Juggling and motion control are all about keeping appointments. Manufacturing and machine building are all about keeping promises. There’s more than one way to skin a cat. Time is money. Most machine builders know these principles by heart.
As the controls engineering supervisor at CBW Automation Inc. in Ft. Collins, Colo., Steve Corwine and his colleagues continually must devise ever-faster, simpler, more-efficient, and less costly ways for CBW’s robots to pull plastic parts out of injection molding and thermoforming machines. Do more with less. Great job—now, what have you done for me lately? We all know this drill.
To make its robots simpler to build and operate, CBW recently spent 18 months redesigning and rebuilding its TS-303 side-entry robot to include PC-based controls of its servo drives via a SERCOS digital network (Figure 1). This replaced the robot’s formerly hardwired communications between its I/O points, PLC, and servo drives. The robot now uses Beckhoff Automation’s TwinCat software to control Bosch Rexroth’s Indramat servo drives. Its PC-based scan time is 2 msec, whereas its former PLC scan time had been 27 msec.
Pulling Plastic
Figure 1: TS-303 side-entry robot’s ability to grab and retrieve parts is based on CBW Automation’s patented end-of-arm-tooling (EOAT) methodologies.
Photo courtesy of CBW Automation
TS-303’s ability to grab and retrieve parts is based on CBW’s patented end-of-arm-tooling (EOAT) methodologies. After building its robots, CBW integrates them into the work cells of its clients’ molding and forming machines, such as those by Husky or Stackteck, and adds capabilities as needed so they can stack parts, transfer parts to packaging machines, or put parts in boxes and tape them.
“The biggest reason for incorporating a fieldbus was its high-speed communications, and the labor saved by doing away with all the hardwiring that used to go to all the I/O points,” says Corwine. “SERCOS also lets us use fiberoptics, so we can put more components nearer to their drives and other remote locations. This means they can be in smaller boxes closer to where they need to be, rather than in big centralized boxes with lots of wiring.”
Zuri Evans, Siemens E&A’s Simotion product manager, adds that motion buses’ ability to reduce wiring and locate components further afield also enables efforts by some builders to develop more modular machines. “Each module now can have its own intelligence, but they all need a common fieldbus for communications that is fast, accurate, and safe,” says Evans. “Mechatronic solutions are replacing mechanical components with electric motors, but higher axis counts and greater synchronization requires more capable fieldbuses that are flexible, have faster cycle times, and less jitter. Basically, users need controllers that can talk motion with other controllers to handle more complex solutions.”
Recent Motion Bus History
Though digital networks communicate more efficiently and with less hardware than their point-to-point predecessors, they historically begin to break down and drop the ball when faced with the higher-speed data requirements of industrial motion control applications.
“In the late ’80s and early ’90s, we had centralized motion controllers and analog interfaces, but these made it difficult to set tuning parameters for drives, and it was difficult to get errors out of the drives,” says Markus Sandhoefner, B&R Automation’s Ethernet Powerlink specialist. “The digital interfaces that came next were better, but their 30-50 msec update times were too slow for motion control. Developers got faster with CANbus and Profibus, and then began looking at Firewire and Ethernet’s 100-µsec updating to achieve the determinism required for motion control.”
Consequently, several heroic efforts were made in recent years to bring the benefits of low-power, two-wire fieldbuses to motion control, and several different methods sprung up to serve users. Just like commandos in a war movie, this usually meant synchronizing watches. The two main camps involve time slicing or time stamping of the signals that deliver messages and software data packets, and most protocols operate via some flavor of Ethernet. The main players include SERCOS, Ethernet Powerlink, Profinet IRT, EtherCat, CIP Motion via EtherNet/IP, SynqNet, and even generic Ethernet TCP/IP itself.
Unfortunately, debates still rage over which is most useful, which should be the industry standard, and which is likely to dominate future motion-related fieldbus markets. Perhaps because there still are relatively few motion bus applications, most supporters of one protocol seem to spend much of their time criticizing the others. Bickering aside, most users just want to build machines faster, integrate motion systems with more axes, drives and nodes, and help their own end users manufacture products faster and better. Some suppliers help them. Others appear to help users, but serve mainly themselves by binding users to unnecessarily proprietary networks that can potentially restrict their flexibility and long-term growth.
