Skip to main content

Full text of "Temperature The Secret To Thermocouples"

See other formats


The Software is the Instrument 

Application Note 043 

Measuring Temperature with 
Thermocouples - a Tutorial 

David Potter 


One of the most frequently used temperature 
transducers is the thermocouple. Thermocouples 
are very rugged and inexpensive and can operate 
over a wide temperature range. A thermocouple is 
created whenever two dissimilar metals touch and 
the contact point produces a small open-circuit 
voltage as a function of temperature. This 
thermoelectric voltage is known as the Seebeck 
voltage, named after Thomas Seebeck, who 
discovered it in 1821. The voltage is nonlinear with 
respect to temperature. However, for small changes 
in temperature, the voltage is approximately linear, 


where AV is the change in voltage, S is the 
Seebeck coefficient, and AT is the change in 


S varies with changes in temperature, however, 
causing the output voltages of thermocouples to be 
nonlinear over their operating ranges. Several types 
of thermocouples are available; these 
thermocouples are designated by capital letters that 
indicate their composition according to American 
National Standards Institute (ANSI) conventions. 
For example, a J-type thermocouple has one iron 
conductor and one constantan (a copper-nickel 
alloy) conductor. 

You can monitor thermocouples with versatile PC- 
based data acquisition systems. Thermocouples 
have some special signal conditioning 
requirements, which this note describes. The Signal 
Conditioning extensions for Instrumentation (SCXI) 
system, depicted in Figure 1, is a signal 
conditioning front end for plug-in data acquisition 
(DAQ) boards. SCXI systems are ideal for 
amplifying, filtering, and even isolating the very 
low-level voltages that thermocouples generate. 

Figure 1. The SCXI Signal Conditioning Front-End System for Plug-In DAQ Boards 

Product and company names are trademarks or trade names of their respective companies. 

340904 B-01 © Copyright 1996 National Instruments Corporation. All rights reserved. November 1996 

J3 (+) 

To DAQ Board 




Figure 2. J-Type Thermocouple 

Thermocouple Circuits 

General Case 

To measure a thermocouple Seebeck voltage, you 
cannot simply connect the thermocouple to a 
voltmeter or other measurement system, because 
connecting the thermocouple wires to the 
measurement system creates additional thermoelectric 

Consider the circuit illustrated in Figure 2, in which a 
J-type thermocouple is in a candle flame that has a 
temperature you want to measure. The two 
thermocouple wires are connected to the copper leads 
of a DAQ board. Notice that the circuit contains three 
dissimilar metal junctions-Jl, J2, and J3. Jl, the 
thermocouple junction, generates a Seebeck voltage 
proportional to the temperature of the candle flame. 
J2 and J3 each have their own Seebeck coefficient and 

generate their own thermoelectric voltage proportional 
to the temperature at the DAQ terminals. To 
determine the voltage contribution from Jl, you need 
to know the temperatures of junctions J2 and J3 as 
well as the voltage-to-temperature relationships for 
these junctions. You can then subtract the 
contributions of the parasitic thermocouples at J2 and 
J3 from the measured voltage. 


Thermocouples require some form of temperature 
reference to compensate for these unwanted parasitic 
thermocouples. The term cold junction comes from 
the traditional practice of holding this reference 
junction at 0° C in an ice bath. The National Institute 
of Standards and Technology (NIST) thermocouple 
reference tables are created with this setup, illustrated 
in Figure 3. 

Figure 3. Traditional Temperature Measurement 
with Reference Junction Held at 0° C 


In Figure 3, the measured voltage depends on the 
difference in temperatures T j and T„f ; in this case, 
is 0° C. Notice that because the voltmeter lead 
connections are the same temperature, or isothermal 
(this term is described in detail later in this note), the 
voltages generated at these two points are equal and 
opposing. Therefore, the net voltage error added by 
these connections is zero. 

Under these conditions, if the measurement 
temperature is above 0° C, a thermocouple has a 
positive output; if below 0° C, the output is negative. 
When the reference junction and the measurement 
junction are the same temperature, the net voltage is 

Although an ice bath reference is accurate, it is not 
always practical. A more practical approach is to 
measure the temperature of the reference junction with 
a direct-reading temperature sensor and subtract the 
parasitic thermocouple thermoelectric voltage 
contributions. This process is called cold-junction 
compensation . You can simplify computing cold- 
junction compensation by taking advantage of some 
thermocouple characteristics. 

By using the Thermocouple Law of Intermediate 
Metals and making some simple assumptions, you can 
see that the voltage the DAQ board measures in 
Figure 2 depends only on the thermocouple type, the 
thermocouple voltage, and the cold-junction 
temperature. The measured voltage is in fact 
independent of the composition of the measurement 
leads and the cold junctions, J2 and J3. 

According to the Thermocouple Law of Intermediate 
Metals, illustrated in Figure 4, inserting any type of 
wire into a thermocouple circuit has no effect on the 
output as long as both ends of that wire are the same 
temperature, or isothermal. 

Metal A Metal C 

Metal A Metal B Metal C 


Figure 4. Thermocouple Law of Intermediate Metals 

Consider the circuit in Figure 5. This circuit is similar 
to the previouly described circuit in Figure 2, but a 
short length of constantan wire has been inserted just 
before junction J3 and the junctions are assumed to be 
held at identical temperatures. Assuming that 
junctions J3 and J4 are the same temperature, the 
Thermocouple Law of Intermediate Metals indicates 
that the circuit in Figure 5 is electrically equivalent to 
the circuit in Figure 2. Consequently, any result taken 
from the circuit in Figure 5 also applies to the circuit 
illustrated in Figure 2. 

