iCC 2017 CAN in Automation

The role of CAN in the age of Ethernet and IOT

Christian Schlegel, HMS Industrial Networks

CAN technology was developed in the 1980s and became available in 1987, just as other
industrial fieldbus systems like PROFIBUS or INTERBUS entered the stage of industrial
communication. Beside the fact that CAN is a success in the automotive industry and
used in all types of cars today, it has also made its way in many other industrial areas.
About 15 years ago, new technologies based on Ethernet started to emerge, with ap-
pealing and sometimes outstanding features. Some six years ago Ethernet also started
to find its way into automobiles. Today, other new communication technologies are
showing up on the horizon driven by the omnipresent Industrial Internet of Things.
But even now, 30 years after their introduction, these “classic” fieldbus technologies
are still alive – with varying success. Since CAN was initially developed with a focus for
use in automobiles, CAN has certain features that still make it the best choice for many
applications in automobiles and industrial areas – even when compared to the newer
technologies.
This paper discusses why CAN is still a valid or even better choice for certain applica-
tion areas than Ethernet-based technologies, not just focusing on the advanced fea-
tures provided by the enhanced capabilities of CAN FD but also highlighting how these
applications benefit from the features of “classic” CAN.

Looking back into history … the requirement to transmit data between

… when CAN was born these ECUs but also to connect sensors
and actuators to them. Using conventional
In the early 80s, the increased use of cabling concepts would have led to a huge
electronics in all areas, especially in increase of cables and consequently of the
automation, required new communication weight of a car. There was also a need for
technologies with more bandwidth and a new way of exchanging data in a reliable
features. As a consequence some of the and secure way between these ECUs,
larger automation equipment manufacturers sensors and actuators.
started to develop new technologies.
These activities resulted in the publication As there are many diverse function areas
of several new fieldbus technologies like inside a car like motor management, breaks,
PROFIBUS, INTERBUS or Sercos in the chassis control or passenger compartment
second half of the 80s. All of them were still and certain functions are optional depending
based on a serial communication principle on the car series or the customers’ wallet,
but with considerably increased wire speed, there is no clearly structured communication
more features and – due to the typical architecture inside a car like it is in the
architecture of an automation system — automation world. At a first glance, the
all of them were based on a master/slave communication structure of a car seems
communication principle. These started to to be chaotic. In addition, the electric
replace the existing serial communication world inside a car is not as stable as in the
technologies. automation world. For example there are
many interferences from the engine which
Similar to the automation world, more disturb the communication. Moreover,
and more electronics were introduced shielding of cables has to be avoided in
in passenger cars in form of Electronic order to safeguard weight. Consequently,
Control Units (ECUs). This also imposed communication strategies and technologies

