Temperature is a parameter that often needs measuring. It may be within an enclosure, on a PCB, remotely or even “wirelessly”. Wireless, in this case, does not necessarily mean a remote sensor communicating using radio frequencies, it can mean a pyroelectric sensor. Pyroelectric sensors are very long wavelength detectors which measure radiation in the 6-15μm range approximately. Bear in mind visible light is 0.4-0.7μm, and you can appreciate the wavelength for pyroelectric sensors is very long. With suitable optics, they can be used to measure at a distance of several feet or more, depending on the size of the target you are measuring. The accuracy is not as good as most other methods of temperature measurement and depends on the emissivity of the source you are measuring. However, it is a genuinely non-contact method and so can measure temperature where other methods are impractical.
Pyroelectric sensors are also the sensor in passive infra-red intruder detectors such as the Murata IRA-E and IRA-S series detectors. Rather than actually measuring temperature in an intruder detector, they rely on the change in temperature as a heat source (e.g. a person) moves across zones created by a Fresnel lens in front of the sensor.
A more common temperature measurement method is a temperature sensing IC. There are plenty of different types around from ones with an analog output such as the Texas Instruments LMT84, to digital output and serial output devices such as the MCP9800/1 from Microchip with I2C interface. The MCP9800 can be purchased with one of 8 I2C addresses whereas the MCP9801 has address pins so that you can select the on the PCB.
Semiconductor sensors usually rely on a “proportional to absolute temperature” (PTAT) sensing method. This is based on the fundamental diode/bipolar transistor equations where for a constant current, Vbe is proportional to kT/q where k is Boltzmann’s constant, T is absolute temperature in Kelvin and q is the electron charge. While voltages from newer devices might be scaled to a useful range, older devices such as the LM334 “current source” have an output that would roughly reach 0V at 0K if the output graph was extrapolated that low.
Semiconductor sensors are fine if the temperature you want to measure is within the range of a typical semiconductor device e.g. -40C to +125C but if you want to measure outside that range you need to look for an alternative sensor. A thermocouple is one option.
Thermocouples rely on the Seebeck effect whereby a junction of dissimilar metals produces a small voltage dependent on temperature. Specific combinations of metals have been categorized to produce the familiar type K, T, and E thermocouples as well as others. The different combinations are chosen for their temperature range, linearity and sensitivity with type K being the most common (chromel-alumel). Something like platinum-palladium would have low sensitivity and poor linearity but are appropriate for some applications. Even the more linear thermocouple types such as type K are not perfectly linear but can be considered linear over a limited temperature range depending on the accuracy required. Linear Technology Application Note 28 has some useful comparisons such as the graph below showing the non-linearity for various types.
When you consider how to wire a thermocouple into your measurement system, you will find that you end up with at least two junctions of dissimilar metals and most likely three. You have the actual thermocouple junction then two other unwanted thermocouples where the thermocouple wires connect to copper, tin or some other metal at you enclosure connector or PCB. The temperature you measure will be the difference between the thermocouple temperature and the temperature of the other two unwanted thermocouples (the “cold junction”). You need to ensure the two unwanted thermocouples are at the same temperature (usually they are close together anyway) and measure that temperature. Your measured thermocouple temperature is then relative to the cold junction temperature. If the cold junction and thermocouple are at the same temperature, then the measured voltage would be zero. Bear in mind the output of a thermocouple is typically less than 50μV/C.
So, a thermocouple system typically requires a separate temperature measurement to calculate the thermocouple temperature accurately. There are ICs to help with this problem such as the LTC2983 from Linear Technology or the Microchip MCP9600. The LTC2983 covers a wide range of temperature sensor types, and for thermocouples, it includes cold junction compensation as well as linearization. The MCP9600 (below) is specifically for thermocouples but also does cold junction compensation and linearization.
These sorts of devices make thermocouple interfacing fairly easy, but the LTC2983 is not cheap — around $28. If you only wanted a thermocouple interface, then the MCP9600 is a lot cheaper.
Thermistors are well established resistive devices which can have negative or positive temperature coefficients (NTC or PTC). They are usually highly non-linear and more suited to alarm or temperature limit functions but can be used to measure temperature. They are usually specified by the resistance at 25C and that resistance can be fairly accurate. This image shows the variation in resistance for the Panasonic ERTJ NTCs.
While it may not look very non-linear, bear in mind that the Y axis is logarithmic and the X axis is linear.
PTCs are used mostly for overcurrent protection rather than temperature measurement. They are also very non-linear. Their use as overcurrent protection relies on self-heating due to current elevating the thermistor temperature, so in turn, the resistance increases rapidly and prevents the current climbing higher. The graph below shows the characteristics of some Vishay PTCCL devices.
As you can see, the resistance increases significantly above 125C. Also, this is on a lin-log scale, so the resistance rise is even more dramatic than it looks on this graph.
A platinum resistance thermometer (PRT) is another resistive sensor but a lot more accurate and linear than NTC and PTC sensors. They are not perfectly linear but pretty good over a reasonable range. The most common is the PRT100 which has 100 ohms resistance at 0C. The resistance varies by approximately 0.385Ω/C although there are slightly different standards for the platinum, so you need to check with your supplier. As they are simply a resistor you would usually include them in a bridge type measuring circuit and if the sensor is remote, you should use a 3- or 4-wire sensing circuit for accuracy. The PRT1000 has a higher nominal resistance and so will be less affected by cable resistance if used in a 2-wire measuring system.