How Hot Is Hot?
Building, using and calibrating thermocouples
by Philip Stein
I intended to use this article to address a specific measurement problem, but it turned out to be an excellent example of how metrology works instead.
Metrologists start with a simple physical principle and show how it can be exploited as a measuring tool. Once the principle is established, a long and sometimes boring process of fleshing out the details begins: making the measurement repeatable, making it accurate, extending the range over which it can be used and determining the many sources of uncertainty that must be considered and dealt with. While this process may seem long and boring to some, it's important to remember that the most significant work lies in dealing with these details.
As a measurement consultant and laboratory accreditation assessor, I have witnessed many scientifically based and correctly executed measurement practices. I also have found several practices that were poorly understood and produced inaccurate results. The worst examples of the latter instance, in my own experience, are temperature measurements using thermocouples.
The thermocouple is one of the best temperature measuring devices known. It's inexpensive, rugged, cheap, easy to build, low in cost, and can withstand temperature extremes. However, the thermocouple is not a thermometer or a digital voltmeter. It can't be purchased at the corner drug store, and its measurement isn't read on a display. The thermocouple is more like a piece of equipment in a high school physics lab--it's useless without additional equipment and an understanding of basic scientific principles.
Connecting two wires made of dissimilar metals generates a voltage proportional to the temperature difference between the connected end and the unconnected ends of the wires. This is known as the Seebeck effect. The Seebeck effect is the key to measuring with a thermocouple.
Making a thermocouple is easy. Take two pieces of wire made from different metals, copper and iron for example, and weld, solder or twist the end of one wire to one end of the other. The temperature of the twisted end (the measuring junction or hot junction) is measured by reading the voltage across the unconnected end (the cold junction) wires. The temperature thus indicated is the difference between the temperature at the hot junction and that of the cold junction end--the one with the electrical connection.
Although the principle remains the same, many of the details about this arrangement need to be considered before these wires can be used as a measuring instrument. Getting accurate, repeatable measurements from a thermocouple can be quite a chore. Many practitioners ignore these details or are ignorant of them, so incorrect results occur. Here are a few things to watch for:
* What is the temperature at the voltage measuring end? In the past, the cold junction was immersed in an ice bath, and the temperature at the hot junction was thereby referenced to zero degrees Celsius. Today, an electronic compensator measures the room temperature (usually with a resistance-temperature device, or RTD) and simulates the cold junction voltage for the ice point.
* The Seebeck voltage is not linear with temperature. Organizations such as the National Institute of Standards and Technology have published tables of voltage vs. temperature for various combination pairs of thermocouple wire. These tables assume that the cold junction is at the ice point. Usually, the electronic compensator also applies these tables to the measurement, resulting in a direct, linear temperature-reading instrument.
* Consider the effects of ambient temperatures. If you combine associated electronics with a thermocouple to measure at or near room temperature, the thermocouple is not really being used at all. The Seebeck voltage is near zero because the hot and cold junctions are at room temperature, and the RTD in the electronics is being used as the sensor. Therefore, the thermocouple is usually the wrong device to use in these circumstances.
* Note the date of your thermo-couple table. The thermocouple tables change as the International Temperature Scale is revised. The most recent scale is from 1990. The previous scale was established in 1968. Any thermocouple electronics (or computer program) built before 1990 will read an incorrect value for temperature. This error amounts to only a few hundredths of a degree near room temperature, but might be as large as 2 C at temperatures above 1,000 C.
* Use an isothermal terminal block. When two thermocouple wires are connected to a voltmeter or compensator, two more thermocouples are created--each where a thermocouple wire connects to a voltmeter terminal (which is made of a different metal from that of the thermocouple wire). If these terminals are at the same temperature, no voltage is generated and no error occurs.
To make sure this happens, use an isothermal terminal block, one where the terminals are engineered to stay as close to each other in temperature as possible. Thermocouples are often located hundreds or thousands of feet from the cold junction and temperature readout; either the thermocouple's actual wire is run that far or extension wires of the same metal as the thermocouple's wire are added. If the thermocouple tip is distant from its cold junction, bring the thermocouple wires to an isothermal terminal block near the hot junction, and extend the circuit back to the meter using high-quality soft copper wire. Once the circuit has connected through the block, any low-resistance insulated wire will do as an extension as long as both conductors are the same.
Generating a voltage
Because the thermocouple has the look and feel of a measuring probe with a junction prominent at the far end, many believe that the temperature is measured at that point. Not so. What generates the Seebeck voltage is a temperature difference or gradient.
Imagine a uniform crucible of molten metal, with the tip of the probe buried in the center of the melt. Most of the thermocouple voltage is generated where the wires enter the metal from the cooler, room temperature air. The total voltage generated is the same as it would be if the measuring were done at the tip of the probe.
If the probe was reading incorrectly, you could attempt a repair by changing the tip and a few inches of wire on each side and not fix the problem. The problem would, in fact, be somewhere farther back along the wire, where the wire enters the melt and the temperature gradient is largest.
Thermal gradients along a wire are not the only factors that will generate a voltage. Gradients in the wire's composition, inclusions, stresses and annealed spots will generate voltage as well.
While the best quality thermocouple wire is uniform in makeup, that will change once the wire is put to use. The hostile atmosphere surrounding the thermocouple corrodes the wire, diffuses in foreign impurities and drives the composition of the wire to drift (decalibration). Frequent bending or kinking also changes the measurement properties.
Because the voltage is very sensitive to the details of the metal, thermocouple wire must be pure and uniform. In fact, you can buy premium wire with a higher-than-normal degree of uniformity.
After you have considered the above factors, it should be clear that the thermocouple is not a precision measuring device. Inexpensive, rugged, easy to build, yes. Precise, no. If someone says he or she is measuring more accurately than ±1 C with a thermocouple, it's time for a closer look. Above 1,000 C, and the error band is more likely to be within ±2 C. Of course this is an oversimplification, but the information provided here should alert you to the need for a more thorough uncertainty analysis.
Calibrating the thermocouple
The easiest way to increase a thermocouple's accuracy is to calibrate it. The actual wire to be used should be calibrated to its expected temperature range and environmental conditions. In this case, a precision of 0.1 degree or better might be expected. As noted earlier, unless in extremely benign conditions, thermocouples will drift with time; therefore, they need to be recalibrated at regular intervals.
Remember to pay attention when having thermocouples calibrated by a service. If you are not specific in your instructions, many commercial services will calibrate your ice point simulator and electronic indicator and won't touch the thermocouple measurement itself. The largest sources of uncertainty in thermocouple measurements come from decalibration of the wire and from the inability of the wire to accurately conduct heat. In such instances, the thermocouple doesn't adequately reflect the heat of the area it's supposed to be measuring. Testing the electronics is usually only a small part of the work.
PHILIP STEIN is a metrology and quality
consultant in private practice in Pennington, NJ. He holds a master's
degree in measurement science from The George Washington University,
in Washington, and is an ASQ Fellow.
For more information, go to www.measurement.com.