In Figure 5, junctions J2 and J4 are the same type 
(copper-constantan); because both are in the 
isothermal region, J2 and J4 are also the same 
temperature. The junctions occur in opposite 
directions, however, so their total contribution to the 
measured voltage is zero. Junctions Jl and J3 are both 
iron-constantan junctions and also point in opposite 
directions, but may be different temperatures. 
Therefore, junctions Jl and J3 are the only two 
junctions with outputs that have any effect on the total 
voltage measured. 

Using the notation V^T^,) to indicate the voltage 
generated by the junction Jjc at temperature T^,, the 





+V J4 - +V, 3 - 


+v J2 - 







Figure 5. Inserting an Extra Lead in the Isothermal Region 


general thermocouple problem is reduced to the 
following equation: 

V ME as=V j1 (T tc ) +V J3 (T ref ) (2) 

where is the voltage the DAQ board measures, 

Ttc is the temperature of the thermocouple at Jl, and 
T re f is the temperature of the reference junction. 

Notice that in equation 2, V Jjc {Ty) is a voltage 
generated at temperature T^, with respect to some 
reference temperature. As long as both Vj] and V J3 
are functions of temperature relative to the same 
reference temperature, equation 2 is valid. As stated 
earlier, for example, NIST thermocouple reference 
tables are generated with the reference junction held at 

Because junction J3 is the same type as Jl but in the 
opposite direction, V J3 (T ref ) =-Vj|(T ref ). Because Vj, 
is the voltage that the thermocouple type undergoing 
testing generates, this voltage can be renamed V xc . 
Therefore, formula 2 is rewritten as follows: 

Vmeas =V tc (T tc ) -VtcCIW) (3) 

Therefore, by measuring and T^, and knowing 

the voltage-to-temperature relationship of the 
thermocouple, you can determine the temperature of 
the thermocouple. 

There are two techniques for implementing cold- 
junction compensation-hardware compensation and 
software compensation. Both techniques require that 
the temperature at the reference junction be sensed 
with a direct-reading sensor. A direct-reading sensor 
has an output that depends only on the temperature of 
the measurement point. Semiconductor sensors, 
thermistors, or RTDs are commonly used to measure 
the reference-junction temperature. For example, 
several SCXI terminal blocks include thermistors that 
are located near the screw terminals to which 
thermocouple wires are connected. 

Hardware Compensation 

With hardware compensation, a variable voltage 
source is inserted into the circuit to cancel the parasitic 
thermoelectric voltages. The variable voltage source 
generates a compensation voltage according to the 
ambient temperature, and thus adds the correct voltage 
to cancel the unwanted thermoelectric signals. When 
these parasitic signals are canceled, the only signal the 
DAQ system measures is the voltage from the 
thermocouple junction. With hardware compensation, 

the temperature at the DAQ system terminals is 
irrelevant because the parasitic thermocouple voltages 
have been canceled. The major disadvantage of 
hardware compensation is that each thermocouple type 
must have a separate compensation circuit that can add 
the correct compensation voltage, which makes the 
circuit fairly expensive. Also, hardware compensation 
is generally less accurate than software compensation. 

Software Compensation 

Alternatively, you can use software for cold-junction 
compensation. After a direct-reading sensor measures 
the reference-junction temperature, software can add 
the appropriate voltage value to the measured voltage 
to eliminate the parasitic thermocouple effects. Recall 
formula 3, which states that the measured voltage, 
Vmeas> is equal to the difference between the 
thermocouple voltages at the thermocouple 
temperature and at the reference-junction temperature. 

Note: The National Instruments Lab VIEW®, 
Lab Windows®, andNI-DAQf® software 
packages include routines that perform the 
required software compensation steps. 

There are two ways to determine the thermocouple 
temperature when given the measured voltage, V^s, 
and the temperature at the reference junction, T^ . 
The first method is more accurate, but the second 
method requires fewer computational steps. 

Procedure 1-Direct Voltage Addition 
Method for Software Cold-Junction 

The more accurate compensation method uses two 
voltage-to-temperature conversion steps. From 
formula 3, you can find the true open-circuit voltage 
that the thermocouple would produce with a reference 
junction at 0° C, as shown in the following equation: 

V TC CT TC ) = Vmeas + VtcCiW) (4) 

Therefore, this method requires the following steps: 

1. Measure the reference-junction temperature, T ref . 

2. Convert this temperature into an equivalent 
voltage for the thermocouple type undergoing 
testing, V TC (T^ ). You can use either the NIST 
reference tables or polynomials that assume a 
reference junction at 0° C. 

3. Add this equivalent voltage to the measured 
voltage, Vmeas, to obtain the true open-circuit 


voltage that the thermocouple would produce with 
a reference junction at 0° C, V TC (T TC ). 

4. Convert the resulting voltage into a temperature; 
this value is the thermocouple temperature, Tjc- 
This compensation method requires a translation 
of the junction temperature into a thermocouple 
voltage followed by a translation of a new voltage 
into a temperature. Each of these translation steps 
requires either a polynomial calculation or a look- 
up table. It is, however, more accurate than the 
following method. 

