I’m in Chicago this week for a series of meetings. The triggering event was a visit from Chuck Lukasik and John Wozniak of the CC-Link Partner Association. What a great meeting. I love a deep dive into networking technology. And..from people who don’t try to mis-characterize the competition. And also people who can take my challenging comments and answer them. Sure keeps me from going to sleep in meetings.
CC-Link has released a controller-level network based on Ethernet. They just announced a device-level network also based on Ethernet.
Now–a cautionary word. When I say “Ethernet”, what do you think? Wrong. In common usage, we say Ethernet when we mean Ethernet plus the entire TCP/IP protocol suite including http, ftp, snmp, smtp, cip and the like. But Ethernet really only applies to the physical layer. So, CC-Link IE Field Network is Ethernet at the physical layer (Cat 5e, RJ-45 connectors and the rest), but the protocols and ASICs are all developed specifically for CC-Link. This is not proprietary–CC-Link is just as “open” a standard as are EtherNet/IP, EtherCat, Ethernet PowerLink and Profibus/Profinet. And the use of custom ASICs (chips) is not unique to CC-Link, either. EtherCat, Ethernet PowerLink and Profinet IRT all use them. I’ve also heard that EtherNet/IP uses (or will use) one for a motion control option.
CC-Link IE (for industrial Ethernet) Field Network is a spec-killer. It blazes at 1 gigabit/sec. Since the network uses custom firmware, network design, implementation and use becomes faster and easier. As they like to say, “It requires no IT knowledge.” For the electrician (technician) installing this network, it is similar to, but easier than, a device network. Essentially just plug it in. The configuration tool is easier than SNMP. It can be daisy-chain (they call it line), star or ring–or a combination. It is deterministic, yet simultaneously transmits control, log and diagnostic data.
It makes no pretense to “standard” Ethernet–as are basic Profinet and EtherNet/IP. However, it is specifically designed for manufacturing networks and targeted at control engineers and technicians. Mitsubishi, an automation supplier very strong in Japan and strong in China, is the systems supplier behind this network. So, third party suppliers have been lining up support since everyone wants to penetrate the China market. While not strong in North America, it is growing and putting pressure on Rockwell Automation (primary automation supporter) and EtherNet/IP.
The next few years should be very interesting regarding the automation networking landscape. (No, I’m not “prognosticating” a winner.)
Some corrections and comments:
Gigabit Ethernet speed is no big deal; it’s common on PCs and high end embedded chips. Ethernet PowerLink is already certified for gigabit.
I’d say certain standards (EtherCAT slave, CC-Link IE, Profinet IRT) use non-standard MACs (Media Access Control). These non-standards MACs typically can be implemented using ASICs, FPGAs (common for EtherCAT), or specialized CPUs (Ertec, Hilscher netX, Freescale MPC8560).
The IEEE-1588 Precision Time Protocol is in a different class, since it is rapidly becoming standard, available on commodity chips from single chip MCUs to multicore MPUs.
Ethernet PowerLink requires only a standard MAC;it doesn’t even require IEEE-1588, although later versions can use it.
I believe Ethernet/IP CIPSync will require IEEE-1588 support.
EtherCAT masters use only a standard Ethernet MAC; slaves require a custom MAC.
I believe building on widely used standards (such as standard Ethernet MACs and IEEE-1588) are in general a better way to go, because industrial customers can piggy back on the volume and R&D of other, much larger markets (such as automotive, telecom, and PC businesses).
got a little behind. Meant to respond last week. Tony, thanks for pointing out the Powerlink info. Maybe that’s why they didn’t include it in their competitive analysis? Or, it just doesn’t have enough traction in the market?
I purposely avoided getting into 1588. EtherNet/IP and Profinet use it.
I’m pretty sure Profinet IRT does not use IEEE-1588. IIRC, EtherCAT says they use a time synchronization scheme similar to IEEE-1588 (but not the same).
I don’t really care about market share. However, my unscientific analysis of the US industrial Ethernet market (based on the companies I know about) is that there are three very different markets:
1. Process, which I don’t know anything about
2. Discrete I/O, which is typically not real time, and I’d say Ethernet/IP and Modbus/TCP are the clear leaders (with Profinet third)
3. Motion control, which requires real time Ethernet, with EtherCAT the leader and Ethernet PowerLink in second.
Ethernet PowerLink has picked up a few US major drive manufacturers, such as Baldor, Parker, and AMC, so it’s definitely a significant player in the servo drive market.
I wonder how well EtherCAT will move to higher speeds; its "take the packet, insert some data and pass it along" approach is definitely more time critical than the rest of them, and may have problems scaling to higher speeds.
