Sensors - Thermocouples - EdsCave

Go to content

Main menu

Sensors - Thermocouples

Sensors

3 JAN 2016


When you need to measure temperature over extended ranges, for instance a few hundred degrees C, thermocouples are often the preferred temperature sensor.  They are rugged, inexpensive, and some varieties can measure temperatures over ranges exceeding 1000 C.

A thermocouple is constructed by connecting two wires of dis-similar metals together at a junction point which is placed in contact with the object whose temperature is to measured. These wires are then run out to the temperature measuring instrument. The output of the thermocouple is a voltage which varies with temperature. The picture below shows a typical thermocouple wire with the sensing junction at one end and a termination connector at the other. The two wires are insulated from each other with woven fiberglass sleeving.


Type 'K' Thermocouple (One of the most common) & Terminator Block


Contrary to popular belief, the voltage developed by a thermocouple is not generated by the junction itself, but by the difference in temperature between the junction point, and the other end of the wire by which is connected to the measuring instrument. For this reason it is possible to make a thermocouple junction in a number of ways - welding, mechanical joining, or even brazing or soldering with a third metal!

Thermocouples are based on the Seebeck effect - which states that when there is a temperature gradient along a conductor, there will also be an associated voltage gradient. For example, if you heat one end of a piece of metal while keeping the other end cool, there will be voltage gradient along the length. Both the magnitude and sign of this voltage gradient are determined by the material's Seebeck coefficient.


Creating a temperature gradient along a conductor also creates a voltage gradient

Demonstrating the Seebeck effect with a single metal is difficult. For example, if you take a wire, heat it in the middle, and measure the voltage at each end, you will note the voltage to be zero, as the voltage drops along each end of the wire will equal each other and cancel out.  To be able to measure the Seebeck voltage requires the use of two conductors, each with a different coefficient, as shown below. The amount of voltage you see at the ends of the wires is proportional to beoth the difference in temperature along their lengths, and the difference in the two materials' Seebeck coefficients.


A thermocouple's voltage is generated by differing Seebeck voltages developed along each of the differing wires - not by junction effects

The Seebeck coefficients for most materials, especially metals are typically not very large.  The following table lists the coeffcients for a few materials. You can see that for the case of metals, pairing chromium with nickel gives the largest difference, which is only 26uV/K  (+18uV/k - -18uV/k).  Since a thermocouple generates microvolt-level signals,amplification, filtering and other signal processing is almost always required to use them in any given application. Note however, that the coefficients for 'N' and 'P' type silicon are pretty large - in the hundreds of uV/K. This makes it relatively straightforward to implement  sensitive thermocouple-type sensors on integrated circuits.

Material

Seebeck coefficient (uV/K)

Chromium

18.8

Gold

1.79

Copper

1.70

Aluminum

-1.70

Platinum

-4.45

Nickel

-18

P-type silicon (mono)

300 to 1000

N-type silicon (poly)

-200 to -500

After Herwaarden & Meijer in [1]

There are a wide range of thermocouple types available for different types of applications.  The output voltage vs temperature curves for a few common types of thermocouples are shown in the graph below.



While more voltage (higher sensitivity) is a desirable feature, it is far from the only factor to be considered in selecting a thermocouple. Some other key factors that should be considered are:

  • Service temperature - Just because you can get a voltage out at a given temperature does not mean the particular thermocouple will operate at that point reliably.

  • Linearity - How straight is the curve in the region of intended use?  Higher linearity requires less downstream signal processing and compensation.

  • Mechanical - Does the thermocouple become mechanically unstable (brittle or excessively soft) at the intended application temperature?

  • Chemical - Will the thermocouple react with the environment (oxidize/reduce/corrode)?

  • Cost - Is the price right? For example, Type K thermocouple wire (chromel-alumel) is considerably less expensive than Type R (platinum/iridium alloy)


You may have noticed that all of the voltage vs. temperature curves in the chart above show 0mV output at 0 degrees C. One way of accomplishing this measurement would be to bring both the end-terminals down to 0C. Another way is to add a secondary 'ice-point' reference junction that is held at zero C, as shown below. This has the advantage of allowing similar metal connections to be made to the measuring instrument.



'Ice-Point' Reference Junction


One downside, however,  of using an ice point reference is that stopping off at the local 7-11 to get a bag of ice every time you need to make a measurement gets old after a while. Modern thermocouple-based measuring instruments dispense with the ice by using a secondary sensor to measure the temperature of the cold-junction block where the thermocouple wires are connected to the instrument (Figure below). The voltage across the thermocouple is amplified up to useful levels, and this secondary temperature measurement is used to compensate the thermocouple measurement.


A Measurement of Termination Block Temperature can be used to compensate the thermocouple without an ice-bath.


Since thermocouples are common sensors, you can buy  integrated circuits (ICs) that implement the key signal processing functions. For example, the Analog Devices AD8495   provides an on-chip amplifier, cold-junction temperature sensor, and compensation circuitry for type-K thermocouples. This chip directly accepts the voltage generated by a type-K thermocouple, and outputs a 5mV/C calibrated output voltage.  Becuase the AD8495's  cold-junction temperature sensor is internal to the IC,  to get the benefit of accurate cold-junction compensation, you must ensure that the cold junction temperature is close to that of the IC. In many cases this is not all that difficult to arrange if the cold junction is on the same PCB, and in close proximity to the IC.  Similar ICs are also available for handling J-type thermocouples.


References:

[1] Semiconductor Sensors, S.M. Sze ed., 1994, John Wiley & Sons, New York.
[2] Transducer Interfacing Handbook, Daniel H. Sheingold, 1981,  Analog Devices, Norwood MA

Back to content | Back to main menu