Professional Documents
Culture Documents
of a Temperature
Measurement System
with worked examples
Gerd Scheller
Note
This reference work has been created to the best knowledge and belief. We assume no liability for
possible errors. The definitive source of information is always the operating manual for the relevant
device.
To make measurements comparable, the quality must be stated by specifying the measurement
uncertainty. The ISO/BIPM Guide to the Expression of Uncertainty in Measurement published in
1993, normally abbreviated to GUM, introduced a unified method of determining and specifying
the measurement uncertainty. This method was adopted by the calibration laboratories worldwide.
Its application, however, requires some mathematical knowledge. The simplified approach used in
the following chapters makes the measurement uncertainty clearly understandable for everyone
who uses temperature measurement systems.
Mistakes made when installing temperature sensors and connecting the evaluation electronics
lead to increased measurement error.
In addition, there are the measurement uncertainty components of the sensor and evaluation elec-
tronics themselves. After an explanation of the measurement uncertainty components, a number of
worked examples are set out.
A knowledge of the measurement uncertainty components and their order of magnitude enables
the user to reduce individual components by changing the installation conditions or selecting a dif-
ferent instrument. The crucial factor is always the measurement uncertainty required for a measu-
rement task. Where, for example, a standard specifies limits for the deviation of the temperature
from the nominal value, the measurement uncertainty of the measurement method used should not
exceed 1/3 of the limit.
This publication, particularly the worked examples set out in Chapter 3, represents an aid in assessing
the measurement uncertainty. If any problems occur, we are happy to discuss specific cases with our
customers and provide practical help and support.
Gerd Scheller
Fe
Compensating
T1 T2
cable
Con
Fig. 1: Features of the temperature measurement system with thermocouple
A thermocouple always measures the temperature difference between the measurement point and
the reference junction. The temperature T2 (reference junction terminal temperature at the transiti-
on from thermocouple to copper cable) must be known in order to be able to measure the tempe-
rature T1. A compensating cable which has the same thermoelectric properties as the thermo-
couple itself is used to connect the thermocouple to the evaluation electronics. This prevents an
additional thermoelectric emf arising at the junction. However, it is essential that the appropriate
compensating cable is always used. With enclosure entries with different internal and external tem-
peratures, plugs and sockets free from thermoelectric emf must be used.
All the thermoelectric emfs quoted in the tables are referred to a reference junction temperature of
0C. This cannot normally be realized with measurement systems and so the terminal temperature
must be measured or held constant. If the terminal temperature is 20C, for example, the result
(from the measured thermoelectric emf) must be corrected by +20C.
EN 60 584 specifies error limits for all standard thermocouples dependent on the measured tempe-
rature. It should be noted that error limits also apply to the compensating cables (Class 1 and 2), so
that an overall tolerance is obtained for each thermocouple, made up of the tolerances of the ther-
mocouple and the compensating cable.
Thermocouples normally show a worse drift behavior than platinum resistance thermometers (ab-
sorption of materials from the surroundings, chemical reaction, recrystallization, mechanical
stress), so it is impossible to make a general statement about the service life of thermocouples.
However, the measurement uncertainty should be checked on a regular basis with a reference ther-
mometer, such as a measurement system calibrated by the DKD (German Calibration Service)
using a resistance thermometer connected in a 4-wire circuit), ideally in a liquid-filled thermostatic
bath.
1.1.2 Adjustment
Adjustment is the setting or balancing of a measuring instrument to eliminate systematic measure-
ment deviations to the necessary extent or as far as possible. Adjustment therefore requires an ac-
tion which permanently alters the instrument.
Temperature measuring instruments are normally adjusted using fixed resistors to EN 60 751. The-
refore, no statement about the overall measurement uncertainty of the measurement system can
be derived from this, as the deviation of the temperature probe connected at a later stage is not
normally considered.
1.1.3 Testing
During testing, the extent to which an object under test fulfills a specific requirement (e.g. tolerance
class) is determined. Testing is thus always linked to a comparison under specified conditions. The
nature and scope of the measurement checks made on the test equipment are left to the discretion
of the respective organization or test laboratory. No statements about measurement uncertainty
are made.
