An industry-wide consortium, founded by companies that include BMW, Volkswagen, DaimlerChrysler and Philips, has driven the FlexRay standard development. Having concluded the critical portion of the standards-making activity, carmakers will begin climbing the learning curve as this technology is introduced in vehicles.
V2.1 BMW, for instance, will introduce FlexRay for active damping control as the successor to BMW's X5 model, while other car companies have indicated definitive Flex-Ray rollout plans. Although, initial features implemented using FlexRay will be limited, it will lead toward an autonomous or nearly autonomous vehicle in the future.
At first, it was envisioned as an enabling technology for automotive X-by-wire (drive-by-wire, steer-by-wire), but now other applications are adopting FlexRay technology. Currently, the primary trend is manufacturing efficiencies, safety regulations and the proliferation of convenience features.
Together, these trends raise the performance bar considerably for the electrical-control architecture and for the automobile's networking protocol. Simply put, cars of the future will require higher bandwidth buses and a deterministic protocol that guarantees fast responses for mission-critical tasks.
FlexRay's 10 Mbps bandwidth, its built-in fault-tolerance, and its deterministic nature were designed for just these reasons. FlexRay was also conceived to ac-commodate multiple network topologies and to be scalable so that it will be able to meet new engineering challenges.
EVOLUTION — NOT REVOLUTION
For the immediate future, the CAN and LIN buses that have served the industry for more than a decade, can still get the job done. It is unlikely, in fact, that FlexRay's adoption will have much effect on LIN applications such as windshield wipers and electric window controls. This is because they are localized control functions that do not need to communicate much with the rest of the vehicle and they have low data bandwidth demands.
CAN also is destined to play a role in automobile control architectures well into the future. Although, it is true that CAN falls short of accommodating the aggregated bandwidth requirements of a modern automobile, clever engineering offers real solutions, for a while at least.
These solutions include multiple, interconnected CAN networks that meet bandwidth requirements by partitioning the network and do not even attempt to put all data on the same bus. In particular, CAN benefits from being a proven, road-tested technology. In the safety-conscious automobile industry, where design changes are planned decades into the future, this is a real advantage.
BMW's use of FlexRay for damping control, for example, demonstrates a well-understood function that could have been performed by CAN. Introducing FlexRay as a straight CAN replacement is as much about learning how the new technology performs in real life situations as it is about superior performance.
Nevertheless, the progressive use of FlexRay is assured because it reduces overall complexity and offers a path for more cost-effective solutions. In the long term, its 10 Mbps bandwidth and built-in fault tolerance provides the most efficient and elegant solution for the cars of the future.
LESS COMPLEXITY, MORE COMMUNICATION
As cars add more safety and comfort features, the number of the electronic control units (ECUs) that implement these features continues to grow. There are now more than 100 ECUs in high-end automobiles and the number is likely to increase unless a new architecture is adopted. Coordinating ECU operations with a collection of CAN buses is becoming ever more complex and a serious impediment to X-by-wire progress. Even if the complexity issues could be solved, CAN lacks the deterministic and fault-tolerant aspects that are mandatory for X-by-wire.
Complexity has resulted in a strong interest in the engineering community to simplify automobile manufacturing and to decrease overall subsystem costs. The best way to do this is by reducing the number of ECUs. This means, of course, that those ECUs that remain will need to provide more services and higher performance. They will also require more bandwidth from the network because they will need to communicate more often with other ECUs. This is particularly true for safety systems.
These considerations, along with the longer-term implementation of X-by-wire, make it clear that FlexRay is succeeding and will continue to do so. The automobile industry is solidly behind FlexRay and the product roadmaps of virtually all major car manufacturers envision its integration. Semiconductor companies active in automobile electronics have FlexRay on their roadmaps as well.
In order to move FlexRay silicon development into the fast lane, Philips Semiconductors and Freescale Semiconductor have agreed to exchange FlexRay intellectual property. The ex-change of IP resulted in a common FlexRay protocol engine design that will be used in both companies' FlexRay controllers and microcontrollers to ensure interoperability of their FlexRay devices. It is this kind of cooperation that will propel FlexRay into cars as expeditiously as possible.
How will FlexRay be implemented when it gets beyond the try-out stage? An example of its use in an advanced automobile network topology is shown in Figure 1.
FlexRay can be configured in passive bus and star topologies as well as combinations of the two. Both can support the dual-channel ECUs that will result from integrating multiple system-level functions to save manufacturing costs and reduce complexity. A dual-channel architecture such as that shown in Figure 1 offers redundancy and doubles the available bandwidth. Each channel is capable of a maximum data rate of 10 Mbps.
The passive bus topology's primary advantage is that it's a familiar architecture for automobile networks and the engineers that design them, resulting in cost efficiency. The passive bus topology is useful when there is a requirement for higher bandwidth, short latency times or deterministic behavior and at the same time fault containment is not mandatory. A typical application area is a straight CAN replacement to accommodate the bandwidth requirements.
With the star topology, fault containment is fully addressed because branches of a star can be switched off selectively in case of undesired behavior. Stars can also be used as replicators in case passive bus cable length would exceed the specified limits.