Procedure 2-Temperature Addition 
Method for Software Cold- Junction 

A second, easier software compensation approach 
makes use of the fact that thermocouple output 
voltages are approximately linear over small 
deviations in temperature. Therefore, for small 
deviations in temperature, you can use the following 

Vtc(T,)-V tc (T 2 ) -VtcCT, -T 2 ) (5) 

This assumption is true if T] is fairly close to T 2 
because the thermocouple voltage- versus-temperature 
curve is approximately linear for small temperature 
variations. Assuming that the thermocouple 
temperature is relatively close to the reference 
temperature, you can rewrite formula 3 as shown in 
the following equation: 

V M eas = V tc (T tc -I",*) (6) 

temperature of T ref . Therefore, you can convert the 
measured voltage into a temperature using the N1ST 
reference tables. This temperature is the difference 
between temperatures Txc and . This simplified 
compensation method is as follows: 

1. Measure the reference-junction temperature, T ref . 

2. Convert the measured voltage, Vmeas, into a 
temperature using the voltage-to-temperature 
relationship of the thermocouple. This 
temperature is approximately the difference 
between the thermocouple and the cold-junction 
reference, T TC - T^. 

3. Add the temperature of the reference junction, 
T ref , to this value. This is the thermocouple 


This method saves a computation step over the first 
method of software compensation but is less accurate. 
A comparison of the accuracy of these two methods is 
included in the application examples later in this note. 

Linearizing the Data 

Thermocouple output voltages are highly nonlinear. 
The Seebeck coefficient can vary by a factor of three 
or more over the operating temperature range of some 
thermocouples. For this reason, you must either 
approximate the thermocouple voltage-versus- 
temperature curve using polynomials, or use a look-up 
table. The polynomials are in the following form: 

T = ao + a,v + a 2 ^ + ... + a„v" (7) 

Remember that if you are using standard NIST 
thermocouple reference tables or equations, the 
voltage V TC is a function of temperature relative to a 
reference temperature of 0° C. By assuming linearity 
and using equation 6, assume that the voltage-versus- 
temperature curves with a reference temperature of 
0° C are identical to curves with a reference 

where v is the thermocouple voltage in volts, T is the 
temperature in degrees Celsius, and ao through a„ are 
coefficients that are specific to each thermocouple 
type. Table 1 lists NIST polynomial coefficients for 
several popular thermocouple types over a selected 
range of temperature. 


Table 1. NIST Polynomial Coefficients for Voltage-to-Temperature Conversion (T = ao + ai v + a 2 v 2 + ... + ^V) 

Thermocouple Type 







0° tn i noo °f 

0° to 7fil) °C 

0° to 500 °C 

-50° to 250 °C 

-50° to 250 °C 

0° to 400 °C 










1.978425E -2 





a 2 

-2.3301759E -7 


7.860106E -8 

-9.3835290E -5 


-7 .60296 IE -7 

a 3 

6.5435585E -12 


-2.50313 IE -10 

L3068619E -7 

1.02237430E -7 

4.637791E -11 


-2.549687E -16 


-2.2703580E -10 


-2.165394E -15 




-1.228034E -17 


1.88821343E -13 

6.048 144E -20 



-5.344285E -26 

9.804036E -22 

-3.8953900E -16 

-1.5908594 IE -16 

-7.293422E -25 

a 7 



-4.413030E -26 

2.8239471E -19 

8.23027880E -20 






a 9 



3.135361 IE -26 

2.79786260E -27 


-3.3187769E -30 


±0.02° C 

±0.05" C 

±0.05° C 

±0.02° C 

±0.02° C 

±0.03° C 

The errors listed in Table 1 apply only to the 
polynomial calculation, and do not take into account 
errors introduced by the measurement system or 
thermocouple itself. 

Recall that Procedure 1 for software cold-junction 
compensation also requires a temperature-to-voltage 
step to convert the cold-junction temperature into the 
equivalent voltage for the specific thermocouple type. 
Again, you can use either thermocouple reference 
tables, or approximate the curve with a polynomial. 
NIST also specifies a set of polynomials for the 
thermocouple voltage as a function of temperature, 

v = c () +c,T + c 2 T 2 + ... + c n T n (8) 

where c through c„ are coefficents that are specific to 
each thermocouple type. Table 2 lists NIST 
polynomials for temperature-to-voltage conversion for 
several popular thermocouple types. 

NIST also specifies sets of polynomials that cover 
temperature ranges other than those lited in Tables 1 
and 2. More examples of polynomials, including 
wider temperature ranges, are listed in the National 
Bureau of Standards Monograph 175 (see References). 

To speed computation time, a polynomial can be 
computed in nested form. Consider the fourth order 

T = a + a ] v + a 2 v 2 + a 3 v 3 -^V* (9) 

If this polynomial is evaluated as written, several 
unnecessary multiplications will be performed to raise 
v to the various powers. If the polynomial is written 
and evaluated instead as shown in the following 
formula, no powers are calculated and computation 
executes much faster: 

T= ao + v(ai + v(a 2 + v(a 3 H-va*))) (10) 


These NIST polynomials are implemented in functions 
included with LabVIEW, LabWindows/CVI, and 
NI-DAQ software from National Instruments. 

Note: As a precaution, check the units specified 
for the voltages. For the formulas in 
Tables 1 and 2, the voltages are in 
microvolts. For some other thermocouple 
tables and linearization polynomials, the 
voltages may be in millivolts or volts. Using 
the incorrect unit yields erroneous results. 

