12 Conditioning of Temperature Sensor Output Signals 12 .1 Introduction Temperature is a physical variable characterising the flow of heat energy . A temperature measurement process attaches real numbers to the degree of hotness of bodies . Although it can only be measured by suitable sensors the output of these components needs to be presented to the human observer by indicators, recorders or other instruments . To accomplish this further task it is necessary to transform the temperature signal, which is in the thermal energy domain, into another more convenient energy form . Transforming to electrical signals is the most convenient for these purposes . This allows more effective forms of processing, which may also be called conditioning, of the temperature signal in systems such as that shown in Figure 12 .1 . Evidently these operations are an important stage of the temperature measuring process . Transformation and conditioning of measuring signals are important in metrology (IOTech Inc ., 1998 ; Lang, 1987) . Nevertheless the critical element in all measuring channels is the sensor itself . It is a widely accepted opinion that no amount of signal treatment can improve an inherently bad signal . The significant developments, which can be observed in transformation and conditioning methods, are also causing fundamental changes in the handling of temperature data . These data are probably the most widely encountered type of data, as temperature is generally regarded as the most frequently measured indicating and recording " instruments SIGNAL temperature PROCESSING y signalling, alarming and control devices Figure 12 .1 Signal processing as a part of temperature measuring chain 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) 230 CONDITIONING OF TEMPERATURE SENSOROUTPUT SIGNALS physical quantity both in laboratories and in industry . Usually, the conditioning units, which enhance the quality of the whole measuring system, allow more easy maintenance . In this chapter the status and trends in temperature measurement channel design are presented . This is achieved by giving an overview and classification of contemporary methods and algorithms used within measuring systems applied in the temperature field . Such a broader theoretical context allows further discussion ofboth traditional and contemporary solutions used in temperature measurement and also shows the trends of development in this field . Typically, a temperature measuring chain consists of a number of different, discernable conditioning steps, which adjust the signal to the requirements of various instruments . The role of initial transformation of temperature into another physical quantity, performed by temperature sensors, has been presented in Section 1 .4 . In the terminology of Figure 1 .4 these further conditioning steps belong to the group of modifiers . Continuing this approach, further conditioning of measured temperature signals will be discussed in this chapter . At present there is no general agreement on the nomenclature which should be used for various methods of signal transformation . However, an appropriate approach, allowing these methods to be grouped and classified is presented by Sydenham (1983), where it is asserted that the three main groups of methods are : " transformation of signal nature, " transformation of signal scale, " transformation of signal shape . Transformation of signal nature includes methods applied, when the physical quantity or energy form of the signal does not suit the requirements of the measuring units or instruments . Initial transformation is always a transformation of signal nature but it can also be found in many other steps of signal processing . Transformation of signal scale changes the proportional values of a processed signal by an increase or decrease . Temperature signals usually require amplification because they are of low energy content . This is true irrespective of whether the signal is obtained from a self- sustaining cross-converter or a modulating sensor . Transformation of signal shape includes algorithms influencing the time domain form of the signal . It usually leads to modification of the spectral power distribution of the signal, elimination of certain harmonics, frequency modulation etc . Each of these methods of signal transformation can be found at various steps of temperature signal processing . Sometimes they are used within one integrated processing unit . It should be noted however that all of them influence the final accuracy of a temperature measuring process . 