Understanding the Challenges of Surface Temperature Measurement Using Thermocouples

Understanding the Challenges of Surface Temperature Measurement Using Thermocouples

Thursday, April 14, 2016

Understanding the Challenges of Surface Temperature Measurement Using Thermocouples

Ronald D. Lucier, ASNT NDT Level III, FLIR Systems, Inc.


Voltage from a thermocouple must eventually be converted to temperature. The potential produced by a thermocouple, as a function of temperature difference between its two ends, is very nearly linear over a very wide range. The following curve is the "standard response curve" for a Type K thermocouple. This is the process of calibration.


The most common surface temperature measurements are made with thermocouples; but since a thermocouple’s reading is actually a measurement of its own current temperature, the challenge has always been to get the thermocouple to correctly match the heat of the measured surface. Unfortunately, few infrared thermographers take this measurement uncertainty into account when relying on thermocouple measurements as a reference in determination of emissivity. This paper will explain the theory behind thermocouples and, through demonstration, illustrate many of the problems in their use. In addition, this paper will highlight situations in which the combination of an infrared camera and a thermocouple is preferred, and cases in which an infrared camera by itself is the superior method for surface temperature measurement.


A vast number of commercial and industrial processes rely on accurate temperature measurements. But are the measurements – as performed – accurate? How temperature is measured and what degree of accuracy is required are two very important questions that need to be answered for every application. Entire textbooks are devoted to this one subject!

This paper focuses on one of the biggest challenges: accurate surface measurements made with thermocouples. The author readily acknowledges that while thermocouples provide accurate readings of liquids and gases, surface measurements present a unique set of problems.


A common question asked of an infrared thermography instructor is, “If we really want to measure temperatures, why don’t we just use thermocouples?” This is amusing considering the student is sitting in class with an infrared camera! When asked about thermocouple mountings, most students suggest using electrical tape because it is inexpensive, easy to install, and easy to remove. One particular student employed in the HVAC industry related that he commonly installed thermocouples on compressors using electrical tape and preferred to rely on the thermocouples’ readings than those of his gauges. (Incidentally, he explained that he was in the infrared class because he was replacing too many compressors and hoped IR would give him the answer.)

Temporary attachment of thermocouples is probably the worst option, as the resulting surface measurements will be inconsistent and inaccurate. Permanent attachment by cementing is the preferred method for those who need consistent results. When permanent attachment is inconvenient or impossible, infrared thermography is the preferred – if not only – choice.


Physicist Thomas Seebeck is credited with the discovery of the “thermoelectric effect” in 1821. This states that any conductor subjected to a thermal gradient will generate a voltage. Seebeck misinterpreted the effect, thinking the current was magnetic rather than electric. Indeed, his reports to the Prussian Academy of Sciences in 1822 and 1823 describe his observations as, “the magnetic polarization of metals and ores produced by a temperature difference.”

Two Italian physicists, Leopoldi Nobili and Macedonio Melloni, continued Seebeck’s work to create a thermoelectric battery. This battery is now called a “thermopile.” When Nobili and Melloni coupled the thermopile with a galvanometer, they were among the first physicists to measure infrared radiation.


The illustration in Figure 1 shows a complete thermocouple circuit. Two dissimilar metals are connected in a circuit and subjected to a thermal gradient that will cause a change in voltage.

Simple Thermocouple Circuit

Figure 1. A simple thermocouple circuit

A change in the temperature gradient results in a change in voltage. These voltages are generally in the microvolt-per-°C range. Higher temperature gradients (indicating higher temperatures) generate higher voltages.

The Seeback voltage change is linearly proportional with temperature when the temperature changes are small. The traditional equation is represented as:


In order to measure the voltage change caused by the temperature gradient, a volt meter must be placed in the circuit. This adds two more electrical junctions: one is copper-to-copper, the other is copper-to-dissimilar metal. Now that we have two thermocouples in the circuit, how does the volt meter distinguish between them? Notice the ice bath in Figure 1, which is assumed to be 0°C. It is used as a “known reference junction” or known temperature. Once one junction temperature is known, the other temperature – the one we are trying to measure – can be determined mathematically.

When you purchase a thermocouple and install it, where do you add the ice? For factory-made thermocouples such as the Extech EA10, the manufacturer uses hardware compensation and an internal temperature sensing resistor in place of the ice bath. This is commonly known as an electronic ice point reference circuit and is unique to each type of thermocouple.


Typical Type K calibration curve

Figure 2. Typical Type K calibration curve

Unfortunately, the temperature-versus-voltage relationship of a thermocouple is not always linear. The equation introduced previously shows an ideal relationship with the Seebeck coefficient alpha being a constant. But this is actually not the case, it is a variable represented by a polynomial.