“Servo companies such as Bosch-Rexroth, Baldor, Siemens, Danaher, Yaskawa, Baumüller and other suppliers—still debating which protocol to use—are evaluating Profinet IRT, Powerlink, CIP Motion, and others,” says Corwine. “So, we machine builders are waiting for an industrial standard to emerge, because more builders and manufacturers in our world see the need for a conformance standard and see the benefit when suppliers support other people’s protocols.”
Slice Your Schedule
Several motion bus protocols use time-slicing methods to avoid message collisions, increase speed, and secure determinism. Time slicing guarantees that critical tasks are assigned a specified time slot to send data. A managing node on the network handles time allocation, so the other nodes can transmit without interference when it’s their turn. Time slicing is used mostly by Profinet IRT, Ethernet Powerlink, EtherCat, and SERCOS.
Even when SERCOS moved from its first version’s dedicated fiber to SERCOS III’s 100-Mbps Ethernet, its developers sought to retain its time-slicing principles, while adopting as much of Ethernet’s physics and transport layer as possible, explains Scott Hibbard, Bosch Rexroth’s technology vice president and a SERCOS N.A. board member.
Coating in a Vacuum
Figure 2: Semicore’s SC1500-01 vacuum coating system uses dynamic motion axes and accurate positioning to coat the inside of pipes with diamond-like carbon, chromium carbide, and titanium nitride coatings.
Photo courtesy of Semicore
For instance, this level of coordination was used by Semicore Equipment in Livermore, Calif., to design and build its SC1500-01 vacuum coating system (Figure 2). This machine uses highly dynamic motion axes and requires very accurate positioning to coat the inside of pipes with diamond-like carbon, chromium carbide, and titanium nitride coatings, so they’ll last longer in gas, chemical, and petroleum in internal surfaces. SC1500-01 uses Beckhoff cabinet-mount PCs and control panels, TwinCat automation software, and distributed I/O networked over Ethernet TCP/IP for its servo motors and drives. “It’s not difficult to apply wear and corrosion-resistive coatings to pipes’ outside surfaces, but coating the inside is where it gets tricky,” says Matthew Hughes, Semicore’s president. To handle SC1500-01’s high-frequency control, arbitrary waveform generation with multi-channel cope feedback, and complex gas management functions, TwinCat software manages parts moving through the coating system’s transport and positioning systems.
“Time slicing is like sending a Fed Ex package that’s guaranteed to arrive the next day,” says Sandhoefner.
For example, Voorwood Co. in Anderson, Calif., uses Ethernet Powerlink to coordinate 11 drives on its new arch-shaping machine, which manufactures cabinet door parts, countertops, and other products. “Our end users really don’t care about different communications systems,” says Adam Britton, Voorwood’s engineering manager. “They only care if it fails. So, now we’re using fieldbus to communicate from the PLC to the I/O rack, which allows us to save space by not having the I/O hooked to the PLC. Instead, the PLC now can send complete cam data to the servo drive, and the drives then communicate directly with each other via the network to execute motion, instead of waiting for incremental commands from the PLC. This configuration seems more reliable, and doesn’t require a super-fast fieldbus system to decrease latency.”
Supporters say it helps to think of devices in a time-slicing network as train stations, which use controller cards and intelligent switches to learn what’s plugged into the network, know the network’s schedule, and compensate ahead of time for any disruptions. For example, in a demonstration at Germany’s Hannover Fair earlier this year, Profinet IRT synchronized 54 axes on one network, which reportedly used up 30% of its bandwidth, and then sent a real-time video signal through the same cable, but still only took up 60% of its bandwidth. Likewise, SERCOS III performed a similar demonstration at the fair by running a full-motion video signal through six or eight drives while they were running, and did it without switches.
IRT’s synchronization is organized by a sync master—one controller with an ultimate clock—which updates other clocks on the network, and delivers a schedule that informs them when they’re supposed to receive packets and where to send them, so they don’t have to read every packet on the network. “IRT creates a very predictable time lag, so it’s easy to compensate for, which results in minimal jitter,” says Evans.
