400 Commonwealth Drive, Warrendale, PA 15096-0001 U.S.A. Tel: (724) 776-4841 Fax: (724) 776-5760 Web: www.sae.

org
SAE TECHNICAL
PAPER SERIES
2004-01-1421
The Development of a RTD Temperature
Sensor for Exhaust Applications
C. Scott Nelson, David Chen,
Joseph Ralph and Eric DHerde
Delphi Corporation
Reprinted From: Diesel Exhaust Emission Control
(SP-1860)
2004 SAE World Congress
Detroit, Michigan
March 8-11, 2004
All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or
transmitted, in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise,
without the prior written permission of SAE.
For permission and licensing requests contact:
SAE Permissions
400 Commonwealth Drive
Warrendale, PA 15096-0001-USA
Email: [email protected]
Fax: 724-772-4891
Tel: 724-772-4028
For multiple print copies contact:
SAE Customer Service
Tel: 877-606-7323 (inside USA and Canada)
Tel: 724-776-4970 (outside USA)
Fax: 724-776-1615
Email: [email protected]
ISBN 0-7680-1319-4
Copyright 2004 SAE International
Positions and opinions advanced in this paper are those of the author(s) and not necessarily those of SAE.
The author is solely responsible for the content of the paper. A process is available by which discussions
will be printed with the paper if it is published in SAE Transactions.
Persons wishing to submit papers to be considered for presentation or publication by SAE should send the
manuscript or a 300 word abstract of a proposed manuscript to: Secretary, Engineering Meetings Board, SAE.
Printed in USA
Copyright 2004 SAE International
ABSTRACT
A RTD (resistive temperature device) high temperature
sensor was developed for exhaust gas temperature
measurement. Extensive modeling and optimization
was used to supplement testing in development. The
sensor was developed to be capable of withstanding
harsh environments (-40 to 1000C), typical of engine
applications, including poisons, while maintaining high
accuracy (< 0.5% drift after 500 hrs of aging at 950C).
The following sensor characteristics are presented:
resistance-temperature curve, accuracy, response time,
and long-term durability. In addition, a system error
analysis program was developed with representative
results.
INTRODUCTION
With emissions regulations becoming more stringent,
improved control on existing and new emissions devices
are becoming necessary. Some new emissions devices
are very sensitive to temperature, they will not work
below certain temperatures, and can be damaged if the
temperature goes too high. Examples of such devices
are: Diesel Particulate Filters (DPF), Selective Catalytic
Reduction (SCR) and Diesel Oxidation Catalysts. The
need to measure temperatures on these technologies
has been discussed previously. [1]
While temperature sensors have been around for
hundreds of years, the ability for them to withstand harsh
conditions has always been difficult to overcome.
Engine exhaust environments are extremely hostile to
sensors. With temperature ranges from 40C to over
1000C, flow velocities from 3 m/s to over 200 m/s,
unburned combustion products, poisons from both oil
and fuel, severe vibrations, and impact from road debris,
it is challenging to attain durability requirements of up to
700,000 km. This paper discusses the development of a
high temperature sensor capable of meeting these
requirements.
SENSOR TYPES
Practically speaking, there are 3 sensor technologies
that are viable in vehicle exhaust: thermocouples,
thermistors, and resistance temperature detectors
(RTD). Each have their own advantages and
disadvantages, and will be discussed in brief detail.
Thermocouples
Thermocouples have been around the longest of the
three sensor types. They are formed by welding two
dissimilar metals together forming a bimetallic junction
that produces a voltage which varies with temperature.
For a vehicle application, a type K (chromel-alumel) or
type R or S (platinum-rhodium) would be used for the
range of temperatures previously mentioned.
Thermocouples can be relatively low cost sensors (if
using type K) as compared with thermistors or RTDs.
However, the electrical system is significantly more
expensive since there must be compensation for
voltages produced whenever there is a change in wire
material (often called cold-junction compensation).
Thermocouples can be made with very little mass which
allows for a fast response with changing temperature. In
order for the sensor to minimize drift and be durable in a
vehicle exhaust environment however, the thermocouple
must be protected by a sheath, and made thicker. Thus
much of the fast response advantage is eliminated.
Thermistors
Thermistors are made from various nonmetallic
conductors (i.e. metal oxides). The types of thermistors
found in a vehicle exhaust environment will typically
produce a negative temperature coefficient (NTC),
meaning the resistance will decrease with increasing
temperature. Thermistors offer a high sensitivity over a
smaller range in temperature than either thermocouples
or RTDs. At 0C the resistance can be over 100,000 O,
at 200C 200 to 500 O, and at 800C 50O. Thus,
thermistors can achieve very high sensitivities over a
2004-01-1421
The Development of a RTD Temperature Sensor
for Exhaust Applications
C. Scott Nelson, David Chen, Joseph Ralph and Eric DHerde
Delphi Corporation
particular range of temperatures. However, achieving
nearly the same accuracy over a large range in
temperatures is not possible (unless several pull up
resistors are used) due to the highly nonlinear
characteristic response.
Thermistors can be made very small for quick response.
However, they are not able to withstand even mild
vehicle exhaust environments without being protected by
a metal or ceramic insulated sheath thus causing the
sensor response to be relatively slow.
Tolerance of a thermistor depends on its intended range
