the range –272°C (cryogenic temperatures) to 3800°C, this being the range most relevant to industrial measurement. Some techniques for making measurements outside this range are covered but only in outline.

1.4 Summary of contents

Section 2 discusses concepts, terms and definitions associated with temperature measurement, as well as identifying temperature units and conversions. Section 3 identifies the main categories of industrial temperature instrumentation (together with some lesser known examples), grouped mainly according to operating principle. A detailed description of each type is given; typical parameters such as operating range, resolution and accuracy are included. A summary of known pros and cons is listed for each of the main instrument types. Section 4 briefly outlines principal selection criteria, such as range of temperature being measured, required ‘accuracy’ in use, cost of installation, environmental conditions, and so on. Relevant British/European temperature standards are listed in

Appendix A, whilst Appendix B details the historical evolution of the thermodynamic and practical temperature scales. Appendix C identifies the role of the platinum resistance thermometer in the ITS- 90. Appendix D lists the tolerance classes for Pt100 thermometers (IEC 751: 1983), and Appendix E provides Pt100 reference tables (IEC 751: 1983). Appendix F lists the tolerance classes for thermocouples (IEC 584-2: 1982), and Appendix G provides thermocouple reference tables (IEC 584- 1:1995). Appendix H provides some data tables for radiation thermometers or pyrometers. Appendix I gives an example instrument specification: Labfacility Tempmaster 100 digital thermometer. Appendix J lists a limited selection of market-leading UK manufacturers and suppliers of temperature instrumentation.

Note: all data, standards and manufacturers specifications contained or referenced within this guide are current at the time of going to press but the user should always check for the latest edition.

2. OVERVIEW OF THE CONCEPT OF TEMPERATURE, TERMS & DEFINITIONS

The accurate measurement of temperature is vital across a broad spectrum of human activities, including industrial processes (e.g. making steel), manufacturing; monitoring (in food transport and storage), and in health and safety. In fact, in almost every sector, temperature is one of the key parameters to be measured.

2.1 What is temperature? Definition of

The Oxford English Dictionary definition of temperature is as follows:

“the degree or intensity of heat present in a substance or object” — ORIGIN originally in the sense ‘the state of being tempered or mixed’: from Latin temperatura, from temperare ‘restrain’.

Another qualitative definition would be:

“the degree of hotness or coldness of a body or environment (corresponding to its molecular activity).”

The everyday concept of temperature is naturally associated with sensations of hot and cold and, although such sensations are not very reliable, the principle is nevertheless correct. A definition in simple terms is that ‘temperature is the potential for heat flow’, with analogies to fluids in pipes. More generally, to include radiative processes:-‘Temperature is the potential for heat transfer’.

Thus an object at a higher temperature will lose heat to an object at a lower temperature when they are placed in contact. We all know what we mean when we talk of 'the temperature' of an object: it is just a measure of how hot or cold it feels. But our sense of temperature is qualitative rather than quantitative. Engineers and scientists can routinely create temperatures from very close to absolute zero, up to a few thousand degrees. To understand what we mean by temperature when we are dealing with situations well outside our experience we need a more specific definition of temperature and a reliable numerical scale. The understanding of how this can be done came with the development of thermodynamics and statistical mechanics in the nineteenth century. One important deduction can be stated as:

‘The temperature of a substance (or of an object) is a measure of the average energy of the microscopic objects of which that substance is made’.

However, one or two points need explaining further:

The "microscopic objects of which the substance is made" are usually the atoms or molecules. However, if the measurement being made is of the temperature of a blackbody, then the microscopic objects are the photons which make up the electromagnetic field within the blackbody cavity.

The "average energy" is slightly loose terminology, but is a fair description for many purposes.

The "temperature" refers to the absolute temperature which is measured in units of the kelvin (K). The temperature of a substance at 0ºC is actually 273.15 K which is clearly far from zero, indicating that the atoms and molecules of a substance at 0ºC still have a great deal of energy.

2.2 Thermal equilibrium and the Zeroth Law of Thermodynamics

Temperature is based on the idea of thermal equilibrium to quantify ‘hotness’ (this is the macroscopic view). If two (or more) objects are in thermal equilibrium then, by definition, they are at the same temperature. On a microscopic level, atoms and molecules within a body are in constant random motion; temperature is a measure of this random kinetic energy.

If a hot object is placed in contact with a cold object, the hot object gets colder and the cold object gets hotter, until they don’t change any more. At this point they are said to have reached ‘thermal equilibrium’ and are at the same temperature (figure 2.1).

Figure 2.1 Illustration of thermal equilibrium

The Zeroth Law of Thermodynamics states: “If body A is in thermal equilibrium with body B, and body B is in thermal equilibrium with body C, then body A is also in thermal equilibrium with body C”. This is illustrated in figure 2.2.

This law allows us to know whether objects are at the same temperature, even when they can’t be placed in thermal contact with each other. It also allows temperature to become a reproducible, quantifiable concept (body B could be a thermometer, for example).

Figure 2.2 Illustration of the Zeroth Law of Thermodynamics

2.3 What is a thermometer?

A thermometer is an instrument that measures the temperature of a system in a quantitative way. The easiest way to do this is to find a substance having a property that changes in a regular way with its temperature. This is known as the thermometric property. The most direct 'regular' way is a linear one which is expressed by the following equation:

t(x) = ax + b

where t is the temperature of the substance and changes as the property x of the substance changes. The constants a and b depend on the substance used and may be evaluated by specifying two temperature points on the scale, such as 0° for the freezing point of water and 100° for its boiling point.

For example, the element mercury is liquid in the temperature range of -38.9°C to 356.7°C. As a liquid, mercury expands as it gets warmer, its expansion rate is linear and can be accurately calibrated.

Figure 2.3 Mercury-in-glass thermometer

The mercury-in-glass thermometer illustrated in figure 2.3 above contains a bulb filled with mercury that is allowed to expand into a capillary. The degree of expansion is calibrated on the glass scale in temperature units.

A thermometer can be any device which has a measurable property which changes with temperature:

• For a liquid-in-glass thermometer, the property measured is the length of the liquid column inside a glass tube.

• For a platinum resistance thermometer or for a thermistor, the property measured is the electrical resistance of a piece of 'sensing' material.

• For a thermocouple, the property measured is the voltage generated along the wires making up the thermocouple.

• For a pyrometer (radiation thermometer), the property measured is the current generated by a photodiode situated at the focus of a lens system.

Of course, nearly everything changes in some way with temperature, but not everything is a thermometer. So to qualify as a useful thermometer, a device must have some other properties too:

• It must be reproducible. This means that whatever the measured property of the device, that property should have the same value (or very nearly so) whenever the temperature is the same. • It must be insensitive to things other than temperature. This means that whatever the measured property of the device, that property should not depend on factors such as the humidity or pressure, or on the materials of which it is made, or on special properties of the thing being measured such as its colour or size. • It must be calibrated. This means that we must know how to convert the measured property (length, resistance, etc) to temperature. To do this, the device must be exposed to some environments where the temperature is known, and the value of its measured property must be recorded in those environments. In some cases, for example in a mercury thermometer, the scale reads directly in temperature, and in this case calibration serves to show how accurate the thermometer scale is. • It should be convenient to use. Factors such as size, cost, speed of response, ruggedness, immunity to electrical interference, etc, will be important to varying degrees in different applications.

2.4 Summary of the five main Temperature Scales

There have been five main temperature scales, each one being named after the person who invented it.

G D FAHRENHEIT (1686-1736) a German physicist, in about 1714 proposed the first practical scale. He called the freezing-point of water 32 degrees (so as to avoid negative temperatures) and the boiling-point 212 degrees.

R A F de REAUMUR (1673-1757) A French entomologist, proposed a similar scale in 1730, but set the freezing-point at 0 degrees and the boiling-point at 80 degrees. This was used quite a bit but is now obsolete.

Anders CELSIUS (1701-1744) a Swedish astronomer, proposed the 100-degree scale (from 0 to 100) in 1742. This was widely adopted as the centigrade scale. But since grades and centigrades were also measures of angle, in 1947 it officially became the Celsius scale. Also, the SI system of units gives preference to naming units after people where possible.

William Thomson, 1st Lord KELVIN (1824-1907) a Scottish mathematician and physicist, worked with J P Joule - about 1862 - to produce an absolute scale of temperature based on laws of heat rather than the freezing/boiling-points of water. This work produced the idea of 'absolute zero', a temperature below which it was not possible to go. Its value is -273.15 degrees on the Celsius scale.

William J M RANKINE (1820-1872) a Scottish engineer and scientist, promoted the Kelvin scale in its Fahrenheit form, when the equivalent value of absolute zero is -459.67 degrees

Fahrenheit.

Nowadays, while scientists use the KELVIN scale, the CELSIUS scale is the preferred scale in our everyday lives. However, the Fahrenheit scale is still widely used and there frequently is a need to be able to change from one to the other.

Table 2.1 Summary of the five main temperature scales

2.5 The International System of Units (SI)

In 1960 the 11th General Conference on Weights and Measures (Conference Generale des Poids et Mesures, abbreviation CGPM) adopted the name Système International d'Unités (or SI), for the recommended practical system of units of measurement.

The 11th CGPM laid down rules for the prefixes, the derived units, and other matters. The base units are a choice of seven well-defined units which by convention are regarded as dimensionally independent: the metre, the kilogram, the second, the ampere, the kelvin, the mole, and the candela. Thermodynamic temperature (the kelvin) is one of the base units. Derived units are those formed by combining base units according to the algebraic relations linking the corresponding quantities. The names and symbols of some of the units thus formed can be replaced by special names and symbols which can themselves be used to form expressions and symbols of other derived units.

Quantity Unit name Unit symbol

Length metre m Mass kilogram kg Time second s Electric current ampere A Thermodynamic temperature kelvin K Amount of substance mole mol Luminous intensity candela cd

Table 2.2 The seven base units of the SI

2.6 Definition of the SI Unit of Thermodynamic Temperature - Kelvin kelvin [K] The kelvin is the basic unit of thermodynamic temperature. It is the fraction 1/273.16th of the thermodynamic temperature of the triple point of water. It is named after the Scottish mathematician and physicist William Thomson 1st Lord Kelvin (1824-1907).

The triple point of water is the unique temperature and pressure at which the three phases of water (solid, liquid and vapour) co-exist in equilibrium. It is fractionally higher than the melting point, being 0.01°C or 273.16 K.

The definition of the unit of thermodynamic temperature was given in substance by the 10th CGPM (1954, Resolution 3) which selected the triple point of water as the fundamental fixed point and assigned to it the temperature 273.16 K, so defining the unit. The 13th CGPM (1967-1968, Resolution 3) adopted the name kelvin (symbol K) instead of "degree Kelvin" (symbol °K).

Because of the way temperature scales used to be defined, it remains common practice to express a thermodynamic temperature, symbol T, in terms of its difference from the reference temperature T0 = 273.15 K, the ice point. This temperature difference is called the Celsius temperature, symbol θ, and is defined by the quantity equation:

θ = T – T0

The unit of Celsius temperature is the degree Celsius, symbol °C, which is by definition equal in magnitude to the kelvin. A difference or interval of temperature may be expressed in kelvins or in degrees Celsius (13th CGPM, 1967-1968, Resolution 3).

The Kelvin and the degree Celsius are also the units of the International Temperature Scale of 1990 (ITS-90) adopted by the CIPM in 1989 in its Recommendation 5 (CI-1989) (PV, 57, 115 and Metrologia, 1990, 27, 13).

2.7 The Multiplying Prefixes of the SI

The SI allows the sizes of units to be made bigger or smaller by the use of appropriate prefixes. For example, the electrical unit of a watt is not a big unit even in terms of ordinary household use, so it is generally used in terms of 1000 watts at a time. The prefix for 1000 is kilo so we use kilowatts [kW] as our unit of measurement. For makers of electricity, or bigger users such as industry, it is common to use megawatts [MW] or even gigawatts [GW]. The full range of prefixes with their [symbols or abbreviations] and their multiplying factors (which are also given in other forms) is: yotta [Y] 1 000 000 000 000 000 000 000 000 = 1024 zetta [Z] 1 000 000 000 000 000 000 000 = 1021 exa [E] 1 000 000 000 000 000 000 = 1018 peta [P] 1 000 000 000 000 000 = 1015 tera [T] 1 000 000 000 000 = 1012 giga [G] 1 000 000 000 (a thousand millions = a billion) mega [M] 1 000 000 (a million) kilo [k] 1 000 (a thousand) hecto [h] 100 (a hundred) deca [da]10 (ten) 1 deci [d] 0.1 (a tenth) centi [c] 0.01 (a hundredth) milli [m] 0.001 (a thousandth) micro [µ] 0.000 001 (a millionth) nano [n] 0.000 000 001 (a thousand millionth) pico [p] 0.000 000 000 001 = 10-12 femto [f] 0.000 000 000 000 001 = 10-15 atto [a] 0.000 000 000 000 000 001 = 10-18 zepto [z] 0.000 000 000 000 000 000 001 = 10-21 yocto [y] 0.000 000 000 000 000 000 000 001 = 10-24

[µ] the symbol used for micro is the Greek letter known as 'mu'. Nearly all of the SI prefixes are multiples (kilo to yotta) or sub-multiples (milli to yocto) of 1000. However, these are inconvenient for many purposes and so hecto, deca, deci, and centi are also used. deca also appears as deka [da] or [dk] in the USA and Contintental Europe.

Table 2.3 The multiplying prefixes of the SI

2.8 Temperature Conversions

For the same temperature, if t, θ and T represent the temperatures on the Fahrenheit (°F), Celsius (°C), and Kelvin scales, respectively, then the relation between them is as follows:

θ = 5/9 (t – 32)

T = θ + 273.15

T = 5/9 (t + 459.67)

Similarly, the relation between the Rankine temperature r and the Fahrenheit temperature t is given by: r = t + 459.67

T = 5/9 r

A temperature interval of one kelvin is identical to that of one degree Celsius, and similarly a temperature interval of one degree Rankine is identical to that of one degree Fahrenheit.

2.9 The Spectrum of Temperature

200 million °C The Joint European Torus (JET) nuclear fusion project, Culham Oxfordshire

15 million °C Temperature of the centre of the Sun

6000 °C Temperature of the surface of the Sun

1200 °C to 1500 °C Molten glass / steel

1064 °C Melting point of gold

100 °C Boiling point at one atmosphere of pressure

0 °C Freezing point of pure water

- 89.2 °C All time coldest point on earth

- 196 °C Cryogenic storage in liquid nitrogen

- 270 °C Cosmic background radiation

2.10 The Evolution of Thermometry and Temperature Scales

The means of accurately measuring temperatures has long fascinated people. One of the differences between temperature and other physical concepts, such as mass or length, is that it is subjective: different people will have different perceptions of what is hot and what is cold. To make objective measurements, we must use a thermometer in which some physical property of a substance changes with temperature in a reliable and reproducible way.

Thermoscopes, the ancestors of modern thermometers, have been around since about 200 BC. The first recognisable, modern thermometers were made in the 16th century by both Galileo Galilei and Santorio Santorio, a physician to the King of Poland. The latter produced a thermometer incorporating a scale, and his writings show that he understood the importance of the temperature measurement in the diagnosis of disease.

The first sealed thermometer was made by the Grand Duke Ferdinand of Tuscany in 1641. This thermometer was more accurate than its predecessors since it wasn’t dependent on atmospheric pressure. Later, the scientists Fahrenheit and Celsius both made glass thermometers containing mercury, and used reference points (the melting point of pure ice and the boiling point of water) to improve the accuracy.

The two temperature scales commonly in use today date from the eighteenth century and are named after Gabriel Daniel Fahrenheit and the Swedish astronomy professor Anders Celsius. Fahrenheit designed his scale to have two reference points that could be set up in his workshop. He originally chose the melting point of pure ice and the temperature of a normal human body, which he took as being 32° and 96° respectively. These conveniently gave positive values for all the temperatures he encountered. Later he changed to using the boiling point of water (212°) as the upper fixed point of the scale.

Celsius also used the ice and steam points, but took them to be 0°C and 100°C respectively. Although the Celsius scale has taken precedence over the Fahrenheit scale, the latter is still familiar in weather

reports in the United Kingdom: a summer’s day temperature of 75°F seems much more pleasant than one of 24°C.

A third, fundamental, temperature scale was proposed in 1854 by the Scottish physicist William Thomson, Lord Kelvin. It is based on the idea of the absolute zero, the point of no discernible energy, which is independent of any particular material substance. The Kelvin scale is widely used by physicists and engineers to determine and apply fundamental laws of thermodynamics.

So why the proliferation of so many temperature scales?

The roll-call of now-obsolete temperature scales includes the Newton, Rømer, Delisle, Leyden, Dalton, Wedgewood, Hales, Ducrest, Edinburgh and Florentine scales. In the 18th century, it was common to have up to four temperature scales (in one example, Newton, De Lisle, Réaumur and Fahrenheit) inscribed on the backing board of a thermometer.

Today, the only scales in everyday use are the Celsius and Fahrenheit scales, and (mainly for scientists and engineers) the Kelvin and Rankine scales. Réaumur may be familiar to readers of 19th- century Russian and French novels; the others are largely forgotten.

The reason for such a proliferation of early temperature scales is that each of the researchers effectively defined temperature according to the thermometric properties of his home-made measurement apparatus rather than in terms of the fundamental physical property we call thermodynamic temperature, which cannot be measured directly.

Among the earliest quantitative temperature scales was that developed in 1692 by Danish astronomer Ole (or Olef or Olaus) Rømer, who had earlier made measurements of the speed of light and developed a standard system of measures for the Danish realm based on the Rhineland foot. He noticed that in summer, his pendulum clock ran slower, and the graduations inscribed on his astronomical instruments were larger than in winter, making it impossible to carry out accurate measurements throughout the year. In order to compensate for this effect, he needed to quantify the thermal expansion of different materials, and to do this, he needed to be able to measure temperature.

In the 17th and 18th centuries, the ‘state of the art’ temperature measurement device was the liquid- in-glass thermometer, a device consisting of a vertical glass tube connected to a closed reservoir or bulb filled with liquid, similar in principle to mercury thermometers except larger. Water is poorly suited for this purpose (one obvious problem: it freezes at a relatively warm temperature), so Rømer used "spiritus vini [approx 40% alcohol], coloured with saffron" in an 18-inch long glass tube of constant cross-section. Isaac Newton, working on the problem at around the same time, used linseed oil instead of alcohol as a thermometric fluid.

To "fix" a temperature scale it's necessary to identify easily reproducible reference points, or "fiduciary points." Newton chose the temperature of melting snow and the temperature of boiling water, and marked off the interval into 33 or 34 "degrees", each corresponding to a certain height of oil in the tube.

Rømer at first used the same reference points, extending his scale to temperatures below freezing, with the freezing point defined as 7.5 Rø and the boiling point as 60 Rø. He later amended his standard for practical reasons so that the reference temperatures were ice water and "blood-warm" (i.e., human body temperature), which he defined as 22.5 Rø. From this we can deduce that on average, one degree Newton is equal to 3.0°C and 1 Rø is equal to approximately 1.9°C.

The Rømer scale was further developed by his successor Horrebow, to take account of the colder temperatures encountered in Iceland and Greenland, and by Daniel Gabriel Fahrenheit, whose modified scale remains in use to this day, mainly in the USA and Jamaica.

Fahrenheit realised that alcohol was unsuitable for precise and repeatable temperature measurements. In 1714, he adopted mercury, which proved an excellent alternative - its coefficient of thermal expansion is highly linear, it doesn't contain dissolved air, and its composition is guaranteed

from one batch to the next, unlike the alcohol of the time. However, it's also much less sensitive to temperature change than alcohol.

Although the use of mercury had originally been suggested by Edmond Halley, the use of mercury thermometers graduated in degrees Fahrenheit became so widespread that the mercury thermometer came to be known as "Fahrenheit's thermometer", and the popularity and longevity of the Fahrenheit scale were assured.

René Antoine Ferchault de Réaumur's scale was developed in 1732, using a thermometer based on a particular mixture of alcohol and water. He also chose as reference points the freezing and boiling points of water, to which he assigned the values 0 R and 80 R respectively. This scale was officially adopted throughout Europe except for Britain and Scandinavia, but beginning with the adoption of the centigrade scale by the French revolutionary government in 1794 it gradually declined in popularity, finally falling into disuse sometime in the 20th century.

The scale of Joseph DeLisle (modified by Weitbrecht in 1738) was an inverse scale, with the boiling point of water set at 0 and the freezing point at 150 degrees. It remained popular in Russia for a century or so.

One temperature scale for everything? A fundamental temperature scale

As for whether just one temperature scale can be used for all purposes, in principle the answer is yes, provided it is an absolute scale or fundamental scale (i.e., one whose zero point is the absolute zero of thermodynamic temperature). Certain physical relationships (such as the equivalence of heat and work, rates of chemical reactions, the ideal gas equations of state, the kinetic theory of gases, and the laws governing heat transfer) depend on multiplying or dividing temperatures. You won't get correct answers if you use a scale with an arbitrary zero point. Therefore around 1860, two absolute temperature scales were developed: the Rankine scale, based on the Fahrenheit degree, and the Kelvin scale, based on the centigrade degree.

Unfortunately, not only do different temperature scales give different numerical readings (because of the arbitrary choice of constant of proportionality by each researcher), but the use of different standard instruments results in different definitions of temperature itself. The relationship between volume and temperature is different for linseed oil, alcohol and mercury. In fact the same is true for any thermometric property available to us, such as electrical resistance or gas pressure. The definition of temperature therefore depends on the instrument chosen as a standard, and two different instruments that read the same temperature at the reference points will in general not agree at other temperatures.

In an attempt to free the modern science of thermodynamics from the ghosts of 18th century instruments, the Kelvin (and hence, Celsius) temperature scale was redefined in 1954 in terms of the interval between absolute zero and a single fixed reference point (namely, the triple point of water, set at 273.16K, or 0.01°C). The thermodynamic temperature of any system could in theory be established using a constant-volume ideal gas thermometer. No longer would the properties of any real substance determine the temperature. This means that the boiling point of water is no longer set at 100°C by definition (in fact the current best available figure is 99.974°C), and so strictly speaking the modern Celsius scale is no longer a "centigrade" scale.

In practical terms, unfortunately, there is no such thing as an ideal gas, and therefore no such thing as an ideal gas thermometer. A real-world temperature scale will therefore necessarily be considerably less elegant. The standard temperature scale now in use is called the "International Temperature Scale 1990" (ITS-90), the result of years of work at the UK's National Physical Laboratory and elsewhere. This uses no less than 16 reference points, from the triple point of hydrogen (defined as 13.8033K), through the melting point of gallium (302.9146K) to the freezing point of copper (1357.77K). Like all its predecessors, it is not perfect; in fact it is an inelegant amalgamation, consisting of 5 discontinuous and overlapping ranges based on three different types of measurement device, and does not yet extend to temperatures below 0.65K. The ITS has been derisively referred to as a "rubber scale," because it stretches or shrinks as improved measurement data becomes available. Nonetheless, it is, for the moment, adequate for our need and ability to measure temperature, and it's the best we've got.

2.11 Practical Realisation of the Definition of Thermodynamic Temperature & The International Temperature Scale of 1990 (ITS-90)

Since 1954 the unit of (thermodynamic) temperature has been defined as the kelvin, and is the fraction 1/273.16 of the thermodynamic temperature of the triple point of water. From this single point it is possible to generate a thermodynamic temperature scale using gas thermometers and radiation thermometers which respond strictly in accordance with physical laws.

Such experiments are not easy and are rarely done, but good values have been established for a series of fixed points: freezing points of pure metals at high temperatures and triple points of gases at low temperatures. These are incorporated into the International Temperature Scale so that standard platinum resistance thermometers and radiation thermometers can be calibrated with excellent reproducibility.

Figure 2.4 Water triple point cell

The National Physical Laboratory maintains the temperature scale (currently the International Temperature Scale of 1990, the ITS-90) in the UK, and compares this with the ITS-90 maintained in other national laboratories. In this way temperature standards around the world can be accurately equivalent, and all manner of thermometers can be reliably calibrated for everyday use.

Direct measurements of thermodynamic temperature can only be made by using one of a small number of so-called primary thermometers. These are thermometers whose equation of state can be written down explicitly without having to introduce unknown temperature-dependent constants. Primary thermometers that have been used to provide accurate values of thermodynamic temperature include:-

■ the constant-volume gas thermometer ■ the acoustic gas thermometer ■ the spectral and total radiation thermometers and ■ the electronic noise thermometer

Uncertainties of a few millikelvins have been achieved with such thermometers up to about 373 K, beyond which the uncertainties increase progressively. The use of such thermometers to high accuracy is difficult and time-consuming and there exist secondary thermometers, such as the

platinum resistance thermometer, whose reproducibility can be better by a factor of ten than that of any primary thermometer.

In order to allow the maximum advantage to be taken of these secondary thermometers the CGPM has, in the course of time, adopted successive versions of an international temperature scale. The first of these was the International Temperature Scale of 1927 (ITS-27); this was replaced by the International Practical Temperature Scale of 1948 (IPTS-48) which in turn was replaced by the International Practical Temperature Scale of 1968 (IPTS-68). In 1976 the Comité International des Poids et Mesures (CIPM) adopted, for use at low temperatures, the 1976 Provisional 0.5 K to 30 K Temperature Scale (EPT-76). On 1 January 1990 the IPTS-68 and the EPT-76 were replaced by the International Temperature Scale of 1990 (ITS-90), adopted by the CIPM in 1989 in its Recommendation 5 (CI-1989). Appendix B shows the full historical evolution of the thermodynamic and practical temperature scales.

The 19th CGPM (1991, Resolution 3) recommended that national laboratories continue their efforts to improve the world-wide uniformity and long-term stability of temperature measurements by rapid implementation of the ITS-90.

The ITS-90 extends upwards from 0.65 K to the highest temperature measurable using an optical pyrometer. The scale is based on:-

1. a set of defining fixed points, and 2. specified methods of interpolating between them

The defining fixed points are the temperatures assigned by agreement to a number of experimentally realisable thermodynamic states and the interpolations are defined in terms of:-

■ the helium vapour-pressure equations from 0.65 K to 5 K ■ interpolating constant-volume gas thermometers from 3 K to 24.5561 K ■ platinum resistance thermometers from 13.8033 K to 961.78°C, and ■ the Planck radiation law at higher temperatures

In several ranges of temperature more than one definition of T90, the temperature defined by the Scale, exists. The various definitions have equal validity.

Advice on the realisation and implementation of the ITS-90 is given in two documents, Supplementary Information for the ITS-90 and Techniques for Approximating the ITS-90, which are approved and updated periodically by the Consultative Committee for Thermometry (CCT) and published by The Bureau International des Poids et Mesures (BIPM).

3. CATEGORIES OF TEMPERATURE INSTRUMENTATION

A thermometer is a device in which some physical property of a substance changes with temperature in a reliable, reproducible and quantifiable way.

There are several different types of practical thermometer, most of which are discussed in some detail within this section. Choosing which one to use depends (amongst other things) on the intended application, temperature range and required accuracy (refer also to Section 4 for general guidance).

3.1 LIQUID-IN-GLASS THERMOMETERS

Figure 3.1 Liquid-in-glass thermometer

3.1.1 Introduction

Liquid-in-glass, in particular mercury, thermometers have been used for almost 300 years in science, medicine, metrology and in industry. They rely on the expansion of a fluid with temperature. The fluid is contained in a sealed glass bulb, and its expansion is measured using a scale etched along the stem of the thermometer.

The most widely used fluid is mercury, covering a temperature range from -38°C to 356°C, although the introduction of a gas into the instrument can increase the range to 600°C or beyond. Other working fluids include ethyl alcohol, toluene and technical pentane, which can be used down to - 200°C.

The specifications of the thermometer are different depending on the required usage. For example, some thermometers can be read to 0.001°C over limited temperature ranges, whilst others can be used up to 550°C with coarser scales.

By the early 18th century, Fahrenheit had made the all-important inclusion of mercury as the filling liquid and he was able to construct the first reliable thermometers which approached very closely the type of instrument with which we are familiar today. They too, were hermetically sealed, and Fahrenheit introduced the cylindrically shaped bulb which for a given volume of mercury increased the available surface area and hence the responsivity.

Shortly before the 20th century exhaustive investigations in Germany determined conclusively the ingredients of the glasses which proved most suitable for thermometric purposes. In France, the International Bureau of Weights and Measures undertook a classic and painstaking programme of work to distribute a scale of temperature called the International Hydrogen Scale (refer to Appendix B). This scale held a position of some importance for about 20 years and was disseminated by calibrating mercury-in-glass thermometers (so-called primary standards) by comparison with a primary hydrogen gas thermometer. Understandably, in the first world war, when the supply of German glass was curtailed, manufacturing interests and techniques had to be established in the UK and the USA.

Liquid-in-glass thermometers continue to offer a measurement accuracy which up to 200°C is only surpassed by the best resistance thermometers and between 200 and 500°C their accuracy is only slightly inferior to the best thermocouples. They offer advantages of stability, simplicity and portability; moreover, they are much cheaper than any comparable system. On the other hand, they require visual reading, and automation of this process has not generally been successful.

Literally hundreds of different patterns of thermometers are currently in use within the limits –200 to 600°C. Over narrow-ranges, at comparatively low temperatures, very high precisions can be obtained. For example, secondary reference thermometers (BS 1900) operating in 10°C ranges between 0 and 50°C can give measurement accuracies of ±0.005°C. Thermometers operating over wider ranges have less finely divided scales, and the best accuracies obtainable at –200°C and +600°C are of the order of 1°C, and roughly proportionate accuracies are attainable at intermediate temperatures.

The lower and upper limits of the total range are governed by, respectively, the availability of suitable thermometric liquids and the softening ranges of the thermometric glasses.

Figure 3.1 above shows an illustration of a liquid-in-glass thermometer. Mercury or some other liquid (alcohol, pentane) fills the glass bulb and extends into the capillary bore of the stem. The space above the mercury column to the sealed top is evacuated, but occasionally it may be filled with an inert dry gas, such as nitrogen, to increase the temperature range. The volumetric expansion of pure mercury is 0.01%/°F (0.005%/°C) and is very linear.

Although bulb and capillary could be made from the same type of glass, it is more convenient to make the bulb from glass with a good stability factor and the capillary from glass which is easier to work. For accurate measurements, the capillary must be properly annealed after it is drawn to the correct bore. Uniformity of bore is desirable but not absolutely necessary if the thermometer is calibrated at a sufficient number of points.

The glass-stem thermometer can be made for a very narrow range in temperature. For instance, consider a clinical thermometer whose full range may be 96 to 102°F (35.6 to 38.9°C) and with a stem 4 in. (101.6 mm) long. Another design feature of some glass-stem thermometers, notably the clinical type, is a restriction purposely placed in the capillary which prevents the liquid from returning toward the bulb when the thermometer is removed from the warmer object. This creates a separation of the column in the stem. In this case, it is a desirable feature because it permits a "highest point" reading or peak point. If the separation occurs inadvertently in any glass-stem thermometer, however, the result is an erroneous reading. The column can usually be rejoined by shaking or tapping.

The design of a glass-stem thermometer requires that the filling material be a liquid over the entire range of temperature desired. Mercury is most suitable and can be used from its freezing point (-38°F, or –39°C) up to nearly its boiling point (over 1000°F, or 538°C). At this upper limit the space above the mercury column must be under great pressure with an inert gas to prevent evaporation of mercury from the top of the column. Alcohol and a few other hydrocarbons may be used for low temperatures. Colourfast dyes are usually added to these liquids to increase visibility.

To minimise accidental breakage a metallic thermowell is sometimes used to protect the bulb. This has no effect on the accuracy but may reduce the speed of response. Such thermometers are ‘ruggedized’ for mechanical strength and are sill used in HVAC (heating, ventilation, and air conditioning) applications. In most other industries their use has been discontinued, mostly because of the concern about the toxicity of mercury.

3.1.2 Principles and materials of construction

The principle of this type of thermometer is that the apparent expansion of the liquid in glass provides a significantly measurable quantity, i.e. length, to indicate temperature. The relationship between the bulb volume change and temperature change is represented by the volumetric expansion relationship Vt2 = Vt1 (1 + K(t2 – t1)) where K is the apparent coefficient of expansion. For mercury-in-glass K is 0.00016 or 1/6250 °C-1 so the bulb contains about 6250 times the volume in the stem equivalent to 1°C.

In constructing a thermometer due consideration must be given to the scale length required to represent 1°C, and the total range to be covered on the stem. Then from a knowledge of the bore dimensions a bulb volume commensurate with these requirements can be calculated and made. In practice, capillary diameters range from about 0.02 mm to 0.4 mm for mercury-filled thermometers, but sensitivity cannot, in general, be carried beyond about 100 mm per 1°C because the movement of mercury becomes too irregular in the very fine capillary.

Additionally, a wide spacing between successive graduation lines leads to difficulty in subdivision by eye. Line thicknesses tend to be between 0.05 and 0.2 mm so that with an insensitive thermometer scale there is a danger that graduations become congested and the line thickness is then too great in relation to the spacing. Normally, spaces between lines are about 0.7 mm allowing a line thickness to spacing ratio of 1:5. It is interesting to note that in order to achieve an accuracy or uncertainty of 0.005°C which is quite commonly available with certain types of thermometer, and remembering that the bulb volume is about 6250 times that of a degree unit volume of the scale, then the bulb volume is required to remain stable to better than 1 ppm. This is certainly comparable with (if not better than) the stability of the best platinum resistance thermometers.

The satisfactory behaviour and stability of a liquid-in-glass thermometer is of course dependent on the behaviour of both the glass and the fluid, but more especially on that of the glass. In general, the thermometric fluids remain comparatively stable and benign, but the nature of glass leads to some instabilities which should be borne in mind throughout a thermometer's life. In fact, changes in bulb volume are a normal feature of such thermometers and are quite manageable.

The bulb is normally cylindrical rather than spherical in shape, because the required volume can then be contained within a radius less than that of the stem, and because the surface area is larger with the result that the thermal response is faster. A somewhat greater sensitivity to external pressure variations has, however, to be accepted.

