c
   






à




ñ
   
  

  
  

  

  

à





 
   

   
×
  

 

à


J   ?

    

  


 
!
"
J

# 

3


©




""!

$$
J$
c
    



à
 
 

3



 
   

à








?
?









c
    










×
 
 
In PTC thermistors the resistance increases with temperature. They are
manufactured using barium titanate and are not normally used in temperature
measurement. These devices are selected when a significant change in the
resistance is required at a specific temperature or current. Examples of
applications include temperature sensing, switching, protection of windings in

electric motors and transformers.


 
   
Semiconductor devices, consisting of either diodes or integrated circuit transistors, have
only been commonly used in industrial applications for a few years, but they were first
invented several decades ago. They have the advantage of being relatively
inexpensive,
but one difficulty that affects their use is the need to provide an external power supply
c
    



to the sensor.

Integrated circuit transistors produce an output proportional to the absolute temperature.

Different types are configured to give an output in the form of either a varying
current (typically 1 ȝA/K) or varying voltage (typically 10 mV/K). Current forms are
normally used with a digital voltmeter that detects the current output in terms of
the voltage drop across a 10 k_ resistor. Although the devices have a very low cost
(typically a few pounds) and a better linearity than either thermocouples or resistance
thermometers, they only have a limited measurement range from _50°C to C150°C.
Their inaccuracy is typically š3%, which limits their range of application. However,
they are widely used to monitor pipes and cables, where their low cost means that
it is feasible to mount multiple sensors along the length of the pipe/cable to detect
hot spots.
In diodes, the forward voltage across the device varies with temperature. Output
from a typical diode package is in the microamp range. Diodes have a small size, with
good output linearity and typical inaccuracy of only š0.5%. Silicon diodes cover the
temperature range from _50 to C200°C and germanium ones from _270 to C40°C.

 


 
Thermography, or thermal imaging, involves scanning an infrared radiation detector
across an object. The information gathered is then processed and an output in the form
of the temperature distribution across the object is produced. Temperature
measurement
over the range from _20°C up to C1500°C is possible. Elements of the system are
shown in Figure 14.14.
The radiation detector uses the same principles of operation as a radiation pyrometer
in inferring the temperature of the point that the instrument is focused on from
a measurement of the incoming infrared radiation. However, instead of providing a
c
    




measurement of the temperature of a single point at the focal point of the instrument,
the detector is scanned across a body or scene, and thus provides information about
temperature distributions. Because of the scanning mode of operation of the instrument,
radiation detectors with a very fast response are required, and only photoconductive
or photovoltaic sensors are suitable. These are sensitive to the portion of the infrared
spectrum between wavelengths of 2 ȝm and 14 ȝm.
Simpler versions of thermal imaging instruments consist of hand-held viewers that
are pointed at the object of interest. The output from an array of infrared detectors is
directed onto a matrix of red light-emitting diodes assembled behind a glass screen, and
the output display thus consists of different intensities of red on a black background,
with the different intensities corresponding to different temperatures. Measurement
resolution is high, with temperature differences as small as 0.1°C being detectable.
Such instruments are used in a wide variety of applications such as monitoring product
flows through pipework, detecting insulation faults, and detecting hot spots in furnace
linings, electrical transformers, machines, bearings etc. The number of applications is
extended still further if the instrument is carried in a helicopter, where uses include
scanning electrical transmission lines for faults, searching for lost or injured people
and detecting the source and spread pattern of forest fires.
More complex thermal imaging systems comprise a tripod-mounted detector
connected to a desktop computer and display system. Multi-colour displays are
commonly used in such systems, where up to 16 different colours represent different
bands of temperature across the measured range. The heat distribution across the
measured body or scene is thus displayed graphically as a contoured set of coloured
bands representing the different temperature levels. Such colour-thermography systems
find many applications such as inspecting electronic circuit boards and monitoring
production processes. There are also medical applications in body scanning.
c
    





  
 
The principle of acoustic thermometry was discovered as long ago as 1873 and uses
the fact that the velocity of sound through a gas varies with temperature according to
the equation:
vD
_
ÞRT/M _14.10_
where v is the sound velocity, T is the gas temperature, M is the molecular weight of
the gas and both R and Þ are constants. Until very recently, it had only been used for
measuring cryogenic (very low) temperatures, but it is now also used for measuring
higher temperatures and can potentially measure right up to 20 000°C. However, typical
inaccuracy is š5%, and the devices are expensive (typically £6000 or $10 000). The
various versions of acoustic thermometer that are available differ according to the
technique used for generating sound and measuring its velocity in the gas. If ultrasonic

generation is used, the instrument is often known as an p



. Further
information can be found in Michalski, (1991).


