Choosing the Right Temperature Transducers For Your Data Acquisition System

Making temperature measurements with a data acquisition system is easy. Making accurate, repeatable temperature measurements, however, is not quite so simple.
Temperature is a deceptively simple measurement quantity. We often think of it as a single number, but it really is a statistical construct whose accuracy and repeatability can be affected by thermal mass, measurement time, electrical noise, and measurement algorithms.

This difficulty was highlighted in 1990 when the committee reviewing the International Practical Temperature Scale adjusted the definition of one reference temperature by nearly one-tenth of a degree Celsius. Imagine discovering that all the current measurements you had been making were off by one-tenth of an amp.
In other words, temperature is difficult to measure accurately in the best of circumstances, and real-life test conditions only make it harder. Understanding the advantages and disadvantages of the various approaches to measuring temperature will help you avoid problems and get better results.
The four most common types of temperature transducers used in data acquisition systems are resistance temperature detectors (RTDs), thermistors, IC sensors, and thermocouples. Choosing the right temperature transducers and using them correctly can mean the difference between misleading results and numbers you can count on.

So Many Transducers

No single transducer is the best for every measurement situation, so you need to know when to use which type. As Table 1 shows, the factors to consider include performance, useful range, cost, and convenience.1
Resistance Temperature Detectors

The RTD operates on the principle that the resistance of all metals is temperature dependent. The choice of platinum in top-quality RTDs offers the most accurate and stable measurements available up to about 500°C. However, platinum makes the RTDs quite costly. RTDs using nickel or nickel alloys are not as stable or as linear as their platinum counterparts but are much less expensive and still provide good accuracy.
On the downside, RTDs are susceptible to self-heating errors. To measure the resistance, you have to apply a current which produces heat that distorts the measurement results.
A second disadvantage is the RTD’s low resistance, which influences the ways you can use it to measure temperature. Because the resistance is so low, the resistance of the leads used to connect the RTD can create large errors.

In the so-called two-wire technique, you measure resistance at the terminals of your data acquisition system, making the lead resistance part of the unknown quantity you are trying to measure (Figure 1a). In contrast, the four-wire technique measures resistance at the terminals of the RTD, thereby removing lead resistance from the measurement (Figure 1b). The penalty is twice as much wiring and twice as many data acquisition channels. The three-wire technique offers a compromise; it is one wire simpler but not quite as accurate.

Thermistors, resistive detectors typically made of ceramic semiconductors, offer much higher impedance than RTDs, so the reduced lead error makes it feasible to use the simpler two-wire technique. Their high output (a large change in resistance with a small change in temperature) yields high-resolution measurements and reduces the impact of lead resistance. Additionally, the thermistor’s very low thermal mass minimizes thermal loading on the device under test.
The low thermal mass also presents a disadvantage—the potential for higher self-heating from the current source used in the measurement. A second disadvantage of the thermistor is its high degree of nonlinearity, requiring a linearization algorithm to reach usable results.
IC Sensors

IC sensors solve the linearity problem and provide high output levels. They also are relatively inexpensive and offer good accuracy at room temperatures.
You do not have as many choices with IC sensors in terms of product configurations or temperature ranges. Plus, IC sensors are active devices and require a power source, making them susceptible to self-heating errors.
IC sensors, by the way, are part of the trend toward “smart sensors.” These are transducers that possess enough on-board intelligence to help with measurement computation and communications that you traditionally had to perform in the data acquisition system.


Thermocouples are popular because they offer a much broader temperature range and more rugged construction than other types. They also do not require any form of power, and their low cost makes them an attractive choice for large data acquisition systems. However, it is important to understand the nature of thermocouples in order to overcome some of their inherent drawbacks and produce quality results.

The behavior of a thermocouple is based on the gradient theory. Figure 2a illustrates this concept. When we heat one end of a wire, we produce a voltage that is a function of the temperature gradient from one end of the wire to the other and the type of metal that composes the wire.
A thermocouple simply is two wires of different metals joined at one end and open at the other (Figure 2b). The voltage across the open end is a function of both the temperature at the junction and the metals used in the two wires. All dissimilar pairs of metals exhibit this voltage, named the Seebeck voltage after its discoverer, Thomas Seebeck. Over small temperature ranges, the Seebeck voltage is linearly proportional to temperature:

V = a Tx

where a , the Seebeck coefficient, is the constant of proportionality. Over larger ranges, however, the Seebeck coefficient is itself a function of temperature, making the Seebeck voltage nonlinear. As a result, thermocouple voltages tend to be nonlinear.

Relative vs Absolute Temperature

RTDs, thermistors, and IC sensors all measure absolute temperatures. The thermocouple, however, only measures relative temperatures.

The reason is obvious when we think about connecting a thermocouple to a voltmeter or data acquisition system. Let’s say we are using the common Type J thermocouple which consists of one wire made of iron and one wire made of constantan, an alloy of 45% nickel and 55% copper.