A ‘contraction chamber’ is an enlargement of the bore of those thermometers in which the mercury would otherwise withdraw into the bulb at room temperature, or in which an auxiliary scale is required, e.g. for checking for secular change (see below). The so-called ‘expansion chamber’ is a safety volume at the top of the capillary which serves to prevent the generation of excessive gas pressures when the thermometer is used near the top of its range. These might otherwise stretch, deform or even break the bulb. The chamber is not a reservoir for expanded liquid - if liquid enters it, then the thermometer has been heated above its specified maximum operating temperature, and its calibration may well have been nullified.

A surprisingly wide variety of liquid-in-glass thermometers is available with possibly as many as 5000 different instruments listed in manufacturer's catalogues. In size they range from the familiar clinical medical thermometer, about 100 mm long, to the less-familiar thermometers used in brewers or dyers vats, up to 3 metres long.

Thermometer glasses

The glass used in thermometers is predominantly silica with the addition of other substances to give the desired properties. Overriding consideration is given to the following properties: the softening point which limits the possible working temperature range, the ability to be worked in a blow-lamp without devitrifying or clouding, the ability to be clearly etched or engraved, freedom from physical imperfections, the ability to weld to an opaque enamel backing and, most importantly, the characteristic of thermal stability. Very few glasses satisfy all these requirements.

Glass, in particular the bulb glass, undergoes continual change toward a state of quasi-equilibrium which is peculiar to the temperature at which it is maintained. This drift with time is called the secular change and in general is found to be most troublesome during the first year or so of a thermometer's life. Secular change is well illustrated by work carried out by Joule who kept two thermometers under observation for a period of 40 years. He noted their indications when they were immersed in melting ice at 0°C and observed that as a result of the bulb becoming progressively smaller, the indicated temperatures became higher. Although the rate of change noticeably slowed, at the end of 40 years the change in bulb volume nevertheless continued.

Modern glasses have superior characteristics so that a new thermometer may exhibit a secular change of 0.04°C in the first year after manufacture and thereafter 0.01°C per year. Secular change is most easily monitored by regular measurement of a fixed temperature within the thermometer range, generally 0°C, although it may be masked by the other more significant instability of glass known as ‘the temporary depression of zero’ (see Section 3.1.5). This effect is simply a hysteresis characteristic of glass which fails to return quickly to its original size after cooling from an elevated temperature.

The glasses most commonly used in thermometry are so-called normal glasses, which contain about 73% silica, the remainder being sodium carbonate and zinc oxide. This glass is used almost exclusively as a bulb glass up to 350°C and also as the stem, although a lead glass containing about 21% lead oxide may be used for the stem where the best performance is not necessary. A considerable improvement in stability at temperatures above 300°C can be achieved by using borosilicate glasses for both bulb and stem. Such glasses are 80% silica with 13% borax and alumina hydrate. Borosilicates are nominally useful to 460°C but the addition of more boron permits extension of use up to 600°C. Generally, high-temperature thermometers are made with the same bulb and stem glass because different thermal expansions occurring between, say, boro-silicate and normal glass, prevent successful fusion of the glasses.

Thermometric glasses have been the subject of approval tests by the national standardising laboratories and are recognised internationally by a scheme of coloured lines running the length of the bulb glass, or by an approved abbreviation inscribed on the stem.

Working fluids

As already stated, mercury is most advantageously used as the thermometric fluid for several reasons. It remains liquid over a wide temperature range nominally bounded at the lower end by its freezing point, -38°C, and at the upper end by its boiling point, 356°C. By increasing the pressure of the inert gas-filling, normally nitrogen, in a thermometer tube the boiling-point may be elevated so that temperatures may be measured up to 600°C and even beyond.

Another important property of mercury is that it does not ‘wet’ the glass of a capillary bore and is free of the drainage difficulties encountered with spirit filled thermometers. Mercury is opaque and readily detected in glass. It is also readily obtainable by distillation in a very pure form and the expansion is very regular. Perhaps its only drawback is that its coefficient of expansion is relatively small when comparison is made with many other liquids. This means that fine-bore tubing must be used to obtain a given scale sensitivity without unduly increasing the bulb size.

Below the freezing-point of mercury as far as about –60°C a eutectic mixture of mercury and thallium is used. The main requirements for the amalgam are that the correct mercury-thallium proportions are observed and it must be kept from contact with air at all times during manufacture to prevent oxidation. The optimum proportion of thallium is about 8.7% by weight and some uncertainty about its exact freezing point exists due possibly to the presence of varying amounts of impurities or to a region of partial freezing as the result of departure from the eutectic mixture proportions.

Organic liquids permit the thermometer range to be extended to nearly –200°C. Pure ethyl alcohol has a freezing point of -112°C; toluene a freezing point of –95°C and technical pentane extends the range to the extreme. Such liquids are at first sight unsuitable for accurate thermometry since they all wet the glass capillary as their surface tension is low. They have coefficients of expansion which are 6 to 7 times greater than that of mercury and while this allows greater sensitivity it also provides more opportunities for irregularities since the expansion is greater at some temperatures than at others.

Some deterioration of organic liquids may also occur even under oxygen-free conditions because of polymerisation (chemical union of like molecules one with another) although properly prepared thermometers, using a nitrogen gas atmosphere have had long periods of use and shown no signs of deterioration.

3.1.3 Common defects

Detecting and remedying defects in thermometers is crucial to their successful use. Maintenance is readily carried out using a magnifying lens of about 5x magnification, a coolant such as solid carbon dioxide (sublimation point -79°C) or an ice/salt mixture, and a flexible source of gentle heat such as that provided by a methylated spirit flame. With these, only a little rudimentary skill or practice is required.

Frequently, and especially after transportation, the mercury column is seen to be fragmented or broken. This condition is easily remedied by cooling the bulb so that an individual length of mercury retracts and enters the contraction chamber either as a droplet on the wall, or a membrane spanning the chamber. The thermometer may then be tapped laterally until the droplet falls into the main volume. Each segment of the broken column may be treated in turn in this way until a homogeneous column is re-formed.

Similarly, a large gas bubble trapped in the shoulder of the bulb may be removed by cooling the bulb until the mercury meniscus enters the bulb and the gas bubble may be liberated into the bore. It often helps to grip the thermometer in the palm of the hand and gently strike the heel of the hand on a bench so as to encourage gas movement by vibration.

Great care must be exercised not to freeze (-38°C) the mercury when of course the bulb may fracture. If the surface of the mercury visible through the bulb glass takes on a crazed appearance somewhat similar to the surface of crumpled aluminium foil, disaster may be imminent.

Another difficulty arises when the bulb is seen to contain a number of very small gas bubbles perhaps widely dispersed. Again, cool the meniscus as far as the bulb allowing a large gas bubble to enter the bulb when by inverting and manoeuvring the thermometer the bubble may be allowed to wash around the bulb to collect all the smaller ones. A timely reversion of the thermometer will allow on warming, a uniform gas-free mercury column to rise up the bore. An unspoken rule so far has been, no heating; because of the depression of zero effect any heating of the bulb is undesirable before use, and the temptation to drive mercury up the bore to remedy temporary faults must be avoided. The exception to the rule is met when droplets of mercury are found in the expansion chamber. Do not drive mercury into the chamber but gently wave the chamber in a meths flame so that the mercury vaporises and re- condenses lower down the capillary tube where it may be swept up by driving the mercury column to it. When the thermometer bulb has been heated in a meths flame, the thermometer should not be put into use until the following day (see later, under temporary depression of zero).

A second exception is permissible, even routinely necessary, with spirit filled thermometers. In such thermometers the spirit is highly volatile even at room temperatures so it is usual to find condensate in the expansion chamber. At the same time these liquids have a low surface tension and persistently wet the glass of the capillary. Before use it is necessary to vaporise the spirit in the expansion chamber and continue to flame the capillary bore progressively downwards so that the vapour condenses lower and lower down the scale. This process is aided by simultaneously maintaining the bulb in, say, solid CO2 so that the chamber and bore may be flamed as far as 20°C on the scale, after which the bulb may again be warmed gently to room temperature and the spirit allowed to rise and sweep up the condensate. Spirit filled thermometers should be inspected carefully for these faults before use, stored vertically, and preferably with the area around the expansion volume maintained above ambient temperature by, for example, leaving them close to an electric lamp.

When a spirit thermometer is cooled in use, it should be cooled very slowly to overcome the drainage of spirit problem; a suitable rate is about 1°C per minute but this may vary for individual thermometers.

3.1.4 Calibration of liquid-in-glass thermometers

The ice-point may be realised in a Dewar (vacuum) flask or insulated vessel, ideally one which has provision for easily removing surplus water, e.g. a draincock. It is important to ensure that the ice- water mixture is a melting mixture and to achieve this condition the ice should be shaved or finely- divided so that each particle is in intimate contact with the surrounding water. To this end the ice particles should not exceed a few mm in diameter. Both the water and the ice should be pure or prepared from de-ionised water although in the UK it is likely that normal supplies of domestic cold water would not cause depression of more than a few hundredths of a degree. For the greatest confidence, however, it should be avoided. The water should also be air-saturated, although this has an even smaller effect. When adequately mixed and consolidated, the surface of the mixture takes on a grey rather than a white appearance and any manual compression of the ice at the surface should just show a corresponding swell of water without any tendency to float the ice.

Care must be taken when inserting thermometers into the ice, so that the bulb is not subjected to any excess pressures. It is advisable to prepare (with a tube) a hole in the ice of sufficient depth and minimal diameter to receive the thermometer. Once the thermometer has been lowered into the prepared water-filled hole it should be momentarily lifted and replaced to ease any pressure build-up, and tapped gently before it is read so that any stiction of the mercury meniscus is also minimised. In theory, accuracies of about ±0.001°C may be achieved by this method if the water is well shaken up with air close to 0°C beforehand, but in practice this and other factors lead to accuracies of ±0.005°C being more realistic. When the highest precision is required the thermometer should be maintained at 0°C for at least 10 minutes.

Measurements may be made at 100°C in a steam bath, but at other temperatures thermometers are calibrated by comparison with standards in temperature-controlled baths of water, oil or other suitable liquids. It is sensible always to use two standards so that they are mutually self-checking and they should also detect the occurrence of any hot or cold spots which may develop in the calibration bath. As a general rule thermometers should have the zero or ice-point checked at least annually and the main scale re-calibrated after five years. It is usual to calibrate a thermometer at no fewer than 5 temperatures covering 80% of the graduated scale and generally at every 100 scale divisions. When the highest precision is required calibration is made at every 50 divisions.

3.1.5 Precautions in use

Immersion

There are 3 modes of immersion, all of which assume that the thermometer is used vertically.

Complete immersion requires the entire thermometer to be immersed so that the expansion chamber also attains the temperature of the medium. Since it invariably contains a gas under pressure, the internal forces are amplified, extend the bulb glass and depress the mercury column.

The preferred mode is known as total immersion. In this the bulb and all of the mercury column are immersed, so that in practice only 1 mm or so of the mercury is visible to enable the viewer to read the thermometer. Above about 150°C there is always a risk of some mercury distilling off so it is preferable to obtain the total immersion condition only shortly before observations are made. At other times the thermometer should be withdrawn a few centimetres to reduce the mercury meniscus temperature.

The third and common type of immersion is partial immersion when the thermometer is always immersed to a specified or prescribed depth, usually marked on the stem. As a result, a length of the mercury column protrudes out of the measurement zone. The emergent liquid column (elc) stands in a temperature gradient whose mean temperature is variable, is substantially different from that of the bulb and so affects the reading of the thermometer. The mean temperature must therefore be measured and the indicated temperature corrected to the conditions which applied on calibration or as stated in the thermometer specification. In general, partial immersion thermometers measure temperatures with an uncertainty about twice that of total immersion thermometers.

The elc temperature is commonly measured with a series of auxiliary thermometers placed alongside the column, the first being positioned with the bottom of its bulb 10 mm above the medium surface and others at not more than 100 mm intervals above it, with the last thermometer bulb at the level of the elc meniscus. For use above 100°C a better method is to use Faden thermometers, which have long bulbs and are generally supplied as a set. One is selected from the set with a bulb length as long as the elc to be measured. It is mounted alongside the elc with the bottom of the bulb just in contact with the medium surface to obtain the average temperature. Faden thermometer bulb lengths range from 50 to 300 mm in 10 mm steps.

Corrections to thermometer indications resulting from differences in elc temperatures between calibration and use are deduced from the relationship:-

Correction = KN(t1 - t2)

where K is the apparent coefficient of expansion of mercury in glass, t1 and t2 the temperatures of the elc given respectively by the calibration certificate and that of the auxiliary (or Faden) thermometers at the time of use, and N is the length of the elc expressed as a temperature in terms of the scale graduated on the thermometer. The advantages derived from the use of Faden thermometers to measure elc temperatures are that the bulb responds to the same temperature profile as the thermometer elc along its complete length, and also that it is in direct contact with the medium whose temperature is being measured. Note that the elc correction is in addition to the scale correction given in the calibration certificate.

Column stiction

As a thermometer is heated and the mercury begins to expand, the surface of the meniscus becomes more convex, the capillary pressure increases and is transmitted to the walls of the thermometer bulb. Momentarily the flexible bulb expands until the restoring forces in the glass overcome the excess pressure, the meniscus suddenly rises or jumps, and the meniscus becomes flat again. The cycle continues in a series of uneven jumps equivalent to 0.005 or 0.010°C and is a source of measurement error in thermometers.

The problem is overcome either by gently tapping the thermometer stem during measurements or employing a mechanical tapper driven by a small solenoid electro-magnet attached to the thermometer stem. As a rule measurements should be made under slowly rising temperature conditions as then thermometer behaviour is most reproducible because the mercury meniscus is kept in a convex form and the capillary forces are kept reasonably constant.

Pressure effects

Since the wall of the bulb glass is intentionally thin (about 0.4 mm) it readily flexes in response to changes in pressure whether they be external or internal in origin, so that thermometer readings may vary with everyday atmospheric pressure changes, with altitude, or by changing the angle of use from vertical to horizontal. Typically, a change in the external barometric pressure of 30 mbar will change a thermometer's reading by about 0.005°C, roughly corresponding to a coefficient of 0.15°C/bar. The external pressure coefficient is dependent on the bulb's internal and external radii and a constant K, representing both the elasticity of the glass and a conversion factor for the volume change in terms of a change on the thermometer scale in degrees. Then the coefficient is given by:-

2 2 2 Cext = K Rext /(Rext – Rint )

For any thermometer the correction can be determined by supporting it in a closed transparent tube (maintained at a constant temperature) connected to a variable pressure device, e.g. bellows and manometer. A series of measurements over a range of temperatures and pressures leads to a determination of the coefficient.

A thermometer used horizontally will have a higher reading than when used vertically at the same temperature because of the changed internal pressure governed by the head of mercury in the bore, the capillary pressure of the meniscus and any gas filling. When the mercury is at the upper end of its scale the correction is typically about 0.1°C and is proportional to the distance from the centre of the bulb at intermediate temperatures.

Parallax errors

Errors in reading thermometers can be introduced by parallax unless due care is taken to view the meniscus normally to the stem. Such errors are dependent on the angle between the correct and wrong lines of sight, and on the thickness of the glass between the meniscus and the graduation lines, so that in the case of a 6 mm diameter thermometer graduated at 0.2°C intervals, an error of 0.05°C would result if it were viewed about 4 degrees from the normal.

Temporary depression of zero and secular drift

Whenever a thermometer undergoes a temperature excursion the bulb glass is subject to a thermal hysteresis effect which is generally observed as a change in the indicated temperature at a reference temperature, commonly at 0°C. If a normal glass thermometer is used to measure successive temperatures of 0, 100, and 0°C in a very short period, the second ice-point reading will be lower than the first by about 0.05°C, i.e. it is depressed. This effect is known as the temporary depression of zero. Having expanded after a temperature rise the glass does not immediately recover its original volume when it is again cooled to 0°C. Recovery begins immediately, but is not complete for some hours or in some cases, several days. As a general rule the depression value per 100°C of rise is about 0.05°C and 0.02°C for bulbs made of normal and borosilicate glasses, respectively.

The nuisance of this effect is the limitation it imposes in a thermometer's daily use, since once it has been used at a high temperature it cannot be used at a lower temperature until the bulb has been allowed to recover. This does not apply, of course, when the anticipated depression is less than the required uncertainty of test; also in any one day the thermometer may be used to measure any number of progressively higher temperatures. Practically, there are two ways of dealing with the problem. The first involves taking an ice-point reading immediately following each measurement of a temperature so that the ice-point reading has a bulb condition or volume corresponding to the state of the thermometer when the higher temperature was read. This approach may be most appropriate when a thermometer is intended as a primary standard, but the second and normal practice is to measure an ice-point only after the thermometer has been at room temperature for a considerable time. In this way, whatever the temperature measured beforehand, the zero value is measured after the thermometer has rested and regained a stable and reproducible condition.

The acceptance of this second method also allows one to separate secular changes (long-term drift of bulb volume) from scale corrections. During calibration the thermometer's readings are compared with those of standard instruments at a series of temperatures. These establish the corrections which must be added to the reading to obtain the correct temperature. If the correction for the thermometer at 0°C is subtracted from each of the calibration corrections, the resultant revised values apply as if the thermometer read exactly 0 at 0°C. These values can be regarded as a series of semi-permanent scale corrections, to which the ice-point correction should be added. Subsequently, at such times as the ice-point correction is re-determined, the new value is adopted. It is recommended that ice-point corrections should be re-determined at least annually, with a complete scale calibration after not more than 5 years.

Both the secular change and the temporary depression of zero are effectively absent from thermometers made from fused quartz. The secular change in such thermometers used up to 100°C may not exceed 0.001°C after several years use and even after prolonged use at 600°C changes may be about only 0.01°C. However, early manufacturing difficulties and the high cost of quartz thermometers has discouraged their use in the UK. A European manufacturer advertises quartz thermometers for use up to 1000°C with a gallium filling. This has a boiling-point in excess of 2000°C and obviates the need for high-pressure gas filling. As expected, ice-point stability of a quartz thermometer is claimed to be excellent, thermal shock resistance is extremely high, mechanical

strength is very good, and the low coefficient of expansion of gallium increases the confidence with which corrections for emergent liquid columns can be measured.

ADVANTAGES AND LIMITATIONS OF LIQUID-IN-GLASS THERMOMETERS

ADVANTAGES:

Low cost, simplicity, and long life if treated properly. Good performance if used correctly.

DISADVANTAGES:

Difficult in reading, confinement to local measurement, and non-adaptability to recording or automatic control. They also break very easily.

Table 3.1 Advantages and limitations of liquid-in-glass thermometers

3.2 DIAL-TYPE EXPANSION THERMOMETERS

Figure 3.2 Dial-type expansion thermometer

A dial-type expansion thermometer (filled thermal system) is basically a pressure gauge connected by small bore tubing to a bulb acting as the temperature sensor. The whole system is gas-tight, and filled with an appropriate confined gas or liquid under pressure.

With this type of thermometer, the pressure inside a closed thermal system is used to measure temperature. The reading of the pressure gauge is affected not only by the process temperature at the bulb but also by other factors that can introduce errors. These include: 1) ambient temperature variation, 2) barometric pressure variation, and 3) the hydrostatic head of the liquid in the capillary, if it is at a different elevation from the readout. In order to obtain accurate temperature readings, these error sources must be minimised, as will be explained in this section.

There are many different types of dial-type thermometer, each having certain peculiarities which give it advantages over others. One way to classify these instruments is according to filling material. As already mentioned, the use of mercury filling has been discontinued in most industrial applications because of health concerns. The use of both mercury and liquid fillings have also lost ground because of the expense associated with compensating for the ambient effects on the capillary and because of the errors caused by elevational differences between bulb and readout. Still extensively used are the gas and vapour fillings, but these too are limited in their usefulness. The gas-filled bulb is large, as is the temperature span required to operate it. The vapour-filled system is also limited due to its non- linearity and its potential for problems caused by cross-ambient operation or by errors due to elevation.

3.2.1 Bulbs, wells, and capillaries

The temperature-sensitive element, the bulb, comes in many sizes and shapes to handle the many different applications. It is good practice to use the largest bulb which will do the job. This will cut down on ambient temperature errors, and permit smaller spans and longer capillaries.

Plain bulbs are used where the measured medium is not under pressure and will not harm the bulb material. If this is not the case then a separate well to protect the bulb from the process medium is needed. This, of course, will slow down the response time.

The speed of response generally doubles with the doubling of bulb diameter and tends to be the fastest with vapour or gas and the slowest with liquid filling. When thermowells are added to the bulb, speed of response typically slows as follows: 12 to 24 seconds for 0.25 in. (6 mm), 20 to 35 seconds for ⅜ in. (9 mm), 25 to 50 seconds for 0.5 in. (13 mm), and 40 to 75 seconds for 0.75 in. (19 mm) diameters.

Higher speed of response can be obtained with a long, thin, bendable bulb used to sense the average temperature in large areas. Another way of presenting a long bulb is by coiling it. For use in low- velocity gas flow temperature measurement the coil is set at the factory and cannot be uncoiled.

Most bulbs are made of stainless steel, which is relatively inert and will withstand high temperatures. Other materials are readily available.

The relatively fragile thin-wall capillary should be protected by a flexible armoured stainless steel or PVC-covered bronze tubing. An extension neck to the bulb prevents the tubing from being immersed directly in the measured medium. Bendable smooth steel tubing might also be used for capillary protection.

3.2.2 Liquid-Filled Systems

These systems are completely filled with a liquid (other than mercury) and operate on the principle of liquid expansion with increase of temperature. The filling fluid is usually an inert hydrocarbon, such as xylene (C8H10), which has a coefficient of expansion six times that of mercury and makes smaller bulbs possible. Other liquids (even water) are sometimes used. The criterion is that the pressure inside the system must be greater than the vapour pressure of the liquid to prevent bubbles of vapour from forming in the spiral. Also, the liquid should not be allowed to solidify even in storage; otherwise, the calibration may be affected.

The minimum operating temperature is usually set by the freezing point of the filling liquid, which is usually between -100 and –350°F (-75 and –210°C). The maximum operating temperature is set by the point at which the filling liquid is no longer stable, usually around 600°F (315°C). The minimum span range is a function of bulb size; the maximum, of linearity. With larger bulbs the span can be narrowed to 22 to 45°F (12 to 25°C), while the maximum limit on the span is about 300°F (167°C) due to the non-linearities caused by expansivity and compressivity of the filling liquids.

The maximum temperature which the bulb can be exposed to without damage is defined as the allowable over-range of the system. Over-range is usually expressed as a percentage of the span over the full range value; in the case of liquid filling, over-range is around 100%.

3.2.3 Vapour Systems

The pressure element, capillary, and bulb of a vapour system have the filling medium in both the liquid and vapour form. The interface between the two must occur in the bulb, and this will move slightly with temperature, affecting the pressure.

The pressure within the system is a function of the vapour pressure of the filling fluid at the operating (bulb) temperature. The filling fluid is usually so selected as to give a 100 PSIG (6.9 bar) change in its vapour pressure as the bulb temperature goes through the span of the thermometer. This much change is necessary to keep the error due to barometric pressure changes to less than 0.5% of full scale. The filling fluids used include methyl chloride, sulphur dioxide, butane, propane, hexane, methyl ether, ethyl chloride, ethyl ether, ethyl alcohol, and chlorobenzene. Each has a different vapour-pressure-temperature relationship. For lower temperatures ethane filling can be used; it changes its vapour pressure from 20 to 600 PSIG (1.4 to 41 bar) as the temperature rises from -73 to 29°C. For higher temperatures one might use ethyl chloride, which goes through the same vapour pressure change as its temperature rises from 38 to 177°C.

Overall, the minimum temperature at which vapour fillings can be used is around -40°C, while the maximum is about 315°C. The maximum temperature is limited by the critical point of the fill, while the

minimum limit is a consequence of the loss of reading sensitivity, as the vapour pressure changes less per unit temperature change at low temperatures. The non-linear nature of the vapour-filled thermal systems is generally considered to be a disadvantage, although if a better reading sensitivity toward the top of the scale is required, it can be a desirable feature.

The speed of response of vapour-filled systems is generally in the range of 1 to 10 seconds. It is faster than the liquid or mercury fills and as fast as the gas-filled systems in most of its configurations.

The over-range limit of the vapour-filled systems is small, because the vapour pressure tends to rise exponentially with temperature. Under some very unique sets of conditions it is possible to obtain higher over-range by designing the thermal system so that all liquid will be exhausted at a temperature slightly over the maximum required reading.

3.2.4 Gas-Filled Systems

The operating principle for gas-filled systems is that in a perfect gas confined to a constant volume the pressure is proportional to the absolute temperature. The gas is not perfect and not all at the same temperature nor is the volume constant, but variances are small enough so that a measurement of pressure can be used to indicate temperature.

Nitrogen is the favourite fill for this class of thermometer because it is inert and inexpensive. It does react somewhat with the steel bulb material at temperatures exceeding 800°F (427°C), and it does act less like a perfect gas at extremely low temperatures. Under these conditions helium should be used. Different ranges are obtained by selecting the correct filling pressure.

In general, bulbs should be as large as practical to lessen the influence of temperature variations along the capillary. One way of avoiding long capillaries is to terminate a short capillary at a small diaphragm chamber. The force due to gas pressure on the diaphragm causes it to compress the spring. This motion is amplified and used to regulate another pressure which is transmitted to the spiral. This arrangement, though more expensive, permits much smaller bulbs than could otherwise be used.

Gas-filled systems approximate Charles's law (absolute pressure of a confined gas is proportional to its absolute temperature) by keeping the bulb volume relatively large compared to the rest of the system. Such thermometers are primarily used for low and high temperatures. On the low side they are limited by the critical temperature of the filling gas (usually nitrogen or helium) corresponding to - 268°C, and on the high side by the temperature limits on the bulb materials, usually corresponding to 760°C. The maximum span can be 1200°F (667°C) and is limited only by non-linearities due to mass flow from the bulb. The minimum span is limited by the pressure at which the Bourdon tube becomes over stressed (usually designed for 400 PSIG, or 28 bar). Therefore, the minimum span with conventional systems is about 400°F (222°C), while with helically wound thin-tube bulbs it can be reduced to 120°F (67°C), or even less.

The speed of response of gas-filled systems is usually good. The time constant is typically 1 to 4 seconds because the ratio between bulb mass and surface area tends to be favourable. These thermometers can usually provide 150 to 300% over-range protection as the maximum temperature is limited only by the permissible pressure and temperature ratings of the bulb. The gas-filled bulb is ideal for measuring the average temperature over large areas, such as in dryers or ovens. This is the case because the bulb volume is not critical and capillary-type bulbs can cover large areas.

3.2.5 Mercury-Filled Systems

Mercury is a liquid and in this respect mercury-filled systems are similar to liquid-filled systems. The two are considered separately because of the unique characteristics of mercury and its importance as a temperature-measuring medium. Mercury provides rapid response, accuracy, and plenty of power for operating control elements. Pressures within the working system are relatively high - as much as 1200 PSIG (83 bar) for the higher temperatures, dropping to 400 PSIG (28 bar) at the low- temperature end of the range. This high pressure cuts down on any head effect error (difference in elevation between bulb and measuring instrument).

Mercury-filled systems can detect temperatures between the freezing and the boiling points of mercury, or from -40 to 649°C. Mercury fillings also provide the widest range of spans, from 28 to 667°C. The speed of response of mercury-filled systems is faster than that of the liquid-filled ones but slower than that of gas or vapour systems. These systems can also have an over-range protection of at least 100%.

3.2.6 Ambient temperature compensation

The measured variable in a filled thermometer is the total internal pressure. This pressure is the result of two factors - the temperature around the bulb and the ambient temperature around the rest of the system. The purpose of the installation is to measure the process temperature around the bulb, and therefore it is desirable to eliminate the effects of ambient temperature variations on the total reading. The error introduced by ambient temperature variations is different for the different types of fills, and it also increases as the bulb or span gets smaller or as the capillary length is increased.

When the ambient temperature effects are compensated for in both the capillary and in the readout instrument, the design is called fully compensated. When the case and the capillary is at the same ambient temperature and the length of the capillary is relatively short, it might be sufficient to leave the capillary portion uncompensated and provide case compensation only.

Vapour-filled systems do not need any compensation because the liquid/vapour interface is always within the bulb. Although in some such systems, it is important to keep all sections of the capillary at a temperature below that of the bulb; otherwise the fill fluid could vaporise in the capillary and introduce substantial errors.

In some systems full compensation can be achieved by duplicating the spiral pressure sensor and the capillary. The compensating capillary is closed, and its length is the same as the active capillary without the bulb. The two capillaries are run parallel to each other and the two Bourdon spirals are connected in such a way as to cancel the ambient effects.

In the case of mercury-filled systems, full compensation can be provided without the expense of duplicating most of the thermal system. Here, an Invar wire can be inserted inside the single active capillary. Invar is a nickel alloy which has a low coefficient of thermal expansion. Therefore it is possible to select the diameters of the Invar wire and of the capillary in such a manner that an increase in ambient temperature will increase the annular space between the wire and the capillary by just about the same amount as the volumetric expansion of mercury. Therefore, a change in ambient temperature around the capillary does not change the pressure inside the thermal system.

When both the capillary and the case are at the same ambient temperature and the capillary volume is very small relative to the total volume of the system, it is possible to leave the capillary uncompensated and limit the compensation to the case. Case compensation can be achieved by the addition of a bimetallic spring which generates a force that is nearly equal to the one caused by the change in ambient temperature. Because the volume of the Bourdon spiral varies as the internal pressure changes with the measured temperature, the case compensation provided by the bimetallic spring is effective at only one bulb temperature. The bimetallic spring is usually set at midrange, resulting in under compensation at the top and over-compensation in the bottom half of the range.

Uncompensated operation is possible with liquid-filled systems when the capillary is very short, and with all types of vapour-filled systems. Gas-filled and mercury-filled units should always be compensated. In gas-filled systems, the capillary error rises rapidly with bulb temperature and can be reduced by increasing bulb volume and/or span.

3.2.7 Effects of bulb elevation

In a filled system a pressure gauge is used to measure temperature. If the thermal bulb is above the readout instrument, the hydrostatic head of the filling fluid will be added to the total pressure at the spiral Bourdon element. If the bulb is below the case, the hydrostatic head will be subtracted from the total pressure. Therefore, when the bulb is high the thermometer will have a positive error and when it

is low, it will have a negative error. The relative size of the error is a function of the size of the hydrostatic head relative to the total pressure.

In the case of liquid- or mercury-filled systems, the hydrostatic head caused by the elevational difference between the case and the bulb is a constant and therefore can be zeroed out. Depending on whether the bulb is above or below the case, the hydrostatic head must be subtracted or added at the readout Bourdon tube. This is done with suppression or elevation springs, respectively. The amount of elevation or suppression that is available is not unlimited. Due to the high density of mercury in the mercury-filled systems, the hydrostatic head due to bulb elevation is much greater than with other fillings. Therefore, with mercury-filled systems the maximum elevational difference that can be tolerated is about 9 m. Needless to say, if the readout of the bulb is moved, the system must be rezeroed.

For most vapour-filled systems, correction of the elevational error is similar to correction for liquid and mercury-filled systems, using elevation or suppression to compensate for the hydrostatic heads and thereby to rezero the system. In some vapour-filled systems, however the height of the liquid column in the capillary varies with temperature and therefore cannot be zeroed out as a fixed quantity. Consequently, such systems should not be used when the case and bulb are at different elevations.

In the case of gas-filled systems, the density of the gas (and therefore its hydrostatic head) is too small to cause any significant errors. Therefore no correction is needed to protect against bulb elevational errors.

3.2.8 Barometric errors

The Bourdon spiral is a differential pressure sensor. Its deflection is a function of the difference between the pressure of the filling fluid on the inside and the atmospheric pressure on the outside. Because the atmospheric pressure is not a constant, its changes can also result in readout errors. The barometric pressure can change by about 0.5 PSIG (25 mmHg). This may or may not be large enough to introduce an error, depending on the filling pressure. The pressure in the readout spiral varies with the type of filling fluid. In liquid-filled and mercury-filled systems it is high enough to make the barometric effect insignificant.

Vapour-filled systems are generally designed for a minimum of 100 PSIG (6.9 bar) pressure change across their span, and therefore the barometric effect can result in a 0.5% error. Gas-filled systems are usually designed to produce a 400 PSIG (28 bar) pressure change across their ranges. Therefore the barometric effect can result in an error of about 0.1%.

ADVANTAGES AND LIMITATIONS OF DIAL-TYPE EXPANSION THERMOMETERS

The dial-type expansion thermometers are superior to bimetallic elements but are generally inferior to electronic thermometers. Their advantages include rugged, simple, self-contained, inexpensive, easily maintained construction and acceptable sensitivity and accuracy without need for auxiliary power sources. They are inherently explosion-proof and can operate displays or recorders at some distance from the measurement using only hand-wound spring drives.

Their disadvantages include the bulky bulbs, which are space-consuming and result in a relatively slow speed of response (about 20+ seconds with well), inability to provide narrow spans, limitation on the distance between bulb and readout (usually under 60 m), and relatively difficult maintenance of the pressurised thermal elements, which is usually done at the factory. Dial-type expansion thermometers are not suited for high-temperature measurement and can be easily misapplied if the user is not familiar with the unique characteristics of each system.

Table 3.2 Advantages and limitations of dial-type expansion thermometers

3.3 RESISTANCE TEMPERATURE DETECTORS (RTDs)

Figure 3.3 Resistance temperature detector (RTD)

3.3.1 Introduction and Overview

In the modern world, mercury and spirit-filled thermometers have largely given way to electronic devices, the output signal of which can be digitised and remotely analysed. Resistance thermometers are electrical thermometers which make use of the variation of resistance (or resistivity) of a metal wire with temperature (note: this effect can also be found in semi-conductors; thermistors are discussed separately in Section 3.4). As electrons move through a metal, they are impeded by the thermal vibrations of the atoms in the crystal lattice. The higher the temperature the greater the impedance and the higher the resistivity. This effect is very marked in pure metals, and the variation is predictable, enabling accurate measurements to be performed. They are sensitive and, with sophisticated equipment, measurements, can routinely be made to better than 0.001°C.

Resistivity in metals, alloys and semiconductors arises because the electrons which are free to respond to an applied voltage and hence to provide electrical conduction, are scattered by lattice vibrations, impurity atoms, lattice defects, and also by other electrons. Scattering by lattice imperfections and impurities dominate the resistivities of alloys such as nichrome used in 'resistance wire' for heater elements, and is largely temperature-independent. However, in a pure and well- annealed metal, scattering by these means is small, and the resistivity is then mainly due to the vibrations in the lattice. This is strongly temperature-dependent, and is the property on which most resistance thermometers rely.