 
 
Temperature-indicating devices known as Seger cones or pyrometric cones are
commonly used in the ceramics industry. They consist of a fused oxide and glass
material that is formed into a cone shape. The tip of the cone softens and bends over
when a particular temperature is reached. Cones are available that indicate
temperatures
over the range from 600°C to C2000°C.

    

Intelligent temperature transmitters have now been introduced into the catalogues of
most instrument manufacturers, and they bring about the usual benefits associated with
intelligent instruments. Such transmitters are separate boxes designed for use with
transducers that have either a d.c. voltage output in the mV range or an output in the
form of a resistance change. They are therefore suitable for use in conjunction with
thermocouples, thermopiles, resistance thermometers, thermistors and broad-band
radiation
pyrometers. All of the transmitters presently available have non-volatile memories
where all constants used in correcting output values for modifying inputs etc. are stored,
thus enabling the instrument to survive power failures without losing such information.
Facilities in transmitters now available include adjustable damping, noise rejection,
self-adjustment for zero and sensitivity drifts and expanded measurement range. These
features allow an inaccuracy level of š0.05% of full scale to be specified.
c
    



Mention must be made particularly of intelligent pyrometers, as some versions of
these are able to measure the emissivity of the target body and automatically provide
an emissivity-corrected output. This particular development provides an alternative to
the two-colour pyrometer when emissivity measurement and calibration for other types
of pyrometer pose difficulty.
Digital thermometers (see section 14.2) also exist in intelligent versions, where the
inclusion of a microprocessor allows a number of alternative thermocouples and
resistance
thermometers to be offered as options for the primary sensor.
The cost of intelligent temperature transducers is significantly more than their
nonintelligent
counterparts, and justification purely on the grounds of their superior accuracy
is hard to make. However, their expanded measurement range means immediate
savings are made in terms of the reduction in the number of spare instruments needed
to cover a number of measurement ranges. Their capability for self-diagnosis and
selfadjustment
means that they require attention much less frequently, giving additional
savings in maintenance costs.

Butter end
3
 
  
3emiconductor-based temperature sensors are used in an increasing proportion
of temperature measurement applications, mainly due to the need to
monitor integrated circuit temperatures in many electronics applications.
3ensors based on simple transistor circuits can be readily incorporated as part
of an integrated circuit to provide on-board diagnostic or control capability.
The typical temperature range for these devices is ±55°C to 150°C. The
temperature range for some smart integrated circuits can be extended to
185°C.
The majority of semiconductor junction sensors use a diode-connected
bipolar transistor. If the base of the transistor is shorted to the collector (see
Figure 6.22) then a constant current flowing in the remaining p±n junction (base
c
    




to emitter) produces a forward voltage that is proportional to absolute

temperature. A constant forward voltage supplied to an ideal p±n semiconductor
junction produces a forward voltage drop that can be modelled by

_r

0F =


_ ln _ F

 _ (6.44)
3

where: k = Boltzmann¶s constant (1.38 Ú 10±23 J/K)

T = temperature (K)
q = charge of electron (1.6 Ú 10±19 C)
IF = forward current (A)
I3 = junction¶s reverse saturation current (A).
In practice, the overall forward voltage drop has temperature coefficient of
approximately ±2 mV/°C.
The basic components of a semiconductor sensor are a bipolar transistor
and a constant current supply. More sophisticated circuitry can be added to
improve linearity, precision and specific features such as high and low set
points. The sensor is produced in the form of a miniature integrated circuit and
enclosed in a standard electronic package such as a TO99 can, a TO-92 plastic
moulding or a DIP plug as illustrated in Figure 6.23.
There are a large number of types of semiconductor temperature sensor.
This is because of their versatility, not only as a temperature sensor but also
c
    




as they can be integrated into a controller. For temperature sensing only,

small-signal semiconductors such as the 2N2222 and 2N2904 transistors are
common choices. Integrated temperature sensors such as the LM35 from
National 3emiconductors, provide a linear output voltage of 10 mV/°C
operating in the range from ±40°C to 110°C. The TMP-01 programmable
temperature controller available from Analogue Devices gives a control signal
from one of two outputs when the device is above or below a specific
temperature that can be set by user-selected external resistors. The gate
threshold voltage of a power MO3FET can be used to determine the junction
temperature of a semiconductor component (Frank, 1999). 3emiconductor
temperature sensors can be incorporated within a larger integrated circuit to
produce a smart sensor that can control regions of the circuitry on a chip at
predetermined temperatures.
3emiconductor-based ICs suffer from a number of sources of uncertainty.
The slope of the sensor output characteristic is only nominally adjusted to the
quoted value. Using these devices without an external calibration circuit
involves significant uncertainties. The sensor characteristic is non-linear,
particularly at the ends of temperature ranges. This can be reduced by the
addition of further electrical circuitry. A self-heating error occurs due to the
forward current. This can be minimized by the use of low currents of the order
of 0.1 mA and below. The uncertainty associated with one semiconductor
device, the LM35, which has an output of 10 mV/°C, is 0.8°C.
The advantages of these devices are their linearity, simple circuitry, good
sensitivity, reasonable price and ready availability. As they are a highimpedance,
current-based unit they can be used for remote sensing with just
a twin copper cable required for connection purposes. 3uitable digital
voltmeters, some scaled in temperature units, are available for use with silicon
transistor sensors. 3uch voltmeters frequently incorporate the necessary power
supplies to operate the sensors. A disadvantage is the need to add extra
circuitry and for calibration to attain a reasonable uncertainty.