In addition to its topology flexibility, FlexRay has numerous advantages over other protocols. It supports both time-triggered (deterministic) communication and event-triggered communications such as initiating a braking sequence.
FlexRay also supports multiple message-passing schemes between buses. This feature will be increasingly important as FlexRay applications spread through the vehicle and the network becomes more homogeneous. Several automakers, for example, have either implemented or proposed a network architecture that includes a gateway for all communication regardless of protocol. To communicate across protocol boundaries, the network will require support for multiple message passing options.
The adoption of FlexRay as an increasingly important part of automotive control architectures has opened a new playing field for semiconductor companies. Chips with higher performance and higher bandwidth are necessary, of course, but important strategic decisions must be made as well.
Chip companies competing for FlexRay market share must decide whether to extend the capabilities of their existing chips, which most likely were designed to support CAN or to follow a platform strategy and create a true FlexRay-specific family.
There are essentially two silicon components that need to be considered, including the transceiver, which moves data on and off the bus, and the logic, which includes basic microcontroller functions and communications control specific to FlexRay.
In the past, transceivers were relatively simple, but FlexRay's topology flexibility and implementation of event-triggered or time-triggered operating modes have serious implications for transceivers. Following are a selection of the implications:
- handling 10 Mbps data rates;
- supporting FlexRay node and active star topologies;
- delivering power management capabilities for ECU efficiency;
- integrating two dedicated control inputs for time- and event-driven modes;
- supporting local and remote wake-up capability;
- providing error detection;
- qualifying under the automotive environment's strict specifications (in the area of ESD and EMC, for example, while respecting all of the aspects mentioned above).
Situated between the transceiver and the basic control functions implemented by the host microcontroller is a control function specific to communications and the FlexRay protocol. There are two ways to handle this function, including implementing it on a communications controller IC or integrating it into the MCU architecture, which requires adopting a platform strategy for FlexRay.
The latter approach has advantages in performance, reliability and cost. Integrating the host MCU and the communications controller on a single chip reduces part count, component costs and engineering time, but there are other benefits, such as permitting a shared memory architecture. When the host microcontroller and the communications control do not require separate memories, designers can wring more performance out of the chip even as they are reducing system costs. Another benefit of integrating the FlexRay controller into the MCU architecture is that partitioning can be implemented in hardware or software. This makes the FlexRay implementation flexible and scalable as was intended by the founders of the FlexRay consortium.
The industry's decision to adopt a flexible approach when drafting the FlexRay standard has important implications when implementing a FlexRay MCU in silicon.
First of all, a considerable amount of computing horsepower is required. Most 8- or 16-bit MCUs that are common in automotive applications today will run out of steam, so a 32-bit MCU architecture is appropriate.
As previously mentioned, the in-vehicle network will be heterogeneous for the foreseeable future. Therefore, in addition to supporting the FlexRay V2.1 specification, the MCU most suitable for gateway applications must integrate multiple LIN 2.0 and CAN 2.0B controllers and be prepared for media-oriented system transport (MOST) bus and Ethernet.
Power conservation will have a high priority as automobiles use electronics and software more and more for product differentiation. MCUs that have the ability to adjust the power consumption to the actual need by providing a broad spectrum of power modes or by selectively switching on and off peripherals are strongly preferred because memory requirements will increase, making a unified memory architecture more important from the perspectives of component cost and performance.
Philips SJA2510 FlexRay microcontroller fits this profile well, with a 32-bit ARM968E core running up to 80 MHz, a fully integrated FlexRay V2.1 dual-channel controller, six CAN controllers, eight LIN master controllers and multiple advanced power modes (Figure 2). The SJA2510 began sampling in November 2005. Philips is also ahead of the competition in adopting a platform strategy, integrating the communications controller and implementing a shared memory architecture.
In addition, the company's TJA1080 FlexRay transceiver is an excellent match for the transceiver characteristics previously mentioned. It is the only transceiver that is capable of operating both in passive bus and in star mode.
Now that a solid spec is in place and some FlexRay-based systems are about to be deployed, automobile and semiconductor companies are already thinking of X-by-wire.
Safety is the top priority consideration in deploying X-by-wire and among FlexRay's specifications will include a bus guardian that monitors access to the bus so that data not lost or corrupted during transmission.
The bus guardian concept is still in the process of being defined by the consortium. Concepts for both a central and a decentralized bus guardian are under consideration. When the specification is complete, it will be put onto silicon, but for the FlexRay applications slated to appear in 2006, a bus guardian is not required.
In the meantime, we will see more and more convenience, control and safety functions performed by FlexRay. By the time X-by-wire arrives, FlexRay will be a well-tested, road-proven technology.
ABOUT THE AUTHOR
Toni Versluijs is business development manager at NXP Semiconductors Automotive Business Line, based in Nijmegen, the Netherlands. His responsibilities include 32-bit automotive microcontrollers and all FlexRay products. Prior to his current role, Versluijs held various positions within NXP Semiconductors' marketing and sales organization. He holds a masters degree in physics and a professional doctorate in engineering in the area of logistics management systems.