01-7
iCC 2017 CAN in Automation

from the automation world were not the message. As all CAN messages have
applicable in cars and a new approach was an inherent message priority, the most
necessary. prioritized message always wins the
arbitration.
In the early 80s, a group of people at Bosch • In combination with the error signaling
were the first to investigate the existing mechanism, CAN ensures that the data
serial communication systems regarding transmitted in the network is always
possible use in passenger cars. However, consistent in all nodes connected to the
none of these systems were able to fulfill the network. Each node which identifies an
requirements and the development of a new error in a message immediately transmits
serial communication system was started in a special error signature (error frame)
1983. Finally, in 1986, at the SAE congress destroying the currently transmitted
“Controller Area Network” was introduced to message and making sure that all other
the public. Only one year later, the first CAN nodes in the network also consider the
controller chip, the 82526 was presented by currently transmitted message to be
Intel and shortly after Philips Semiconductor erroneous and, as a consequence, reject
presented the 82C200. it. Error situations are e.g. checksum
error, message frame delimiter errors
Ingenious features made CAN different (special bits of a message always have
to have a defined value), bit stuffing
Designed for use in passenger cars, the error (after 5 consecutive bits with the
inventors gave CAN several clever and same value, an additional inverted bit
sophisticated features to which none of the is inserted into the message). An error
new upcoming industrial fieldbus systems can also be signaled by the transmitting
could compete with. node if the currently transmitted bit seen
on the network is different from the bit
In order to understand why CAN is sometimes information the node actually transmits.
a better choice compared to Ethernet, some If there is an error in a message
of the most important features need to be transmitted on the bus, the transmitting
understood: node immediately stops the transmission
• The first and most outstanding feature to and restarts a new transmission but with
mention is the bus access method. CAN re-executing the arbitration procedure. If
uses the Carrier Sense Multiple Access there is a message with higher priority at
(CSMA) principle which was not a new another node which shall be transmitted,
principle: for example, it was already the message with the higher priority will
used by Ethernet. With this principle, win the arbitration and be transmitted
all nodes of a network which want to first.
transmit a message, listen to the network • The arbitration mechanism including
and if there is no transmission, they can message prioritization and error signaling
start to transmit. This leads to collisions is only possible because of the physical
in case more than one node start to representation of a bit on the bus line. A
transmit. Consequently the messages bit is transmitted in form of a differential
are destroyed and all transmitting signal on two twisted lines, ISO11898-2
nodes have to resolve this situation (High Speed) defines a differential signal
by abandoning their transmission for a of 0V as recessive bit and a differential
random time which might lead to longer signal of 2V as a dominant bit. The
delays, especially when there are many symmetric transmission makes CAN
nodes which want to transmit data. In immune to common mode interferences
order to avoid this situation, the arbitration and the twisted pair lines compensate
mechanism of CAN avoids collisions and electromagnetic interferences. This
allows one node of all the nodes which makes CAN very resistant against
start to transmit their message at the external disturbances in general. For
same time to continue its transmission special purposes, fiber optic cables can
and all others immediately start to receive also be used.

01-8
iCC 2017 CAN in Automation

CAN and its way into industrial Looking at these application examples and
applications others from industrial CAN pioneers, CAN
was used very early in different application
The first company to use CAN in a mass areas and markets — and not only focusing
produced car was Mercedes who started to on industrial factory automation like the
use CAN in the S class in 1991. “classic” fieldbuses. For sure, this was the
result of ease of use, reliability and the
At the same time as CAN was introduced, simple concept of CAN.
also the “classic” fieldbus systems like
PROFIBUS, Sercos or INTERBUS were Because CAN was a really new technology,
being introduced, especially for industrial there was also a need to connect PCs (at that
applications. Although the major companies time running MS-DOS) to a CAN network
driving CAN were from the automotive for either running a control application or for
market and hence have been focused on running an analyzer software which allowed
passenger cars, there has been a few other, users to monitor what was happening on the
rather small companies which started actively CAN bus.
promoting CAN for industrial applications.
Some of these pioneers were. Kvaser AB, This lead to the development of first
I+ME GmbH, Softing and STZP (Steinbeis CAN products with focus on industrial
Transfer Center Process Automation, applications. Like CAN interface boards
which became IXXAT Automation GmbH in as shown in figure 2 or the very first CAN
1998 and was acquired by HMS Industrial repeater as shown in figure 3.
Networks AB in 2013). The distinctive
features of CAN were also very interesting
for industrial applications and so very early
first applications were developed using CAN
as a dedicated communication network.
These applications used an own way of
defining which data should be used and how
data was to be transmitted within the CAN
messages. Some of these first applications
were developed by STZP for customers
paving the way for CAN. These included a coil
winding machine, a machine for validating
and counting paper money, a security door
control system for banks, a CAN-controlled Figure 2: First PC CAN interface board
endoscope system or a cow milking system with ISA bus hosting the Philips 82C200
with decentralized controllers for milking, (BasicCAN) and already the Intel 82527
feeding and weighting (see fig. 1) (FullCAN) CAN controller chips (Source:
STZP / HMS)

Figure 1: CAN-controlled cow milking and Figure 3: First CAN Repeater presented at
feeding system developed 1992 (Source: the INTERKAMA fair in Dusseldorf in 1992
STZP / HMS) (Source: STZP / HMS)

