In the past, systems in cars were autonomous. A brake failure meant a brake system fault, and a steering failure was a steering system problem.

Not so anymore. A key fob that won't open the door could be related to various other systems, like antilock brakes. As electronics increasingly populate cars, in-vehicle communications networks are reaching the their limits of speed and reliability.

"There have been examples of a fault in one system triggered by an issue elsewhere," says Chris Webber, vice president of the automotive practice at research and consulting firm Strategy Analytics. "Carmakers have found that they get a certain number of reliability issues coming from having all those modules connected together."

Auto recalls due to hardware and software glitches have resulted in financial losses and brand name stigma. In 2002, BMW recalled 15,000 7 Series cars equipped with iDrive, a master dashboard gizmo that controls hundreds of car functions. The iDrive may have been responsible for a 2003 incident in which Thailand's finance minister had to be freed from his BMW by a guard with a sledgehammer after the car suddenly stalled, locking the doors and windows.

The car industry sees one answer with FlexRay, the highly touted next-generation network protocol that debuted quietly this year in the BMW X5 SUV to support a suspension application.

FlexRay provides bandwidth that's 20 times higher than the existing Controller Area Network (CAN) communications protocol and offers the predictability and fault tolerance that advanced electronic controls require. It also has built-in redundancy, running two parallel channels at 10Mbps each.

"If one fails, there are no consequences," says Uwe Schumann, manager of automotive software architecture at STMicroelectronics.

CAN, by comparison, lacks fault tolerance and has a relatively low data rate, up to 1Mbps, making it inadequate for higher level electronics. Moreover, its messaging procedure (lower priority messages will always be delayed by higher priority messages) could cause unresponsive controls no matter how many times you push the button.

Developed in the 1980s, before the era of car electronics, the CAN protocol can take some blame for electronic system vulnerability, especially as more controls, sensors, actuators and software fire huge amounts of data through the vehicle, Webber says.

FlexRay, therefore, is important as an enabler of advanced car electronics. Safety systems, on-board diagnostics, in-car entertainment, "black boxes" and whatever else can be crammed into a car now have a far more robust communications bus to run on.

"The whole scope of what is achievable in a vehicle is expanded," says Andrew Lee, an analyst at Frost & Sullivan.

FlexRay is also expected to deliver the missing element of high reliability. Its time-triggered architecture guarantees that all electronic signals will go through to critical systems. As a deterministic bus, FlexRay also creates a network that is predictable, regardless of outside influences.

Auto OEMs also see FlexRay as the car's communications backbone, tying all the existing car networks together, says Klaus Lange, head of the networks and diagnostics division at Volkswagen.

The FlexRay protocol was conceived by European car manufacturers BMW and DaimlerChrysler. In 2000, they joined forces with NXP (formerly Philips Semiconductor) and Freescale Semiconductor (then part of Motorola) to start the FlexRay consortium. Now all the world's carmakers and key systems and component suppliers are on board. Even Austria-based TTTech, developer of the Time-Triggered Protocol, which was a rival to FlexRay, eventually signed on to the consortium.

A lot hinges on FlexRay, which ideally benefits all. For the consumer, FlexRay promises to deliver cars that are more safe, reliable and feature-packed than in the past. For the semiconductor manufacturer, it opens up opportunities by acting as a bigger Christmas tree on which to hang more car electronics.

But big question marks remain. Consumers will have to feel comfortable relying on cars to parallel park themselves, for example. Certain critical functions may require multiple backup systems. "As a consumer, how much will you ultimately trust computers in the car to carry out functions?" asks Lee.

Moreover, the technology is relegated to high-end cars right now because it is too expensive for volume carmakers with low margins, which are sticking to traditional CAN bus systems. "Ultimately, the consumer has to pay for this technology," Lee says.

The unanimous global support for FlexRay highlights the urgency the auto industry places on ensuring that in-vehicle systems work right the first time.

Joachim Langenwalter, director of the automotive network unit at Mentor Graphics, points out that cars can be ordered with some 30,000 different combinations of electronics-related features, from electric seat adjustment and audio systems to climate and cruise control. "With so many combinations, testing is a huge challenge," he says.

Traditional methods include testing to correctness, which entails using a measurement tool to see if the system performs to specifications. The problem is that faults are discovered too late to make changes before the car is ready.

"You'll find out something is not working correctly only when you have it loaded onto an electronic control unit [ECU] and test it in driving scenarios," Langenwalter says. "If this is several months or years after the first design, you will not be able to test out possible combinations."

Testing to correctness is not enough , so testing continues after the consumer buys the car. Langenwalter points to the 2004 recall of 680,000 Mercedes-Benz E-Class vehicles that was related to an electronic brake-by-wire system (see story on page 48). "The problems were discovered much too late," he says. "Recall costs reached the original development costs of that model."

FlexRay could play a role in fault prevention after the car is on the road, but Langenwalter believes guaranteeing effective network design at the design phase is key. Specifications are plugged into the network design software tool to analyze if the specific function is reliable.

"The software can measure if this is good communication or not and tell which node is a problem," he says. Once software is integrated in the ECU, it's tested to see if it meets the original spec. Suppliers then meet the original spec, which is an easier task than testing thousands of combinations.

The AUTOSAR initiative, formed by carmakers and software developers, is also aimed at providing better test controls by creating a standardized software platform for cars.

"Manufacturers have underestimated problems that can occur," says STMicroelectronics' Schumann. "A unified software structure in the car will increase the quality of the system dramatically."

AUTOSAR is a library of pretested and preapproved software components from which a supplier can select.

FlexRay and AUTOSAR complement each other, says Frost & Sullivan's Lee. "For new electronics [enabled by FlexRay], you need more software applications," he says. "AUTOSAR allows them to redevelop the software to be used in a FlexRay system and get it to market a lot quicker than in the past."

FlexRay is welcomed by the semiconductor industry, which stands to benefit, Lee says. Semiconductor content will increase. To ensure the safety integrity level in critical car systems, more than one microprocessor is required. A 32-bit microprocessor could be used for the primary system and a 16- or 8-bit one for backup.

Freescale and NXP have produced microcontrollers with embedded FlexRay. Renesas Technology is sampling microcontrollers with a FlexRay controller and bus driver. The company is also developing a system-in-package FlexRay device integrating a microcontroller and driver, says Paul Fox, automotive marketing director at Renesas Technology America.

A veil of secrecy surrounds FlexRay road maps because automakers don't want to divulge their plans to competitors. Volvo and BMW would not comment on FlexRay, and the FlexRay consortium itself does not talk to the press, referring inquires to its Web site, www.flexray.com. Volkswagen has a timeline for adoption of the new protocol but will not disclose details, Lange says.

FlexRay-enabled vehicles will enter the market in a few high-end European models to support advanced vehicle chassis applications, Strategy Analytics' Webber says. For example, BMW announced it will introduce FlexRay in its 7 Series in 2008, and Audi plans to use it in its A8 model in 2009.

The Japanese are also followers of FlexRay. Toyota, expected to be the world's largest carmaker next year, plans a prototype car using the new protocol.

The low-end car will have to wait. FlexRay's cost per node is too expensive for volume carmakers with low margins, and financial troubles at General Motors and other automakers have crimped spending, Webber says. Strategy Analytics says FlexRay will remain at the premium end until 2010. "FlexRay is here to stay," Webber says. "But beyond the luxury segment it is really cost-dependent."