Usually platinum wire is used in the construction of the thermometer, since it is a noble metal which is unreactive over a wide range of temperatures. But copper, nickel and rhodium alloy may also be used in various temperature ranges. Usually a coil of the pure wire is wound onto an alumina former or placed in the bores of an alumina tube, and this assembly is mounted in a steel tube.

What is a platinum resistance thermometer?

A platinum resistance thermometer (PRT) is a device which determines the temperature by measuring the electrical resistance of a piece of pure platinum wire. The piece of platinum wire is referred to as a temperature sensor. When manufactured carefully, these devices offer an excellent combination of sensitivity, range and reproducibility.

Platinum resistance thermometers were first applied for temperature measurement by Sir William Siemens more than 100 years ago. Siemens' thermometers suffered from instability due to contamination, and it was only after the designs had been improved by Callendar and others that it was possible to achieve the reliability needed to use platinum resistance thermometers in temperature standards and industrially. Standard platinum resistance thermometers (SPRTs) have been specified as the interpolating devices in all the International Temperature Scales. In the first of these, the ITS- 27, the range extended from -200 to 630°C, but this was widened in the IPTS-68 to -259.34°C (the

triple point of hydrogen), and in the ITS-90 the upper limit was raised to 961.78°C (the freezing point of silver).

How do PRTs work?

The electrical resistance of many metals (e.g. copper, silver, aluminium, platinum) increases approximately linearly with absolute temperature and this feature makes them useful as temperature sensors. The resistance of a wire of the material is measured by passing a current (AC or DC) through it and measuring the voltage with a suitable bridge or voltmeter, and the reading is converted to temperature using a calibration equation.

The most reproducible type of sensor is made from platinum because it is a stable unreactive metal which can be drawn down to fine wires but is not too soft. Using very pure wires, thermometers can be made with closely similar resistance characteristics and achieve good reproducibility in use.

The resistivity to temperature conversion can be easily handled by a digitiser, but it is described in the industrial standard IEL 751, where the temperature/resistance relationship above 0°C is approximated by:

W(t) = R(t)/R(0) = 1 + At + Bt2

where R(t) and R(0) are the resistances at temperatures t and 0°C respectively, and A and B are constants. At lower temperatures, more complex functions must be used.

Figure 3.4 The resistivity of five metallic elements plotted on a linear scale as a function of temperature. The data for platinum is the result of many closely-spaced measurements and is plotted as a continuous curve. The data for Al, Cu, Ag and Au consists of just a few points with lines drawn to connect the data points.

The length and diameter of the platinum wire used in a thermometer are often chosen so that the resistance of the device at around 0ºC is 100 ohms. Such a sensor is a called a Pt100 sensor, and its resistance changes by approximately 0.4 ohms per degree Celsius. Using a typical 1 mA measuring current, at around 0ºC a PT100 sensor would have a voltage drop of around 100 mV across its terminals and this would change by approximately 0.4 mV per degree Celsius, which thus makes sensitive thermometry available to anyone with a high resolution voltmeter or resistance bridge. In many instruments the measurement is converted so that the reading is directly in temperature.

Platinum sensors are manufactured to comply with a standard curve within a specified tolerance. The international standard, IEC 751, is published in the UK as BS EN 60751. The Class A tolerance for a Pt100 sensor is ±0.25ºC over the range 0ºC to ±200ºC. However, better uncertainties can be achieved if the sensor is suitably protected in a steel probe and calibrated individually.

For the highest accuracy, special glass-sheathed standard PRTs, usually of 25 ohms at 0°C, are calibrated at the fixed points of the International temperature scale 1990 (figure 3.4). The ITS-90 specifies equations to relate the resistance to temperature and, using these, uncertainties can be achieved of 0.001°C or better. Standard PRTs can be used from temperatures as low as -259°C up to 660ºC, or even, 962ºC, with some increase in uncertainty and loss of reproducibility.

No single type of thermometer can cover the complete range, and in practice small capsule-type thermometers are used at the lowest temperatures, 'long-stem' thermometers of 25 ohm at 0°C are used from about -200 to 660°C, and 0.25 to 2.5 ohm thermometers are used up to the silver point. The various designs, and those of industrial thermometers (usually 100 ohm and known as RTDs or Pt100s), are discussed later.

3.3.2 Materials and construction

The conductors used for resistance thermometry include platinum, nickel of various purities, 70% nickel/30% iron (Balco™), and copper, listed in order of decreasing temperature range capability. No other candidates are generally available in commercial thermometers. These conductors are all available as fine wire for sensor winding. Platinum is also available as a deposited film sensor, and nickel and Balco are available in foil-type sensors. All share in varying degrees the characteristics of repeatability, high temperature coefficient, long-term stability, and linearity over a useful temperature range. While no sensor material surpasses platinum in overall performance, each has at least one characteristic that may encourage its selection.

Platinum is almost invariably used as the sensing resistor in resistance thermometry because it is a noble metal (which means that it is relatively unreactive over a wide range of temperatures), it is easily workable, it is a very reproducible metal from the point of view of its refinement and manufacture and because the relationship between its resistance and temperature has been extensively studied over the years. However, it is contaminated by a number of materials, particularly when heated, so it is important to use carefully chosen materials for the coil support and sheath, and to clean all parts of a thermometer before assembly. Below 250°C contamination is not usually a problem, but at higher temperatures, materials used in the construction, or accidentally introduced, are likely to react with, or dissolve in, the platinum. Examples include base metals (particularly iron) some kinds of mica, borosilicate glass and hydrocarbon residues from improper cleaning or handling, all of which have been found to cause trouble above 500°C. Thermometers which are hermetically sealed require the presence of some oxygen in the filling gas in order to keep base metals, trace impurities in the platinum, hydrocarbon residues and less stable oxides in an oxidised state so that contamination of the platinum is minimised.

Copper is another metal which is used in the construction of resistance thermometers because of its low cost and its highly linear resistance/temperature relationship. Its temperature coefficient is slightly higher than that of platinum but it has a lower resistivity, which is a disadvantage in many industrial applications, where a high resistance element is often desired. Also, it has a poor resistance to oxidation above moderate temperatures and is generally less stable and reproducible than platinum. Copper thermometers are often used where it is desired to measure average temperatures (e.g. in large vessels) using long sensors in which convenient resistance values can easily be achieved.

Nickel is sometimes used in industrial resistance thermometers over a limited range of temperatures, mainly because of its lower cost and its high temperature coefficient. However, its resistance/temperature relationship is far from linear and it undergoes a metallurgical phase transition around 320°C, which affects the resistance and thereby limits its useful range. A recent addition to the industrial resistance thermometer market involves the use of molybdenum film resistors. The manufacturers claim that they show good linearity and stability over the range -50 to 200°C.

The basic construction of a resistance thermometer has changed only a little over the years; Callendar's recognition of the need for a resistance element of very high purity platinum supported so

that the wire is subjected to minimum strain, still holds good for the most precise instruments. Callendar’s original design consisted of a mica cross around which he wound the platinum coil - this has drawbacks when the thermometer is used at temperatures around 500 to 600°C, because the mica is dehydrated and then, in time, becomes very weak mechanically. If the thermometer is gas- filled and sealed, then moisture from the mica tends to collect in the head of the thermometer and can cause problems when it is next used at a lower temperature (say, the ice point). On the other hand, if the thermometer is heated and pumped sufficiently to remove most of this moisture before it is gas- filled and sealed then the mica is liable to become too weak and brittle to support the element.

Other designs which followed Callendar's early mica ones included the use of a porcelain cross - this was heavy and had a considerably increased lag constant; quartz was tried but was hard to make in cross form and tended, therefore, to be expensive. Another design consists of a flat strip of silica which was twisted to form a helical support for a platinum coil, anchored at the bottom and top. And another consists of a machined ceramic former which was made from a ceramic rod with a groove cut into it around which the coil was fitted. The National Physical Laboratory designed some years ago a device consisting of a thin-walled glass tube (o.d. about 2 mm) into which the platinum coil was placed. The tube was then bent into a U-shape, 4 platinum leads being sealed into the top of the U- tube (2 in each limb), and the whole assembly suspended in a hard glass or silica outer sheath. This formed a particularly stable thermometer, which is still the basis of the current design which is made commercially in this country. A present day modification of this design is that the U-tube has become two separate tubes each containing a platinum coil, connected together at the bottom by a thick platinum wire sealed into the glass and welded to the two platinum coils. To improve the amount of support for the coils the glass tubes are twisted together.

In all cases, for the most precise use in the range above –189°C, the resistance element is carefully cleaned then mounted in a long glass or silica tube with the 4 platinum leads passing through a glass seal at the top. After assembly the thermometer is heated and pumped out for many hours before being finally filled with dry air or a mixture of high purity argon with a few percent of oxygen.

For low temperature use (down to 13.80 K) the thermometer is normally of the capsule type. This typically consists of a thin-walled platinum tube containing a resistance coil wound on a former of one of the types already described. The platinum tube, closed at the bottom end, is about 5 cm long by 5 mm o.d. with a glass head through which the 4 platinum leads are sealed. The glass head also provides a tube through which the assembled thermometer can be evacuated and then filled with helium before being permanently sealed off. The helium is necessary because air would freeze out at low temperatures leaving the platinum resistor in very poor thermal contact with its surroundings.

In all the laboratory thermometers so far considered the resistance at 0°C is approximately 25 ohms; this gives a sensitivity equivalent to about 0.1 ohm/°C. For high temperature resistance thermometers, insulation leakage becomes a problem at elevated temperatures, causing a shunting effect across the resistor. To minimise the effect, the element resistance is usually reduced to between 0.2 ohm and 5 ohm at 0°C depending upon the design of the thermometer and the materials used to support the wire. Suppose, for example, that there is a 5 megohm dc insulation resistance at 962°C. With a 25 ohm sensor this would cause an error of 25 mK; in a 0.25 ohm sensor the error would be only 0.25 mK.

Several designs of high temperature platinum resistance thermometers have been extensively studied. One of the early ones originated at NBS and is known as the ‘bird-cage thermometer’. This design has a resistance element consisting of 8 straight parallel lengths of 0.4 mm diameter platinum wire threaded through holes close to the edges of 4 silica discs, each 0.5 mm thick and 5 mm in diameter. The ends of the wires are bent over and then welded together in series, the 4 discs supporting them being spaced by means of short lengths of thin-walled silica tubing threaded on another piece of 0.4 mm wire, which is connected as a fifth lead for insulation checking purposes. The thermometer has a nominal resistance of 0.2 ohm at 0°C, and 1 ohm at 1000°C.

A good compromise has been achieved by mounting the helical resistance coil in the bores of a multi- bore alumina tube. The coil is anchored down by a small amount of glaze so that whilst the greater part of it is free, a small portion of each turn is attached. Another method involves embedding the platinum coil in alumina powder to reduce the effects of vibration. By careful use of these techniques, it is possible to construct thermometers with R(0°C) stabilities, of a few hundredths of a degree when

used over the range -200 to 660°C or even 800°C.

Another technique which has been introduced during recent years consists in depositing a film of platinum onto a suitable substrate. Some such thermometers are of thin film construction (whereby the film is evaporated onto the substrate), others are thick film devices in which the platinum is printed onto the substrate by a silk screen process. These thermometers have a performance almost equal to that of wire-wound glass-coated detectors over the range from about -50 to 500°C; they are highly insensitive to vibration and can be produced at very modest cost. They are particularly suited to applications such as surface temperature measurement and air temperature monitoring, where their fast time response is an advantage. This is due to the intimate contact of the film with the substrate and the surroundings, and the lower mass of metal that needs to be heated.

Construction of Industrial RTDs

Industrial resistance thermometers differ from the laboratory pattern in that they have to be constructed to withstand the vibration and shock associated with an industrial environment. For high stability, the platinum wire should be mounted in such a way that it is free to expand and conduct without strain. For maximum vibration resistance in an industrial thermometer the wire should be fully supported - so some compromise is necessary. Where vibration levels are such that it is essential to firmly attach the platinum to the former, the wire is usually wound upon a glass or ceramic rod which is then coated with glass. The glass is selected in an attempt to match the expansion properties of the platinum, but although the thermometer is extremely robust it has somewhat poorer stability and greater hysteresis than with a partially supported coil. There is also a smaller temperature range over which it can be used.

All resistance thermometers require the following considerations in their manufacture. Wire-wound sensors must be supported on mandrels closely matching the wire in thermal expansion to minimise strain effects. Additional assembly materials, such as cements, should not introduce additional strain in the operating temperature range. The final assembly must be in a stable, annealed condition, trimmed to the required resistance tolerance. Only high-purity materials and clean assembly methods should be used to avoid sources of contamination that might degrade the sensor. All internal connections should be welded, and connecting leads should be chosen for the required temperature capability and avoidance of thermoelectric junctions. To realise the ruggedness of fully supported elements in the total sensor assembly, all internal connections should be anchored and isolated from effects of thermal and mechanical strains, including shock and vibration. The same requirements apply when deposited film or foil-type resistance elements are used.

For equivalent performance in their respective temperature ranges, base metal RTDs cost the same as 100 ohm platinum RTDs. Construction requirements and materials cost are similar. Base metal RTDs have a materials cost advantage at higher resistance values compared to wire-wound platinum sensing elements. Thin film platinum elements erase this advantage.

Platinum RTDs

In the case of platinum RTDs, the fully supported rugged construction using "reference grade" wire has a temperature coefficient (alpha) over the interval 32 to 212°F (0 to 100°C) of between 0.00387 and 0.003915, depending on the manufacturer. Compared to 0.003927 on a SPRT, the reduction in sensitivity is insignificant. For best accuracy, the user should be aware of or specify the actual temperature coefficient. One common value available from most manufacturers is 0.003902. This is the result obtained for windings on pure alumina mandrels.

Standard platinum industrial RTD curves based on slightly doped platinum wire have been adopted by several countries. These curves are all substantially identical with an alpha of 0.00385. This result is reproducible by manufacturers everywhere, and the so-called international grade platinum curve is the most widely used curve, even in the United States. An exception is temperatures below -196°C, where only reference grade wire RTDs are well characterised.

Platinum RTDs using thick or thin films are also available to the same curve as international grade wire-wound sensors. Performance is often equivalent to wire-wound sensors except maximum

temperature may be reduced. Wire-wound platinum RTDs are most common at 100 ohms ice point resistance, with 200 and 500 ohms available at additional cost. Using thin films, ice point resistances of 100 and 1000 ohms are available at the same cost with slightly lower alpha specified at 1000 ohms.

Base-Metal RTDs

Second in usage to platinum is high-purity nickel, which offers the highest temperature coefficient, second-highest temperature range, and lower assembled cost than wire-wound platinum at high resistance values. The most common are 120 and 500 ohm resistances, with 1000 ohms available. Nickel is non-linear, with an increasing temperature coefficient at increasing temperature. Nickel is also highly strain-sensitive and requires great care by the manufacturer to obtain good interchangeability. The temperature coefficient of nickel is highly influenced by both purity and state of anneal. In addition, lower purity nickel such as 99% nickel (no longer available) or ballast nickel has been used and provides a somewhat lower temperature coefficient. The maximum temperature of nickel sensors should not be much more than 260°C. There is no internationally recognised standard curve for nickel sensors, although there are national standards, and several manufacturers can provide sensors to a common curve characterised by an alpha temperature coefficient between 0 and 100°C of 0.00672.

Third in usage is the 70% nickel/30% iron alloy tradenamed Balco. The sole basis for its use is a very high specific resistance which makes possible very high resistance windings without increasing size. Ice point resistances are commonly 2000 or 10,000 ohms. It has the second highest temperature coefficient, third-highest temperature capability, and, like pure nickel, is non-linear with an upward bending R versus T curve. There is no recognised standard curve for Balco sensors.

Last in usage is pure copper, which is generally available only at 10 or 100 ohm ice point resistance values due to low specific resistance of the winding wire. Copper's temperature coefficient is almost the same as platinum and it is very linear above the ice point. Copper in ‘bifilar’ windings is used in electrical machinery due to very low inductive or capacitive reactance, but platinum can also be used. Some traditional applications have also exploited the linearity of copper sensors in narrow range temperature difference measurements where two sensors are connected on opposed arms of a bridge. There is no internationally recognised standard curve for copper although some national standards exist.

Sensor construction

The most popular industrial designs are the fully encapsulated RTDs. In these units a 0.025 mm diameter or smaller platinum wire is wound into a coil and is inserted in the multiple bores of a ceramic tube or is wound directly on the outside of a ceramic tube. The ceramic material is usually 99.7% pure aluminium oxide, and the winding is completely embedded and fused onto or into the tube. The RTD thermometers are in direct competition with thermocouples and therefore are available with the same features.

The packaging of the RTD sensor depends on the application. For example, the measurement of temperature inside glass-lined chemical reactors requires special configurations.

One limitation of RTDs (relative to thermocouples) is size. RTDs tend to be relatively bulky, because in order to obtain the required resistance (usually 100 ohms) the length of the sensor wire must be relatively long, frequently in excess of a metre. This limitation has been overcome by the film-type designs which are suited for miniaturisation.

The design and construction of a successful detector is, of course, only part of the total problem of producing a good industrial thermometer. The design of the sheath and mounting arrangements, and, particularly the support of the internal leads are of great importance. In practice, the vast majority of failures of resistance thermometers in service are due to faults which occur in either the leads or the sheaths, rather than in the temperature sensors themselves. Apart from complete failure due to obvious causes, such as mechanical mishandling or heating the thermometer well beyond its intended temperature range, problems can occur due to gradual deterioration following long-term

exposure at temperatures near to the maximum design temperature. This is usually the result of gradual 'poisoning' of the platinum by contamination from its surroundings. For example, above 550°C, iron will poison platinum, and so special sheathing techniques are required. Nickel alloys can be used up to about 800°C; above that temperature ceramic sheaths are necessary. As with the high precision standard platinum thermometer, it is essential to ensure that the sheath contains oxygen in order to prevent reduction of metal oxides. This is particularly important when glass is involved in the construction of the detector, because any lead oxide present may be reduced to lead, which can form a low-temperature eutectic with platinum.

3.3.3 Sources of error

Various sources of error may be encountered in the application of resistance thermometers. Some of them are dealt with in this section, which refers particularly to precision platinum resistance thermometers, although applying to some degree to industrial instruments, also.

Strain

As mentioned earlier, the temperature independent component of resistance is affected by dislocations in the crystalline structure of the platinum, so it is necessary to mount the wire in as strain-free a manner as possible. A new thermometer must be stabilised by heating it to a temperature a little above its maximum intended operating temperature, to anneal out any strains imposed during the manufacture of the wire and in the winding of the resistance element. This has the effect of reducing R(0.01°C) and of increasing the -coefficient. In practice R(0.01°C) is re-measured periodically to check the thermometer stability, experience having shown that the calibration constants determined in the original calibration will generally remain valid as long as R(0.01°C) remains unchanged. An increase in R(0.01°C) usually indicates an increase in the residual resistance of the thermometer due to strain (which can be annealed out) or to chemical contamination of the wire (which cannot be removed). In the latter case the thermometer calibration constants will generally require re-determination.

Thermometers which are used at temperatures above about 600°C require special care to avoid rapid cooling, which may cause strain or vacancy-quenching. If removal from a furnace at such high temperatures is unavoidable, additional annealing may be necessary before R(0.01°C) is re- determined.

Self-heating and thermal response

Measurement of the resistance necessitates passing current through the thermometer. Although most bridges are designed to minimise the sensing current (often limiting it to 1 mA or less), the resulting heating that occurs in the resistance element raises its temperature above that of the surroundings until the heat can be dissipated. There are two components in the self-heating effect - firstly, the internal heating effect, which is due to the difference in temperature between the platinum wire and that of the outside of the protecting sheath, and is therefore a function of the thermometer construction. Secondly, there is an external heating effect where the heat emitted by the sensor into the protecting sheath is not absorbed immediately into the thermometer's surroundings; this is a function of the thermal contact between the outside of the sheath and its surroundings. To keep the internal effect to a minimum, thermal contact between the element wire and the sheath is made as good as possible, while the external effect is minimised by designing the apparatus so that the thermometer sheath is in good thermal contact with its surroundings.

The magnitude of the self-heating is found by making measurements at more than one current and assuming that the effect is proportional to the power dissipated, i.e. to the square of the current. The difference between the two readings (which should only be made when the value has fully stabilised) therefore allows a calculation of the self-heating coefficient - the temperature rise per unit power - and hence of the self-heating at the normal measuring current. Most resistance bridges simplify this test by including a facility to increase the current by a factor of √2, so doubling the power dissipated. The difference between the readings at the two currents is then the self-heating at the original current. In situations where the temperature is not steady it will be necessary to alternate between the two currents a number of times before a realistic estimate of the self-heating can be established.

Some self-heating is of course unavoidable. For the most demanding uses of resistance thermometers it is desirable to extrapolate the measured resistance to that at zero current, but it is generally sufficient to calibrate and use a thermometer at the same current. However, the current and self-heating should not be too large or the measurements will be too dependent on the constancy of heat flow away from the element. It is advisable to make periodic checks of self-heating as these can provide early warning of faults in the thermometer or changes in the contact with its environment.

For a glass or silica-sheathed thermometer of laboratory pattern, where R(0°C) = 25 ohms, the heating effect of 1 mA with the thermometer immersed in water at the ice point varies between about 0.5 and 3.5 mK, depending upon its design. An industrial thermometer with an R(0°C) of 100 ohms under similar conditions may show a self-heating effect of 20 mK or more for 1 mA, depending upon the design of the thermometer element and its sheath assembly. Even larger errors can arise in static air temperature measurement, unless the measuring current is kept low. For example, in still air, a 100 ohm element with 1 mA measuring current may show a self-heating error of 50 mK or more.

Closely allied to the self-heating effect is the time response of a thermometer. In order that this should be short, one requires a light-weight design with which good contact can be established. This conflicts with industrial requirements for rugged, vibration-insensitive devices which may have to be inserted into massive pockets in the walls of pressure vessels. Response times for resistance thermometers can be a few seconds for laboratory instruments, but several minutes for heavy industrial units, especially in static media. Special techniques may be applied to measure, and then compensate for, the effects of thermal lag.

Conduction along the sheath and leads: immersion

In any measurement of temperature it is necessary to ensure that the thermometer is sufficiently immersed to prevent errors due to the effects of heat conduction along the leads and sheath. Early thermometers were often constructed with heavy gold leads (chosen because of their low electrical resistance), but the usual practice with precision thermometers nowadays is to use platinum leads, which are arranged to be reasonably close to the walls of the silica or glass sheath. In some designs the first few centimetres of the leads immediately above the resistance element are of reduced diameter wire to further reduce thermal conduction between the element and the leads.

A check upon the effect of conduction along the leads and sheath is usually made by measuring at a steady temperature (for example at the ice point) and then gradually reducing the depth of immersion to see what effect this has on the resistance of the element. An alternative method is to immerse the thermometers in ice so that the part of the sheath containing the element is just covered. The resistance is measured and then more ice is packed up around the sheath while measurements continue until further packing causes no significant change in resistance. In this way the minimum immersion necessary for the required precision at that temperature is determined. However, immersion tests require considerable care, with enough time being allowed for a steady reading to be obtained at each position, and if the temperature is not steady only a crude indication of the effect will be obtained. Poor immersion, or poor contact at the point of measurement, is a major source of error in temperature measurement at all levels.

Thermometers for use below the argon point (84 K) are usually made in capsule form (as described earlier), the capsule being a close fit in the cryostat copper block. The short platinum leads which emerge from the glass seals at the top of the thermometer are soldered to copper leads which are coiled around some sort of thermal anchoring point in the cryostat before being brought out to the measuring apparatus. In this way conduction of heat into the thermometer is suppressed.

Radiation errors

Where the resistance element is enclosed in a transparent glass or silica sheath, as for example in a precision laboratory thermometer, two main sources of error can occur. Firstly there is the transmission of radiation from the element up the sheath by internal reflection. This 'light piping' effect can cause errors of up to about 10 mK at 600°C, the resistance element indicating a low temperature due to loss of heat by radiation along the sheath to the cold outside. The effect is easily overcome by

coating the outer sheath of the thermometer with a suitable black paint or by roughening the outer surface (e.g. by sand-blasting).

The second radiation effect occurs when the sensor can "see" a surface which is appreciably hotter or colder, the power gained or lost from the resistance element resulting in an increase or decrease of its temperature compared with that of its surroundings. The answer here is to shield the sensor by enclosing the thermometer in a metal tube, for example.

Insulation Resistance

As mentioned earlier, the effect of insulation leakage between a resistance element and its supporting former or between the leads and the outer sheath of a thermometer is to reduce the apparent resistance of the sensing element. In laboratory thermometry this does not usually present problems below 660°C, but in high temperature thermometry great care has to be taken over the selection and preparation of materials, and it is usual to use a low resistance element. Thermometers for industrial application, in which the element is not usually hermetically sealed and is contained in a metallic sheath, are particularly liable to errors due to insulation leakage unless care is taken in the selection and purity of the materials used, especially if they are to be used above 600°C. Such thermometers usually have R(0°C) values of 100 ohms, so an insulation resistance of 1 megohm at 0°C would cause an error approaching 0.03 K.

3.3.4 Measurement of RTD resistance

The methods used to measure the resistance of a thermometer may be classified as either potentiometric (in which, at balance, no current flows in the thermometer potential leads), or bridge circuits (where, in principle, current flows at balance although its effects may be negligible). Both dc and ac currents are employed with either resistors or inductive dividers to establish the ratio of the thermometer resistance to that of a standard resistor. For all accurate work, four-lead thermometers are used in order that lead effects may be minimised.

By measuring the resistance of the RTD element one can determine the process temperature if the change in total resistance measured is affected by nothing but the process temperature. In real installations the RTD element is connected by lead wires to the readout or transmitting instrument.

Assuming that the resistance of the RTD (at 0°C) is 100 ohms and assuming that the element is platinum, the resistance of the 100 ohm RTD element will change by 0.385 ohms for each degree Celsius change in temperature. If 500 ft (152 m) of 20 gauge copper lead wire was used to connect the RTD to the bridge, this adds 10 ohms to the total resistance (5 ohms per leg), which will contribute a measurement error (as an increase in process temperature) of 10/0.385 = 26°C.

The previous example illustrates the relatively large lead-wire error in a two-wire RTD installation, and for this reason such installations are not used if accurate temperature measurement is desired and the length of the lead wires is more than a few centimetres. When the transmitter is mounted directly on the thermowell the lead-wire length is so small that the resulting error is not significant. Yet even in these configurations the better suppliers will provide a three-wire RTD to minimise the lead-wire error.

Three-Wire RTDs

The introduction of a third lead wire acts as a ‘sense’ lead and is part of both halves of the bridge, and therefore cancels out at balance. The other two lead wires are in different halves of the bridge. Therefore, now the lead-wire error is no longer the total lead resistance, but only the difference between their resistances. This is a major improvement in reducing the lead-wire error and is sufficient for the needs of most industrial applications where the lead-wire lengths are short. However, it is not a complete solution because wire resistances are guaranteed only within a 10% tolerance; therefore, if the lead wires are rated as being identical wires of identical lengths their resistances can still differ within the 10% tolerance. So if nominally they both are 5 ohms, in reality one could be 4.5 and the other 5.5 ohms. If this were the case, the difference of 1 ohm would still introduce an error. With a 100 ohm platinum RTD that error would correspond to 1/0.385 = 2.6°C.

If the purpose of the temperature measurement is to calculate the exothermic heat release of a batch reactor, this error might still be too much. In that application the temperature rise through the reactor jacket is about 2.8°C and the span usually selected for the differential temperature transmitter is 5.6°C. In order to identify the end point accurately, the total heat release must be determined to within 0.5% maximum error. Because the total heat release is calculated by multiplying coolant flow with its temperature rise, the flowmeter itself will contribute 0.25% in error and therefore one must measure the temperature rise within 0.25%. An error of 0.25% over an actual measurement of 2.8°C is 0.007°C. This is such a small error limit that even three-wire RTD transmitters may not meet it (their usual error limit is about 10 times higher). For this reason, in laboratory situations or for other high- precision measurements one might consider the use of four-wire systems which completely eliminate the lead-wire effect.

Four-Wire RTDs

Four-wire RTDs can be connected either to a null-balance bridge or to a constant current source (CCS). The four-wire null-balance bridge operates by switching a triple-pole double-throw switch and making alternate null-balance measurements in the two configurations. In the first configuration one of the lead wires is measured together with the RTD resistance, while in the second configuration the other lead wire is used instead; so they cancel out completely and the actual value of the RTD resistance can be determined. Microprocessors and advanced electronics make it feasible to provide this level of sophistication, but complexity still costs money, so these designs are relatively expensive; in addition, they are still limited by contact resistance considerations. Even the best (gold-plated) switching contacts contribute some contact resistance, and the difference between these resistances does introduce some minuscule errors whenever one uses a switching configuration to make a resistance measurement.

Another way to eliminate the lead-wire error is to use a constant current source (CCS) in a four-wire RTD configuration. These miniaturised CCS packages are available at relatively low cost and provide an accurately constant current flow of about 2 milliamperes or less to avoid self-heating errors. In this configuration the bridge itself is replaced by a digital voltmeter which measures the resistance of only the RTD and is insensitive to lead-wire effects as there is no current flow through the connecting wires.

In general, two-wire RTDs are used in HVAC-type secondary applications, three-wire RTDs are used in the processing industries, and four-wire RTDs are used in high-precision services or in the laboratory.

A simple dc potentiometric method would be one in which the resistance thermometer RT is connected in series with a known standard resistor RS and a stable current source. A digital voltmeter can, of course, be used instead of the potentiometer; the whole system then lends itself to microprocessor control. Thermal emfs are eliminated by reversing the current and the potentiometer polarity and taking an average of the two sets of readings. The method is particularly suitable where several thermometers have to be measured because their current leads can be connected together in series and the potential leads scanned rapidly.

Various dc bridges have been used over the years and are still commonly found in industrial situations with 2, 3 or 4 lead configurations. Many are of the 'balanced bridge' types, in which the circuit is maintained in a balanced condition by manual or automatic adjustment of resistance in one or more arms of the bridge. The adjustable resistance often takes the form of a slidewire which is linked mechanically or electrically to an associated temperature scale. Another form of instrument used industrially is the fixed bridge, in which only the sensing resistor is allowed to vary, the other bridge resistance being chosen so that the bridge is in balance for one value of the temperature sensing resistor only. Provided that the bridge supply voltage is stable, the out of balance voltage developed across the bridge is a measure of the temperature.

An example of a high precision 4-lead dc instrument which is still in use is the Mueller Bridge. This uses a commutator to interchange the thermometer leads in such a way that the average of two balances is independent of the lead resistance. Other examples of 4-lead high precision dc instruments are the Smith Bridge and the Kelvin double bridge. There is also a dc bridge which uses

the ratio transformer principle. This is the Kusters current comparator bridge, in which the ratio of the currents through the thermometer and a standard resistor is measured when the potentials across them are equal.

Over the last 25 years or so exploitation of the inductive voltage divider or ratio transformer, and advances in electronics have caused increasing application of ac measuring equipment. The overwhelming advantages of the inductive voltage divider are its accuracy (1 in 108 is possible without too much difficulty), and its stability, which has removed the need for periodic re-calibration. Modern electronics have produced both high resolution phase-sensitive detectors and self-balancing devices.

One of the earlier ac bridges to be used for resistance thermometer measurement is a Kelvin double bridge. The outer and inner dividers being mechanically coupled to give the same ratio at all times. The effects of lead resistances are small, and for the most precise work, the lead resistances are measured and corrected for. An automatic, seven decade version of this bridge was available commercially for many years with an operating frequency of 375 Hz, and many such bridges are still in use.

In order to eliminate lead resistance errors entirely, bridges based upon the ac potentiometer, using multi-stage transformers have been designed. One such bridge was designed by Thompson and Small; a later version by Knight in 1977 used a three-stage transformer rather than the two-stage one of Thompson and Small.

The present generation of ac bridges, which has been on the market for several years now, is self- balancing and can be computer-coupled to provide a direct temperature read-out, using the calibration constants of the thermometer. The principal feature of these bridges is the use of high- precision, low-noise voltage followers driving a small multi-stage, multi-decade ratio transformer. In the most accurate version, the voltage followers drive the energising windings of a multi-stage transformer in order to 'boot-strap' the input impedance of the main winding. A particular feature of these bridges is the lower operating frequency (25 Hz or 75 Hz), which has the advantage that for many applications traditional designs of dc resistors can be used as standards.

For laboratory use to a more modest accuracy, and for some industrial applications, direct-reading instruments are finding increasing use for temperature measurement in the range from about -200 to 650°C. They are self-balancing and can accommodate two, three or four-lead sensors, the most straightforward approach being from the use of a digital multimeter. Its resistance readings can be converted into temperature either manually or by means of a microcomputer.

Direct-reading instruments often use linearising techniques, which may take into account either the average characteristics of a particular class of platinum resistance thermometers, or the individual characteristics of a particular sensor. In the latter case, of course, much better accuracy can be achieved but the thermometer sensor cannot be replaced without recalibration of the instrument. Another approach is to supply the calibration characteristics of each thermometer sensor in a read- only memory, which is plugged into the linearising and indicating unit together with the sensor. More recently, instruments in which the thermometer calibration constants can be entered via a keyboard have become available.

Digital thermometers are available with resolutions of 10 mK or better; however, care must be taken to ensure that the measuring current is not excessive and that it is sufficiently constant to guarantee reproducible self-heating. Also, in many instruments the linearisation is to the accepted industrial thermometer standard (IEC 751, or BS EN 60751) rather than to the interpolation equations used to define ITS-90. As a result, the readings are likely to be in error by several hundredths of a degree in the -200 to 600°C range, and by more if the range extends to higher temperatures. For the best accuracy, it may be necessary to calibrate the unit and to use external correction tables.