01-9
iCC 2017 CAN in Automation

In order to broaden the know-how, V2 (today’s POWELRINK) in 2006.

acceptance and use of CAN, a supporting This was only possible because all
ecosystem is necessary, consisting of the companies behind these standards
product suppliers, consultants, system started working on these standards
integrators and the machine builders already by the end of the 90ies.
implementing it. As learned in the first
Keynote paper from Holger Zeltwanger, Why was everyone looking into Ethernet
this requires much more than only a as replacement of the fieldbus networks?
system and technology which allows to There are some aspects which are
transmit data. An application layer with obvious: Ethernet is used widely in
device and application profiles based office networks, public infrastructure
on commonly agreed definitions and and the Internet. Consequently Ethernet
specifications becomes indispensable. components like PHYs or Transceivers are
The work to do that started immediately produced in huge quantities which brings
when CAN in Automation was founded in the price down. Furthermore, Ethernet
1992 and some years later, the first higher brings the possibility of transmitting
layer standard for CAN became available. messages with a larger amount of data
Today CAN in Automation has the most in one message and in general a much
comprehensive set of standards for larger bandwidth due to the fact that it
applications and device profiles compared runs at 100 Mbit/s whereas CAN and
to all other industrial communication the classic fieldbuses are limited in
technologies. speed and data per message. Furthermore,
the number of nodes in an Ethernet
On the other hand, the passenger car network is almost unlimited. This is very
manufacturers have been applying a interesting especially in the area of factory
completely different approach for the use automation: here we have large machines
of CAN in cars. Each manufacturer started requiring a considerable amount of
to define an own catalog specifying for data to be transmitted for controlling
each data (or signal as they call it) which the machine. The machines can be large
ECU transmits the data, which ECUs and so the network extension becomes
receive the data, and how often the data large. Also networks between machines
shall be transmitted. Years later, additional on the factory floor interconnecting
common definitions started to became machines become more and more
standards e.g. in AUTOSAR for diagnostic important and also require a high data
purposes. capacity and long network extensions.
Finally, as machines become more and
Ethernet - A threat for CAN? more sophisticated, they impose high
demands on motion control. This results
Around the year 2000, Ethernet entered in requirements to much faster cycle times
the stage in the industrial automation and also more data to be transmitted
world. Already some years before, there in shorter times guaranteeing real-time
was a discussion about using Ethernet behavior.
in industrial automation systems. As
a consequence, the first organization This may apply for the use of Industrial
targeting common specifications for Ethernet in larger machines, fast running
applying Fast Ethernet in industrial machines and basically in the area of
automation systems was founded in factory automation. But what about
1999 — IAONA. In 2001, Rockwell applications and markets which do not have
Automation launched EtherNet/IP as a such requirements? Like small machines
first real industrial Ethernet specification (ticket machines, vending machines,
and then other standards were introduced labelling machines,…), machines in
one after the other: Powerlink V1 in medical, laboratory automation, test &
2001, EtherCAT and PROFINET in 2003, measurement, mobile machines or utility
Sercos III in 2004 and POWERLINK and farming machines (figure 4).

01-10
iCC 2017 CAN in Automation

and star topologies with short branches.

