The editors of the CAN Newsletter have followed the discussion about whether Ethernet will make field bus systems obsolete or not. Some engineers seem to think that one protocol must be the "better" one and will thus supersede all the others. More important for the CAN Newsletter is the relation between CAN and Ethernet. The editors have addressed some experts on this topic.

Where do you see the strengths of CAN as opposed to industrial Ethernet?

Prof. Dr. Konrad Etsch-berger (Ixxat Automation): The CAN data transmission protocol comprises a simple but highly sophisticated layer-2 multimaster protocol, which beside the collision-free bus arbitration scheme provides message-priority-oriented data transmission, a powerful error-recovery mechanism and a very high error detection capability. Also Ethernet represents a simple layer-2 multimaster protocol. Without an additional higher layer protocol the bus access implies possible bus collisions, a node-oriented message transmission scheme and only moderate error detection mechanism.
Since Ethernet does not provide error recovery, Ethernet-based networks require an additional higher layer protocol (such as TCP) to provide error-recovery. The CAN and Ethernet protocols are completely integrated in the respective communication controllers as stand-alone chips or as part of a micro-controller. Due to the only moderate resources (processing power, interrupt frequency, memory) required by CAN, this can be implemented already on very low-cost micro-controllers. CAN therefore allows the implementation of very low-cost communication systems. Due to the much higher data rate of Ethernet, many more resources are required. Ethernet communication controllers therefore are mainly found on powerful and more expensive 32-bit micro-controllers.
The application of CAN-based communication is most appropriate in applications with event-oriented transmission of short messages, up to some thousand messages per second, latency times no shorter than a few milliseconds and a network extension up to some hundred meters. With CANopen and DeviceNet, worldwide-accepted CAN-based comprehensive higher layer protocols, device and application profiles are available for distributed automation systems.

Martin Jetter (Jetter): The obvious strengths of CAN are its low costs and its handy size. It is these strengths that make CAN especially apt for combining with smaller and less cost-intensive devices. In this area, CAN is an interesting counterpart of Ethernet. Although Jetter AG applies Ethernet company-wide, our controllers have nevertheless been equipped with a system bus the functions of which are transparent to the user. This system bus is CAN-based, which makes standardized distributed solutions possible without needing further field bus nodes.

Martin Rostan (EtherCAT Technology Group): Compared with generic industrial Ethernet, CAN is very cost effective, well suited for small data packets such as I/O and drives, and ideal for networking both embedded systems and typical field bus applications. However, when you take EtherCAT into account, some of these advantages are not so unique any more...

Viktor Schiffer (Rockwell Automation): CAN is optimized for messages up to 8 byte. These short messages are transmitted with much higher protocol efficiency over CAN as compared to Ethernet. Another advantage lies in the much simpler physical layer. Due to the lower speed and the different bus access mechanism, CAN networks are built with a simple passive physical layer that is much easier to set up than Ethernet.

Would you like to read more? Please subscribe to the CAN Newsletter. It's free!