22 Calibration and Testing of Temperature Measuring Instruments 22 .1 Definitions and Terminology The following main terms are used in the calibration and testing of temperature measuring instruments, necessary for the maintenance and dissemination of ITS-90 . " Calibration of a thermometer is the sum of activities concerned with the determination of its thermometric characteristics . These characteristics define the function correlating the chosen property of the thermometer with the temperature . If a thermometer directly indicates the measured temperature, its calibration depends on correlating certain numerical values with the scale graduation . For example, this concerns liquid-in-glass thermometers . " Testing a thermometer is the sum of activities concerned with verifying that the thermometer complies with the relevant regulations . " Primary standards are thermometers used for reproduction of ITS-90, as well as for international comparisons . " Transfer standards are thermometers used for the transfer of temperature units to other thermometers, which thus have lower accuracy than these standards . They comprise secondary, tertiary and other standards, which occupy important transfer levels in what is called the chain of traceability of standards . " Working standards are thermometers destined for the calibration of other working standards, situated lower in the traceability hierarchy . They are also used in the calibration of industrial thermometers . " Industrial thermometers are thermometers used in the day-to-day practice of temperature measurement . " Laboratory thermometers are thermometers used in laboratories . Calibration and testing procedures for thermometers comprise a general scheme which defines the hierarchy of thermometers . They also determine the methods for and accuracy Temperature Measurement Second Edition L. Michalski, K. Eckersdorf, J. Kucharski, J. McGhee Copyright © 2001 John Wiley & Sons Ltd ISBNs: 0-471-86779-9 (Hardback); 0-470-84613-5 (Electronic) 420 CALIBRATION AND TESTING of, transferring the temperature unit from primary standards to industrial thermometers (Richardson, 1962 ; Gray and Chandon, 1972 ; Gray and Finch, 1972) . Most industrialised countries have established national standard laboratories which are equipped for reproducing ITS-90 through the calibration and testing of standard thermometers . The following are among the more well known Laboratories : " Institut National de M6trologie (INM), France, " Institute of Metrology "Mendeleiew", Russia, " Istituto di Metrologia "G . Collonetti" (IMGC), Italy, " National Institute of Standards and Technology (NIST), USA, " National Institute of Metrology (NIM), China, " National Physical Laboratory (NPL), UK, " National Research Council of Canada (NTC), Canada, " Physikalisch-Technische Bundesanstalt (PTB), Germany . Industrial thermometers are tested in regional and industrial laboratories . 22 .2 Fixed Points of ITS-90 22 .2 .1 General information Fixed points which define ITS-90 are given in Chapter 1 . Cryogenic fixed points in the range from 13 .8033 K (-259 .3467 °C) to 83 .8058 K (-189 .3442 °C) are used for calibration of capsule-type platinum resistance sensors . They consist of two boiling points of H2 and the four triple points of H2, Ne, OZ and Ar . Although the triple point of Ar is often used in the calibration of long-stem platinum resistance sensors, the triple point of OZ is very rarely used . Cryogenic fixed points, below the Ar triple point, are not dealt with, as this book does not cover extremely low temperatures . Fixed points in the range from 234 .3156 K (- 38 .8344 °C) to 1234 .93 K (961 .78 °C) are used for calibration of long-stem and high temperature platinum resistance sensors . In this range there are two triple points, of Hg and H2O, one melting point, of Ga, and five freezing points, of In, Sn, Zn, Al and Ag . Fixed points in the range from 961 .78 °C to 1084 .62 °C, which are used for calibration of pyrometers, consist of the three freezing points of Ag, Au and Cu . Construction of a fixed point has to be adapted to the kind of calibrated sensor . When calibrating resistance thermometer sensors, it is extremely important to provide as good a heat transfer as possible between the sensor and the medium applied in the fixed point . In the calibration of pyrometers, the fixed points used should have all the properties of a black body . FIXED POINTS OF ITS-90 421 22 .2 .2 Realisation of fixed points Some of the more important and commonly used fixed points of Table 1 .1 will now be described . The triple point of argon 83 .8058 K (-189.3442 °C), for the calibration of long-stem platinum sensors is shown in Figure 22 .1 (Bonnier, 1987) . Argon of 99 .9999 % purity is kept in a sealed cell of