ADVANTAGES AND LIMITATIONS OF RTDs

RTDs are among the most accurate, reproducible, stable, and sensitive thermal elements available. Some of the precision platinum RTDs can measure within a few thousandths of a degree, and this precision is the reason such instruments are used to define parts of the International Temperature Scale (ITS-90). Other advantages include their relatively good sensitivity, and their ability to use conventional copper lead wire (instead of more expensive thermocouple wire). An advantage of copper RTDs is that since both the element material and the lead-wire material are the same, the thermocouple effect is minimised at their junction. Yet another advantage of RTDs is the convenience of using a single bridge to measure the temperature difference between two RTDs.

The disadvantages of RTDs include their higher cost, more fragile construction, and larger size, relative to thermocouples. Because of their size, their thermal response time is also relatively slow. RTDs are also subject to self-heating. The size of this error rises with RTD size and resistance and is reduced by heat transfer and by minimising (about 2 milliamperes in CCS) or eliminating (null-balance) the current flow through the RTD. Errors can be introduced if the RTD insulation resistance is affected by moisture being sealed in the sheath or due to contact between element and sheath. Some RTDs are more vibration- sensitive than others. RTDs also depend for their precision on stable (insensitive to temperature changes) and constant resistances and power supplies in the associated bridges.

In terms of installation, RTDs require the same precautions as thermocouples. The best installation practice is to place all electronics directly on top of the thermowell and thereby eliminate lead-wire and noise effects. If for some reason this cannot be done, the lead wires should be twisted and shielded; also, the wires should not be stressed, strained, or made to go through steep temperature gradients. The extension wire should be low resistance (large diameter), and the readout instrument should be guarded.

Table 3.3 Advantages and limitations of RTDs

Note: additional information concerning the use of the platinum resistance thermometer can be found in the Appendix section of this guide:

Appendix C: The role of the platinum resistance thermometer in the ITS-90 Appendix D: Tolerance classes for Pt100 thermometers (IEC 751: 1983) Appendix E: Pt100 reference tables (IEC 751: 1983)

3.4 THERMISTORS

3.4.1 Overview

A thermistor is simply an electrical resistor whose resistance changes rapidly with temperature. Usually thermistors are made from small beads of complex materials and although they are not as reproducible as platinum resistance thermometers, they are much more sensitive. Commonly the resistance falls exponentially with increasing temperature and such devices are said to have a negative temperature coefficient (NTC). The strong change in temperature is useful for establishing sensitive temperature control of a system.

Thermistors are semiconductors made from specific mixtures of pure oxides of nickel, manganese, copper, cobalt, tin, uranium, zinc, iron, magnesium, titanium, and other metals sintered at temperatures above 1800°F (982°C). Their distinguishing characteristics are a high temperature coefficient and the fact that their resistance is a function of absolute temperature. Bell Laboratories in the USA manufactured them some 60 years ago, naming them from the term "thermally sensitive resistors". While their use as temperature sensors was recognised over 100 years prior to the Bell investigation, the sensors were unstable and not reproducible.

Their temperature coefficient is usually negative, although it can be positive as well. In some cases, with the negative coefficient type, thermistor resistance decreases at the rate of over 3% for each °F temperature rise. This makes them very good for narrow span measurement, but rather hard to handle for wide span applications.

As a consequence of the high sensitivity, the range of use of an individual thermistor is comparatively narrow, and a variety of thermistors are needed to suit different ranges and applications. They are popular in digital thermometers for everyday, medical and industrial use. The following graph compares the characteristic of a low-temperature thermistor with that of a PRT.

Figure 3.5 Schematic variation of the electrical resistance of a platinum resistance thermometer (PRT) and a thermistor. Notice that the resistance of a PRT increases gradually with increasing temperature. In contrast, the resistance of a thermistor falls dramatically as the temperature is increased.

Thermistors are widely used in medical applications and for control purposes where the sensitivity is more important than long-term stability. Their characteristics are, however, very non-linear, and interchangeability has been something of a problem in the past, because it is difficult to trim the small beads of material to precise resistance values. However, thermistors with close interchangeability are

now available, as are many linearising devices for use with them. Thermistors for use at high temperatures (even up to 1000°C) are also available.

3.4.2 Sensor types

The Bell project resulted in the development of thermistors stable and reproducible enough to make their large-scale use worthwhile in telephone work around 1940. However, industry in general did not accept these sensors until the 1950s. Bad experiences with commercially available thermistors hampered their acceptance. Variances in resistance at a given temperature and in rate of change with temperature made individual calibration a requirement. Overcoming this difficulty, one manufacturer patented a process for interchangeable thermistors, resulting in production of probes interchangeable to 0.05°F (0.03°C).

A number of configurations are possible. Most familiar is the bead type, which is usually glass-coated. However, thermistors can be made into washers, discs, or rods. They can also be encapsulated in plastic, cemented, soldered in bolts, encased in glass tubes or needles, or a variety of other forms.

These assemblies serve to support the sensor, protect against damage to the wires, direct flow across the unit uniformly, permit sealing of conduits or flowlines, and provide for easier handling.

Thermistors can be used in a number of applications other than temperature measurement. Some of these are voltage regulation, flowmeters, power level controls, gas analysers, and vacuum detectors.

The I2R temperature increase is a direct function of the dissipation constant in its mounting environment, and, unlike resistance thermometers or thermocouples, thermistor resistance values are varied by varying their composition to suit the temperature span, range, and sensitivity desired for a given application. If a small current flows through the thermistor, there is a negligible increase in its temperature due to this flow. The resistance can thus be measured and, since resistance is proportional to absolute temperature, the temperature can be inferred.

3.4.3 Self-heating effect

If the current flow is slowly increased, the heat generated within the sensor will gradually begin to raise its temperature above that of its environment. This in turn lowers its resistance and more current will flow. Eventually, the current input will reach a level where it is balanced by the heat output of the thermistor. At this point, sensor temperature would be 300 or 500°F (149 or 260°C) above ambient and sensor resistance would have dropped to a value approximately 10-3 times its value at the original low current.

For most temperature-measuring applications, self-heating is not a problem, since thermistor currents used are relatively low. If a current of higher magnitude is used, under fixed conditions, an offset allowance may be made for the self-heating effect. However, anything that affects the dissipation constant will change the offset. This could be the result of a number of things, including flow changes in the measured medium, changes in fluid composition, and the like.

The power dissipation factor of a thermistor element can vary from a few microwatts to several watts per degree of resultant temperature rise. In industrial units a few degrees of self-heating can be expected for each milliwatt of resistive heating. As thermistors are made smaller and smaller in order to increase their response times and to lower their thermal shunting, the effect of self-heating also rises.

3.4.4 Measuring bridges

To measure temperature with a thermistor, a Wheatstone bridge circuit can be used as mentioned earlier. But, a simple circuit consisting of a battery, a sensor, and a microammeter can also be used. In such a circuit, the sensor will have a very high resistance. As long as the voltage is constant, the current flow will be determined only by changes in resistance of the thermistor. The sensor can be mounted fairly far from the meter and ordinary copper wire can be used for transmission. With a

sensor resistance on the order of 105 ohms, meter temperature changes and leadwire resistances can be neglected.

Assuming that the thermistor changes its resistance by 4% (by 4000 ohms) for each degree of temperature change and assuming that the total lead wire used was 500 ft of 20 gauge copper wire having a resistance of 10 ohms, the lead-wire error would amount to only 0.0025 degrees of temperature error. Even if the thermistor had a total resistance of only 5000 ohms, and therefore its 4% temperature coefficient were only 200 ohms/°C, the lead-wire error would still only be 10/200 = 0.05°C. For this reason lead-wire compensation is not a serious problem, and three- or four-wire bridges are not used with thermistors.

Locating the thermistor in one leg of a bridge circuit with a centre zero galvanometer enables very narrow temperature spans to be displayed relatively inexpensively. Range depends upon galvanometer sensitivity and can be as low as 2°F (1°C).

Very accurate temperature measurements can be made with a differential circuit. With two thermistors in different halves in a bridge, the unbalance will be determined by resistance difference caused by temperature difference of the sensors. With a high gain amplifier, differentials of a few mK can be measured.

3.4.5 Thermistors combined with resistors

Combinations of thermistors and wire-wound resistors can be connected in networks to produce either a varying voltage or resistance which is linear with temperature. The basic equation for a divider network of R1 and R2 in series is Eout = EinR1/(R1 + R2), where Eout is the voltage drop across R1. When R1 is a thermistor and Eout is plotted against temperature, the total curve is non-linear and S-shaped. If R1 is modified by the addition of another thermistor and resistor of proper values, linearity of the centre section of the S curve can be extended to cover a relatively wide temperature range. This section then is considered to follow the general equation of a straight line. For the resistance mode, this would be R1 = MT + b, where M is slope in ohms per degree, T is temperature in degrees F and b is value of R1 when T = 0°F. The advantages of such a system are obvious.

3.4.6 Specialised applications

Thermistors can be connected to a microprocessor-based memory element and packaged as a portable element. These small micropacks can memorise the temperature history of batch reactor products or can travel with the process material through several steps in processing. When the temperature memory-pack unit is retracted from the process it can be plugged into a computer for interrogation. This allows the plant to store the temperature history of each batch. The micropack weighs less than 100 g, has a battery life of about 500 hours, and has a temperature range of –40 to 302°F (-40 to 150°C).

Another specialised application for thermistors is to use them as regenerative temperature switches. These units are produced through the use of thin film technology where a change in the resistance of a metal oxide (Moxie) film is used as a temperature switch. In the extended scale mode, Moxies behave like thermistors but exhibit a resistance change of up to 300%/°C. In the switching mode, they operate as thermally actuated switches. The switching temperature is solely determined by the programming voltage. On passing through their actuation temperature, Moxies abruptly switch to their low impedance state and latch.

3.4.7 Calibration and testing

The National Bureau of Standards has offered a limited calibration service for thermistors over a temperature range of approximately -150 to 200°F (-101 to 93°C). Calibrations in the order of ±0.02°F (±0.01°C) are the rule. Users have claimed better than ±0.002°F (±0.001°C) stability over a two-year period.

The following should be checked to test the operating parameters of thermistors:-

1. Zero power resistance. This test is done under conditions that produce negligible heating of the sensor by test current. Most common sources of error are measurement of ambient temperature, self- heating error, thermocouple effects at junctions of dissimilar metals, and accuracy of test equipment. An ordinary ohmmeter is too crude. A five-dial Wheatstone bridge and a good mirror galvanometer will do the job.

2. Temperature coefficient of resistance. The rate of change of thermistor resistance versus temperature at the desired temperature.

3. Voltage developed across the thermistor under conditions of thermal equilibrium with a constant current. A current-limiting resistor should be used. Maximum current should not be exceeded, even for short periods. The sensor should not be moved to a medium of lower thermal conductivity during a test. The problem is similar to that of measuring voltage across ordinary resistors; the difference is that stabilisation times are longer and small currents are used.

4. Time required for the thermistor to pass a certain current after the voltage is applied.

5. Dissipation constant measurements, which have been defined as the ratio at a certain temperature of a change in power dissipation in a thermistor to the resultant body temperature change.

6. Thermal time constant, which is the time for a 63.2% change from initial to final temperatures when subjected to a step change.

We have spoken mainly of negative coefficient thermistors, since they are most widely used. However, positive temperature coefficient thermistors are also used to protect motors from overheating, for very accurate measurements at high temperatures, and for non-expendable current limiters, for example. In these applications, the positive increase in resistance with temperature and rapid change in sensitivity at switching temperatures are useful.

ADVANTAGES AND LIMITATIONS OF THERMISTORS

Thermistors have the desirable characteristics of small size, narrow spans, fast response (their time constant can be under a second), and a very high sensitivity (about 4%/°C), which usually increases as the measured temperature drops. Thermistors do not need cold junction compensation because their resistance is a function of absolute temperature, and errors due to contact or lead-wire resistances are insignificant because of their relatively small values. Thermistors are available in a great variety of configurations, they are inexpensive, they are not affected by polarity, and their stability increases with age. They are the most sensitive differential temperature detectors available.

The disadvantages of thermistors are also many. They are neither interchangeable nor linear, although in modern data acquisition systems linearisation is no problem. They are also fragile and are not suited for wide spans. Their high resistance necessitates the use of shielded power lines, filters, or DC voltage. One of the most serious limitations of thermistors is their lack of stability (drift and de-calibration) at higher temperatures. This is because semiconductors in general change their composition and develop migration, diffusion, or other decomposition properties at high temperatures. Therefore, their use is limited to temperatures up to 600°F (316°C), and even at such temperatures extended exposure can cause drifts. (It should be noted that if a thermistor is operated at temperatures far below its maximum it can be a very stable device). Thermistors also have a low temperature limit, because as the temperature drops their resistances rise to such levels that measuring it becomes difficult.

Table 3.4 Advantages and limitations of thermistors

3.5 THERMOCOUPLES

Figure 3.6 Illustration of a simple thermocouple

3.5.1 Overview

Thermocouples are the most common sensors in industrial use. They have a long history, the original paper on thermoelectricity by Seebeck being published in 1822. They consist of two dissimilar metallic conductors joined at the point of measurement. If the voltage across the unconnected ends of these wires is measured then it is found to vary according to the temperature at the junction between the two wires. This is the simplest form of thermocouple circuit.

What is a thermocouple?

When a conductor is placed in a temperature gradient, electrons diffuse along the gradient and an emf, or thermovoltage, is generated. The magnitude of the emf depends on the material and also on its physical condition. To measure the generated thermal emf (or Seebeck emf), the circuit must be completed using a second different conductor. This is joined to the first conductor at the point of measurement and passes through the same temperature gradient, forming a thermocouple. The thermocouple emf is then the difference between the emfs generated in the two conductors.

In practice, thermocouples have two junctions. One of the junctions is held at the temperature, t1, to be measured, for example in a furnace. The second reference junction is held at temperature t2 which is usually the melting point of pure ice. This can be done with real melting ice or electronically, but with some loss of accuracy. The difference in the emfs generated by the two conductors is then given by:

t1 t1 t1 E = ∫ Sa dt - ∫ Sb dt = ∫ (Sa – Sb) dt t2 t2 t2 where E is the net emf generated and Sa and Sb are the Seebeck coefficients of conductors a and b.

Many different thermocouple combinations have been used, but only 8 are standardised in IEC 584-1. These include 3 noble metal thermocouples using platinum and platinum-rhodium alloys, widely used for temperature measurement up to 1600°C. The remaining 5 mainly use nickel-based alloys, which are cheaper and more suitable for industrial use up to about 1200°C. Other refractory alloys can be used up to and beyond 2000°C.

The simplicity, ruggedness, low cost, small size and wide temperature range of thermocouples make them the most common type of temperature sensor in industrial use. However, there are a number of potential problems in use and in practice, several techniques to moderate these are nearly universally employed:

• The wires are insulated. This prevents the wires touching except at the tip. The ability of the insulation to withstand temperature is often the limiting factor in working out the highest temperature at which a thermocouple can be used. • The wires are usually welded together at the tip. Any kind of contact at the tip will create a thermocouple. Soldering is possible, as is mechanical crimping, but welding produces the smallest and most robust tips. • Standard combinations of wires are used. Many different materials can be used in the manufacture of thermocouples. These include platinum and platinum-rhodium alloys for use up to 1600 °C; cheaper, mainly nickel-based alloys for industrial use up to 1200 °C; and refractory alloys up to and beyond 2000°C. Standardised letter-designated thermocouple types (e.g. Types K, N, T, B, R, S) are manufactured to meet the specifications of IEC 584 (BS EN 60584). This standard gives tables and equations for the EMF (or voltage) generated at every degree in the range, and the tolerances with which they should be manufactured. Thus, for example, a K-type thermocouple has wires made from carefully selected alloys of Nickel-Chromium and Nickel-Aluminium, for which the EMF at 1000 °C should be 41.276 mV, within about 0.16 mV. Summaries of the tables can be found in many reference books, such as (Kaye and Laby, Tables of physical and chemical constants). • Cold junction compensation. Because the voltage is generated along the wires where there are temperature gradients (see below), it follows that the temperature of the 'free ends' of the wires is important and must be controlled. The 'reference' or 'cold' junctions are commonly held at 0 °C, using melting ice for best accuracy, but in most digital instruments the cold junction effect is compensated by sensing the reference temperature and adding the appropriate voltage to the signal, thereby allowing digital multi-meters to display the temperature directly.

The voltage generated by thermocouples is not usually very large, typically between 10 µV and 40 µV per degree Celsius.

Figure 3.7 Illustration of a common configuration for the use of a thermocouple: The thermocouple junction is placed where the temperature needs to be measured, and the other ends of the wires are joined to copper wires in a reference bath held at a known, stable temperature (typically that of melting ice). The thermocouple voltage (typically 40 µV for each degree Celsius difference between the junction temperature and the reference temperature) is measured by a high-resolution voltmeter or analogue to digital converter.

How thermocouples work

Although the voltage measured by a thermocouple reflects the temperature at the junction between the two dissimilar metals, surprisingly, none of the thermovoltage is actually generated there. Instead

it arises all along the wires and depends on the product of the so-called Seebeck coefficient S(x) for a material and the temperature gradient dT/dx at each point on the wire.

Figure 3.8 Demonstrating the principle of operation of the thermocouple

In the figure above, points A and C are at the same reference temperature TR, and point B is the measurement junction with temperature TJ. A is connected to B by wire with Seebeck coefficient S1 and C is connected to by wire with Seebeck coefficient S2. The voltage generated in the circuit, VAC , is built up along the two wires, and mathematically this can be written as an integration between points A and C:

C VAC = ∫ S(x) dT/dx dx A

If the Seebeck coefficients are constant along each type of wire, then it can be shown that the measured thermovoltage is proportional to the temperature difference between TR and TJ and given by:

VAC = [TR – TJ] x [S2 – S1]

The role of calibration

Buying standard thermocouple wire from a reputable company should allow you to construct thermocouples which are accurate within the specified tolerances. If measured correctly, the tip temperature inferred is unlikely to be in error by more than 5ºC over the range from -100ºC to over 1000ºC. However, if you want to be sure that this is so you must make or buy a standard thermocouple, and have it calibrated at a laboratory whose thermometers can be traced back to the ITS-90.

3.5.2 Thermocouple materials

Many combinations of materials have been used to produce thermocouples: the American Society for Testing and Materials (ASTM) Handbook and the monograph by Kinzie give descriptions of a large number of them. Probably because of a combination of economic pressures and the general move towards technological standardisation, only a few are readily available. The International Standard IEC 584-1:1995 (soon to become BS EN 60584-1, replacing BS 4937 Parts 1 to 8), gives polynomials and reference tables for the eight standardised thermocouples. These standard thermocouples are most conveniently described by letter designation - a system originally proposed by the Instrument

Society of America and now used internationally. The system also has the advantage that it avoids the use of sometimes confusing trade names for base metal thermocouples.

A description of the properties of standard thermocouples follows. Note that in thermocouple terminology the positive element is designated first.

Type S - Platinum - 10% Rhodium/Platinum

This is the original Le Chatelier couple of 1886 and was used to define IPTS-68 over the temperature range 630.74°C to 1064.43°C (the freezing point of gold). It may be used in oxidising or inert atmospheres continuously at temperatures up to 1400°C and for brief periods up to 1650°C. For high temperature work it is advisable, and indeed usual, to use insulators and sheaths made of high purity recrystallised alumina. In all but the cleanest laboratory applications the thermocouple must be protected by an impervious sheath as small amounts of metallic vapour etc. can cause deterioration - invariably leading to a reduction in emf for a given temperature. Continual use at high temperatures can cause excessive grain growth and weakening of the pure platinum arm. There is also the possibility of diffusion of rhodium from the platinum-rhodium conductor into the pure platinum conductor, again leading to a reduction in output.

Type R - Platinum - 13% Rhodium/Platinum

This thermocouple may be used under the same conditions as Type S. However, it does have the advantages of slightly higher output and rather better stability than Type S. There are no real technical (as opposed to historical) reasons for the continued existence of both types, though in the UK Type R has been preferred because of its better stability.

Type B - Platinum - 30% Rhodium/Platinum - 6% Rhodium

Developed in the 1950s by the Degussa Company in Germany this thermocouple may be used continuously up to about 1600°C, and intermittently up to about 1800°C. Restrictions and conditions of use are otherwise similar to Types S and R. The output is lower, however, and this thermocouple would not normally be used below about 600°C. It has the advantage that since the output is negligible over the range 0 to 50°C, cold junction compensation is not normally necessary.

Type J - Iron/Copper-Nickel

Introduced around 1910 and known as iron/Constantan, this is one of the few thermocouples that can safely be used in reducing atmospheres. It may also be used in vacuum or in inert atmospheres. The maximum continuous operating temperature is around 800°C although for short term use one can go up to 1000°C. Under oxidising conditions the iron element oxidises rapidly at temperatures above 550°C. Although the BS table goes down to -210°C caution needs to be exercised when using this couple at sub-ambient temperatures because of the risk of condensation occurring with subsequent rusting of the iron arm. Some low temperature enbrittlement can also occur.

Type K - Nickel-Chromium/Nickel-Aluminium

Introduced around 1906 this thermocouple is usually called 'Chromel-Alumel', although this is in fact the trade name under which it was introduced by the original manufacturers - Hoskins Manufacturing Co, USA. It is by far the most common thermocouple in industrial use today. It may be used continuously at temperatures to about 1100°C in oxidising atmospheres, and intermittently at up to 1200°C, and it is also suitable for cryogenic use down to about -250°C. Great care should be taken before use in other than oxidising conditions, although above about 800°C oxidation increasingly causes 'decalibration'. While Type K is widely used it is not as stable as some of the other base metal couples in common use. A reversible metallurgical instability in the nickel-chromium arm at temperatures between 300°C and 550°C can cause errors of several degrees, in addition to the slow drift in calibration at high temperatures due to oxidation. The couple is extensively used in nuclear installations because of its relative immunity to radiation damage. However, in this respect also, the Type N thermocouple would now be preferred.

Type T - Copper/Copper-Nickel

This thermocouple, copper/Constantan, is much used for laboratory temperature measurement over the range -250°C to about 400°C, the upper temperature limit being imposed by the increasingly rapid oxidation of the copper arm. Several workers have found that excellent repeatability is possible (±0.1°C over -200 to +200°C). The calibration section at NPL formerly used this thermocouple as a comparison standard over the oil-bath temperature range (up to 200°C). The high thermal conductivity of the copper arm should be taken into account - it can be either an embarrassment or a virtue, depending upon the application.

The copper-nickel alloy used for the negative arm of this couple is not of the same composition as that used in Type J, and is therefore not interchangeable.

Type E - Nickel-Chromium/Copper-Nickel

The Nichrome/Constantan thermocouple has the highest output of the commonly used thermocouples (up to 82 V/°C at 400 to 600°C) - this is of perhaps less importance in these days of digital voltmeters with nanovolt resolution. The usable temperature range extends up to about 900°C in oxidising or inert atmospheres. It may also be used in the cryogenic region down to about –250°C. This thermocouple combination seems to be more stable than Type K, and is therefore more suitable for accurate measurement.

Type N - Nickel-Chromium-Silicon/Nickel-Silicon

Because of the defects in Type K thermocouples, much work has been carried out to try to develop improved alloys. This resulted in the production of the nickel-chromium-silicon and nickel-silicon alloys that have been given the names Nicrosil and Nisil. This is a thermocouple combination which has a much improved metallurgical stability than Type K and much greater resistance to drift caused by oxidation at high temperatures. Because of the oxidation resistance the Nicrosil/Nisil thermocouple is usable to somewhat higher temperatures than Type K. A full description of the development and properties of this thermocouple is to be found in the NBS Monograph. It was standardised in IEC 584- 1 in the late 1980s as Type N.

This is probably the most significant development in base metal thermocouple materials for many decades. It must be said that, so far, it has not been much used in industry - partly due to simple human inertia, but also because in sheathed mineral-insulated form the oxidation resistance of the Type N system is less pronounced compared with sheathed Type K. New sheath alloys have now been developed so as to take full advantage of the potential for use at high temperatures.

Note - For base metal thermocouples generally, the maximum operating temperature is related to the diameter of wire used - as well as to the environment and the required life. For these reasons figures quoted in the literature for maximum working temperatures are subject to quite wide variation. This question is discussed further in BS 1041 Part 4.

Non-standard thermocouple types

Over the years literally hundreds of non-standard thermocouples have been produced for particular purposes, but few have been widely used. However, the following are important in special applications.

Gold/Platinum and Platinum/Palladium

These elemental thermocouples have received attention in recent years for the most precise uses in air up to 1000°C and about 1300°C, respectively. They may be prepared as for platinum-rhodium thermocouples, except that the gold wire must be supported during annealing, and some provision may need to be included to allow for the differential expansion between gold and platinum. Under the best conditions reproducibilities of better than 0.03°C have been reported for gold/platinum, not far short of results obtainable with the much more expensive high temperature standard platinum

resistance thermometer. Standard tables are being developed for incorporation in the international and national standards.

Tungsten-Rhenium Thermocouples

Tungsten-rhenium alloys have been developed as thermoelements for use at temperatures beyond the range of platinum-rhodium types. Three combinations are presently in use:

Tungsten/Tungsten - 26% Rhenium Tungsten - 3% Rhenium/Tungsten - 25% Rhenium Tungsten - 5% Rhenium/Tungsten - 26% Rhenium

Of these, the first listed is the cheapest but suffers from enbrittlement problems in the pure tungsten arm. These thermocouples may be used continuously up to 2300°C and for short periods as high as 2750°C in vacuum, pure hydrogen, or pure inert gases, though selective vaporisation of rhenium may occur at temperatures in excess of 1800°C. The ASTM has recently published a standard for the 3/25 and 5/26 thermocouples including tables from 0°C to 2315°C (4200°F) with a suggested tolerance of ±1% of the temperature. Beryllia or thoria insulators have been recommended for these thermocouples, although some reaction may occur between wires and insulator at the upper end of the temperature range.

Iridium - 40% Rhodium/Iridium

Thermocouples made from varying proportions of these two elements are the only noble metal thermocouples that can be used without protection in air up to 2000°C - albeit only for limited periods. They may also be used in vacuum or inert atmospheres. The wires become very brittle and fragile due to grain growth after exposure to high temperature. There are no standard reference tables for these thermocouples, batch calibrations being supplied by the manufacturers.

Platinum - 40% Rhodium/Platinum - 20% Rhodium

This thermocouple may be used instead of Type B where a slightly higher upper temperature limit is required. It may be used continuously up to 1700°C and for short periods up to 1800°C or even 1850°C. Conditions of use are as for the other platinum alloy thermocouples. Although an unofficial reference table exists it is usual to obtain a batch calibration from the manufacturer.

Nickel-Chromium/Gold - 0.07 atomic % Iron

At low temperatures most thermocouples exhibit low sensitivity whereby the generated emfs decrease markedly, but small amounts of magnetic impurities in certain metals can give rise to substantially enhanced values. Gold-iron is one such alloy, and coupled with Chromel the output is about 15 V/K at 4 K. The thermocouple may be used down to 1 K or lower, but the generated emf falls off rather rapidly below 4 K, which is a more realistic lower limit. A reference table has been published by the National Bureau of Standards. In Europe, a gold - 0.03% iron alloy is more commonly used.

3.5.3 Hardware and fabrication

A practical thermocouple must consist of at least two wires or conductors, made from dissimilar materials, usually insulated, and joined at one end to form the, 'measuring junction'. The wire diameter may range from 0.2 mm or less for fine applications, to 3 mm or more for industrial use. Junctions are best made by welding, although soft or hard soldering can be adequate. The welding, or soldering, technique used should be such that minimum change is caused to the composition of the thermocouple wires. Base metal couples are usually welded electrically in an argon atmosphere. Platinum thermocouples may be welded using a small oxy-hydrogen flame. Base metal thermocouple wires are usually supplied in the fully annealed state and may be used after welding without further treatment. Platinum metal thermocouple wires however, are usually annealed before use. For the most precise work, rather complicated annealing schedules have been suggested, but for most

purposes it is sufficient to anneal the platinum wire at 1100°C for 15 mins and the alloy wire at 1450°C for the same time by passing a suitable electric current through the wires, in air. Usually 12 to 13 A suffices for 0.5 mm diameter wires and the current is reduced gradually over a period of a few minutes at the end of the anneal. After inserting the wires into the insulators and making up the junction the completed couple should be annealed at 1100°C.

Insulators

There is a wide choice of insulating materials for thermocouples. The insulation, where practicable, is colour coded to indicate the thermocouple type. There is, at present, no internationally agreed standard. The following materials are those most commonly used:

1. PVC may be used over the temperature range -10°C to +105°C. It is available either in 'figure-of- eight' or 'flat-pair' configurations.

2. PTFE can be used from -75°C to +250°C (or to 300°C for short periods). It is available in 'flat-pair' or 'twin-twisted' forms.

3. Glass-fibre, varnish impregnated is used from -50°C to +400°C.

4. Plain (i.e. untarnished) glass-fibre is usable up to 500°C.

5. Ceramic insulators are available in various forms and materials. Porcelain 'fish-beads' may be used on the larger sizes of base metal thermocouple wires (say 1 mm dia. upwards). Mullite (aluminium- silicate) twin-bore insulators are extensively used with Type K thermocouples in industrial furnaces, either used unprotected or housed in metal or ceramic sheaths, depending on the application. Platinum metal thermocouples are best used in single piece twin-bore high purity alumina insulators in order to minimise the risk of contamination. These are available in lengths of up to 600 mm.

6. Mineral insulated metal sheathed (MI or MIMS) thermocouples consist of a metal sheath enclosing highly compacted mineral insulant in powder form (usually magnesium oxide) supporting the thermocouple wires. Various forms of construction are possible. Cables are made having from two to six cores, in diameters from 0.5 to 8 mm. The junction may either be insulated from the sheath, or bonded to it. The insulated form of junction ensures freedom from 'ground-loop' effects on the associated instrumentation, whilst the grounded junction type has a faster thermal response.

MI thermocouples have many advantages - small size, ease of installation, mechanical strength, good isolation of the junction from hostile environments, choice of outer sheath to suit wide range of operating conditions, high stability, good insulation resistance, reasonable first cost and ready availability 'off-the-shelf'. The thermocouple wires are available in the usual material combinations, including platinum metal types. Sheath materials available include mild steel, stainless steels, Inconel, and cupro-nickel. For high temperature use, particular care must be taken to select a sheathing material which will withstand the harsh environment and will not contaminate the thermocouple element. Special alloys have been developed for use with type N thermocouples, and platinum- rhodium alloy sheaths are usually used with platinum-type thermocouples.

Connectors

A completed thermocouple may be made up in various ways. For the simplest application the thermocouple will end in bare wires which are simply twisted around copper wires to form the reference junction which will be inserted into glass tubes immersed in a crushed-ice/water mixture. However, most applications will need something rather more robust. Many commercially supplied complete thermocouples incorporate a special connector, the contact parts of which are made from the same materials as the thermocouple itself. This helps to avoid the generation of spurious thermal emfs if the connector is in a region of non-uniform temperature (which in the case of a typical connector near to a hot furnace is highly likely). The connector is used to connect the thermocouple to an extension cable.

Protection tubes and sheaths

In some applications the thermocouple can be used without a protective sheath. Typical examples might be a heavy gauge Type K couple used in an electrically heated metal treatment furnace, or a platinum-rhodium couple in a clean research apparatus where the refractory components are made from high-purity alumina. However, in most applications the thermocouple must be protected from the environment whose temperature is being measured. The protective tube also serves the secondary purpose of shielding the thermocouple from mechanical damage.

Metal protection tubes may be used with base metal couples up to about 1150°C, or with a high-purity alumina liner with platinum couples to the same temperature. Ordinary carbon steel tubes can be used in oxidising atmospheres up to about 700°C. Austenitic stainless steels (300 series) can be used at temperatures up to about 850°C, again in oxidising atmospheres. Ferritic stainless steels (400 series) may be used at up to around 1150°C in both oxidising and reducing atmospheres. Inconel may be used in oxidising conditions at temperatures up to about 1150°C.

Thermocouples for use in pressure vessels are usually inserted into thermowells, for which various standards exist, mostly originating from the chemical and petroleum industries. These are currently being standardised within Europe.

Ceramic protection tubes are frequently used at the higher temperatures - although they may also be used at lower temperatures in atmospheres which are corrosive to metals. The material most commonly used is mullite, which may be used at temperatures up to 1600°C. It has good mechanical strength and reasonable resistance to thermal shock. Because of its silica content it should not be used with platinum thermocouples. An outer sheath of silicon carbide is sometimes used where greater resistance to thermal shock, abrasion, and chemical attack is required.

For high temperature use (say above about 1200°C) with platinum thermocouples, or where the use of a gas-tight tube is essential, recrystallised alumina is the most suitable sheath material.

3.5.4 Emf measurement

The emfs generated by thermocouples range from around 10 V/°C for Type S to a maximum of about 80 V/°C for Type E. Hence the emfs to be measured range from a few hundred microvolts (as for Type T at around ambient temperature) to about 75 mV (Type E near the upper end of its temperature range). The resolution required ranges from a fraction of a microvolt for precision measurements with platinum couples to about 100 V for simple moving-coil indicators used with Type K couples on industrial furnaces.

It is important to realise that thermocouple emfs must be measured absolutely and therefore that some sort of reference voltage source is needed, such as a standard Weston cell (for laboratory use) or a solid state Zener diode (for more robust applications).

The simplest form of emf measuring device is the moving-coil meter (in reality of course a current measuring instrument). These are simple, cheap, and robust, and for temperature indicating purposes are often made with a horizontal scale. The non-linear emf/temperature characteristic of thermocouples can be accommodated by means of a non-linear scale, by means of non-linear deflection characteristics, or by a combination of the two. In addition, a form of 'cold-junction compensation' can be introduced by the use of bimetallic devices acting on the meter hairsprings. These indicators were formerly much used on laboratory and industrial furnaces and were commonly calibrated in 5 or 10°C divisions. The thermocouple/meter combination is often referred to as a 'furnace-pyrometer', a usage that tends to cause confusion with radiation pyrometers.

Potentiometric chart recorders are electro-mechanically operated self-balancing potentiometers and are much used both in laboratories and in industry for the continuous recording of temperature. In order to measure small temperature changes, an adjustable 'backing-off' emf is applied to oppose most of the thermocouple emf. The recorder can then be operated on a more sensitive range, allowing an expanded scale to be used over a narrow temperature range. In theory at least, with

sensitive modern recorders it is possible to record changes of a fraction of a degree at temperatures of 1000°C or more in this way. In practice, however, it is difficult to find a sufficiently stable backing-off emf source to be able to do this with any degree of accuracy for more than a short period of time.