In case of a daisy chain of CAN nodes,
the CAN cable is routed in a passive
way through the CAN node (just directly
connecting the cables inside the device).
Failing of such a device will not result in
losing connection between all the other
nodes in the network.
• In automation and control applications,
the data to be transmitted typically
consists of short data packages (e.g.
command, parameter, status) with a few
bytes or even bits. Ethernet becomes
rather inefficient when only this type of
data is transmitted because the minimum
payload of an Ethernet frame is 42 bytes.
However, considering the speed of
100 Mbit/s, this is still faster than CAN with
1 Mbit/s. But a lot of performance is lost
and processing of Ethernet frames and
protocols on transmitting and receiver
side requires more performance.
• Looking at robustness, CAN has several
mechanisms directly incorporated which
makes it well-suited to handle electroma-
Figure 4: Today’s application areas of CAN gnetic interferences. Erroneous messa-
based communication networks ges are automatically repeated and its
residual error probability (probability that
Would it make sense to deploy Ethernet- a transmission error remains undetected)
based communication networks in general is several times lower than for Ethernet.
also in these areas? In Ethernet, additional software effort is
required when the capability for detecting
The answer is: Definitely not, and there are erroneous data shall be improved and
good reasons for staying with CAN as long messages with erroneous data shall be
as the available data capacity, bandwidth repeated. Since CAN operates with lower
and real-time capability of CAN meet the frequencies for data transmission and ty-
requirements of these applications: pically uses smaller microcontrollers, its
EMC behavior is better in general.
• Ethernet and the derived Industrial • Another important advantage of CAN is
Ethernet systems are based on a point- the possibility to monitor the whole data
to-point connection. This means that traffic from any point in the network which
active components in form of switches is not possible with Ethernet. Here you
are required to interconnect nodes. only see a snapshot of the data which
Switches can be implemented with two passes by the point of the network which
external ports on a device allowing daisy- is connected for monitoring purposes.
chain (line) topologies or as external Root cause analysis in case of commu-
switches allowing star topologies. If one nication problems becomes much more
device with an integrated switch or if an difficult and typically, it is necessary to
external switch device fails, larger parts have more knowledge to investigate
of the network may not be accessible problems.
anymore. CAN on the other hand is using • The power consumption required for the
a passive connection method. A CAN Ethernet interface of a device is about 3
node can be connected anywhere to the times higher compared to a CAN interface
cable since CAN allows line topologies (considering physical network interface

01-11
iCC 2017 CAN in Automation

and required microcontroller interfaces information in a car, this requires a split of the
and performance). In case of a stationary transmitted data into two or more messages
machine with a permanent power supply sent successively. As a consequence, not
from the electrical network this is not a all data which belongs together arrives at
problem (when not considering heat dis- the same time. Also diagnostic data usually
sipation or sustainability aspects). How- consists of longer data packages and
ever, for battery-driven mobile machines therefore needs to be transmitted segmented.
and applications, this is a critical aspect.
• Ethernet is more expensive. Just Especially for the car manufacturers, there
by comparing costs for the physical was a need to increase the bandwidth of the
interfaces and cables. Running an communication networks in the car as more
Industrial Ethernet protocol stack and and more electronic functions and ECUs were
quite often also a TCP/IP protocol stack added. One solution would have been to add
in parallel, requires a more performant more CAN networks and the other solution
CPU in order to meet required execution to use faster networks with higher bandwidth
time, response time and real-time and larger messages. BMW was the first to
demands. Furthermore, more resources introduce higher bandwidth by using FlexRay
are necessary for implementation. This in its cars. Daimler and Audi started using
will result is more costs for CPU and FlexRay somewhat later. However, Opel and
memory, especially for frequently used General Motors wanted to avoid introducing
smaller devices with limited functionality. a new network technology and have been
Today, the CAN controller itself is not a searching together with Bosch for a suitable
cost factor anymore. It is a commodity way to improve CAN in that area.
interface on microcontrollers like a serial
interface. In general it can be said that an The result was CAN FD (flexible data-
Ethernet interface is 3 to 5 times the cost rate) presented in 2012. CAN FD has two
of a CAN interface. essentials improvements: CAN messages
• Of course, CAN also has one important can have up to 64 data bytes and bit rates
disadvantage due to its bus arbitration of more than one Mbit/s. In order to keep the
principle: the maximum extension advantages and reliability of the arbitration
of a CAN network depends on the principle but also to be backwards compatible
transmission speed. At one Mbit/s it is to the “classic” CAN, the transmission
typically less than 40m. Consequently. speed is only increased after the arbitration
for larger extensions of the network, e.g. phase, when the payload data of a message
in larger machines, where 100m or 250m is transmitted. This principle results in
are required, the baud rate needs to be considerably increased bandwidth also for
lowered to 500 kbit/s or 250 kbit/s. CAN networks with longer line lengths like
250m where the arbitration bit rate still needs
The conclusion which needs to be drawn to be no more than 250 kbit/s but the payload
from these considerations is that CAN data then can be transmitted with e.g. two
still has its eligibility in many applications, Mbit/s, five Mbit/s or eight Mbit/s.
especially when power consumption, price,
up-time (MTBF) as well as diagnostic and Today, CAN FD is still not a topic for industrial
maintenance capabilities are important applications. A main reason for this is that
requirements. there are no suitable microcontrollers with
integrated CAN FD support available yet.
CAN FD – The booster for CAN However, this is only matter of time, possibly
one or two years from now. Meanwhile, all
One obstacle which makes the use of CAN in European and American car manufacturers
today’s automotive and industrial applications as well as other car manufacturers
difficult, is its limitation to no more than 8 data have plans or have already started
byes per message. For some applications introducing CAN FD in their next series
and functions like electrical drive status and of cars. Consequently, the semiconductor
control data in a machine, or wheel rotation manufacturers are already working on new