The process of calibrating a thermocouple produces an ideal curve such as presented in Figure 2. The blue line represents an actual millivolt output versus temperature and the dashed line is a “best fit” of the data. This type of thermocouple is relatively linear over a wide temperature range although when the actual data is examined there may be significant non-linear data at specific points. This curve is for illustration purposes only.

The coefficients of the polynomial representing the calibration curve are used with the input of millivolts to determine the temperature readout on the thermocouple.


Like many other 19th century Industrial Revolution inventions, the thermocouple has many everyday uses.

Figure 3. Extech EA10 dual-input thermometer

The thermocouple shown in Figure 3 is typical of general purpose instruments. It’s reliable, economical, and available through multiple outlets. Many manufacturers produce instruments such as this, and the following comments are not unique to Extech (a FLIR company).

The stated accuracy of the EA10 in the user manual is +/- 0.3% + 2°F. The two Type K thermocouples read within 0.4°F to 0.8°F of each other over a five minute period, which is considered a good correlation.

The instruction manual was very easy to understand, showed all of the features of the instrument, and presented the instructions in English, Spanish, and French.

The author downloaded the instruction manuals from several competitors of Extech and found similar manuals for general purpose instruments. All manufacturers included clear, concise instruction in multiple languages.

Despite the copious information provided in the instructions, one very important piece of information was missing: how to mount the thermocouple. Manufacturers offer instruction for operating the instruments, but not for the most important step in the measurement mission: how to make the measurement. This step requires an understanding of the thermocouple environment. The only information on mounting a thermocouple was found on a sheet of instructions that came from Omega Engineering and referred to their epoxy adhesive. Refer to the latter section on permanent thermocouples.


Thermocouples by their very nature only indicate the temperature they have reached. In order to measure a solid, liquid, or gas the challenge is to normalize the thermocouple to the same temperature as the solid, liquid, or gas. This paper explores the biggest challenge – solid surfaces.

Surface (noun)

  • The exterior or upper boundary of an object or body
  • A plane or curved two dimensional locus of points
  • The external or superficial aspect of something

Merriam-Webster’s Dictionary (www.m-w.com) provides a simple definition of a surface important to review here, as it is the very nature of the surfaces involved that present the biggest challenges in temperature measurements. Describing a surface as “two dimensional locus of points” relates best to the problems of thermocouples as seen with infrared thermal imagers. Technically, points are infinitely small, but, when collected in an area, define a surface.

Type J & K beaded wire-to-surface configuration

Figure 4. Type J & K beaded wire-to-surface configuration

Two common forms of thermocouples are Type J and Type K beaded wire thermocouples, which are typically attached to surfaces temporarily. These thermocouples can also be permanently cemented in place.

When using Type J or K thermocouples, either temporarily attached or cemented, an obvious measurement issue is the problem of heat conduction. That’s because the primary mode of heat transfer between the object surface and the thermocouples is conduction. The Fourier Law of Conduction for a one dimensional surface is:

This equation and terms should be familiar to all Level I, II, and III infrared thermographers. What value do you use for A? Technically, since it is a point, it must be zero. In reality, it cannot be zero since we get a reading once we place a thermocouple on a flat or curved surface. Try this and note how slowly the temperature increases. The area available for heat conduction is very small.

In the case of very complicated conduction heat transfer processes, simple equations no longer work. Here we have to modify the above equation and introduce the concept of the conduction shape factor. The conduction shape factor is used to account for the specific geometry not accounted for in general equations such as the one above. Therefore, we can rewrite that general equation into a new form in which the term S is our conduction shape factor:

In calculations involving a sphere and a plane, as represented in Figure 4, we get a rather complicated formula where D is the diameter of the solder ball and r is the radius:

It is obvious that if the value S is large, the heat transfer will be larger and more heat will transfer into the thermocouple, giving a more accurate response. So far we have accounted for the conduction of the heat from the flat surface to the spherical surface. Now we have to get the heat from the spherical surface to that dissimilar junction so a thermal gradient can be established. This involves another equation:

In the above equation, r1 is the radius of the solder ball, r2 is the radius of the thermocouple wire and k is the thermal conductivity of the material in the solder ball.

The transfer of heat from the surface we wish to measure into the device that will be measuring it – even when approximated under steady state conditions and equations – involves some interesting mathematical concepts. If this situation is extended into transient conditions, where the temperatures are changing very rapidly, the math becomes very complicated and the measurement challenges become extremely difficult.