What happens when we connect to the two test leads which probably are made of copper? We create two more thermocouples, each of which contributes a voltage to the circuit (Figure 3). Now we have three thermocouples and three unknown temperatures.

The classical solution to this dilemma involves adding an opposing thermocouple and a reference junction at a known temperature (Figure 4). In this example, the opposing thermocouple is another copper-iron junction to match the copper-iron junction created when we attached a copper lead to the iron lead of the “real” thermocouple. These two junctions will effectively cancel each other if they are isolated in an isothermal (constant temperature) block.

We now have just two junctions, the original junction from the thermocouple (Tx) and the reference junction (Tref) we have just added. If we know the temperature at the reference junction, we can compute Tx. Many data acquisition systems and voltmeters that offer thermocouple measurements take care of this computation automatically.
Unfortunately, the nature of temperature makes things a bit tricky here because there are so few practical, inexpensive reference points for temperature. The freezing and boiling points of water at (0° and 100°C) are about the only easy ones Mother Nature offers.

A common way to determine the temperature of Tref is to physically put the junction in an ice bath, forcing the temperature to 0°C. In fact, all thermocouple tables are referenced to an ice bath.
Simplifying the Setup

The ice-bath approach delivers accurate readings, but an ice bath is not the most convenient accessory for a data acquisition system. And, we still have to connect two thermocouples.

The first step toward simplification is to remove the ice bath. If we just measure Tref with an absolute temperature device, such as an RTD, and compensate for it mathematically, we do not have to force it to 0°C.

The next step is to eliminate the second thermocouple (Figure 5). By extending the isothermal block to include Tref, we set the temperature of the isothermal block to Tref since the other two thermocouples in the block still cancel each other. Finding Tref now is just a matter of measuring the temperature of the isothermal block with the RTD or another absolute temperature device.

Getting to the Answer

Once we know Tref, we can compute its equivalent voltage and subtract it from the measured voltage V, thus simulating a Tref of 0°C. Now we can compute Tx using ice-bath-referenced thermocouple tables or equations. As mentioned earlier, voltmeters and data acquisition systems that do thermocouple measurements typically perform these computations for us.

Seebeck coefficients and the resulting output voltages are small numbers, making it difficult to accurately measure both absolute levels and relative changes (Table 2). This is where electrical noise can affect the accuracy of your temperature measurements.

Reduce magnetic and electrostatic coupling by using twisted-pair wiring, minimizing the length of your leads, and staying away from strong magnetic and electrical fields. Last, but certainly not least, you need instrumentation capable of clean, low-level measurements.


With proper care given to selecting the right sensor, accurate and reliable temperature measurements are not difficult. Consider transducer cost, temperature range, accuracy, sensor output, and error modes, such as self-heating and thermal settling times before choosing. Also, pay careful attention to the measurement system. The correct sensor is useless if the data acquisition system cannot measure its output accurately and repeatably.


1. “Practical Temperature Measurements,” Application Note 290, Hewlett-Packard Publication Number 5965-7822E, July 1997.

2. NIST Monograph 175, National Institute of Standards and Technology, Washington, D.C., 1993.

About the Author

Barry Scott is a product manager at Hewlett-Packard’s Electronic Measurements Division. He received a B.S.E.E. from Oregon State University and has more than eight years experience supporting and managing data acquisition products. Hewlett-Packard, Test and Measurement Organization, P.O. Box 50637, Palo Alto, CA, 94303-9512, (800) 452-4844.





IC Sensor





· Most stable

· Most accurate

· More linear than thermocouples

· High sensitivity

· Fast

· Two-wire measurement

· Most linear

· Highest output

· Inexpensive

· Self-powered

· Rugged

· Inexpensive

· Wide variety of physical forms

· Wide temperature



· Expensive

· Slow

· Current source required

· Small resistance


· 4-wire


· Self-heating

· Nonlinear

· Limited temperature range

· Fragile

· Current source required

· Self-heating

· Limited to 250°C

· Power supply


· Slow

· Self-heating

· Limited configurations

· Nonlinear

· Low voltage

· Reference required

· Least stable

· Least sensitive

Table 1



Seebeck Coefficient


Output Voltage at




at 0°C


at 100°C





-0.25 µV/°C

0.90 µV/°C

0.033 mV


58.7 µV/°C

67.5 µV/°C

6.32 mV


50.4 µV/°C

54.4 µV/°C

5.27 mV


39.5 µV/°C

41.4 µV/°C

4.10 mV


5.40 µV/°C

7.34 µV/°C

0.65 mV

Table 2

Copyright 1997 Nelson Publishing Inc.

September 1997

Sponsored Recommendations


To join the conversation, and become an exclusive member of Electronic Design, create an account today!