The Future of CAN / CANopen and the Industrial Ethernet Challenge

by Wilfried Voss, President esd electronics, Inc USA

Industrial Ethernet technologies are a formidable challenge to CANopen as the low-cost

industrial networking technology of choice. Ethernet technologies will eventually replace
the majority of CANopen applications, at least in regards to new developments.

For many years, Controller Area Network (CAN) and CANopen, a higher-layer protocol
based on CAN, represented the best choice for low-cost industrial embedded networking.
However, since the official introduction of CAN in 1986, there has been a quest to
replace CAN and CANopen to overcome the most obvious shortcomings such as limited
baud rate and limited network length.

Industrial Ethernet technologies are currently the most formidable challenge to CANopen
as the low-cost industrial networking technology of choice. Ethernet technologies will
eventually replace the majority of CANopen applications, at least in regards to new
developments, starting at this very moment in certain areas such as industrial control
including motion control and, especially, robotics. Ironically, CAN - the underlying
hardware layer of CANopen - has a far greater lifetime expectancy in the North American
market than CANopen as a higher layer protocol.

However, there can be too much of a good thing, and that is definitely the case when it
comes to Ethernet-based fieldbus technologies. There are currently more than 20 different
industrial Ethernet solutions available, all with their distinctive advantages and
disadvantages, making a pro/contra decision difficult. The major question, besides the
technical aspect, is which of these technologies will survive in the market, and how do
they support the current need for control components. In all consequence, the battle
between these technologies is a mere marketing battle.

This article will elaborate on the future of CAN and CANopen in the North American
market. It will also address the need for faster fieldbus systems and narrow the field of
potential winners in the Ethernet race.

Distributed vs. Central Control

Over the years, CAN and CANopen, due to their economical implementation combined
with an extremely high level of reliability, have proven and established their dominance
as a fieldbus system for embedded solutions. They were, however, never intended to
effectively penetrate the huge North American market for simple industrial I/O, namely
the PLC market.

Conventional industrial controls, especially PLC, are based on Central Control where a
single CPU is responsible to control the application tasks, resulting in high performance
requirements for the CPU. Another downside is the great amount of wiring resulting in a
considerable impact on service and maintenance.
 Central Control
(Distribut vs Central Control 1.jpg)

Distributed Control
(Distribut vs Central Control 2.jpg)

Established serial communication technologies for embedded solutions such as RS-485,

CAN, TCP/IP, etc. support a Distributed Control architecture in form of Master/Slave or
Client/Server configuration. The most obvious advantages are the saving in wiring plus
the increased performance of a multi-processor system. There are more benefits resulting
from the savings in wiring such as increased system reliability, improved service and
maintenance, and, most importantly, reduced downtime and reduced operating costs.

The established Ethernet TCP/IP protocol, theoretically 100 times faster than CAN or
CANopen, never had the potential to be fully accepted as an industrial protocol, because
it is non-deterministic, i.e. it does not support real-time control. Consequently, Ethernet
TCP/IP was never a threat to existing CAN and CANopen applications. New Ethernet
technologies, such as EtherCAT, Powerlink, Ethernet/IP, etc., however, do have the
potential not only to replace existing fieldbus systems, but also to play a major role in the
North American PLC market.

Advantages and Disadvantages of CAN and CANopen

CAN and CANopen, used as fieldbus systems for embedded solutions, combine a number
of advantages that cannot be matched by Ethernet TCP/IP. They are:
 • Extreme Reliability and Robustness
• No Message Collision
• Very Low Resource Requirements
• Low-Cost Implementation
• Designed for Real-Time Applications
• Very Short Error Recovery Time
• Support of Device Profiles (CANopen only)

However, there are some disadvantages of using CAN and CANopen, the biggest being
the limited network length (~120 feet at a 1 MBit/sec baud rate). The disadvantages are:

• Limited network length (depending on baud rate)

• Limited baud rate of 1 MBit/sec
• Limited bandwidth

The following picture demonstrates the relation between baud rate and supported network
length:

CAN Bus 4.jpg

CANopen is basically a software add-on to provide network management function to

CAN. The side effect is a reduced CAN bandwidth. The degree of bandwidth loss
depends primarily on the use of Service Data Objects (SDO) and Process Data Objects
(PDO). Only a meticulous “housekeeping” can guarantee the best possible performance.

In all fairness, the limited bandwidth is not a major problem, since CANopen handles
only the communication means between multiple processors (nodes); the major control
tasks take place within the nodes, and they do not necessarily effect the bus
communication.
Houston, we have a solution…

(Uhm… We’re looking for the problem…) In all consequence, the current excitement
about new Ethernet technologies is based on a mere marketing hype. The matter of the
fact remains that more than 95% of all industrial control applications are sufficiently
covered by technologies that were established many years ago. It is nevertheless also the
case that these new Ethernet technologies are developing steadily closer toward the lower
price level of conventional components, and “why buy a compact car when you can get a
large hybrid SUV with four-wheel drive and better gas mileage for nearly the same
price?”

While it is true that the costs for Ethernet technologies show a tendency to lower prices,
the statement “nearly the same price” can only be regarded as another marketing
statement. A tough reality check will reveal that the costs are currently several times
higher than those of a conventional system such as CANopen (the estimated factor is
somewhere between 3 and 20).

However, the technological advantages of Ethernet based fieldbus systems, whether

needed or not, are compelling, and they contribute to a higher acceptance level
worldwide. The advantages are:

• Speed! (Currently 100 MBit/sec)

• Unlimited network length
• Use of established HW components (RJ-45, Switches, Hubs, etc.)
• Increased system performance

Note: There is currently no information on how Ethernet based fieldbus technologies

compare to the high level of reliability and robustness as provided by Controller Area
Network.

