Making sense of sensors

The use of sensors in products has had a profound effect on the way many engineering problems are solved. Not only do sensors enable totally new applications – such as automated character recognition or gas-based leak detection – they can also optimise the efficiency of pre-existing processes. The modern car is a great example of this. The inputs from dozens of sensors are integrated to improve control of fuel combustion and help maintain stability in difficult driving conditions. Despite the possibilities that sensors offer, their successful application can be fraught with problems. Practical sensor technology requires a technique for measuring the physical phenomenon of interest; a design that ensures the chosen technology will work and continue to work in real-world conditions; and a method for integrating the sensor’s outputs into the rest of the system. For engineers who find themselves without a basic sensing technology to suit their needs, help is available from a fresh source. Regional Development Agency Advantage West Midlands has joined forces with defence and security technology company Qinetiq to provide access to sensor technology and funding to develop new commercial opportunities. The Advanced Sensors Project attempts to help businesses adopt and commercialise new sensor technologies, with the help of Qinetiq’s intellectual property base and development experience in sensor technologies. According to Christine Cambridge, director of the project at Qinetiq, the aim of the project is a straightforward one. “There is a high level of risk in adopting any brand new technology and trying to get it to market,” she says. “The project is designed to help companies minimise that risk by taking new technologies forward to the point where they can obtain commercial funding.” Through the scheme, Qinetiq aims to find around 300 new ideas for sensor applications. These ideas, selected on the basis that the problem matches the skill set of the technology company, will undergo rigorous market and technology evaluation. The result of this process may be, says Cambridge, that a pre-existing sensor technology is identified that will solve the customer’s problem. If the problem can’t be solved, it will be put forward for possible further funding. Around 15 projects will be selected to enter a technical feasibility phase, worth up to £250,000 and managed by Qinetiq. Matter of integration John Golby, who heads up sensor capabilities at technology company Sagentia, believes the key challenge in sensor use often comes from the integration process, rather than the choice of sensor technology itself. Golby is responsible for developing and commercialising Sagentia’s sensor technologies, while the company as a whole offers consultancy services to a wide range of clients – which can often involve helping them to develop and integrate sensor technologies into particular applications. “The first question a design engineer should ask themselves when thinking about a sensing application is ‘what is the value of my application?’” says Golby. “This gives you an indication as to whether you are likely to be looking at an off-the-shelf product or a bespoke solution.” The pressure to go bespoke comes at both ends of the market: in low-volume, high-cost applications where the particular demands of the problem may drive a need for a customised solution; and in low-cost, high-volume applications where a pared-down, application-specific sensor might be the only way to deliver effective performance at the right price. Between these extremes, says Golby, the chances are that discussion with sensor manufacturers will lead to something off-the-shelf. “It costs around £50,000 to develop customised packaging for an automotive sensor, while, if you need an Asic (application specific integrated circuit), you may be talking about £125,000,” says Golby. “If your application won’t support those sorts of costs, you are probably better off going for a standard unit.” Whether standard or bespoke, modern technology has a host of advantages for the sensor user. At the heart of many of these is cheap computer power. “The ability to do quick complex calculations for just a few pence has made a big difference to the way sensors are designed,” he adds. “In the past, you needed a huge amount of effort to achieve a linear output signal. Today, as long as you know the output signal is repeatable, it doesn’t matter what shape it is – you can use signal processing to give you the output you want.” Cheap computer power also helps to make sensors more robust by allowing them to include self-diagnostic features and to report any problems to the control system, allowing for safe, quick intervention. The right formula Sectors that have difficult environments to deal with adopt bespoke solutions as a matter of course and a whole industry has evolved specifically to serve them. Christchurch-based Active Sensors was set up 14 years ago to meet the demands of the motor sport industry for off-the-shelf and bespoke linear and angular position sensors, used in telemetry systems and controllers in all forms of motor sport – from private enthusiasts to the rarefied heights of Formula One. “Our business is all about fast turnaround,” explains operations manager Ian Miles. “It is common for us to design and manufacture a sensor to a new specification in only two weeks.” Like Sagentia, Active Sensors has seen a move to higher levels of electronic sophistication and increasing integration. While its basic products still use linear and angular potentiometers, with wipers moving across an electrically resistive base material, more advanced units are increasingly contactless LVDT designs, which require external conditioning to deliver a usable signal. The trade-off is better accuracy and a longer life. “With contactless designs, you may pay twice as much for a sensor, but it will probably last 10 times longer,” says Miles. The latest variants of Active Sensors LVDT systems now have full, integrated signal condition electronics to reduce the number of components that motor sport designers need to package, and to simplify assembly, wiring and maintenance. Another company with extensive experience of sensor application in the demanding motor sport environment is Norfolk-based Beru F1 Systems. Simon Roberts, the company’s strain gauge systems manager, is in charge of developing instrumented components to monitor stresses in the racing environment. “Extreme demands are placed on all motor sport components, including high temperatures, extraneous forces and vibration. These demands are a serious threat to load cell accuracy, but high accuracy and repeatability is still demanded for consistent race car set-up,” he says. “The effects of these extreme demands can be minimised, if they are taken into account when the design process is in its very infancy.” Beru F1 makes extensive use of finite element analysis techniques when designing instrumented parts, both to ensure that load cells and strain gauges are positioned properly to record the right loads, and to maximise signal quality by concentrating the flow of stress into the area where the strain gauges are bonded. Roberts and his team also develop novel techniques for environmental protection. The weakest link In a world of cheap, accurate, self-diagnosing sensors, the network connecting them can become the weakest link. Sagentia’s Golby says: “More sensors can mean that the amount of wiring goes up dramatically – which has knock-on effects for assembly costs and maintenance.” The use of bus systems to minimise wiring, or the adoption of wireless networking technologies, varies widely by industry, he notes, “The automotive sector uses bus systems extensively, but they are far from widespread in aircraft and almost unheard of in white goods.” Using a large number of sensors in an application can create other challenges, too. While a rich picture of the environment is a good thing, as that picture becomes more complex, it will inevitably include more noise and uncertainty. Cutting-edge software algorithms use Bayesian statistics and inferential calculus to make ‘sensible’ decisions in an environment of complex, noisy data. Perhaps there is room for sensor integration to go a step further, particularly in environments where cost is at a premium, Golby says. “Most of our clients ask for a sensor with a simple – often analogue – output signal. With a modern sensor, that might mean taking the original signal, digitising it in an Asic, processing it to get the desired measurement and then re-converting it to a varying voltage.” With some customers, Golby and his team are exploring an idea of a radical piece of simplification. “The customer almost always has spare computing capacity in the controller, so we are now talking about embedding the sensor signal processing software directly in the central control unit – and allowing that to process the raw signal for the sensor. If designers can only think a bit differently about the way different parts of their systems come together, the potential for savings is considerable.”