Jeremy Bryant, networking specialist in Siemens’ Automation and Motion division, adds that, “The key to Profinet’s simplicity is that all its functions—motion, PLCs, distributed I/O, HMI, safety, engineering, maintenance, and MES—can coexist on the same network and cable.”
Floating Glass
Figure 3: To avoid imperfections, Profinet IRT handles vision system data from Grenzebach’s float glassmaking line to create cam profiles that are sent to controllers in charge of the X-Y cutting axes.
Photo courtesy of Profinet IRT
For example, Profinet IRT recently was used to synchronize three Simotion controllers at Grenzebach’s float glassmaking and cutting line in Asbach-Baumenheim/Hamlar, Germany. This process floats molten glass on a layer of phosphorous, and two PCs running Windows XP and a vision system take and process images of the glass to find imperfections (Figure 3). “This data is used to create cam profiles, which are sent via Profinet to the controllers in charge of the X-Y cutting axes,” explains Evans. “This level of control allows Grenzebach to avoid imperfections, while creating the largest pieces of glass possible. Ethernet provides fast cycle times and bandwidth, but it’s Profinet that lets all these functions happen on the same cable.
Stamping Your Tags
The other primary method for coordinating motion bus traffic is known as time stamping, and it’s based on IEEE 1588 Standard Precision Time Protocol (PTP). IEEE 1588 defines a method for sub-microsecond synchronization of the clocks in sensors actuators, and other terminal devices on a standard Ethernet-based network or other distributed application. It’s used primarily by EtherNet/IP with CIP Sync and CIP Motion. IEEE 1588’s basic function is to have the most precise clock on a network synchronize all the others, and then time stamp each data packet moving on the network. Each packet is tagged with a statement about when it was captured or when a specific activity needs to begin, and the network’s receiver processes times tags according to a schedule. CIP Sync synchronizes its network’s clocks, while CIP Motion defines the data packets needed to perform motion control.
“EtherNet/IP is unique because it uses standard Ethernet connections and TCP/IP stack to work with motion,” says Steve Zuponcic, applications engineer for Rockwell Automation’s field marketing services group, and a member of ODVA’s Disributed Motion Joint Special Interest Group (JSIG). “Other protocols use Ethernet’s physical layer, but bastardize other layers of the OSI stack, such as the transport and data link layers, which can create a new beast and cause some potential problems. For example, if a user tries to employ some standard Ethernet switches and routers, they might be unable to connect to that altered network. We think it’s better to use a standard implementation of Ethernet that isn’t limited by proprietary changes.”
However, critics claim that time stamping just notes the time that a message was sent, but doesn’t guarantee that it actually will arrive at a particular time. “Ethernet’s standard 64-byte messages waste overhead, and limit users to 19 axes with nothing else on the network,” says Hibbard. “This restricts the nodes, drives and devices they can use at a time when they want to add I/O points, HMI, and data logging to their networks.”
Conversely, time stamping’s supporters counter that time slicing is too proprietary and relies too much on hardware, such as an ASIC chip needed to boost a protocol’s performance up to real-time levels. Again, they claim that dedicated hardware can make a motion bus too rigid and limit its openness and ability to interoperate with other devices.
Software vs. Hardware
While hardware traditionally is used to increase motion bus speeds to deterministic levels, some of this assistance also comes from software. For instance, SERCOS uses a chip, but, says Hibbard, its software list is programmed into a field-programmable gate array (FPGA) that any user can load into the process.
Meanwhile, Ethernet Powerlink is software based, and doesn’t require an ASIC chip to operate. Sandhoefner says this means it can reside in any processor or Ethernet-enabled port, and this logically makes it easier to implement on end users’ machines and systems. In addition, Powerlink networks can be arranged in a variety of topologies, such as daisy chains, trees, stars, or central hubs, and this helps them more easily meet the needs of various machines.
“Powerlink’s decentralized communications can broadcast a master position from one drive to all the others at the same time, which means a slave driver can pick up that master position, and calculate its own position in one cycle,” says Sandhoefner. “A centralized, hardware-based controller’s communication might only come along on its schedule, and this could take three cycles.”
For example, Eagle Manufacturing, Shelby Township, Mich., built Eaglematic, a 22 servo-axis, notch-and-cut-to-length fabrication system for multi-durometer extrusions for automotive window trim, which required a 400 μs communication cycle time to ensure 10 position updates per 0.04 in. at a 50 ft/min throughput. To achieve this speed, Eagle uses Ethernet Powerlink’s synchronous channel, and its time slices are assigned to nodes on the network. This helps split up formerly centralized motion controller jobs into smaller, faster decentralized tasks in the drive, which reduces network and CPU load because the network only needs to broadcast a master position once every 400 μs. This enables any of Eaglematic’s four fabrication stations to be moved onto the line within five seconds to meet two-door and four-door production changes with less waste. “This machine combines six different tools in one operating setup that moves the required tool into place when called on by the controller,” says Brent Short, Eagle’s owner and Eaglematic’s designer. “The tools are changed by a combination of the 22 servos without interrupting the flow of extruded material upstream.” Besides eliminating the expense of one or more motion controllers, Short adds that Eaglematic’s automation costs were 30% less than a conventional system.
Sandhoefner adds that the Ethernet Powerlink Standardization Group (EPSG) presently is developing a 1 gigabyte/sec (Gbps) version of Powerlink, and claims that its first nodes will be available in early 2008.
In addition, though hardware might initially be thought of as imposing some restraints, Hilscher North America launched its NetX 1588-compliant ASIC chip, which lets users switch between many of the motion bus protocols as needed, or use two chips to translate and convert data types. NetX uses a series of ARM microprocessors, downloads software to emulate motion bus processors, and has built-in pulse-width modulation (PWM) for motion control. “Because this chip can handle motion buses, it helps get automation manufacturers out of the protocol business,” says Phil Marshall, Hilscher’s COO. “This chip’s PWM allows users to build a motion controller with it, but they don’t have to build the board by themselves, and components are way down because there are no daughter boards.”
Motion buses and Ethernet have reduced former point-to-point hardwiring, and now wireless soon might reduce it even more. John Guite, division engineering manager at Parker Hannifin, reports that self-healing, wireless mesh networks also might soon be fast and well-switched enough for multiple motion axes.
Do-It-Yourself Testing
Because users and builders still face an unsettled, Wild West, buyer-beware environment for motion-based fieldbuses, they’re also forced to independently evaluate digital networks for their machines and applications.
To find the right controls for its TS-303 robot, for example, CBW evaluated several servo supplier’s solutions, and tested how well each met its own criteria. “We set up a test machine with a 10-lb payload on our EOAT, a 66-in. required distance and a target time of less than 140 msec, and then ran each solution over an eight-hour run time,” says Corwine. “Three firms met our criteria, but Indramat seemed to have better added features, support, and off-the-shelf components that didn’t need as much tweaking.”
Corwine adds that the hardest job in developing TS-303 was going from the PLC Ladder Logic programming to TwinCat’s sequential function charts programming. “We had to learn all of TwinCat’s software commands, but we had a lot of support, and it became easy over time,” he says. “Before starting this whole process, we decided we wanted IEC 61131-compliant components, so in the future we could use any other drive following the same standard, and be able to plug it into our controller. As an OEM, 61131 is a real benefit to us because it keeps suppliers competitive.” In fact, Corwine adds that 61131 already has enabled CBW to build a machine using Profibus and Siemens E&A’ Simatic servo controllers.
More Sound Advice
Though they disagree on which protocol to use, motion bus advocates advise users to let the needs of their applications direct their search for the right fieldbus.
“Ethernet fieldbuses have high bandwidth and fast cycle times, and many have low jitter as well, so make sure your fieldbus gives you the flexibility you need, such as various topologies, safety capability, diagnostics, compatibility with existing systems, and openness to the MES world,” says Evans. “Also, keep your network simple. You don’t need dedicated motion buses. Your fieldbus should be able to handle all the communication your machine needs. And, use one cable for motion, PLC, I/O, HMI, safety, engineering, and maintenance. Finally, a fieldbus is just an enabler, so you shouldn’t ask for the best way to use it. Instead, you should ask, ‘What’s the best way to configure my machine?’ The right fieldbus will be the one that gives you the robustness and flexibility to achieve this.”
Beckhoff’s Rawlyk adds, “Don’t be told, ‘This is the bus you have to use.’ Look at your application, add up its devices and nodes, and determine how coordinated they need to be. If you’ve got one gear axis, you can pretty much leave it alone, but continuously changing dynamic profiles will need a lot more coordination.”
Sandhoefner suggests users calculate the update time they’ll need on their networks, especially because this becomes more difficult as capabilities multiply and more sophisticated components proliferate on systems.
"There’s been an explosion in the number of nodes on many networks because there are so many more things users want them to do. For example, applications are going from variable-frequency drives to servos and continuous servos, I/O points are getting IP addresses, and format-adjustment axes are being added,” says Sandhoefner. “We’re seeing typical machines going from two to four axes to 20 to 40 or 100 axes. One of our users, a blow molding manufacturer in New Jersey, wanted to decentralize I/O points so his application could compete better with Asian manufacturers. However, he couldn’t afford to wire everything back to his main cabinet, plus his temperature control needed a faster response than the 30-40 msec his initial fieldbus provided. So, he calculated his required update time, Ethernet Powerlink was pre-scaled to meet it, and it was able to precisely maintain this application’s temperature.”
Less Wire + Less Time = Big Payoff
Despite the learning curve required to implement fieldbus, CBW has secured other substantial returns on its investment in digital networking. “Our previous servo and PLC suppliers didn’t think we’d ever be able to switch, but we did it. We went to a faster solution, and gained a lot of cost savings. We eliminated so many wiring and screw terminals and spent so much less time at power-up searching for mismatched wires and troubleshooting that we were able to save about 30% on labor to build and commission our machines.”
For example, TS-303 previously had 120 hardwired I/O points, but they now communicate via SERCOS. “We now have eight I/O points that go to a hub, and then back to the PC, and this saves even more wire,” says Corwine. “Our machines used to require 400 wiring hours, but now they only take about 220 hours.”
Early in TS-303’s development, this time savings was especially crucial because Corwine and another engineer were rushing to get it ready for a trade show. “We had a lot of late nights anyway, but the reduced wiring meant we were able to meet our deadline,” he says. “We’ve increased the number of machines we can build with the same amount of labor. We used to make four to six machines per month, and now we can build up to 10 per month.
Looking to the future, CBW already is testing Ethernet-based networking to its I/O points, but not to its servo drives just yet. Corwine says CBW is in the process of switching to Beckhoff’s EtherCat software for its I/O devices, and expects it to be even more robust than SERCOS.
“Also, less wiring and hardware means less maintenance for our end users. Just as we do, they appreciate having fewer terminal connections, less screws vibrating loose, and fewer wires breaking. We all experience less wear and tear over time.”
Motion Fieldbus To-Do List
So, you want to use a motion-related fieldbus in your equipment or discrete manufacturing application? Well, as with any control and automation project, there’s a laundry list you’ll need to complete to select the most appropriate protocol, software, and hardware. These primary tasks include:
- Recruit an in-house, multi-disciplinary research and implementation team.
- Evaluate capabilities of existing machines, systems, and applications in light of present and future business goals.
- Make sure any proposed fieldbus gives you the flexibility, topology, diagnostics, legacy compatibility, and openness to MES that your specific application needs.
- Specify the new network for safety, so you can migrate from hardwired safety.
- Set up machine performance criteria, and define equipment capabilities and benefits needed to meet them.
- Research all available components that can fulfill those criteria, and contact other end users for input.
- Secure local support from suppliers and/or system integrators needed to make the transition from old machine design and capabilities to the new one.
- Establish tests based on needed criteria, and conduct testing to determine which solution meets them best.
- Maintain long-term evaluation to make sure it continues to meet application’s evolving needs.
Full-length web version at www.ControlDesign.com/motionbuses
