Sensors - Infrared Thermopiles - EdsCave

6 JUNE 2016

Non-Contact Temperature Sensing

Many common temperature measurements are made by bringing the temperature sensor into contact with the object being measured. Examples include sticking a thermometer into a turkey one is cooking, and gluing a thermocouple onto a piece of equipment being monitored.  Sometimes, though, it can be inconvenient to try to measure temperatures via physical attachment - a non-contact temperature sensor is what is really needed.


Temperature and Radiation

A non-contact temperature sensor relies on the phenomenon of radiation.  All objects with a temperature greater than absolute zero (-273C) emit electromagnetic radiation.  The hotter the object, the more radiation it emits.

The amount of radiation emitted by an object increases of the fourth power of its absolute surface temperature. An object with a surface temperature of 200K would therefor emit 16 times the energy as would be emitted by the same object at 100K.  This fourth-power relationship means that small differences in temperature will result in large changes in radiated energy.  To measure an object's temperature at a distance, we need a detector that is sensitive to this emitted radiation.

When referring to 'radiation' in the context of temperature measurement, we are generally concerned with electromagnetic radiation with wavelengths ranging from a few tens of microns (infrared) down to approximately half a micron (visible light).  This range of wavelengths allows for useful measurements from roughly room temperature to a few thousand degrees C.   While temperature measurements of exotic astronomical objects may require detecting x-rays or gamma rays, these applications are a bit too exotic for this article.


The Infrared Thermopile

An infrared thermopile is a sensor IC  that is designed to measure the amount of incident radiation. The device is directed so that the 'window' on the top of the IC package is pointed at the object whose temperature is to be measured.

Melexis MLX90247 Infrared Thermopile

How the Thermopile Works.

The figure below shows a very simplified view of what is inside a thermopile sensor. On the right is the input window, which is also an optical filter that rejects visible light- this is why you can't see inside the device.  Under this filter is a silicon die which contains the thermopile sensor itself.  Unlike most IC's this die has had a cavity micromachined in the rear so as to produce a thin diaphragm on which the incoming radiation falls.

The micromachined diaphragm is critical to making this sensor work. Because it is very thin relative to the rest of the silicon die, it has a very low thermal mass, and heat is not conducted away to the case as quickly as it would be from the rest of the silicon die. This allows even small amounts of incoming radiation to create a temperature difference or gradient between the diaphragm and the rest of the die.

The actual sensing of this temperature gradient is performed by a thermopile - which is a string of thermocouples connected in series, as shown in the figure below.  One set of junctions (hot junctions) is fabricated on the diaphragm while another set of junctions (cold junctions) is fabricated on the thicker part of the die.  While the voltage generated by a single hot-cold junction pair (or actually the interconnections between them) is very tiny, a series string of dozens or hundred of junctions can generate voltages in the range of a few millivolts, which while still small, is measurable.

Some Practical Considerations

While the simplified transducer described above can result in a sensor that is responsive to the temperature of remote objects, there are a number of issues that must be considered when attmepting to implement a generally usable sensor.

The first of these considerations is the use of the input filter. As mentioned before, this filter passes radiation around 10um. Why this wavelength?  Although a hot object emits radiation over a range of wavelengths, the distribution of energy over these wavelengths is a function of temperature. More specifically, for any given temperature there is a wavelength at which meissions are highest.  For example, an object at 1000C will emit most most of its energy in wavelengths around 2um, while one at 25C will emit most strongly around 10um, as shown in the plots below. Note, however, that the plots have been normalized to the peak emission for that temperature - because of the 4th power emissions law, the 1000C object will emit 300 times the energy it would at 25C.  If one is designing a thermopile sensor to measure temperatures around 25C, it becomes very important to keep out other radiation such as near-infrared or visible light which could interfere with the measurement.

Because the thermopile element itself measures temperature gradients, the measurement is sensitive to the IC package's ambient temperature. Consider the case in which the package is at 25C, and the object being measured is at 50C.  Radiation from the object will warm the transducer's diaphragm to some temperature slightly above 25C, creating a temperature gradient and an output voltage, which is what would be expected.  Now consider the case where the IC package is at 50C - the same temperature as the remote object.  In this situation, both the diaphragm and the remote object are in thermal equilibrium - meaning that the diaphragm radiates infrared energy back out as fast as radiation enters. This results in the diaphragm remaining at 50C, hence no temperature gradient and no output voltage.  

To get around the problem of the transducer's ambient temperature, the thermopile may also incorporate a separate thermistor dedicated to measuring the die temperature.  By combining the die temperature measurement with the measurement of the temperature gradient created by incoming radiation it now becomes possible to account for the effects of the transducer's ambient temperature.

A major application issue is that of field-of-view.  The thermopile's window may accept radiation coming in from a wide range of angles, often more than +/- 45 degrees off from its central axis.  Because the thermopile measures the total energy entering through the window, it can't distinguish between a very small, very hot object that only fills a small part of this field of view from a much larger, but cooler object.  To get consistent measurements, you need to ensure that the field of view is filled in a consistent way. One way of doing this is to collimate the field of view through a smaller hole, so that the transducer only 'sees' a well-defined region of the object whose temperature is being measured.  Of course, the transducer will now also be sensing the radiation emitted by the collimator, so this effect needs to be accounted for as well in the measurement process.

Finally, not all surfaces are equally efficient at emitting radiation.  This emission efficiency is referred to as the surface's emissivity, and varies between 0 and 1, with 1 being a characteristic of the ideal case called a black-body radiator.  There can be substantial variation in emissivity between different materials. For example, concrete surfaces can have emissivities in the range of 0.9, while a bright polished metal surface can have an emissivity of only 0.1.  Emissivity also varies as a function of temperature and surface preparation (roughness, oxidation, etc.) so it is not something that can be taken for granted as constant from sample to sample.  Variations in emissivity from one material to the next can have a dramatic effect on the accuracy of a thermopile sensor, so one must be very aware of how this property varies in the course of the target application.


An Application

One particularly successful application for infrared thermopile sensors is in medical thermometesr, specifically those which are inserted into a patient's ear for the temperature measurement.  There are several reasons that make this application a good match for this sensor.

• Although high precision (0.1C) is needed, the required measurement range is rather limited to about a +/- 10C range. Much beyond this and you are either looking at a corpse or someone on the way to becoming one.
  

• Inserting the thermometer into one's ear effectively fills the transducer’s field-of-view, eliminating this source of potential error
  

• The emissivity of human skin is typically very high (~1). Additionally, the ear canal is a cavity with a small opening. This makes it look like a cavity radiator, which behaves very closely to an ideal black body (emissivity =1) regardless of the material of which the cavity is made.  The combination of skin's high emissivity and the cavity radiator effect combine to increase measurement consistency.
  

• Since the thermopile element is tiny, it comes to thermal equilibrium in a matter of seconds, making for a fast measurement.  In contrast, it takes a lot longer to get a good temperature reading with a traditional oral thermometer.