Industrial Ethernet has exhibited rapid growth, but the reality is that fieldbuses and sensor networks still retain a large position and many potential network nodes remain hardwired. End users understand and seek the advantages of a harmonized network based on Ethernet, IP, and the related open ecosystem.

Benefits include reduced complexity and cost by minimization of gateways and elimination of hardwiring; expansion of the qualified labor pool; and improved optimization and maintenance opportunities via Cloud connectivity and analytics. Unfortunately there remain barriers to the single network vision.

The mix of EtherNet/IP, fieldbuses, and hardwired nodes persists due to application constraints near the network edge. Fortunately, the open market is bringing new techniques to bear. Unprecedented IoT opportunities led IETF to create IP-stack optimizations for constrained devices and networks (RFC 7228). Enhancements are applicable to both low power wireless (6TiSCH) and wired networks, eliminating TCP overhead (UDP-only), compressing messages (6LoWPAN), expanding the address space (IPv6), optimizing security (OSCORE), and shrinking the Web server (CoAP).

Numerous industries flooded into IEEE to develop enhancements for enabling Ethernet to displace other networks at the edge.

The resulting Single Pair Ethernet suite offers reduction in wiring, node cost, size, and power consumption, delivering communication and power over a single pair. A 1000 meter variant targets process plants and other large sites.

This deterministic Ethernet bus-variant targets very constrained devices, such as in-cabinet components. This article proposes a set of enhancements, adopted from or inspired by IETF and IEEE, to extend EtherNet/IP into constrained applications, further enabling the single network vision.

Candidate constrained application areas for EtherNet/IP systems.

Proposed enhancements

This article presents a set of proposed enhancements, many adopted from or inspired by IETF and IEEE, making it possible to use EtherNet/IP on constrained devices and networks, thus enabling the single-network vision–where all devices in an industrial plant can communicate with the same set of protocols.

Industrial Ethernet solutions have been growing rapidly, with EtherNet/IP emerging as a leader. However, traditional fieldbus solutions and sensor networks still have advantages, retaining a large position. Additionally, many potential network nodes remain hardwired.

The single network vision

Many end users understand and are seeking the advantages of a harmonized network-based on Ethernet, IP, and the related open ecosystem. This desire is expressed within organizations across many industries.

The Automotive industry has been working toward an all-Ethernet vehicle. Multiple, incompatible networks (including Ethernet) were added to vehicles, until finally the cable harness became the third heaviest component in a vehicle – adversely affecting the mileage. The industry formed the OPEN (One Pair EtherNet) Alliance to promote a variety of Single Pair Ethernet (SPE) solutions, first creating a specification around a proprietary solution and the later developing open standards solutions within the IEEE. Automotive continues to push SPE solutions to both higher rates (e.g., LIDAR and radar) and low bit rates (e.g., door switch and headlamp assemblies). Ethernet now pushes to the edge to displace other networks in the vehicle.

The Digital Ceiling is a partner ecosystem promoting smart LED lighting and associated applications. The transition to LED enables recent Power over Ethernet to both power the lighting and to communicate with the lighting over a single cable. Once smart lighting is in place, sensors (occupancy, security, etc.) can be integrated, and wireless communication can be added, to service both the room and the occupants.

Process Automation end users from NAMUR promoted Ethernet (and IP-based automation protocols including EtherNet/IP) for instruments and related devices. Vendors have responded to deliver the technology, organizing in the Advanced Physical Layer organization, and involving ODVA and their peer organizations. Work in IEEE on Single Pair Ethernet (SPE) is anticipated to deliver Ethernet communication and power over cables as long as 1000 meters (including cables) into intrinsically safe environments.

These initiatives exist due to compelling advantages of a single network:

• Higher performance for a similar cost (compared to the displaced networks)
• Elimination of costly application-specific gateways
• Leverage of a large existing ecosystem (protocols, security, network switches, etc.)
• Reduced installation, maintenance, and management complexity
• Simplified integration with cloud applications
• Reduced interoperability issues

Example UDP-only stack.

Barriers to single network vision

The mix of Industrial Ethernet, fieldbuses, and hardwired nodes persists due to application constraints near the network edge. Many of the edge devices are constrained by very low cost. It does not make economic sense to have the network cost more than the rest of the device. Some devices are constrained by size. They are very small (photo-sensors) and the network components must fit. Other devices are constrained by power. They are either densely packed or battery powered. Low power consumption becomes a requirement.

Total system cost may be constrained by the amount of wiring, which includes running both communication and power wiring. Wiring cost may exceed the cost of certain devices. Delivery of adequate power plus communication over a single cable becomes a necessity. Certain applications are constrained by the disadvantages in having any wiring at all. Low power wireless solutions exist, but not as part of a harmonized system.

For other applications, long cable distances are required (1000 m or more for industrial fieldbus) and constrained by existing Ethernet length limitations. This may be associated with a requirement for compatibility with Intrinsic Safety for explosive environments.

Addressing the barriers

These barriers are in part addressed with the development of new constrained Ethernet and IP technologies, for wired and wireless usage, emerging from standards organizations such as IEEE and IETF. This does not automatically imply that Industrial Ethernet protocols can be used on such technologies. The industrial Ethernet protocols have assumed operation over traditional platforms. So, protocols such as EtherNet/IP must be adapted to match with new constrained devices.

IETF contributions

We adopt terminology from the IETF paper named “Terminology for Constrained-Node Networks”. Nodes and networks have separate characteristics that make them constrained.

Constrained Node characteristics:

• Low cost
• Small size
• Limited memory [Flash, RAM], and processing resources
• Limited power and energy [battery size or scavenging]
• Limited upper layer services
• Low weight

Constrained Network characteristics:

• Low bitrate or throughput
• High packet loss
• Variability delivery rate
• Asymmetric traffic
• Small packet size
• Limited availability [device sleeps]
• Limited upper layer services

The IETF has introduced a number of related standards and drafts for constrained nodes and networks. These include RFC 7252 (CoAP), RFC 7400 (6LoWPAN), and derivative documents. Techniques are introduced and are applicable to both low power wireless and wired networks, and include:

• Elimination of TCP overhead (UDP-only)
• Compression of message headers
• Expansion of the address space (IPv6)
• Optimization of security (OSCORE)
• Shrinking the Web server (CoAP)

The IETF suite of IP standards is intended to solve IoT needs, but these needs are similar to Fieldbus needs.

IETF 6TiSCH stack supporting EtherNet/IP.

IEEE contributions

Within IEEE, a family of Single Pair Ethernet (SPE) standards have been developed. These enable communication and optional power over a single pair, facilitating reduction in wiring, node cost, size, and power consumption. A number of the standards are already complete, including 100BASE-T1 (100 Mb/s), 1000BASE-T1 (1000 Mb/s), and optional power known as PoDL.

Another family member, IEEE P802.3cg, is in development and has estimated completion in late 2019. The standard will introduce a pair of 10 Mbit/s SPE PHYs targeted for constrained applications. Numerous industries sought Ethernet enhancements to displace edge networks and are contributing to the standard.

IEEE P802.3cg includes the following two PHYs.

10BASE-T1L:

• Addresses long distance
• Targeted at process automation instruments
• 1000 m, intrinsic safety compatible, legacy wiring

10BASE-T1S:

• Addresses low cost control
• Targeted at replacing:
• CAN, CAN FD, MOST, and FlexRay protocols in automotive
• Hardwired components for in-cabinet industrial automation
• I2C and SPI in data centers
• 25 m multidrop option
• eterminism by PHY-level Collision Avoidance (PLCA) protocol

Stack modification summary for EtherNet/IP for constrained nodes and networks.

Constrained EtherNet/IP apps

Several new application areas for EtherNet/IP at the network edge are constrained applications for field devices. From basic control, up through the enterprise, 100BASE-TX Ethernet and emerging 1000BASE-T Ethernet is suitable and likely to remain in place. At the field level, these are not well suited to meet the listed constraints.

One important application area is Process Automation. Users are requesting Ethernet to the instruments. Constraints include achieving long distance communication and power over single pair cables while maintaining compatibility with intrinsic safety.

A related application area is Low-power Wireless. This is complementary for Process Automation, but has wider usage in Discrete and Batch. The constraints are related to achieving long battery life and the related constraint of fitting into small packet sizes.

Another important application is In-cabinet components. Here the transition is from hardwiring to networked devices. Very strict constraints exist for low cost, small size, and low power.

Another potential application area is On-machine components. The challenge is to reduce the number of cables, while delivering adequate power. Often the component constraints are low cost and small size. Higher performance SPE may be applicable. This application is not discussed further in this article.

UDP-only option

EtherNet/IP currently requires the use of both TCP and UDP in the network stack of a device. While TCP does have advantages e.g. when it comes to the guaranteed delivery of messages, it also adds additional resource requirements. TCP being connection-oriented, the stack needs to keep track of each single client connected to the device and needs to buffer messages in case packets are lost on the network and need to be retransmitted. This means TCP needs additional memory and processing compared to UDP.

Using UDP-only can result in a substantial reduction in stack complexity and messages. A UDP-only prototype demonstrated about 30% savings in Flash and RAM in a constrained device. Such a reduction can mean a device uses one size smaller MCU, reducing cost, size, and power.

One way to enhance performance is the elimination of TCP-based connected messages and the associated encapsulation session. Besides the reduction in stack size, the reduction in messages benefits low power wireless applications by extended battery life.

Note that other constrained stacks, such as CoAP from IETF, also implement and rely on UDP exclusively. Support of UDP-only would align EtherNet/IP with IoT requirements and evolution.

We propose to add optional support into the EtherNet/IP adaptation for UDP-only. Since CIP Security requires both TLS and DTLS, we also propose to add optional support for into the EtherNet/IP adaptation for DTLS-only. It is also proposed that ListServices be extended to identify the capabilities of UDP-only, TCP+UDP, or both.