Measurements with a DAQ 

Thermocouples are commonly used to monitor 
temperature with PC -based DAQ systems. For 
example, the National Instruments SCXI product line 
is especially well-suited for thermocouple 

SCXI is a front-end instrumentation system for plug-in 
DAQ boards. A shielded, rugged chassis houses 
signal conditioning modules that amplify, filter, 
isolate, and multiplex analog signals from 
thermocouples or other transducers. The amplified 
analog signal is then cabled to the plug-in digitizing 
DAQ board, or passed to a local SCXI digitizing DAQ 
module. Figure 6 shows some of the hardware in the 
SCXI product line. 

For some applications, you may be able simply to use 
Table 2. NIST Polynomial Coefficients for Temperature-to- Voltage Conversion (v = c + c, T + CjT 2 + ... + CnT") 

Thermocouple Type 








0° to 1,000 °C 

■210° to 760 °C 

0° to 1372°C 

-50° to 1,064 °C 

-50° to 1,064 °C 

















4.503227558E - 

3.047583693E -2 




3.32922279E -2 


2.890840721E - 



-2.388556930E -5 

-2.324779687E -5 

2.06182434E -4 


-3.30568967E -7 


3.18409457E -7 

3.5691600106E -8 


-2.18822568E -6 


6.50244033E - 


-5.607284E -10 





-1.9197496E -13 

2.09480907E -13 


5.007774410E -14 

2.557442518E -14 

-3.0815759E -11 




-3.202072E -16 

-3.73105886E -17 


4.54791353E - 

c 8 

2.14892176E - 

1.56317257E -20 



2.714431761E -21 

-2.7512902E -17 



-1.210472E -23 



3.59608995E - 


NOTE A: The equation for type K is v = c + c, T + CjT 2 + ... + c 9!" +118.5976e ( 1183432E-4XT- 126.9686)' 


Figure 6. SCXI Signal Conditioning and Data 
Acquisition System 

a plug-in DAQ board with termination accessories 
equipped with temperature sensors for cold-junction 
compensation, such as the AMUX-64T multiplexer 
board and the SC-207X Series termination panels. 

Thermocouple Measurements 
with SCXI 

The SCXI product line includes a variety of analog 
and digital signal conditioning modules for various 
types of signals, including thermocouples, RTDs, 
thermistors, strain gauges, voltage, and current 
sources. The SCXI-1 102, SCXI-1 120, and 
SCXI-1 122 modules in particular are well suited for 
use with thermocouples. The SCXI-1 120, SCXI- 
1121, and SCXI- 11 22 provide isolated measurements, 
which prevents problems caused by ground loops, 
high common-mode voltages, and high voltages in 

Signal Conditioning for 

The following sections describe some general 
measurement system capabilities that can be important 
when measuring thermocouples. 

Low-Noise System 

Low-level thermocouple signals are very susceptible 
to noise corruption. Therefore, it is very important 
that the signal conditioning and DAQ hardware are 
well shielded with very low-noise performance. For 
example, the SCXI chassis and modules are fully 
shielded. The backplane analog bus, the SCXIbus, is 
optimized for low-noise performance and avoids noise 
corruption caused by high-speed digital signals. In 
addition, the cable connecting the SCXI system to the 
plug-in DAQ board uses shielded, twisted-pair cable 
for the best possible noise performance. 

You can also improve the noise performance of your 
system significantly by amplifying the low-level 
thermocouple voltages at as short a distance as 
possible from the actual thermocouple. Thermocouple 
wire can act like an antenna and pick up unwanted 
noise from the environment. Because thermocouple 
voltages are extremely small, typical electrical noise 
levels can seriously corrupt the measurement. 

SCXI modules, however, amplify the thermocouple 
voltage near the thermocouple before the signal is 
passed over the cabling through a potentially noisy 
environment, and into the noisy computer chassis. 
Because the thermocouple signal has been amplified 
to a higher level, the noise that is picked up in the 
wiring and in the computer has much less of an effect 
and causes a relatively smaller measurement error. 

The SCXI- 11 02 is a nonisolated thermocouple 
amplifier module. Table 3 summarizes the capabilites 
of these input modules. 

Table 3. SCXI Signal Conditioning Modules for Thermocouples 




Number of inputs 


8 (SCXI-1 120) 
4 (SCXI-1 121) 


Amplifier gains 

lor 100 

- software-programmable 

- selectable per channel 

1 to 2,000 

- jumper-selectable 

- selectable per channel 

0.1 to 2,000 

- software-programmable 

- selectable per module 

Filtering options 


4 Hz or 10 kHz 

4 Hz 1 or 10 kHz 



250 Vrms 

450 Vrms 

Recommended terminal 
block for thermocouples 

SCXI-1303 or 

SCXI-1328 or 


' This filter setting requires slower acquisition rate; check product data sheet for full details. 


High Amplification 

Because thermocouple output voltage levels are very 
low, you should use as large a gain as possible for the 
best resolution and noise performance. Amplification, 
together with the input range of your analog-to-digital 
converter (ADC), determines the usable input range of 
your system. Therefore, you should carefully select 
your amplification so that the thermocouple signal 
does not exceed this range at elevated temperatures. 

Table 4 lists the voltage ranges from several standard 
thermocouple types; you can use this table as a guide 
for determining the best gain and input range settings 
to use. 