The Future of CAN and CANopen

The future of CAN - as the physical layer - and CANopen - as a higher layer protocol
based on CAN - in the North American market must be seen separately.

The use of Controller Area Network is still dominated by its vast use in the automobile
industry, and there are no indications that CAN will be replaced in short-term. Another
stronghold is the use as a physical layer for the SAE J1939 protocol, and CAN will
remain the most cost-sensitive fieldbus solution for small, embedded systems.
 CAN Nodes Sold in Millions
Source: CAN-in-Automation
(CAN Nodes Sold.jpg)

In summary, the use of CAN will continue in:

• Automobiles and Trucks

• Aerospace (e.g. satellites)
• SAE J1939
• Small Embedded Solutions
• Legacy Applications

CANopen, however, is facing a much tougher battle, since its major application range is
now being attacked by the new Ethernet technologies. These CANopen legacy
applications are:

• Motion Control
• Industrial Machine Control

Other applications not necessarily effected by Ethernet technologies are:

• Niche Applications (Lifts, Escalators, Gambling Machines, Telescopes, Specialty

Vehicle Systems, Cost-Sensitive I/O Control, etc.)
• MilCAN

The MilCAN standard, mainly used by the British Army for their military vehicles,
allows CANopen and SAE J1939 devices to access the same bus.
The Medical industry is still the biggest supporter of CANopen in the North American
market, but here, too, are tendencies to look into Ethernet technologies due to the vastly
increased data throughput that these technologies offer.

The Ethernet Challenge

There are currently more than 20 different industrial Ethernet solutions available in the
marketplace, all with their distinctive advantages and disadvantages, making a pro/contra
decision for the end-user difficult. The major question, besides the technical aspect, is
which of these technologies will survive in the market, and how do they support the
current need for control components. As a matter of fact, the battle between these
technologies is not necessarily based on technological advantages; it is a mere marketing
battle.

Out of the 20+ Ethernet based industrial protocols remains a much shorter list of
candidates that realistically have a chance to penetrate the American market significantly.
They are:

• EtherCAT
• Powerlink
• Ethernet/IP
• Modbus/TCP
• ProfiNet

The following picture demonstrates the differences between these systems in an

overview:

Ethernet Protocols Comparison.jpg

The picture also demonstrates the major difference between Ethernet/IP and
Modbus/TCP on one side and EtherCAT, Powerlink, and ProfiNet on the other.
Realistically speaking, neither Ethernet/IP nor Modbus/TCP is deterministic (i.e. not
suited for hard real-time control); they do not support a direct connection between the
physical layer and the application.
Especially Modbus/TCP – basically Modbus over Ethernet - with a cycle time between 5
and 10 milli-seconds must be considered too slow for real-time control. It is nevertheless
extremely easy to implement, which explains its wide usage.

While Ethernet/IP may not be the optimum choice in regards to real-time control, it is
nevertheless the strongest player in the North American market due to the “political”
support by Rockwell Automation. The same is true for ProfiNet (supported by Siemens).
A downside for ProfiNet may be the high performance requirements by the standard,
making an implementation for third-party vendors somewhat difficult.

There are only three protocols that provide a direct hardware connection between the
physical layer and the application, i.e. they require a specific controller chip. These
protocols are:

• EtherCAT
• Sercos III (limited presence in North America)
• ProfiNet V3 (current status unknown)

EtherCAT and Powerlink also support CANopen device profiles to provide a certain level
of portability for existing CANopen applications.

The current problem with the Powerlink standard seems to be the transition from
Powerlink V1 (a B&R proprietary version; B&R is the creator of the Powerlink protocol)
to the Powerlink V2 standard as agreed by the members of the Ethernet Powerlink
Specification Group (EPSG), and there are strong indications that the support for the
Powerlink standard is waning.

The following provides an overview of the various Ethernet protocols.

EtherCAT Ethernet/IP Powerlink Modbus/TCP

Vendor EtherCAT Open Ethernet Modbus-IDA
Organization Technology DeviceNet Powerlink Group
Group Vendor Specification
Organization Group
Homepage www.ethercat.org www.odva.org www.ethernet- www.modbus-
powerlink.org ida.org
Availability of Members signing Free Members Free
Specification an NDA
Availability of Example Code, Example Code Standard Example Code
Technology ASIC, FPGA Ethernet Chips
Products 2003 2000 2001 1999
Available Since
Interaction Master/Slave Client/Server Master/Slave Client/Server
Structure
Communication One frame for all Message Message Message
Method communication Oriented Oriented Oriented
 partners
Ether Data 100 MBit/sec 100/10 100 MBit/sec 100/10
Transfer Rate MBit/sec MBit/sec
Physical Line, Daisy, Star Star Star, Tree
Topology Chain, Tree
Logical Open Ring Bus Bus Ring Bus
Topology
Infrastructure Switches Switches Hubs, no Hubs, Switches
Components between different (Hubs are switches
segments possible, but
not efficient)
Device Profiles CANopen, DeviceNet, CANopen None
SERCOS ControlNet

Note:
• Ethernet/IP and Modbus/TCP are open to increased speed standards, i.e. there
would be no need for hardware upgrades. All others would be subject to hardware
changes.
• The use of switches and/or hubs increases wiring costs and lowers system
reliability.
• The use of hubs may impact real-time performance (Does not apply to Powerlink
due to timing control preventing message collision).