Most laboratory thermocouple emf measurements today are made using a digital voltmeter (DVM) - really a convenient form of automatic potentiometer. Many DVMs are capable of 0.1 V resolution with an accuracy of a few microvolts over a period of twelve months or so (under laboratory conditions). Much depends on the quality of the internal voltage reference, and this should periodically be checked, for example with a standard cell.

Digital voltmeters have largely replaced the traditional potentiometers, though these are still useful as calibration sources for checking digital voltmeters - throughout their range, not just at a specific point.

In any precise measurement of small D.C. voltages the possibility of stray thermal emfs must be considered. These are best guarded against by shielding all connections, terminals etc, from draughts, and by working in an environment held at a uniform steady temperature. A precision potentiometer set-up always included a multi-pole reversing switch which enabled all connections to the instrument to be reversed. A mean was then taken of the normal and reversed readings for each measurement, so as to cancel out the effect of stray thermal emfs. It is good practice also to reverse the inputs to a DVM, though in this case the instrument must be calibrated to read positive and negative emfs.

Reference junctions

There are four approaches to treating the reference junction (or junctions):-

1. It may be maintained at 0°C.

2. It may be maintained at some other fixed temperature.

3. Its temperature may be allowed to vary, some compensation being built into the measuring circuit to offset this variation.

4. As in 3, but with compensation being affected mechanically in the indicator.

Maintenance of the reference junction at 0°C is the best method for precise work. For laboratory purposes an ice-water bath is reasonably convenient in use - and has the considerable advantage that thermocouple reference tables (and most calibrations) have 0°C as their reference point. With care the ice-point can be realised to 1 mK - a degree of precision that is ample so far as thermocouples are concerned. For most thermocouple work crushed tap-water ice is quite suitable. The junctions are inserted into glass tubes which are then immersed in the crushed-ice/water mixture contained in a wide-mouth Dewar vessel. The depth of immersion needed for the junctions will obviously vary with the material and diameter of wire used. As an example we have found that an immersion of 15 cm is satisfactory for 0.5 mm diameter platinum/rhodium wires joined to 0.5 mm copper wires.

Some care needs to be taken with the ice-bath if errors of up to several degrees are to be avoided. One source of error is the 4°C error possible if enough ice melts for the junctions to be immersed in water covered by a layer of floating ice. Another is the error caused by the use of ice which has been stored in a freezer and may initially be at a temperature considerably below 0°C. These conditions may be avoided by checking from time to time that the bath actually consists of a well-packed ice- water mixture.

Automatic ice-point devices are more convenient to use than ice/water mixtures, and they are readily available They make use of semiconductor thermo-electric (Peltier) cooling devices - using the large expansion of water on freezing to actuate the control system. They can operate very effectively, giving temperature errors of less than 0.1°C but some care needs to be taken with thermal loading and immersion errors as they do not have the thermal capacity or uniformity of an ice-bath.

With any reference junction operating below ambient temperature the possibility of condensation on the thermocouple wires leading to a wet junction, and hence the possibility of a galvanic cell being set-up, must be guarded against.

As an alternative to maintaining the reference junction at 0°C it may instead be held at a temperature somewhat above the highest likely ambient temperature. This method is very suitable where large numbers of thermocouples are being used, as it is easier and cheaper to make an enclosure of fair size operating at say 50°C than a cooler of the same capacity. The temperature uniformity and control within the enclosure ('zone box' in US terminology) must obviously be such that the overall accuracy of the whole system is not prejudiced. When using reference junctions held above 0°C in this way with reference tables or instruments referred to 0°C, the emf corresponding to the reference temperature must be added to the thermocouple reading.

With cold-junction compensation the temperature of the reference junction (typically the terminals of the measuring instrument) is allowed to vary, the variation being sensed by a suitable device (e.g. thermistor) situated as near as possible to the junction. By means of suitable circuitry an emf is produced which varies with temperature in such a way as to compensate for changes in the temperature of the reference junction. This system can be made to work well over the normal range of ambient temperatures and is widely used in electronic temperature controllers and indicators where an accuracy of a few °C is adequate. Electronic reference junction compensators are also available as discrete modules, either mains or battery powered. This is not very well suited to systems using large numbers of thermocouples as separate sensing and compensation is really needed for each junction.

Mechanical compensation is used in the simple 'furnace pyrometer' type of instrument where the current produced by the thermocouple emf is used to drive a galvanometer indicator directly. The reference junction here is at the terminals of the instrument, and compensation is applied by means of a bimetallic element acting directly on the hairspring of the indicator. Hence, with no input, such an instrument will indicate the ambient temperature. This is not a particularly accurate method, but is adequate for the precision of the total system.

Extension and compensating cables

Extension and compensating cables incorporate wires having roughly the same (net) thermoelectric properties as the corresponding thermocouple. Extension cable is made from nominally the same materials as the thermocouple itself, while compensating cable is made from a different pair of alloys. They are only required to match the emf/temperature characteristic of the thermocouple wire over a limited temperature range, normally no wider than -40°C to 200°C. Thus Type K extension cable is made from nickel-chromium and nickel-aluminium (possibly slightly sub-standard thermocouple material), while compensating cable for platinum-rhodium thermocouples is usually a copper/copper- nickel or copper-nickel/copper-nickel combination.

Extension and compensating cables are used in industry for several reasons:

1. Cost saving. Where the thermocouple wire is expensive, as in the case of platinum metal couples, or of heavy gauge as with base metal couples in large furnaces, it would be very expensive and wasteful to run them all the way to a remote indicator. Instead, there would be a junction box close to the furnace - probably at the thermocouple head - where the transition from thermocouple wire to extension or compensating cable takes place.

2. Mechanical properties. Extension and compensating cables are usually made of relatively light multi-stranded wires. This is cheaper, more flexible, and less likely to fracture than if larger diameter single strand thermocouple wires are carried for any distance.

3. In the case of the simple 'furnace pyrometer' systems the extension cable can have an important part to play in determining the total electrical resistance of the circuit – a matter of some importance when a current indicator is used to measure emf.

Manufacturing tolerances for extension and compensating cable are given in IEC 584-3 and BS 4937 Part 30, to which reference should be made for the current specifications. These give the emf tolerances which apply when extension or compensating cable for the various standardised thermocouples is included in the circuit, in addition to the tolerance for the thermocouple itself. The standard also specifies the temperature range over which the cable is expected to meet the tolerance.

It has long been the practice to use colour codes for extension and compensating cables and other thermocouple accessories (such as plugs and sockets) so that they may easily be distinguished and identified. Unfortunately the codes adopted in the past in various countries were different, but IEC has recently adopted a completely new colour code for all standardised thermocouples in IEC 584-3 with (at the time of writing) the exception of Type N. This has also been adopted in BS 4937 and is strongly recommended for national and international use.

Linearisation

Except over very narrow ranges, the emf/temperature relationships for thermocouples are, to varying degrees, non-linear. For example for Type K thermocouples over the range 0 to 1372°C the relationship is expressed by a 9-term power series expansion plus an exponential term, with constants expressed to eleven significant figures. For Type R, to cover the range -50°C to 1768.1°C three separate polynomial functions are used, ranging from fourth order to ninth order.

When a moving-coil galvanometer indicator is used linearisation may easily be effected by the use of a non-linear scale, sometimes used in conjunction with a meter specially designed to have a non- linear characteristic. However, with the increasing use of microprocessors and electronic digital indicators, better methods of linearisation have become available. A commonly used technique replaces the continuous but non-linear variation of emf with temperature from the thermocouple with a set of linear segments, continuous at the break points but varying in slope between them. The accuracy of this form of linearisation depends mainly on the number of segments used. With digital instruments this can be large, and instruments with characteristics very close to those given by the thermocouple reference tables can be produced. Modifying such an instrument to work with a different type of thermocouple involves mainly the replacement of the ROM containing the sequence of segments. Alternatively, linearisation may be achieved with a continuous representation of the thermocouple characteristic.

In analogue circuitry a number of non-linear circuit elements are available which allow logarithms, exponentials, powers and roots of signals to be obtained. A suitable combination of these elements with linear elements enables a continuous representation of the temperature to be derived. It is difficult however, to achieve an overall accuracy much better than 0.2 - 0.5% over an appreciable temperature range by this method. Some modern instruments employ continuous linearisation after A- to-D conversion, with a functional representation of the thermocouple characteristic whose coefficients are stored in a ROM and evaluated by a microprocessor. A high conversion accuracy can be achieved. This method also avoids the discontinuities in slope associated with the linear segment technique and is simpler to implement if the instrument design already includes a microprocessor for other purposes.

Finally, mention should be made of temperature transmitters. These are electronic units which convert the millivolt signal from a thermocouple (or other electrical transducer) to a low impedance source current signal, usually 4 to 20 mA. This signal is of course much less sensitive to electrical interference when transmitted over long leads in an electrically noisy industrial environment than is the original thermocouple signal. These devices are usually used in large industrial plants, in conjunction with controllers and indicators designed to operate with a standard current input signal. Cold-junction compensation is usually incorporated in the transmitter or in the mounting rack for a group of transmitters. Linearisation however, is not usually incorporated in thermocouple transmitters.

3.5.5 Calibration

Thermocouple wires as supplied are usually guaranteed to give an output within certain limits of that quoted in a standard reference table. For example Types R and S are usually guaranteed to ±1°C at

the gold melting point (1064.18°C), whereas Type K might be supplied to limits of ±4°C or ±7.5°C at 1000°C (Class 1 or 2 of IEC 584-2 and BS 4937 Part 20). Where greater accuracy is required it is sometimes possible to purchase wire from the manufacturer to tighter tolerance as a special order. Usually however the wire or the complete thermocouple will have to be calibrated. By calibration we mean the measurement of emf against temperature as defined by the ITS-90. This may be done at a few fixed temperatures (e.g. gold and palladium melting-points for a platinum thermocouple) or over a larger number of points covering the entire working range of the thermocouple. In either case a curve may be fitted through the calibration points and used to construct a table of differences or corrections from the standard reference table, or a full calibration table for the thermocouple may be produced.

Calibration is usually best carried out by comparison against a standard thermometer whose calibration is traceable to the ITS-90. The standard thermometer could be another thermocouple, a platinum resistance thermometer, or at high temperatures a radiation pyrometer. The calibrating environment used will depend upon the temperature region. Over the range -150°C to +200°C stirred liquid baths are usually used. From about 180°C to 550°C salt baths may be used, containing a mixture of sodium and potassium nitrates and nitrites. For the most precise work over this whole range of -150°C to +550°C a platinum resistance thermometer would be used as the standard thermometer. For less precise work secondary standard platinum resistance thermometers or a standard thermocouple would be used. Over the range 100°C to 600°C NPL for many years used silver-palladium thermocouples as the standard. These are excellent reference thermocouples, being very stable over long periods, although they have the disadvantage of having no standard reference table.

Above about 600°C calibration is best carried out in a horizontal tube furnace of suitable length for the gauge of thermocouple wire under test. The reference thermometer could be a platinum-rhodium thermocouple attached to the junction of the test couple by binding with thin platinum wire. The furnace used would probably need a metal liner to reduce temperature gradients - nickel being a suitable material for this purpose up to about 1100°C. Accuracies of 1 or 2°C may be achieved with care.

For more precise calibration above 1100°C one must use radiation thermometry. The thermocouple under calibration is inserted into a blackbody cavity contained in a furnace capable of providing minimal temperature gradients over the working zone. The temperature is measured using a suitable calibrated radiation thermometer.

As an alternative to calibration by comparison methods, platinum-rhodium thermocouples may be calibrated by fixed point techniques. For the highest accuracy attainable with these couples (about ±0.3°C) at temperatures up to about 1100°C, calibration would be made at the freezing points of zinc, silver and gold or copper (419.527°C, 961.78°C, 1064.18°C and 1084.62°C, respectively), using substantial ingots of the metals.

For calibrations to higher temperatures the so-called wire-point method may be used, with some sacrifice of accuracy. In this method a small piece of wire of a metal whose melting-point is accurately known (usually gold or palladium, 1064.18°C and 1553.5°C, respectively) is used to bridge the two wires of the thermocouple under calibration. This is then placed at the centre of a small vertical tube furnace which contains a second platinum-rhodium thermocouple which is used to monitor the furnace. The temperature of the furnace is allowed to rise slowly and alternate readings are made of the two thermocouples. At the moment when the bridging wire begins to melt the emf of the test couple becomes static for a minute or two and then falls to zero as the circuit breaks, whilst the output of the monitor couple continues to rise at a uniform rate. It is quite possible by this simple and inexpensive technique to achieve a reproducibility of ±0.5°C at the gold point and ±1°C at the palladium point.

In all calibrations it is important that sufficient immersion of the thermocouple is obtained, so that one can be sure that the junction region of the couple is at the true temperature of the calibration environment, be it bath or furnace. This is obviously more difficult the thicker the wire being calibrated - and for this reason also sheathed industrial-type thermocouples usually have to be dismantled and removed from the sheath before calibration. The larger sizes of mineral insulated thermocouple can be a problem to calibrate for this reason, as obviously they cannot be dismantled.

Cold junctions for calibration purposes are usually ice-water mixtures as described earlier in this paper. If the thermocouple is used with an extension cable this may be treated as a separate thermocouple to be calibrated over a limited temperature range (e.g. 0 to 50°C). However, it is more usual (and essential for integral MI thermocouples) to calibrate the thermocouple and extension cable together. It is increasingly common also to calibrate thermocouples with an associated indicator or digital thermometer. Since these will include some form of cold-junction compensation, it is essential to test that the effect of changes in ambient temperature is acceptably small.

As an alternative to laboratory calibration as described above it is sometimes more practical, especially in the case of large furnaces, to carry out an in-situ calibration. This is probably the best method of checking the calibration of a thermocouple that has been in use for some time, as the thermocouple is unlikely to be homogeneous - see the note below. The best method of in-situ calibration is to make sure that a calibrated reference thermocouple can be inserted into the furnace alongside the couple to be checked. Readings are then compared over the working range of temperatures and the reference couple removed.

The level of accuracy one can hope to achieve in the calibration of thermocouples varies widely according to the type of couple, the temperature range, the equipment available and the skill of the user. Platinum thermocouple Types R, S and B may be calibrated to ±0.3°C in the range up to 1100°C; and ±2°C up to 1553.5°C (the palladium point). Base metal types may be calibrated to ±0.1°C over a limited temperature range, say -80°C to +100°C, and to a few degrees at temperatures up to 1000°C, for types K and N. However, because of the inherent instability of base metal couples at high temperatures, it is not normally worth while to attempt, or pay for, accurate calibrations of these couples at temperatures anywhere near to their upper working limit. If accurate measurement of temperature is required for prolonged periods above the 600 to 800°C region it is usually more satisfactory to use platinum-rhodium thermocouples.

Calibration by the methods described may be carried out by the user, but where he does not have the facilities required, calibration services are available from both the NPL and laboratories accredited by UKAS, the United Kingdom Accreditation Service.

Any thermocouple which is to be calibrated must be homogeneous. New thermocouple wire from a reputable manufacturer should be satisfactory in this respect, but any thermocouple that has been in use for a long period at high temperature must be suspect. Platinum-rhodium thermocouples can suffer from the effects of rhodium migration and evaporation, although the effects are usually small. They may also have become contaminated. Base-metal thermocouples can suffer large changes due to oxidation at high temperatures. Any non-homogeneous thermocouples, if calibrated, and then used under conditions of temperature gradient different from those present during calibration will give false readings - possibly even tens of degrees in error. For this reason, the only reliable method of calibrating base metal couples that have seen high-temperature service is the in-situ calibration method described earlier. In general, precise calibration of base-metal couples for high temperature use is to be discouraged.

Thermocouple reference tables and tolerances

In order to facilitate the practical use of thermocouples as interchangeable units they are made to conform to a given emf/temperature relationship. This is usually expressed as a power series polynomial from which tabulated values of emf against temperature (or vice versa) are obtained. The standard tables give emfs resolved to 1 V at 1°C temperature intervals - and the converse relationship at corresponding intervals.

In practice, no thermocouple can be made to conform exactly to the published table, and they are therefore supplied to an agreed limit of tolerance. Thermocouple tolerances are published in IEC 584- 2 and BS 4937 Part 20. Other types of thermocouple are usually supplied to manufacturers (batch) tables.

The thermocouple tables are based upon a reference junction temperature of 32°F (0°C); therefore, direct conversion from the tables can be made only when an ice bath is used at the reference junction.

If it is not possible to maintain the reference junction temperature at 32°F (0°C), a correction factor must be applied to the millivolt values shown in the thermocouple tables. Note that the millivoltage produced by a given thermocouple is decreased when the temperature difference between the measuring junction and the reference junction is decreased. Correcting for reference junction temperatures other than 32°F (0°C) is described below.

Converting Millivoltage to Temperature

To apply the reference junction correction factor to a given potentiometer millivoltage reading, proceed as follows:

1. From the appropriate thermocouple table, obtain the millivoltage (based upon a 32°F R/J) corresponding to the actual temperature of the thermocouple reference junction.

2. Add algebraically the value obtained in step 1, above, to the millivoltage read on the potentiometer.

3. The corrected millivoltage may then be converted into terms of temperature directly from the same table.

Converting Temperature to Millivoltage

To determine the proper millivolt input required to check the calibration of an instrument, proceed as follows:

1. From the appropriate table, obtain the millivoltage based upon a 32°F reference junction corresponding to the actual temperature at the input terminals of the instrument to be checked.

2. From the same table, obtain the millivoltage based upon a 32°F reference junction for the temperature to be checked.

3. Subtract algebraically the value obtained in step 1, above, from the value obtained in step 2.

ADVANTAGES AND LIMITATIONS OF THERMOCOUPLES

Thermocouples are low-cost, small size, rugged, convenient, and versatile (can be welded to a pipe), wide range, reasonably stable, reproducible, accurate, and fast. The emf they generate is independent of wire length and diameter. While RTDs are more accurate and more stable and while thermistors are more sensitive, thermocouples are the most economical sensors and they can detect the highest temperatures.

The main disadvantage of the thermocouple is its weak output signal. This makes it sensitive to electrical noise and limits its use to relatively wide spans (usually the minimum transmitter span is one mV). It is non-linear, and the conversion of the emf generated into temperature is not as simple as in direct reading devices. Thermocouples always require amplifiers, and their calibration can change due to contamination, composition changes due to internal oxidation, coldworking or temperature gradients. Thermocouples cannot be used bare in conductive fluids, and if their wires are not homogeneous, this can cause emfs that are difficult to detect.

In general, one should use the largest size thermocouple wire possible, avoid stress and vibration, use integral transmitters whenever possible (and otherwise use twisted and shielded wires with the shield connected to the guard of the integrating analogue-to-digital converter), avoid steep temperature gradients, and be careful in the selection of sheath and thermowell materials.

Table 3.5 Advantages and limitations of thermocouples

Note: additional information concerning the use of thermocouples can be found in the Appendix section of this guide:

Appendix F: Tolerance classes for thermocouples (IEC 584-2: 1982) Appendix G: Thermocouple reference tables (IEC 584-1:1995)

3.6 RADIATION THERMOMETERS OR PYROMETERS

Figure 3.9 Illustration of the radiation thermometer or pyrometer

3.6.1 Overview

Radiation thermometers, or pyrometers, make use of the fact that all objects emit thermal radiation, as seen when looking at the bars of an electric fire or a light bulb, for example. The amount of radiation emitted can be measured and related to temperature using the Planck law of radiation. Temperatures can be measured remotely using this technique, with the sensor situated some distance away from the object. Hence it is useful for objects that are very hot, moving or in hazardous environments.

All objects emit radiation in the infrared and visible parts of the spectrum, the intensity of which varies strongly with temperature. The fundamental relationship governing thermally emitted radiation is the Planck law. This relates the intensity of the radiation from a perfect radiator (or blackbody) to the temperature and wavelength:-

-5 -1 L d = (c1/π) [exp(c2/ T) - 1] d Planck's law

-5 L d ≈ (c1/π) exp(-c2/ T) d Wien's approximation

E = ( /π) T4 Stefan’s law

In the above L is the spectral radiance (the intensity of thermal radiation at wavelength ), T is the temperature of the source in Kelvin, E is the total radiant energy, c1 and c2 are known as the first and second radiation constants, respectively, and is the Stefan-Boltzmann constant. The values of the constants are:-

2π5k4/15c2h3 = 5.67051 x 10-8 W m-2 K-4 2 -16 2 c1 2πhc = 3.7418 x 10 W m -2 c2 hc/k = 1.4388 x 10 m K

where k, c and h are the Boltzmann constant, the speed of light, and Planck's constant, respectively.

For real surfaces the intensity is reduced by the emissivity factor, which is between 0 for a perfect reflector and 1 for a blackbody.

How does a radiation thermometer work?

Radiation thermometers gather and focus the thermal radiation onto a detector. Semiconductor detectors are usually used: the most common are silicon, lead sulphide, indium antimonide or indium gallium arsenide. The output can be easily digitised and continuously monitored.

They measure the thermal energy emitted by a source and relate this to its temperature by means of the Planck law of radiation. They consist of optics (generally lenses) to collect and focus the emitted energy onto a detector. The signal from the detector can either be measured directly, or it can be converted to a temperature using a system of electronics. Filters are usually used to define the wavelength or wavelength band over which the emitted energy is measured.

Radiation thermometry has long been used for measuring temperatures in industrial and manufacturing processes, and it is now being increasingly used to make measurements at more ordinary temperatures. However their apparent ease of use masks the equal ease with which they can be used incorrectly. Field-of-view effects, reflections, atmospheric absorption and unknown emissivities are all potential problems.

Many types of radiation thermometer are available for different applications. For measuring high temperatures a thermometer should be chosen that operates at a short wavelength, where the rate of change of emitted radiation with temperature is very high. However, for low temperature applications where the amount of emitted radiation is low, a broad-band device operating at longer wavelengths is required. NPL can calibrate infrared thermometers between -40°C to 3000°C.

What is emissivity and why is it important?

All objects at temperatures above absolute zero emit thermal radiation. However, for any particular wavelength and temperature the amount of thermal radiation emitted depends on the emissivity of the object's surface. Emissivity is defined as the ratio of the energy radiated from a material's surface to that radiated from a blackbody (a perfect emitter) at the same temperature and wavelength and under the same viewing conditions. It is a dimensionless number between 0 (for a perfect reflector) and 1 (for a perfect emitter). The emissivity of a surface depends not only on the material but also on the nature of the surface. For example, a clean and polished metal surface will have a low emissivity, whereas a roughened and oxidised metal surface will have a high emissivity. The emissivity also depends on the temperature of the surface as well as wavelength and angle.

Knowledge of surface emissivity is important both for accurate non-contact temperature measurement and for heat transfer calculations. Radiation thermometers detect the thermal radiation emitted by a surface. They are generally calibrated using blackbody reference sources that have an emissivity as close to 1 as makes no practical difference. When viewing 'real' more reflective surfaces, with a lower emissivity, less radiation will be received by the thermometer than from a blackbody at the same temperature and so the surface will appear colder than it is unless the thermometer reading is adjusted to take into account the material surface emissivity. Unfortunately, because the emissivity of a material surface depends on many chemical and physical properties it is often difficult to estimate. It must either be measured or modified in some way, for example by coating the surface with high emissivity black paint, to provide a known emissivity value.

What is a blackbody source and what is it used for?

A blackbody source is an ideal, 'Planckian', radiator, i.e. it emits thermal (visible and infrared) energy whose intensity at a given wavelength and temperature is given by the Planck Law of radiation. Blackbody sources, whose temperatures are known or can be measured, are therefore extremely useful standards for the calibration of radiation thermometers.

An ideal blackbody source is a completely enclosed cavity held at a uniform temperature. The radiation field inside the cavity depends only on the temperature, and not on any physical property (size, shape, construction material). It completely absorbs and emits all radiation and has an emissivity of 1. For practical purposes, in order to view the radiation field inside the cavity, it is

necessary to introduce a hole or aperture. Since this means a departure from the 'ideal' situation it is not possible to have a practical blackbody cavity with an emissivity of 1. However, by careful design cavities can be made with emissivities that closely approach 1.

3.6.2 Optical systems

In any radiation thermometer the optical system must be designed so that the radiation emitted from the source is collected as efficiently as possible by the detector. This means that interference from radiating sources outside the nominal field of view must be eliminated or a means found to compensate for the effect of the source of interference. An optical filter will normally be necessary to isolate the required waveband.

Various optical systems are employed, each with its own advantages and disadvantages, and these are now briefly described.

Aperture optics

Radiation thermometers for use at high temperatures, or for the measurement of the mean temperature of large areas where the radiance is high, may use aperture optics. The basic principle is that all radiation incident on the detector arises from some defined area of the source.

Lens optics

A lens is used to improve the radiation collection efficiency so that a smaller target area can be measured with the same detector system. The sensitivity of the radiation thermometer and the field of view depend on the area of the aperture stop and the solid angle subtended to it by the field stop. When calculating the minimum target area the effects of chromatic and spherical aberrations in the lens must be considered. Glass lenses are normally suitable for wavelengths out to 1 m, but for temperatures below about 700°C where longer wavelengths are required, their transmissions may not be adequate, and silica or other materials must be used. It is essential to keep all optical surfaces free from dust and films.

Mirror optics

Mirror optics using single or Cassegrain systems are often used in order to avoid the problems of aberrations in lenses and the increasingly poor transmission of glass at long wavelengths. They are well suited for broad-band or 'total' radiation thermometers. In dirty environments special windows may be needed to protect the mirror surfaces from deposits.

Fibre optics

All the systems described so far require a direct unimpaired view of the object under measurement, but there are many situations where this is either a limitation or is impractical. Fibre optics can then be advantageously used to transmit the radiation to a remote detector. Thus objects in hostile environments, in process plant or in aeroengines, for example, can be measured by focusing the radiation onto a fibre while the detector is kept at a convenient location. Direct line-of-site is no longer a requirement and the optical head can be built to withstand temperatures of 200 to 400°C, much higher than those at which detectors can operate. If these temperatures would otherwise be exceeded, air or water cooling may be employed. Small low temperature targets can also be measured by placing the head of a fibre close to or in contact with the target.

Sighting

For successful operation it is necessary to ensure that the optical system is correctly aligned and focused on the target. A direct optical sighting system is normally provided if the lens or window is transparent in some part of the visible spectrum. Otherwise an auxiliary telescope or sight may have to be used. Radiation thermometers designed for use in fixed positions can normally be correctly aligned once and for all, without sighting systems.

3.6.3 Selecting a radiation thermometer

In selecting the appropriate type of radiation pyrometer to be used in detecting the temperature of a particular process, one would prefer to select a wavelength-band in which the transmittance (T) is near zero (the material is opaque). In addition, one would like to select a wavelength-band that will not be absorbed by the atmosphere the radiation has to pass through. These two considerations result in a requirement for matching the pyrometer wavelength to the particular process.

Because of the different needs of the various processes, each radiation pyrometer manufacturer has developed a range of products operating in a variety of wavelength-bands.

Short wavelength units can be used in iron and steel processing. Long or broad wavelength pyrometers are useful at lower temperatures where short wavelength radiation can be undetectable. Materials like glass are transparent to shorter wavelengths and therefore short-wave pyrometers would not detect the surface temperature, but would measure the temperature below the surface. In some processes like lamp sealing this is actually desirable, and therefore the 3 to 4 micron band is used.

As stated earlier, the rate of change of intensity with temperature is greater at shorter wavelengths. For the most precise temperature measurement a monochromatic thermometer operating at as short a wavelength as is practicable would be recommended. The optimum spectral range of operation therefore principally depends on the temperature range of interest, and the detector must be selected to suit this.

Another factor to consider when selecting the spectral region is the absorption of radiation of certain wavelengths by the atmosphere between the target and the instrument. A normal atmosphere contains a fairly small but constant amount of carbon dioxide, and a variable amount of water vapour. To avoid variations particularly with humidity and target distance, these wavelengths should be avoided. Typical wavelengths and wavebands used are 0.65 to 1.0 m, 3 to 5 m and 8 to 14 m.

Apart from the above, perhaps the factor most affecting the choice of radiation thermometer is how it can cope with uncertainties or variations in emissivity. Serious errors may result if an incorrect value is assumed, but as has been discussed, emissivity varies with temperature, wavelength and surface condition, and the correct value may not be easily determined. The selection of the 'best' radiation thermometer for any particular application may be difficult, but there are certain guidelines which can help.

One of the most important parameters to consider when selecting a radiation thermometer for any particular application is that of target size. It is also one of the most confusing, since there is not usually a sharp cut-off to the field of view of a thermometer. The field of view (or target size at given distances) is defined in different ways by different manufacturers, and in some cases the instrument may respond significantly to radiation outside the nominal target area. Therefore before purchasing a particular instrument one should make certain that it has a field of view suitable for the application in mind.

If the measurement of differences or fluctuations of temperature is required, rather than the actual temperature of the object, or if repeatability is more important than absolute values, the emissivity problem is greatly alleviated. One only requires the emissivity to remain constant and not vary strongly with temperature or wavelength, but the actual value need not be known.

If the actual temperature must be measured then it should be remembered that most radiation thermometers are calibrated in terms of blackbody radiation. Most modern instruments provide adjustment to take the emissivity into account, but it is better if the object can be measured under conditions of high emittance. This can be achieved by selecting a radiation thermometer that operates in a spectral region where the emissivity of the object is intrinsically high or otherwise by increasing the apparent emissivity of the object, for example by coating it with special matt black paint, oxidising it, or increasing the effective emissivity of the surface by placing a reflector so that it is surrounded as far as possible by its own radiation field. It may also be possible to sight onto a cavity in the material, which will have the effect of increasing the effective emissivity.

Finally the radiation thermometer must be able to sight within the total area of the target. Large errors will result if the target does not fill the field of view of the instrument.

Various types of radiation thermometer are now briefly described.

The disappearing filament pyrometer

The disappearing filament (visual) pyrometer is still sometimes used for high temperature measurement where there is enough visible radiation. The eye is a very sensitive comparator of adjacent radiating sources and in this type of instrument the centre of a filament in a reference lamp is matched with an image of the source. The temperature and brightness of the filament is varied by changing the current through it until at balance the filament 'disappears' against the background. The current is measured potentiometrically to produce a reading, directly in terms of temperature. The matching is performed within a narrow, about 30 nm wide, wavelength range centred in the visible red, on 650 nm.

Temperatures in the range 700 to 3000°C can be measured and it is possible to attain reproducibilities of about 1°C at source temperatures around 1000°C with an uncertainty of about ±5°C. Targets as small as 0.01 mm can be measured with some instruments and the calibration can be checked against calibrated standard tungsten-ribbon lamps. All optical surfaces must be kept clean, particularly the front or objective lens. The operator should make a sequence of brightness matches for each measurement approaching the balance point, alternately from higher and lower temperatures. The mean value of these observations should be taken. As the eye response of each person differs, the calibration should ideally be performed by the normal operator of the instrument. Working at short wavelengths, errors due to emissivity are reduced and the effects of atmospheric absorption can be ignored. Fumes and dust in the optical path should be avoided. However, using the eye as a detector does require a subjective decision on the part of the operator and so this type of radiation thermometer has largely been superseded by direct reading and automatic instruments.

Infra-red thermometers

The application of radiation thermometry has been enormously increased by the availability of good quality, not-too-expensive and convenient-to-use infrared instruments, and infrared thermometers using silicon photodiodes or pyroelectric detectors are now almost always preferred over visual instruments. A silicon photodiode is generally suitable for temperatures between 600 and 3000°C, and will operate at wavelengths between 0.6 m and 1 m. It has a very fast response and is suitable for the control of automatic processes where measurement of temperature variation is required. Uncertainties (relative to a blackbody source) may be 1°C or 0.1% of temperature. Again, dirt and dust in the optical path should be avoided, and target size and distance will be important. Below 600°C pyroelectric or photodetectors, such as InGaAS, InSb or HgCdTe, are available for operation at longer wavelengths.

Narrow-band spectral radiation thermometers are available for specific applications such as measuring the temperature of glass and plastics. By appropriate selection of the wavelength either the surface or the interior temperature may be measured. Operating ranges and uncertainties vary enormously.

Ratio or two-colour thermometers

If the emissivity of a surface is the same at two different wavelengths this parameter can be eliminated, with some sacrifice in sensitivity, by taking the ratio of the signals at each wavelength. This is the principle on which 'two-colour' or 'ratio' thermometers operate. In cases where the ratio 1/ 2 is known and remains constant, a correction may be applied, and some two-colour thermometers have an adjustment control for this purpose. Since the sensitivity is less than for a single-wavelength instrument (because the signals in the two channels both increase with temperature, their ratio must increase more slowly), the accuracy required for the emissivity ratio is correspondingly greater than for each value independently. Unfortunately good emissivity data is often not available and, as has been mentioned, is specific to particular surfaces. However, the

technique allows some compensation for absorption in the optical path by smoke or dust, although this also tends to be wavelength dependent. A real advantage is that if the source is smaller than the target area of the pyrometer, a reading of reasonable accuracy is still possible as the signal in each channel is reduced by a similar amount, provided that the source is much hotter than its surroundings. The technique can also be advantageous in situations where the emissivity is likely to change, for example with aluminium surfaces.

In general the accuracies obtained using two-colour and more complicated multiwavelength instruments, are difficult to assess because of the uncertainties arising from the emissivity, and they should only be used with caution.

Thermal imaging

Thermal imaging systems are specialised forms of radiation thermometer. A complete picture of the target is built up by a 'camera' using either an array of detectors or by scanning as in a television picture. Such systems are very useful for detecting 'hot spots' in electrical circuits or installations, heat losses from buildings, corrosion in pipes and boilers etc., and also in medical diagnostics and night vision. Although they are primarily used for displaying temperature differences, actual temperatures are shown on the display in some instruments. Calibration normally requires a source of known temperature, preferably a blackbody, to be included in part of the display.

Total radiation thermometers

The measurement of low temperatures by radiation thermometry (near or even below ambient temperatures) is complicated by the difficulty of distinguishing the signal from background radiation and the need to cool photodetectors. Thermal detectors such as thermopiles, in which a temperature difference is set up by the incoming radiation, are commonly used and can resolve 0.1°C. However, they are generally delicate and slow. In pyroelectric detectors the temperature rise is converted to an electric charge, which when rapidly chopped gives a signal in the form of an alternative voltage.

