February 2011, Vol. 238 No. 2

Features

Applying Wireless To EtherNet/IP Pipeline Automation

The use of Ethernet for industrial networking is growing rapidly in process control and SCADA systems such as oil and gas pipeline applications. The Open Device Net Vendor Association (ODVA) EtherNet/IP network standard is gaining popularity as a preferred industrial protocol. Engineers are recognizing the significant advantages that Ethernet-enabled devices provide such as ease of connectivity, high performance and cost savings.

While EtherNet/IP has many advantages, cable installation is often expensive, and communications to remote sites may not be reliable or cost-effective. Wireless Ethernet technologies have emerged that can now reliably reduce network costs while improving efficiencies.

However, applying these technologies is not a simple matter because industrial Ethernet systems vary greatly in terms of bandwidth requirements, response times and data transmission characteristics. This article will only briefly provide brief excerpts from a much more detailed discussion by the authors about applying IEEE 802.11a/b/g and proprietary frequency hopping wireless technologies to EtherNet/IP based networks for pipeline and other industrial automation systems.

Ethernet Industrial Protocol (EtherNet/IP) is a network protocol defined by ODVA. As an open standard, vendors may implement EtherNet/IP communications in their devices without licensing fees. Many vendors have adopted EtherNet/IP including Rockwell Automation, who selected the protocol as one of three preferred networks on their popular Logix controllers (DeviceNet and ControlNet are the other two).

An important part of the EtherNet/IP standard is definition of Common Industrial Protocol (CIP) messaging. CIP defines the information packet with recognition that the message attributes will vary as applications do. Thus CIP message definition takes into account a wide range of applications including programming/diagnostics, data collection, PLC data exchange and I/O communications. EtherNet/IP uses the standard 7 layer OSI model for protocol definition as shown in Figure 1.

Implicit Vs. Explicit Messaging. CIP defines two different types of connections. The first type is Explicit CIP which uses TCP/IP for its communications protocol. Explicit messages are unscheduled and use a request/response communications procedure or client/server connection. Examples of Explicit: executing a MSG statement between PLCs, HMIs, device diagnostics and program uploads/downloads.

The second type of CIP is Implicit. Implicit uses UDP/IP for its communication mechanism. Implicit connections are time critical, are scheduled and use a requested packet interval (RPI) parameter to specify the rate at which data updates.

Implicit connections use UDP packets for produce/consume data over an Ethernet/IP network. The UDP packets are usually multicast if there is more than one consumer of the data. This multicast address is assigned by the Ethernet/IP interface and is unique for each produced tag. Multicast IP addresses are used to make the network more efficient. A producer of data can produce data for multiple consumers. By using multicast packets, many devices can receive or consume this packet without the producer having to send it to each individual consumer.

EtherNet/IP I/O blocks may support two major implicit connection types: direct and rack optimized. A direct connection is a real-time, data transfer link between the controller and a single I/O module. Rack optimization is a connection option where multiple discrete I/O modules in a chassis can be consolidated to use a single connection. Analog modules typically can not be rack optimized and each analog channel uses a separate CIP connection.

Proper network design is critical for implicit networking systems in order to achieve predictable “deterministic” I/O performance and to ensure that I/O traffic does not “leak” outside of the automation network causing network degradation. ODVA recommends specific design strategies to ensure optimized network performance such as segmentation (isolating sub-networks), the use of managed layer three switches (IGMP snooping and multicast packet filtering) and high speed network infrastructure (100Base-T or faster). Wireless design is particularly critical as the wireless media is by nature slower than wired networks.

CIP Safety
CIP safety is an extension of standard CIP. It simply extends the application layer by adding a CIP safety layer to it. CIP safety has generally been used where reliable communications is a must or stop on a failure is required. It has many triggers in place to detect critical and non-critical errors and to close the connections in order to ensure a safety condition.

New specifications enhancements have been added to allow safety applications to have longer fault tolerances and the ability to assist in maintaining operations over wireless networks. Some of these enhancements include extending the RPI multiplier and the ability to configure the packet time expectation. These parameters are especially helpful in the wireless world where latency tends to be higher. This setting also could allow for retransmission of RF packets if required to ensure the safety packets get through. These changes lend themselves to make CIP safety well-suited for wireless communications.