stainless steel, immersed in liquid nitrogen . This cell should be able to withstand pressures up to a maximum of 10 MPa, which occurs at maximum ambient temperature . The triple point of argon is attained in the melting process from the solid state . After supplying each pulse of energy to the cell with frozen argon, the cell temperature is measured . Until the melting process commences, only smaller and smaller steps of energy are supplied . Pavese et al . (1984) point out that the temperature of the triple point of argon can then be determined by observing the temperature as a function of time during the meltingprocess . The protecting tube, in which the calibrated sensor is placed, is in direct contact with the liquid and solid phase of argon . The triple point of mercury, 234 .3156 K (-38 .8344 °C), is realised in a glass cell as shown in Figure 22 .2 (Preston-Thomas et al ., 1990) . A cell with mercury is thermally insulated and placed in a vacuum stainless steel sheath . The degree of vacuum between the cell and the sheath is controlled to achieve the desired insulation between the cell and the surroundings . For calibration the RTD sensor is placed in the protecting tube, which is filled with ethyl alcohol to enhance the thermal contact between both of them . Very high purity mercury (1 to 10 8 ratio) allows the triple point of mercury to be obtained during either melting or solidifying . The realisation of the Hg triple point, during solidification, proceeds as follows . The steel sheath is cooled down by a mixture of solidified CO Z and ethyl alcohol . At the moment when the ampule temperature attains the mercury solidifying point, the air contained between the cell and its sheath is pumped off . A rod, cooled in liquid HELIUM GAS INLET LONG-STEM PLATINUM RESISTANCE SENSOR MANOMETER \ ^/ VALVE FILLING TUBE FOR LIQUID NITROGEN SENSORTUBE POLYURE- THANE FOAM - STAINLESS STEEL VESSEL LIQUID NITROGEN - - SOLID -LIQUID ARGON - CRYOSTAT 7 7 7 - Figure 22 .1 A cell for the triple point of argon (Bonnier, 1987) 422 CALIBRATION AND TESTING VACUUM SYSTEM VALVE RING TUBE SEAL SENSORTUBE ETHYL ALCOHOL INDIUM SEAL PAPER INSULATION STAINLESS STEEL JACKET CONNECTION FOR CLEANING AND FILLING PAPER INSULATION Cu FOIL CYLINDER BOROSILICATE GLASS CELL MERCURY QUARTZ WOOL STAND AL-SILICATE INSULATION 50mm Figure 22 .2 Triple point of mercury (Preston-Thomas et al ., 1990) nitrogen, is then introduced in place of the sensor . This creates a layer of solidified mercury in the ampule . In the final stage of this procedure, the calibrated sensor, which has been cooled down beforehand in a mixture of solidified C0 2 and ethyl alcohol, is then introduced in place of the rod . The triple point of mercury in solidification is reproduced with a precision better than f0 .1 mK . Different waysof realising the Hg triple point as well as their precision are given by Hermier and Bonnier (1992) and by Furukawa(1992) . The triple point of water, 273 .16 K (0 .01 °C) , which is shown in Figure 22 .3, is realised in a sealed glass cell filled with distilled water under vacuum . After cooling the water down to about 0 °C a layer of some millimetres of ice is formed around the inside of the tube by means of the powdered solid C02 . With the solid C02 removed, the inner tube is filled with water at about +20 °C for a short time, until a thin layer of ice is melted and replaced bya thin layer of water . The inner tube is then filled by water at 0 °C to enhance the heat transfer to the calibrated thermometer . Keeping the cell in the ice-water mixture maintains the triple point temperature with an accuracy better than t0 .1 mK to f0 .3 mK for FIXED POINTS OF ITS-90 423 at least 24 hours . A more detailed description is given by Stimson (1956), Hall and Barber (1964) and Furukawa and Bigge (1982) . The melting point of gallium 29 .7646 °C, is realised in the apparatus built in NPL (Chattle and Pokhodun, 1987) shown in Figure 22 .4 . Owing to an increase in the volume of Ga on solidification, it is placed in an elastic teflon container . Standardised thermometers are introduced into a nylon tube with a lining of Al . The sealed cell which is immersed in melting Ga, is filled with an atmosphere of pure argon . Each cell is supplied with recommendations from the manufacturer specifying the necessary immersion depth in the bath . An accuracy in the reproduction of the melting point of gallium of about f0 .4 mK can be attained (Chattle and Pokhodun, 1987) . A commercially produced gallium melting point cell, which can be removed for freezing and then returned to the apparatus, holds the melting temperature for many hours, ensuring a precision of about fl mK . The freezing point of indiuM 156 .5985 °C, consists of high purity indium (better than 99 .999 %) in either a graphite crucible (Chattle and Pokhodun, 1987) or in a teflon cell (Mangum, 1989) . The indium container is placed in a special furnace . After the ingot is melted, the furnace temperature is stabilised about a degree below the freezing point . When the temperature indicated by a resistance sensor in a protecting sheath, placed in the cell, has fallen close to the freezing point, the sensor is withdrawn and allowed to cool for up to 1 min before being put into the cell again . The loss of heat is sufficient to form a thin layer of solid indium around the sensor well . The plateau corresponding to the freezing point temperature is then quickly reached . The freezing point of 99 .999 % pure indium is reproducible to about f0 .1 mK (Mangum, 1989 ; Hanafy et al ., 1982) . The freezing point of tiny 231 .928 °C, when based on tin of very high, 99 .9999 %, purity, is reproducible to about 0 .1 mK (Preston-Thomas, 1990) . In realising this point, the phenomenon of large supercooling at the beginning of freezing, should be taken into PUMPING TUBE NYLONCAP - - TEFLON CONTAINER OUTERNYLON CASE GALLIUM GLASS CELL - - NYLON TUBE WATER VAPOUR - WATER Al SENSOR TUBE ICE LAYER ,~ WATER 1 \ ICE-WATER - - MIXTURE _ hti^1 DEWAR VESSEL - - U_ Figure 22 .3 Triple point of water Figure 22 .4 Melting point of gallium (Chattle and Pokhodun, 1987) 424 CALIBRATION AND TESTING consideration (McLaren and Murdock, 1960) . The construction of this freezing point, which is described by Marcarino (1992), is similar to the freezing point of zinc, shown in Figure 22 .5 . It is important to achieve a high degree of supercooling (over 4 K) to attain the plateau temperatures by means at outside slow freezing . To attain this supercooling the ingot should be kept in an inert atmosphere and not topped with graphite powder . The freezing point of zing 419 .527 °C, is realised with a reproducibility of about 2 mK if zinc of very high purity (99 .9999 %) is used (Preston-Thomas et al ., 1990 ; McLaren, 1958, Ma and Lawlor, 1992 ; Furukawa et al ., 1981 ; Marcarino, 1992) . The apparatus used to produce the fixed point should ensure the necessary zinc purity as well as a uniform temperature distribution during its solidification, with equality of the temperatures of the sensor and the metal . A tubular furnace with a copper block, which is shown in Figure 22 .5, can ensure this uniformity of temperature . Inside the block there is a graphite crucible with a lid, through which a sheath for carrying calibrated sensors is inserted . At the beginning of metal solidification, the observable under-cooling by some hundredths of a degree may be eliminated by withdrawing the calibrated sensor for a short time . After reaching ambient temperature the sensor is reinserted . The temperature which then quickly rises to the freezing point, stays constant at the freezing point temperature for a long period of time . Zinc purity should be periodically tested . The freezing point of aluminiuiA 660 .323 °C, is produces in a similar way as for zinc . Graphite fibre insulation and a graphite crucible are used combined with an atmosphere of argon gas to prevent oxidation . The reproducibility of this point is about 1 mK with the temperature stability at the beginning of freezing of 0 .2 mK (Ancsin,1992 ; McAllan and Ammar,1972 ; Furukawa, 1974) . No contact of molten aluminium with moisture, oxygen and silicon ceramic can exist . The freezing point of silver, 961 .78 °C, is realised in a similar apparatus to that shown in Figure 22 .5, but with a higher temperature furnace . To get high temperature uniformity in the middle part of furnace, where the crucible with the silver is placed, it is advisable to install a heat pipe screen around the crucible . Even though silver does not oxidise easily, it should still be protected from air contact, since it absorbs oxygen in the molten condition, SENSOR TUBE A12 0 3 -POWDER ELECTRIC FURNACE LID GRAPHITE CRUCIBLE MOLTEN ZINC COPPER BLOCK CERAMICS THERMAL INSULATION t Figure 22 .5 The freezing point of zinc PRIMARYSTANDARDS 425 resulting in depression of the freezing point . As oxygen absorption starts at a temperature of about 30 °C above the melting point, any unnecessary overheating of the metal should be avoided . If the ingot is kept in the molten state in an inert gas, the surrounding graphite will effect the complete removal of the oxygen within a few hours . The constant temperature level, corresponding to the freezing point of silver, is reached in a time period, covering 20 to 60 % of the total solidifying time, with an accuracy of 1 mK (Preston-Thomas et al ., 1990) . The freezing point of silver, 961 .78 °C, gold, 1064 .18 °C, and copper, 1084 .62 °C which are intended for the calibration of pyrometers, are each