When I was a new manager (way back when) I had a troublesome employee – Derris was a bit curmudgeonly [Encarta definition: curmudgeonly: somebody who is irritable or stubborn: somebody considered to be bad-tempered, disagreeable, or stubborn]. I consulted with the president of the company trying to figure out how to deal with him. The president started by asking a very insightful question: “Does he get angry for no reason or is it more like righteous indignation?” It was definitely the latter. The moral of the story as it relates to this post: I get righteously indignant when I see misinformation about PROFIBUS and PROFINET.
So today let’s move on to the second page of the FUD piece that was handed out at a Rockwell lunch-and-learn and that fails to accurately compare Ethernet/IP (EIP) and PROFINET. (You can see this second page here. Page 1 is deFUDified here.)
This section is divided into three columns: PROFINET claim, Facts, and EIP Response. The first alleged PROFINET claim is that standard Ethernet is not suitable for motion control. That is not our claim.PROFINET IRT (Isochronous Real-Time) that is used for motion control is in fact standard Ethernet. Look at a PROFINET IRT frame with an Ethernet sniffer and it is just a plain old Ethernet frame. We do reserve bandwidth on the network to ensure that the motion control data gets through no matter what other network activity may be loading the network. We do use hardware to accomplish this, but their alleged “fact” that it is proprietary is just wrong. Just like any of the Industrial Ethernet standards that are part of IEC61158 (including EIP), the specification is open and available to anyone for purchase from the IEC. EIP relies on IEEE1588 to synchronize the times of the various nodes. The EIP approach makes no provision for the network becoming flooded with traffic that could disrupt the flow of data. (That’s why we reserve bandwidth.) The EIP response is that Rockwell has demonstrated motion control. “Demonstrated” is the key word; they have no shipping product and no application stories to show that it works in the real world. They respond to the “proprietary” label they mistakenly apply to PROFINET by commenting that their stuff was developed in an “open” working group. PROFINET motion control was not developed secretly in a smoke-filled backroom at the corner bar. Several of the 50 PROFIBUS and PROFINET Working Groups work openly on motion control.
The second claim they claim we make: “PROFInet is an extension and relative of Profibus.” Ok, we make that claim. Their contrary “fact” is that PROFIBUS and PROFINET are different protocols. (They used to claim that PROFINET was PROFIBUS on Ethernet, so they are getting closer to being correct.) PROFIBUS and PROFINET are different protocols… by design. When we developed PROFINET we could have taken the same route that Modbus and DeviceNet took and just wrapped our old protocol in a TCP or UDP wrapper and been done. But, that approach has a penalty when it comes to determinism. We chose a more deterministic approach but protected our existing users’ investments by using a proxy to connect PROFIBUS to PROFINET. This allowed us to specify proxies for other protocols as well.
They get the third claim right. We do think “Embedding switch technology into control products is good.” It allows a bus type topology instead of a star. Wasn’t that one of the big benefits of a fieldbus in the first place? If your application permits it, use a bus type connection with the switches embedded in the devices. There is no reason not to use bus topology if a single failure that disrupts the network is no worse than the single failure by itself. The EIP responses here seem only tenuously relevant, so let’s treat them separately.
Ethernet technology is constantly improving. Can’t argue with that; that’s one of the reasons why we all chose to base our Industrial Ethernets on Ethernet in the first place. But let’s be realistic here, too; 100Mb full duplex, switched Ethernet will be adequate for decades to come in control environments. The speed of the Ethernet is not the limiting factor in achieving real-time speeds; it’s the processing time in the devices. Higher speeds for Ethernet will be very valuable at the backbone level with trunking, but that will have minimal impact on the control devices.
This page’s final claim about the node count difference was dealt with in FUD 1. But I thought they had a lot more than 3 million DeviceNet nodes shipped so I don’t know why they think that we think there are only 3 million. I have to admit being amused by the notion in their response that they understated the announced number of DeviceNet nodes .No one understates those kinds of numbers… except us, I guess. We count chips used in devices, but some devices don’t use the chip; they use a standard RS485 chip and implement the protocol in software. But you know what I mean…
They claim that “All Profibus networks do not share a common protocol. ”That’s wrong, too. This false claim arises because we have PROFIBUS DP and PROFIBUS PA. DP is used with discrete IO and PA for process instruments. But the protocol is the same; only the physical layer is different. One set of configuration and diagnostic tools handles both.
Their last “fact” spills over to page 3, so let’s carry this through to the end. Here’s page 3. They make a comparison from 2004 between Rockwell and Siemens respective shipments of EIP nodes and PROFINET nodes. Firstly, 2004 is ancient history in the Industrial Ethernet market place. So, who cares what happened then? The more important thing to note is that they focus on Rockwell for EIP nodes and we focus on multiple vendors for PROFINET nodes. As an open standard, PROFINET devices are available from many suppliers including Phoenix Contact, WAGO, Beckhoff, and many more. The interesting comparison will be 2007. We will report on total PROFINET nodes early in 2008.
They say that we claim that PROFINET is an international standard. Of course it is. In fact, so is EIP.They claim that 3 European automakers standardized on the “future” of PROFINET. Well, it was 4 automakers actually and they have already deployed and continue to deploy PROFINET.
Finally, and I’m getting just as tired of this as you are, Gentle Reader, they say we claim PROFINET works with standard Ethernet. We do claim that because it is true. Their mistaken” facts” say some things about “V3” of PROFINET. Version 3 of PROFINET is just the latest version of the spec. We do have conformance classes that differentiate the performance level required. So, for example, conformance class C is used for motion control using bandwidth reservation. One of the points they are missing is that the bandwidth reservation system still uses standard Ethernet. Another point they are missing is the meaning of the word “proprietary.” The opposite of proprietary is not open source it’s open standard.PROFINET is no more proprietary than EIP is. Both are based on open standards.[Encarta definition: proprietary – used with exclusive legal right: used, manufactured, or sold by a person or company with an exclusive property right such as a patent or trademark.] I’ve already covered the switch arguments in FUD 1, so no need to repeat them here.
I’m not going to jump to the defense of the footnoted technologies although I bristle at the word “proprietary.” Others are welcome to jump to their defense by commenting. In fact, comments, refutations, defense of Rockwell’s position – all are welcome.