To work efficiently the detector surface is coated with a black paint, and the response is then almost flat over the 'total' wavelength range. Instruments may be unfiltered, but in practice a protective window is often included and this restricts the spectral response. Mirror optics, or simple aperture systems, are generally preferred, and the required target size is usually comparatively large so as to collect enough radiation and increase the sensitivity. Emissivity corrections need to be applied, using values appropriate to the wavelength response of the instrument.

The broad-band pyrometer is used generally in industry for readout and automatic control. It can cover wide temperature ranges and is the least expensive of the three types. Narrow-band and two- colour pyrometers are used, where necessary, to minimise the emissivity effects and for special applications where it is desirable to select the particular band pass.

The prospective user of a radiation pyrometer should consider the following points during selection:-

1. Target temperature, low, normal and high limits. 2. Minimum target size and distance factors. 3. Target material and emittance. 4. Angle of observation. 5. Is target stationary or moving? If moving, will the speed of response of the pyrometer be fast enough? 6. Atmospheric conditions between target and detector. 7. Ambient temperature. 8. Can pyrometer sight directly on target or must it sight through a sealed auxiliary window such as required for vacuum or pressure? 9. Is scale to be read directly in temperature units or will an arbitrary reading be satisfactory?

In addition, in designing or choosing the radiation pyrometer for a particular application, consideration

must be given to the following characteristics:-

1. The field of view, or the size-distance relationship.

2. The transmission qualities of the collector system and any windows or filters in the optical path.

3. The band pass and sensitivity of the detector.

The physical shape of the optical system (lens or curved mirrors) and its mounting within the pyrometer housing control the sighting path, while the material from which it is made determines the optical properties. Glass does not transmit well beyond 2.5 microns and is suitable only for the higher temperatures where plenty of output is available. Other popular optical materials are quartz (fused silica) to 4 microns and crystalline calcium fluoride out to about 10 microns. Lesser used (and more expensive) materials will increase the transmission even more.

Windows and filters in front of or behind the optical system can alter the transmission properties greatly. A plate glass window in front of a calcium fluoride lens, for instance, will very effectively stop the longer wavelengths which would have passed through the lens.

A band pass filter might be purposely placed in front of the detector to cut off unwanted wavelengths.

3.6.4 Calibration

The calibration of a radiation thermometer requires the use of a reference source of known calibration. For disappearing filament pyrometers standard tungsten ribbon lamps are used, but for other types of radiation thermometer it is usually necessary to use a blackbody source, as this allows the radiation to be characterised at all wavelengths, using Planck’s law, from a measurement of its temperature.

Blackbody sources

Painting a surface black can be effective in increasing the emissivity, but it can also be deceptive since the infra-red emissivities of some paints are often markedly different from those in the visible. For a good reference blackbody source it is preferable to employ the fact that radiation inside a completely enclosed cavity is truly blackbody ('Planckian' - independent of the surface) provided that the temperature of the cavity is uniform. The radiation field is perturbed if a hole is made in the cavity for observation of the radiation, but this perturbation can be kept small by careful design, by blackening the interior and by using a cavity which is large compared with the size of aperture required. Good approximations to blackbody sources can be made and are available commercially.

A cylindrical cavity is more easily constructed and controlled than a spherical one. The ratio of cylindrical length to aperture diameter should be as large as practicable, and not less than 5. The higher the emissivity of the material used in the construction the better, and in some applications special black paints may be used to coat the surfaces. In others, such as those made from stainless steel for use at high temperatures, the surface should be oxidised. A conical or grooved rear surface together with roughened side walls or baffles will also increase the effective emissivity.

The best way of estimating the emissivity of a cavity is from first principles based upon the geometry of the enclosure and surface emissivity of the wall material. Checking its performance is best achieved by comparison with a reference cavity at more than one wavelength. However as this may involve the use of two calibrated radiation thermometers it may not be the most practical. It is also desirable to vary the target distance and aperture size. For a fixed cavity temperature there should be no change in the radiation thermometer reading provided that the target size/distance specifications are still met. If it is possible to reduce the size of the cavity aperture and still keep within the target size requirements of the radiation thermometer, no change in reading should be recorded for a fixed cavity temperature.

Making sure that a cavity is at a uniform temperature is important, but may not be easy. Immersion in a stirred liquid bath or circulating a temperature controlled liquid around the outside of the cavity are probably the best techniques for temperatures below 200°C. At higher temperatures furnaces are used. The provision of 'end-heaters' is recommended to supplement the main heater and to make up for end-losses. Measuring the temperature profile to ensure adequate uniformity can be problematical: moving a thermocouple along the cavity wall may not give a true indication of the gradients and a small resistance thermometer may be better in this respect.

As was mentioned above, the characterisation of a blackbody requires only that its temperature is known. It is usual to measure this with either resistance thermometers (up to about 500°C) or thermocouples (up to 1200°C and beyond). However there are several requirements which must be fulfilled if this is to give an adequate calibration of the cavity as a blackbody source:-

1. The radiation thermometer calibration must be accurate enough and the device stable enough for use in the temperature range and environment required.

2. It must be located so as to measure the temperature of the surface on which the radiation thermometer is sighted. Hence it should be immediately behind the surface opposite the cavity aperture. Even then it may be difficult to be sure that the probe registers the temperature of the radiating surface. Probes are sometimes placed within the cavity itself, but the line of sight must not be obscured and again the probe will not necessarily be at the required temperature.

3. Accurately measuring the radiating surface temperature will only lead to an accurate characterisation of the cavity if its temperature is uniform.

Since all these factors need to be assessed and investigated it is often preferable to calibrate a blackbody cavity as a complete system rather than to rely on the external calibration of the probe. Thus the radiance temperature would either be measured at an appropriate wavelength using a calibrated radiation thermometer, or compared with that from a reference cavity - the choice of which method to use depending on which standard is most conveniently available.

Blackbody sources are available from various manufacturers, but care should be taken that they are suitable for use with a specific radiation thermometer, for example with regard to target size, and that the calibration has been properly established.

Standard lamps

Tungsten-ribbon lamps are convenient standard sources of radiation for the calibration of radiation thermometers at short wavelengths. They are available in two classes, normal and high accuracy, each class being divided into two groups, evacuated ('vacuum') and gas-filled. They normally operate within the temperature range 700 to 1700°C for the vacuum types and 1600 to 2300°C for the gas- filled versions.

Since the tungsten ribbon is not a blackbody radiator, the calibration of a lamp is in terms of its apparent, or radiance temperature as a function of the current through the ribbon, rather than its actual temperature. The (spectral) radiance temperature is equal to the temperature of a blackbody which has the same radiance at the wavelength specified. The actual temperature is generally much higher, but one is only interested in how hot the ribbon appears to be.

All lamps must be tested and calibrated before they can be used. The calibration is usually made at an effective wavelength of 655 ± 10 nm. Lamps can be calibrated at other wavelengths provided that the transmission of the viewing window or envelope of the lamp is high enough and independent of wavelength. For small variations in the calibration wavelength in the region of 660 nm, a correction can be applied.

Although most lamps are of the tungsten-ribbon type as discussed, blackbody lamps are available in which the tungsten is formed into a tube containing at its centre a bundle of tungsten filaments. When viewed along the axis through an aperture of 1 mm diameter, resulting cavity has an emissivity of

about 0.992, for wavelengths in the range 400 nm to 800 nm. The radiance temperature is then much closer to the actual temperature, and the gas-filled blackbody lamp provides a useful source up to about 2650°C. However, because of alignment problems these lamps are only recommended for use at these high temperatures, where ribbon lamps are unsuitable.

The correct alignment of a lamp with respect to the radiation thermometer is very important, and should be as stated in the calibration certificate. A direct current (DC) supply of the correct polarity is essential as any reversal of the current flow through the filament will almost certainly invalidate the calibration. All lamps should be shielded from extraneous light sources which may result in reflections from the filament and from the effects of convection currents (e.g. due to air conditioners). The latter is most important for gas-filled lamps, and for vacuum lamps at low temperatures where the temperature of the base is an important factor.

3.6.5 Precautions necessary in the use of radiation thermometers

Some of the factors that can affect temperature measurements made using a radiation thermometer have already been mentioned. Apart from the temperature range required, the following factors should be taken into account when selecting a radiation thermometer for a particular application.

Emissivity

An emissivity control is included with most instruments to correct the measured temperature to its true value. However this assumes that the correct value of the emissivity is known.

Because the emissivity will change with temperature, composition and surface conditions, a range of values is given for each waveband. The appearance of the surface is not always a reliable guide to its emissivity as the slightest oxide film may affect the value.

Radiation reflected from neighbouring hot sources

Radiation thermometers respond to all radiation which reaches the detector within the range of its spectral response. Radiation reflected from the surface under view will add to the emitted radiation and hence lead to an overestimate of the temperature. This is a particular problem in furnaces where the target is being heated radiatively. It is then necessary to estimate the temperature of the surroundings (or make an ancillary measurement of it) and to make a correction to the radiation thermometer reading on the assumption that the furnace acts as a blackbody cavity.

Absorption by the intervening medium

An error will result in the temperature as measured by a radiation thermometer if there is any absorption or scattering of the radiation by the intervening medium. Fumes, smoke and spray usually result in fluctuating readings but other effects such as absorption by atmospheric carbon dioxide and water vapour are not always obvious and can produce serious errors especially at long pathlengths. Fortunately atmospheric absorption is confined to known spectral regions and by using detectors having spectral responses in one of the many atmospheric 'windows' (e.g. 8 to 14 m) this effect can largely be avoided.

If it is necessary to have an optical window between the source and the radiation thermometer (e.g. to maintain a controlled atmosphere, etc) this must be taken into account. The window transmission at the radiation thermometer wavelength must be measured, or the instrument must be calibrated together with the window. The presence of flames between the radiation thermometer and the source may also affect the reading due to emission by water and carbon dioxide in the flame. Non-luminous flame effects can be avoided by selecting detectors operating outside these emission bands. Luminous flames may make accurate readings very difficult to achieve.

Installation

The field of view of a radiation thermometer must be fully filled by radiation from the target, and the field stop or detector must be completely illuminated by radiation from the source. The field of view

may not be very well defined so it is advisable to ensure that the target size is always greater than the minimum specified. At low temperatures, particularly when using broad-band detectors, care must be taken to avoid stray radiation from the surroundings being reflected or diffracted into the optical system. Reflections from the object itself can be particularly troublesome, as discussed above.

Intrusions into the field of view especially when sighting through apertures or tubes must be avoided. The radiation thermometer should be sighted at right angles to the target as the specifications regarding target size are usually quoted for viewing at 90° and there may be considerable differences in the emissivity, especially at angles greater than 45° to the normal. All the optical surfaces must be clean. Any protecting windows must be kept free of deposits due, for example, to fumes or splashes of molten materials and care must be taken to avoid scratching the optical surfaces. Air purging can be used for this but the air must be clean and free from any oil or water. Only the minimum amount necessary for a clear view along the sight path and to obtain steady readings should be used, as too much air may cool the surface or cause oxidation. If the working temperature of the radiation thermometer itself is likely to exceed that specified, either by the manufacturer or in its calibration, water or air cooling can be used. Excessive cooling may cause water to condense on lenses or windows and must be avoided. Manufacturers will be able to supply suitable cooling jackets and give full instructions regarding their use.

Electrical connections between the radiation thermometer and any indicating or recording equipment must be made with suitable cables with precautions taken to avoid interference from other electrical sources by the use of screened or twisted pairs of leads.

Process control

In many situations, such as process control, the radiation thermometer reading is often sufficient for monitoring the temperature even if this is different from the true temperature. This will of course depend, in the short term, upon the constancy of the difference between the true and indicated temperatures from one series of measurements to the next. This will, in turn, depend upon the constancy of the surface emissivity, selective absorption of the radiation by fumes, water vapour etc. and other measurement conditions, some of which have been highlighted in this section.

ADVANTAGES AND LIMITATIONS OF RADIATION THERMOMETERS OR PYROMETERS

The major advantages of radiation thermometry arise from the absence of physical contact between the instrument and the system under study. The speed of response is therefore largely determined by that of the detector, and may even lie in the microsecond range. Hence it is possible to measure rapidly changing temperatures or the temperature of a moving object. Exposure of thermocouples or resistance thermometers to prolonged high temperatures often produces large changes in their calibration, with consequent uncertainty in the measurements. Radiation thermometers on the other hand are affected only by variations in the ambient temperature, and the long-term drift of their component parts at this temperature. If the surface emissivity is stable, so that it is known or can be measured, or if the source may be made to resemble a blackbody, radiation thermometers may achieve very high accuracy, e.g. ±1°C up to 1500°C or higher.

ADVANTAGES

1. Does not require physical contact with material whose temperature is being measured. 2. Fast speed of response - can be used on moving targets. 3. Can look at small targets (1/16 in. or 1.6 mm in diameter) or measure the average temperature over a wide area. 4. Measures much higher temperatures than thermocouples.

DISADVANTAGES

1. More fragile and costlier than thermocouples, RTDs, or thermistors. 2. Non-linear scale shape, approximating the 4th power of the temperature.

3. Emissivity of target may cause a low temperature reading if not corrected. 4. Relatively wide temperature span required.

Table 3.6 Advantages and limitations of radiation thermometers or pyrometers

Note: additional information concerning the use of radiation thermometers or pyrometers can be found in the Appendix section of this guide:

Appendix H: Data tables for radiation thermometers or pyrometers

3.7 BIMETALLIC THERMOMETERS

Bimetallic thermometers make use of two fundamental principles. (1) metals change volume with temperature and (2) this coefficient of change is not the same for all metals. If two different straight metal strips are bonded together and heated, the resultant strip will tend to bend toward the side of the metal with the lower expansion rate. Deflection is proportional to the square of the length and the temperature change, and inversely proportional to the thickness. A bimetallic spring can be calibrated to produce a predictable deflection at a preset temperature This is the basis of operation of the many bimetallic temperature switches in various household appliances.

The motion produced by a bimetallic spring is small; to amplify it in a reasonably sized space the bimetal strip may be wound in the form of a spiral or a helix.

Knowing the coefficients of expansion of the two metals, their thickness, and the desired scale length and range, the total length of the spiral can be computed. A favourite combination of metals is low- expanding invar (64% Fe, 36% Ni) against high-expanding nickel-iron alloy with chromium or manganese added.

Most industrial or residential bimetal thermometers use a helical coil which can be designed to fit into a stem more easily than the spiral. The element is surrounded by a protecting tube or well. The device can be mounted to measure the temperature of the gas or liquid inside a duct. The design is frequently used on domestic furnaces and over the years has replaced most of the glass stem thermometers used earlier. When only periodic checks are needed on a particular temperature point, a thermowell only is installed and is protected from dirt accumulation by a cap.

A single helix moves axially as it winds or unwinds with heat and cold. This requires clearance for a vertical movement of the pointer. The difficulty can be overcome, if desired, by using a multiple element, wound coaxially so as to form coils within coils. This construction is more costly but has an advantage in requiring less immersion depth.

The thermometer is usually either back- or bottom-connected, depending on which orientation allows the operator better visibility of the dial face. Bimetal thermometers are also made in types that can adjust the dial face at any angle, with respect to the axis of the stem - this can even be a swivel type. Any of these constructions requires a bend in the motion transmission from coil to pointer. This is done with an edgewound helical spring, which eliminates backlash and requires little torque to operate.

Another feature usually included is complete sealing. A dry gas is in the dial face portion of the assembly while silicone fluid fills the stem and surrounds the coil to dampen vibration and accelerate heat transfer.

Readout dials are available varying from 25 to 125 mm in diameter and with stem lengths up to 610 mm. Wells, made of carbon steel, stainless steel, or other materials, are available to facilitate removal or to protect against corrosive environments.

Bimetallic elements can be made sufficiently sturdy to actuate a recording pen. A chart, driven by mechanical clockwork behind the pen can form a complete measuring and recording system which is independent of outside electric power.

THE ADVANTAGES AND LIMITATIONS OF BIMETALLIC THERMOMETERS

The advantages over glass stem thermometers include that the bimetallic design is less subject to breakage and is easier to read. The advantages of the bimetallic thermometers relative to the filled or electronic temperature indicators are their lower cost and their simplicity.

A disadvantage is that rough handling changes their calibration The overall accuracy is not as good as that of the glass stem design. The bimetallic thermometers are confined to local measurement.

Table 3.7 Advantages and limitations of bimetallic thermometers

3.8 CALIBRATORS AND SIMULATORS

Temperature calibrators range from simple hand-held instruments to large, permanently installed baths and chambers. Calibrators reproduce temperatures with an accuracy and stability adequate for the range of devices to be checked. The thermometers are normally calibrated at one specific temperature at a time, although a series of baths can be assembled for sequential immersion to enable a range of temperature points to be checked. Correspondingly, the temperature can be reset in stages or as a ramp function for multipoint checking.

Simulators duplicate the outputs of temperature sensors enabling measuring instruments to "see" a simulated but precise temperature value. This output may be a millivoltage to simulate thermocouples, a resistance to simulate RTDs, or a light or radiant energy level to calibrate instruments based on optical or infrared (IR) energy. Simulators can usually be connected locally (at the sensor) and then at the readout instrument to check the operation of transmitters, multiplexers, and cabling.

A fixed resistor connected to a multipoint monitor is an example of a simulator that continuously checks the calibration of indicators, recorders, and alarms.

Some calibrators incorporate both temperature environment and sensor output functions.

Temperature calibration baths

A bath or chamber creates a temperature environment suitable for the immersion of temperature sensors such as thermocouples, RTDs, or bulbs. There are two bath types: fixed and adjustable. The earliest fixed-temperature calibrator used was the ice bath. This was supplemented by other so-called fixed points, such as the steam and sulphur points, and later by a full range of freezing, boiling, and triple points. These are based on the principle that materials change state (freeze and boil) at certain fixed temperatures. Many of these points have become reference values for defining the temperature scale and are therefore especially appropriate for calibration.

Useful examples of such fixed points are the triple point of water (0.0100°C) and the melting point of zinc (420°C). The triple point of water is obtained in a sealed container in which the solid, liquid, and vapour states of water are in equilibrium. In contrast to the triple point, the ice point can be more easily obtained and used to an accuracy suitable for industrial calibration. RTDs in particular are defined in terms of their Ro value, or resistance value at the ice point (e.g., 100 Pt). Immersion in an ice point bath is a useful way of compensating for calibration drift even though the thermometer is normally used at higher temperatures.

Adjustable temperature baths contain a fluid (liquid or fluidised solid) circulated through a chamber in which the thermometers can be immersed. Controllers maintain a temperature at the desired setpoint. A block establishes temperature uniformity amongst sensors at temperatures below 800°C, or where conduction is the primary means of heat transfer.

Simulators

Simulators for thermocouples and RTDs consist of voltage sources and/or resistors having values that correspond to the required temperature readings. Thermocouple simulators require provision for reference junction compensation and alloy terminals.

Simulators for optical or total radiation pyrometer calibration may be as simple as a 25 watt lamp or as sophisticated as a controlled output arc furnace. It is usual to use these simulators with comparison standards. By comparing one optical pyrometer with a standard or with a laboratory calibrated optical pyrometer one can avoid errors that may arise due to their large ambient temperature coefficients. Such calibrations can, for example, be performed in an outdoor shed or pyrometer shop adjacent to a steel mill.

A typical laboratory-type thermocouple calibrator/simulator combines both the simulation and calibration functions in one unit.

A simulator for optical pyrometry incorporates a copper fixed point to provide a reference temperature 1085°C independent of a standards laboratory. A very high temperature 1650°C thermocouple and optical/radiation pyrometer calibration system incorporating a palladium freezing point, molybdenum block, and standard lamp was built for a steel mill in Brazil where traceability to a national laboratory was unavailable.

Field and laboratory temperature calibrations are best performed with calibrators and simulators that are adapted for the particular application.

3.9 INTEGRATED CIRCUIT (IC) TRANSISTORS AND DIODES

Diodes and integrated circuitry (IC) transistors have been available as temperature sensors for several decades, but their use in routine industrial applications is relatively recent. The silicon (and germanium) transistors are small and inexpensive, but before they can be used as part of an industrial system (such as intelligent thermostats), they need to be packaged. One of the first applications of the IC transistor was to provide cold junction compensation for thermocouple circuits. They are also used in multipoint temperature-sensing cables and in other applications. Today both the diode and the IC sensors are available as fully packaged, off-the-shelf units; some of them are also integrated with microprocessors, resulting in a "smart" sensor. Diodes and transistors can both provide sensitivities approaching 0.1°F (0.06°C) on narrow span applications, but in the case of the diodes calibration is needed to achieve that sensitivity.

Integrated circuit temperature sensors

Transistors are sensitive to temperature variations. It has been found that if two identical transistors are operated at a constant ratio of collector current densities, then the difference in their base-emitter voltages will be directly proportional to absolute temperature. Therefore, as temperature decreases the base bias must be increased to maintain the collector current constant. The base bias voltage is usually converted to a current by a low-temperature coefficient thin film resistor. This temperature- proportional characteristic allows IC transistors to produce output signals that are proportional to absolute temperature. The IC temperature sensors are available in both voltage and current output configurations. The current output units are usually set for a one microampere output change per degree Kelvin (Celsius), while the voltage output configuration generates 10 millivolts per degree Kelvin. If a digital voltmeter is used as the readout, the current output is detected as the voltage drop through a 10,000 ohm resistor. The temperature-proportional characteristics of IC sensors are highly linear and in the range of -67 to 302°F (-55 to 150°C) it is superior to all electronic sensors, including RTDs.

Because they produce an analogue voltage that is proportional to temperature, IC transistors have been used for some time for cold junction compensation in thermocouple circuits. IC sensor modules have also been packaged in flat cables, where they are attached to the cable at regular (several feet) intervals. The cables can be as long as 10,000 ft (3050 m) and can have as many as 1000 sensing modules. These cables are economical ways of detecting pipe surface temperatures or cold or hot spots. IC sensors are also being used in the HVAC (heating, ventilation, and air conditioning) industry and in other applications where their low cost and strong linear output outweighs their low accuracy and limited range. Other disadvantages of IC sensors include that they require an external power source (4 to 30 volts) and that they are fragile and subject to errors due to self heating.

Diode-type temperature sensors

Diodes are highly sensitive and linear temperature sensors. Silicone and germanium diode temperature elements are available from -458 to 395°F (-272 to 202°C). They are accurate to 0.2°F (0.1°C) for temperatures that are above ambient and to 0.2% of their full scale range below that. The current through the silicon diode thermometer readout could be anywhere from 10 microamperes to 10 milliamperes, but because higher currents result in higher errors due to self-heating, the currents are usually kept between 100 and 500 microamperes. The meters used with germanium diodes usually have a range of 0 to 50 microamperes. Because of the low source impedances, a simple microampere indicating meter can be used. This sensor can also be used to detect small temperature differences. The main advantages of the diode-type sensors are their high accuracy, particularly at cryogenic temperatures; their small size; their low cost; and their good linearity. The diodes are small enough for most applications, but where sensor size is a problem, microdiodes can be used. One of the disadvantages is that the variations between diodes require a calibration procedure for absolute accuracy, which also increases their costs.

3.10 QUARTZ CRYSTAL THERMOMETRY

The effect of temperature on the frequency of quartz crystal controlled oscillators has been known for some time, work on this subject having first been published in 1946.

Temperature measurement using quartz thermometers is based on the change in resonant frequency in response to a temperature change. This temperature sensitivity is of the order of 0.0005°F (0.0003°C) under ideal conditions. Temperature deviations of the order of 10-6 °F have been measured in the laboratory.

A commercial application of the quartz crystal thermometer includes a unique angle of cut, which exhibits a very linear relationship between resonant frequency and temperature. In this case a reference oscillator and a sensor oscillator are employed. The reference oscillator frequency is selected to provide zero beat with the sensor oscillator when the probe is at 32°F (0°C). Sensor oscillators operate to obtain a frequency sensitivity of 500 Hz/°F. Dividers are arranged to obtain a sample period of 0.01 seconds. A resolution of 0.2°F (0.1°C) is obtained with a digital readout. A digital-analogue converter changes the output for external use. A differential measurement can also be made using similar circuitry, with the principal difference being that gating circuits are added to enable the oscillator circuits to be heterodyned against each other.

Probes are provided with the quartz crystal hermetically sealed in a stainless steel cylinder, similar to a well for a thermocouple or a resistance thermometer capsule. This makes the sensor unit larger than either of the two sensors mentioned; a 3/8 in. (9.5 mm) outside diameter is typical. However, the probes can be used to pressures of 3000 PSIG (21 MPa) and can stand shocks of 10,000g without changing calibration.

Response time is stated as being 1 second for a step change in a well-stirred water bath. A strong feature is stability, which for short term is in the order of 0.0001°F (0.000056°C). Long-term stability, for periods of a month or more, is in the order of 0.02°F (0.01°C). These can be degraded by oscillator drift, but, in an environmentally controlled area, this is not generally a problem.

ADVANTAGES AND LIMITATIONS OF QUARTZ CRYSTAL THERMOMETRY

ADVANTAGES

1. No lead resistance or noise problems because temperature is converted to frequency. 2. Excellent short term stability. 3. Good accuracy. 4. One-second response time. 5. Accurate differential measurements possible. 6. Rugged - can withstand shocks without changing calibration.

DISADVANTAGES

1. Expensive. 2. Accuracy somewhat lower than resistance thermometer or thermistor. 3. Probe size larger than thermocouple or RTD. 4. Best used in laboratory environment.

Table 3.8 Advantages and limitations of quartz crystal thermometry

3.11 ULTRASONIC THERMOMETERS

The relationship between sound velocity and temperature has been known for nearly 100 years, but until recently it has been little used. Temperature dependence of velocity in an ideal gas is expressed as:-

2 v = RT/MW

where v is sound velocity, is the ratio of specific heats, R is the gas constant per mole, MW is the molecular weight, and T is the absolute temperature.

A method of temperature measurement in a plasma jet involves the use of two quartz probes set a fixed distance apart. The sound velocity is determined by circuitry for the continuous measurement of the ultrasonic wave transit time.

Making the measurement one of time is an advantage resulting in good resolution (and in well- implemented designs, high accuracy). Because the gas is the thermometer element, errors such as leakage are absent and fast changes can be followed. Disadvantages include high cost, non-ideal gas behaviour, pressure correction, accurate determination of , and the inability to make point measurements.

Acoustical temperature sensors can measure temperature from the cryogenic range to plasma levels (20,000°C, or 36,000°F). The accuracy can approach that of a primary standard as in the example of the ultrasonic interferometer developed by Harmon Plumb and George Cataland of the NIST. A fluidic oscillator in which the resonant frequency of the cavity varies with temperature can be employed as a high-temperature pyrometer.

Ultrasonic thermometers are based on the effect of the temperature on the velocity of sound waves in the medium transmitting the sound waves. Temperature measurements can be made in gases, liquids, or solids.

The acoustic velocity can be detected by immersing a rod or wire into the fluid or by using the medium itself as an acoustic conductor. The sensor rod can measure the temperature at a point or, by means of a series of constrictions or indents, can profile or average the temperature within the medium. This is done by measuring the time lags of the sound waves reflected from the consecutive indents.

When the medium is used as the conductor of the ultrasonic pulse, the transducer is located within the vessel or on its external shell. This latter configuration is useful when measuring the temperature of solids or extremely hot or corrosive materials such as molten sodium.

The acoustic pulse is generated by a piezoelectric crystal cut to resonate at a frequency ranging from 0.5 to 3 MHz or by means of magnetostrictive materials.

The thin wire sensor is installed like a thermocouple. A lead-in wire carries the pulse to the thin wire made of a material suitable for the process medium and its temperature range. Reflection from the beginning and the end of the thin wire provides the time lapse information for temperature determination. The selection of materials is more flexible than for thermocouples as only one material is involved.

Non-invasive applications employ a transmitting transducer to send acoustic energy through the medium. Receiving transducers detect the energy and the time delay as measured by an intervalometer, which determines the velocity of sound and therefore the temperature of the process.

Ultrasonic thermometry can be used at temperature extremes, in high electrical fields, or when the medium being measured is inaccessible. It is also useful for averaging the temperature of bulk materials or for profiling furnace temperatures.

3.12 MISCELLANEOUS TEMPERATURE SENSORS

In this section several unrelated methods of temperature measurement will be briefly discussed. Some of these are still in the experimental stage, while others have not been exploited commercially because they are not yet competitive with the more common sensors. This fast-changing field is so broad that it is not easy to provide complete coverage.

3.12.1 Self-measuring devices

Sometimes temperatures can be measured by allowing a material to serve as its own thermometer. This is done when determining the pressure of a confined gas to indicate its mean temperature, or the resistance of heater coils for the same purpose.

Actually, any property having a consistent rate of temperature variation will serve to indicate temperature. The frequency of the chirp of crickets on a summer night is an indication of their temperature environment. The rate at which a viscous substance such as oil drips through a small hole in the bottom of its container is just as much an indication of its temperature (provided the time- temperature relationship is known) as would be obtained by inserting a glass-stem thermometer.

In addition to such self-measuring devices, there are also many new configurations of older systems. For example, the bimetallic spring can now be used as an optical temperature switch. In this design the bimetallic element is upwardly convex until the set temperature is reached. At that point it becomes downwardly convex, which blocks an optical light path. The number of devices such as this which exploit old concepts in new ways is very large, and no attempt will be made to cover them here.

3.12.2 Acoustic time domain reflectometry

This thermometer operates on the principle that ultrasound pulses travel in solids at speeds which are a function of the temperature of the solids. The measurement is made by detecting the time needed for the acoustic pulses to travel from the transducer to the impedance demarcation point (which may be the junction between the wire and the wall of the tank) and back to the transducer. This device is in the development stage and shows good potential, although some drift in the measurement has been reported. The rhenium sensor used in the system will require sheathing material in more hostile environments.

Gas temperature can be measured by detecting the time needed for an ultrasonic pulse to travel through a fixed distance in the hot gas.

3.12.3 Carbon resistors

Commercially available carbon resistors have been used as temperature sensors in the cryogenic temperature area near absolute zero (from about –424°F, or –253°C, downward to below –458°F or – 272°C). The problems below –458°F (-272°C) have been noted in connection with paramagnetic salt measurements.

Resistor sizes of 0.1 to 1 watts and ambient resistance values up to 150 ohms exhibit a large increase in resistance below –424°F (-253°C). Reproducibilities on the order of 0.2% are obtainable.

Small size, low cost, and general availability make their use attractive in cryogenic work.

The influence of stray radio interference and a loss of sensitivity are drawbacks. Variations between resistors make calibration difficult.

In addition to carbon resistors and conventional RTDs or thermistors there is a variety of special resistors used as temperature sensors. One example consists of a thin nickel film deposited onto an electrically insulating substrate in a reducing atmosphere. The sheet resistance is a function of the heat treating temperature, cycle time, and thickness. The different designs of special temperature- sensitive resistors are too numerous to mention.

3.12.4 Capacitance cable for detecting hot spots

A cable consisting of an electrical conductor, a dielectric, and a conductive plastic can be used to detect hot spots. The polymer in the over-temperature locator cable increases its electrical resistance with temperature. Therefore when a hot spot evolves anywhere along the length of the cable, this will result in a drop in capacitance, as measured from one end of the cable. The location of the hot spot can be determined by comparing the measured capacitance from the control end of the cable with the capacitance of the full length of the cable.

3.12.5 Fluidic sensors

The fluidic sensor is a device for converting gas temperatures into gas pressure. Actually, it is a beat- frequency detector system which contains no moving parts other than the gas.

One type of design uses a two-chamber resonator (the oscillator) in which the entering gases are split by a knife edge. The gases are reflected from one chamber into the other, setting up oscillations whose frequency is proportional to the square root of the absolute temperature.

A reference signal input from a temperature-sensitive resonant oscillator is compared with the unknown in a beat-frequency detector made of beam-deflection fluid amplifiers. The frequency of its output is the beat frequency of the combined reference and oscillator signals. The components in the frequency converter create a steady pressure proportional to the beat frequency. Accuracies of about 2% and temperatures up to 2000°F (1093°C) are claimed for the system.

3.12.6 Johnson noise thermometer

Another thermometer under development capitalises on the phenomenon that electrical resistors provide a voltage related to thermal noise. These devices also use rhenium sensors. They are referred to as the Johnson noise detection-type thermometers.

3.12.7 Liquid crystals

Used in non-destructive testing for surface temperature measurements, liquid crystals undergo a series of colour changes as temperature varies. They are an organic compound which is physically liquid, but which exhibits optical properties similar to those of a crystalline solid.

A number of solutions are available, from the minimum temperature of about 68°F (20°C) to a maximum of approximately 340°F (170°C). The solutions are packaged in kit form for various ranges within these limits. Mixtures are made covering spans as narrow as 4°F (2.2°C) within the selected range.

Temperature is read by comparing the colour exhibited when a thin coating is subjected to the conditions under question to a standard reference colour. Response speed is less than one second. The indication is continuous and reversible. Cost is low. Disadvantages are manual preparation and limited range, in addition to lack of automatic readout.

Liquid crystals have also been used in fibre-optic thermometers. A pellet of liquid crystal is inserted at the tip of an optical fibre, and as light at different wavelengths is sent through the fibre, the reflection peaks from the liquid crystal are related to temperature.

3.12.8 Paramagnetic salts

Magnetic thermometry has been developed chiefly to measure temperatures near absolute zero (below –458°F, or – 272°C). The temperatures themselves are obtained by adiabatic demagnetisation of a paramagnetic salt. An isothermal magnetisation at the lowest attainable liquid helium temperature (about –458°F, or –272°C) followed by an adiabatic demagnetisation is used. The entropy is decreased, with a simultaneous heat flow from the sample, when the magnetic ions are oriented

parallel to the field. During subsequent adiabatic demagnetisations the entropy of the salt remains constant, if demagnetisation is reversible, and temperature decreases.

To obtain the temperature, some temperature-dependent quality of the salt under investigation is used, such as the magnetic susceptibility.

