As IC device dimensions shrink and heat management and dissipation become tougher-than-ever challenges, one simply cannot overestimate the importance of sensing IC temperature. In particular, temperature sensing has become ubiquitous, playing key roles in process-control, environmental, test-and-measurement, and communications applications. In addition, its use in electronic circuit design continues to expand throughout large-volume automotive, medical, and consumer applications.
Thermocouples, resistance temperature detectors (RTDs), and thermistors are the most common temperature sensors for electronic circuits (Fig. 1). Mostly used in industrial measurement and control, they produce analog signal outputs that must be digitized before they’re placed into computer circuits. Semiconductor IC sensors, another common technology, dominate pc boards. They’re available with analog as well as digital outputs. Besides their relatively smaller physical size, they’re much more amenable for use in digital electronic circuitry, since they can be fabricated on the same piece of silicon where other electronic functions reside (see “Typical Characteristics Of Contact Temperature Sensors” at www.electronicdesign.com, Drill Deeper 18912).
“Choosing the right sensor type and understanding its thermodynamic properties as they relate to its mounting and use are the two critical issues that determine the success or failure of the temperature sensor’s application,” says Jim Williams, senior scientist at Linear Technology Corp.
“Packaging and mounting of a sensor are crucial, no matter what type of sensor is used. The most common problem designers encounter is properly mounting the sensor to the measurand for optimal thermal performance,” he says.
THERMOCOUPLES
Thermocouples are the most commonly used devices for temperature measurement. They operate via the Seebeck Effect in which two dissimilar metals, welded or joined together at one end, produce a voltage output at the two open ends of the metals for a given temperature. That temperature is measured at the point where the two metals are joined. A thermocouple’s output voltage increases as the temperature rises.
Widely used in industrial applications, their ruggedness, accuracy, and very wide temperature range are key attributes. “We still see a lot of thermocouples in heating, ventilation, and air-conditioning (HVAC) applications, an area where semiconductor IC temperature sensors are trying to compete,” says Susan Pratt, an applications engineer with Analog Devices.
Thermocouples are flexible—they can be constructed in just about any manner and from many materials to suit any application. They feature many advantages over other temperature sensor types.
For example, they’re very rugged, inexpensive, and highly responsive. They don’t require any excitation source. And best of all, they feature the broadest temperature range of all contact- type sensors.
Type J or iron-constantan (constantan is alloy of copper and nickel) thermocouples are the most widely used devices for thermocouple calibration. Other popular versions include types B, E, K, R, S, T, and N. The Instrument Society of America (ISA) compiled the standard ISA thermocouple calibration table.
However, thermocouples are “tip” sensitive, measuring temperature at a very small point of reference. Their outputs are also quite nonlinear, which means they require external linearization in the form of cold-junction compensation. Cold-junction compensation is crucial if accurate temperature measurements are needed.
Also, the thermocouple output voltages are quite low, in the tens to hundreds of microvolts, requiring careful wiring layout techniques to minimize noise and drift. One way to reduce noise is to place resistors in series with the thermocouple and a capacitor across the thermocouple leads to form a filter.
One common mistake with thermocouples is to use copper from the thermocouple connection to the measurement device. This introduces another thermocouple in the measurement process.
RTDS
For the most accuracy and stability, try an RTD. Most RTDs use platinum (in wire or film form) wound on a small ceramic tube, though some are made from nickel, a nickel/iron alloy, or copper.
Also, RTDs are very stable and offer fairly good linear outputs. A platinum RTD can be thermally shocked from boiling water to liquid nitrogen (–195°C) 50 times with a resulting error of less than 0.02°C. Typical RTD stabilities are on the order of ±0.5°C/year. But they do require some linearization circuitry, typically via a lookup table in a microcontroller, to correct for some nonlinearities.
RTDs are more expensive than thermistors and thermocouples, though. They require a current source to operate (a current that causes self-heating). And, they feature a low resistance-value change to temperature change, as an RTD might change by just 0.1 O in response to a 1°C change in temperature.
Because RTDs are self-heating devices, measurement inaccuracies can occur if the RTD self-heats under the test current. In general, currents should be kept to 1 mA or less. Self-heating errors can also be reduced by using an extremely low bias current or a 10% duty-cycle current instead of a constant bias. Too low a bias, however, introduces some noise that can affect the RTD’s measurements. Nonetheless, designers can minimize this noise by using differential, ungrounded, and shielded RTDs.
Continue to page 2
Reprints