Now that a thermocouple has been placed on a surface and the complexity of the heat conduction shape factor has been introduced, the problem becomes truly complex with addition of another process—convection. All surfaces are subject to the three forms of heat transfer: conduction, convection, and radiation. The exception is when the surface exists in a vacuum, where convection cannot occur.

The thermal image in Figure 5 shows a Type K thermocouple attached to a surface with a piece of electrical tape (the entire test setup is shown in Figure 7). The thermal gradient causing the Seebeck effect is clearly shown, as are the emissivity differences between the tape and the metal surface.

Figure 5. Thermal image of a Type K thermocouple on a warm surface

The differences in the radiation heat transfer are apparent. The piece of tape securing the thermocouple to the surface has the effect of insulating the flow of heat from the surface of the can. Because thermal resistance to conduction heat transfer has slightly increased in that area, this area will radiate and cool faster. You can see the results in Figure 5: the tape appears considerably brighter than the metal surface.

To further confuse matters, the effects of convection have to be recognized. Traditional infrared cameras have limited abilities to see small temperature effects caused by convection. However, the FLIR GF320 camera has a High Sensitivity Mode (HSM) with sequential image frame subtraction, allowing for the detection of minute quantities of trace gases. Look at the very small differences in the images in Figure 6. The surface in question was 112 °F (44.4 °C); temperature variation caused by convection across the surface was as much as 0.7°C.

Figure 6. HSM depiction of convection at a thermocouple surface


The net effect of all the above is shown in Figure 7. One thermocouple was placed on the inside of the can and secured to the surface and one on the outside, in approximately the same place. The expectation is that the two readings should be the same, but they aren’t.

Figure 7. T1 and T2 measurements don’t match up

The two thermocouples agreed within 0.5 °F most of the time, but there was variation. The variation can be explained from the room convective currents, as indicated in Figure 6. There are stronger convective currents inside the can of water due to the density of water, but that would cause the temperature indication of T2 to drift versus T1, but not cause the difference.

The difference in the readings here is caused by all the effects previously mentioned, but most significantly by conduction heat transfer. The shape of the solder ball at the end of the beaded thermocouple provides very little area for heat conduction. The magnitude of the conduction shape factor indicates the same.

Now that we’ve concluded that temporary attachment of thermocouples is a poor practice, what about permanently mounted thermocouples?

Figure 8 is an image of a crude oil heater taken with a FLIR GF309 furnace camera. The tube temperatures are measured to determine tube life and to determine operating parameters of the heaters. One thermocouple was permanently fixed to the surface of one newly installed tube and is in the extreme upper left of the image, just out of view.

Figure 8. Crude oil heater

The thermocouples in the heaters are subjected to an extreme environment: gas temperatures are in excess of 2,800°F and a radiation environment in excess of 1,800°F (1,537°C and 982°C, respectively).

Unfortunately, the thermocouples measure at a single point. While they provide useful data, at least for a while, they do not indicate whether coke is building up inside the tubes. Severe coking will eventually lead to tube ruptures. You can see the coke build-up in Figure 8 as white hot spots in the thermal image.

These severe environments obviously affect the thermocouples’ long term performance. Unfortunately, they are not the only problem.


The results from using electrical tape as shown in Figure 7 are unacceptable. A better method was investigated: cement-on-thermocouples. The author expected that at nearly five times the price, the results would be much improved—and they were.

The author chose two Omega CO1 Type K fast response thermocouples. The foils are 0.0005” thick and bonded between a very thin polymer/glass laminate. These thermocouples are very thin and flat, ideal for mounting on curved surfaces. Omega recommended OB200 Epoxy for mounting.

The epoxy was mixed (two parts resin and catalyst) according to the manufacturer’s recommendations, applied to both surfaces, and cured close to the recommended temperature. Again, the author notes that the only place indicating how to mount a thermocouple was found in the instructions for the epoxy, not for the thermocouple itself.

Infrared results while curing

Figure 9. Infrared results while curing

Infrared results while curing

Figure 10. Visible (thermocouple) results while curing

The results present a puzzle: what to believe? Each thermocouple was mounted in the same fashion at the same
location inside and outside the container. The can was heated with a hot plate and based on simple convection heat transfer—different film coefficients between an enclosed space and an open one—one wouldn’t expect the two readings to match up. There were many times during the day when the two thermocouples read exactly the same, but there was an 8°F temperature difference with the thermocouple attached to the outside of the container, and a 13°F reading difference for the thermocouple inside of the container at the time that the thermal image was obtained. The outside thermocouple was experiencing additional room convection from the air conditioning. There was little consistency in readings between the camera and the thermocouples, particularly the outside one (smaller case reading T2 on the thermocouple readout).