of use. Thermistor tolerances in manifold air
temperature sensors (MAT) or coolant sensors are very
tight over the relatively narrow range of measurement
(ex. 0.6C from 0 to 100C). However in a vehicle
exhaust that can vary between -40C and 1000C,
thermistors have a fairly poor tolerance depending on
the temperature range (2% to > 6% of temperature).
As previously mentioned, thermistors typically have a
very high resistance below 100C. This makes it difficult
to meet requirements of being able to read the sensor at
-40C, or being able to perform OBD II start up
diagnostics at 20C.
Unlike thermocouples or RTDs, each manufacturer of
thermistors have their own characteristic temperature vs.
resistance curve. These curves can vary significantly
and thus are often not interchangeable without pull up
resistor hardware changes [2].
RTD
Resistance temperature detectors or RTDs are based on
the natural change in a metals resistance with
temperature. In vehicle exhaust applications the
material of choice is platinum due to its capability to
span the entire range of temperatures while having a
near linear output. Industry standard for a RTD is 100O
at 0C, however the standard resistance for vehicle
exhaust applications is 200O at 0C. The higher
resistance doubles the sensitivity and makes vehicle
wiring even less significant.
Both thermistors and RTDs can be affected by self
heating or Joule heating. This is caused by power loss
across the sensor element caused by the measurement
current. The lower the pull up resistor used to measure
the sensors resistance (to be covered later), the larger
the measurement current, and thus the self heating.
However, self heating is generally a concern only in still
air; even at idle flow velocities, the power loss generated
in the sensor element is carried away by the exhaust
flow making the self heating affect insignificant.
There are two types of RTDs: wire wound and film type.
Wire wound is made by winding a thin platinum wire
around a mandrel until the desired resistance is reached
being careful not to stress the wire. While wire wound
has the highest accuracy, they typically have a relatively
large thermal mass causing them to have a slow
response. The film type RTD is typically made by
applying a thin film of platinum onto a ceramic forming a
serpentine. The element is then passivated for high
temperature environments. Thin film RTDs cost less
than wire wound, have fast response times (if the
element does not use a metal sheath for passivation),
and are the selection of choice for vehicle exhaust
environments.
COMPARISON OF SENSOR TYPES
Table 1., shows a summarized comparison of the three
different sensor types. The RTD is the most accurate of
the three sensor types over the entire measurement
range over time. It is important to note that zero hour
accuracy is only one portion of total accuracy; how the
sensor performs in the application over time at
temperature is vital.
Characteristic Thermocouple Thermistor RTD
Linearity Poor Poor Good
Accuracy Fair Fair to
poor
Excellent
Sensitivity
<600C
>600C
Poor
Poor
Excellent
Fair
Good
Good
Signal Level
<600C
>600C
Poor
Poor
Excellent
Fair
Good
Good
Response
time
Poor
(3 mm dia)
Poor Good
(No
sheath)
System
Complexity
Poor Good Excellent
Standardized
output signal
Excellent Poor Good
-40C
measurement
Excellent Poor Excellent
20C OBD II
measurement
Excellent Poor Excellent
Table 1. Comparison of three sensor technologies
[2, 3, 4, 5]
Figure 1., shows two thermistors and a typical RTD
designed for vehicle exhaust use (thermocouples are not
shown since they generate an output voltage instead of
measuring a resistance).
As can be seen from the graph, the thermistors show a
very non-linear curve as compared to the RTD. Further,
there is a significant difference between the two different
thermistor manufacturers, with one not being practical
for measuring temperatures below 100C. The
tolerances indicated on the graph show that the RTD
has a much tighter tolerance than thermistors over the
measurement range.
Thermocouples have the lowest sensitivity, followed by
RTDs or thermistors depending on the temperature
range. Below a range of 500 to 600C, thermistors have
very high sensitivity. Above 600C the RTD has the
highest sensitivity. Signal levels are again lowest with
thermocouples followed by RTD and then thermistors up
to 200C to 600C (depending on the type of thermistor).
Linearity of the sensor is also a desirable trait. The
platinum RTD is the most linear followed by
thermocouples. Being linear provides the ability for only
one pull up resistor to be used over the entire
measurement range without losing accuracy.
TEMPERATURE SENSOR REQUIREMENTS
Temperature
Typically the low-end temperature requirement is
consistent at -40C with a need to accurately measure
temperature at 20C for OBD II diagnostics. High-end
temperature requirements depend on the application.
For example, any of the following can have a significant
impact on the maximum temperature requirement:
Engine type (Diesel, or spark ignition), turbo charging,
location of sensor in exhaust pipe, the size and load of
the engine, type of emission devices and control
mechanization. Typical temperature maximums for
diesel engines are < 850C and for spark ignition
<1050C.
The time spent above 800C is critical to sensor
accuracy after aging. Even short durations above
1000C can significantly change the sensor accuracy.
When specifying the maximum temperature of the
sensor, a time at temperature specification should be
indicated for temperatures above 800C, 900C and
1000C showing both duration for a single instance as
well as a percent of vehicle life.
0
100
200
300
400
500
600
700
800
900
1000
-40 60 160 260 360 460 560 660 760 860 960
Temperature (C)
R
s

