Electronic Measurement and Control of Temperature
by Jerry Casebeer
Among the many hot ideas in electronics has been the development of electronic temperature sensors. This is not surprising, since temperature has so much to do with the way we go about our business now. Concerns about global warming, energy consumption, pollution, costs, and disease all involve exacting temperature measurement and control.
In medicine, the doctor may need to measure temperature, not just to see if you have a fever, but also to look for infections, circulatory problems, or other disease processes. Of course, the people at the National Weather Service use balloons equipped with radio to measure air temperature at higher levels in the atmosphere.
Since most chemical reactions involve temperature, chemists often measure and record the temperature in a reaction vessel when they are investigating new processes. And automobiles usually come equipped with either an engine temperature gauge or the less helpful “idiot light.” In all these examples, and in many more that could be mentioned, electronic temperature sensors are used routinely.
Electronic temperature control, often with the aid of a digital computer, has become an important part of our lives, as well. The production of many of the new “high-tech” materials is dependent on precise control of temperature during the manufacturing process. Where process yield, cost, or pollution are concerns, automatic temperature control can be absolutely necessary.
For example, oil refineries pump crude oil through a steam heater to vaporize it. This vapor rises in a column inside` a tank where the temperature lessens as it rises. At the higher levels of the tank, lighter products such as naptha and gasoline condense and are collected. At lower levels kerosene, diesel fuels, and different “weights” of lubricating oil are recovered.
At the very top, dangerous hydrogen sulfide gas is vented off and burned immediately. To a certain extent, the refinery can be “tuned” to yield more or less gasoline, heating oil, and other products, depending on the expected need for them in the market. Without electronic temperature control, this “tuning” would be very difficult, if not impossible.
Electronic process controls are also used in such familiar applications as the automobile. At one time a bimetal spring controlled the automatic choke, and set the carburetor for a richer or leaner mixture, depending on the temperature of air around the exhaust manifold.
Today, a computer measures the coolant temperature and combines this information with signals from other sensors that measure exhaust oxygen content, intake manifold pressure, and cylinder detonation to adjust the timing of the spark, fuel injectors, and other devices. This allows the engine to deliver better performance while using less fuel and causing less pollution. Similar control computers are used to improve the performance of the engines in trucks, aircraft, locomotives, and ships.
Electronic power plants, chemical plants, and even food preparation plants all benefit from the use of controls. In large buildings the heating, ventilation, and air conditioning (HVAC) are improved with controls, too.
The type of sensor chosen for a particular application depends on the temperatures to be measured, as well as the precision and accuracy required. Although the words, “precision” and “accuracy” are often used interchangeably, they have special importance in metrology - the science of measurement. These special meanings are easily understood, as the following illustrates.
Suppose there were two people stranded on a desert island where, with some effort, they were able to make a living. Let’s say what one day a thermometer washes up on the shore. Unfortunately, it has been floating around in the ocean for so long that all the markings on the board to which it is attached have been washed away.
Now both people agree that it would be useful to have a thermometer. Suppose that there are mountains on the island, and that there’s snow at the tops of these mountains. So our two people, after some adventures perhaps, climb a mountain, put the thermometer in an ice bath, and mark that point on the thermometer board. Then, after perhaps some more adventures, they return to their hut where they have a supply of fresh water and a pan to boil it in. So they put the thermometer in the pan, bring the water to a boil and mark the thermometer board again.
They now have calibrated their thermometer to standards. As far as they are concerned, the markings on the board are THE standard temperatures. Now they must consider the issue of dividing the distance between their calibration marks into degrees. Since they’re both Americans, they decide to use the Fahrenheit scale. They begin by reasoning that the distance between the boiling point (212 deg. F.) and the freezing point (32 deg. F.) is 212 degrees - 32 degrees: 180 degrees.
Suppose that they tried different methods to divide the scale, and that for some reason they decided to use six major divisions. These should fall at 30-degree intervals: 32 degrees, 32+30, or 62 degrees, 62+30, or 92 degrees, and so on up to 212 degrees. Figure 1 shows their first attempt. This scale is “nonlinear” because the markings are bunched at one side and spread apart at the other.
Figure 2 shows their next attempt, with each of them trying their own method of spacing the marks. The scale in Fig. 2-a is more “accurate” than the one in Fig. 2-b because the markings are closer to their “true” positions. Since marks at 30 degree spacings are not very useful, they both try their own method of dividing up the scale again.
Fig. 3 shows their results. Although both scales are equally accurate, the one in Fig. 3-a is more “precise” because you can make closer measurements with it.
The users of electronic temperature sensors must consider the linearity, accuracy, and precision of their devices to make sure they suit their applications. We will now look at each of the major kinds of electronic temperature sensors, keeping these considerations in mind as we go.
