Chapter 3

Temperature Sensors
1. Thermoresistive sensors
2. Thermoelectric sensors
3. PN junction temperature sensors
4. Optical and acoustic temperature sensors
5. Thermo-mechanical sensors and actuators
 A bit of history
Temperature measurements and
thermometers
1600 - thermometers (water expansion, mercury)
1650 - first attempts at temperature scales (Boyle)
1700 - standard temperature scales (Magelotti,
Renaldini, Newton) these did not catch
1708 - Farenheit scale (180 div between freezing
and boiling points of water)
1742 - Celsius scale (100 div between freezing and
boiling points of water)
1848 - Kelvin scale (based on Carnots
thermodynamic work)
1927 - IPTS - International Practical Temperature
Scale
 More history - sensors
Temperature sensors are the oldest sensors
1821 - Seebeck effect (Thomas Johann Seebeck)
1826 - first temperature sensor - a thermocouple - based
on the Seebeck effect (by Antoine Cesar Becquerel)
1834 - Peltier effect (Charles Athanase Peltier).
First peltier cell built in 1960s
Used for cooling and heating
1821 - discovery of temperature dependence of
conductivity (Sir Humphrey Davey)
1871 - William Siemens builds the first resistive sensor
made of platinum
 Temperature scales
Temperature scales
Centigrade (Celsius (C)) scale: 0C at freezing point of water,
100C at boiling point of water
Standard scale (K) based on the Kelvin scale
Absolute zero at -273.15 C
Freezing point of water at 273.15 K,
Boiling point of water at 373.15 K
Fahrenheit scale (F)
Freezing point of water at 32 F
Boiling point of water at 212 F
Absolute zero at -459.67 F
Conversion between scales
From C to K: N [C ] = (N + 273.15) [K]
From C to F: P [C] = (P 1.8 + 32) [F]
From K to C: M [K] = (M 273.15) [C]
From F to C: Q [F] = (Q 32)/1.8 [C]
From K to F: S [K] = (S 273.15) 1.8 + 32 [F]
From F to K: U [F] = (U 32)/1.8 + 273.15 [K]
 Heat
Heat is a form of energy units of joule [J]
1 J = 1 W.s, 1 kWh = 3.6 MJ, 1 cal = 0.239 J
(the kWh and the calorie (cal) are not SI units)
Thermal conductivity: the ability of materials to conduct heat.
Denoted as k, units: (W/m/K)
Heat capacity: Amount of heat (energy) necessary to change the
temperature of a substance by a given amount. Denoted as C,
units: (J/K)
Molar heat capacity: energy necessary to change the
temperature of one mole of substance. Units: (J/mol/K)
Specific heat capacity: energy necessary to change the
temperature of 1 kg of substance by 1 K. Units: (J/kg/K)
Volumetric heat capacity: energy necessary to change the
temperature of 1 m3 of substance (gas) by 1 K. Units: (J/m3/K)
Often specific heat capacity is given in units of (J/g/K) and that
of volumetric heat capacity in units of (J/cm3/K)
 Temperature sensors -
general
Temperature sensors are deceptively simple
Thermocouples - any two dissimilar materials,
welded together at one end and connected to a
micro-voltmeter
Peltier cell - any thermocouple connected to a DC
source (polarity defines heating or cooling)
Resistive sensor - a length of a conductor
connected to an ohmmeter
More:
Some temperature sensors can act as actuators as well
Can be used to measure other quantities that can affect
temperature (electromagnetic radiation, air speed, flow,
etc.)
Many newer sensors are semiconductor based
 Temperature sensors - types
Thermoelectric sensors
Thermocouples and thermopiles
Peltier cells (used as actuators but can be used as sensors)
Thermoresistive sensors and actuators
Conductor based sensors and actuators (RTDs)
Semiconductor based sensors - thermistors, diodes
Semiconductor junction sensors
Others
Based on secondary effects (speed of sound, phase of light)
Indirect sensing (infrared thermometers chapter 4)
Expansion of metals, bimetals (actuators or sensors-
actuators)
 Thermal actuators
A whole class of thermal actuators
Bimetalactuators
Expansion actuators
Thermal displays
Sometimes sensing and actuation is
combined in a single device
 Thermoresistive sensors
Two basic types:
Resistive Temperature Detector (RTD)
Metal wire
Thin film
Silicon based
Thermistors (Thermal Resistor)
NTC (Negative Temperature Coefficient)
PTC (Positive Temperature Coefficient)
 Thermoresistive effect
Conductivity depends on
temperature
Conductors and
semiconductors
Resistance is measured, all
other parameters must stay R= L
constant. sS
Resistance (R) of a length of R = L
sS
wire
T is temperature [C ]
Conductivity (s) of a s = s0
s0 - conductivity of the
1 + a T - T0
conductor is: conductor at reference
temperature T0.
Resistance as a function of R T = L 1 + a T - T0 T0 usually 20C but
temperature R(t): s0 S may be any convenient
temperature.
a - Temperature Coefficient a - given at T0
of Resistance (TCR) [C]
 Example
Copper: s0=5.9x107 S/m, a=0.0039 C at T0=20C.
Wire of cross-sectional area: 0.1 mm2, length L=1m,
Change in resistance of 6.61x10 /C and a base
resistance of 0.017 at 20C R=
L
=
1
s S 5.9 10 0.110
7
= 0.017W
-6