(
o
h
m
s
)
RTD
Thermistor A
Thermistor B
Not intended for
use below 100 C
Tolerance:
+/- 6 C @ 150 C
+/- 5 C @ 300 C
+/- 13 C @750 C
Platinum RTD
Tolerance:
1% of resistance
Tolerance:
+/-10 C -40 to 600 C
+/- 30 C 600 to 700 C
+/- 40 C 700 to 800 C
+/- 60 C 800 to 1000 C
@ -40 C =
96,065 ohms
@ 100 C =
156,000 ohms
Exposure to Exhaust Contaminants
Certain exhaust constituents, found in either the fuel or
oil, can contaminate the temperature sensor and cause
a resistance shift over the life of the sensor. The
concentration of the constituent, air/fuel ratio of the
exhaust, and temperature of the exhaust all contribute to
how much the temperature sensor will be poisoned and
cause a resistance shift. Typically poisoning of the
exhaust sensor starts to be a concern above 600C
depending on the level of passivation.
Time Response
Time response is typically defined as the time it takes to
reach one time constant (63% of final value) as
temperature increases. The need for how fast the
sensor responds depends on the application. In general,
the faster a sensor can respond, the sooner decisions
can be made which might enable emissions
improvement (tighter control) or provide protection to
emission devices. The time response is highly
dependent on flow velocity. Some applications may not
be concerned with idle conditions, thus it is important to
note the time response for the flow velocity at which the
sensor information will be used. Typical response times
for sensors range from 3 to 16 seconds.
Figure 1. Comparison of sensor technologies
Sensor Durability
Heavy duty diesel engines have a requirement of
700,000 km, while light duty diesels and spark ignition
engines are typically 240,000 km or less. Since
temperature sensors will be exposed to the same
physical conditions as other exhaust sensors, the
requirements would be the same. For spark ignition
engines these requirements would be equivalent to
oxygen sensors which have been used for 30 years and
have a very well developed durability requirement.
Dimensions
Dimensions are dependent on vehicle application. If the
application temperatures are lower, the sensor package
will be cooler, thus the temperature sensitive
components can be moved closer towards the exhaust,
and the sensor height can be reduced. Often, the
exhaust system will have other exhaust sensors
packaged into the same pipe. This could mean that the
seat to seal height of the sensor might be the same as
other exhaust sensors, depending on location. Likewise,
the thread size of exhaust sensors has been
standardized to M18 thread in spark ignition engines,
thus it would make sense to keep this thread size
consistent. Internal (female) pipe thread on the exhaust
pipe has long been preferred so as to not damage the
threads during vehicle assembly.
Insertion Depth
Insertion depth is again dependent on the application,
with the largest variable being the size of the pipe and
the flow velocity of interest. There is typically no need to
locate the sensor in the center of the exhaust flow, since
the temperature profile (as will be shown later) has a
fairly flat shape depending on Reynolds number. At low
Reynolds number the profile is less flat and can affect
the exhaust temperature measurement. Oftentimes
however, low flows are of little interest in temperature
measurement. As will be shown later, it is fairly simple
to compensate for insertion depth if the exhaust flow is
known.
Another decision to be made is whether the peak
temperature, or bulk temperature is desired. Peak
temperature is the maximum temperature in the pipe
cross section, typically located in the center of the pipe.
The bulk temperature is the average temperature of the
gas over the cross section. The bulk temperature can
be calculated by:
m
Tdr r
T
R
g
bulk