One of the oldest temperature sensors is based on the simple principle that when two “dissimilar," (unlike) metals are joined, a voltage appears between them. In general, the higher the temperature, the higher the voltage. This is known as the “Seebeck Effect” names in honor of its discoverer, Thomas J. Seebeck.
In 1821 Seebeck was experimenting with metals when he constructed the apparatus shown in Figure 4. It consisted of a long wire joined to a shorter one. The longer wire was made of copper and the shorter was bismuth. In the middle Seebeck places a long magnetized needle supported on a pivot in the middle. Basically, this was just a king-size compass needle. By allowing the needle to orient itself to magnetic north, he has a simple, but very sensitive sensor of magnetic fields.
Seebeck found that when he heated one of the metal-to-metal junctions (J 1 and J2 in Fig. 4) the compass needle would deflect either left or right. But this only happened when the junctions were at different temperatures. If he heated one junction and left the other at room temperature, the needle would swing in one way. If he heated the other junction instead, the needle swung the other way.
Seebeck found that this happened with any two metal wires, so long as they were made of different metals. Further, the strength of the magnetic field depended on which metals were joined. He decided that somehow the heat was being converted directly into magnetism. Years later he realized that a voltage was appearing at each junction. This voltage produced an electric current in the wires, and the magnetic fields were the result of the current.
The hitch in using the thermocouple as a temperature sensor is illustrated in Figure 5-a. Ideally, we would like to be able to simply join two dissimilar metals together and measure the voltage between them. But when you connect a sensitive voltmeter to the two wires, you must create two more thermocouples, as in Fig. 5-b. The voltages from these junctions interfere with the voltage from the measurement junction. Instead, the arrangement shown in Fig. 5-c must be used. Junction J1 is used as the measurement junction, while J2 serves as the “reference junction.”
Over the years, it has become standard to use the “triple point” of water, 0 deg. C., as the temperature at the reference junction. A very good temperature standard can be made with distilled water and ample amounts of ice mixed together in an insulated container. For more accuracy, the water can be highly purified and the barometric pressure taken into account.
Often this gives enough accuracy for all but the most demanding applications. In these situations, the thermocouple may be calibrated against an instrument whose accuracy is traceable to THE standard maintained by the National Institute of Standards and Technology (formerly the National Bureau of Standards.)
Over the years a number of different metals have been developed that are best suited for use in thermocouples. The various thermocouple types have been given different letter designations, as shown in Figure 6.
There are two problems with thermocouples. One problem is that they produce relatively small voltages, on the order of tens of microvolts per degree Celsius. This requires highly sensitive voltage measuring instruments. Since standard analog-to-digital converters (ACDs) require inputs on the order of 5V, specialized “instrumentation” op amps, along with low-pass filters to reduce noise, are frequently used.
The second problem is that thermocouples are nonlinear devices. In bygone years, the user measured the voltage and then used a table to the corresponding temperature. For intermediate values between the table entries, the differences were worked out by using a proportion. This is known as “interpolation” With digital computers, a program can be set up to refer to a table stored in memory, and interpolation can then be done by calculation software. Where better accuracy and precision are required, the software can use a special formula in the form T= aV7 + bV6 + cV5 + dV4 + eV3 + fV2 + fV + h where V is the voltage reading.
Depending on the thermocouple type, certain numbers are substituted for a, b, c, etc. in the formula. These have been determined and published by the NIST and are available in computer software form, too.
To eliminate the need for an ice bath, the reference junction is mounted on a special block where its temperature can be measured by lower-temperature devices such as the thermistor. The reference junction’s temperature can then be used to compensate the temperature reading. Special ICs containing internal reference junction compensation, a customized instrumentation op amp, and other circuitry for thermocouple use are available from Analog Devices Corporation.
Resistive Temperature Devices
About the same time that Seebeck published his work on thermocouple voltages, Sir Humphry Davy discovered another temperature sensor. He found that the electrical resistance of metal wires increases linearly with temperature. Although the percentage increase is small, it is very accurate. As long as the metal wire is not chemically contaminated. sensors having very good linearity, accuracy, and precision can be made. Because it exhibits a higher percentage increase than gold, can withstand high temperatures, and is chemically inert. platinum was soon established as the standard. Platinum Resistive Temperature Devices (PRTDS) are often used in demanding laboratory work, especially where strong acids, bases, or corrosive chemicals are involved.
The mass of a temperature sensor can cause "thermal shunting." For example. suppose you had a big outdoor thermometer intended to be nailed to a barn. lf you tried t0 use it to measure the temperature of a cup of Coffee, its mass would interfere with your measurement. Accordingly, PRTDS are made with very fine wire encapsulated in tiny glass beads. Thin-film types made with processes like those used in making ICs are made to reduce thermal shunting even more.
Most PRTDs are made to have 100 Ohm's resistance at 25 deg. C., the standard "room temperature" used in electronics. Special devices with as little as 10 Ohms or much as Ohms at 25 deg. C. are available, though. In the past, Wheatstone bridges and other circuit arrangements were used to determine the PRTD`s resistance. Today, a current regulator is usually employed. Since Ohm’s law says that E=IR, the voltage across the device varies linearly with its resistance, and thus indicates the temperature. Attention must be given to the resistance of the connecting leads and the effect of temperature on their own resistance.
Unlike the thermocouple, the PRTD is not self-powered. To determine the resistance, it is necessary to pass some current through it. Since the power dissipated in any resistance is I2R, the measurement current produces some “self-heating” in the device. One solution is to use the lowest current possible, and to use an instrumentation op amp to boost the small voltage being measured. A second remedy is to apply the current in brief pulses lasting only long enough for the amplifier to stabilize, and allow enough time between pulses for the PRTD to give up its heat.
Special precautions must also be taken when such low-level voltage signals are involved. At every connection on the circuit board, two dissimilar metals may be joined. This results in unwanted thermocouple junctions at various points in the circuit. To minimize their effects, a differential amplifier must be used as shown in Fig. 7. If identical metals are used to form the connections, and if the connections are close together (thus keeping them at the same temperature) their voltages balance each other out. The differential amplifier responds to the difference between its input voltages. If these unwanted thermocouples are arranged to present “common-mode" voltages, they will be greatly reduced at the amplifier’s output. In addition to these measures, special shields called “guards” also help to reduce electrical interference. The same techniques are often used in amplifying the output signals from thermocouple temperature sensors, too.
The thermistor is basically an inexpensive device whose resistance changes with temperature. Its main advantage is that it offers a large change in resistance for a given change in temperature. The circuitry needed to operate it can thus be simple and inexpensive. Most thermistors have a negative temperature coefficient (NTC,) meaning that their resistance goes down as the temperature rises. Some, however, are made with positive temperature coefficients (PTCs.)
Thermistors are normally rated according to their nominal values at 25 deg. C. As with ordinary resistors, 5% and 1% types are available. Some kinds come in the standard resistor values, while others are made just in the decade values (l00Ω, 10000Ω, 10K, etc.)
Thermistors have very poor linearity, although special “linear” types are available. Ordinary thermistors are useful up to about 150 deg. C., although special nickel-chromium types are made that can be operated at up to 1000 deg. C. Thermistors are often used as part of the engine controls in automobiles. They are mounted in a sensor that threads into a hole in the engine block, and the thermistor itself is immersed in the coolant.
As in ordinary resistors, when current passes through a thermistor heat is developed. Normally this "self-heating” is a disadvantage, But some thermistors are deliberately made and operated so that self-heating will make their resistances change. One of the most common applications of this effect is in the automatic demagnetizing (“degaussing”) coil around the picture tube of a color TV. A PTC thermistor is connected in series with this coil so that when the set is turned on, the thermistor is cold, its resistance is low, and the current is high. As self-heating warms the thermistor, its resistance increases. This lessens the current, but increases the voltage dropped across the thermistor. This in tum increases the self-heating. The process continues, with the thermistor heating up and the current going down, until the current through the degaussing coil has been reduced to a very low level. So for all practical purposes, the coil operates for only the first thirty seconds after the set has been turned on. Thereafter it remains off until the set has once again been turned off, and the thermistor has cooled down.
The opposite effect is also used in the so-called “life extenders” for ordinary filament type light bulbs. Again, the thermistor is wired in series with the lamp, thermistor is wired in series with the lamp, but an NTC type is used, instead. When the lamp is first turned on, the thermistor is at room temperature, and its resistance is high. Thus the current through the lamp is relatively low. As self-heating occurs in the thermistor, its resistance falls, and more current flows through the lamp. Eventually, the thermistor comes up to temperature, and the current through the lamp is at its maximum. The small voltage drop that remains across the thermìstor helps by reducing the drop across the lamp. Lamp life is greatly extended by operating it at a slightly reduced voltage, and by avoiding the sudden “thermal shock” when it is first turned on.
Thermistors are made of various materials, such as strontium carbonate, titanium dioxide, maganese oxide, iron oxide, or even uranium oxide. These are heated in a furnace until they disintegrate into flakes. The flakes are then cooled, mixed with other materials, and fired under pressure to form ceramics. Semiconductor materials are used as well. Generally, they are made into beads, rods, washers, and other shapes that make them easier to attach to objects or fitted into threaded housings. In applications where thermal shunting is a problem, thin-film thermistors are also made. Again, processes similar to those used in making ICs are employed.