For example, a J-type thermocouple outputs -8 mV at - 
2 10° C and 33 mV at 600° C. If you are using a J-type 
thermocouple over this temperature range with SCXI 
and an AT-MIO-16E-2 board, set your SCXI module 
gain to 100. In this configuration, the effective input 
range of the system is ±5 V/100 = ±50 mV. Because 
the AT-MIO-16E-2 uses a 12-bit ADC with 2 12 binary 
levels, the resolution of this system with a total gain of 
100 is 10 V/(100 x 4,096) = 24 U.V. If you are using 
the thermocouple over a more limited temperature 
range, you can use a higher gain to obtain higher 

Input Filtering 

To further reduce noise, the SCXI-1 102, SCXI-1 120, 
and SCXI-1 121 include a lowpass resistor-capacitor 
(RC) filter for each input channel. The SCXI-1 102 
includes a fixed 1 Hz filter on each channel. The 
SCXI-1 120 and SCXI-1 121 filters are 
jumper-selectable for 4 Hz or 10 kHz. These filters 
are useful for removing the 60 Hz power line noise 
that is prevalent in most laboratory and plant settings. 
The SCXI-1 122 module includes a single lowpass 
filter that can be enabled or disabled. Because this 
filter is applied to the output of the module, after the 
multiplexer, the filter should be used only for very low 
scanning rate applications. 

Broken Thermocouple Detection 

It is useful to be able to detect a break, or open circuit, 
in the thermocouple circuit. You can use the 
characteristics of SCXI amplifiers to detect broken 
thermocouples. A differential amplifier saturates to 
either full-scale positive or full-scale negative output 
if the input leads are open circuit. Therefore, when a 
thermocouple is broken, the SCXI amplifier saturates 
and produces a full-scale output in either direction. 
You can check for this saturation condition with 
software and take the necessary actions. 

Temperature Sensor for 
Cold-Junction Compensation 

As discussed earlier, thermocouple measurements 
require sensing of the cold-junction, or reference, 
temperature at the point where the thermocouple wire 
is connected to the measurement system. Therefore, 
signal connection accessories should include an 
accurate cold-junction sensor, and should be designed 
to minimize any temperature gradients between the 
cold-junction sensor and thermocouple wire 

Shielded terminal blocks connect thermocouples to the 
SCXI signal conditioning modules. Table 3 lists 
which terminal blocks provide the most accurate cold- 
junction sensing. These terminal blocks use a high- 
precision thermistor to measure the cold-junction 
temperature. These terminal blocks also use an 
isothermal design to minimize temperature gradients 
across the screw terminals. For example, the 
SCXI-1328 uses screw terminals embedded in an 
aluminum isothermal plate that helps maintain all the 
screw terminals and the temperature sensor at the 
same temperature. With these terminal blocks, you 
can measure the reference temperature with 0.5° C 
accuracy. Other general purpose terminal blocks, as 
well as the AMUX-64T multiplexer board and SC- 
207X Series termination boards, use the National 
Semiconductor LM-35CAZ temperature sensor. The 
LM-35CAZ temperature sensor produces a linear 
voltage output of 10 mV/°C. 

Table 4. Thermocouple Voltage Output Extremes (mV) 




Voltage Range 

Seebeck Coefficient 




Range (°C) 






-270° to 1,000° 

-9.835 to 76.358 

58.70 at 0° C 




-210° to 1,200° 

-8.096 to 69.536 

50.37 at 0° C 




-270° to 1,372° 

-6.548 to 54.874 

39.48 at 0° C 




-270° to 400° 

-6.258 to 20.869 

38.74 at 0° C 





-50° to 1,768° 

-0.236 to 18.698 

10.19 at 600° C 





-50° to 1,768° 

-0.226 to 21.108 

11.35 at 600° C 


High-Performance Scanning 

Although thermocouples are generally used for low- 
bandwidth applications, you may need to use high 
scanning rates to minimize sampling delays between 
channels and for applications involving large numbers 
of channels. For example, a 10 Hz scanner monitoring 
300 thermocouples can sample each thermocouple 
only once every 30 s. An SCXI system with the AT- 
MIO-16E-2 DAQ board, on the other hand, can scan 
input channels at up to 3 |is per channel. Therefore, an 
SCXI system can sample each of the 300 
thermocouples every 0.9 ms. Also, higher scanning 
rates minimize the sampling delay from channel to 

Differential Measurements 

If you are using any thermocouples that are grounded, 
a differential measurement is important. With 
differential measurements, each channel uses two 
signal leads; only the voltage difference between the 
leads is measured. The differential amplifier rejects 
ground-loop noise and common-mode noise, which 
therefore do not corrupt the measurement. With 
single-ended measurements, on the other hand, the 
negative leads of all the input signals are connected to 
a common ground. SCXI analog input modules use 
differential input measurements. 
If you are connecting thermocouples directly to your 
DAQ board and are not using SCXI, use the 
differential input mode of the board, if possible. You 
can configure the MIO Series boards, Lab Series 
boards, and DAQPad units for differential inputs. 

Thermocouple Examples 

The following section includes some examples to 
assist you in making your own thermocouple 

Example 1-Monitoring 64 
Thermocouples with SCXI-1102s 

For this example, assume that you want to use 
64 K -type thermocouples to measure the temperature 
of a high-pressure boiler system with a known 
maximum temperature of 250° C. Assuming that 
electrical isolation is not important for this application, 
you can use two SCXI -1102 modules to monitor the 
64 thermocouples. Figure 7 diagrams the complete 
DAQ system. 

