Hirose: Connecting the future
Industrial Ethernet Book Issue 73 / 98
  Print this Page   Send to a Friend  

Deterministic Ethernet: examining the fine print

In 1675, while working with fellow astronomer Giovanni Cassini on a table of eclipses for Jupiter's four moons, Ole Christensen Rømer noticed that they reached their predicted eclipse positions later than expected when Earth was distant from Jupiter, and earlier when it was closer.
He realised that the discrepancy, about 10 minutes, must be due to the finite time that light took to reach Eartømer went on to make the first preliminary calculation for the speed of light. Somewhat later, the world's automation company marketing departments discovered realtime Ethernet.

IT REALLY OUGHT to be said but nobody wants to say it. Profibus International's Karsten Schneider or Carl Henning couldn't say it, EPSG's Stefan Schönegger wouldn't say it, the ODVA's Katharine Voss or Rockwell's Paul Brooks have never, ever said it, Ethercat's Martin Rostan might say it but not in public...
That Ethernet is a stochastic pile of ordure which should never have been considered as a medium for realtime industrial motion control... But never mind, folks, there is great business to be had in clearing up the mess!

A comparison of realtime Industrial Ethernet protocols. Browse by protocol:

CC-Link IE

EtherCAT

Ethernet Powerlink

EtherNet/IP

Profinet

SafetyNet p

Sercos III

SynqNet

TTEthernet

Varan

Nobody would deliberately set out on a scheme for deterministic network communication which used Ethernet as its starting point without good reason. But this is exactly what all the protocol camps did a decade or so ago, and with a fierce justification. The predatory marketing departments of the world's major machine and system automation companies forced the development of Industrial Ethernet. They predicted that a uniform network space from top to bottom of an enterprise would present an utterly compelling sales message, a decision which has taxed the hours and efforts of industry working groups ever since.

It does seem that the corporate marketeers have proved themselves to be right and that the end justifies the means.

Making a universal network vehicle for realtime plant with, if not exactly square, then irregular polygon-shaped wheels has provided the automation world with fresh marketing opportunities in prodigious quantity. Ethernet, used as an homogenous network technology right across the plant offers a reliable promise of manufacturing production control, stock control, delivery logistics and efficient general management. The quid pro quo for integrators is highly effective customer lock-in. For this, protocol SIG development hours are a price worth paying. Furthermore, the nuances of design arrived at by these talented and driven people becomes easy fodder for PowerPoint sales bullets. This was the ammunition used in the old fieldbus wars. The same battle has long moved on to Industrial Ethernet, particularly since the technology is actively displacing fieldbus-level stuff in recent-build plant.


'What do you mean "not round"? The ride will be a whole lot better once we've fitted those new 1588 suspension units...'

So what should a potential industrial automation customer take from this? Probably that the big picture protocol fit for a specific application and the customer's existing plant enterprise environment easily trumps technology differences between individual protocols; we don't know of anything out there which doesn't do what it says on the packet. It is true though that some technologies might be easier or cheaper to apply than others. For example, the cost of software coding to integrate heterogeneous network layers could tip the balance away from a particular hardware and protocol combination.

What did you do in the war?

Having established that the protocol developers have spent the last decade in circling the square to force good old Office IT IEEE802.3 to run motion applications, it is interesting to look at the strategy thumbnails for their solutions. Essentially these divide into three major variations depending how you classify the action: timed and signed for datagram delivery (Profinet IRT, Powerlink and CIP Motion/EtherNet/IP); the network acting as a giant clocked shift register (Ethercat, Sercos); timed special delivery (Varan, SafetyNet, CCLink IE, TTEthernet, etc).

A common factor in all of these groups and subgroups is the need to take account of propagation delay in the cabling - Ole Rømer would have worked it out to be about 4.5ns/metre had he known about velocity factor - to which needs to be added the latency of the node electronics (50ns looks like a good number). Most of the early developers struggled with IEEE1588 PTP in its original version to stabilise datagram delivery in which each node was, in effect its own timed domain. Lack of decent silicon assistance and, more important, a big hike in jitter with increasing number of distributed nodes meant that daisychains were out, the topology being automation's favourite.

The advent of 1588v2 three years ago with its new transparent clocks and use of UDP signalling changed everything. It meant that nodes could now be accurately timestamped on the fly regardless of topology. Further, ready availability of off-the-shelf silicon enabled network devices to be made more cheaply. It also led to major system spec revisions for Powerlink, Profinet IRT, EtherNet/IP-CIP Motion and other 'standard' Ethernet proponents which must have left potential motion customers pondering issues of specification stability.

Turning Ethernet into a specialised fieldbus is not necessary for most factory automation jobs: the completely standard variety can turn in a fairly sharp performance simply through use of segmentation and strict traffic management under 802.1 priority regimen. For instance, Profinet RT, which uses UDP without any stack bypass tricks, has been shown to handle all manner of time-sensitive industrial application tasks with nothing more than strict network management. The number of applications needing or able to run Profinet IRT are very few by contrast. The same may be equally said of EtherNet/IP vs. E/IP with CIP Motion.

The protocol pages

In an area where it is difficult, if not impossible, to discuss protocols fairly or comprehensively in relative terms, we have instead sought the views of the gatekeepers about themselves and their technology. The pages that follow offer the unedited (except for reasons of clarity) results of a standard questionnaire sent out to the protocol managements. To make it a bit easier to digest the answers in context, we have added a brief system description taken mostly from edited information supplied by the gatekeepers themselves. We have occasionally added our own clarifications but these should not materially change the essence of the presentations.

Browse by question:

1) What is your definition of 'realtime' in the context of Industrial Ethernet?

2) What cycle times/latency does your protocol achieve, and under what conditions?

3) What would be the shortest response time that your protocol could achieve to an asynchronous consumer (or slave) event, best case/worst case?

4) By what mechanism do protected realtime network segments connect with the higher level network, for instance in the transfer of system status/ERM data?

5) What are the permissible topologies for realtime network segments?

6) What are the synchronisation performance limits between individual realtime network segments, for instance in an applicationwhichmight concatenate individual automation cells?

7) What are the failover arrangements in the event of realtime network segment failure¡; and are there circumstances where the mechanism might be problematic?

8) Is operation at SIL3 possible within realtime network segments?

Realtime Ethernet protocols:

CC-Link IE

EtherCAT

Ethernet Powerlink

EtherNet/IP

Profinet

SafetyNet p

Sercos III

SynqNet

TTEthernet

Varan


Source: Industrial Ethernet Book Issue 73 / 98
   Print this Page    Send to a Friend  

Back

Sponsors:
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: 20.11.2018 · 24 User online · Privacy Policy · Contact Us