In the following section we would like to consider the individual measurement uncertainty compon-
ents:
80 C
6
1/3 DIN Class B
Class A
5
Class B
Class 0,5
Tolerance C
0
-400 -200 0 200 400 600 800 1000
Measured temperature C
Fig. 5: Permissible deviation limits for platinum temperature sensors
1.0
0.9
Step response of
temperature probe
0.5
0
t 0.5 t 0.9
Because of the ever-present thermal resistance (the magnitude is dependent on the materials
used), the temperature sensor never takes on the temperature of the measured medium immedia-
tely - there is always some delay.
This delay is determined by the heat transfer coefficients sensor - fill material - metal protection
tube - measured medium, and is thus a design parameter of the temperature probe.
The step response provides information on the overall response. As the response time depends to
a large extent on the flow rate, the measured medium used and the immersion depth, EN 60 751
lays down conditions for recording the step response in air and water (e.g. surrounded by a flow of
air, flow rate 3.0 0.3m/sec). The data sheet for the temperature probe normally gives figures for
the half-value time t05 (measured value has reached 50% of the final value) and the ninety percent
time t09 (measured value has reached 90% of the final value). For measurements in air, t09 times of
five minutes and more are quite possible here.
The measures already described for reducing the heat conduction error also lead to a reduction in
response times. In addition, the user must ensure that an adequate stabilization time is allowed be-
fore, for example, any measurements are recorded.
2
R
t = I------------ (2)
EK
For a given measuring current, the self-heating error is reduced by using a sensor with a lower no-
minal resistance.
In addition, the self-heating error can be reduced by ensuring a good heat transfer between the
measured medium and the temperature probe (e.g. highest possible flow rate). This means that,
particularly with measurements in gases, account must be taken of the self-heating error.
Any thermoelectric emfs occurring can be determined by taking two measurements (resistance
measurement with a digital multimeter) with the measuring current reversed for the second measu-
rement. The bigger the absolute difference between the two indicated values, the higher is the
thermoelectric emf in the measurement circuit.
Insulation
resistance
Fig. 7: Equivalent circuit for the insulation resistance inside a resistance thermometer
For a Pt 100 resistance thermometer and a measured temperature of 100C, the resistance values
are:
R (100C) = 138.5055 R (101C) = 138.8847
t = 1C and R = 0.3792 CS = 2.637C/
2-wire circuit
The connection between the resistance thermometer and the evaluation electronics is made with a
2-core cable. The evaluation electronics measures the total resistance, made up of the temperature
sensor, internal wiring of the temperature probe (in the protection tube) and the connecting cable,
leading to a systematically higher indication.
The sum of the resistances of the internal wiring and the connecting cable are therefore subtracted
from the measured resistance. However, these resistances vary depending on the manufacturing
process, and are exposed to the operating temperature to some extent, so that further uncer-
tainties occur. The user should take particular care to ensure that the connecting cable is not in
contact with any sources of heat, as this gives rise to an increase in lead resistance. Regular check
measurements show that the fluctuation in lead resistance is approx. 5% of the loop resistance
(each meter of the connecting cable must be regarded as part of the loop). As an example, the fi-
gure quoted in the data sheet for the loop resistance of the connecting cable with a cross-section
of 0.22mm2 is 162m (per meter), so, using the 5% figure, the variation in lead resistance is esti-
mated to be 8m/m.
3-wire circuit
To minimize the effects of the lead resistance and their fluctuations with temperature, a 3-wire cir-
cuit is normally used. For this, an additional lead is run to one contact of the resistance thermome-
ter. This then forms two measuring circuits, one of which is used as a reference. With the 3-wire cir-
cuit, it is possible to compensate for both the value of the lead resistance and its variation with
temperature. An essential requirement is that all three conductors have identical properties (resi-
stance) and are all exposed to the same temperatures.
4-wire circuit
The measuring current is fed to the thermometer via two supply leads. The voltage drop at the resi-
stance sensor is picked up via the two measuring leads. If the input resistance of the connected
electronics is many times higher than the lead resistance, the lead resistance can be neglected. So
for temperature measurement systems using a 4-wire circuit, the measurement uncertainty for the
lead resistance is not applicable. However, it should be noted that the 4-wire circuit does not al-
Measurement
Fig. 9: Example of measured values with 100 repeat measurements
If the frequency of the measured value is plotted against the measured value, a distribution func-
tion is obtained (Fig. 10, arising from the measured values in Fig. 9) in which certain values occur
more frequently than others. The distribution function reflects the probability of the same measure-
ment result being obtained when a measurement is repeated. If the measured values are scattered
purely at random, the distribution function corresponds to a Gaussian normal distribution. It must
be pointed out here that, in theory, the normal distribution only applies for an infinite number of
measured values. However, even with a small number of measured values ( 30) the pattern is al-
ready very close to the normal distribution.