12 .2 Methods of Signal Processing in Temperature Measurement Methods of temperature signal processing have been classified following the methodical approach presented in Section 1 .4 . The basis of this classification is a space of all physical quantities in which both non-electric and electric quantities are distinguished . In addition, analogue and digital forms of electric quantities are considered, as shown in Figure 12 .2 . METHODS OF SIGNAL PROCESSING IN TEMPERATURE MEASUREMENT 231 Such an approach allows all types of signals which can be used for carrying information about measured temperature, within measuring systems, to be taken into account . Temperature is usually captured from the body, or object, under measurement by various sensors or transducers, which may be regarded as initial energy transformers or converters . This initial transformation leads to a change of the signal from the thermal energy domain either to another non-electric form or to an electric form as shown in Figures 1 .6, 1 .7 and 1 .8 . This is usually accompanied by shape transformation due to nonlinearity of the functional characteristics of temperature sensors . The resulting signal is rarely used directly by indicators, recorders or other measuring instruments . Hence, it undergoes subsequent further transformation, called signal conditioning by Sydenham (1983) . Signal conditioning used for temperature measurement is often a multistage process, which can be performed both for non-electric and electric signals . Methods applied within various temperature measurement systems can be successfully grouped following the classification proposed in Figure 12 .2 . Traditional temperature measurement equipment, which is still used widely in industry, often includes non-electric temperature sensors and/or electromechanical measuring instruments . These systems, which are mostly based on processing their inherently analogue types of non-electric quantities, exclusively use analogue signal conditioning methods . An example of such a method is shown in Figure 12 .3 . Although these methods are not very popular it is worthwhile to stress that they can also be classified following the general approachproposed in Section 12 . l . Most of the conditioning performed within temperature measurement systems is now based on electric signals both in analogue and digital form . These signals are commonly regarded as a more convenient form . This trend is also stimulated by the rapid development of microprocessors and computers . For these important reasons, methods of conditioning of electric signals used in temperature measurement systems are discussed in detail below . 12 .2 .1 Transformation of signal nature Transformation of analogue electric signals is an important group among the methods of transformation of signal nature . For example, it is evident that in determining the actual resistance, RT, ofan RTD, based on measuring the voltage, VT, developed across it due to a known current flowing in it, represents a transformation of resistance to voltage drop Also at the final stage of this measuring channel, voltage or current can often be transformed into the movement of the pointer using the electromagnetic torque in electromechanical indicating instruments . Transformation between main groups of physical quantities, as shown in Figure 12 .2, also represents an important group of methods of transforming the nature of analogue signals . It concerns mainly the transformation of non-electric to electric analogue signals and the converse . This group includes, for example, opto-isolating elements, used to provide galvanic separation of different parts of a measuring system, as well as electromechanical indicating instruments . Transformation of digital signal nature is illustrated by a changeof digital signal code, which is often applied within microprocessors-based systems . This operation often allows the elements of digital measuring systems to be simplified or to facilitate signal transmission . The manner of signal coding is determined by its type, values and also its initial conditioning transformation N RANSFORMATION OFNATURE t change of movement type N ji ; optical transducers pneumatic transducers E L E non electric C temperature sensors T 'I' . VA274ti1'FFC7lY~Ok' :SCt1t~~ :~ :~RAN~FDRAMY3U}~T#IF :StL48E : " : ek mechanical gears qk mechanical dampers I qk ambient temperature C correction electric sensors uANSF AT~ :o~ :n~s~uRS : :: mutual transformation PHYSICAL of non-electric and electric QUANTITIES quantities SF(SRMA :T#lN :O1 . . sFtbtRFA'Btt3N~ .(lE' :NA'i'tIR'LC .~ :~ :?? E change of electric quantities - ok change of code L types * change of tram E TRANSFORMATION protocol C OF SHAFE A/D R and I D/A C SFOR rATIa : Facu ;~es : :? : rsscmMA oN :UFa V> : : conversion analogue amplification 411 ; analogue filtering unification analogue linearisation 4 digital amplificatie t unification 4E correction " characteristic values . t signalling ANALOGUE DIGITAL Figure 12 .2 Classification of signal transformation methods METHODS OF SIGNAL PROCESSING IN TEMPERATURE MEASUREMENT 233 ANALOG INDICATING 'C INSTRUMENT SIGNAL NATURE TRANSFORMATION LINEAR-ROTATIONAL MOVEMENT SIGNAL SHAPE TRANSFORMATION BIMETALIC CORRECTION OF AMBIENT TEMPERATURE SIGNAL NATURE TRANSFORMATION LIQUID VOLUME - SPRING SHAPE SIGNAL TRANSMISSION INITIAL TRANSFORMATION TEMPERATURE - .CHANGE OF LIQUID VOLUME Figure 12 .3 Transformation of non-electric signal in a liquid-filled manometric thermometer origin and destination . For example, the binary code, which represents numbers in radix 2, is the basic code used to represent integer numbers within digital systems . The integer value, X, within the range 0<_ X<_ (2 ° -1) can be represented by a binary word of digital information consisting of n-bits in the form : an -1 an -2 . . .ap aic10,1} where the value of X may be written as : X = 2 °-l a -,+2 °-Zan_Z+ . . .