Change of 0.38% per C .

DR = R aDT = 0.017 ( 0.0039 1) = 6.6310 W
0
-5

Conclusions from this example:

Change in resistance is measurable (and fairly large)
Base resistance must be large - long and/or thin
conductors or both
Other materials may be used (platinum, nickel, etc.)
depending on requirements
 Temperature Coefficient of
Resistance
Material Conductivity s [S/m] Temperature Coefficint of
Resistance (TCR) C-1
Copper (Cu) 5.8 x 107 0.0039
5
Carbon (C) 3.0 x10 -0.0005
6
Constantan (60%Cu,40%Ni) 2.0 x10 0.00001
Cromium (Cr) 5.6 x 10 6 0.0059
Germanium (Ge) 2.2 -0.05
7
Gold (Au) 4.1 x 10 0.0034
7
Iron (Fe) 1.0 x 10 0.0065
6
Mercury (Hg) 1.0 x 10 0.00089
6
Nichrome (NiCr) 1.0 x 10 0.0004
7
Nickel (Ni) 1.15 x 10 0.0069
6
Platinum (Pl) 9.4 x 10 0.003926 (at 0C)
-6
Silicon (Si) (pure) 4.35 x 10 -0.07
7
Silver (Ag) 6.1 x 10 0.0016
6
Titanium (Ti) 1.8 x 10 0.042
7
Tungsten (W) 1.8 x 10 0.0056
7
Zinc (Zn) 1.76 x 10 0.0059
7
Aluminum (Al) 3.6 x 10 0.0043
Notes:
1. Instead of conductivity s [S/m], some sources list resistivity r, measured in ohm.meter r
= 1/s [Wm). 1S/m=1/Wm

2. TCR given at 20C unless otherwise indicated.

 Other considerations
Tension or strain on the wires affect resistance
Tensioning a conductor, changes its length and cross-sectional
area (constant volume)
has exactly the same effect on resistance as a change in temperature.
increase in strain on the conductor increases the resistance of the
conductor (to be discussed in Chapter 6)
Resistance of the sensor should be relatively large (25 and up)
RTDs transfer functions are not strictly linear. For short lengths
of wire, the linear representation in Eq. (3.4) is sufficient:

For more accurate sensing, the Callendar-Van Dusen equation

is used:
For T 0C
For T < 0C
The coefficients can be calculated from experiment. They are available in
tables for various materials.
 Construction - wire RTD
A spool of wire (length)
Similar to heating elements
Uniform wire
Chemically and dimensionally stable in the sensing range
Made very thin (<0.1mm) for high resistance
Spool is supported by a glass (pyrex) or mica
supports
Similar to the way the heating element in a hair drier is supported
Keeps strain at a minimum and allows thermal expansion
Smaller sensors may not have an internal support.
Enclosed in a glass, ceramic or metal enclosure
Length of the sensor is from a few cm, to about 50cm
s
u m wire
n
plati

