Hirose: Connecting the future
Industrial Ethernet Book Issue 71 / 48
Request Further Info   Print this Page   Send to a Friend  

Working successful motion control via standard Ethernet

Until recently, coordinated motion and similar difficult real-time applications could not be achieved using EtherNet/IP, but the introduction to the market of CIP Motion products has now extended EtherNet/IP to such demanding applications. Viktor Schiffer explains the basics of real-time communication with EtherNet/IP and the set of standard Ethernet features that have been used to expand its capabilities so that such motion applications can be resolved without abandoning common Ethernet standards.

STANDARD ETHERNET IS in extensive use in office and manufacturing environments, yet it is often seen as unsuitable for real-time control applications. However, standard Ethernet has been used for real-time fieldbus applications by EtherNet/IP in many installations around the world - about three million devices in the field. Real-time communication using EtherNet/IP, plus the set of standard Ethernet features, allows standard Ethernet products, such as cameras, printers, servers, IP phones etc to operate in the same network without having to use special hardware.

Extending the network's speed and capacity, e.g. via a gigabit backbone, is no problem either since CIP Motion is neither tied to special extensions of the lower communication layers nor linked to specific data rates.

It should be noted that the rising popularity of Ethernet for control is driven by the users' desire for simplification and future-proofing of their installations. In the case of networks, simplification means reducing the number of networks and using standard tools that are abundantly available. Users typically want one single network throughout the facility that needs only one skill set to maintain it. Users also envision that adopting standard Ethernet will future-proof their infrastructure much more than networks that deviate from the standard.

This article examines the principles of CIP Motion, how real time for motion is achieved, and how CIP Motion is easily integrated into a control system using standard, unmodified Ethernet.

What is Ethernet?

Although this might seem like an extremely basic question to ask of Industrial Ethernet Book readers, most of whom are highly experienced, it is important because it tends to vary according to the differing perspectives of IT people and shopfloor engineers. When IT people answer this question, they usually imagine the office and computer environment with Ethernet networking and all associated applications such as HTTP, FTP, email, Voice over IP, Video over IP on the upper layers, as well as a large number of management services on the lower layers - see Figure 1, the ISO OSI layer model.

Fig. 1: The ISO/OSI layer model. Ethernet networking and all associated applications are on the upper layers, plus management services on the lower layers.

However, if you ask a factory automation person, then he or she will relate Ethernet more and more to networked control applications that complement or (increasingly) replace existing fieldbus applications.

The factory automation Ethernet world can be subdivided into those Ethernet variants that still use the same addressing and transport mechanisms as the office and computer world (e.g. Modbus TCP, EtherNet/IP, Profinet NRT) and those that have modified standard Ethernet to achieve their desired real-time capabilities.

These Ethernet variants, such as EtherCAT, Ethernet Powerlink, SERCOS III, Profinet IRT etc are significantly different from standard office Ethernet, and can only be connected via gateways to standard office Ethernet. This often results in structures as shown in Figure 2.

Fig. 2: Separate Ethernet worlds. Ethernet variants (EtherCAT, Ethernet Powerlink, SERCOS III, Profinet IRT etc.) can only be connected via gateways to standard office Ethernet - though it is generally not sensible to directly connect motion to ERP. A suitable gateway would be needed to isolate non-RT traffic except that destined for the motion nodes.

The individual gateway to the enterprise area is the problem; the issue is that there is more and more interaction between the office and the factory world, and that more and more office technologies are migrating into the factory floor. To reap the benefits from cost reductions through mass production, it is important to deploy these technologies on the factory floor without modifications ('keep it standard'). This means that either there must be compatible channels available on the factory floor, or the network must be designed using standard Ethernet for compatibility to the office and computer. EtherNet/IP offers a solution that is completely built on standard Ethernet yet offers real-time capabilities that allow solving high-precision motion applications.

Real-time on the basis of TCP/IP and UDP/IP, extended by IEEE 1588 and QoS Ethernet, is still often viewed as the communication system as it was in the early days, i.e., a system that cannot guarantee the transmission time for a specific message. This was because of the shared media and the associated frame collisions, which lead to random wait times before a new transmission can be started. This collision behaviour does not allow the prediction of a transmission time; it is made worse when multiple collisions occur.

Replaced by proprietary layers

This is why several consortia were formed during the early development of industrial Ethernet and solutions were created that increasingly modified the original Ethernet system, so that it was no longer based on TCP/IP and UDP/IP. These components are replaced by proprietary layers as shown in Figure 3.

Fig. 3: Ethernet without and with special layers. The original Ethernet system has been modified and is no longer based on TCP/IP and UDP/IP.

Since full duplex switching is the norm for most standard Ethernet installations, with each port running at wire speed, collisions and their negative effects occur much less often so that the need for intermediate layers becomes less pronounced. The only kind of 'congestion' that still might exist after collisions have disappeared is when large amounts of data have to be transmitted, but this can easily be resolved by switching to higher data rates (100 MBit/s, 1 GBit/s and more).

EtherNet/IP and CIP Motion are based on a full duplex Ethernet infrastructure supporting at least 100 Mbit/s.

TCP/IP on the transport layer offers a confirmed transport service, yet it is unsuitable for real-time because its complexity. UDP/IP is much simpler and faster and this is why it is used within EtherNet/IP for the transmission of all time-critical information - the transport time for a message is mainly governed by the available bandwidth in the network that is shared among the messages in the network.

