September 1, 1998

Cars and complexity 

Graham Prophet, EDN Europe Editor In Chief

Today's typical automobile is more "environmentally friendly" than its predecessors of just a few years ago. Its growing electronics content plays a major part in such efficiency. Similarly, engine control units run power plants at higher efficiencies and with greatly reduced emissions, and more efficient control of transmissions further boosts the mileage obtained from a tank of fuel. You can argue, too, that if GPS navigation systems become widespread, they will also reduce the negative environmental impact of cars by getting us to our destinations more directly—thus using less fuel.

All the same, we shouldn't minimise the contribution of all of the other engineering disciplines involved. Better mechanical design and aerodynamics contribute to efficiency, and chemistry plays a big part, too. Obviously better fuels, oils, coatings, and paints extend the efficient working life of a typical vehicle. After all, a significant part of a vehicle's overall environmental impact is the energy and raw material cost of building and maintaining it.

But there's a downside to all of this electronic "progress"; the newest car models are very complex, and although remarkably reliable, they do have a significant failure rate. Consider the following scenario: You own a car for a few years and are tens of thousands of kilometres into its lifespan. Although the car still runs efficiently and hasn't yet corroded, it suddenly develops an obscure and intermittent fault that your dealer has trouble pinning down. Somewhere in its dozens of microprocessors and sensors, something has gone wrong.

When dealing with electronics, many car mechanics have no ability (or inclination) to diagnose beyond the major subsystem level. Even worse, each of those subsystems, as replacement parts, can be very expensive. Now, all of a sudden, you realise that your few-years-old vehicle, still in fine condition, is in fact economically unrepairable. It can only be scrapped and replaced, which is not at all an environmentally friendly outcome.

So, what are we designers to do? It turns out that our industry may already have a number of necessary solutions available in its bag of tricks. The problem is one of limited fault observability in a very complex system—rather like some of the problems that IC designers have been wrestling with for some time. It's one part of the thinking behind Hewlett-Packard's decision to create an entire division, based in Lyons, France, dedicated to serving the needs of the automotive industry. With it, HP plans to sell carmakers everything from LEDs to test gear, including test methodologies. Expect to see many of the same techniques that have evolved in IC testing: fault simulations run against vehicle models' subsystems (maybe even of the complete vehicle, but that's a tall order); fault dictionaries; and probabilistic analysis in which the mechanic gives the diagnosis system the fault symptoms and the system generates a list of the possible causes, thereby providing a degree of likelihood for each known symptom.

If you talk to automotive electronics designers, you'll soon realise that some of them believe that the solution lies in layering on yet more complexity—such as built-in self-testing or even redundancy and self-repairing mechanisms within electronic subsystems. As an increasing number of industrial, commercial, and domestic systems follow the path of escalating complexity, that sort of thinking may have to be applied much more widely. Or, to stand the argument on its head, there's nothing totally unique about chip design. We are entering an era in which very complex system design will be pervasive, and the techniques available to ensure that such design remains manageable are just starting to emerge.

You can reach EDN Europe Editor In Chief Graham Prophet at +44 118 935 1670 or graham.prophet@rbi.co.uk.

EDN Access | Feedback | Table of Contents |