form
e
osur
encl
 Construction (cont.)
Platinum - used for precision applications
Chemically stable at high temperatures
Resists oxidation
Can be made into thin wires of high chemical purity
Resists corrosion
Can withstand severe environmental conditions.
Useful to about 800 C and down to below 250C.
Very sensitive to strain
Sensitive to chemical contaminants
Wire length needed is long (high conductivity)
Nickel and Copper
Less expensive
Reduced temperature range (copper only works up to about 300C)
Can be made into thin wires of high chemical purity
Wire length needed is long (high conductivity)
Copper is not suitable for corrosive environments (unless properly protected)
At higher temperatures evaporation increases resistance
 Example
Required: Platinum wire RTD: pure platinum, 0.01 mm in diameter.
Required resistance: 25 at 0C.

At 100C:
Resistance (using the linear formula):

Resistance ranges from 25 at 0C

to 35.652 at 100C
At 0C:

Change in resistance per C:

Length:

Sensitivity: 0.1065 /C:

 Thin Film RTDs
Thin film sensors: produced by depositing a thin layer of a suitable
material (platinum or its alloys) on a thermally stable, electrically non-
conducting, thermally conducting ceramic
Etched to form a long strip (in a meander fashion).
Eq. (3.1) applies but much higher resistance sensors are possible.
Small and relatively inexpensive (some are only a few mm2)
Often the choice in modern sensors especially when the very high
precision of Platinum wire sensors is not needed.
Some Parameters
Temperature range: -250 C to 700 C
Resistances: typically 100 (higher available)
Sizes: from a few mm to a few cm in length
Compatibility: glass, ceramic encapsulation
st rip
Available in ready made probes
at in um
Accuracy: 0.01 C to 0.05 C pl
Calibration: usually not necessary beyond
manufacturing
tra te
subs
 Self heat in RTDs
RTDs are subject to errors due to rise in their temperature
produced by the heat generated in them by the current used to
measure their resistance
Wire wound or thin film
Power dissipated: Pd=I2R ( I is the current (RMS) and R the
resistance of the sensor)
Self heat depends on size and environment
Given as temperature rise per unit power (C/mW)
Or: power needed to raise temperature (mW/ C)
Self-heat errors are of the order of 0.01C/mW to 10C/mW
(100mW/C to 0.1mW/C)
Given in air and in water
In water values are lower (higher if mW/C used)
Self heat depends on size, construction and environment
Lower in large elements, higher in small elements
Important to lower the current as much as possible
 Example
A 100 (at 0C) platinum RTD operates between -200C and +850C and has a
TCR of 0.00385/C. Its self heat is 0.08C/mW. To sense a 1 mA current passes
through the RTD. We wish to calculate the error due to self heat.

The resistances at the extremes are calculated using the Callendar-Van Duzen Equation
with the coefficients listed for platinum (see. Eqs. (3.6) and (3.7)

Power dissipated in the RTD at the temperature extremes

At 850C the error is 0.57%

 Connection of RTDs
RTD measurements are subject to errors due to the connecting
wires resistance and temperature variations along these wires
In low resistance RTDs the connecting lead wires may
introduce significant errors
Compensation schemes take into account the wiring
To do so there are various (standard) lead wire connection
configurations available

R lead R lead R lead

A A A
R lead
A A A
R lead R lead R lead
B
R lead R lead R lead R lead
B B B A
B B
a. b. c. R lead d. R lead

a. uncompensated, b. three wire connection, c. four wire connection, d. two wire with compensation loop
The three and four wire style allow compensation for temperature variations and resistance.
Lead wires must be the same length. The method of compensation will be discussed in Chapter 11
 Response time in RTDs
Response time
Provided as part of data sheet
Given in air or in water or both, moving or stagnant
Given as 90%, 50% (or other) of steady state
Generally slow
Wire RTDs are slower
Typical values
0.5 sec in water to 100 sec in moving air
 Silicon Resistive Sensors
Conduction in semiconductors
Valence electrons
Bound to atoms in outer layers (most electrons in pure semiconductors)
Can be removed through heat (band gap energy)
When removed they become conducting electrons (conduction band)
A pair is always released - electron and hole
Conductivity of semiconductors is temperature dependent
Conductivity increases with temperature
Limited to a relatively small temperature range
(Much more will be said about semiconductors in Chapter 4)
Conductivity of a semiconductor is:

(e is the charge of the electron, n, p are concentration of electrons and holes and me, mp, their
mobilities)
In a doped semiconductor, with a dominant species
concentration:

(e is the charge of the electron, nd is the concentration of dominant species and me its mobility)
Note: Much more on this will be discussed in Chapter 4
Silicon Resistive Sensors (cont.)
Pure silicon: negative temperature coefficient (NTC)
Produces NTC device - Resistance decreases with
temperature
Resistance in pure silicon is extremely high
Need to add impurities to increase carrier density
N type silicon - add arsenic (As) or antimony (Sb)
Behavior changes through doping:
N-type silicon (doping with arsenic (As) or
antimony (Sb) produce positive temperature
coefficient (PTC) sensors
Above about 200C, the behavior of pure silicon
changes to NTC
 Silicon resistive sensors
Silicon resistive sensors are somewhat nonlinear and offer
sensitivities of the order 0.5-0.7 %/C.
Can operate in a limited range of temperatures like most
semiconductors devices based on silicon
Maximum range is between 55C to +150C.
Typical range: - 45C to +85C or 0C to +80C
Resistance: typically 1k at 25 C.
Can be calibrated in any temperature scale
Made as a small chip with two electrodes and encapsulated in
epoxy, metal cans etc. 2.4
Resistance is calculated using a

normalized resistance
2
form of the Callendar-Van Dusen
Equation 1.6

1.2
1
0.8
HOT stands for higher order terms indicating
0.4
that additional coefficients may be used to -50 0 25 50 100 150
temperature [C]
improve representation
 Thermistors
Thermistor: Thermal resistor
Became available: early 1960s
Based on oxides of semiconductors
High temperature coefficients
NTC or PTC
High resistances (typically)
Small, simple, inexpensive
Most are NTC devices
Some are PTC devices (achieved by use of specific materials)
PTC thermistors are not as common
Advantageous when runaway temperatures are possible
 Thermistors (cont.)
Transfer function: Two approximations
Simple exponential model

(K) is a constant (called material constant)

R(T): resistance of the device
T: temperature in K, T0 the temperature at which R0 is given
Relation is nonlinear but:
Only mildly so (/T is small)
Approximate transfer function
The Steinhart-Hart model
The coefficients , or a, b and c are calculated from experiment or known
resistances at specific temperatures. The Steinhart-Hart model is more accurate

The inverse relation (calculating temperature from known resistance) is often used
 Construction
Beads
Chips
Deposition on substrate

PCB deposited thermistors (dark rectangles)

Structure of bead and chip thermistors

Some small bead and chip thermistors
 Thermistors - properties
Self heating errors as in RTDs but:
Usually lower because resistance is higher (low current)
Typical values: 0.01C/mW in water to 1C/mW in air
Wide range of resistances up to a few M
Can be used in self heating mode
To raise its temperature to a fixed value for specific applications (flow sensing)
Repeatability and accuracy:
0.1% or 0.25C for good thermistors
Temperature range:
50 C to about 600 C
Ratings and properties vary along the range
Linearity
Can be consideredlinear for narrow range applications
Nonlinear for wide temperature ranges
Available in a wide range of sizes, shapes and also as probes of various
dimensions and shapes
Some inexpensive thermistors have poor repeatability - these must be
calibrated before use.
Overall much less expensive than RTDs
 Example
An unknown thermistor is rated as 10 k at 25 C. Its resistance is measured at
0 C as 29.49 k. We wish to evaluate its resistance at 50 C.

Using the simpler model, we can now evaluate the constant b. From the general
relation (Eq. (3.12)):

For the values given (T0 = 25C = 298.15 K, R0 = 10 k)

That is:

At 50 C = 323.15 K:

Note the large change. The resistance changes fro 10 k to 4 k in 25 C.