Of all physical quantities, temperature is the most commonly measured. There are temperature sensors all around me—in my oven, refrigerator, water heater, and wall thermostat. Even the furnace has its own internal temperature sensor. Given their ubiquity, you'd think it would be easy to find a good, but really low-priced one for various kinds of embedded or lab measurements.
One of the tricks to making a sensor is to find a readily measurable phenomenon that varies consistently as a function of the quantity you want to sense. Because nearly everything varies to some extent with temperature, and because it is a simple matter to accurately monitor electrical properties—voltage, current, resistance, etc.—phenomena that change these properties are good candidates for sensing mechanisms. For example, the voltage generated across copper and steel nails inserted into a lemon is probably temperature dependent. We will leave verification of this as an exercise for the reader.
If you don't want to pursue fruit-based sensor technology, you can investigate the component racks (now hidden in drawers) at Radio Shack. Although electronics manufacturers go to great lengths to ensure that their parts don't fluctuate with varying temperatures, some dependencies are unavoidable. One of the best-known is a diode's forward voltage drop, which is about 0.6–0.8 V at room temperature and varies roughly –2 mV/°C. One candidate device I found in a parts drawer was the 1N914 silicon diode (see Figure 1). It's small, it's cheap (a pack of 50 for $2.59), and it's rated for operation from –40°C to 150°C, convenient for many kinds of temperature measurement. One particular diode showed 0.611 V of drop at 1 mA of bias current at room temperature (~20°C).
To properly bias a diode for use as a temperature sensor requires a constant-current source. Because the diode voltage will vary so little (~400 mV) over its operating temperature range, a simple almost-constant current bias circuit can be implemented with a 12 V power supply and a single 10 kΩ resistor (see Figure 2). This results in a nominal room-temperature bias current of 1.14 mA that can be expected to vary ~3% over the operating temperature range.
This bare-bones interface circuit, although simple, does have a few shortcomings. The first is calibration—every diode is going to be somewhat different, both in roomtemperature voltage and sensitivity. The second problem is that a –2 mV/°C change riding on 600 mV is difficult to interpret from a glance at the voltmeter. One solution to both snags is to build a simple front-end amplifier for adjusting the temperature signal's gain and offset. You can make this with an op amp and a few resistors, some of which are variable (see Figure 3). Note the inverting amplifier configuration—it is convenient for the voltage to go up with rising temperature, as opposed to going down. By using a 10 kΩ pot for span adjust, you can achieve sensitivities up to 20 mV/°C.
Even though the parts are inexpensive and the design is simple, there are many ways to improve this interface circuit. While separate adjustments are provided for span and offset, there is some interdependence that requires you to adjust span first and offset afterward. You also need a well-regulated 12 V current to bias the offset adjustment and prevent offset drift. The biggest obstacle might be having to calibrate span and offset for each diode you want to use. Calibration must be performed at no fewer that two temperatures, so calibrating more than a few diodes can turn into a chore.
Next month we will look at another affordable temperature sensor that is more convenient from a calibration standpoint.
Ed Ramsden, B.S.E.E., a member of the Sensors Editorial Advisory Board, designs sensors for the heavy-truck industry in Portland, OR; [email protected].