If a sphere or rotational ellipsoid of an isotropic paramagnetic salt is located in the homogeneous part of the magnetic field of a coil of a mutual inductance or a self-inductance, the inductance of the coil is a function of the temperature. Inductance can be measured with an AC bridge whose balance is independent of frequency. A galvanometer can be used for detection. Effective shielding is a requirement.

In a paramagnetic salt with a coil surrounding it, self-inductance is related to temperature. An Anderson AC bridge has been used to measure magnetic temperature in such a situation. The relationship between self-inductance and susceptibility of a salt has been found to be linear when the ellipsoidal or spherical salt piece is placed in the homogeneous part of the measuring field.

Accuracy of the magnetic method has been estimated in the order of 0.001°F (0.00056°C). The method is the best available for measurements near absolute zero.

3.12.9 Spectroscopic temperature measurement

Spectroscopic methods are often used to measure the temperature of hot gases. They are, in fact, the only possible way to measure the surface temperature of stars.

The spectroscope in its simplest form is the familiar triangular glass prism which breaks up light from a hot object into its constituent colours (its spectrum). The chemical composition of glowing gas is determined from the pattern of dark (Fraunhofer) lines which appear across the spectrum.

Many procedures for temperature determination from the spectrum have been developed, such as measurement of brightness and actual colour, reversal temperatures, population temperature estimates, measurements made of spectral line shifts in ionised gases, and many others. These are all laboratory techniques seldom employed industrially because of their complexity and relatively high costs. For further information on this subject the reader is referred to Chapter 56, "Spectroscopic Methods of Temperature Measurement", of the book Temperature - Its Measurement and Control in Science and Industry, Volume III, Part I.

3.12.10 Thermography

The strong temperature dependence of the brightness of certain luminescent materials may be converted into a pattern of colour which can be recorded photographically. A thin layer of this material is placed on the surface to be investigated and is suitably excited with ultraviolet radiation in a darkened room. The temperature is implied by the brightness of the coating in comparison to the brightness of the same coat at a known temperature.

The sensitivity of the phosphors used gives a 10% brightness change per °F and this can be picked up with a relatively crude system of photometry. Temperature range for this type of measurement is from 32 to about 750°F (0 to about 400°C).

3.12.11 Colour Indicators, Crayons and Pellets

A number of temperature-related physical changes have been used to produce simple thermometers. Crayon marks on heated work-pieces or pellets placed in furnaces change from solid to liquid when their melting point is reached; paints and heat-sensitive labels change their colour; luminescent materials change their brightness; and liquid cholesteric crystals detect skin temperature.

For centuries, manually operated furnaces have been temperature-controlled by the operator's placing a heat-sensitive object inside the furnace and observing the status of that indicator through a peephole. The most easily observed physical changes were found to be melting and changes in

colour. Coloured crayons and paints have been used as temperature indicators and are still used as end-point indicators in a number of high-temperature batch processes.

Colour indicators are a class of sensors which have the property of changing their original colour when a certain temperature is reached. The change is distinct, not just an alteration in shade. For instance, an indicator may change from yellow to grey, or from light blue to light brown. Some can go through several colour changes at different temperatures.

Paints and crayons are familiar forms of these indicators which are applied directly to a solid object either when it is cold and about to be heated or when it is already hot. Some indicators can determine the temperature of solid objects immersed in oil. They are not recommended for use in hot gases.

Temperature is indicated by a chemical reaction, where a molecule of a gas, such as ammonia, CO2, or water vapour, is driven off the basic stock (colourful salts of metals like nickel, cobalt or chromium), thus changing its colour. The change is usually permanent after the object cools down. An exception occurs when the gas is water vapour; the indicator may slowly reabsorb this gas from the air and revert to the original colour.

Change in colour of these types of indicators is not only a function of temperature but also of time. For this reason, the immediate past temperature history of the indicator will influence the exact point at which it will change colour. The indicators are usually rated for a specific temperature over a certain time period, for instance 60°C in 30 minutes. This means that if held at a constant 60°C the colour change will occur in 30 minutes. If the colour change occurs in less than 30 minutes, the average temperature is higher than 60°C, and vice versa. On such an indicator, if the temperature does not exceed 54°C the change will never occur, because the indicator is stable below this temperature.

Many different temperature ratings are available. They can be obtained in a series for every few degrees to the maximum offered (about 2500°F, or 1371°C). This class of indicators is quite inexpensive and is used in industry where only an end point is needed and someone can be present to watch for or interpret the results. A disadvantage of these sensors is that the material adheres tightly to the object on which it is placed and presents a problem if it must be removed later.

3.12.12 Pyrometric Cones

Like coloured crayons and paints, pyrometric cones have also been used as temperature indicators and are still used as end-point indicators in such batch processes as firing in pottery furnaces. They can be small, expendable, plugs, chips, or geometrically shaped objects whose purpose is to accompany the products through a heating cycle. The physical or metallurgical changes that occur indicate the temperature reached in the process.

The ceramics industry

The German ceramist Herman Seger invented the first pyrometric cone in 1886. The individual cones look like truncated pyramids. Because the pyrometric cone measures the effects of both temperature and the length of firing time, it acts as a heat integrator, a function that cannot be easily reproduced by thermocouple-type pyrometers. For this reason the ceramic industry continues to use cones, in addition to recording pyrometers, even in its most modern kilns.

The indicator material is generally quite similar to the substance of the work under test. Pyrometric cones are actually composed of ceramic materials very carefully blended to soften at a certain temperature. The slender cone is slightly tilted from the vertical; when its softening point is reached the tip bends over and may actually touch the base. This action can be watched through the window of the firing furnace or its condition can be studied after cooling. Observation of a fired cone will show the experienced operator if the furnace atmosphere was oxidising, reducing, or carbonising. If the later has taken place the cone will have formed a shell less dense than the interior. Presumably the work will have taken on the same characteristic.

The cones are set in a plaque of fireclay (called a "cone pat") close together and tipped at about 8 degrees from the vertical toward the cone that is expected to bend first. The cone pat is located in

front of the peephole where the operator (or a closed-circuit TV camera) can-observe its status. Usually three cones are placed on a pat: the controlling target cone is in the centre, a cone that melts at a lower temperature is in front of it and one that melts at a higher temperature is behind it. The softening points can be selected very accurately, and accuracies of 2 to 5°F (1 to 3°C) can be obtained.

As an alternate to cones, the indicator may take the shape of a long cylindrical bar. The bars are supported at their ends with axes horizontal. On temperature rise they soften and sag at the middle under gravity. The deformation serves as a measure of temperature.

Another group of indicators operates by shrinkage rather than deformation. After removal from the furnace the diameter of a hole in the indicator, or perhaps the indicator's length, is measured and compared with the original dimension.

Like colour indicators, pyrometric ceramics should not be considered exact temperature measuring devices. The fusion, bending, and/or shrinking that they undergo is a time-temperature relationship and, as such, it is only useful to determine the end point of the specific job. This property is frequently more important than an exact measurement of the instantaneous temperature. The use of this type of indicator may almost be considered an art.

While pyrometer cones are not well suited to automatic process control, they are inexpensive and valuable quality control tools in guaranteeing repeatable qualities of ceramic and similar batch products from kilns and furnaces.

Engine test research

An entirely different material, used in a similar manner, is the metal test plug. This small device can tell temperature by a change in hardness that results from the heat treatment it has received. One use is to have it located carefully in an operating engine, in an otherwise inaccessible spot, where it will respond to the temperatures that occur during operation. When the test is over the plug is removed and carefully analysed to determine the change in hardness along the horizontal axis.

Time is again a factor, but metal responds much faster than ceramic material. Exposures of less than one second duration can be detected.

Advantages of this class of temperature sensors are their relative economy and design for a very specific lob. Their shortcomings are obvious.

4. GENERAL SELECTION CRITERIA

A detailed description of each of the main categories of temperature instrumentation (together with some lesser known categories) has already been discussed in Section 3. Within Section 3, some specific selection criteria relevant to each class of instrument has been included, and known advantages and limitations are summarised at the end of most sub-sections.

In addition, the following table summarises some of the main features of the principle types of temperature instrument.

THE MANY WAYS OF MEASURING TEMPERATURE

It is believed that Galileo invented the liquid-in-glass thermometer around 1592. The principle behind the thermocouple, the existence of the thermoelectric current, was discovered in 1821 by Thomas Seebeck. The same year Sir Humphry Davy noted the temperature dependence of metals, but the first resistance temperature detector (RTD) was not constructed until 1932, by C. H. Meyers.

The difference between the thermal expansions of two strips of metals produces the bimetallic spring, used in ranges from -80 to 800°F (-62 to 427°C) to build thermostats, bimetallic thermometers, and temperature switches. The latter can be manufactured at unit costs of a couple of dollars and serve to control various household appliances, from washing machines to dryers to dishwashers.

The other traditional thermal element, the filled system, which has a range from -300 to 1000°F (-184 to 538°C) is still used to a limited extent in HVAC (heating, ventilation, and air conditioning) and industrial applications, although there is something about nearly all of the filling materials that makes them less than ideal. Wax and other solid filling materials have just about disappeared, mercury filling is in disfavour because of its toxic characteristics, gas filling is disliked because of the requirement for a large size bulb, and vapour filling can be a problem if the process temperature crosses through the ambient. Liquid filling is still commonly used, but it also has problems with head effects, limited capillary lengths, and compensation.

Electronic thermometers

These days the most often used industrial thermometers use sensors that change either their output voltage or their electric resistance as a function of temperature. Of these, probably the most popular are the thermocouples (TC). They have a wide temperature range (-440 to 5000°F, or -262 to 2760°C), relatively low cost, and many physical forms, making them highly versatile. When used on high- temperature combustion processes, it is necessary to continuously detect the loop resistance of the thermocouple, because a sudden drop in that resistance signals thermocouple burnout.

The main limitation of thermocouples is the small amount of voltage that is generated by a unit change in temperature. For example, a one degree Fahrenheit temperature change on a platinum thermocouple results in an output voltage change of 5 microvolts. In contrast, some of the temperature transmitters might have a minimum span of 3 millivolt and an inaccuracy of 0.25% of that full range, or an error of 7.5 microvolts. Consequently, thermocouples are not recommended for use on narrow span or small differential temperature measurement applications.

Similarly, because of the weak output signal, multiplexer contact resistances (over 1 microvolt) can also contribute relatively large errors. Because the thermocouple generates the smallest signal of all thermometers, it is the most affected by noise and lead-wire errors. Because the thermocouple is a temperature difference sensor, it is also subject to cold junction compensation and lead-wire junction errors. Most of these problems can be overcome by placing the converter electronics directly on top of the thermowell and thereby eliminating the problems associated with the transmission of the weak thermocouple signal.

Resistance temperature detectors (RTDs) have a narrower range, from -430 to 1200°F (-257 to 649°C), but otherwise are equally as popular as the thermocouples. They are more linear and more stable and

can accurately measure smaller temperature spans or differentials than can the thermocouples. As with thermocouples, it is recommended to mount the electronics directly on top of the RTD and thereby eliminate the need for lead wires. If this is not done, the resistance of equal length lead wires can still differ by 10%, and even the three-wire installations can cause substantial errors. (A standard 100 ohm platinum RTD changes its resistance by 0.385 ohms/°C, while if it has 500 ft long leadwires, they can be mismatched by a full ohm). Other potential error sources include the ageing of reference resistances, change in the voltage generated by the power supply, and moisture penetration.

Thermistors are miniature metallic oxide elements, which generate the most sensitive and the fastest response of any of the thermometers. For each degree Centigrade, their resistance changes by about 200 ohms. Their range is only from -150 to 600°F (-101 to 316°C) and their output is very non-linear. Because the thermistor element usually has a resistance of about 5000 ohms, even small current flows can cause self-heating.

Thermistors can be packaged in a small, portable element, which can be inserted into the process and can memorise the process temperature while it is travelling with the product. When this temperature memory package emerges at the discharge point, it can be plugged into and interrogated by a computer to thereby obtain the full temperature history of the process.

Integrated circuit (IC) sensors have a limited temperature range (67 to 257°F, or -55 to 150°C) and are not yet available in the variety of packaging that thermocouples and RTDs are. Their main advantages are their low costs, and their linear and strong output signals. Diodes are mostly used in the cryogenic temperature range of –271 to 202°C (-455 to 396°F) and when calibrated can be accurate to 0.05 to 0.1°C.

Non-contact pyrometers

In combustion processes, the temperature can be measured on the basis of the radiation emitted by the hot body. Total radiation pyrometers operate by measuring the total amount of energy radiated by a hot body. Their temperature range is from 0 to 7000°F (0 to 3890°C). These pyrometers have gradually been replaced by infrared (IR) pyrometers, which detect the dominant wavelength of the radiation received from a hot object. The basis of infrared thermometry is the fact that as temperature increases, the dominant wavelength of hot body radiation gets shorter. Infrared thermometers have about the same range as total radiation ones.

Newer developments in temperature pyrometry involve the automatic measurement of changing surface emissivities of hot objects, using laser beams to automatically correct these changes, The other major development is the coupling of infrared pyrometry with fibre-optic (FO) technology, which can focus the temperature measurement on inaccessible points in unfriendly environments that are a long distance from the optical/electronic readout instruments. Some of these fibre-optic thermometers are capable of high accuracy, about 0.1% at around 1000°C and can operate over a range of from 932 to 3632°F (500 to 2000°C). Fibre-optic systems can also be multiplexed so that the cost of expensive electronics is shared between several temperature measurements.

Non-contacting temperature measurement can also be achieved by the measurement of radiated heat using thermopiles. These devices are unaffected by colour, finish, or emissivity changes as they depend only on the convection of heat.

Table 4.1 The many ways of measuring temperature

The development of temperature sensors was a slow process until the middle of the twentieth century. Today there are over 20 different types of thermometers, each employing a different principle of operation; and many thousands of different makes and models available for each thermometer type. The traditional practice of using only dial-type expansion thermometers, RTDs or thermocouples throughout a particular industrial plant is giving way to the practice of selecting each temperature sensor for a particular application. This requires a better understanding of the features and capabilities of the many thermometers on the market.

The ‘best’ thermometer to use in a particular application, however is something that only the user should decide. The choice will depend on many factors and a selection of these is listed below (a few of which are briefly discussed in this section). However, the procedure for deciding on the optimum choice for any particular situation cannot be written down concisely.

Things to consider:-

• The temperature range to be covered (maximum/minimum operating range of the instrument) • The smallest temperature change likely to be measured (required resolution) • How well the temperature must be known (required accuracy, including hysteresis and reproducibility) • Process media characteristics (e.g. temperature fluctuations, corrosion, composition and density) • External environment (e.g. ambient pressure, temperature and humidity changes; corrosion, vibration etc.) • Use of thermowells • Type of use (e.g. indication only, control, measurement of solid samples, fluids, furnaces, moving objects etc.) • Signal conditioning, outputs and displays • Cost of purchase; ease and cost of installation & maintenance (including servicing and calibration) • Safety considerations • Storage and transportation • Other factors, e.g. instrument characteristics - ruggedness, size, speed of response, etc.

For example, if the temperature to be measured is around 2000°F (1143°C) that requirement alone narrows the choice to either pyrometers or thermocouples. On the other hand, if the temperature to be measured is 200°F (93°C), just about all the temperature sensors can be considered and therefore one can take other factors, such as size, speed of response, linearity, or cost into consideration.

4.1 Temperature range

Figure 4.1 Selection of thermometer type by measurement range

The range in temperature within the universe varies from near absolute zero of empty space to the billions of degrees in the nuclear fusion process deep within the stars. But the practical range on earth can be considered as extending from around 0.5 K upward about five decades to around 10,000°C. This is still a tremendous range, and no single sensor could possibly cover it. Therefore, one of the restrictions on the temperature sensor concerns the temperature range over which it can stay reasonably accurate.

The temperature range is probably the single most important criteria to consider – can the thermometer operate within the required range of interest? Is it suitable for the extremes of temperature to be measured? Can it withstand long durations at a given temperature without affecting its calibration?

The above table shows general ranges for the main classes of temperature instrument. Typical range details of each instrument type are included in each of the main sections in Section 3. In addition, the manufacturers technical specifications should be consulted prior to selection.

4.2 What is the smallest temperature change you need to record? (required resolution)

For many applications knowledge of the temperature to within 1ºC is sufficient. The more accurately you wish to know the temperature, the more care and the greater expense you will incur. Typical resolution details of each instrument type are included in each of the main sections in Section 3.

4.3 How well the temperature must be known (required accuracy)

The factor most often considered after instrument range is accuracy.

If one needs to know the temperature only to within a few degrees, then it may be possible to use uncalibrated sensors of many sorts to produce estimates of temperature. For example, for many purposes an essentially uncalibrated off-the-shelf platinum resistance sensor or thermocouple should be able to determine temperature to within a few degrees from around -200ºC to + 200ºC with relative ease. However, if you wish to be confident that the temperature of your process or experiment is within, say, 0.5ºC of one specified, then in general you will need to have your sensors (and your entire measurement system) calibrated at a UKAS accredited laboratory.

In making a selection, one should evaluate the different sensors considered against the actual accuracy requirement of the installation. It should also be recognised that the total error of the temperature measurement or control system will be more than just the error contribution of the sensing element. Added to this error will be the inaccuracies contributed by the installation (lead wires, junctions, noise, etc.) and by the converter, transmitter, and readout devices. Most importantly, the errors associated with a failure to achieve thermal equilibrium between the sensor and the process, are likely to be very large.

The typical error contribution of an integral thermocouple transmitter is 0.1% of span or 5 microvolts, whichever is larger (in the case of platinum thermocouples, this amounts to an error of about one degree Fahrenheit). For integral platinum RTD-type transmitters the error is usually 0.15% of span or 0.15°F (0.08°C), whichever is larger, and for nickel RTDs it is 0.25% of span. The best obtainable digital-to-analogue (D/A) conversion error is 0.025% of span. In addition, supply voltage variation can contribute 0.02% of span and ambient temperature changes can cause additional zero, span and reference junction errors.

Typical accuracy details of all instrument types are included in each of the main sections in Section 3.

4.4 External environment (atmospheric effects)

Adverse atmospheric conditions can cause problems with the measurement of temperature. For example, in a highly humid or very moist environment it is essential that the element of a resistance thermometer or the bead of a thermistor be well insulated electrically. If moisture contacts the resistance element or thermistor bead, it may cause a short. Thermocouples in general are less sensitive to moisture than are resistance thermometers. If null balance potentiometric or high input impedance readout devices are used, insulation resistance between legs of the couple as low as 10,000 ohms can be tolerated without serious error in the indicated temperature. If low input impedance current measuring readout devices are used, a high insulation resistance between legs of the couple becomes as important as with resistance thermometers and should be in the megohm range.

Corrosive, reducing, or oxidising environments also create problems. If a sensing device is exposed to such environments, it must be protected by some form of envelope or coating. It should be remembered that protection applied to a sensor will increase its mass and thus, in general, adversely affect its response characteristics.

4.5 Type of use

Temperature sensors should be selected to meet the requirements of specific applications. The most difficult temperature measurement applications are those where high temperatures are to be detected within a hostile environment, such as that which exists within a fluid-bed coal gasifier.

High-temperature service

At high temperatures in the 2000 to 3000°F (1100 to 1650°C) range, the problem of thermowell drooping becomes a problem when using thermocouples. If direct temperature measurement is to be used (where the temperature sensor is physically inserted into the process), metallic thermowells are likely to droop, resulting in problems of replacement and maintenance. If ceramic thermowells are used instead of metallic thermowells, erosion becomes a problem in fluid bed-type environments due to the relatively low surface strength of the ceramic materials.

If the thermowell is flush with the inside diameter of the reactor, its life is likely to be increased, but only at the expense of having a less representative temperature measurement because of the greater influence of wall temperatures. Another way of increasing the reliability of the overall installation is to provide automatic, scheduled, preventive maintenance replacement of the thermowells. This can be done on a redundant basis, so that at any one time two or even three temperature sensors are in operation. A voting system can be arranged in critical locations so that the reading of the one sensor which disagrees with the other two is disregarded. While it is desirable to replace the thermowells without shutting the process down, and while it is desirable to replace them before they fail, all this effort is not inexpensive, and over a period of years of operation it can easily approach the initial cost of non-contact sensors, such as infrared pyrometers.

The trend in high temperature measurement favours the indirect, non-contacting type measurements. The unit cost of such installations can be lowered by making multiple readings, using fibre optics to connect many sensing points to the same infrared source and the same receiving electronics package. This seems to be the most promising solution to the problems of high temperature measurements in hostile environments.

Measuring the temperature of solids

Determination of the allowable size and configuration of the sensor requires some knowledge of the heating or cooling conditions together with an estimate of the magnitude of the temperature gradients that are likely to exist in the region in which the measurement is to be made. A simple rule-of-thumb indicator to determine if significant gradients are likely to be present is the magnitude of the Biot modulus (hL/K), where h is the surface heat transfer coefficient, L is the smallest dimension of the solid, and K is the thermal conductivity of the solid. If this modulus is over 0.2, significant temperature gradients are likely to exist in the solid, and care should be exercised in choosing the size, location, and orientation of the sensor within the solid. If the Biot modulus is less than 0.2, no significant gradient is expected and a measurement anywhere on or within the solid should give identical results regardless of size or configuration of the sensor.

Measuring the temperature of fluids

The fundamental problem of measuring the temperature of a fluid is one of assuring strong thermal coupling. For a fluid temperature measurement to have meaning, the sensor must come to equilibrium with the temperature of the fluid. The difference between the equilibrium temperature of the sensor and the fluid temperature is a direct error. Consequently, with rapidly changing temperatures the rate of heat transfer between the sensor and process fluid must be sufficient to overcome the thermal capacity of the sensor in order that it can follow the fluctuations in fluid temperature.

A sensor, to respond rapidly to changes in fluid temperature, should have a large surface-area-to-

mass ratio. Furthermore, the heat transfer coefficient, which is a direct function of the fluid mass flow rate over the sensor, should also be as large as possible. Forced convection or rapid flow of the fluid over the sensor is therefore desirable.

A practical measure of sensor sensitivity is its time constant. Since the sensor approaches change in fluid temperature asymptotically, it is difficult to determine when it has reached that temperature. The time constant of a sensor is the time required to accomplish 63.2% of the step change.

When measuring the temperature of a gas stream in a heated duct or furnace where temperature differences between the sensor and its surroundings exceed 1000°F (538°C), significant errors can occur due to radiation exchange between the sensor and its surroundings. Under such conditions the sensor must be shielded against thermal radiation exchange. This can cause disturbances in the flow of the fluid around the sensor and hence affect the directional response characteristics.

Another thermal effect in measuring the temperature of gas streams at high velocities is the recovery factor, which results from an increase in the temperature of the gas at the sensor due to compression heating as the gas is brought to stagnation against the sensor.

Temperature of moving surfaces

Measuring the surface temperatures of webs of paper, plastics, textiles, or metals or of rotating cylinders (such as calendar rolls, rotary kilns, or drier cans) requires special thermometer designs. These instruments are either non-contacting or travelling-brush skid designs. Some use thermocouple elements, while others use RTDs or even bimetallic springs. Still others are of the conventional or fibre-optic infrared variety or of the convective null-heat balance design. The brush design is available for use from -300 to 1000°F (-184 to 538°C) and the contact bar can be lead, copper, brass, or mild steel. The feather-touch units operate up to 1600°F (870°C) in intermittent, and up to 1000°F (538°C) in continuous services. The non-contacting designs use type J thermocouples, are constructed mostly of stainless steel, and can be used from -150 to 1000°F (-100 to 538°C).

Depending on the environmental conditions, radiation pyrometers can also be used to detect the temperature of moving surfaces. The portable units are convenient for maintenance purposes, such as the detection of evolving shorts in cables, the spotting of leaking steam traps or insufficient thermal insulation, and so on. The main limitations of IR thermometers are caused by changes in the emissivity of the surface and by contamination and dirt build-up. Laser based compensation is claimed to overcome the emissivity problems, and purging can be useful to reduce contamination in fibre-optic infrared thermometers. The radiation pyrometers are usually either portable or permanently installed infrared thermometers. When the temperature of a hard to reach area is required (say the detection of the surface temperature of turbine blades inside combustion turbine engines), the fibre- optic infrared sensors are likely to be used.

The convective null-heat balance concept involves the null-balance of convective heat transfer between a sensing head of known temperature and a surface of unknown temperature, in which temperatures are compared rather than directly measured. If, for example, two surfaces are placed in close proximity to one another, their temperatures can be compared by measuring the rate and direction of heat transfer between them. The direction of heat transfer is a definite measure of which surface is at the higher temperature; if the heat transfer between them is zero, their temperatures must be equal.

If one of these surfaces contains a heat transfer sensor and its temperature (T) is known, one can readily determine if the temperature of the other surface is either below, equal to, or above T by simply observing whether the heat transfer is zero or not and what its direction is. It is not necessary to measure the magnitude of the heat transfer. The technique, of course, works only if the point at which heat transfer is being sensed is isolated from surrounding influences. On a flat surface, isolation is accomplished by locating this point at the centre of a relatively large isothermal plate (typically 6 in/150 mm in diameter) spaced approximately 1/16 to 1/8 in (1.5 to 3 mm) from the surface being measured.

The basic concept of null-heat balance can be used either to monitor or to control surface temperature. In the monitoring mode, the temperature of a sensing head is automatically varied to

keep it at null-heat balance (at the same temperature) as the moving surface. The head temperature, which is equal to the surface temperature, can then be measured.

In the control mode, the head is held at the desired setpoint temperature at which the surface is to be controlled and the signal from the heat flow sensor is employed to manipulate the surface temperature.

In another arrangement, the signal from the heat flow detector is calibrated for temperature difference between the sensor, which is at a constant known temperature, and the surface being measured.

The convective null-heat balance type surface thermometers can detect temperatures up to 500°F (260°C) and can follow changing temperatures up to the rate of 15°F/min (8.3°C/min); their measurement error is about 2% of full scale.

4.6 Use of thermowells

The purpose of thermowells is to protect the thermal elements from mechanical damage and corrosion or to act as sighting tubes for radiation thermometers.

Thermowells permit elements to be removed for calibration, replacement or repairs, or the use of portable sensors. The well or protection tube is fixed into the pipe or vessel secured by threads, flanges, or welds allowing the element to be inserted without stress.

Thermowell types

Thermowells are usually metallic and may be coated with other materials to provide additional corrosion protection. Integral flanges or threaded sections enable the wells to be secured to the pipe or vessel. Internal threads secure nipples which extend the head beyond any insulation. It is important to avoid contamination of temperature sensors by oil (such as cutting oils) left on the inner surface of drilled wells or accumulations of foreign materials within wells.

Protection tubes can also be ceramic when used to protect noble metal thermocouples or as sighting tubes for radiation pyrometers. While not as strong as metallic thermowells, they do not droop and can withstand higher temperatures; in addition, they are free of contamination which can cause thermocouple drift due to vapour deposition of elements at higher temperatures. Mullite and high- purity alumina are commonly used as ceramic thermowell materials. In addition to being corrosion- resistant, mullite can operate up to 3200°F (1750°C) and alumina up to 3540°F (1950°C). When platinum thermocouples are used at temperatures exceeding 2200°F (1200°C), mullite should not be used, because it contains impurities which can contaminate platinum. For such applications, high- purity alumina is the proper choice.

Dual protection tubes may be used where the outer tube provides mechanical protection and the inner tube provides corrosion or permeation protection.

The sheaths are either extruded or woven and are used to protect thermocouples or other wires. Woven sheaths are similar to those used for electrical insulation and may be made of stainless steel, Inconel, tinned copper, fibreglass, or ceramics. They can withstand temperatures up to 2200°F (1204°C) continuously or 2800°F (1538°C) for short periods of time.

Extruded sheaths can be plastic or metallic. The metallic sheaths are commonly used for mineral- insulated thermocouples in which the wire elements are surrounded by ceramic oxides such as MgO. After being packed with powdered oxides, the sheaths are swaged or rolled under pressure to reduce their diameter and tightly pack the powdered insulation. While this process serves to reduce intrusion of atmosphere within the sheath, it does not protect against humidity or moisture. Moisture intrusion is rapid and deteriorates insulation resistance to the point of thermocouple failure. Procedures for insulation resistance testing are to be found in ASTM documentation. Such testing is vital for mineral- insulated, metal-sheathed thermocouples stored in the open at construction sites or in repair shops for periods of time. Sheathing can also cause vapour transfer at elevated temperatures, which can result in thermocouple drift. This is the case with stainless steel sheaths used for nickel-bearing

thermocouples such as types K and N. The offending element is manganese, and the problem can be eliminated by specifying low-manganese, Inconel or modified Nicrosil sheathing.

Thermowell installation

Well installation should be planned so that the wells are readily accessible for servicing and so that the thermal elements can be withdrawn without obstruction by adjacent structures. It is desirable to install spare wells for calibration purposes so the working sensor can be compared to a standard temperature sensor.

Vertical installation is preferred for very high temperature use to prevent sagging. Horizontal installation, on the other hand, avoids some contamination by foreign materials. In horizontal installations, consideration should be given to the possibility that construction or repair personnel might misuse the wells as steps.

The thermowell immersion depth (U) should be sufficient to eliminate conduction error. A general rule is to use an insertion length equalling a minimum of 10 times the diameter of the protection tube or well. Another rule of thumb is to have the sensitive portion of the sensor immersed to a depth of a minimum of 3 in. (75 mm) plus the length of the sensitive portion. In the case of expansion bulbs, the immersion depth may be specified by the supplier or can be indicated on a calibration report, if furnished.

For pipe installations such as steam or hot water lines, insertion in an elbow on the axis of the pipe can permit an appropriate immersion depth should the diameter of the pipe be inadequate.

When installing thermowells at an angle or in elbows it is hard to maintain sanitary conditions, which is an important consideration in the food and pharmaceutical industries. Thermowells inserted into small pipes can also cause excessive flow restrictions, pressure drops, or even plugging.

Other considerations

The time constant of a thermal element increases with its mass. If time constants of a couple of seconds or less are required, unprotected, small diameter, bare thermal elements must be used. The addition of even a thin-walled sheath or thermowell increases the time constant to about 5 seconds as a minimum, and with larger bulbs it can reach 30 seconds. It is also important to maintain positive contact between the thermal element and the well in order to make sure that the temperature being measured reflects the temperature of the process on the outside and not the air temperature on the inside of the well.

If the process gases include hydrogen and a filled thermal element is used, it might be a good idea to vent the well, because hydrogen can damage the filling material.