)
=
0
v 2 t
Accuracy
Accuracy has several components. The accuracy of the
sensor itself is defined as the ability to read a uniform
temperature, where the entire sensor is exposed to the
same temperature. This accuracy will vary with thermal
and chemical (poisons) aging.
Where

g
: density of the gas
v: velocity of the gas
m : mass flow rate of the gas.
The second component of accuracy is the difference in
temperature between the location actually measured and
the location desired. As previously mentioned this can
be compensated for if the exhaust flow velocity is known,
and the sensing element is not located in the boundary
layer.
Since the area increases with the radial distance from
the center of the pipe, there are larger quantities of
exhaust gas the further from the pipe center. The region
of the bulk flow and temperature varies with flow rate but
is roughly encompasses 2/3 to 3/4 of the pipe diameter.
The desired location of the temperature measurement
may be dependent on what the sensor is being used for.
If the sensor is used for applications such as over
temperature protection, peak temperature may be
desired. However, in other applications, the bulk
temperature may be desired.
The third component of accuracy is the system level
accuracy. The pull up resistor value and its tolerance as
well as the analog to digital (A/D) resolution can play a
large part in the overall accuracy of reading the sensor.
Lastly, whether a look up table or an equation is used to
convert the resistance measured to the temperature, can
make some difference since an equation rarely can
perfectly describe nature.
Performing a computational fluid dynamics (CFD)
analysis, on two different pipe diameters, the difference
between peak and bulk temperatures can be observed.
Figure 2, shows the percentage difference in
temperature between the bulk and peak temperatures,
along a 24 mm location point in a 127 mm ID exhaust
pipe. Table 2, describes the input and boundary
conditions.
SENSOR DEVELOPMENT
Based on the previous comparisons between the
different sensor types, it was decided that a RTD would
best fit vehicle exhaust applications. Further, a platinum
thin film planar sensor with a resistance of 200O at 0C
was chosen due to its stability and standardization.
Bulk
Temperature
Ambient
Temperature
Ambient
velocity
127 mm
pipe
500C 0C 22.5 m/s
57 mm
pipe
250C 0C 22.5 m/s
Insertion Depth
Insertion depth is dependent upon the unsupported
cantilever length of the element, thickness, and element
cost/mm of length vs. performance. The unsupported
cantilever length must be designed to withstand vibration
levels (that have enough amplitude to excite the
element) higher than seen in vehicles.
Table 2. CFD model boundary conditions
The longer the sensing element, the higher the cost.
This is partially due to the cost of the platinum on the
sensing element, and partially due to packaging. To
optimize the cost, it is important to know the exhaust
temperature profile under the conditions in which the
temperature information will be used. The temperature
profile is mainly dependent on flow velocity and pipe
diameter. In most situations, the velocity profile is very
uniform except in the boundary layer close to the pipe
wall. Only at a very low idle condition in a large diesel
exhaust pipe, will the exhaust temperature profile have a
significant difference in temperature between peak and
measured locations of the sensor.
Figure 2. Comparison with bulk temperature
for 127 mm ID pipe
120 40 60 80 100
Internal Flow velocity (m/s)
20
10%
0%
0
2%
4%
6%
8%
Peak temperature
24 mm Sensor Location
127 mm ID Pipe
D
i
f
f

