01 September 2003

RTDs vs. thermocouples: Measuring industrial temperatures

By H.M. Hashemian

Resistance temperature detectors (RTDs) and thermocouples each have their own distinct place in industrial temperature measurements. Thermocouples will almost always make air or gas temperature measurements because of the self-heating error inherent in temperature measurement with RTDs. As of 2002, thermocouples measured 50% to 60% of all industrial temperature, RTDs 30% to 40%, and thermistors and optical pyrometers measured lower and higher temperatures respectively. The share of RTDs versus thermocouples has been growing steadily over the past three decades as people perceive RTDs as better than thermocouples in most applications.

In applications where high accuracy is the main concern, RTDs are almost always a better choice than thermocouples if temperature is in the range of the RTD operation. RTDs can be calibrated to yield accuracies of as good as a few tenths of a degree, while thermocouples cannot be trusted to produce accuracies of better than a degree, especially at high temperatures.

Researchers have made much progress in developing new process instrumentation systems over the past three decades. This includes the advent of smart temperature sensors and digital electronics to condition the sensor signals and provide digital read out, computer control, and the like. Still, thermocouples and RTDs reign supreme in the conventional industrial temperature arena.

In theory, you can use RTDs for measuring temperatures up to about 1000°C, but in practice, it is difficult to get accurate measurements if the temperature is greater than 400°C. Similarly, you can use thermocouples up to about 3000°C or more, but accurate measurements beyond 1000°C are a challenge. Fortunately, a majority of industrial temperature measurements often fall between 200°C to 400°C-where RTDs and thermocouples yield great performance.

The main problem these two sensors have with high-temperature measurements is limited material properties to construct sensors. Most material can degrade or change at high temperatures and cause the sensor to produce erroneous readings. The insulation material in industrial temperature sensors cannot normally tolerate temperatures near 1000°C for any significant period of time.

RTDs can consist of platinum, copper, nickel, and other wires that have a large temperature coefficient of resistance. Nickel has the best sensitivity but is the least linear, and copper has good linearity but a small temperature range. Today, almost all industrial RTDs use platinum wire. In the past forty years, platinum cost more than copper or nickel. Today it does not cost much more than an overall temperature measurement channel in an industrial process. The same argument applies to thermocouples.


You should use a small electric current (about 1 mA) to measure the resistance of an RTD. This current, although small, causes the platinum element in the RTD to heat up above the temperature of the RTD environment. This method-Joule heating-is proportional to the electric power (P= IR2) in the RTD and the heat transfer between the RTD sensing element and environment. If the RTD is in a poor heat transfer medium such as air, it will heat up more than if it is in a fluid such as water.

Yet RTDs are not always the best choice for temperature measurement in poor heat transfer media such as gaseous process media, because Joule heating causes an error (self-heating error) in RTD temperature measurement. It is inherent in all RTDs. For these applications, thermocouples are often better, provided other process conditions support using a thermocouple.

The self-heating error of an RTD is normally less than a tenth of a degree in a fluid, but it could be as much as a degree or more in air or gaseous processes.

A thermocouple can develop extraneous junctions along its wire because of cold working of the wire, a temperature difference between the portion of the thermocouple that intrudes into the process and the remainder of the thermocouple assembly. These effects can produce inhomogeneity along one or both thermocouple wires. If the inhomogeneity falls in a temperature gradient, it produces an erroneous output voltage that can add to the normal output of the thermocouple or subtract from it, depending on the temperature gradient and the inhomogeneity's location. This means the thermocouple can indicate an erroneous temperature-sometimes even a negative temperature. That is probably what happened at the 1979 Three Mile Island nuclear power station in Pa.


Normally, an RTD includes a thin platinum wire (sensing element) coiled around a support structure-mandrel. The extension wires are normally welded to the platinum element. RTDs also fail from mechanical stress that the platinum element may experience during construction as the element is bent and secured on its mandrel. So it could help to anneal the RTD after construction to relieve the stress. Because of RTD sensing elements' mechanical weaknesses, thermocouples are generally better for those applications that involve vibration or mechanical or thermal shock of the temperature sensor.RTDs are more immune to noise than thermocouples, because they have a larger relative output, which you can amplify and control better. In terms of noise pickup, thermocouples can sometimes act like an antenna, and their output can become overwhelmed with extraneous noise. Electronic fitters can help alleviate this type of noise pickup as long as response time requirements for the thermocouple are not critical.

Thermocouples usually have better response times than RTDs, but not always. Generally, a bare thermocouple has a faster response time than a bare RTD. Yet when you use it in a thermowell, the response time depends strongly on the air gap between the sensing tip of the sensor and thermowell.

RTDs are generally more accurate and maintain their calibration better and longer than thermocouples. You can also remove an RTD from the process and recalibrate it, which you cannot do with thermocouples because they can develop inhomogeneity along their wires while installed in a process-possibly interfering with thermocouple accuracy. IT

H.M. Hashemian is president of Analysis and Measurement Services Corporation in Knoxville, Tenn.