Value Of Wireless EtherNet/IP
While EtherNet/IP networks grow in popularity, Ethernet infrastructure is not always easy, practical or cost-effective to install. The proper implementation of wireless Ethernet can reduce costs and in many situations improve reliability.

Wireless Alternative To Cable Installation. The cost of installing cable along a pipeline is a function of the material costs plus the labor charges. Cable installation costs have been estimated as ranging from $20-2,000 per foot, depending on installation challenges (distances, obstacles), environment and local labor costs. Factors that impact total cable installation costs include: 1) distance and number of locations, 2) conduit design and installation, 3) trenching, 4) fiber-optic cable and infrastructure (e.g. fiber switches), and 5) hazardous location regulations.

Once the total cost of cable installation is calculated, a comparison can be made to wireless. Similar to cable installation though, the total cost of wireless should be examined. This not only includes the wireless hardware (including wireless nodes, antennas and cables), but also antenna installation (if applicable) and personnel training. However, even when factoring in these additional costs, the savings realized by wireless is often dramatic and significant.

Additionally, wireless (when implemented properly), offers better reliability than cabled systems because there are fewer mechanical connections to fail. If the cable is severed during construction or damaged by environmental circumstances, production may be down for hours until the problem is located and corrected. Wireless also offers electrical isolation as fiber optics do, eliminating potential surge damage from ground-plane transients.

Finally, a significant benefit of wireless is reducing project time. Wireless systems are typically much quicker to install than wired systems.

Wireless As A Leased Telephone Line Alternative. For distant sites, leasing phone lines for Ethernet communications is common. Most phone companies offer a range of digital services from 56 Kbps up to multiple Mbps.

Unfortunately, many remote industrial sites (such as pipelines, well heads and storage tanks) may be too far away from the telecomm infrastructure to support higher speed services. This can sometimes lead to telephone line reliability issues, which often frustrates production and maintenance engineers as the phone companies are notorious for slow service to industrial customers. Private RF systems (such as spread spectrum) are managed by the end user and do not rely on any third-party services.

Cost is another problem when leasing of telephone digital circuits. Monthly charges may be as high as several hundred dollars per site, or even higher. Because these re-occur every month, a significant budget just for communication must be set. The initial cost of a private RF system may be higher, but there are no significant re-occurring costs. Return on investment for a complete wireless system may only take several months.

Wireless EtherNet/IP Reliability And Performance. Many factors affect designing reliable wireless EtherNet/IP systems that will meet performance requirements of automation systems. Perhaps the most important consideration is selecting (and correctly implementing) the best technology for the automation network. This section will focus on wireless design of specific EtherNet/IP network architectures.

Industrial wireless applications can be divided into two broad categories: Those requiring high-speed, low-latency performance, and those permitting slower speed with longer packet latency. Wireless technologies are available to accommodate both.

A common error, though, is assuming that faster technologies are better. If the application can handle slower speeds, then using relatively slower frequency-hopping technology may be the best approach. Frequency hopping is the most robust, especially regarding communications in high RF noise areas, and easier to implement. As applications demand higher speeds, more considerations and engineering challenges are typically encountered.

Wireless For Explicit Messaging Between Plcs. One of the most popular uses of wireless is in sharing I/O information between PLCs. As previously discussed, Explicit Messaging uses TCP/IP-based communications. Because these messages are unscheduled at the protocol layer, slower wireless Ethernet technologies may be used. MSG blocks in PLC ladder code may be programmed to accept long delays in transmission. If the application (process) is not time-critical, then a slower (but robust) frequency hopping technology may be the best choice.

There are many factors influencing how fast an explicit MSG may be executed. Generally, applications requiring 200 ms response time or slower are a good candidate for FHSS. Faster response times may require faster technologies such as OFDM available in IEEE 802.11a and 802.11g standards.

Figure 2: Wireless TCP/IP System using Explicit Messaging for Data Exchange.

Remote SCADA Systems. One of the most popular industrial uses of RF is for remote SCADA communications. Oil and gas extensively use remote SCADA networks where pipelines, remote pump stations and tanks are controlled and monitored from a central site.