Frequency
Measured value
Fig. 10: Frequency of the individual measured values from Fig. 9
2 2 2 2
= 1 + 2 + 3 + ... + N (3)
When measurement uncertainties are quoted on calibration certificates, they are generally based
on a normal distribution, as the results of the calibration are not derived from a one-off measure-
ment, but from several repeat calibrations. The measurement uncertainties are quoted with a 95%
confidence level, so that this figure must be divided by 2 to calculate the standard measurement
uncertainty (68.3% confidence level).
If only a tolerance interval of a quantity is known and no information on the distribution is available,
a rectangular distribution must be assumed. Examples of this type of distribution are the state-
ments made in data sheets. If the quantity is measured again, the only certainty is that the measu-
red value will once more lie within the tolerance interval and there is also zero probability of measu-
ring a value outside the interval. Within the interval (x0 - a, x0 + a) the probability of occurrence is
constant for each value. So the probability of occurrence for the total interval is 100%.
Probability
Measured values
Fig. 11: Rectangular distribution
It is assumed that the temperature at the measurement point is 400.50C. The transmitter supplies
an output signal of 16.816mA.
Deviation of the thermocouple as per EN 60 584 (MD)
The deviation limit for tolerance class 2 is 0.0075 t; this gives a deviation limit of 3C for a mea-
sured temperature of 400C.
Measurement signal deviation caused by insufficient stabilization (MA)
It is assumed that the measured value has stabilized, so that this uncertainty component can be
neglected.
Measurement signal deviation caused by deviation of the reference temperature (t0S)
The data sheet for the transmitter quotes a figure of 1C for the reference junction accuracy.
Measurement deviation caused by the compensating cables (VLX)
The full deviation limit of 2.5C must be assumed.
Deviation caused by supply variations (V) (effect of supply)
According to the data sheet, a supply in the range of 20.4V to 25.4V can be connected to the
transmitter. The uncertainty component is 0.01% per V deviation from 24V (referred to the max.
output signal of 20mA). If we assume an actual supply voltage of 22V, this gives a percentage error
of 0.02% and thus an uncertainty component of 0.004mA. This corresponds to 0.125C at a given
sensitivity of 0.032mA/C. (The sensitivity is given by the output signal span of 16mA divided by
the range span (500C)).
Deviation caused by ambient temperature (tM) (effect of temperature)
According to the data sheet, the effect of ambient temperature is 0.005% per C deviation from
22C. Let us assume that the ambient temperature is 30C. This gives an uncertainty component
It is assumed that the temperature at the measurement point is 100.50C. The transmitter supplies
an output signal of 12.040mA.
Geometric addition gives us a measurement uncertainty of 0.52C assigned to the measured tem-
perature (this corresponds to 0.042mA with a sensitivity of 0.08mA/C); for K = 2 the measurement
uncertainty is 1.04C (0.083mA).
So the component of the insulation resistance causes an uncertainty in the temperature of 0.3m
2.717C/ = 0.0008C.
Variations in lead resistance (RAL)
This component is omitted as the resistance thermometer uses a 4-wire circuit.
Supply (V)
The instrument is supplied from a stabilized power supply unit. Unacceptable voltage fluctuations
are indicated as error messages, so this component is omitted.
Ambient temperature (tM)
For this, extensive measurements were carried out in a climatic cabinet. With an ambient tempera-
ture between 10C and 30C the uncertainty component is 0.010C.
Linearization and processing error (tW)
According to the data sheet, the accuracy of the instrument is 0.03C.
This gives us a measurement uncertainty of approx. 0.024C assigned to the measured tempera-
ture; for K = 2 the figure is 0.048C.
It should be noted that, in the last example, unlike the worked examples with transmitters, a com-
ponent is included for the long-term drift of the temperature probe.