01-12
iCC 2017 CAN in Automation

versions of their microcontroller families with not in sensors and actors inside machines.
CAN FD controllers inside instead of “classic” The purpose of the second technology - OPC
CAN controllers. When these will become UA – is to have a standardized protocol for
available, it can be expected that also other communicating data between senders and
applications outside of automotive will start receivers. Of course, there are already many
using CAN FD. other protocols available for this purpose, but
their nature is that sender and receiver already
CAN and the Internet of Things have to know the meaning of the data. There
is no information provided by the protocol
A big buzzword these days is the “Internet about the nature and meaning of the data.
of Things” (IoT). At first glance, it is all about Hence OPC UA also provides semantics for
“Ethernet/Internet everywhere”. Ethernet each data point. This is a strong requirement
from IT down to the sensor level, the when data must be exchanged between
integration of the automation networks with automation systems and IT systems.
the IT networks and Cloud services.
As also already discussed, there is a tradeoff
When looking closer at the use-cases of between the use of CAN or Ethernet for
IoT, it is mainly about improving diagnostics, automation and control purposes inside
maintenance and management capabilities a machine. For many applications, CAN
of assets (machines and systems) and also remains the preferred solution because of
about interconnecting machines and systems efficiency, costs, reliability, performance,
in order to optimize production and making and maintainability. However, related to IoT
production more flexible. In the automotive and the additional data which needs to be
world, it is about communication between gathered from sensors and actuators inside
cars, between cars and infrastructure and the machine, CAN would not be a suitable
between cars and humans. solution because of its limited bandwidth with
up to 8 data bytes in one message. CAN
The nature of data which is transmitted for FD will transmit up to 64 data bytes in one
these purposes is different than the data message even at a much higher bit rate.
transmitted for automation and control This will allow for collecting the necessary
purposes (e.g. for controlling a machine). As additional IoT-related data from the sensors
already discussed in this paper, data used and actuators inside the machine in parallel
for automation and control is typically short to the automation and control data. It will
data transmitted more frequently whereas also make it possible to transmit the OPC UA
IoT related data is detailed diagnostic protocol in case devices inside a machine
data consisting of longer data packages provide OPC UA services.
transmitted less frequently.
Therefore, CAN FD is also the perfect sub-
Within the context of Industrie 4.0 and network for IoT-related applications, because
Industrial Internet of Things, two technologies inside the machine, it will be able to handle
are considered as core communication automation and control data in parallel to IoT-
technologies: TSN-based Ethernet and OPC related data, resulting in overall cost savings.
UA. TSN (Time Sensitive Networking) will
extend Standard Ethernet (802.1) with real-
time capabilities. In principle (or should we
say theoretically?), the real-time can go down
below the microsecond but there are still
many open questions and unclear aspects.
What seems to be clear is that the faster the
real-time will become, the more complex
and more expensive the technology will be.
Therefore, the primary focus is today to use
this technology between IT systems and as
“backbone” in areas like the factory floor, and Figure 5: IoT architecture with CAN FD

