Drive-by-wire revs 'zero defect' reliability push

Austin, Texas — As carmakers push "by-wire" braking and steering control further out on their road maps, automotive suppliers are steering hard toward the zero-defect reliability levels that car companies say are needed before drivers will feel confident handing the keys over to the electronics.

To be sure, electronically controlled braking and steering offer real gains in performance and fuel savings — but automakers want to know they'll be able to sell consumers on drive-by-wire's safety. That has led to automotive-industry initiatives and a broad engineering push to improve component- and system-level reliability.

As the Convergence 2004 show gets under way today in Detroit, Freescale Semiconductor Inc. and Texas Instruments Inc. are among the companies expected to announce plans for silicon that supports the FlexRay Consortium's approach to drive-by-wire hardware and software.

The FlexRay standard defines a fully deterministic, time-triggered communications system for safety-critical applications, including the fail-safe modes that ensure that a driver can "limp home" even if the electronics fail.

Freescale will demonstrate engineering-prototype modules that support the FlexRay specification. Freescale "is really close, from a timing standpoint, to implementing the FlexRay standard and getting silicon out there," said Peter Schulmeyer, director of strategy and marketing at Freescale's automotive products group.

The "gating factor for FlexRay-compliant silicon," said TI marketing manager Matthias Poppel, "is the FlexRay conformance testing specification" that will guarantee a supplier's silicon meets the spec. The conformance suite is expected in the fourth quarter of 2005, he said, at which time TI will release the silicon now under development at the company.

"No one can begin to really sell FlexRay-compliant silicon until that compliance software is ready," Poppel said.

The carmakers are doing "huge amounts" of R&D on by-wire technology, he said, but "from our discussions with customers, the carmakers had a lot of practical problems with the electromechanical braking systems they have tried to bring out to the market. That has caused delays with exclusively electronically controlled steering or braking."

Error correction techniques, redundant CPUs, better testing algorithms and the fail-safe modes in the FlexRay modules all appeal to the engineering community. But drivers want to know that the joystick that might someday replace the hefty hydraulic steering column is safe. That will require a selling campaign to consumers, probably starting with those that can afford the S-class sedans from Mercedes Benz and the 7-series autos from BMW, sources said.

Meanwhile, companies are moving closer to the zero-defect goal set by the Automotive Electronics Council (AEC).

"There is a huge drive in the industry to make a quantum jump to zero defects," said Ray Cornyn, operations manager at Freescale's transportation products group. "But there is always a finite chance that a component will fail, and that is why the FlexRay specification includes fault-tolerant or fail-safe modes. "

As electronic components have become more complex, testing has become more extensive. Cornyn said Freescale engineers take "snapshots" of their customers' code to run exhaustive verification cycles of Freescale's engine controllers, for example.

And error correction techniques for flash and SRAM bit cells have become more sophisticated.

Some companies are adopting "parts average testing" (PAT), endorsed by the AEC, as a means of reaching the zero-defects goal. "Oftentimes, some change in the process within the factory can be the root cause of later failures," said Jason Cole, automotive business development manager for the Americas at TI. "Parts average testing treats all minor changes as major."

Dallas startup Pintail Technologies Inc. is working with TI, STMicroelectronics and other suppliers to implement the PAT approach, said Taylor Scanlon, Pintail's CEO. Other companies are working on in-house PAT solutions aimed at meeting the AEC's zero-defect standard.

The goal is early detection of the faults and process problems that can lead to failures out on the highway. By comparing one wafer lot with the next, PAT can help identify "outliers" — parts that meet a target spec but lie at the edge of the quality requirements. Often, it is such parts that later fail. Outliers can provide hints about weaknesses in a design or process, Scanlon said. Weeding them out early can help a component maker meet the higher reliability standard mandated by the AEC's Q100 standard for drive-by-wire components.

Pintail's approach can be applied both at the wafer-probe level and during packaged-part testing, Scanlon said.

Zero-defects reliability is no less a concern for the hybrid vehicles that have proved a hit with consumers. "If a 50-cent battery controller fails, that is going to stop a hybrid car in its tracks, and the Japanese companies just can't let that happen," said Mark Gary, TI's business development manager for the Japanese automotive market.

TI supplies Toyota and other hybrid-vehicle manufacturers with battery control silicon and other mixed-signal ICs. Before the company makes any changes to its process, exhaustive testing is required to make sure the process tweak doesn't result in parts failures on the road, Gary said.

He noted that the AEC's quality standards take a two-tiered approach, aimed at stricter testing of both the parts and the circuit boards used in the drive-by-wire subsystems.

For military vehicles, TI's Cole said, manufacturers are employing drive-by-wire approaches with redundancy at the system level — a hydraulic backup to the electronic control system — in case of an electrical failure. Such an approach also may be used in commercial automobiles and trucks at some point, but for now the cost is an inhibiting factor, Cole said.

Paul Grimme, general manager of Freescale Semiconductor's transportation and standard products group, said some car designers are looking to the FlexRay standard initially for the much higher bandwidth that it offers.

FlexRay offers 10 times more bandwidth than CAN-based networks. And FlexRay's approach to specific time slots means that a vehicle's network can be split into smaller pieces, with fewer gateways, each having more bandwidth into the control modules.

"FlexRay won't come in at lightning speed. But all the European car designers are working on by-wire braking," he said, with pilot vehicles being tested today for possible introduction in 2006. That will require dual-core processors so that a spare CPU can kick in if one core fails, an approach also being adopted for engine controllers.

Freescale's Schulmeyer believes steer-by-wire will see widespread adoption before brake-by-wire. "We might see mechanical backups at first. If the electronics fail, then the mechanical controls would take over," he said, predicting there will be a transitional phase for both electronically controlled steering and by-wire braking.

The TI managers said electronically controlled power steering saves about 200 pounds of weight, which results in a half-mile-per-gallon improvement in fuel efficiency. By 2008 or later, the by-wire approach will extend to other parts of the car, they predicted.

"I don't think we will see by-wire coming in during the next two or three years, but it will happen in seven to 10 years, for sure," said TI's Poppel. "The trend toward by-wire is just being pushed out; it is not being reversed."