 


 
The forward voltage drop across a p n junction increases with decreasing
temperature. For some semiconductors, the relationship between voltage and
temperature is almost linear; in silicon this occurs between 400 K and 25K
c
    



with a corresponding sensitivity of approximately 2.5 mV/K. The two most
commonly used semiconductors for thermometry are GaAs and 3i. The
typical voltage sensitivities for these devices are illustrated in Figure 6.24. For
silicon at temperatures below about 25 K, when the forward voltage
approaches 1.1V, the characteristic function for the voltage temperature
relationship changes. As can be seen in Figure 6.24, 3i diodes give a lower
output than GaAs diodes but have better stability and are cheaper and easily
interchangeable. Generally rectifying diodes are used and these can be potted
in a small container. Commercial versions are available, for example the
1.25 mm in diameter and 0.75 mm long DT-420 device, from Lake 3hore
Cryotronics. Zener diodes have also been used to indicate temperature
(3zmyrka-Grzebyk and Lipinski, 1995).
The virtues of diode thermometers are their low price, a simple voltage±
temperature relationship, a relatively large temperature range (1 to 400 K), no
need for a reference bath or junction, relatively high sensitivity, an uncertainty
lower than ±50 mK (Krause and Dodrill, 1986) and simplicity of operation
with a constant current source and a digital voltmeter. Measurements down to
1K do, however, require careful calibration. Errors can occur if the supply
current is not a true DC but has an AC component due to, say, noise induced
in the circuit from improper shielding, electrical grounds or ground loops. To
minimize noise effects, all the instrumentation should be electrically shielded

and proper grounding techniques used (Morrison, 1998). The diode current
supply should have a single ground, generally at the voltmeter, which then
requires a floating current source. A current between 1_A and 100_A can be
used but generally needs to exceed 10_A in order to overcome noise
problems. However, this can cause problems associated with self-heating at
very low temperatures.

w  
   

c
    



Infrared thermometers can be classified in a number of groups based on the
principle of operation:
_ spectral band thermometers
_ total radiation thermometers
_ ratio thermometers
_ multiwaveband thermometers
_ special-purpose thermometers and methods
_ thermal imagers.

w  
Thermography or thermal imaging involves determining the spatial distribution
of thermal energy emitted from the surface of an object. This information
can be manipulated to provide qualitative and quantitative data of the
distribution of temperature on a surface. In their usual form they comprise an
optical system, a detector, processing electronics and a display. A typical
handheld thermal imager is shown in Figure 9.27 and images resulting from
the use of such devices in Figures 9.28 and 9.29. Thermal imagers do not
require any form of illumination in order to operate and this makes them
highly attractive to military and surveillance users. However, these devices
tend to optimized to produce an image rather than quantitative information on
the distribution of temperatures.
It is possible to make use of a single detector with some form of scanner to
transmit the radiation signal from specific regions of the optical system to
enable a two-dimensional image of the temperature distribution to be built up,
c
    



c
    




(Figure 9.30). This principle can be extended by the use of a linear detector
array as illustrated in Figure 9.31. An alternative is to use an array of detectors
and this form is known as a staring array (Figure 9.32). It is usually necessary
to cool single and linear array detectors because of the level of incident
c
    



energy. This can be achieved in a number of ways ranging from introducing
liquid nitrogen into a dewar incorporated into the device to the use of a
miniature 3terling engine.
The optimum waveband for a thermal imager, as for most other infrared
thermometers, is dictated by the wavelength distribution of the emitted
radiation, the transmission characteristics of the atmospheric environment
between the imager and the target and by the characteristics of the available
detector technology. The optical windows for air shown in Figure 9.15
between 3 and 5_m, with a notch at 4.2_m due to CO2 absorption, and
between 7.5 and 14_m make these a common choice for the selection of
detectors. The band 3±5_m is commonly referred to as medium-wave
infrared (MWIR) and that between 7.5 and 14_m as long-wave infrared
(LWIR). The emissivity of most naturally occurring objects and organic paints
is high in the long-wave infrared but is lower and more variable in the
medium-wave infrared. Metallic surfaces tend to have lower emissivity in
either band. As such, the use of a thermal image to provide quantitative
information for the temperature distribution, particularly of a surface
c
    