01-13
iCC 2017 CAN in Automation

The future of CAN / CAN FD in automotive However, the way that data is communicated
applications inside cars has also started to change. In
recent years, new mechanisms have been
Today, CAN is very well established, not introduced in the AUTOSAR standard
only in passenger cars but also more or like Intelligent PDU Multiplexing (IPDU),
less in everything moving. Due to higher Secure on-board communication (SECOC)
demands for new functions inside a car, or E2E (End-to-end protection profiles).
like driving assistance and security, camera These mechanisms require more data to be
systems providing surround views of the car transmitted than before. In addition, there
or infotainment, the Ethernet technology also are authentication mechanisms for ensuring
slowly started entering the cars already at the that safety-critical data is only accepted
end of the last decade. by receivers when the data comes from a
known and authorized sender.
For Ethernet, there are specific requirements
for use inside cars. First of all, regular CAT5 or Therefore, CAN FD comes at the right time.
CAT6 cables are not appropriate for use inside It will be the bridge between the “classic”
cars. CAT5/6 cables are too thick and stiff. CAN world and Ethernet. It is the most
Therefore it was necessary to develop a new cost-effective solution for many use cases
way of transmitting data at 100 Mbit/s in cars. inside the electrical car infrastructure. And
The solution was BroadR-Reach. Identical to there is one more very important aspect:
Fast Ethernet BroadR-Reach is a point-to- CAN is known to everyone and there is a
point connection providing a mechanism of lot of expertise at the car manufacturers.
100 Mbit/s full-duplex transmission by using Ethernet is new and there are usually
only two wires. Consequently, Ethernet wiring only a few people with the required
inside the car is similar to CAN wiring. For knowledge. It will take time until everyone
the mentioned application areas in which has the same knowledge level for Ethernet
Ethernet is already used or is going to be as for CAN.
used, Ethernet also needs to have a specific
feature for ensuring that data arrives in time All German car manufacturers will introduce
at the receivers. Initially intended for the CAN FD in their next generations, but CAN
infotainment or camera systems, the AVB will not yet be completely replaced by CAN
standard (Audio Video Bridging) was used FD. This will take some time. Ethernet will
consisting of time synchronization based only be used where the performance and
on the IEEE1588 technology and defined capabilities are absolutely required.
bandwidth reservation for data streams. For  
new application areas like driving assistance The future of CAN / CAN FD in
and security, the focus is on TSN (Time non-automotive applications
Sensitive Networking) as this standard is
more general and will be part of 802.1. Its cost structure, flexibility, ease of use
and low power consumption still makes
Will Ethernet replace CAN in cars? - Definitely CAN the best choice for the automation
not within the next 10 or even more years. A and control networks in many different
major aspect for cars are costs. Today, the car application areas. Especially when the
manufacturers consider an Ethernet interface network extension is limited like in smaller
based on BroadR-Reach to be about six machines, for extensions and sub-systems
times more expensive than a CAN interface. to larger machines or in all the various mobile
Another cost factor is the fact, that Ethernet is applications, and especially when they are
a point-to-point connection requiring switches. battery powered. With CAN FD there is four
This restricts the topology and since switches times or even higher bandwidth available at
are only available in a certain granularity, shorter transmission times, and the network
unused ports are an unnecessary cost. It is length can also be extended. With the longer
also not acceptable to use more powerful CAN FD data packages, CAN also enables
microcontrollers with MII interface for simple the connection of IoT applications down to
ECUs like those used in doors. the sensor level.

01-14
iCC 2017 CAN in Automation

Summary
Christian Schlegel
Form today’s point-of-view, the assumption HMS Technology Center Ravensburg GmbH
that Ethernet will become the only Helmut-Vetter-Straße 2,
communication system used for automation 88213 Ravensburg
and control purposes is definitely wrong. +49-751-56146-121
There are requirements Ethernet will not meet +49-751-56146-29
and the price level and ease-of-use of CAN chsc@hms-networks.de
remains unbeatable. CAN FD comes exactly http://www.hms-networks.com
at the right time to make CAN fit for the future
meeting demands for higher bandwidth and
performance by keeping the ease-of-use,
reliability, flexibility and cost level.

01-15