Tony, unfortunately, you have to define Profinet since it has "flavors." Regular Profinet, which is probably the bulk of implementations right now, is "standard" Ethernet. Yes, adding IRT does add some additional time-syncing technology (which I don’t pretend to understand). In fact, I’ve heard through conversations that even EtherNet/IP, which held out for a long time trying to get a standard real-time, deterministic implementation without firmware, has realized that it, too, needs silicon to implement its long-awaited CIP-motion.
I’m not sure how much Ethernet is used in process. There exists a Foundation Fieldbus High Speed Ethernet implementation, but I seldom hear anything about it. In the US otherwise, it’s probably DeviceNet/EtherNet/IP for most market share. However, Profibus sells a lot of device level Profibus worldwide, a fact which probably actually retards Profinet adoption. I’m told the reasons to switch to Profinet from Profibus are less than the reasons to switch from DeviceNet to EtherNet/IP.
Of course, that comment should provoke others, I suppose.
EtherCat momentum in the US seemed to pick up recently, but Ethernet Powerlink is far from dying off. This will be interesting to watch.
Gigabit Ethernet is not really required in the industrial world because the network is not the limiting factor – it’s the processing time in each node. The additional chip costs and wire costs of gigabit Ethernet provide inconsequential incremental speed. Some day it will be cheaper and ubiquitous and then it will make sense in industry only because it will be the minimum standard, not because the speed is needed.
To Tony’s comments:
Market share: I have not seen US-specific market share information, but IMS Research recently published a study. Slightly oversimplifying their global chart, it shows one third each for Ethernet/IP, PROFINET, and other.
The different markets:
1. Process does not use much Industrial Ethernet. Although Fieldbus Foundation has created HSE as their Ethernet standard, I think only ABB uses it. Other DCS vendors use proprietary Ethernet protocols above FF. PROFINET of course works with PROFIBUS in both discrete and process applications.
2. Discrete, process, and motion applications all require a certain speed – just not the same speed. For a process instrument, 100ms may be fine; for discrete IO, ~10ms; for motion, 1ms or less. All of them have to have repeatable timing – determinism.
3. I have to say I never hear about Powerlink or SERCOS, but occasionally about EtherCAT. Of course, in my completely unbiased opinion, PROFINET is the technology leader in motion control.
PROFINET IRT does use IEEE1588 for time synchronization but that is only part of the reason PROFINET IRT can achieve high speed motion control. (Unlike Gary, I do pretend to understand this.)
To Gary’s comment:
PROFIBUS is just so darn powerful compared to DeviceNet (244 byte packets vs 8 byte; 12Mbaud vs 0.5Mbaud) that the reasons to move from PROFIBUS to PROFINET may be less compelling than the move from DeviceNet to Ethernet/IP. There are still lots of reasons to move to PROFINET though: greater bandwidth, larger address space, remote diagnostics, and wireless to name a few. And PROFINET integrates PROFIBUS, DeviceNet, and many other buses through its proxy concept.
Thanks, Carl. I like your way with words — compelling case is a phrase I was searching for. One thing for sure, Ethernet is growing everywhere except maybe in process instrumentation where the bandwidth just doesn’t quite justify it…yet.
OK, Profinet IRT does use IEEE-1588 for time synchronization, although it did take a while for me to confirm that fact (quite a few articles mention Profinet IRT and its custom switch protocol without mentioning IEEE-1588, so it took a bit of digging). That’s good.
As far as what you see, well, a lot depends on where you are and what you do. I’ve never seen or heard of anyone using Profibus or Profinet… and only two of the vendors I use or have looked at support Profi***.
Tony, I do not know of ANY PLC or DCS that does not offer PROFIBUS connectivity. Some are just a little shy about telling anyone. 🙂
Actually, I was thinking motion control (servo drives and stepper drives); I forgot to make that clear….yes, I could add Profibus to the PLCs we do use, but it wouldn’t add anything except cost for our applications (PLC is normally connected directly to a PC).
The only time I’d consider using Profi*** is if I had an application that was a good match for the Siemens S7-1200 series micro PLCs (they support Profinet out of the box); these PLCs will look very attractive after Siemens adds Structured Text support
Final comment, I hope. For small machines, I think there are several distributed motion control systems that often make more sense than a centralized solution. But, distributed I/O systems currently do not make sense in most cases (for small machines).
OTOH, if I were doing another conveyor system, I’d look long and hard at AS-i because its self-healing insulation displacement wiring system looks perfect for a conveyor.