The AT-MIO-16E-2 board is used to acquire the 
readings because the board has a higher scan rate, self- 
calibration, and high-performance direct-memory 
access (DMA) capabilities. The two SCXI-1102 
modules are housed in an SCXI -1000 four-slot chassis 
and connected to the AT-MIO-16E-2 board with one 
SCXI-1349 shielded cable assembly. This SCXI-1349 
cable assembly is available in lengths up to 10 m. 

An IBM PC/XT/AT or compatible computer running 
Windows controls the temperature monitoring system. 
Therefore, application software choices for controlling 
the system include Lab VIEW, LabWindows/CVI, or 
ComponentWorks. Alternatively, you can use a 
standard programming language in Windows, such as 
C, Pascal, or Visual Basic, and control the DAQ 
hardware with the NI-DAQ driver software that is 
included with National Instruments plug-in DAQ 

You connect the thermocouples to the SCXI-1102 
modules with SCXI- 1303 or TBX-1303 terminal 
blocks. Each terminal block includes an accurate 
sensor for cold-junction compensation. In addition, 
you can configure the SCXI- 1303 and TBX-1303 

64 2SCXI-1303 2 SCXI-1102 SCXI-1000 

Thermocouples Terminal Blocks Multiplexer Amplifier Modules Chassis 

Cable Assembly 
(1-10 m) 

DAQ Board 

Figure 7. SCXI System for Monitoring 64 Thermocouples 


terminal blocks with a high-impedance ground 
reference on each channel. In this configuration, you 
can connect either floating or ground-referenced 

A thermocouple reference table shows that the output 
voltage for temperatures less than 250° C never 
exceeds 10 mV (recall that the voltage measured is a 
function of the reference temperature as well as the 
temperature being measured). Therefore, a ±5 V input 
range with a total system gain of 500 (±10 mV 
maximum signal) accommodates the maximum 
expected signal from the thermocouple. With this 
gain setting, the resolution is 
10 V/(500 x 4,096) = 4.9 U.V, or 0.12° at 250° C. 

Both the AT-MIO-16E-2 board and SCXI-1102 
module include amplifiers that you configure via 
software. Because using the largest gain possible near 
the signal source yields the best noise performance, 
configure the SCXI-1 102 module for a gain of 100 
and the AT-MIO-16E-2 board for a gain of 5. With 
Lab VIEW, you can simply indicate the high and low 
input limits (±10 mV in this example), and the driver 
software calculates the necessary gain for both the 
plug-in board and the SCXI module. 

The following sections detail two software 
compensation procedures for this example. The first 
method is more accurate, but the second method is 
faster. Consult your software reference manual to 
determine the actual functions you need. For example, 
Lab VIEW, LabWindows, ComponentWorks, and 
NI-DAQ implement thermocouple conversion routines 
that convert measured voltage values into 
temperatures for different thermocouple types and 
perform cold-junction compensation. 

Procedure 1-Direct Voltage 
Addition Method of Software 
Compensation for Example 1 

1. Read the voltage from the temperature sensors 
(MTEMP) located on each SCXI-1303 terminal 
block. The thermistor used for cold-junction 
sensing in the SCXI-1303 will output a voltage 
ranging from 0.58 V to 1.91 V. Therefore, the 
gain on the AT-MIO-16E-2 should be either 1 or 
2 (2 for the best resolution). After measuring the 
voltage output of the thermistor, use the 
thermistor conversion routine provided with 
Lab VIEW and NI-DAQ to convert this voltage to 
temperature. For example, an MTEMP thermistor 
reading of 1.25 V from the SCXI-1303 
corresponds to a cold-junction temperature of 
25° C. 

2. Translate this cold-junction temperature reading 
into the corresponding voltage for a K-type 
thermocouple at that temperature using either a 
look-up table or an NIST polynomial. Notice that 
the polynomials required here are the inverses of 
those given in Table 1. Using a look-up table, or 
temperature to voltage conversion routine (such 
as the one provided with Lab VIEW), your reading 
of 25° C translates to a K-type thermocouple 
voltage of 1.000 mV. 

3. Read the voltages on the thermocouple channels. 
For the example given, assume that you get a 
reading of 2.930 mV on Channel 1. 

4. Add the voltage from step 2 to the voltage 
measured in step 3. You then have 

1.000 + 2.930 = 3.930 mV 

5. Translate the result into a temperature using either 
a look-up table or a polynomial such as one from 
Table 1. For the reading from the K-type 
thermocouple of 3.930 mV, for example, you can 
calculate that the temperature is 96.15° C. 

Procedure 2-Temperature Addition 
Method of Software Compensation 
for Example 1 

1. Read the voltage from the temperature sensors 
(MTEMP) located on each SCXI-1303 terminal 
block. The thermistor used for cold-junction 
sensing in the SCXI-1303 will output a voltage 
ranging from 0.58 V to 1.91 V. Therefore, the 
gain on the AT-MIO-16E-2 should be either 1 or 
2 (2 for the best resolution). For example, an 
MTEMP thermistor reading of 1.25 V from the 
SCXI-1303 corresponds to a cold-junction 
temperature of 25° C. 

