(www.lantronix.com). The two main ones, transmission control protocol (TCP) and user datagram protocol (UDP), are both built on top of Internet protocol (IP). Other protocols—simple mail transfer protocol (SMTP) and hypertext transfer protocol (HTTP), for example—are based on TCP as well as file transfer protocol (FTP), he says, adding that trivial file transfer protocol, or TFTP, is based on UDP.
In the control systems domain, Fieldbus Foundation’s High-Speed Ethernet (FF HSE), the Open DeviceNet Vendor Association’s (ODVA) EtherNet/IP and Profibus International’s ProfiNet are the top three applications that run on Ethernet, says Dick Caro, principal of CMC Associates (www.cmc.us), Acton, Mass. That lets many different services run simultaneously on the same network, he adds. “They all result in messages that are framed as either TCP/IP messages for the non-real-time functions, such as status or setup data, or UDP/IP for real-time data exchanges.” Concurrently, the user could share that Ethernet network with any other IP, including FF HSE or ProfiNet, Caro says, because they all use communications based on TCP/IP and UDP. “[And] because TCP/IP and UDP are point-to-point protocols, then devices would only talk to other devices that understand their protocols.”
Some protocols compete in a technical sense, Shayegani points out. In TCP, verification of sending and receiving of the data packet is built into the protocol, he explains, but in UDP, the verification is the responsibility of the application, not the protocol. The impact, he says, is that TCP has a larger network overhead, whereas UDP is more optimized on traffic flow. In UDP, the receiver replies with a simple acknowledgement of message receipt. But TCP requires the receiver to acknowledge receipt of the sender’s message, and then the sender to acknowledge receipt of the receiver’s acknowledgement, Shayegani notes. Thus, on average, TCP has triple the communication requirements of UDP, he says—and that’s why TCP has a larger network overhead.
Overall though, the protocols don’t compete with each other—that is, none has priority—but rather for bandwidth and what they can do, Shayegani explains. “There’s no differentiation of which protocol gets its data packet on the network first—it’s simply who’s in line first.”
The resulting traffic jams raise the main issue in industrial Ethernet networks, which he says is time latency or delays in receiving messages. On a hard-wired network, protocols are deterministic, Shayegani explains. But Ethernet is non-deterministic, and this is where the ODVA’s common industrial protocols CIP Sync and CIP Safety come into play, he says.
Sync clocks
For example, with CIP Sync, there’s a master clock that synchronizes the network’s slave clocks. Thus, data packets have a time stamp, Shayegani says. “That means we can send data packets across the network, that even though they arrive at different times, we know when they were sent and when they’re supposed to arrive, so we can synchronize actions involved.”
And because of the traffic on an industrial automation Ethernet net, it is important to isolate and clean the factory-floor nets, Shayegani says. Cleaning the network means keeping non-essential traffic such as e-mail and Web browsing off of the network, he explains.
“The really neat part about the use of Ethernet as a transmission base is not so much Ethernet itself, but the fact that the worldwide network runs on IP,” Caro states.
C. Kenna Amos, [email protected], is an Automation World Contributing Editor.