Solid State Sensors
Soon after the transistor was invented it was found that the forward-biased PN junction has a negative temperature coefficient. In other words, as the temperature increases, the voltage decreases. Even in ordinary solid-state diodes, this is true. This is equally true of the base-emitter junction of a bipolar (PNP or NPN) transistor. Anyone who has studied basic solid state theory has learned that diodes and separate base emitter junctions are sometimes used to stabilize the bias in transistor amplifiers.
The voltage across the base-emitter junction of a small-signal transistor such as the MPS-A20 or the 2N2222 decreases at a rate of about 2.5 mV per degree C. This makes them good temperature sensors for simple electronic thermometers, as shown in Figure 8. The potentiometer is set to provide about 0.6V so that when the sensor is placed in an ice bath, the meter reads 32 deg. F. This calibrates the thermometer well enough for use as a simple indoor or outdoor electronic thermometer. These were very popular in the 70’s.
Since the temperature coefficient is very linear in some IC bipolars, semiconductor manufacturers have developed good temperature-sensing ICS. Some produce a voltage that rises at a rate of l0mV per degree C. or F., while others produce a current that rises at luA per degree C.
One the first widely available temperature sensor ICs was National Semiconductor’s LM335. The IC is operated as though it were a Zener diode, as shown in Figure 9. Its output is proportional to the Kelvin scale. This scale uses the same divisions as in the Celsius (centigrade) scale. But in the Kelvin scale zero is absolute zero (273.16 deg. C.) Accordingly, the voltage drop across the device in Fig. 9 is 273mV at 0 deg. C. Fahrenheit or Celsius signals can be obtained, simply by using the proper amplification factors and offset voltages in circuits like Figure 10.
Among their many uses, these sensors can also be used as compensators for the reference junction temperatures of thermocouples. A voltage is produced which is proportional to the difference between the junction’s output voltage at the present temperature and the voltage the junction would produce at 0 deg. C. The sensor is mounted on an “isothermal” (same-temperature) block along with reference junction and the other voltmeter lead. The newer LM34 and LM35 sensors produce Fahrenheit and Celsius calibrated voltages directly and do not require the 273 mV offset needed with the LM335.
These and other devices are also used in digital thermostats for controlling the HVAC in homes and offices. The user can program different temperature settings for different times of the day. saving on energy costs.
Using a circuit similar to the VBE bandgap reference described in the previous article on IC voltage references, Analog Devices AD59O produces a current that increases at l uA per degree C. The advantage to this sensor is that the current signal does not change with variations in the resistance of the connecting wires and any thermocouple junctions that may be present in the amplifier circuit`s input. Current-mode sensors are favored in control circuits for this reason a current-mode signal can easily be converted into a voltage for an ADC with the op-amp circuit in Fig. ll.
Everyone is familiar with the fact that things like the heating elements in a toaster glow when hot. The temperature of hot objects determines the color of the light they emit. According to quantum theory, the wavelength at which the light intensity peaks is determined by the temperature of the hot object. (See Fig. 12.) The filament in a flashlight is a familiar example. With fresh batteries, the lamp filament is hotter und the light it emits is white. When the batteries weaken, the light from the filament is a reddish orange. Red light has a longer wavelength which corresponds to a lower temperature.
Silicon phototransistors are most sensitive to infrared light, making them suitable for rough temperature measurements. Other electronic infrared sensors are used to make images suitable for “night-vision” cameras. In electronic and industrial applications, they can help the user locate “hot spots” in machinery, circuit boards, and in evaluating the effectiveness of insulation in the walls and roofing of buildings.
For more exacting applications, the intensity of the light at different wavelengths is measured. Light at any wavelength is converted into heat when an object absorbs it. Although the amount of heat produced this way is very small, highly sensitive devices are made which can measure these tiny amounts of heat. Using the same techniques employed in making ICs, very small, low mass thermistors or PRTDs are fabricated. These are then mounted in cases having transparent lids to admit the light being measured. To help thermally isolate these sensors from their surroundings, they are mounted in a sealed vacuum compartment.
Often, two identical sensors are mounted side-by-side. One sensor is shielded from the light so that the difference in their temperatures is the result of the light being absorbed. A differential amplifier can then be used to amplify the difference in their resistances. A system of lenses and other optics focuses the light onto the sensors, and filters the wavelengths being measured. This kind of temperature measuring device is called a "pyrometer" Pyrometers are frequently used to determine the temperatures in such things as blast furnaces and the combustion beds of coal-fired power plants where most sensors would be destroyed by the heat.
This run-down of electronic temperature sensors should give you a better idea of the kind of technology that exists today. In your career in electronics it is likely that you will encounter these devices more than once. Their ease of use, the existence of standardized products for their application, and their compatibility with digital computers make it certain that they will find more widespread use in the future. Although we have only scratched the surface of this very broad subject we hope that we have added to your store of knowledge about electronics.