comprising different materials, has be carefully managed. Without correction

for local emissivity values the thermal imager will assume a default value and
apply this to the whole image.
Thermal imagers remain expensive devices although their cost is falling.
Prices range from over £71 500 ($100 000 year 2000 prices) for highperformance
military imagers to a few thousand pounds for uncooled imagers.
The price reflects the performance, ruggedness and image-processing
capability. 3ome compact imagers can be readily hand held whilst other
systems are designed to be mounted on a platform with the imager weighing
c
    



as much as 100 kg. The uncertainty associated with the temperature



measurement is specific to the device but typical figures are 2 K or 2% fullscale
output (Runciman, 1999). Table 9.7 provides some information about
commercial imagers available at the time of press. Most of the devices now
commercially available are particularly easy to use. They must simply be
aimed at the target of interest and the image captured by pressing a button on
the camera. Alternatively, some devices allow the image to be continuously
streamed to memory on board the camera or via a PC link.

 3

  

An alternative technique for monitoring temperature in gases is the

observation of spontaneous Rayleigh and Raman scattering. 3cattering in this
context is the absorption and re-emission of electromagnetic radiation by
atoms and molecules.

Rayleigh scattering is the elastic scattering of light by molecules or

very small particles less than about 0.3_m in size. The intensity of
Rayleigh scattering is proportional to total number of particles, x, and the
irradiance L:

If the Rayleigh cross-section is kept constant, the temperature of the probed

volume can be found. The calibration constant, î, can be determined by
measuring a reference temperature under known conditions and the measured
temperature related to this, assuming that the probe volumes for the reference
and test conditions are equal, by

Rayleigh spectra can be obtained using continuous wave and pulsed lasers to
c
    



excite the flow. The principal components for scattering based measurements
consist of an optical system and spectrometer to observe the gas sample and
a pulsed or continuous wave laser to excite the particles in the gas. A typical
system is illustrated in Figure 10.5. In Rayleigh scattering the collected signal
will typically be a factor of 109 smaller than the pump signal from the laser,
making it susceptible to corruption from other processes such as Mie
scattering, optical effects and background radiation. Also, in order to analyse
the spectra it is usually necessary to know the individual concentrations of the
species in the flow. As a result this technique must be undertaken very
carefully. The range for Rayleigh scattering is from approximately 293 K to
9000 K (e.g. see Farmer and Haddad, 1988). Applications have included
plasmas (Murphy and Farmer, 1992; Bentley, 1996), combustor flames (Barat
‘
, 1991), sooting flames (Hoffman ‘
, 1996), and supersonic flows
(Miles and Lempert, 1990).

When a molecule is promoted by incident radiation from the ground state

to a higher unstable vibrational state it can either return to the original state,
which is classed as Rayleigh scattering as discussed above, or to a different
vibrational state which is classed as Raman scattering. Raman scattering
gives rise to 3tokes lines on the observed spectra (see Kittel, 1986).
Alternatively, if a molecule is in an excited state it can be promoted to a
higher unstable state and then subsequently return to the ground state. This
process is also classed as Raman scattering and gives rise to an anti-3tokes
line on the observed spectrum. Raman scattering involves the inelastic
scattering of light from molecules. There are two basic methods for
determining the temperature by Raman scattering: the 3tokes±Raman
method and the 3tokes to anti-3tokes ratio method. The 3tokes±Raman
method is based on measurements of the density of the non-reactive species
assuming uniform pressure and ideal gas conditions. The 3tokes to anti-
3tokes ratio method involves measurement of the scattering strengths of the
3tokes to anti-3tokes signals of the same spectral line. The temperature can
then be calculated utilizing the Boltzmann occupation factors for the lines in
question (Edwards, 1997). This process is generally only suitable for high
c
    



combustion temperatures due to the relative weakness of the anti-3tokes
signal. The uncertainties in temperature measurement utilizing Raman
spectroscopy are discussed by Laplant ‘

(1996). Raman scattering
spectra can be observed by an optical system and spectrometer and using a
pulsed or continuous wave laser to excite the flow. The setup is similar to
that for Rayleigh scattering as illustrated in Figure 10.5. The range and
accuracy for Raman scattering are approximately 20 to 2230°C and 7%
respectively. Applications have included: reactive flows (Dibble ‘
,
1990); flames (Burlbaw and Armstrong, 1983); atmospheric temperature
observation (Vaughan ‘
, 1993).