Image Processing Reference
In-Depth Information
the software of the generating task) can slow down and even block the whole network. Such a kind
of failures cannot be detected by the fault confinement unit embedded in CAN chips, as they do
not depend on physical faults but are due to logical errors. Although bus guardians can be used to
overcome these errors, they cannot be implemented efficiently for conventional “event-driven” CAN.
15.4.3 Solutions
The drawbacks highlighted in Section .., despite serious, are not going to rule out CAN com-
pletely. In many cases, it is (and will remain for several years) the solution of choice to implement
cheap and robust networked embedded systems (including those used in automotive applications).
Among the solutions conceived to enhance the behavior of CAN there is the so-called time-
triggered CAN (TTCAN) protocol [ISO], for which network controllers are already available. By
adopting a common clock and a time-triggered approach, it achieves very-low jitters and provides a
fully deterministic behavior.
Concerning fault tolerance, several interesting proposals have appeared in the literature that can
be effectively adopted in real-world systems. For example, in [PRO] a novel architecture for CAN,
which is based on a star topology, is described. Unlike the conventional bus topology, such an arrange-
ment helps avoiding partitioning faults (i.e., a broken link only affects one node and does not prevent
the others from communicating).
Finally, it is worth remarking that little can be actually done in order to improve the network
performance. Despite some solutions have appeared in the literature—such as, for example, Wide-
CAN [CEN] that provides higher bit rates while still relying on the conventional CAN arbitration
technique—their interest is mainly theoretical.
15.5 Time-Triggered CAN
The TTCAN protocol was introduced by Bosch in with the aim of making CAN suitable for the
new needs of the automotive industry, and, in particular, to satisfy (at least in part) the requirements
of the upcoming x-by-wire systems. However, TTCAN can be profitably used in all those applica-
tions characterized by tight timing requirements that demand for a strictly deterministic behavior. In
TTCAN, in fact, it is possible to decide exactly the point in time when messages will be exchanged,
irrespective of the network load. Moreover, composability is noticeably improved with respect to
CAN, so that it is possible to split a system in the design phase into several subsystems that can be
developed and tested separately.
The TTCAN specification is now stable and has been standardized by ISO [ISO]. The main
reason that led to the definition of TTCAN was the need to provide improved determinism in com-
munication while maintaining the highest degree of compatibility with existing CAN devices and
development tools. In this way, noticeable savings in investments for the communication technology
canbeachieved.
15.5.1 Main Features
One of the most appealing features of TTCAN is that it allows both event-driven and time-triggered
operations to coexist in the same network at the same time. In order to ease a graceful migration from
CAN, TTCAN foresees two levels of implementation that are known as level and , respectively.
Level implements only basic time-triggered communication over CAN. Level , which is a proper
extension of level , also offers a means for maintaining a global system time across the whole network,
irrespective of tolerances and drifts of local oscillators. his enables accurate synchronization up to
the application level, and hence true time-triggered operations can take place in the system.