+ao . Another very popular coding method is the BCD (Binary Coded Decimal) form in which a 4-element binary number is determined for each digit of the decimal number . This method is very often used in digital indicators . The comparison of these two digital coding methods for representing the decimal number 123 is given below . Code Codeword Binary 01111011 BCD 0001 0010 0011 The modification, or transformation, of the nature of a digital signal is performed by digital-to-digital (DID) converters, which may be a part ofan integrated processing unit or, rarely, a separate element . 234 CONDITIONING OF TEMPERATURE SENSOR OUTPUT SIGNALS 12 .2 .2 Transformation of signal scale Amplification This is the most important example of the modification of signal scale in a temperature measurement system . It can be performed both in analogue and in digital form . Analogue amplifiers are usually based on integrated circuit operational amplifiers or specialised integrated circuits . A good example of such a module is the AD 594/595 thermocouple amplifier by Analog Devices, which matches type J or K thermocouples . It provides a high level signal sensitivity of 10 mV/°C and resolution of 0 .5 °C . Omega Inc . offers the OMNI-AMP series of amplifiers for various signals including those for thermocouples (Omega Inc ., 1999) . The portable OMNI-AMP 1 amplifier, whose circuit diagram is shown in Figure 12 .4, is dedicated for use in indicating and recording instruments . It ensures seven fixed gains of ix, 2x, 5x, 10x, 25x, 50x and 100x set by a rotary selector switch . The bandwidth of the amplifier, which ranges up to 100 Hz at the highest gain, increases at lower gains . This amplifier can also be used when the input impedance of indicating or recording instrument limits the thermocouple lead length or the total circuit resistance . When the amplifier gain is set at unity the input resistance of the thermocouple circuit is virtually independent of the indicating device . Thermocouple leads of length up to about 400 m can be used . The amplifier is supplied by two batteries with a lifetime of 100 hours or more . The OMNI-AMP HB, which is a laboratory amplifier model also designed for thermocouples, is equipped with a reference junction compensation circuit while the OMNI-AMP III'offers a gain up to 1000x . The industrial model OMNI-AMP IV is additionally epoxy encapsulated with a built-in supply . Digital amplification is performed by software multiplication of digital signal values . The multiplication factor, corresponding to the gain, which can easily be set and changed, is not influenced by thermal instability of electronic elements as are analogue circuits . Unification . The scale of a measured signal can also be adjusted to one of the unified ranges of electrical signals . The most popular unified signals used in measurement systems are : " unipolar current signals : 0-20 mA, 4-20 mA, " voltage signals : 0-10V, 0-5 V, ±10V,±5 V . SUPPLY INPUT GAIN OUTPUT 0 Figure 12 .4 Circuit diagram of OMNI-AMP amplifier METHODS OF SIGNAL PROCESSING IN TEMPERATURE MEASUREMENT 235 Since current signals are highly insensitive to disturbances, the information can be carried for quite long distances . Voltage signals are more sensitive . Moreover, it is very convenient to be able to detect any damage to the sensor, using signals oflow limit value different from zero . The unified 4-20 mA signal, which also allows the structure of a measurement system to be simplified by applying a current loop, is described in Section 12 .5 . Unification of signals, which can be performed both by analogue and digital methods, is usually the final element of various signal processing units . It allows the whole measurement system to be configured easily and also allows easy reconfiguration by replacing selected parts . 