Smart sensors

Citation

Bogue, R.W. (2004), "Smart sensors", Sensor Review, Vol. 24 No. 2. https://doi.org/10.1108/sr.2004.08724baa.002

Smart sensors

Smart sensors

Robert W. Bogueis an Associate Editor of Sensor Review

Keywords: Smart sensor, Microprocessor, Electronics

It is particularly timely to be considering “smart” sensors in this issue of Sensor Review, as it was just over 20 years back, in 1983, that Honeywell launched its “ST 3000” range of smart pressure transmitters, arguably the world's first family of smart sensors. However, the term “smart” has since been applied to numerous classes of sensors and as is so often the way in the sensor industry, the term has been applied somewhat indiscriminately. Ten years back, Professor Julian Gardner, currently head of the Warwick University Sensor Research Laboratory, stated “There are differing views at present over the definition of a smart sensor” (Gardner, 1994), and this is as true today as it was then. Some authorities argue that for a sensor to be “smart”, the signal processing electronics must be an integral part of the sensing element, which implies that only sensors based on silicon technology can be smart. This requirement eliminates most of the sensors that are now deemed to be smart and it is perhaps better to take a more pragmatic approach and think of a smart sensor in terms of its functionality. A workable definition is a sensor whose directly associated electronics can undertake logic functions and communicate with the host system or other devices, irrespective of what form the electronics takes.

Implicit in this definition is the use of microprocessors or perhaps programmable logic devices and although the research community had been considering the application of the microprocessor to sensor technology, ever since Intel's launch of its “4004” in 1971, the Honeywell sensors took the process instrumentation industry by surprise. During that period some industry commentators doubted whether the capabilities offered by these products could justify their (then) higher costs. However, the benefits of “smart” soon became clear: the sensors could be addressed remotely by connecting a hand-held communicator to any point in the control loop; rangeability was greatly increased; thermal coefficients were reduced; accuracy was increased; digital output options were offered; and most importantly, ownership costs fell. This arose principally from the remote electronic access which, according to Honeywell, allowed 96 per cent of the cost of re-ranging a differential pressure sensor during installation, start-up and operation to be eliminated. Further, it was claimed that the improved stability, accuracy and temperature performance reduced calibration costs by 67 per cent. As it was, Honeywell's bold move set a trend that was to dominate all future generations of process sensors and analysers and within just 5 years of the launch of the ST 3000, all major process instrumentation manufacturers had followed this trend and the markets for smart sensing products in Europe alone had reached over US$ 160 million per annum (Frost and Sullivan, 1989). Growth has been dramatic and by 2001, this figure had reached almost US$ 1 billion.

The dramatic advances in, and falling cost of, digital electronic technologies have allowed smart functionality to become the norm across many sectors of the sensor industry. Features such as self-fault diagnosis, auto-ranging, remote calibration, digital outputs, data storage and active compensation for temperature and other cross-sensitive factors are now a commonplace. Further, the availability of sensors with digital outputs has progressively driven the move away from the long established industry standard 4-20 mA analogue control systems to digital data highways. In addition, a modern smart process transmitter is capable of digital communication over an existing 4-20 mA loop using protocols such as HART. Another benefit of having real processing power at the sensor is the ability to exploit inherently non-linear sensing phenomena or those requiring significant signal processing; good examples being Coriolis effect mass flow sensors and gas sensor arrays.

The original Honeywell sensors were developed in the era of “286” series processors which had around 120,000 transistors; today's “Pentium 4” contains 42 million, giving access to almost unlimited processing power. This opens-up all manner of advanced sensing capabilities which are only limited by the ingenuity of the sensor technologists.