Wireless is a very good alternative to leasing phones lines as previously discussed. The main challenge with remote SCADA is distance and terrain between the sites. Wireless technologies that support EtherNet/IP speeds require unobstructed line-of-sight (LOS) and adherence to the Fresnel Zone for optimal performance. Analyzing the terrain and LOS obstructions is vital in determining the feasibility of wireless EtherNet/IP. Fortunately, repeater sites are often available to achieve LOS and many FHSS radios have store-and-forward repeating capabilities.

Most EtherNet/IP SCADA systems do not require very fast communications as explicit messaging is used to communicate from the remote PLCs back to the central plant. Frequency hopping is often the best choice here because FHSS offers the longest range due to excellent receiver sensitivity and support of the 900-MHz frequency band.

Bandwidth is limited in FHSS systems, so careful examination of network traffic is prudent. There is sometimes a temptation to add in other types of communication (such as voice-over IP, surveillance video, Internet connectivity, etc.) which will quickly exceed the capabilities of an FHSS wireless system. IEEE 802.11-based systems offer the highest speeds for multiple use remote communications, but are limited in distance as their receiver sensitivity is lower and they do not support the 900-MHz band.

Figure 3: Remote Wireless SCADA System using EtherNet/IP TCP/IP.

Wireless For Ethernet I/O (Implicit Messaging). An emerging application for wireless is communications to distributed I/O blocks using EtherNet/IP. Wireless offers many advantages in these applications including elimination of mechanical coupling methods used in moving systems (e.g. rails, slip rings) and general cost savings due to reduction of Ethernet infrastructure.

Communication to EtherNet/IP I/O blocks can also reduce automation costs compared to using remote PLCs. Programming is simpler using I/O instead of remote PLCs because MSG blocks are not required in the main controller’s ladder program. But remember that implicit messaging is UDP/IP-based, not TCP/IP. Wireless networks must be carefully designed, and the plant RF environment more closely managed, to ensure reliable communications.

Several factors should be carefully considered before choosing this architecture including: 1) lack of remote PLC control (intelligence) in case of communication failure, 2) amount of I/O and required scan times (network traffic), 3) packet-handling ability of wireless technology, 4) efficiency of the RF technology with multicast UDP packets, and 5) 802.11 clear channel availability.

If circumstances are right, wireless EtherNet/IP I/O can be a significant cost saver and actually improve system reliability when correctly implemented, especially in moving systems.

ProSoft Technology has performed extensive tests of its Industrial Hotspot technology with EtherNet/IP and has many customer installations. The following information is provided only as a guideline toward predicting wireless performance and should not be relied upon for any other purpose. ProSoft Technology always recommends field testing to confirm wireless performance and reliability.

Predicting Wireless I/O Performance. Because EtherNet/IP I/O messages are scheduled, it is possible to predict scan time performance over industrial wireless systems if the following conditions are met: 1) packets per second performance of wireless technology, 2) wireless behavior in multipoint systems (handling of UDP Multicasts), and 3) number of CIP connections.

The first step in designing a wireless EtherNet/IP system is calculating packets per second which determines minimum wireless bandwidth requirements. Start by counting the number of CIP connections.

To calculate how many connections are in an I/O system, sum up all direct connections and rack optimized connections. To determine how many packets per second the system will be using, multiply each connection by two. It is multiplied by two because each CIP connection is bi-directional, meaning that during every Requested Packet Interval (RPI), a produced packet is sent by each end of the connection.

For example, if there are five direct connections and two rack-optimized connections (with six digital modules in each) equals 7 total CIP Connections, the total number of packets is then calculated:

7 CIP Connections x 2 = 14 Packets.

Note that the six modules in each rack that are rack-optimized only count as one connection. Rack optimization (if available in the I/O hardware) can significantly reduce wireless traffic.

Then multiply the packets by scan time (derived from RPI setting) to calculate packets per second (pp/s). Let’s assume that in the above example, the required RPI time is 20 milliseconds (actual RPI time is application dependant), we know there are 50 packets per second at an RPI time of 20 ms (1 / 0.02). We then multiply the 14 connections by the 50 packets per second to get the over-all packets per second rate:

14 Packets x 50 per second = 700 packets per second (pp/s).

The overall packets per second rate for 802.11 a/g radios can be in the thousands. However, it is best practice to not operate the radio network at maximum capacity. Rockwell Automation suggests reserving 10% of each adaptors bandwidth so it is possible to use its RSLogix 5000 software for remote programming.

It is also suggested that 30% of the radios packet per second rate be reserved for RF overhead. In a congested RF environment a radio contending for the RF medium will use valuable time if the radio determines the channel is busy by using its carrier-sense mechanism. If the radios determine the medium is busy the radio will not send any packets while it runs its back off algorithm and then re-accesses the channel. All this accounts for time when the radio could be sending packets but can not. A radio network that is in a highly congested RF environment can easily use 10% of its packets doing RF retries. RF retries can occur if a packet is lost or corrupt due to poor signal to noise ratio, antenna placement or multipath fading problems. Point-to-multipoint systems consume higher amounts of bandwidth Selecting a “clear channel” is good practice in wireless EtherNet/IP I/O networks.

The next step is to determine a reliable packet per second (pp/s) rate that the wireless technology will reliably support while keeping in mind that at least 40% should be reserved for other applications and RF overhead.

Impact Of Multicast UDP Packets On 802.11 Systems. Produced CIP packets are multicast over the Ethernet network to accommodate multiple consumers. In wired systems, managed switches with IGMP querying are recommended to direct multicast UDP traffic to only the segments that need them. This ensures that high-speed I/O data does not reduce performance of the plant Ethernet network, a major concern of IT managers.

Similarly, 802.11-based access points will rebroadcast multicast UDP packets to all active wireless clients. This represents a major problem because the unnecessary broadcast of high-speed UDP packets will quickly clog an 802.11 channel significantly reducing performance and even dropping UDP packets which will cause system errors.

One way to correct this problem and optimize wireless performance is to invoke IGMP Snooping and multicast packet filtering at the RF layer. By determining which devices are actually consuming the packets, the radio can build a consumption table and eliminate needless rebroadcasts. In point-to-multipoint systems, this feature can improve throughput up to 30% while significantly reducing dropped packets.

By applying multicast filtering to the 802.11 standards, it is possible to predict packet-per-second performance even in multipoint systems. For example, ProSoft Technology’s 802.11abg Industrial Hotspot has been determined to support 1,800 packets per second. This equates to a little over 1,000 packets per second available for EtherNet/IP CIP packets after subtracting 40% for RF overhead and other applications.

Wireless I/O Design Steps Summary: 1) calculate required packets per second of application, 2) select wireless technology capable of meeting pp/s requirement (with 40% reserve), 3) ensure good line-of-sight antenna placement, 4) select a “clear channel” – perhaps consider 802.11a 5 GHz band if 2.4 GHz is crowded, 5) utilize IGMP Snooping/Multicast filtering at RF layer to optimize performance, and 6) test system performance before commissioning.

Emerging Wireless Technologies
While this article briefly focused on widely available FHSS and IEEE standards such as 802.11a and 802.11g, there are several wireless standards on the horizon that promise higher performance and connectivity options for EtherNet/IP networks.

IEEE 802.11n. Not a ratified standard yet at this time, 802.11n promises several features that are attractive for EtherNet/IP communications including dual-band (2.4 GHz, 5 GHz) support, significantly faster packet transfer rates with a reported throughput up to 300 Mbps and RF propagation that actually takes advantage of reflected signals (quite common in industrial plants with lots of metal), using multi-input, multi-output (MIMO) antenna systems.

IEEE 802.16 (WiMax). While popularly known as an emerging cellular technology, WiMax technology will soon be available in the spread spectrum (license-free) bands including 2.4 GHz. WiMax offers high speed (up to 70 Mbps) at potentially long range. WiMax technologies may dramatically improve data rates to remote industrial sites and SCADA systems.

ISA100.11a. The ISA is working on the ISA100.11a standard for wireless enabled devices such as sensors. Operating in the 2.4 GHz band, the technology will “sense” existing 802.11b/g systems and work around them. While designed primarily for embedded devices, EtherNet/IP adapters and gateways will likely be supported.

Authors
Gary Enstad has a B.S. degree in electrical engineering and has been involved in wireless design and technical support for over nine years. He is Wireless Application Development Engineer for ProSoft Technology in its Madison wireless division. He can be reached at genstad@prosoft-technology.com.

Jim Ralston has been involved with the design and support of industrial wireless systems for over 12 years. He is Northeast Regional Sales Manager for ProSoft Technology and lives in the Pittsburgh, PA area. Jim may be reached at jralston@prosoft-technology.com.