12 .2 .3 Transformation of signal shape This group of conditioning methods is characterised by the miscellaneous nature of its numerous members . Signal filtering . The most popular type of transformation of signal shape is the rejection of noise frommeasured temperature signals . Stochastic, also called random, noise, which may strongly influence the accuracy of the measured temperature signal, is an inevitable interference in any real temperature measurements . It occurs in industrial applications where strong nearby electromagnetic fields exist, such as occur in induction heating appliances . The signal-to-noise ratio (SNR) can be improved by applying both analogue and digital filters which usually modify the spectrumof the measured signal The simplest type of analogue noise filter is a passive I st order low-pass RC-filter shown in Figure 12 .5(a) . The bandwidth of such a filter is determined by the corner frequency, f, = 1/(21rRC), and by the asymptotic attenuation, which is 20 dB/decade beyond f, More effective filtering of high frequency components can be obtained using higher order filters such as Butterworth, also called maximally flat, Bessel filters, also called Thomson or maximally linear phase, or Tchebishev . In some cases notch filters can be very effective when narrow band disturbances from power supply units need to be rejected . An example of an integrated circuit active filter, which can also be used, is shown in Figure 12 .5(b) . However, when choosing the type of filter both its amplitude and phase characteristic should be taken into consideration so that no undesired amplitude or phase (°I PASSIVE FILTER (c) 'S R T C n ACTIVE FILTER " f I (b) = I I ( I I I ~ t 2 . . . n Figure 12 .5 Noise filtering methods are typically the analogue passive filter in (a) with its analogue active filter realisation in (b) and a digital filtering procedure in (c) 236 CONDITIONINGOF TEMPERATURE SENSOR OUTPUT SIGNALS distortion of the measured temperature signal is introduced . This applies especially to the measurement of rapidly changing temperature signals . Digital methods of noise filtering are also often used in temperature measuring systems . They usually process the measured signal in the time domain . An averaging filter, whose output is calculated as an average value from a set of neighbouring temperature samples, as illustrated in Figure 12 .5(c), is a simple but effective type of digital filter . Such a filter, which is usually supplemented by stochastic analysis of the measured signal, can be a convenient tool for a typical temperature measurement system . The comparison of the effectiveness of analogue and digital noise filtering in a computerised temperature measurement system is presented in Section 13 .2 .2 . Additionally, in temperature measuring systems, the problem of aliasing may occur during the conversion from an analogue to a digital signal form . Aliasing errors, also called fold-over errors, are caused by the appearance of high frequency components as false low frequency components in the signal spectrum . They occur when the bandwidth of a measured analogue signal is above half the sampling frequency . It will be recalled that the Nyquist frequency, which equals one half of the sampling frequency is the theoretical maximum frequency which can be captured in a sampling system . These errors are often hard to detect and difficult to remove in software . Hence anti-abasing analogue filters can be placed before the A/D converter to eliminate the high-frequency signal components which would be folded over . Thus, they are prevented from causing aliasing errors . Numerical example A temperature signal is acquired by an 8 channel data acquisition system with a sampling rate 100 samples/second . Calculate the sampling rate for one channel, specify the Nyquist rate and comment . Solution : The sampling rate for one channel is 100/8 = 12 .5 samples/second . In this case the Nyquist frequency can be calculated as 12 .5/2 = 6 .25 Hz . Any signal with a component above 6 .25 Hz will cause aliasing errors . It should be filtered out . Figure 12 .6 shows the amplitude frequency response of a typical anti-aliasing filter . There are three commonly used types of filters : Butterworth, Bessel and elliptic filters . Elliptic filters have the sharpest cut-off, but their transient response is not good . Bessel filters, which have the slowest attenuation above the cut-off frequency, have the best transient response . As a general conclusion it can be stated that the steeper the attenuation of the filter the slower is its dynamic response . It should also be noticed that analogue-to-digital conversion introduces some filtering especially in the case of dual slope, or double integrating, converters . The influence of noise can also be reduced by shielding, earthing and channel separation (Omega, 1999) . In recent years more and more sophisticated methods, including artificial intelligence, are applied for data and signal filtering (Russo, 1996) . An example of the application of fuzzy logic for noise rejection from the temperature signal has been described by Kucharski (1999) . METHODS OF SIGNAL PROCESSING IN TEMPERATURE MEASUREMENT 237 CUT-OFF FREQUENCY CUT-OFF CHARACTERISTIC w J a s a FREQUENCY Figure 12 .6 Amplitude frequency response of typical anti-aliasing filter Experimental example