Temperature sensors play a critical role in automotive applications such as climate control and engine monitoring. Opportunities to improve performance by moving to new technologies – without paying a price premium – can be missed in the rush to finish other parts of a design.
Traditional sensors such as thermistors, resistance temperature detectors (RTDs) and thermocouples, have analog outputs, hence their outputs need digitising before they can be used in a digital-control loop. Alternatively, temperature sensors can be fabricated in silicon that achieve unbeatable reliability because the sensor's behaviour is as stable as the silicon itself. Silicon-based sensors experience very little drift over a long time period (more than 50 years). Meanwhile, using the precise manufacturing techniques that characterise semiconductors make the sensors highly reproducible.
Being able to take advantage of the packaging and high volume manufacturing disciplines developed for integrated circuits is particularly important for sensors that are used in today's automotive applications because miniaturisation and encapsulation are such dominant trends. Silicon temperature sensors also have a positive temperature coefficient, which provides fail-safe operation as the device's resistance rises with temperature.
One means of manufacturing a stable, highly linear, long-lived silicon temperature sensor is based on the spreading resistance principle (Figure 1).
Figure 1. “Spreading resistance” devices provide a conical current distribution.
The chip size is approx. 500 x 500 x 240m. The upper plane of the chip is covered by a SiO2 insulation layer, in which a metallised hole with a diameter of approx. 20 m has been cut out. The entire bottom plane is metallised.
This arrangement provides a conical current distribution through the crystal, hence the name 'spreading resistance.' A major advantage of this arrangement is that the dependency of the sensor resistance on manufacturing tolerances is significantly reduced.
The dominant part of the resistance is determined by the area close to the metallisation hole, which makes the setup independent of the Si crystal dimension tolerances. An n+ region, diffused into the crystal beneath the metallisation reduces barrier-layer effects at the metal-semiconductor junctions.
This configuration is polarity dependent, however, and requires a package with radial leads. Sometimes that can cause problems when the sensor is installed because its polarity is not always evident. To successfully address this issue, two single sensors can be connected in series but with opposite polarity. With this configuration, the sensor's resistance is independent of current direction.
The single-sensor arrangement, however, still offers advantages in some applications. Its simplicity allows sensors to be produced in the compact SOD68; DO-34 package, for example. Another important advantage is its ability to operate at temperatures up to 300°C, rather than a more usual 150°C of silicon sensors. This can be accomplished if the single-sensor device is biased with its metal contact positive. Increased maximum temperature is due to the fact that a positive voltage on the gold contact severely depletes the hole concentration in the upper n+ diffusion layer, effectively insulating holes spontaneously generated within the body of the crystal. As a result the holes are prevented from contributing to the total current, and hence from affecting the resistance.
Spreading resistance technology is the basis for NXP Semiconductors' KTY series of silicon temperature sensors. They provide accurate temperature measurements because they display a virtually linear temperature coefficient over their entire range (Figure 2).
Figure 2. Linearity of spreading resistance sensors (NXP Semiconductors' KTY 81/82)
Resistance over temperature can be calculated with the type-dependent constants A and B. A linearisation resistor can be easily added where further linearisation is required.
As the temperature coefficient is positive, the sensors exhibit fail-safe operation when a system overheats. Furthermore silicon is inherently stable, so KTY sensors are not only reliable, but have long operational lifetimes.
Stress testing of KTY temperature sensors has verified that when sensors are pushed to operate near their maximum temperature for 10,000 hours, they exhibit typical drifts of 0.2K. These sensors are typically used at half of the specified maximum temperature, however, and extrapolating the test data under these real world conditions predicts that low drift will be experience for 450,000 hours (51 years).
NXP offers automotive design engineers one of widest choices of silicon temperature sensors available today. The product families are specified according to package, nominal resistance, tolerance and operating range.
The KTY81 and KTY82 families use a twin-sensor technology for polarity-independent sensing. Available in a variety of packages, including hermetically sealed glass for use in fluids such oil or water, they can operate at temperatures up to 300 C and are ideal for use in exhaust and heating systems.
In conclusion, the KTY series of silicon sensors provide an attractive alternative to more conventional sensors based on NTC or PTC technology.
Author profile: Guenter Reiniger is marketing manager for the Sensor, Safety and Comfort Market Sector Team, Automotive Business Unit, NXP.