IMAGE:NICHOLAS EVELEIGH; PHOTO MANIPULATION: LAURA HOFFMAN

Chances are, your health and happiness rely on sensors, those ubiquitous little devices that tell us if a fridge is too cold, a nuclear reactor's safety systems are operating, or a factory production line is processing components correctly. But sensors have a dirty little secret: it's all too easy for them be in perfect working order, reporting all is well when, in fact, your milk is turning into a frozen block, the reactor's safety system is impotent, and that factory has filled a warehouse with useless—and possibly dangerous—products.

Fortunately, help is on the way with a new standard for analog sensors, the most common kind in use today. The dirty little secret of sensors is calibration, the process by which data from a sensor are mapped to real-world conditions, and the new standard should help make miscalibration a thing of the past. Miscalibrated sensors can cause problems ranging in severity from a wasted morning's research to what happened at the Bruce B nuclear generating station near Toronto in 2002. There it was discovered that a backup reactor shutdown system that had been operating for weeks, in what appeared to be working order, was actually incapable of catching a dangerous rise in radiation, owing to an incorrectly calibrated neutron detector.

Like most standards, the new standard goes by an unlovely name—in this case, IEEE 1451.4. But 1451.4 marks a huge advance in sensor technology and is already being applied in research and industrial laboratories. This new standard marries the tried-and-true robustness and cost-effectiveness of analog sensors with the intelligence of digital equipment. Now, what does that mean in practice? A lot of things—in the long term, one of the most important aspects of 1451.4 is that it offers a standard interface and protocol by which a sensor can describe itself over a network. With the advent and adoption of intelligent networked and wireless sensors, the notion of self-identifying devices may seem fairly elementary, but this has taken more than a decade to happen with analog sensors. Most commercially available sensor networks today are based on proprietary communications protocols, limiting their usefulness and hampering their adoption. IEEE 1451.4 could change all that.

We'll return to the long-term promise of 1451.4, but for now let's stay with the immediate problem of calibration. IEEE 1451.4 will eradicate one of the most common sources of sensor errors today: incorrectly transcribed calibration information from sensor data sheets. To understand how these errors arise and why they're such a big problem, look at how sensors are traditionally used. Analog sensors typically output a voltage that is proportional to the magnitude of whatever it is they've been designed to measure—be it temperature, pressure, or something more exotic.

Analog sensors persist in a digital world because they are cheap, extremely reliable, and rugged. Simply put, they can take a beating that would damage or destroy a digital sensor that can output a number describing the measured quantity directly. And punishing environments—such as the inside of a car engine or the depths of an oil well—are often exactly the places where we most want to put sensors.

Although analog sensors are easy to hook up to a data acquisition system that monitors output voltage, converting that voltage back into degrees Celsius, pascals, or whatever else, is trickier. In other words, when a temperature sensor registers 2.5 volts, we need to know how to translate that voltage into the actual temperature, be it 100 °C, 50 °C, or 2500 °C [see illustration, "Curve Confusion"].

Until now, the only way to know was by looking up the sensor's data sheet, a manufacturer-supplied document that details how to calibrate the sensor correctly. Someone has to enter information from this sheet into the data acquisition system, which is usually based on a personal computer. A single moment of human error here can make a sensor—and all the data it gathers—worse than worthless.