The Thermocouple
Many of the transducers we’ve looked at over the last few weeks really are not transducers in the strictest sense, they don’t change a form of energy into another. Other items we’ve looked at, tend to exploit electro-magnetism and changes in fields to produce measurable results, all interesting and fascinating in their own way but fairly straight forward, well known (if less understood) physical effects. This week we’re looking at a far stranger energy transformation – heat to electricity.
We’re used to electricity producing heat, many of you will remember that passing a current through a wire will result in the wire heating up, an effect called joule heating, but we all know this can’t be reversed.
There is a far less well known effect, called the Peltier or Seebeck effect that’s much… stranger.
If you get two metals that aren’t the same as each-other, connect them at one end, and then produce a temperature gradient between the joined and non-joined ends you find a difference in electrical potential between the unconnected ends of the two metals. What on earth is going on?
There can’t be any potential difference between the metals at the joined point, it would equalise as no siginificant resistance exists at their connection. Normally this would hold true for two metals that are electrically conductive throughout the whole mass, and yet a potential difference is maintained. If you connect the hitherto unjoined ends but maintain the temperature gradient, a current begins to circulate. It is this phenomenon that is called the seebeck effect, or thermo-electric effect.
Despite this being such a peculiar and lesser known effect, it’s discovery goes all the way back to Volta himself using frog-leg osciloscopes, although Thomas Johann Seebeck re-discovered it in 1821 and described it.
But what is actually going on here? When a conductor is heated, the electrons within it become ‘excited’. If only one end of the conductor is heated excited electrons make their way from the heated end to the cooler end and stay there, becoming less excited. This concentration of charge generates a potential difference between the hot and cool ends of the conductor. The bit that gets the whole effect going is that the amount of voltage built up in response to a particular temperature difference varies between different metals (and different alloys of metal). So in our two pieces of metal the build-up of potential at the ends is slightly different. That difference is what produces the current. This effect can be used to generate power, but it can also be measured with a voltmeter. If we have a handy table with the difference in voltage in response to temperature gradient we’ve got ourselves a temperature sensor: a thermo-couple.
Thermocouples work because for certain pairs of dissimilar metals we know in great detail the potential difference in response to a particular temperature difference, usually expressed in milli-volts per kelvin. A big limitation is that the response is not to temperature itself, but the gradient. To work out the temperature at the ‘probe’ end, you need to know the temperature at the measured end. This is often called T-ref, or the reference temperature. This means you need to run the two metals all the way from the probe end to the point of t-ref, which is usually at the point where the voltage measurement is conducted. Special thermo-couple wire is used to do this which is identical in composition to the thermo-couple type being used.
A contentious issue in thermo-couples, is where exactly the voltage potential is generated, with some saying that it’s over the whole wire, others at the point of metals meeting (often called the bead). Technically speaking the potential is actually generated over the region of temperature gradient, and so both could be true, depending on how the thermocouple is being used. In laboratory settings quite often t-ref is taken as 0 degrees Celsius (or 273.15K) as the thermo-couple is terminated in a bath of ice water which produces a gradient over the entire length of the thermo-couple. If being used in a gas boiler to sense if the pilot light is lit, then the gradient is almost entirely concentrated at the bead of the thermocouple and therefore the potential generation is highly localised.
You could also argue that as the hot point is the bit that is energising electrons and causing them to migrate, it is the part that causes the potential. You could also argue that the cool part is the bit that actually causes the gathering of those electrons and so that is the part that generates the potential. On top of this, the probe end is not always the hot end, it can measure temperatures below t-ref also which would reverse which bit is energising electrons and which end is gathering them. You can see why, when it gets technical, arguing where the potential is generated can get complicated.
At the end of the day what matters is how we need to use thermocouples to get the results we want from them. This is why we bother running thermo-couple cable all the way to the point of t-ref when most of the wire is normally at the same temperature. Whilst in most engineering settings most the wire beyond the hot point is near room temperature, we often don’t know the exact region of gradient and over the period of measurement the region of the temperature gradient can change as conduction heats the surroundings and the wires. Whilst the wire might be essentially at the same temperature for 95% of its length, we still need to know what temperature it is. There are very few environments in which we will measure the temperature of the wire, but not also conduct the voltage measurement at the same point, although there are several exceptions.
So how do we measure the temperature at the reference point? Typically we use another temperature sensing method that is accurate over smaller temperature ranges, usually a resistance based sensor. T-ref is not going to vary much, maybe up or down by 50 degrees, and we have absolute temperature measurement devices (rather than relative as a thermo-couple is) that function very well within that range. Why use thermo-couples then? A typical type K thermo-couple can happily accurately measure temperature across a range from 0 to 1100 degrees centigrade, and unhappily go down to -180 degrees below and up to 1370 degrees whilst maintaining an accuracy in the region of a single degree hence why they are used so readily. Some can go to significantly higher temperatures.
But what if we’re using a thermocouple that’s say, platinum based, suitable for extremely high temperatures up to 1800 degrees centigrade. Getting hold of several metres of platinum wire can get rather expensive. What we can do instead is get something that matches the seebeck behaviour of platinum over a smaller temperature range maybe +-70 degrees; the gradient we might expect along the wire away from the extreme heat of the target. This behaves like platinum wire, but saves a large amount of costs. But only works if the majority of the temperature gradient is within the platinum region of the thermo-couple. If we’re measuing the core of a gas turbine, or a furnace temperature, once we’re outside the wall of the combustion chamber the majority of the gradient is from thermal conduction along the wire itself, which is considerably smaller than within the chamber, allowing this approach to be used.
There are loads of internationally recognised thermo-couple types, the most common being the aforementioned type K. There are also at least three different standards of colour codes for each one, which makes my engineer blood boil, but I won’t knock this article off the rails complaining about poor standardisation practices.
Thermo-couples are incredibly wide spread in our modern world, for many applications, you only need to know if something is hot and knowing t-ref with any more detail than near room temperature is irrelevant. One example is detecting flames in gas cookers and boilers, thermocouples are perfect for this. The electrical energy they create, although small, is enough to open valves, creating automatic sensor-actuator systems that are robust and simple. This is what the vast majority of thermo-couples in the world are used for.
In engineering applications they remain the go to measurement device for high temperatures as they possess good accuracy, high range, and are simple to use. Their only downside is that they are not particularly linear and so non-linear corrections are necessary to ensure good data.
It’s these qualities that mean thermo-couples, even with a transition away from gas and combustion based technologies, will be with us for decades and centuries to come.
Do you have a job that requires measurement or characterisation of high temperature environments? Here at Flintmore we have extensive experience working with thermocouples in the heart of gas turbine combustion chambers. Get in touch with us today to talk about your challenge.
Thermocouple Further Reading
If you want to delve further into thermocouples, the wikipedia article is extensive and gives good reference material for temperatures of operation whilst being a little unfriendly to someone just learning the subject.
https://en.wikipedia.org/wiki/Thermocouple
Omega Engineering have a variety of good articles on a number of transducer types, we’ve linked to them before. Their article gives good detail and covers a lot of the same ground as ours, useful to get a different perspective on the subject:
For an excellent resource for working with thermocouples, I highly recommend thermocouple info. Run by Reotemp instruments it provides extensive information for working with these sensors.
https://www.thermocoupleinfo.com/thermocouple-reference-tables.htm
Other articles:
Are you curious to find out more about transducers?
The Galvanometer – before the advent of digital to analogue converters, the galvanometer would have been used to detect the small currents induced in a thermocouple.
Strain Gauges – for a completely different sensor (and not technically a transducer), have a look at last weeks article – strain gauges.