f
r
o
m

b
u
l
k

t
e
m
p
As shown in Figure 2, there can be a significant
difference between peak temperature at the center of the
pipe and the bulk temperature, which is located at about
25% of the pipe diameter. The 24 mm insertion depth
location follows the peak temperature shape quite well.
A sensor located between the boundary layer and the
pipe center can determine peak or bulk temperature by
compensating for the difference in temperature since the
shapes are very similar.
Figure 3, shows a CFD analysis on a 57 mm ID pipe. As
shown, the peak temperature and the temperature at a
24 mm insertion depth lie almost directly on top of each
other, indicating no practical difference in temperature
between the two locations.
Element Thickness
The thickness of the element depends on the cantilever
length (which was determined in the previous section)
and the desired response time of the element. For a
given width, the thickness will determine the mass of the
element, which will in turn be directly related to the
response time of the sensor.
Response Time
Several factors influence the response time of the
temperature sensor; they include the exhaust gas flow
conditions such as temperature, flow velocity, lower
shield design and pipe geometry. In addition, the
element thermal conductivity and thickness also impact
the response time.
An analytical correlation was developed for the response
time of the sensor in exhaust gas. The heat transfer
equations can be expressed as [2]:
The solution to the above equation is:
nd the convection coefficient h can be obtained from
here
sity of the element
ent
ent
f the gas
s
rface
he properties of the gas and the element are a function
to the
is shown from above, that the response time is
A
the Nusselt number Nu, which is expressed as:
3 / 1 n
hd
g
k
Pr Re C Nu = =
v
= Re
Vd
( ) ( )
Pr Re
2 2 +
=
+
=
g
k
C
th w th w h
t
3 / 1 1
- -
n s s s s
d th w c th w c
W

s
: den
c
s
: specific heat of the elem
w: width of the element
D
i
f
f

f
r
o
m

b
u
l
k

t
e
m
p
57 mm ID Pipe
Peak temperature
24 mm Sensor location
200 100 150
Internal Flow velocity (m/s)
50
10%
8%
6%
4%
2%
0%
0
th: thickness of the elem
Re: Reynolds number
Pr: Prandtl number
V: average velocity o
d: diameter of the lower shield
v: dynamic viscosity of the gas
k
g
:
thermal conductivity of the ga
T
g
:
temperature of the gas
T
s
:
temperature of the element su
T
of temperature. Here the gas properties are evaluated
at temperature T = 1/3*T
g
+2/3*T
s
and the element
properties at temperature T = 2/3*T
g
+1/3*T
s
.
The coefficient C and exponent n are specific
Figure 3. Comparison with bulk temperature for a
57 mm ID pipe
lower shield selected and can be deduced from
measured data.
It
proportional to thickness of the sensor. Other gas
properties also affect the response time. A typical trend
of the response time vs. average pipe flow velocity is
shown in Figure 4. It is shown that the response time is
shorter for heating than for cooling. Also, heating and
cooling to different temperatures also have a slightly
different response time.
9.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
8.0
0 10 20 30 40 50
Flow Velocity (m/s)
R
e
s
p
o
n
s
e

T
i
m
e

(
s
e
c
)
Heating, from 100 C to 300 C
Heating, from 100 C to 500 C
Heating, from 100 C to 700 C
Cooling, from 700 C to 500 C
Cooling, from 700 C to 300 C
Cooling, from 700 C to 100 C
Figure 4. Response time vs. flow velocity at
various flow conditions
) )]( ( 2 [ ) (
s
T
g
T th w h
dt
s
dT
th w
s
c
s
+ = -
|
|
.
|