Once the thermocouples were adequately cured and checked for integrity, a third thermocouple was taped on the surface and hot water was added. Results below:

Infrared results while curing

Figure 11. Infrared results with hot water in can

Infrared results while curing

Figure 12. Thermocouple reading with hot water in can

Notice that now with the water inside the can, the cemented-on thermocouple almost exactly agrees with the infrared measurement of the tape. Both are within 1°F of the inside water temperature (the water had to be mixed vigorously to limit natural convection and decrease the convective boundary layer) but significantly disagree with the taped-on thermocouple. Obviously, temporarily attached thermocouples are completely unreliable and should never be used. In contrast, the cemented-on thermocouple agrees with the infrared camera measurement, but only when there is water in the can.

These results illustrate the difficulty of seemingly simple applications. The images in Figures 6, 9, and 11 provide clues as to the reason for the discrepancies and why thermocouples and IR cameras sometimes don’t match: convection. In Figure 11, the can was naturally cooling and was on a surface at room temperature. In Figures 9 and 10, the can was empty and being heated from the bottom by a very hot plate. Therefore, the two convective heat transfer conditions are very different. Secondly, the thermal inertia of the two cans (volumetric heat capacity) was significantly larger in the can with water as opposed to the can with air. Lastly, as evidenced in Figures 9 and 10, there was a much larger thermal gradient in Figure 9, making the selection of measurement points all that more important!


Now that it has been established that surface measurements with thermocouples are, at best, very complex, what else is there to consider? There are several issues concerning the long term performance of a thermocouple.

  • Poor junction connection – How the thermocouple is mounted is critical. Temporary connection via electrical tape is a poor practice in just about every situation. A conductive cemented connection or a thermocouple specifically designed to be cemented on is probably a better choice, but its environment must be considered. In high temperature environments, specialized thermocouples are welded to the surfaces. These welds can separate under stress, making their readings no longer reliable. This can be a sudden or gradual process.

  • Decalibration – Operating a thermocouple at very high temperatures or in corrosive atmospheres can cause diffusion of atmospheric particles into the thermocouple metals, changing the entire characteristics and calibration of the thermocouple. A high temperature annealing can occur along the section of the thermocouple that has the highest thermal gradient causing calibration issues.

  • High temperatures – High temperatures can affect the insulation around the thermocouples. In the case of heaters, the thermocouple may actually measure gas resulting from insulation breakdown rather than the surface itself.

  • Galvanic action – If the insulation contains certain dyes that leach out in the presence of water, electrolytes will form. The resulting current can be hundreds of times greater than the Seebeck effect.

  • Thermal lag – Ideally, we want to have a small thermocouple so that it will not affect the surface temperature. However, small thermocouples generally have small wires, which may significantly affect response time.


Lessons learned:

  • Thermocouples reliably measure the temperature of themselves, within the accuracies stated by the manufacturers.

  • The accuracies stated by the manufacturers are not necessarily the accuracies of measurement.

  • While the ambition is to get the thermocouple to the same temperature as the surface, in reality, that is impossible.

  • Thermocouples offer the opportunity for an accurate measurement of liquids and gases, but fall short of providing an accurate measurement of surface temperature.

  • The thermocouple greatly affects the convective and radiation heat transfer locally at the surface, thus affecting
    the reading.

  • Temporarily attached thermocouples should be avoided.

  • If installed correctly, cemented-on thermocouples are most likely reliable.

  • Proper placement of thermocouples should be done with an IR camera to avoid strong thermal gradients.

  • Avoid placing thermocouples where strong convective currents will affect the readings.

  • Assume nothing. Periodically check the calibration and integrity of permanently installed thermocouples.


Extech EA10 Dual Input Thermometer Instruction Manual

Omega Engineering OB-200 Epoxy Adhesive Instruction Sheet

Omega Engineering CO1, CO2, CO3 User’s Guide

ASHRAE Handbook – 1985 Fundamentals (pages 13.6 & 13.7)



“Introduction to Infrared Systems Engineering”, ©1969 Richard D. Hudson, reprinted in SPIE Milestone Series MS116

“Selected Papers on Temperature Sensing: Optical Methods”, Ronald D. Lucier, Editor (1995)


“Introduction to Infrared Systems Engineering”, ©1969 Richard D. Hudson, reprinted in SPIE Milestone Series MS116


Ron Lucier is a Mechanical Engineer with more than 32 years of engineering experience, 27 of which have been devoted almost exclusively to the use of infrared thermography. He holds an ASNT NDT Level III certificate (number 48004) and is a Senior Instructor for the Infrared Training Center of FLIR Systems. Ron is also a licensed pyrotechnician and a Master Magician.

Recent Posts: