Motion Control Resources
5 Real-Time, Ethernet-Based Fieldbuses Compared
IntervalZero Posted 03/18/2016
WHICH STANDARD STANDS APART?
By: Jeffrey Hibbard, CEO, KINGSTAR
The opinions expressed in this article are those of the author, and do not reflect in any way those of this website or the association.
Ethernet-based technologies have long been game changers. When first introduced, Ethernet innovated local area networks, and by extension broadband networks. Now, key components of the Ethernet ecosystem are being used to innovate industrial fieldbus networks. Typically, industrial automation and machine controls relied on proprietary networks to connect machine controllers to remote components. Now new protocols have emerged that capitalize on the Ethernet standard to deliver breakthrough price and performance.
Machine builders that adopt the right real-time Ethernet fieldbus standard will enjoy a marked competitive advantage in both price and performance when compared to any other fieldbus.
The challenge for these equipment manufacturers is that there are so many different Ethernet fieldbus standards competing to be the most valuable and viable. In a crowded market, it’s hard to know which standard to select. Adopting the wrong standard means unnecessary cost and sacrificing competitive advantage due to slower performance.
This paper examines five different important protocols that have emerged as contenders to offer the best price/performance open standard for real-time Ethernet fieldbuses. In alphabetical order, the standards being compared are EtherCAT, EtherNet/IP, Ethernet Powerlink, PROFINET IRT, and SERCOS III. There are other technologies that leverage Ethernet as well, but their components are not sufficiently published, downloadable or promulgated in the open source community to be considered standard and open. This paper does not address these proprietary technologies, including Mitsubishi’s CC Link Field and Yaskawa’s Mechatrolink III.
One of the five real-time Ethernet fieldbus standards has achieved a tipping point of acceptance. Cutting through all the marketing clutter, it becomes clear that EtherCAT offers both superior performance and market acceptance. Its performance is an order of magnitude better than Ethernet IP and Powerlink. And while PROFINET IRT and SERCOS III offer nearly equivalent performance characteristics, EtherCAT offers a more “open” solution at far lower cost than both PROFINET IRT and SERCOS III. From a technology and price/performance standpoint, EtherCAT is far superior. And the market agrees. EtherCAT has been adopted by 10 times more servo drive and IO suppliers than any of other standard.
Any machine builder that is considering the adoption of a real-time fieldbus technology should choose EtherCAT to deliver the best price/performance and the best long-term value. This white paper offers the evidence for this recommendation by providing a description of distinguishing features of each standard, comparing the performance and price/performance, and by tabulating market adoption rates.
Selecting the Most Viable Real-Time Ethernet Fieldbus Standard
The Evolution of Real-Time Ethernet Fieldbuses
For many years, Ethernet and the TCP/IP protocol have been used in the manufacturing arena to network control systems, management systems, and manufacturing cells on a manufacturing shop floor, but not for the controlling communications inside the actual machines and equipment. The machine controller itself and the communications to the actuators invariably demand the use of deterministic fieldbus, so TCP/IP is not suitable. Trying to use traditional TCP/IP protocol from machine control to the sensors and actuators has failed due to the inability to satisfy deterministic, real-time demands. It is simple as that.
However, the machine-builder industry (companies that build equipment like CNC, semiconductor, medical imaging, test bench, etc.) could see the value of reusing the hardware components typically used in a TCP/IP network setting. The explosive growth of the Internet created a universal standard for communication cards and cabling. A network interface card (NIC) and a TCP cable cost a mere fraction of the cost of an industrial fieldbus cable and DAQ card. Reusing this hardware could save companies 50% of the cost or more versus a traditional proprietary fieldbus configuration.
The economics of adopting Ethernet as a fieldbus are compelling because Ethernet components offer dramatically lower costs and are universally available. Relying on a proprietary fieldbus, a robot manufacturer, for example, must purchase the entire proprietary motion assembly from a single vendor: IO card at up to $400, proprietary cables at up to $30/linear foot and servo drive and motor at a premium. By using Ethernet-based standard protocol, the IO card can be replaced by the onboard NIC card ($0 additional cost) that comes installed on the PC, the proprietary cables can be replaced by inexpensive CAT5 cables, and the servo drives will be dramatically lower if the standard is strong enough to support multiple vendors. Equipment assembly for Ethernet components is much simpler too. Rather than having cable harnesses that are 4 inches in diameter at the PC interface, a simple CAT5 cable similar to the one that connects to your home PC is far more manageable. These are just a few of the economic benefits for Ethernet.
The industry soon realized that while TCP/UDP/IP protocol could never deliver the real-time, deterministic response required by industrial machine controllers, the ubiquitous and inexpensive hardware – network interface card (NIC) and CAT5 Ethernet cables –could be used to deliver the real-time response required by machine control applications. All that was needed was a new real-time protocol that was designed from the ground up to use the physical layers of the hardware, but could deterministically connect and communicate the machine controller to all the sensors and actuators in a machine. In some cases, companies could potentially use some augmenting hardware that delivered the needed determinism.
Since 2001, reputable industrial titans have introduced no less than five different Industrial Ethernet real- time protocols. They have been promoted as potential standards that would enable the reuse of the Ethernet protocol or Ethernet hardware, and promise to lower the cost to build a real-time controller for equipment. This proliferation of so many potential standards stems mainly from the fact that there are clearly different technical approaches to making it possible to reuse Ethernet hardware components like NICs and CAT5 cables to dramatically lower the cost and improve the performance of a machine, while making it possible for messages to be capable of supporting real-time, deterministic applications.
There is another reason that so many potential standards were being promulgated. Leading industrial control manufacturers were attempting to capitalize on the cost savings from Ethernet by extending their own standard to include Ethernet. In doing so, their customers receive some of the benefit from Ethernet but are still locked into the proprietary networks for the long term. Frequently forgotten in these discussions is the fact that it’s not just technical properties such as performance and transfer rates that count, it’s also the soft facts like easy implementation, openness, independence, risk avoidance, conformity, interoperability, long-term availability, and overall distribution that makes a standard gain acceptance and even thrive.
Next, we will examine the key features of the following five real-time Industrial Ethernet protocols, measure their efficacy as a standard real-time Ethernet-based fieldbus, and predict relative long-term viability:
- PROFINET IRT
- SERCOS III
The Long-Term Strategic Importance of Selecting the Most Viable Ethernet Fieldbus Standard
The IEEE states, “In technology development, open standards are the fundamental pillars for the worldwide economic growth and progression in all sectors of the economy.”
So what, really, is a standard, and why is it important to select the “best” most viable standard? Standards are published documents that establish specifications and procedures designed to ensure the reliability of the materials, products, methods, and/or services people use every day. Standards address a range of issues, including but not limited to various protocols that help ensure product functionality and compatibility, facilitate interoperability and productivity, and support consumer safety and public health.
Proprietary products deliver competitive differentiation in early stages of technology development. But there comes a point when open, standards-based solutions are necessary to establish the technological foundation on which more innovators can participate at lower cost, toward the goal of growing a richer and more robust market. Communications, computers, energy, and healthcare are among the many, many technology areas that have all demonstrated this pattern.
Standards form the fundamental building blocks for product development by establishing consistent protocols that can be universally understood and adopted. This helps fuel compatibility and interoperability and simplifies product development, and speeds time-to-market. Standards also make it easier to understand and compare competing products. As standards are globally adopted and applied in many markets, they also fuel international trade.
It is only through the use of standards that the requirements of interconnectivity and interoperability can be assured. It is only through the application of standards that the credibility of new products and new markets can be verified. Standards fuel the development and implementation of technologies that influence and transform the way we live, work, and communicate.
Most importantly, successfully adopted standards wring the cost out of products and drive breakthrough efficiencies, productivity, and price/performance. If the standard is not well adopted, the optimal savings and efficiencies will never be realized. In other words, there is a risk for any company that adopts a standard that is not broadly adopted by the supplier and user ecosystem because that company will always be at a competitive cost and efficiency disadvantage; it will never enjoy the maximum cost savings, performance gains, productivity, and efficiencies as compared to the best or most broadly adopted standard. Further, to stay competitive, a machine builder will usually be forced to eventually adopt the most strategic standard and this has all sorts of avoidable cost implications.
Introduction of Real-Time Ethernet-Based Standards
This section will briefly introduce the standard, the standards body history, and the key philosophical and architectural underpinning that distinguishes each standard from the other. This paper does not describe the entire protocol but shines a light on the technological component that simultaneously capitalizes on Ethernet yet distinguishes it from other open standards that also seek to capitalize on Ethernet. Following that, the standards can be compared side by side to uncover the relative strengths of each approach.
As we will see in the review of each respective protocol, there are profound differences in the technical approaches. This creates real separation between the contenders as to which standard is the most valuable and enduring. Some approaches rely on traditional Ethernet protocols and thus have a limited real-time capability and bandwidth, while others reuse the hardware components of the Ethernet and deliver breakthrough performance.
Specifically in the context of real-time deterministic fieldbuses, there are three different approaches that have emerged that allow the standard to deliver determinism on an Ethernet-based infrastructure:
Standard Software / Standard Ethernet
Based on TCP/IP: Protocols are based on standard TCP/IP layers with real-time mechanisms embedded in the top layer. These solutions usually have a limited performance range.
Open Software/ Standard Ethernet
New standard protocols are implemented on top of standard Ethernet layers. These solutions benefit from Ethernet evolution without further investment. However, to deliver the determinism, the standard must include a proprietary software controller at the OSI Layer 3&4 to reserve time on the network, otherwise latency can occur.
Open Software / Modified Ethernet
These solutions effectively create a new standard to take advantage of the existing Ethernet hardware, but require a new protocol and some hardware that guarantees determinism. The software is published freely and in the public domain. The hardware can be as complex as a special switch or as simple as an ASIC that fits into the slave device.
With that as context, each Ethernet standard will be described and assigned to one of these three architectural categories.
STANDARDS BODY, TITAN AND PHILOSOPHY
The EtherCAT protocol was initially developed by Beckhoff, and the standard has now been handed off to the EtherCAT Technology Group (ETG). Real-time results have shown that EtherCAT delivers the most deterministic response of any industrial real-time Ethernet system available. Being able to process 1,000 I/Os in 32.5 µs or 100 axes in 125 µs offers machine builders the opportunity to deliver breakthrough in machine performance at a lower price.
ARCHITECTURAL PHILOSOPHY: OPEN SOFTWARE/MODIFIED ETHERNET
With EtherCAT, all devices are networked with the bus master in a ring formation. During each cycle, relevant output data is extracted by the devices from the Ethernet data packets sent by the bus master. Input data is also stuffed into packets “on the fly“; these packets arrive again at the bus master upon reaching the end of the ring. This system was designed for centralized controller architectures with simple field devices in a Master/Slave configuration.
To guarantee that devices from different manufactures can work with one another, ETG has run plug fests to make sure that different drives and IOs from different manufacturers can all work seamlessly on a single installation. The fact that more than 500 I/O and drive vendors have now adopted EtherCAT for their slave device configurations speaks volumes about the power and efficacy of the protocol. The adoption was driven by large vendors in the semiconductor and robotics industries who refused to buy a drive unless it supported EtherCAT.
EtherCAT uses the telegram structure of Ethernet, but with an entirely different basic mode of operation. Within a communication cycle, a telegram is not sent to each station separately as in other Ethernet approaches, but rather a single Ethernet telegram runs through all stations/slaves. The data area in the Ethernet telegram divides into sections for real-time and general data. In the real-time data area the headers and process data of all stations are defined in consecutive sub-telegrams, thereby increasing the user data rate in the protocol (in motion control applications the user data rate in the shortest Ethernet frame of 64 bytes is usually below 15%). The slaves feature special ASICS or FPGAs which convert the incoming Ethernet-framed data into an internal so-called E-bus. As the EtherCAT slaves are only able to interpret EtherCAT frames, general data is tunneled in EtherCAT frames in order to guide it through slaves.
If a general data packet is too large to be transferred in one cycle, it is distributed across multiple EtherCAT frames. Tunneling and de-tunneling takes place in the master (Virtual Ethernet Switch) or in slaves with gateway functionality. The entire protocol processing is hardware-based. The slaves do not treat incoming Ethernet telegrams in the normal fashion, interpreting the contents and then copying the process data for forwarding transmission. Instead, the EtherCAT slaves read and write their process data from and to the predetermined location in the telegram while the telegram is passing through the slave. The EtherCAT mechanisms permit extremely short cycle times to be implemented.
STANDARDS BODY, TITAN AND PHILOSOPHY
Initially released in 2000, EtherNet/IP is an open industrial standard developed by Allen-Bradley (Rockwell Automation) and the ODVA (Open DeviceNet Vendors Association). The “Ethernet Industrial Protocol” is essentially a port of the CIP application protocol (Common Industrial Protocol), which was already used by ControlNet and DeviceNet, to the Ethernet data transfer protocol. EtherNet/IP is particularly well established on the American market and is often used with Rockwell control systems.
ARCHITECTURAL PHILOSOPHY: STANDARD SOFTWARE/STANDARD ETHERNET
EtherNet/IP is an application-layer protocol on top of TCP/IP. EtherNet/IP uses standard Ethernet physical, data link, network, and transport layers, while using Common Industrial Protocol (CIP) over TCP/IP. CIP provides a common set of messages and services for industrial automation control systems, and it can be used in multiple physical media. For example, CIP over CAN bus is called DeviceNet, CIP over dedicated network is called ControlNet, and CIP over Ethernet is called EtherNet/IP. EtherNet/IP establishes communication from one application node to another through CIP connections over a TCP connection, and multiple CIP connections can be established over one TCP connection. EtherNet/IP uses the standard Ethernet and switches, thus it can have an unlimited number of nodes in a system. This enables one network across many different end points in a factory floor. EtherNet/IP offers complete producer-consumer service and enables very efficient slave peer-to-peer communications. EtherNet/ IP is compatible with many standard internet and Ethernet protocols but has limited real-time and deterministic capabilities.
Ethernet/IP is the only one of the real-time methods described to be based entirely on Ethernet standards. In contrast to the other protocols, Ethernet/IP is not cycle-based but time-based, meaning that it merely requires that control commands are received by the field stations in time. This means the performance of the overall system can be made independent of network performance. Real-time delivery is safeguarded by three mechanisms which are all standards-based: UDP, Quality of Service (prioritization), and IEEE1588. To achieve real-time capability, accessible bandwidth is limited in order to avoid contention and latency. In other words, because EtherNet/IP is an application-layer protocol, it has a limited performance range that can deliver in real time.
STANDARDS BODY, TITAN AND PHILOSOPHY
ETHERNET Powerlink was introduced by B&R in 2001. Its goal was to provide standard Ethernet with real- time properties and allow universal solutions all the way down to demanding motion applications. Since that time, the EPSG (ETHERNET Powerlink Standardization Group) has promoted ETHERNET Powerlink and taken responsibility for its openness, continuous improvement, and independence.
ARCHITECTURAL PHILOSOPHY: OPEN SOFTWARE/STANDARD ETHERNET
ETHERNET Powerlink is a strictly cyclical protocol that organizes the access to a network as well as the synchronization of the devices. The communication cycle is divided into an isochronous phase for time-critical data as well as an asynchronous phase for transferring ad-hoc data. All of the devices on the network can always directly read all of the data from the other devices; detouring over a central bus master is not necessary. This protocol is equally suitable for local as well as remote control designs.
Because ETHERNET Powerlink is a cycle-based real-time system, it superimposes a time slot mechanism over the CSMA/CD mechanism. The master (controller) successively polls the slaves (drives) within an allocated communication cycle period. The remaining cycle time is left over for asynchronous data traffic, such as for configuration of the devices. Data transport occurs via a standard Ethernet telegram, with the Ethertype set to ‘Powerlink’ for the real-time data and to ‘IP’ for the general data. All stations (master and slaves) within a real-time segment interconnect via a standard Ethernet hub.
A special gateway links the real-time segment to the standard Ethernet environment. The IEEE1588 time synchronization method permits different real-time segments to be synchronized in a highly precise way, such as might be required for the control of multiple robots.
The idea behind ETHERNET Powerlink is to find the right balance between common automation demands and those demands that are specific to each application area. This has led to a widely accepted solution that gained some initial traction in the market. However, only a handful of servo drive vendors adopted Powerlink in spite of the fact that countries like China adopted it as a standard. This means that the standard is not actually broadly adopted. The resulting economic benefits were not as compelling because the price for servos remains high. In response to these conditions, the Chinese government has now updated its preferred standard and adopted EtherCAT. Having more vendor choices drives costs down.
STANDARDS BODY, TITAN AND PHILOSOPHY
PROFINET (“Process Field Network”) is differentiated into different performance classes to address various timing requirements: PROFINET RT for soft real-time, or no real-time requirements at all, and PROFINET IRT for hard real-time performance. The technology was developed by Siemens and the member companies of the PROFIBUS user organization, PNO. The Ethernet-based successor to PROFIBUS DP, PROFINET I/O specifies all data transfer between I/O controllers as well as the parameterization, diagnostics, and layout of a network.
ARCHITECTURAL PHILOSOPHY: OPEN SOFTWARE/MODIFIED ETHERNET
In order to cover the different performance classes, PROFINET makes free use of the producer/consumer principle and resorts to various protocols and services. High-priority payload data sent directly via the Ethernet protocol travels in Ethernet frames with VLAN prioritization, whereas diagnostics and configuration data, for instance, is sent using UDP/IP. That enables the system to achieve cycle times of around 10ms for I/O applications.
Clock-synchronized cycle times below one millisecond, as required for motion control applications, are provided by PROFINET IRT, which implements a time multiplex mode based on specially managed, hardware-synchronized switches. So-called dynamic frame packing (DFP) will in the future give users a new PROFINET variant designed to optimize cycle times, making use of the summation frame principle for a certain set of devices in the network.
PROFINET IRT for hard real-time also uses a time slot mechanism. Accordingly, a certain bandwidth is reserved for the real-time data traffic (IRT = isochronous real time), and the remaining bandwidth is available for asynchronous communication. The stations are interfaced via special switches integrated into the field devices rather than by standard Ethernet switches. These special integrated switches comprise a special ASIC controlling two or four ports at a data rate of 100Mbps.
PROFINET is widely used industrial Ethernet by Siemens and GE, and since it is embedded in their controllers and equipment, the market share is slightly inflated. Not many drive vendors support PROFINET IRT so it has not been broadly adopted by machine builders outside of Siemens and GE.
STANDARDS BODY, TITAN AND PHILOSOPHY
A freely available real-time communication standard for digital drive interfaces, SERCOS III not only specifies the hardware architecture of the physical connections, but also a protocol structure and an extensive range of profile definitions. For SERCOS III, effectively the third generation of the SERCOS Interface (SERCOS, Serial Realtime Communication System) that was originally introduced to the market in 1985, Standard Ethernet according to IEEE 802.3 serves as the data transfer protocol. This communication system is predominantly used in motion control-based automation systems. A registered association, SERCOS International, supports the technology’s ongoing development and ensures compliance with the standard.
ARCHITECTURAL PHILOSOPHY: OPEN SOFTWARE/MODIFIED ETHERNET
SERCOS-III uses the Ethernet physics (100Mbps) and the Ethernet telegram while retaining the existing SERCOS mechanisms. SERCOS-III is likewise based on a time slot mechanism in which bandwidth is reserved for the isochronous (real-time channel) and asynchronous (IP channel) data traffic. SERCOS-III works without hubs or switches. Each station has a special integrated ASIC or FPGA with two communication ports, enabling it to be connected via line or ring topology. Eliminating the switches means shorter cycle times can be implemented, though at the cost of flexibility in the network topology.
While specific hardware is categorically needed for the slave, a software solution is also feasible for the master. The SERCOS user organization provides a SERCOS III IP core to support FPGA-based SERCOS III hardware development. SERCOS III uses a summation frame method. Network nodes must be deployed in a daisy chain or a closed ring. Data is processed while passing through a device, using different types of telegrams for different communication types. Due to the full-duplex capability of the Ethernet connection, a daisy chain actually constitutes a single ring, whereas a proper ring topology will in effect provide a double ring, allowing for redundant data transfer. Direct cross-traffic is enabled by the two communication ports on every node: in a daisy chain as well as a ring network, the real-time telegrams pass through every node on their way back and forth, i.e.; they are processed twice per cycle. Hence, devices are capable of communicating with each other within one communication cycle, with no need to route their data through the master.
Other Real -Time Fieldbuses that Support Ethernet but Are Not “Open”
There are other fieldbuses, such as CC-Link from Mitsubishi and Mechatrolink from Yaskawa, that can use Ethernet components but do not meet our definition of open. These protocols and others were not considered in this white paper because the key elements of the protocol still demand the use of technology from the originating vendor. Even though they use Ethernet components, there are aspects of the protocol that are proprietary and the source code is not freely available.
Fieldbus Scorecard – Price/Performance and Market Acceptance
Now that the fundamental architecture of the various standards for delivering real-time Ethernet fieldbus has been defined, let’s compare, contrast, and ultimately assign a score to each standard.
When evaluating the viability of a standard, there are two primary dimensions by which a standard should be measured. First, the standard needs to compete favorably with proprietary and competing standard alternatives, and thereby deliver the performance, reliability, and efficacy that are typically ascribed to a proprietary solution. These lead to the economic breakthrough from lower costs of the components of the standard. Second, it is important to understand the long-term viability and market acceptance of the standard by both suppliers and consumers.
That being said, it is often difficult to clear away the marketing hype to understand the key distinguishing technological and price/performance differences and to measure market acceptance.
With respect to market acceptance, long-term viability depends on a standards body’s ability to make progress advancing the standard. Additionally, a prerequisite to successful adoption is that the standard must lower the barriers that allow vendors of the ecosystem to enter the market. This is done by making the standard independent and interoperable. In other words, the standard needs to be published and “open”. Equally important, however, vendors that adhere to the standard require low risk and low cost to making its solution comply with the standard. Regardless of the openness and successful adoption of the standard, success is really a measure of how interoperable the various vendor devices are. The strongest of these fieldbuses will be the one that has a very well supported compliance and interoperability program.
As is the case with most industry standards bodies, the Ethernet fieldbus standards bodies all stem from industry titans so they are well funded and capable. However, some of the resulting standards remain too dependent on the original titan, and the “standard” has limited efficacy because it incorporates too many legacy components from the proprietary precursor to the standard. Because of the dependence of proprietary components, the standard often fails to lower the risk and cost sufficiently to get many entrants to adopt the standard. So while the standard delivers the desired definition that enables interoperability, there are certainly differences in the costs of implementation that impact long-term viability.
As stated above, long-term viability is a function of how many vendors in the ecosystem support a given standard. The lower the risk and costs and the better the interoperability, the more suppliers will offer a solution. The more venders that offer a solution, the stronger and stickier the standard becomes. Once again, some standards bodies have done a better job of creating an independent, interoperable, low risk and low cost standard, and this has produced unstoppable momentum that will carry the standard far into the foreseeable future.
Performance and Efficacy Scorecard Comparison
Normally, there are countless technical dimensions of a standard that a customer/consumer must take into account when considering the adoption of a standard. All of these technical dimensions are really only significant if the fundamental performance delivered by the standards is nearly the same. But the performance of the standards varies widely, because the architectural differences render a wide performance difference.
Further, if a standard doesn’t compare from a price/performance standpoint, the standard will unlikely become viable in the long term. In other words, if the stan