Therefore, the real-time performance of EtherNet/IP is comparable to that of typical fieldbus systems. Like them, there is an amount of jitter in the transmission of messages, which is why this is not good enough for very accurate motion. What can be done to make standard EtherNet real-time capable for motion without deviating from standard Ethernet (TCP/IP, UDP/IP, every host can start sending at any time)?

Most demanding real-time applications require determinism (knowing when things happen ahead of time) to perform accurately. This kind of real-time behaviour can be achieved by the introduction of a common notion of time, achieved by the mechanism of synchronised distributed clocks in the devices needing deterministic behaviour. All system actions then make reference to the same time indicated by these clocks. This is sufficient to meet all real-time criteria. The clock synchronisation mechanism used for EtherNet/IP is the international standard IEEE 1588 (Standard for a Precision Clock Synchronisation Protocol for Networked Measurement and Control Systems)[1], see Figure 4.

Fig. 4: Clock Synchronisation via IEEE 1588. Most demanding real-time industrial automation applications require determinism in order to perform correctly.

By using this mechanism, the overall control system has a common understanding of time and the actions of the overall system are not determined any more by when a specific message was received or sent, but by a time-stamp embedded in the message.

Time-synchronised input data get a time stamp that is generated at the time of data capture; time-synchronised output data get a validity time stamp which tells an output device when to apply this data. Therefore, the time at which data is transmitted is of secondary importance; all that matters is that the transmission must take place in time.

To ensure that real-time data has priority over other traffic, EtherNet/IP uses Quality of Service (QoS). The QoS standard IEEE 802.1 D/Q 'Differentiated Services' plus prioritisation is used within EtherNet/IP, further details in [2] and [3]. This allows prioritised transmission of all time-critical frames within EtherNet/IP, i.e. the sync messages of IEEE 1588 and the application data of CIP Motion; both get priority over all other messages within an Ethernet network.

Making distributed real-time clocks available in distributed CIP devices is called CIP Sync. The base technology IEEE 1588 is extended within CIP so that all relevant data is accessible through an object (Time Sync Object, Class Code 43hex). The CIP Sync functionality alone is sufficient as the basis of a number of time-critical applications, such as Sequence of Events Recording or synchronous outputs.

A number of products have already been on the market for a while; e.g. modules sitting in a rack (see [5]) or I/O blocks communicating across EtherNet/IP (see [6]).

CIP Sync has been used as a basis to create a communication profile for motion devices. This was achieved by defining the data sets that both sides need to exchange for time-synchronised motion and how to handle this exchange. These specifications have been published as a device profile within the CIP specifications (see [4]). These specifications have been deployed in certain servo drives for EtherNet/IP and these products have undergone thorough testing both on a device and on a system level.

The availability of this technology, allows users to combine commercial applications (web access, Voice over IP, Video over IP etc.) with industrial control components including coordinated motion in one open system. This can be achieved without having to accept the limitations of specialised Ethernet systems (strict separation of real-time components and commercial technologies) so that integrated systems can be built as shown in Figure 5.

Fig. 5: An integrated control system. Users can combine commercial applications with industrial control components, including coordinated motion in one open system.

For the user, the configuring and commissioning of such a system is extremely simple; all he or she has to do is to carry out the following steps:

Insert the servo drive in the I/O structure - like any other I/O device;

Activate CIP Sync in the EtherNet/IP interface as well as in the drive (Figure 6);

Fig. 6: Activation of CIP Sync. Activate CIP Sync in the EtherNet/IP interface as well as in the drive.

Associate drive to a Motion Group (Figure 7);

Fig. 7: Configure the drive within the motion group. Setting the communication cycle time (coarse update period).

Configure the drive within the Motion Group, mainly setting the communication cycle time; the rest is identical to existing drives (SERCOS).

The advantages of using EtherNet/IP with Ethernet on the basis of TCP/IP and UDP/IP does not look like being real-time capable at first sight. However, the developments within ODVA and the products now available are clear proof that real-time performance and open Ethernet are not a contradiction.

Since all technology details have been published within the CIP Specifications, it is expected that further CIP Sync / CIP Motion products will be available soon.

[1] 1588-2008 IEEE Standard for a Precision Clock Synchronisation Protocol for Networked Measurement and Control Systems
[2] Brian Batke: Implementing and Deploying QoS for EtherNet/IP. Paper and Presentation from the ODVA Conference 2009
[3] The CIP Networks Library, Volume 2, EtherNet/IP Adaptation of CIP, Edition 1.10, November 2010
[4] The CIP Networks Library, Volume 1, Common Industrial Protocol (CIP), Edition 3.9, November 2010
[5] Publication 1756-TD002A-EN-E - May 2009, downloadable from http://literature.rockwellautomation.com/
[6] Publication 1732E-UM002A-EN-P - March 2010, downloadable from http://literature.rockwellautomation.com/

Viktor Schiffer works for Rockwell Automation in Haan, Germany.


Source: Industrial Ethernet Book Issue 71 / 48
Request Further Info    Print this Page    Send to a Friend  


Analog Devices: Time Sensitive Networking
DINSpace fiber optic and Cat 6 patch panels
Japan IT Week Autumn

Get Social with us:

© 2010-2018 Published by IEB Media GbR · Last Update: 21.11.2018 · 38 User online · Privacy Policy · Contact Us