A number of implementation options exist. One option is to base the solution on the existing Volume 8 – Secure UDP-only. Extension would be necessary to support both secure and standard UDP-only, and to support a more complete set of services via UCMM. It would also be beneficial to support a unified capability discovery method. Encapsulation/header compression

In relation to typical CIP application payloads, the encapsulation and CPF headers create a significant overhead in EtherNet/IP messages. Low power wireless networks, such as IETF 6TiSCH, have constrained packet size, which is not able to carry the existing encapsulation header in a single wireless packet. Extra messages (multiple per message) substantially reduce battery life.

The EtherNet/IP encapsulation and CPF headers can be compressed by adopting the same technique as is used in 6TiSCH. The process is called “eliding”. Many of the header fields are options that rarely change. Addition of bit fields to indicate the optional presence of byte or word fields allows a compression on the order of 8 to 1. Eliding is lossless in function and backward compatible.

Prototyping has shown excellent compression for ListIdentity, ListServices, General UCMM, and Class 0/1. Note the ability to compress multiple headers in a single message.

Existing Volume 8 – Secure UDP-only introduced the “CPF for UCMM” to replace and reduce the Encapsulation header from 40 bytes to 16 bytes. The proposed compression could reduce this even further.

It is proposed to evaluate compression options for:

• Encapsulation header
• CPF for Class 0/1
• CPF for UCMM
• CPFs used within CIP services

Constrained physical layers

EtherNet/IP does not support appropriate physical layers for several constrained domains:

• Process Automation wired instruments for APL
• Process Automation companion wireless instruments
• In-cabinet components

We extend three new PHYs:

• IEEE P802.3cg 10BASE-T1L PHY for Process Automation wired instruments
• IEEE P802.3cg 10BASE-T1S PHY for In-cabinet component usage
• IEEE Std 802.15.4-2015 PHY for Process Automation wireless instruments

Constrained communication profile

Current EtherNet/IP communication does not support constrained device and network requirements. It is proposed to develop a constrained EtherNet/IP communication profile as shown. Note that this differs from the concept of a device profile.

The minimum device object model uses the same base objects for constrained EtherNet/IP, but minimizes the implementation of the base objects. There is optional compression of Encapsulation and CPF headers. There are minimized CIP transports over UDP, supporting only UCMM + Class 1.

As part of reducing the overhead in a device, the objects are minimized by limiting the optional features. The Connection Manager is shown as an example. Attributes and services are minimized. The communication methods are minimized. The simplifications still retain the required interoperability.

Constrained over 6TiSCH network

Quoting the EtherNet/IP adaptation of CIP: 9-2 Data Link Layers. Though this specification is called “EtherNet/IP”, Ethernet is technically not required. The EtherNet/IP protocol may be used on any media that supports the transmission of the Internet Protocol.

Wired EtherNet/IP for Process Automation (under APL) calls for a complementary low power wireless solution. Best suited is IETF 6TiSCH with its open UDP support. EtherNet/IP over 6TiSCH requires constrained features and 802.15.4 link support. EtherNet/IP lack of IPv6 complicates 6TiSCH solutions, and precludes other usage.

We suggest to specify necessary enhancements for EtherNet/IP over 6TiSCH. We can leverage the Constrained EtherNet/IP enhancements:

• Add 802.15.4 MAC and PHY
• Add 6TiSCH router and network management
• IPv4/6 mapping to integrate 6TiSCH devices into IPv4

A full support of IPv6 requires substantial changes. Hence, we suggest revisiting full IPv6 EtherNet/IP enhancements and security additions at a future date.

There is good motivation for adopting the IETF 6TiSCH stack for EtherNet/IP. It is an open (not industry specific) wireless standard, supporting IP-based communication. It provides a self-organizing mesh network that is robust, reliable, and reduces engineering. Furthermore it is based on IEEE Std. 802.15.4 radio hardware, which leads the market.

Enhancements are important for integration of constrained EtherNet/IP over 6TiSCH. The physical layer needs to be converted between Ethernet and 802.15.4e. The network and transport layers need to be converted between IP/UDP and 6LoWPAN. The addressing needs to be converted between IPv4 and IPv6.

Security is required for constrained EtherNet/IP over 6TiSCH. One option is to utilize two separate security systems. Link layer security is mandatory for 6TiSCH. Additionally, 6TiSCH requires IETF OSCORE security for network join. CIP Security can provide optional security terminating at the 6LBR. Note that CIP end-to-end security is precluded by any 6LBR processing of the application layer (Encapsulation Layer compression). While CIP Security could operate end-to end, DTLS increases the wireless node overhead. Double processing would have a negative impact on battery life.

Two new objects are required for support of 6TiSCH networks.

804.15.4 Link Object

• Utilized in both 6LBR and WSNs
• Similar to Ethernet: Interface Speed, Flags, Counters, State, Label, Capabilities, Physical Address
• New: RF characteristics

6TiSCH Wireless Network Management Object

• Utilized in 6LBR
• Includes network status information; network topology information; network routing information; device join & leave

David D. Brandt, Senior Principal Engineer, Rockwell Automation, Dayin Xu, Senior Research Engineer, Rockwell Automation, Dr. Jörg Hähniche, Head of Development & Integration Services, Endress+Hauser Process Solutions and Matthias Dietrich, Software Engineer, Endress+Hauser Process Solutions