Selecting the Right Temperature Sensor
The most important question to ask when selecting a temperature sensor and instrumentation system is "What needs to be measured?" A simple question, but it can be surprisingly easy to answer incorrectly.
Design requirements will dictate the choice of temperature sensor and instrumentation. Not all applications require the same choice, and even within an application, different temperature sensors will be required. Selecting the appropriate sensor requires prioritizing the most important design attributes. Some attributes are not exclusive to others: The most stable sensors also have a very slow response rate and can be very expensive. Design requirements can be divided into:
- Quality of measurement.
- Environmental constraints.
- Utility requirements.
Quality of measurement includes accuracy, repeatability and resolution. In many situations, accuracy is not as important as resolution because processes are controlled from changes from a normal value and not the absolute temperature. Repeatability and reproducibility are important when comparing changes to operations over a period of time.
Environmental constraints include the range of operation and the operating conditions to which the sensor is exposed. This can include corrosive or oxidizing environments or ultra-high vacuum and clean room conditions. Utility requirements are primarily costs. This includes direct costs, lifetime and installation costs, and instrumentation costs. For sensors, the direct costs are typically the most important. Installation and instrumentation costs will depend on the design requirements such as number of sensors to be measured. Other utility requirements are ease of use and reliability.
In cryogenic applications, the sensor most often is physically separated from the measurement instrument, and it is easier to consider them separate components. Instrument requirements that need to be considered include the number of sensors to monitor, reading update rate, datalogging and data transmission, and cost.
ThermocouplesThe most common temperature sensor is a thermocouple. It is small and durable and can be used from -454 to 5,432oF (-270 to 3,000oC). A thermoelectric voltage results at the junction of two dissimilar metals in the presence of a temperature gradient. In practice, there are two junctions: the sensing junction and the reference junction. The potential difference between the known reference junction and the unknown sensing junction indicates the temperature.
There are numerous combinations of metals that can be used to make thermocouples. The most common for cryogenic applications are Type E, T and K, which are chromel-constantan, copper-constantan, and chrome-alumel, respectively. Of the three, Type E has the largest sensitivity: 60 microvolt/C at ambient and 25 microvolt/K at -324oF (-198oC). While Type E is good down to -450oF (-268oC), special thermocouples made from gold-iron alloys provide larger signals below -324oF (-1 9 8oC). Chromel-AuFe has a sensitivity of 16 microvolt/C at liquid helium temperatures -450oF (-268oC). Type K is commonly used because of its wider range (-418 to 2,012oF [-250 to 1,100oC]) and its stability in an oxidizing environment.
Advantages of thermocouples are their wide range, low cost, ease of use and fast thermal response time. They are best for applications that require a simple, low accuracy temperature measurement, and for those that span a wide temperature range.
Because of the small signal (~1 mV) and sensitivity (60 microvolt/C), thermocouples are not useful for high resolution, high accuracy measurements. Typical resolutions are limited to about 0.25oC with a good voltmeter. Thermocouples also are prone to errors that affect the overall measurement accuracy. External electrical and magnetic fields (ground loops) can couple to the signal leads. With the addition of other junctions to the circuit, cold-junction errors can cause voltage changes. They are also susceptible to drift if used at high temperatures. Typically accuracies are, at best, 1 percent of absolute temperature or +/-3oC at ambient.
Resistive Temperature SensorsTemperature sensors based on the changing resistance with temperature can be classified as positive temperature coefficient (PTC) or negative temperature coefficient (NTC). Platinum RTDs are the best example of PTC resistance sensors. Other PTC-RTDs includes nickel and copper RTDs.
Thermistors are the name given to negative temperature coefficient RTDs. A PTC-RTDs has a fairly linear response; NTC-RTDs are very nonlinear but much more sensitive to change in temperature.
The most common PTC-RTD is platinum. A platinum RTD is the industry standard due to its accuracy and reproducibility over a wide temperature range as well as its interchangeability. Measurements in the range from -432.4 to 1,112oF (-258 to 600oC) are made routinely with a high degree of accuracy using platinum RTDs.
Industrial-grade platinum RTDs are wire-wound devices that are encapsulated in glass or ceramic. They also are made in thin-film elements that are smaller and faster than a wire-wound, glass-encapsulated sensor. While smaller and cheaper, thin-film sensors do not have the thermal stability or repeatability of a wire-wound device.
Platinum RTDs follow a standard response curve to within defined tolerances (DIN IEC 751). The industry stand a rd for class B accuracy is specified as +/-0.3oC and +/-0.75% variation in the specified 0.00385 C-1 temperature coefficient of resistance at 32oF (0oC). Class A accuracy is better by a factor of two. Below -321oF (-196oC), a platinum RTD is still usable but requires an individual calibration. Typically values for platinum RTD are 25, 100, 500 and 1,000 ohm where the resistance is defined at 32oF (0oC).
Like all resistors, platinum RTD can be measured by current excitation or voltage measurement. Common configurations are two-, three- and four lead measurements. Two-lead measurement does not correct for lead resistance, so it can only be used in applications where the sensor is close to a temperature transmitter.
Platinum RTDs are replacing thermocouples for many applications that require good accuracy and stability. However, RTDs' thermal stability comes at the expense of fast thermal response. Platinum RTDs are more expensive than thermocouples, and thermocouples have a faster response time and wider temperature range -- both lower and higher -- than platinum RTDs.
Thermistors are a special type of resistive temperature detectors. These NTC devices have a large, nonlinear change in resistance with change in temperature. They are made from metallic oxides -- typically iron, nickel or titanium -- and are much smaller than platinum sensors. A thermistor can be as small as 0.25 x 0.25 x 0.15 mm.
The nonlinear behavior follows an exponential rise in resistance with decreasing temperature. Devices are classified by the room temperature (77oF [25oC]) resistance value and range from 100 ohm to 100 kohm. Because of the nonlinear behavior, thermistors have a limited temperature range of operation. The lower resistance thermistors can be used for process cooling applications down to -103oF (-75oC) and up to 572oF (300oC). The sensitivity of a thermistor can be very high, capable of detecting changes in temperature less than 1 millidegree C. The small size and large sensitivity gives a thermistor a faster response time than a platinum RTD.
A thermistor does not follow a standard, universal curve like platinum RTDs. For thermometry applications, thermistors can be curve matched over their full range. These are presorted sensors, all of which are within a specified tolerance band. Thermistors that have a close resistance tolerance easily can be substituted and are considered interchangeable. Tolerance can be as good as +/-0.2 C from -58 to 302oF (-50 to 150oC).
A thermistor can be read similarly in two-, three-, or four-wire configurations. Because the sensor resistance is typically much larger than the lead resistance, a two-wire method is the most common.
Thermistors are ideal for applications that need a very high temperature resolution for a narrow temperature range. The small size lets them be adapted to a variety of probe designs, including surgical needles and direct solder mount to circuit boards. They are not as stable or accurate as platinum RTDs, but they are often much less expensive, depending on the tolerance.
Diode Temperature SensorsDiode temperature sensor is the general name for a class of semiconductor temperature sensors. They are based on the temperature dependence of the forward voltage drop in a p-n junction. The voltage change with temperature depends on the material. The most common is silicon diode, but gallium-arsenide also is sometimes used.
Silicon diodes can be used from -456 to 441oF (-271 to 227oC). From -418 to 441oF (-250 to 227oC), a silicon diode has a nearly constant sensitivity of 2.3 mV/C. Below -418oF (-250oC), the sensitivity increases and is nonlinear. The temperature response curve is shown in figure 1. Diode temperature sensors typically are mounted in a special semiconductor package (1 x 3 mm). The semiconductor packaging is robust and allows for solder mounting for probes or circuits.
Diodes sensors typically are excited with a 10 microA current. The output signal is fairly large: 0.5 V at room temperature and 1 V at -328o F (-200oC). This can be compared to platinum where a 100 ohm PRT with a 1 mA excitation is only a 100 mV signal. Like thermistors, diodes from a particular manufacturer are interchangeable, or curve matched over their whole range. They are classified into different tolerance bands with the best accuracy being +/-0.25oC from -456 to -280oF (-271 to -173oC) and +/-0.3oC from -280 to 32oF (-173 to 0oC).
The large temperature range, nearly linear sensitivity, large signal and simple instrumentation make the diode useful for applications that require a better accuracy than thermocouples. Also, because of the large signal, a diode can be used in a two-lead measurement with little lead error. Disadvantages are diodes are more expensive per sensor and diodes, while interchangeable within particular manufacturer lot, do not follow a universal curve.
InstrumentationAfter selecting the proper sensor, it needs to be installed into the application, connected to instrumentation and integrated into the monitor and control system. While it is important to match sensor requirements with the appropriate instrumentation, some design decisions are independent of the choice of sensors.
To measure temperature, the output signal from the sensor can be sent directly to a dedicated temperature monitor, or a temperature transmitter can be used.
Dedicated monitors and controllers are used as bench-top instruments for laboratory and quality control applications or process monitors integrated into stand-alone systems like cryopumps or chillers. Benchtop monitors and controllers typically have better resolution -- up to six-digit for some instruments -- and accuracy. Temperature controllers have output heaters to adjust the temperature to a programmed setpoint. Control can be on/off or full proportional-integrative-derivative (PID) control.
Panel-mount monitors also vary in capability. Most are three-digit displays without additional analog output, but others can have five-digit displays with an additional transmitter output and serial interface. Panel-mount monitors are available for all classification of sensors.
Sidebar: Know the LingoBelow are some common terms for thermometry and temperature measurement.
Accuracy: The closeness of agreement of the measured value to the true value.
Resolution: The smallest change in measured parameter that can be detected.
Repeatability: Often called precision, this is the closeness of the agreement between the results of successive measurements.
Reproducibility: The closeness of the agreement between the results of measurements under changed conditions. Changed conditions could include method of measurement (changing the same class of sensors, i.e., using a different thermocouple) or principle of measurement (using a diode sensor instead of a thermocouple).
Temperature Scales: The SI unit for temperature is the Kelvin. Depending on the temperature range and application the more common unit is degree Celsius.
Kelvin to Celsius: TC = TK - 273.15
Celsius to Fahrenheit: TF = 9/5 TC +32