The electronic content in future automobiles is constantly rising, providing electronic IC and component manufacturers with new challenges to meet tough performance and cost requirements. We asked Scott Anderson, senior vice president, general manager, Transportation Systems at Motorola Semiconductor, for his views on how certain types of devices and designs will be influenced by the automotive environment.
ED: Automotive processors differ in bit width and environmental capabilities, depending on the application. What trends do you see in future automotive processors?
Anderson: There has been a dramatic increase in processing requirements as systems have taken advantage of both advanced engine-control strategies and the latest auto-coding tools. Automotive products must have good fault coverage with increasing logic densities. That's because drain-to-drain quiescent-current (IDDQ) testing becomes less reliable due to the high current leakage of the sub-0.2-µm transistor. The combination of high-performance design and low-power processes gives us the necessary performance increases with functionality at 150°C junction temperatures.
Performance projections from leading engine-control companies show clock speeds as high as 300 MHz are needed within the next decade to meet processing requirements. Future processor requirements will move to 200 to 1400 MIPS, up from the present 100 MIPS. Major processor performance drivers are 3D graphics for navigation, speech recognition, text to speech, Java virtual machines, and a desire to have "headroom" for unknown future applications. There will also be a need for more flash memory for embedded applications, up to 4 Mbytes by 2006. Although they're still a long way from supporting harsh automotive environments, magneto-resistive memories with infinite durability, long data-retention times, and remarkable bit densities may be the future memory "cover all" for automobiles.
ED: What will the breakdown between digital and analog functions look like in automobiles?
Anderson: The complexity of the electronic control unit (ECU) influences the partitioning between analog and digital IC functions. For large and complex ECUs, it's optimal to have several ICs, each made on a different process—typically one microcontroller IC with most of the digital logic, the nonvolatile memory (NVM), and analog-to-digital converters (ADCs), one or more mixed-signal ICs, and the sensor ICs. For very simple ECUs on the other hand, a single chip or package is the optimal solution.
ED: Will current data buses be able to keep up with higher processor demands and rising clock rates?
Anderson: Today's cars pack in multiple processors linked via standard interconnect buses like LIN for low-bandwidth signals, CAN for connecting modules, and MOST for high-end multimedia and telematics. Beginning in model year 2004, most new US models will use CAN as their standard interconnect bus between electronic modules within the car body. Finally, new additions to electronic vehicle safety systems will make it necessary to introduce a new, highly safe and secure time-triggered interconnect, such as FlexRay.
Interconnect systems will evolve into the real "backbone" of the car, where new components can be added with plug-and-play methods. This requires software to play an enhanced role in automotive electronics development and allows higher flexibility in development, production, and after-sales service support. This evolution will include remodeling of the network from a "flat" to a hierarchical structure.
A significant increase in bus throughput is necessary to keep pace with increasing processor performance. Today's 25-Mbit/s MOST and 10-Mbit/s FlexRay buses will have throughputs of 400 Mbits/s by 2004 to 2006. The CAN bus, which now runs at 125 to 250 kbits/s, can potentially reach 1-Mbit/s throughputs by then.
ED: What technologies are either in place or are needed for "drive-by-wire" systems in future cars?
Anderson: We worked closely with automaker BMW to develop a Byteflight protocol that supports passive-safety, body, and convenience functions that require fault tolerance, deterministic reliability, and interference immunity. FlexRay, to be used for brake-by-wire and steer-by-wire systems, is in development. Its protocol will address the needs of in-vehicle communications for innovative high-speed applications.