2. Read the voltages on any thermocouple channels. 
For the example given, the gain is at 500 for 
Channel 1. 

3. Translate the reading into a temperature using 
either a look-up table or a polynomial such as one 
from Table 1; you get a reading of 2.930 mV. By 
applying formula 3 and the coefficients from 
Table 1, you can calculate that the temperature is 
71.6° C. 

4. Add the cold-junction temperature from step 1 to 
the temperature obtained in step 3. The result is 
the temperature at the measuring end of the 
thermocouple. For the example given, the 


temperature of the boiler system is 

71.6° + 25° = 96.6° C. In this case, the error 

introduced by using this faster method is 0.45° C. 


Procedure 1 is more accurate but requires two 
temperature-to- voltage conversions. Procedure 2 is 
faster but introduces an error of 0.45° C. To 
determine if the error from procedure 2 is acceptable 
in your application, you can work through some 
examples both ways. Use a thermocouple reference 
table and consider voltages and temperatures close to 
those in your application. 

Example 2-Temperature Control 
with SCXI-1120 

Consider another application that uses a Macintosh 
computer with SCXI to control the temperature of 16 
environmental chambers in a testing facility. A J-type 
thermocouple in each chamber monitors the 
temperature, which never exceeds 200° C. Each 
chamber includes a heater that turns on and off to 
maintain the temperature at the desired setpoint. 

Another requirement for this application is that the 
controlled systems be electrically isolated from the 
computer that is monitoring and controlling the 
station. This protects the computer and DAQ 
hardware from large voltage spikes and prevents the 
ground-loop voltages that can occur. 

Figure 8 shows the complete system for this 
application. Because isolation is required, you should 
use the SCXI- 1 120 isolated amplifier module. This 
module includes eight isolation amplifier channels, 
which can operate with a common-mode voltage of 
240 Vrms between channel and earth or between 
channels. The 16 thermocouples are connected to the 
two SCXI-1120 modules using two SCXI-1320 

terminal blocks. This terminal block includes a 
temperature sensor for cold-junction compensation. 

For more accuracy, you can substitute the SCXI- 1328 
terminal block, which includes a high-precision 
thermistor and an isothermal aluminum plate to 
minimize temperature gradients. 

To control the heaters, you can use the SCXI-1 160 
single-pole double-throw (SPDT) relay module. This 
module includes 16 independent SPDT latching relays 
that can switch up to 2 A at 250 VAC or 30 VDC. 
You can use this module with an SCXI- 1324 terminal 
block to control the on-off state of the 16 heaters. 

In this example, an NB-MIO-16L-9 multifunction 
plug-in DAQ board controls the two SCXI-1 120 
isolation amplifier modules and the SCXI-1 160 relay 
module. The modules are housed in an SCXI- 1000 
chassis and cabled to the NB-MIO-16L-9 board with 
an SCXI- 1345 shielded cable assembly. 

You can control the DAQ hardware of this example 
using Lab VIEW on the Macintosh. Alternatively, you 
can use traditional programming languages with the 
NI-DAQ driver software included with National 
Instruments plug-in DAQ boards for the Macintosh. 

Again, you must first determine the configuration of 
the SCXI-1120 modules and NB-MIO-16L-9 board 
that yields the maximum resolution possible. Because 
the chamber temperature will never exceed 200°, 
configure the jumpers of the SCXI-1 120 modules for a 
gain of 500. This yields an input range of ±10 mV 
forthe system. Next, configure the NB-MIO-16L-9 
for an input range of ±5 V and a gain of 1. 

The steps for measuring and compensating the 
thermocouple readings are identical to those of 
example 1. 



16 Thermocouples 2SCXI-1328 
Terminal Blocks 


2 SCXI-1120 
Isolation Amplifier 

16 Heater 

SCXI -1324 
Terminal Block 

SCXI-1 160 
Relay Module 

SCXI-1 000 

SCXI-1 345 
Cable Assembly 
(1-10 m) 

DAQ Board 

Figure 8. SCXI Temperature Control System 


Sources of Error 

When making thermocouple measurements, the 
possible sources of error include compensation, 
linearization, measurement, thermocouple wire, and 
experimental errors. 

Cold-junction compensation errors can arise from 
two sources-inaccuracy of the temperature sensor 
and temperature differences between the sensor and 
the screw terminals. The LM-35CAZ sensor on the 
SCXI-1300, SCXI-1320, and SCXI-1321 terminal 
blocks has an effective accuracy of 0.9° C. In 
addition, temperature gradients between the sensor 
and the screw terminals can be as high as 0.5° C, for 
a total accuracy of 1.4° C. The SCXI-1303, 
SCXI-1322, SCXI-1328, TBX-1303, and TBX-1328 
isothermal terminal blocks, however, limit 
temperature gradients and use a high-precision 
thermistor for better accuracy. 

Table 5 lists the cold-junction sensing accuracy for 
the various terminal blocks. 

Linearization error occurs because polynomials are 
approximations of the true thermocouple output The 
linearization error depends on the degree of the 
polynomial used. Table 1 lists the linearization errors 
for the NIST polynomials. 

Table 5. Cold- Junction Sensing Accuracy 

Terminal Block 

CJC Accuracy 
(15° to 35° C) 

SCXI-1303, TBX-1303 

0.65° C 

SCXI-1328, TBX-1328 

0.50° C 


0.8° C 

SCXI-1300, SCXI-1320, 
SCXI-1321, SCXI-1327 

1.3° C 

Measurement error is the result of inaccuracies in the 
DAQ board and signal conditioning modules. These 
include offset error, gain error, nonlinearities, and 
resolution of the ADC. Although SCXI modules are 
calibrated to rmnirnize offset errors, you can 
completely remove the offset error of any module by 
grounding the input, taking a reading, and subtracting 
this offset error from subsequent readings. 