realised in water cooled electric resistance furnaces with a graphite chamber inside as shown in Figure 22 .6 (Ohtsuka and Bedford, 1982) . The chamber has all the properties of a black body . The graphite black body and crucible containing the metal are made of graphite of highest purity so that an attainable emissivity of about 0 .99999 is achieved (Lee, 1966) . The crucible is placed in a sodium heat-pipe liner, giving a uniform temperature of the crucible walls . Additional blocks and rings of graphite keep oxygen from reaching the crucible . The thermal insulation of the furnace is made of quartz wool and inconel heat shields . During operation nitrogen or argon flows slowly (0 .1 litre/min) along the furnace length to inhibit graphite oxidation . The temperature is measured by a type S thermocouple . The calibrated pyrometer is aimed at the bottom of the graphite chamber . CERAMIC SODIUM GRAPHITE FREEZING METAL (A9,Au,Cu) TUBE HEAT PIPE CRUSIBLE THERMOCOUPLE r r r r ~ ~ r r r ~ ^ r r r r z r ~ ^ t r r 2 ~r 1 -*ARGON _ _ r r r 1 r Z t 1 J r r 1 t r rr I t ^ . QUARTZ WOOL GRAPHITE BLACK INCONEL WATER COOLING BODY CAVITY RADIATION SHIELD 60 an Figure 22 .6 Metal freezing points for calibration of pyrometers (Ohtsuka and Bedford, 1982) 22 .3 Primary Standards Primary standards are instruments specified in the text of ITS-90 for interpolation between the fixed points . Standard resistance thermometers are used for interpolation from -259 .3467 °C to 961 .78 °C, which is the freezing point of silver . Following ITS-90 the thermometer resistor must be strain free, annealed, pure platinum and wound from 0 .05 to 0 .5 mm Pt wire . It is advisable to use resistors of 25 S2, at 0 °C . In the upper temperature range, 0 .1 to 2 .5 S2 resistors are recommended . The resistor should be enclosed in a hermetic sheath filled by 426 CALIBRATION AND TESTING dry, neutral gas with an addition of oxygen . At the lower temperature range, up to 13 K, it should be helium filled . The resistor should be annealed at a temperature higher than the highest expected working temperature, but in any case never below 450 °C (except for cryosensors) . The quality of a sensor, its design and annealing are verified during calibration, determining the constants from interpolation equations and checking the stability of the resistance (Curtis, 1972 ; Foster, 1972) . Depending on the working temperature range, there are three types of resistance temperature sensors : " -260 °C to 0 °C - low temperature capsule-type sensors, " -190 °C to 600 °C - normal long-stem sensors, " +600 °C to 960 °C - high temperature sensors . Capsule-type resistance sensors, which are beyond the scope of this book, are described by Curtis (1972), Hust (1970) and Sparks and Powell (1972) . Long-stem-type resistance sensors, which are used as interpolation standards of ITS-90 from -190 °C to 600 °C, have undergone many modifications to increase their accuracy and stability, to reduce their size and to intensify the heat transfer between the resistor and sheath and between the sheath and environment . A typical contemporary design is presented in Figure 22 .7 . Platinum wire, wound in a spiral of about 1 mm diameter is placed in a thin-walled Pyrex tube matching the spiral diameter and shaped as shown in Figure 22 .7 . Platinum terminals, which are soldered to both spiral ends, are sealed in glass in such a way that the spiral is totally strain free . It is also important for the spiral to remain strain free during its subsequent working life . These terminals are extended by low resistance gold wires in ceramic insulation . The whole assembly, which is encapsulated in a glass sheath, is hermetically sealed after careful drying at about 400 °C . The resistance of the sensor is about 25 0 at 0 °C . As an example, the standard Pt resistance sensor produced by Rosemount Inc . (USA) (Berry, 1982) is hermetically sealed in a metal sheath containing a helium-oxygen atmosphere . Its stability within the specified temperature range of -200 °C to +650 °C is better than 0 .01 °C per year . The self-heating increase in temperature is less than 0 .002 °C with an insulation resistance from the resistor to the outside sheath greater than 5000 MQ at 100 V dc, while its nominal resistance is about 25 S2 at 0 °C . High temperature resistance sensors, operating from 600 °C to 960 °C replace the S-type thermocouple as a primary standard of ITS-90 . Design of high temperature resistance sensors is the subject of many publications (Arai, 1997 ; Anderson, 1972, Evans and Burns, Au-WIRES Pt-WIRES GLASS SHEATH d=6mm CERAMIC INSULATOR Pt-SPIRAL PYREX TUBE Figure 22 .7 Standard long-stem-type resistance sensor PRIMARYSTANDARDS 427 1962 ; Chattle, 1972 ; Curtis, 1972 ; Evans, 1972 ; Furukawa et al ., 1981 ; Strouse et al ., 1992) . One of the designs proposed by Nubbemeyer (1992) is shown in Figure 22 .8 . The resistor is composed of a bipolar spiral winding of 0 .4 mm diameter platinum wire, supported by a notched quartz blade . Two Pt wires of the same diameter, which are welded to both ends of the spiral, are extended by two 75 cm long, 0 .35 mm diameter Pt wires . These wires are insulated by quartz tubes passing through 9 quartz disks placed along the sensor . After the sensor has been annealed at 700 °C, the external 7 mm diameter quartz sheath is hermetically sealed and the tube filled with a gas mixture of 90 % Ar and 10 % Oz . The sensor resistance is 0 .25 SZ at 0 °C . Resistance sensors for interpolation in ITS-90, whose resistance is measured by an ac bridge of highest precision in a four-wire circuit, are calibrated at relevant fixed points within the sensor application range . An example of the bridge is the F 18 bridge produced by Automatic Systems Laboratories Ltd (1999) having the following technical data : " accuracy better than + 0 .25 mK, " resolution : 0 .75 pK, " measuring range : 0-390 fl (for R a = 2 .5-100 S2 in the temperature range : 13 K-960 ° C), " frequencies : 25/75 Hz or 30/90 Hz, " possibility of measuring temperature difference, " automatic or manual balancing, " interface : IEEE-488 . Standard pyrometers and tungsten strip lampsare used for reproducing ITS-90 above the freezing point of silver, 961 .78 °C . They are calibrated at the three fixed points of Ag, Au and Cu, simulating black body radiation Pt 0, 35 mm QUARTZTUBE QUARTZ 4-HOLE DISK QUARTZ SHEATH d=7mm WINDING Pt 0,4mm QUARTZ BLADE QUARTZ BLADE QUARTZ DISK JUNCTION OF TWO PARTS OF WINDING Figure 22 .8 Standard high temperature resistance sensor (Nubbemeyer, 1992) 428 CALIBRATION AND TESTING Standard pyrometers and their calibration methods have undergone many modifications in recent years . The photoelectric spectropyrometer, which was the early type, is based on the principle of the disappearing filament pyrometer, where a photoelectric detector replaces the human eye, as described by Hahn et al . (1992), Kandyba and Kowalewski (1956), Lee (1966), Lee et al . (1972) and Nutter (1972) . This early type has now been replaced by a narrow-band photoelectric pyrometer with either a photomultiplier detector or by the increasingly popular Si detector, which is characterised by high sensitivity, high stability andgood linearity (Coslovi and Righini, 1980 ; Jung 1979) . Standard photoelectric pyrometers, which have an optical system like that shown in Figure 22 .9 (Rosso and Righini, 1985), also have an accuracy better than 0 .1 K in the temperature range of 800 to 1400 K and about 1 K at a level of 2000 K similar to others (Zhao et al ., 1990, 1992) . The silicon detector of the pyrometer, which is placed in a thermally stabilised housing, gives an equivalent stability of 0 .1 K per month and some tenths of kelvin per year . Methods of pyrometer calibration at the fixed points are considered by Bussolino et al . (1987) . Although pyrometers are mainly calibrated for measuring temperature, some pyrometers may also be used as instruments for comparing heat fluxes (Preston-Thomas et al ., 1990 ; Zhao et al ., 1992) . Standard tungsten strip lamps are used for interpolation in the temperature range from 1337 to 2600 K . Vacuum lamps can be usedup to about 2000K, whereas above this temperature the use of gas-filled lamps, shown in Figure 22 .10, is advised . The strip length must be big enough to prevent any substantial influence of ambient temperature on the strip temperature . A `place' is also marked on the strip where the measurements should be made . The sighting angle of the pyrometer is given by two points . One point is on the sighting window and another is on the strip . These two points should coincide during measurements . To prevent any reflection of the radiation, which might be a source of errors, both lamp windows are situated at an angle of 5° to the lamp axis . The dependence of the temperature of the lamp's strip upon the lamp current is called the thermometric characteristic of the lamp . To achieve high stability of this characteristic the lamps are degassed many times during the production process before being finally glued and annealed (Quinnand Lee, 1972) . Calibration of tungsten strip lamps is made by a comparison method based on the readings of a photoelectric spectropyrometer and simultaneous measurement of the lamp current at different strip temperatures . OBJECTIVE SHUTTER IN TEM E RA R LENS MIRROR APERTURE PETURE "' CONTROLLED DIAPHRAGM STOP ENCLOSURE TARGET INTERFERENCE EYE PIECE MIRROR I r MIRROR LL - Figure 22 .9 Standard photoelectric pyrometer - optical arrangement (Rosso and Righini, 1985)