\
|
=

t
t
g s
e T T 1
TESTING
Element Construction
was constructed by sputtering a
ompletely sealing the sensing element from the
rotective Shield rotective Shield
Sensor Calibration Curve The sensing element
When the sensor is in an application, it is expected that
the sensor will have a temperature gradient from the
sensing tip to the connector wire. This temperature
gradient varies with different gas conditions inside and
outside the pipe and changes the resistance of the
leads. It was decided to include the resistance of the
conductor lead in the overall resistance measurement.
This will simplify the calibration electronics with a
negligible error since the lead is a very small fraction of
the total sensor resistance.
thin film of pure platinum onto an alumina substrate with
leads that have been platinum thick filmed. Creating the
serpentine pattern and trimming to a precise resistance
uses a proprietary method that allows for greater stability
of the platinum during thermal aging. After trimming, an
alumina protective cover plate is placed on the element
and sealed with a proprietary high temperature glass
sealant.
C
exhaust gases is critical to maintaining accuracy of the
sensor. Even if the sensor is hermetically sealed, during
high temperature excursions, the sealing glass can
become soft allowing poisons to pass through the
glass, contaminating the sensing element. Using this
special proprietary glass sealant allows the temperature
sensor to be directly exposed to the exhaust stream
without the need for a metal sheath, which would
drastically slow the sensors response time.
to pass through the
glass, contaminating the sensing element. Using this
special proprietary glass sealant allows the temperature
sensor to be directly exposed to the exhaust stream
without the need for a metal sheath, which would
drastically slow the sensors response time.
Two different calibration systems were used to calibrate
the sensor resistance vs. temperature curve. From
-40C to 240C, the sensors were set in a circulating
liquid bath to measure the resistance. At elevated
temperatures above 300C, a thermocouple calibrator
was used to measure the resistance.
The resistance vs. temperature curve is shown in Figure
5, which is represented by the best-fit quadratic equation
of the measured data. The values in the inserted table
are calculated based on this curve. As shown, the o
value is slightly lower than typical (i.e. 3.825 x 10
-3
) due
to incorporating the lead resistance into the curve fit.
PP
er shield in a temperature sensor is
ensor Packaging
The protective low
only useful prior to installation. Unlike other exhaust
sensors, which use the shield to limit exhaust flow past
the element, in a temperature sensor, the shield must
allow a maximum amount of flow past the element
without first absorbing or releasing heat. Therefore,
maximizing hole size and minimizing mass are important
factors in minimizing the affect the lower shield will have
on response time. Further, since the temperature
sensor does not have a heating element, and is
therefore almost always at the same temperature as the
exhaust, there is no need to be concerned with water
impingement on the sensor like there is with other
electrically heated exhaust sensors.
s
S
hat adapting an existing high volume
he current production packaging does not allow the
It was determined t
production planar exhaust sensor package would have
several advantages. Production packaging has already
demonstrated high physical durability and was designed
to meet most application temperatures in the same
exhaust environment that the temperature sensor will be
exposed to. This packaging, along with the previous
discussion determined the geometry of the sensing
element.
Figure 5. The calibration curve of the temperature
sensor
0
100
200
300
400
500
600
700
800
900
1000
-40 60 160 260 360 460 560 660 760 860 960
Temperature (C)
R

(
o
h
m
s
)
Temp (C) Rs (ohms)
-40 170.31
0 200.00
25 218.42
200 344.49
400 482.42
600 613.80
800 738.61
1000 856.85
R
s
= R
o
(1 + oT + |T
2
)
R
o
= 200 ohms
o = 3.6943x10
-3
| = - 4.1007x10
-
7
Thermal durability
Thermal durability without the influence of poisons was
performed in an air atmosphere oven, held at a constant
temperature for a given amount of time. Figure 6, shows
the effect of time vs. temperature for 750, 850, and
950C. As shown, almost no shift in resistance occurs at
either 750C or 850C up to the test time of 1,000 hours.
At 950C, the resistance shifts about 0.32% after 500
hours and 1.32% after 1,000 hrs. Thus temperatures
below 850C do not significantly affect the performance
of the temperature sensor. At temperatures above
850C, the time the sensor is exposed at these elevated
temperature becomes important. Therefore it is critical
to know the temperature-time histogram to determine
how much the sensor will change during the vehicle life.
T
temperature sensor to make a 90 bend in the package
outside the exhaust pipe. However making a 90 bend
may not be advantageous. Depending on the location of
the bend, oftentimes the bent sensor has the same
overall height as a production straight sensor with bends
in the wire. Further, installation can be more difficult
since having a bend does not allow for the use of a
socket wrench, which is typically used in high volume
manufacturing.
n
c
e
a

i
n

R
e
s
i
s
t
e
a
s
e
c
r
I
n
750 C Aging
850 C Aging
950 C aging
Aging hours
1200 1000 800 600 400 200
2.0%
1.5%
1.0%
0.5%
-0.5%
0.0%
0
-1.0%
2.0%
Aging Hours
300 250 200 150 100 50 0
Thermal Cycling with Poisons
A test was constructed to rapidly inflict a lifetime worth of
poisons on the temperature sensor in a way that might
be seen on a vehicle. A diesel exhaust simulation was
used for this test due to its tendency to burn larger
amounts of oil and for its 700,000 km durability
requirement. From previous experience with other
exhaust sensors, it is known when continuously running
at high temperatures, some poisons form a hard particle
and tends not to attach to the sensing element. Thus
the test cycle was designed so that for the majority of the
time, the sensing element would collect soft deposits
that would attach readily to the sensor. Then a high
temperature exhaust spike would drive the poison
particles on the surface into the glass seal. Figure 7,
shows a graph of the cycle.
This cycle was repeated so that the sensor would be
exposed to an equivalent amount of oil used in a large
diesels life, with an estimated equivalent time spent at
high temperature. Figure 8, shows that the sensors had
an average change of only 0.38% over an equivalent
700,000 km.
I
n
c
r
e
a
s
e

i
n
R
e
s
i
s
t
a
n
c
e
1.5%
1.0%
0.5%
0.0%
-0.5%
-1.0%
Figure 6. Oven aging at three different temperatures
Figure 8. RTD change in resistance after a
lifetime amount of exhaust poisons
Electrical System Integration
The electrical interface configuration for the temperature
sensor is shown in Figure 9.
U
p
= 5V
U
o
: output voltage
Temp Sensor
R
s
: sensor resistance
Connector
R
p
: Pull-up resistor
ECU interface
T
e
m
p
e
r
a
t
u
r
e

(
C
)
1000 900 800 700 400 500 600
Time (Sec)
300 200 100 0
900
800
700
600
500
400
300
Figure 9. Electrical Interface Configuration
5 volts are applied to a pull up resistor in series with the
temperature sensor. The voltage drop is measured
across the temperature sensor. The corresponding
voltage is the result of the following equation. Uo = Up *
Rs/(Rs + Rp). After the voltage is read by the engine
contol unit (ECU), the corresponding temperature can be
derived by calculation or by using a look up table. The
calculation method would involve a second order
polynomial equation.
Figure 7. Thermal cycling with poisons temperature
profile
SYSTEM RESOLUTION
The system resolution and system errors were evaluated
using an error analysis model. This model incorporates
system wiring, connectors, pull up resistor effects,
element accuracy, and A/D converter resolution. This
model demonstrates the effects on signal output from
various system configurations.
The accuracy of the temperature sensor is affected by
the value of the pull up resistor labeled as Rp. This value
needs to be chosen to maximize the resolution of the
signal over the temperature range. Figure 10 shows
sensor output with three pull up resistor values. A 2700
ohm resistor will make the sensor have an output of 0.3
volts at 40C and 1.2 volts at 1000C. This corresponds
to a 0.9 volt output change over the whole temperature
range. The 500 ohm resistor will generate a sensor
output of 1.28 volts at 40C and 3.16 volts at 1000C.
Total output is almost double the amount compared to a
2700 ohm resistor. The 500 ohm resistor provides the
best temperature resolution for this sensor. While a 500
ohm resistor will allow higher currents leading to more
self heating, in moving exhaust flows this affect should
be minimal.
The wiring and connections for the system were also
included in the error analysis model. The effects on the
system are extremely small. At 0C the wiring and
connectors account for 0.02% of the system resistance,
this is equivalent to 0.05C. As the resistance of the
sensor increases, the effects from system wiring further
decrease.
System Suggestions
It is recommended that the pull up resistor of 500 ohms
be used for best overall system resolution with a
tolerance of 0.5% or less, and an A/D converter of 10 bit
resolution.
Pull-up Resistor Value Comparison
0
0.5
1
1.5
2
2.5
3
3.5
0 200 400 600 800 1000
Temperature (C)
S
e
n
s
o
r

O
u
t
p
u
t

(
V
)
2700 O
1000 O
500 O
CONCLUSION
This paper, presented the development of a temperature
sensor suitable for vehicle exhaust applications. The
sensor is based on the RTD technology. Several
analytical methods were used in the development of the
sensor. As the result of this study, several conclusions
can be drawn:
1. The RTD type temperature sensor has the ability
for the broadest range of applications for
measuring vehicle exhaust temperature quickly
and accurately over the life of the vehicle.
Figure 10. Various pull up resistor output
voltage vs. temperature
The A/D converter also impacts system resolution.
Table 3, shows the effects of A/D converter selection in
conjunction with the pull up resistor. The increments
represent the number of slices the sensor signal will be
broken into over the temperature range.
2. The result of the CFD modeling demonstrates
that the temperature and velocity distributions in
the exhaust pipe are similar. The temperature
distributions for most applications are very flat
except in the boundary layer. The thickness of
this layer is about 7 mm for most applications.
For these applications, the developed sensor will
have sufficient insertion depth for accurate
measurement. In the extreme case of the very
large pipe with very low flow, such as 7 m/s in a
127 mm ID pipe, the insertion depth of the
developed sensor is not sufficient to measure
peak temperature; however, the peak
temperature can be calculated based on the flow
condition.
Sensor Increments Increments
Ohms Voltage Range 8 - bit 10 - bit
2700 .3 - 1.2 45 184
1000 .73 - 2.3 81 327
500 1.28 - 3.16 96 388
Table 3. A/D Converter Effects
An 8-bit converter has a minimum voltage step of 19.6
mV for a 0-5 volt signal. A 10-bit has a minimum voltage
step of 4.89 mV. This equates to 20C resolution for 8 bit
and 5C for 10 bit around 800C
3. The response time can be expressed by a
formula, which is dependent on the exhaust gas
properties, sensor geometry, and the material
properties of the sensing element. The
response times are different for heating vs.
cooling, with the response time for heating being
faster.
System Errors
The pull up resistor tolerance has an impact on system
errors. The tolerance chosen for the resistor adds to the
tolerance for the sensor. It is suggested to use a resistor
with a tolerance of 0.5% or less. Table 4, shows the
maximum possible error of sensor temperature indicated
due to resistor tolerance with a 500 ohm resistor.
4. The sensor has been exposed to high
temperature aging with and without chemical
contaminants. The RTD sensor developed has
been demonstrated to show high accuracy
throughout the life of the sensor against the
effects of thermal aging and poisons.
Tolerance Sensor Temp. Signal error from Tolerance
0.5% 0C 1.45C
800C 6.55C
1% 0C 2.86C
800C
13.07C
5. A system model was developed for system error
analysis. It was shown that the error from the
Table 4. Pull up resistor tolerance effects
sensor is only part of the total error of
temperature measurement. The overall
measurement accuracy must include the error
from other component in the system, such as
the pull-up resistor and A/D converter.
REFERENCES
[1] T. V. Johnson, Diesel Emission Control in Review
The Last 12 Months, SAE 2003-01-0039.
[2] R. Desmarais, et al., How to Select and Use the
Right Temperature Sensor, Sensors Magazine, Jan.
2001.
[3] J. LeGare, et al., Temperature Sensors for On-board
Diagnosis of LEV/ULEV Systems, Automotive
Engineering, April, 1995.
[4] D. Mathews, Choosing and Using a Temperature
Sensor, Sensors Magazine, Jan. 2000.
[5] A. Volbrecht, Temperature Measurement: Making
Sense of it All, Sensors Magazine, June 1998.
[6] Holman, J. P., Heat Transfer, McGraw-Hill. 1972.