Using a computerised data acquisition system, also discussed in Section 13 .2 .2, the temperature signal measured by two different temperature sensors was recorded as shown in Figure 12 .7 . The following sensors were used : " MI thermocouple of type K and diameter 3 mm, " infrared temperature sensor IRt/c by Omega Inc . (Omega Inc ., 1999) whose output is a voltage signal corresponding to the characteristic ofa type K thermocouple . It can be noticed that the infrared sensor is more sensitive to noise than the MI thermocouple because of its higher input resistance (ranging about kQ) . Thus the signal from the IR sensor requires shielding as well as analogue low-pass filtering . In the case of the thermocouple such filtering is usually effective on its own . The problem of shielding becomes of paramount importance in industrial applications where strong electromagnetic fields exist as, for example, in induction heating processes . IRt/c SENSOR IRt/c SENSOR*ANALOGUE FILTERING xs 250 zoo NON - SHIELDED SHIELDED 200 NON-SHIELDED SHIELDED u u +5 0+50 (r +00 +00 Q 50 tR 50 w 0 w 0 d d IIII ., I, III' w -so -so ~ -+00 ~ -+ oo -+ so -+s0 +0 20 30 4o 50 so 10 so s0 +00 +0 20 30 40 so so 70 so so +00 TIME , s TIME , s MI-THERMOCOUPLE d-3mm MI - THERMOCOUPLE d .3mm+ ANALOGUE FILTERING ao 40 o 35 NON-SHIELDED SHIELDED ; 3s NON-SHIELDED SHIELDED CX 30 ~ 30 7 Q 25 I Q 25 20 ~ 20 ~I +s I^III Ilrtl,h'111~~~1'~r ~~ I~ ~I~i, "p, I I . il r- +s . ~IP~ITII , PP r P ,_° III , . r'tl+o +o +0 20 3o 4o 6o 6o 7o e0 9o +oo +0 2o 3o 4o 6o 6o 7o so 9o +oo TIME , s TIME , s Figure 12 .7 Effectiveness of shielding of connecting leads compared to analogue filtering of noise 238 CONDITIONING OF TEMPERATURE SENSOR OUTPUT SIGNALS Linearisation of sensor characteristics . Temperature sensors are usually non-linear elements . Even if this non-linearity is insignificant it should be corrected to realise or simplify the whole measuring system . This can be done using the inverse characteristic of the sensor as shown in Figure 12 .8 . Linearisation procedure in principle leads to the determination of the real temperature value but usually an electric signal proportional to this temperature is obtained as a result . Analogue linearisation can be performed by an electronic circuit whose input-output characteristic corresponds to the inverse characteristic of a given temperature sensor . The accuracy of such a linearisation depends on the quality of the approximation of the real characteristic and on the thermal stability of the electronic elements . An example ofsuch a circuit is the XTR103 module by Burr-Brown (information from Burr-Brown website) whose block diagram is presented in Figure 12 .9 . The XTR103 circuit, which is dedicated for RTD Pt-100 temperature sensors, ensures the linearisation of a second order polynomial characteristic . It is based on two operational amplifiers with a linearisation procedure performed by changing the measuring current passing through the sensor and producing a 4-20 mA unified output signal . TEMPERATURE a INVERSE SENSOR t= CHARACTERISTIC CHARACTERISTIC 0 LINEARISATION I f f TRUE INDICATED TEMPERATURE TEMPERATURE Figure 12 .8 The idea of linearisation of temperature sensor characteristic XTR 103 0 .8-1 mA 0 .8-1 mA 'Rl ' R l R L V . + 1 4 0 A OUT `'OUT R, PtR 100 Rz R CM Figure 12 .9 Circuit diagram of XTR103 analogue linearisation module by Burr-Brown [...]... discharging when the input signal is decreasing Transistors, T1 and TZ, and amplifier, 2, convert the capacitor voltage into a proportional output current A capacitor discharge time can be set by a potentiometer, P1, so that the decay rate for the maximum value can be adjusted " " 1 " " ~ ' 1 '~ '7" r - 1 ' " t : ~y t© ~, 0 , , ., 1 1 ~ - - 1 - 1 1 _ 1 .- - : - - - " 11 _... have a typical measuring range of -80 to +600 °C with an accuracy of ±2 % FSR (Full Scale Range) They are robust and vibration resistant Electric analogue recorders include deflection type recorders and potentiometric recorders Deflection type recorders are based on millivoltmeters or quotient instruments Their small deflection torques do not allow the recording pen to be attached directly to the pointer,... speed between 20 to 1200 mm/h is accompanied by a recorder accuracy of 1 5 to 2 5 % Recorders with an input amplifier allow a continuously writing pen to be directly attached to the pointer Automatic potentiometric recorders for thermocouples and pyrometers as well as automatically balanced bridges for resistance sensors comprise the most important family of analogue recorders The idea of such instruments . input signal is decreasing . Transistors, T1 and TZ, and amplifier, 2, convert the capacitor voltage into a proportional output current . A capacitor discharge time can be set by a potentiometer, P1, so that the decay rate for the maximum value can be adjusted . " 1 " " " ~ ' '7 " 1 '~ r " - 1 ' t : ~y t© . ~, . 0 ., ,