Selection Guide for Thermowell Materials

Application Protecting Tube Material

Heat treating: Annealing Up to 704°C black steel Over 704°C Inconel 600a, Type 446 SS Carburizing hardening Up to 816°C black steel, Type 446 SS 1093°C Inconel 600, Type 446 SS Over 1093°C ceramicb Nitriding salt baths Type 446 SS Cyanide nickel (CP) Neutral Type 446 SS High speed ceramicb

Iron and steel: Basic oxygen furnace quartz Blast furnaces Downcomer Inconel 600, Type 446 SS Stove dome silicon carbide Hot blast main Inconel 600 Stove trunk Inconel 600 Stove outlet flue black steel Open hearth Flues and stack Inconel 600, Type 446 SS Checkers Inconel 600, Cermets Waste heat boiler Inconel 600, Type 446 SS Billet heating, slab heating and butt welding Up to 1093°C Inconel 600, Type 446 SS Over 1093°C silicon ceramic carbideb Bright annealing batch Top work temperature not required (use bare Type J thermocouple) Bottom work temperature Type 446 SS Continuous furnace section Inconel 600, ceramicb Forging silicon carbide, ceramicb Soaking pits Up to 1093°C Inconel 600 Over 1093°C silicon ceramic carbideb

Non-ferrous metals: Aluminium Melting cast iron (white-washed) Heat treating black steel Brass or bronze not required (use dip-type thermocouple) Lead Type 446 SS, black steel Magnesium black steel, cast iron Tin extra heavy carbon steel Zinc extra heavy carbon steel Pickling tanks chemical lead

Cement: Exit flues Inconel 600, Type 446 SS Kilns, heating zone Inconel 600

Ceramic: Kilns ceramicb and silicon carbidec Dryers silicon carbide, black steel Vitreous enamelling Inconel 600, Type 446 SS

Glass: Fore hearths and feeders platinum thimble Lehrs black steel Tanks Roof and wall ceramica Flues and checkers Inconel 600, Type 446 SS

Paper: Digesters Type 316 SS, Type 446 SS

Petroleum: De-waxing Types 304, 310, 316, 321, 347 SS, carbon steel

Towers Types 304, 310, 316, 321, 347 SS, carbon steel Transit lines Types 304, 310, 316, 321, 347 SS, carbon steel Factioning column Types 304, 310, 316, 321, 347 SS, carbon steel Bridgewall Types 304, 310, 316, 321, 347 SS, carbon steel

Power: Coal-air mixtures 304 SS Flue gases black steel, Type 446 SS Preheaters black steel, Type 446 SS Steel lines Types 347 or 316 SS Water lines low carbon steels Boiler tubes Types 304, 309, or 310 SS

Gas producers: Producer gas Type 446 SS Water gas Carburettor Inconel 600, Type 446 SS Superheater Inconel 600, Type 446 SS Tar stills low carbon steels

Incinerators: Up to 1093°C Inconel 600, Type 446 SS Over 1093°C ceramic (primary) silicon carbide (secondary)a

Food: Baking ovens black steel Charretort, sugar black steel Vegetables and fruit Type 304 SS

Chemical: Acetic acid 10 to 50% 21°C Type 304, Hastalloy Cd, Monel 50% 100°C Type 316, Hastalloy Cd, Monel 99% 21 to 100°C Type 430, Hastalloy Cd, Monel Alcohol, ethyl, methyl 21 to 100°C Type 304 Ammonia, all concentration 21°C Types 304, 316 SS Ammonium chloride, all concentration 100°C Type 316 SS, Monel Ammonium nitrate, all concentration 21 to 100°C Type 316 SS Ammonium sulphate, 10% to saturated 100°C Type 316 SS Barium chloride, all concentration, 21°C Monel, Hastalloy C Barium hydroxide, all concentration, 21°C low carbon steels Barium sulphite Nichromea, Hastalloy C Brines Monel Bromine tantalum, Monel Butadiene Type 304 SS Butane Type 304 SS Butyl acetate Monel Butyl alcohol copper, Type 304 SS Calcium chlorate, dilute, 21 to 66°C Calcium hydroxide Type 304 SS 10 to 20% 100°C Type 304 SS, Hastalloy C 50% 100°C Type 316 SS, Hastalloy C Carbolic acid, all, 100°C Type 316 SS Carbon dioxide, wet or dry 2017-T4 aluminum, Monel, nickel Chlorine gas Dry, 21°C Type 316 SS, Monel Moist, 7 to 100°C Hastalloy C Chromic acid, 10 to 50%,100°C Type 316 SS, Hastalloy C (all concentrations) Citric acid 15% 21°C Type 304 SS, Hastalloy C (all concentrations) 15% 100°C Type 316 SS, Hastalloy C (all concentrations) Concentrated, 100°C Type 316 SS, Hastalloy C (all concentrations) Copper nitrate Types 304 SS, 316 SS Copper sulphate Types 304 SS, 316 SS Cresols Type 304 SS Cyanogen gas Type 304 SS Dow thermf low carbon steels Ether Type 304 SS Ethyl acetate Monel, Type 304 SS Ethyl chloride, 21°C Type 304 SS, low carbon steel Ethyl sulphate, 21°C Monel Ferric chloride, 5% 21°C to boiling tantalum, Hastalloy C Ferric sulphate, 5% 21°C Type 304 SS Ferrous sulphate, dilute, 21°C Type 304 SS Formaldehyde Types 304 SS, 316 SS

Formic acid, 5% 21 to 66°C Type 316 SS Freon Monel Gallic acid, 5% 21 to 66°C Monel Gasoline, 21°C Type 304 SS, low carbon steel Glucose, 21°C Type 304 SS Glycerine, 21°C Type 304 SS Glycerol Type 304 SS Hydrobromic acid, 98% 100°C Hastalloy B Hydrochloric acid 1%, 5% 21°C Hastalloy C 1%, 5% 100°C Hastalloy B 25% 21 to 100°C Hastalloy B Hydrofluoric acid, 60% 100°C Hastalloy C, Monel Hydrogen peroxide, 21 to 100°C Types 316 SS, 304 SS Hydrogen sulphide, wet and dry Type 316 SS Iodine, 21°C tantalum Lactic acid 5% 21°C Type 304 SS, 316 SS 5% 66°C Type 316 SS 10% 100°C tantalum Magnesium chloride 5% 21°C Monel, nickel 5% 100°C nickel Magnesium sulphate, hot and cold Monel Muriatic acid, 21°C tantalum Naptha, 21°C Type 304 SS Natural gas, 21°C Types 304 SS, 316 SS, 317 SS Nickel chloride, 21°C Type 304 SS Nickel sulphate, hot and cold Type 304 SS Nitric acid 5% 21°C Types 304 SS, 316 SS 20% 21°C Types 304 SS, 316 SS 50% 100°C Types 304 SS, 316 SS 65% 100°C Type 316 SS Concentrated, 21°C Types 304 SS, 316 SS Concentrated, 100°C tantalum Nitrobenzene, 21°C Type 304 SS Oleic acid, 21°C Type 316 SS Oleum, 21°C Type 316 SS Oxalic acid 5% hot and cold Type 304 SS 10% 100°C Monel Oxygen 21°C steel Liquid SS Elevated temperatures SS Palmitic acid Type 316 SS Pentane Type 340 SS Phenol Types 304 SS, 316 SS Phosphoric acid 1%, 5% 21°C Type 304 SS 10% 21°C Type 316 SS 10% 100°C Hastalloy C 30% 21 to 100°C Hastalloy B 85% 21 to 100°C Hastalloy B Picric acid, 21°C Type 304 SS Potassium bromide, 21°C Type 316 SS Potassium carbonate, 1% 21°C Types 304 SS, 316 SS Potassium chlorate, 21°C Type 304 SS Potassium hydroxide 5% 21°C Type 304 SS 25% 100°C Type 304 SS 60% 100°C Type 316 SS Potassium nitrate 5% 21°C Type 304 SS 5% 100°C Type 304 SS Potassium permanganate, 5% 21°C Type 304 SS Potassium sulphate, 5% 21°C Types 304 SS, 316 SS Potassium sulphide, 21°C Types 304 SS, 316 SS Propane Type 304 SS, low carbon steel Progallic acid Type 304 SS Quinine bisulphate, dry Type 316 SS Quinine sulphate, dry Type 304 SS Seawater Monel or Hastalloy C Salicylic acid nickel

Sodium bicarbonate All concentration, 21°C Type 304 SS 5% 66°C Types 304 SS, 316 SS Sodium carbonate, 5% 21 to 66°C Types 304 SS, 316 SS Sodium chloride 5% 21 to 66°C Type 316 SS Saturated, 21 to 100°C Type 316 SS, Monel Sodium fluoride, 5% 21°C Monel Sodium hydroxide Types 304 SS, 316 SS, Hastalloy C Sodium hypochlorite, 5% still Type 316 SS, Hastalloy C Sodium nitrate, fused Type 316 SS Sodium peroxide Type 304 SS Sodium sulphate, 21°C Types 304 SS, 316 SS Sodium sulphide, 21°C Type 316 SS Sodium sulphite, 30% 66°C Type 304 SS Sulphur dioxide Moist gas, 21°C Type 316 SS Gas, 302°C Types 304 SS, 316 SS Sulphur Dry molten Type 304 SS Wet Type 316 SS Sulphuric add 5% 21 to 100°C Hastalloy B, 316 SS 10% 21 to 100°C Hastalloy B 50% 21 to 100°C Hastalloy B 90% 21°C Hastalloy B 90% 100°C Hastalloy D Tannic acid, 21°C Type 304 SS, Hastalloy B Tartaric acid 21°C Type 304 SS 66°C Type 316 SS Toluene 2017-T4 aluminum, low carbon steel Turpentine Types 304 SS, 316 SS Whiskey and wine Type 304 SS, nickel Xylene copper Zinc chloride Monel Zinc sulphate 5% 21°C Types 304 SS, 316 SS Saturated, 21°C Types 304 SS, 316 SS 25% 100°C Types 304 SS, 316 SS

aTrademark of the international Nickel Co. bDue to susceptibility to cracking, sudden thermal shocks should be avoided. cDue to susceptibility to cracking, sudden thermal shocks should he avoided. dTrademark of the Cabot Corp. eTrademark of the Driver-Harris Co. fTrademark of the Dow Chemical Corp.

Table 4.2 Selection guide for thermowell materials

4.7 Other factors

Over the years solutions have been found to temperature measurement problems in an astonishing variety of challenging situations. For example, to measure the internal temperature of a person or animal there are 'pills' which are swallowed and which send radio signals indicating the temperature inside the person or animal. Similarly, it is possible to measure the temperature of molten steel at around 1600ºC to within a few degrees Celsius using disposable thermocouple probes. Thus it is likely that a solution to your particular temperature measurement problem already exists.

5. REFERENCES

1. Changing to the Metric System: Conversion Factors, Symbols and Definitions. Anderton P & Bigg P H. National Physical Laboratory

2. Preston-Thomas H., The International Temperature Scale of 1990 (ITS-90), Metrologia, 1990, 27, 3-10 and 107

3. Supplementary information for the International Temperature Scale of 1990, approved and updated periodically by the CCT and published by the BIPM

4. Techniques for approximating the International Temperature Scale of 1990, approved and updated periodically by the CCT and published by the BIPM

5. NPL Measurement Services Booklet: Temperature

6. Temperature, T J Quinn, Academic Press, 2nd edition, 1990

7. Traceable Temperatures, J V Nicholas and D R White, John Wiley and Sons, 1994

8. Instrument Engineers’ Handbook (Third Edition) – Process Measurement and Analysis. Liptak B G

9. Evans, J. P. Temperature, Its Measurement and Control in Science and Industry (American Institute of Physics, New York, USA 1982) Vol. 5, Part 2, p.771

10. Berry, R. J. Temperature, Its Measurement and Control in Science and Industry (American Institute of Physics, New York, USA 1982) Vol. 5, Part 2, p.743

11. Connolly, J. J. Temperature, Its Measurement and Control in Science and Industry (American Institute of Physics, New York, USA 1982) Vol. 5, Part 2, p.815

12. Bass, N. M. Temperature, Its Measurement and Control in Science and Industry (American Institute of Physics, New York, USA 1982) Vol. 5, Part 2, p.813

13. Wood, S. D., Mangum, B. W., Fillen, J. J., Tillett, S. B., J. Res Nat. Bur. Stand. 83 (3), May- June 1978, p247-263

14. Kusters N. C. and MacMartin M. P., 1970 IEEE Transactions IM-19 291-97

15. Hill J. J. and Miller A. P. 1962 Proceedings I.E.E. 109B, 157-62

16. Kibble B.P. and Rayner G.H., 'Coaxial a.c. Bridges', Adam Hilger Ltd, Bristol, 1984

17. Thompson A. M. and Small C. W., 1971 Proceedings I.E.E. 118 1662-66

18. Wolfendale, P. C. F, Yewen, J. D., Daykin, C. I, Temperature, Its Measurement and Control in Science and Industry (American Institute of Physics, New York, U.S.A. 1982) Vol. 5, Part 2, p.729

19. Manual on the use of thermocouples in temperature measurement. ASTM Special Technical

Publication 470A, American Society for Testing and Materials, Philadelphia

20. Thermocouple Temperature Measurement: P A Kinzie, John Wiley & Sons Inc. 1973

21. The Nicrosil versus Nisil Thermocouple: Properties and thermoelectric reference data, NBS Monograph 161 National Bureau of Standards, US Dept of Commerce, 1978

22. Applications of Radiation Thermometry, ASTM Special Technical Publication 895, 1985

23. Theory and Practice of Radiation Thermometry, edited by D P DeWitt and C D Nutter, Wiley Interscience 1989

24. Tables of Physical and Chemical Constants, 16th Ed., Longmans. 1995. Kaye G W C and Laby T H

APPENDICES

APPENDIX A

RELEVANT BRITISH/EUROPEAN STANDARDS

BS 1041-2.1:1985 Code for temperature mearement. Expansion thermometers. Guide to selection and use of liquid-in-glass thermometers

BS 1041-2.2:1989 Code for temperature mearement. Expansion thermometers. Guide to selection and use of dial-type expansion thermometers

BS 1041-3:1989 Temperature measurement. Guide to selection and use of industrial resistance thermometers

BS 1041-4:1992 Temperature measurement. Guide to selection and use of thermocouples

BS 1041-5:1989 Temperature measurement. Guide to selection and use of radiation pyrometers

BS 1041-7:1988 Temperature measurement. Guide to selection and use of temperature/time indicators

BS 1900:1976, Secondary Reference Thermometers. High stability thermometers suitable for standards, total immersion only

BS 593:1989, Laboratory Thermometers. Good quality instruments, medium accuracy, total and partial immersion

BS 791:1990, Calorimeter Thermometers. High precision, 760 mm long, 6°C ranges between 9°C and 45°C only; total and partial immersion

BS 1365:1990, Short-range, short-stem thermometers, 220 mm long. Includes 'Anschutz' type sets enabling 0 to 360°C to be covered by 7 thermometers, total and partial immersions

BS 1704:1985, General Purpose thermometers. An assortment of temperature ranges within the span –120°C to +510°C

The Institute of Petroleum publishes a handbook including about 90 thermometers made to specialised requirements within the range –80°C to +400°C, for total and partial immersion. This is also published as BS 2000:Part 0:Addendum 1

British Standard 1904: 1984, now superseded by BS EN 60751:1996

International Electrotechnical Commission standard IEC 751:1983, including Amendments 1 and 2

IEC 584 'Thermocouples': Part 1 (1995) 'Reference Tables', Part 2 (1982) 'Specification for Thermocouple Tolerances', Part 3 (under review) 'Extension and Compensating Cables. Tolerances and Identification System'. To replace the various parts of BS 4937 as BS EN 60584

Annual Book of ASTM Standards: Vol. 14.01

Temperature-EMF Reference Functions for the Letter-designated Thermocouple Types based on the ITS-90, National Institute of Standards and Technology Monograph 175, 1993

APPENDIX B

HISTORICAL EVOLUTION OF THE THERMODYNAMIC AND PRACTICAL TEMPERATURE SCALES

The normal hydrogen scale

1878 Preparation of metre prototypes was begun. Each national prototype metre was furnished with two mercury-in-glass thermometers, calibrated at the BIPM. The thermometers were made, to the order of the BIPM, by an instrument maker in Paris called Tonnelot. The thermometers were made of verre dur, a particularly good glass from the point of view of stability, and a reproducibility of measurement of a few thousandths of a degree was possible. It became urgent to establish a uniform scale of temperature against which they could be calibrated.

1884–87 Chappuis, at the BIPM, worked to relate the readings of the very best mercury-in- glass thermometers to absolute (i.e. thermodynamic) temperatures. In the first part of his study he considered in detail the constant-volume gas thermometer, using, in turn, hydrogen, nitrogen and carbon dioxide as the working fluids. The estimated uncertainty of his measurements was better than one-hundredth of a degree over most of the range studied, to 100°C.

1887 The CIPM adopted the constant-volume hydrogen scale (called the normal hydrogen scale), based upon fixed points at the ice point (0 °C) and the steam

point (100 °C) as the practical scale for international metrology. This decision

was ratified by the 1st CGPM in 1889.

Chappuis continued his work at the BIPM with an investigation of the constant- 1888–89 pressure gas thermometer using the same three gases. He concluded that the

constant-volume thermometer provided a more convenient practical standard than did the constant pressure thermometer. At the instigation of Griffiths, of Kew Observatory, UK, the work was pursued, using a constant- volume thermometer, to extend the temperature range to higher temperatures. In collaboration with Callendar, Griffiths had been developing a platinum resistance thermometer which was stable to at least 600 °C. Callendar and Griffiths used the boiling point of sulphur, which they deduced to be 444.53 °C, as a third fixed point for calibration, and proposed to the BIPM that a comparison be made between their platinum resistance thermometers and the constant-volume gas thermometer of Chappuis.

This comparison was carried out by Chappuis, in collaboration with Harker of the 1897 Kew Observatory. It involved the establishment of a constant-volume nitrogen scale up to the boiling point of sulphur. The Chappuis/Harker measurement of the sulphur point led to a value of 444.70 °C, in very close agreement with the earlier result of Callendar and Griffiths.

The International Temperature Scale of 1927

1889 Many freezing and boiling points were measured during the last two decades of the nineteenth century. Callendar gave a detailed review of gas thermometry at the 1899 meeting of the British Association for the Advancement of Science

100

(BAAS), when he made a proposal for a practical temperature scale. He proposed that a platinum resistance thermometer be adopted as the defining instrument of the scale, and that it be calibrated at the freezing point of water and the boiling points of water and sulphur. Further, he proposed that a particular batch of platinum wire be selected from which the thermometers defining the scale be manufactured. It was his intention that such a scale be called the British Association Scale of Temperature and that it be related to the ideal temperature scale through chosen gas thermometer measurements of the sulphur point. It is not clear why the British Association did not take up his proposals.

1911 The Physikalish-Technische Reichanstalt (PTR, later to become the PTB), Berlin, addressed a circular letter to the BIPM, the National Physical Laboratory (NPL), Teddington, and the Bureau of Standards (BS, which in 1934 become the National Bureau of Standards, NBS, and in 1986 the National Institute of Standards and Technology, NIST), Washington, suggesting that the

thermodynamic scale be adopted as the International Temperature Scale, and

that a practical realization of it be the 1899 proposal of Callendar. Both the NPL

and the BS agreed, the constants of the platinum were specified, and it was

proposed that above the upper limit (1100 °C) the scale be defined in terms of

the optical pyrometer.

At the 5th CGPM, every encouragement was given to this initiative and a 1913 Resolution was adopted, asking the Directors of the three laboratories to meet with the aim of coming to a firm agreement on such a scale. The planned meeting did not take place, however, owing to the outbreak of the First World War.

By the time discussions resumed, the three national laboratories had put into 1923 operation a platinum resistance thermometer scale covering the range from –38 °C, the freezing point of mercury, to 444.5 °C, the boiling point of sulphur, using a quadratic interpolation formula. During the course of a visit to the NPL and the PTR by a representative of the BS, the basis of an international scale was agreed upon. It was to consist of a platinum resistance thermometer to cover the range up to 650 °C, calibrated at 0 °C, 100 °C and the boiling point of sulphur at 444.5 °C. Between 650 °C and 1100 °C the scale was to be defined by a Pt-10 % Rh/Pt thermocouple calibrated at the freezing points of zinc, antimony, silver and gold and using a cubic interpolation formula. Above the gold point, 1063 °C, an optical pyrometer was proposed.

This informal agreement was followed by wider discussions in which the BIPM and the University of Leiden also participated.

In 1925 a draft proposal was drawn up, to be put to the CIPM in 1927. In this the 1925 range of the platinum resistance thermometer was extended down to 193 °C, and the cubic equation of the thermocouple was replaced by a quadratic equation with calibration points at the freezing points of antimony (630 °C), silver (960 °C), and gold.

The 7th CGPM adopted the International Temperature Scale of 1927, which 1927 differed very little from the draft of 1925. It was planned to hold an International Thermometry Conference in 1928, at which the question of the status of the International Temperature Scale would be examined in more detail. This Conference, however, did not take place.

101

The Evolution of ITS-27 and its modifications in 1948

1937 The CIPM established a Consultative Committee on Thermometry and Calorimetry to advise it on matters concerned with these subjects. Since then, it has been the Consultative Committee on Thermometry (CCT) that has largely taken the initiative in matters concerned with the evolution of the International Temperature Scale.

1948 The first revision of the ITS took place in 1948. In this revision, the only change below 0 °C was the disappearance of the extrapolation below the oxygen point, to –190 °C, which had been found to be unreliable. The IPTS-48 extended down only to –182.97 °C. The junction between the resistance thermometer and the thermocouple was changed from 660 °C to the freezing point of antimony, 630.5 °C, and the temperature assigned to the silver point was increased slightly, from 960.5 °C to 960.8 °C. It was also decided to drop the name "degree Centigrade" for the unit and replace it by degree Celsius.

The 1958 4He and 1962 3He vapour pressure scales

The saturated-vapour pressure/temperature relation for liquid helium provides such a good and reproducible scale that its use as such long pre-dates any internationally agreed scale in the helium range; in fact, it even pre-dates ITS-27. However, it proved difficult to reach international agreement on a helium vapour pressure scale.

1958 The CIPM adopted a Table of 4He vapour pressure against temperature data, proposed to it by the CCT. The Table was based upon gas thermometry data smoothed by magnetic thermometry and, below 2.2 K, by thermodynamic calculations. It covered the range from 0.5 K to 5.23 K and became known as the 4 1958 He Scale, temperatures measured on it being denoted by T58.

1962 Shortly after the adoption of the 1958 4He Scale, a further proposal was made in 3 respect of a vapour pressure scale for He. This was based upon comparison of the vapour pressures of 3He with the 1958 4He Scale above 0.9 K, and with thermodynamic calculations below 0.9 K. The Scale was accepted by the CIPM and become known as the 1962 3He Scale, temperatures measured on it being denoted by T62.

Complete revisions have since been made of both the 3He and 4He scales, making them consistent with the results of gas, noise, acoustic and magnetic thermometry carried out since the original versions of the scales were adopted.

The 1976 Provisional 0.5 K to 30 K Temperature Scale (EPT-76)

1976 In 1976 the CIPM approved a new low-temperature scale called the 1976 Provisional 0.5 K to 30 K Temperature Scale, or EPT-76. Its purpose was to provide a unified Scale upon which temperature measurements could be made in this range, pending the revision and downward extension of the IPTS-68. It was defined in terms of the temperatures assigned to eleven fixed points within the range 0.5 K to 30 K, together with the differences between T76 and the following existing scales: IPTS-68; the 4He-1958 and 3He-1962 vapour pressure scales; NPL-75 and the NBS version of IPTS-68 which was defined by difference from NBS-55. In contrast to IPTS-68, the EPT-76 could thus be realized in a number

102

of ways; either by using one of the above scales and the tabulated differences given in the text of EPT-76 or by using a thermodynamic interpolating thermometer, such as a gas thermometer or magnetic thermometer, calibrated at one or more of the specified reference points of EPT-76.

The International Practical Temperature Scale of 1968 (IPTS-68)

1954 The 10th CGPM finally adopted a proposal that Kelvin had made back in 1854, namely that the unit of thermodynamic temperature be defined in terms of the interval between the absolute zero and a single fixed point. The fixed point chosen was the triple point of water, which was assigned the thermodynamic temperature of 273.16 °K.

The proposal had already been made in 1948, but at that time there was still a divergence of view as to what value should be assigned to the absolute zero. The question was finally resolved by the CGPM in 1954.

1961 In 1961 it was agreed that the NPL and the Physicotechnical and Radiotechnical Measurements Institute (PRMI), Moscow, would undertake a comparison of platinum resistance thermometers calibrated on four of the most important gas thermometer scales. These were the NPL(1961), NBS(1955), PRMI(1954) and Pennsylvania State University PSU(1954) scales. The results of the comparison provided the basis for the eventual low-temperature part of IPTS-68.

The NBS-55 scale is of particular note since it is an example of the way in which

a so-called "wire-scale" can be successfully operated. NBS-55 is a scale based

upon gas thermometry carried out in 1939. It was originally maintained on a

group of six platinum resistance thermometers, and was known as NBS-39. In

1955 an arbitrary shift of 10 mK was made over the whole of the scale and the name was changed to NBS-55. The successors to these six original NBS-39 thermometers continued to be used to maintain an NBS version of IPTS-68.

The CCT defined a reference function W for interpolation between a number of 1964 low-temperature fixed points. The CCT-64 was published as a table under the title "Provisional reference table CCT-64 of W against T for platinum resistance thermometers in the range 12 K to 273.15 K".

A CCT Working Group proposed a 1966 Provisional Scale, taking into account 1966 further gas thermometry results for the oxygen boiling point and the hydrogen triple point.

The second revision of the Temperature Scale took place, and resolved the 1968 curious situation that thermodynamic temperatures were defined in quite a different way from International Practical Temperatures. In the IPTS-68, both thermodynamic and practical units were defined to be identical and equal to 1/273.16 of the thermodynamic temperature of the triple point of water. The unit itself was renamed "the kelvin" in place of "degree Kelvin" and designated "K" in place of "°K".

In outline the IPTS-68 was made up of four parts: (a) between 13.81 K and 273.15 K; (b) 0 °C to 630.74 °C; (c) 630.74 °C to 1064.43 °C; and

103

(d) above 1064.43 °C.(*) In part (a) the Scale was defined in terms of a set of six low-temperature fixed points together with a reference function. In the range (b) the Scale was defined in terms of the old Callendar quadratic equation, but modified to take account of new gas thermometry values for the fixed points. In part (c) the defining instrument was the Pt-10 % Rh/Pt thermocouple, calibrated at 630.74 °C and the freezing points of silver and gold, and using a quadratic interpolation formula. Part (d) was defined in terms of the radiation emitted by a black body and described by Planck's equation.

(*) Thermometrists generally refer to temperatures below 0 °C in kelvin, and those above in degrees Celsius.

The International Temperature Scale of 1990 (ITS-90)

1990 The International Temperature Scale of 1990 (ITS-90) came into effect on 1 Janurary 1990, replacing the IPTS-68 and the EPT-76.

The ITS-90 differs from the IPTS-68 in a number of important respects:

it uses the triple point of water (273.16 K), rather than the freezing point of water (273.15 K), as a defining point;

it extends to lower temperatures: 0.65 K instead of 13.8 K;

it is in closer agreement with thermodynamic temperatures;

it has improved continuity and precision;

it has a number of overlapping ranges and sub-ranges;

in certain ranges it has alternative but substantially equivalent definitions;

it includes the helium vapour pressure scales;

it includes an interpolating gas thermometer as one of the defining instruments;

the range of the platinum resistance thermometer as defining instrument has been extended from 630 °C up to the silver point, 962 °C;

the Pt/10 % Rh-Pt thermocouple is no longer a defining instrument of the scale;

the range based upon the Planck radiation law begins at the silver point instead of at the gold point, but options exist for using any one of the silver, gold or copper points as reference points for this part of the scale.

For further details please refer to the following BIPM publications: Preston-Thomas H., The International Temperature Scale of 1990 (ITS-90), Metrologia, 1990, 27, 3-10 and 107; Techniques for approximating the International Temperature Scale of 1990; Supplementary information for the International Temperature Scale of 1990.

104

APPENDIX C

THE ROLE OF THE PLATINUM RESISTANCE THERMOMETER IN THE ITS-90

From the triple point of equilibrium hydrogen (13.8033 K) to the freezing point of silver (961.78°C) the ITS-90 is realised using standard platinum resistance thermometers. The interpolation equations are written in terms of T90/K or t90/°C, and the resistance ratio:-

W(T90) = R(T90)/R(273.16 K) or, equivalently, W(t90) = R(t90)/R(0.01°C)

Note that this ratio differs from that used in IPTS-68 in which the denominator was R(0°C) or R(273.15K): W(t90) is equal to 0.999960 x W(t68).

To qualify for use in a realisation of ITS-90, the resistance ratio of a platinum resistance thermometer must be not less than 1.11807 at the melting point of gallium (29.7646°C), or alternatively not greater than 0.844235 at the triple point of mercury (-38.8344°C). A thermometer for use to the freezing point of silver must also satisfy the requirement that W(961.78°C) is not less than 4.2844. Two reference functions are specified, one which relates Wref(T90) to T90/K below 273.16 K (0.01°C), and the other which relates Wref(t90) to t90/°C above 0°C. These are given overleaf.

The calibration of a qualifying platinum resistance thermometer is expressed in terms of the deviation of the resistance ratio for the thermometer from that given by the appropriate reference function. The coefficients of the deviation equation are calculated from measurements of the thermometer resistance at fixed points specified according to the range of the calibration. The full text of the ITS-90 should be consulted for details. The deviation equations are:-

1. For ranges above 0°C

2 3 2 W - Wref = a(W - 1) + b(W - 1) + c(W - 1) + d(W - W(660.323°C)) where W and Wref are the thermometer and reference resistance ratios, respectively. The coefficients a, b and c are determined from measurements at the freezing points of tin, zinc and aluminium and d, which is only required for the range above 660.323°C and is zero below this temperature, is determined from a measurement at the freezing point of silver.

ITS-90 also permits two-point realisations, with c = d = 0 and upper limits at the freezing points of zinc or tin, and single-point realisations, with b = c = d = 0 and upper limits at the freezing point of indium or the melting point of gallium.

2. For the range 83.8058 K to 273.16 K (-189.3442°C to 0.01°C)

W - Wref = a(W - 1) + b(W - 1)InW where the coefficients are determined from measurements at the triple points of argon and mercury.

3. For the range 13.8033 K to 273.16 K 5 2 (i+2) W - Wref = a(W - 1) + b(W - 1) + ci (InW) i=1 where the coefficients are determined from measurements at seven specified fixed points. ITS-90 also provides for realisations extending down to 54.3584 K and 24.5561 K, with fewer coefficients.

The calculation of temperatures from measured values of resistance ratio requires first the calculation of Wref from the appropriate deviation equation, and then the calculation of t90 (or T90) by iteration from

105

the appropriate reference function (or directly from its inverse approximation). The reference functions are:-

1. For the range 13.8033 K to 273.16 K 12 i In(Wref) = A0 + Ai ((In(T90/273.16 K) + 1.5)/1.5) i=1 or its inverse (accurate to ±0.0001 K) 15 1/6 i T90/273.16 K = B0 + Bi (((Wref) - 0.65)/0.35) i=1 2. For the range 0°C to 961.78°C 9 i Wref = C0 + Ci (((t90/°C) – 481)/481) i=1 or its inverse (accurate to ±0.00013°C) 9 i t90/°C = D0 + Di (((Wref) - 2.64)/1.64) i=1

The constants are as follows:

i Ai Bi Ci Di 0 -2.135 347 29 0.183 324 722 2.781 572 54 439.932 854 1 3.183 247 20 0.240 975 303 1.646 509 16 472.418 020 2 -1.801 435 97 0.209 108 771 -0.137 143 90 37.684 494 3 0.717 272 04 0.190 439 972 -0.006 497 67 7.472 018 4 0.503 440 27 0.142 648 498 -0.002 344 44 2.920 828 5 -0.618 993 95 0.077 993 465 0.005 118 68 0.005 184 6 -0.053 323 22 0.012 475 611 0.001 879 82 -0.963 864 7 0.280 213 62 -0.032 267 127 -0.002 044 72 -0.188 732 8 0.107 152 24 -0.075 291 522 -0.000 461 22 0.191 203 9 -0.293 028 65 -0.056 470 670 0.000 457 24 0.049 025 10 0.044 598 72 0.076 201 285 11 0.118 686 32 0.123 893 204 12 -0.052 481 34 -0.029 201 193 13 -0.091 173 542 14 0.001 317 696 15 0.026 025 526

The reference resistance ratios at the defining fixed points are:

T90/K Wref (T90) t90/°C Wref (t90) 13.8033 0.001 190 07 0.01 1.000 000 00 17.035 0.002 296 46 29.7646 1.118 138 89 20.27 0.004 235 36 156.5985 1.609 801 85 24.5561 0.008 449 74 231.928 1.892 797 68 54.3584 0.091 718 04 419.527 2.568 917 30 83.8058 0.215 859 75 660.323 3.376 008 60 234.3156 0.844 142 11 961.78 4.286 420 53 273.16 1.000 000 00

106

APPENDIX D

TOLERANCE CLASSES FOR Pt100 THERMOMETERS (IEC 751: 1983)

107

APPENDIX E

Pt100 REFERENCE TABLES (IEC 751: 1983)

108

109

APPENDIX F

TOLERANCE CLASSES FOR THERMOCOUPLES (IEC 584-2: 1982)

110

APPENDIX G

THERMOCOUPLE REFERENCE TABLES (IEC 584-1:1995)

111

112

113

114

115

116

117

118

119

120

121

122

123

124

125

126

127

128

129

130

131

132

APPENDIX H

DATA TABLES FOR RADIATION THERMOMETERS OR PYROMETERS

Typical emissivity values (based on BS 1041 : Part 5)

Surface Emissivity at the wavelengths indicated 0.65 m 0.9 m 2 m

Aluminium – polished - 0.05-0.25 0.04-0.20 Aluminium – oxidised - 0.20-0.40 0.20-0.40 Steel – polished - 0.40-0.50 0.20-0.25 Steel - bright rolled - 0.45-0.55 0.25-0.30 Steel - 'blued' - 0.75-0.85 0.45-0.75 Steel – oxidised 0.80-0.90 0.80-0.90 0.75-0.85 Steel – liquid 0.36-0.40 0.32-0.36 0.25-0.30 Copper – polished 0.10-0.12 0.08-0.12 0.06-0.10 Copper – oxidised 0.60-0.80 0.50-0.75 0.20-0.70 Carbon and graphite 0.75-0.90 0.75-0.90 0.70-0.90 Silicon carbide 0.70-0.85 0.75-0.90 0.80-0.90 Red brick 0.70-0.80 0.70-0.80 0.80-0.90 Silica brick 0.40-0.70 0.40-0.70 0.40-0.70 Ceramics 0.20-0.50 0.30-0.50 0.40-0.60

The total emissivity of almost any opaque plastic sheet will fall within the range 0.7 to 0.95

Corrections in °C to be added to radiation thermometer readings for departures from blackbody conditions

Emissivity at the wavelengths indicated Thermometer 0.65 m 0.9 m 2.0 m reading/°C 0.4 0.6 0.8 0.4 0.6 0.8 0.4 0.6 0.8

200 ------30 16 7 500 - - - 36 20 8 84 45 19 1000 71 39 17 100 54 23 246 127 52 1500 140 76 32 201 107 45 - - - 2000 236 126 53 341 178 74 - - -

133

Errors in temperature arising from an error of 10% in emissivity

Thermometer Error in temperature at the wavelength indicated reading/°C 0.65 m 0.90 m 2.0 m 8-14 m

200 - - 3 17 500 - 3 8 46 1000 7 10 23 - 1500 14 20 44 - 2000 23 32 72 -

These corrections should be added to the reading if the emissivity is lower than supposed.

134

APPENDIX I

EXAMPLE INSTRUMENT SPECIFICATION: LABFACILITY TEMPMASTER 100 DIGITAL THERMOMETER

135

136

137

138

APPENDIX J

PARTIAL LIST OF UK MANUFACTURERS AND SUPPLIERS OF TEMPERATURE INSTRUMENTATION

A J Thermosensors Ltd,(UK) West Sussex " Wherever a temperature measurement is required, AJT can supply both the sensor and the instrumentation to suit the application"; Thermocouples; RTDs, and more. ARi Industries Ltd(UK) Biodata: Measurement and Control (UK) Branom/LITERATURE/thermocouple Brearley (U.K.) Brearley, part of RM&C Ltd, is a leading ISO 9001/2000 manufacturer of all types of thermocouples, MIs RTDs, thermowells and associated accessories such as extension cable, plugs and sockets. Contact [email protected] or visit their website at www.controlsdirect.com. Hewitt Industries, Manufacturers of thermocouples I2R New Products makers of thermocouples, measuring and heating systems and controls. Jumo Process Control Food thermometers, thermocouples, control and more Labfacility Temperature Newsletter from the UK Peak Sensors (UK) A TC maker and instrumentation supplier in Chesterfield, Derbyshire . SMI - The Medical Marketplace - Cryosurgical Instruments TC Ltd (UK) Omega Engineering (USA,UK and elsewhere) The largest temperature sensor seller in the USA, famous for their huge catalogs and wide variety of products. Land Infrared in The UK (nr Sheffield) with offices and service centers in the USA, Germany, France, Italy and Japan (to name a few) Maker and reseller of a wide range of Spot, Line and Area measuring thermometers including the precision handheld spot units (called by the name of an ancient one-eyed Greek mythological monster) made by Minolta Camera Company of Japan. Also has a wide range of calibration equipment. Thermotenix UK Provides Scanner software, a new line scanning RT and a hand-held thermal imager with manfeatures. Hanna Instruments products Provides the Model HI 9040, a hand-held digital thermometer that uses an interchangeable probe with thermistor sensor to deliver fast response and high accuracy. for the range:-58.0° to 302°F (-50.0 to 150.0°C). A wide variety of interchangeable probes for applications such as air, liquid and penetration are available.

139