For example, the SErial Realtime COmmunication System (Sercos) network is optimized for the high-speed, deterministic messaging required by integrated motion control. This network standard, IEC 61491 and EN61491, is based upon a ring topology of fiber optic media.
Ethernet, on the other hand, has become the de facto standard for manufacturing information networking, and is characterized by high bandwidth but is not deterministic.
As in the case of the hypothetical engineer quoted above, Hibbard says that the drive toward reducing the number of networks in a plant toward one is something like driving manufacturing processes toward zero defects. You’re probably not going to get down to the goal. In networking, there is a layer where different goals come into conflict with each other.
This is where the idea of a Sercos III came to life. The challenge, sort of like putting chocolate and peanut butter together, was how to put the deterministic Sercos together with high-bandwidth, nondeterministic Ethernet. Solving that challenge would mean that engineers could have synchronized, distributed motion control while simultaneously providing the real-time manufacturing information craved by the information technology department to feed its databases.
Just a cursory look at the two network types reveals some of the difficulties to be overcome to bring this new network to life. Hibbard notes that Sercos started out at 2 megabits/sec and evolved through 4 mbs to today’s 8 and 16 mbs. These throughputs seem slow compared to 100 mbs Ethernet, but Sercos was finely tuned to optimize the messaging requirements of motion control systems. For instance, sending just one bit of information over Sercos requires about four bytes (eight bits to a byte) of overhead. Ethernet, by contrast, requires about 84 bytes of bandwidth to send that one bit of data.
The Sercos (www.sercos. com) organization in Stuttgart, Germany, initiated a working group about a year ago to investigate a solution. According to Hibbard, the group looked at all the existing buses and figured that standard Ethernet TCP/IP (transmission control protocol/Internet protocol) would allow links to just about all of the existing infrastructure in manufacturing. With that decision out of the way, the next problem was how to combine the networks.
The way most technologies have been added to the Ethernet stack is to simply add a new protocol. With Sercos, that approach would in effect, be putting motion on top of Ethernet. But Ethernet is not deterministic. To get determinism, companies would have to add an ARM processor to each connection and would still need a switching device as part of the network. This added cost would eliminate the benefit of a low cost network economical for small drives.
“What we have found,” reports Hibbard, “was why not look the other way around? Rather than put motion under TCP/IP and suffer with less determinism and more expense, flip it and put TCP/IP under the motion bus. This achieves the objectives without the drawbacks. And this is what we did with Sercos III. From there, we use the transfer and synchronization mechanism of Sercos on the transport layer with 100 mbs Ethernet. Sercos will control the bus for all the motion critical demands and will work cyclically, as it does today. A switch will monitor the network like a traffic cop and allow non-critical data on the network after the motion commands are done.”
The standard is close to final adoption with components to be available in 2005.
Gary Mintchell, [email protected]