Both the plug-in board and the SCXI module can also 
introduce gain error. For example, a calibrated 
AT-MIO-16XE-50 has a gain error of less than 
0.01%, and the SCXI-1102 introduces a typical gain 
error of 0.02% at gain =100. If the input range is 
±5 V and the total system gain is 500, gain error 
contributes a maximum of (0.02% + 0.01%) x 20 
mV, or 6 U.V of error. However, you can use an 
external calibration source to completely remove this 
gain error. 

Other sources of measurement error include the 
nonlinearity and resolution of the digitizing DAQ 
board or module. However, these errors are typically 
minimized by the high gain of the SCXI module. For 
example, the specified relative accuracy of the 
AT-MIO-16E-2, which includes resolution and 
nonlinearity, is ±0.5 least significant bits (LSB). 
Used with an SCXI module that provides a system 
gain of 500, however, this error is equivalent to only 
0.5(20 mV/4,096) = 2.4 jj.V. With a J-type 
thermocouple, for example, this is equivalent to 
0.05° C at room temperature. 

Often, the major source of error is the thermocouple 
itself. Thermocouple wire error, for example, is 
caused by inhomogeneities in the thermocouple 
manufacturing process. These errors vary widely 
depending on the thermocouple type and even the 
gauge of wire used, but a value of ±2° C is typical. 
Check with the thermocouple manufacturer for exact 
accuracy specifications. 

Another potential source of error is noise picked up 
by the thermocouple leads. Software averaging of 
the acquired data can remove noise from your 

Table 6 lists the expected measurement error and 
resolution in degrees Celsius, subject to the following 

• A calibrated SCXI-1 102 , SCXI-1 122, or 
SCXI- 1120 module, configured for gains of 100 
and 500. 

• A calibrated AT-MIO- 16XE-50 configured for a 
gain of 1 and an input range of ±10 V 

The errors listed in Table 6 include the measurement 
error of a calibrated AT-MIO -16XE-50 and SCXI 
module, and the cold-junction sensor error of the 
SCXI-1303, SCXI-1328, or SCXI-1322 terminal 
block. The linearization errors of the NIST 
polynomials and thermocouple wire error is 
neglected because of dependence on several factors 
as previously listed. 


Table 6. Thermocouple Measurement Accuracies with SCXI- 1 102, SCXI- 1 120, or SCXI- 1 122 



Thermocouple Type 







Temp. Range 2 

Accuracy 3 
SCXI- 1102 

-210° to 366° C 
0.05° C 

-270° to 485° C 
0.06° C 

-199° to 307° C 
0.04° C 

-270° to 406° C 
0.06° C 

0.02%+ 0.72 °C 
0.02%+ 0.81 °C 
0.15% +0.55 °C 

0.02% +0.76 °C 
0.02% + 0.82 °C 
0.15% + 0.58 °C 

0.02%+ 0.70 °C 
0.02% + 0.81 °C 
0.15% + 0.54 °C 

0.02% + 0.76 °C 
0.02% + 0.82 °C 
0.15% + 0.59 °C 


Temp. Range 2 


Accuracy 3 
SCXI- 1102 

-2 10° to 1,200° C 
0.24° C 

0.02% + 0.75 °C 
0.02% + 0.88 °C 
0.15% + 1.1 °C 

-270° to 1,254° C 
0.31° C 

0.02% +0.81 °C 
0.02% + 0.92°C 
0.15% + 01.4°C 

-270° to 1,000° C 
0.21° C 

0.02% + 0.73 °C 
0.02% + 0.86 °C 
0.15% + 1.0 °C 


1 With SCXI-1102, gain = 500 is achieved with gain of 100 on SCXI-1102 and gain of 5 on AT-MIO-16XE-50 

'Assumes the cold junction temperature is 25° C. 

Includes the measurement error of the SCXI module, terminal block (0.65 ° max for SCXI-1303, 0.80 °C max for SCXI-1322 
and 0.5° C for SCXI-1328). Does not include polynomial linearization error or thermocouple inaccuracy. 

The errors listed in Table 6 are of the format: 

H x% of reading + 

For example, an SCXI- 1 122 configured for a gain of 
500 and used with an AT-MIO-16XE-50 has an 
accuracy of ±(0.02% of reading + 0.81° C). 
Therefore, at a temperature reading of 70° C, the 
accuracy is ±(0.02% of 70° C + 0.81° C) = ±0.82° C. 
This accuracy includes the potential error of the 
cold-junction compensation reading. 


Thermocouples are inexpensive temperature sensing 
devices that are widely used with PC -based DAQ 
systems. Thermocouple measurement requires signal 
conditioning, including cold-junction compensation, 
amplification, and linearization. The National 
Instruments SCXI signal conditioning system is a 
low-noise front end for PC -based DAQ systems when 
you use thermocouples to measure temperature. 


G. W. Burns, M. G. Scroger, G. F. Strouse, et al. 
Temperature-Electromotive Force Reference 
Functions and Tables for the Letter-Designated 
Thermocouple Types Based on the 1PTS-90 NIST 
Monograph 175. Washington, D.C.: U.S. 
Department of Commerce, 1993.