Network technologies have existed for many decades, and one of the most famous by far has been Ethernet. However, Ethernet is quickly becoming unsuitable for large-scale industrial networks that require reliability and low latency.
In this article, we will look at how Single Pair Ethernet (SPE) and time-sensitive networking can help.
Introduction
When it comes to choosing a network technology, there are many to choose from — some network technologies are focused on the physical layer, while others are focused on the protocols used on that physical layer. For example, Ethernet describes connector types, voltages, and data rates, while TCP/IP describes how data should be transmitted over Ethernet. Other examples of commonly used networks include CAN, I2C, ProfiNet, EtherCAT, Wi-Fi, Bluetooth, and 5G.
All of these technologies have their own respective advantages and disadvantages, which is why choosing the right network technology for a given scenario is extremely important. For example, Wi-Fi is very good for domestic environments offering high data rates and low ping, but it cannot operate over great distances, whereas wired Ethernet can provide high data rates and low ping over great distances. However, Ethernet is wired, meaning that only devices with access to Ethernet ports can access the network.
Why does traditional Ethernet fail in industrial environments?
Ethernet is a good wired networking solution if high volume and high data transfer speeds are needed, but it is very difficult to use in industrial environments. To understand why, we first need to look at how industrial processes differ from standard home and office networks.
Homes and offices can have large bandwidth requirements when users transfer files or stream video, and Ethernet is very good at providing this. However, industrial processes (such as a milling station or an automatic press) produce real-time data that needs to be streamed to some endpoint with minimal delay. Standard Ethernet networks do not account for time-sensitive information and simply send frames whenever they can. This means that real-time systems can very quickly loose synchronization between a source and a receiver.
Timestamps can be added to frames, but real-time processes cannot afford to spend the time trying to figure out what data frames match the current time. Furthermore, standard routers do not prioritize data frames based on whether they are time-sensitive or just standard data frames, so time-sensitive frames can have random amounts of latency added to them.
To make matters worse, industrial systems can have thousands of devices all producing data in real time, and any centrally controlled network would need to efficiently differentiate between time-sensitive packets and standard data packets, all while ensuring that the time delay between a device sending data and the receiver is well-defined.
Another challenge that Ethernet cables bring to industrial environments is their size. Typical Ethernet cables use multiple twisted pairs (eight wires/four twisted pairs), which creates a heavy cable that has a large bend radius. Trying to route many hundreds of cables around cable trays can become challenging, and trying to identify cables can also be a major challenge for any technician. To add to the difficulty, those same cables will need to be rated for harsh industrial environments, including resistance to vibration, temperature, and potentially moist and/or corrosive atmospheres.
What solution do current industrial environments use?
Ethernet is highly desired by industrial sites, and there have been multiple solutions based around Ethernet to try and bring the advantages of Ethernet to industrial applications.
One compromise commonly used is to create bridges between localized device networks and the larger industrial network. For example, a machine station could deploy a two-wire network bus between several PLCs and sensors, while a single bridge between this bus and the greater industrial site provides a method for remote access and monitoring of that local bus.
This method allows for devices inside the local network to communicate with each other over a time-sensitive network while allowing for remote access. However, it still faces the challenge of time sensitivity between the local network and any other system on the wider industrial network. It also has the challenge that latency is added as a result of the bridge, which needs to convert outgoing and incoming data packets.
What would a solution need to do?
The problem with creating an industrial network is the need for solutions in both hardware and software. The use of large heavy cables makes it difficult to route and manage, while off-the-shelf solutions are unable to create a network with high bandwidth and time-sensitive frame prioritization. Furthermore, the method in which Ethernet works means that the path between a sender and receiver is not well-defined, and this can add a latency with a size that cannot be accurately determined. This is where another important factor comes in: determination.
By their nature, Ethernet networks are not deterministic, meaning they send messages out without any consideration for the path that the message takes, what routers handle those messages, and the latency between nodes. This is one of the major reasons why standard Ethernet cannot be reliably implemented in industrial environments.
Meet Single Pair Ethernet — the hardware solution
Single Pair Ethernet is one solution for implementing Ethernet protocols in industrial environments. As the name suggests, it consists of a single twisted pair for both transmission and reception of data.
The use of a single twisted pair means that data rates are slower over SPE compared with standard Ethernet, but the removal of six wires allows for cables to be made thinner and lighter. Furthermore, the use of only two connectors simplifies cable installation, and the reduced cable size allows for tighter bends, which allows for more compact cabling.
Another advantage to a single-twisted-pair cable comes from the lower data rate; cable lengths can be much longer.
Many industrial processes produce real-time data, but they do not produce vast quantities of data (unlike video streaming). For example, a system that produces 1,000 data points per second at 8 bits in size would require a bandwidth of only 1 kB/s (while this does not include control frames, the amount of data being sent is still extremely small).
The low data rate means that voltage transition times (between high and low) can occur more slowly. This not only helps to increase data transmission reliability but also allows for larger voltages to be used, which allows for longer cable runs. This is especially advantageous in industrial applications in which factories can span hundreds of meters in size.
SPE cables use less wiring than traditional Ethernet cables. They have tighter bend radiuses and are also lighter and cheaper to manufacture, too. This is advantageous in applications that need to install large numbers of cables in tight spaces (such as cabinets and overhead cable trays). The smaller cable also allows for reduced connector sizes, which helps to reduce the size of devices connecting to the wired network.
Finally, SPE (just like standard Ethernet) can provide power over its data lines, meaning that connected devices can also draw power from the cable. The combination of Ethernet and power into a single twisted pair creates a highly convenient cable for use in industrial applications that can see many thousands of devices needing a wired connection.
Time-sensitive networks over SPE — the software solution
Single Pair Ethernet solves the physical challenges faced by Ethernet cables, but it does not solve the software challenges of high latency, non-determinism, and prioritization. This is where time-sensitive networking comes to the rescue.
Something is said to be time-sensitive if the passage of time has an impact on the quality or usefulness of data. For example, the speed of a vehicle to the driver is time-sensitive, as the driver needs to know what the speed of the vehicle is in that instance. Any delay in receiving this information makes the recorded speed useless, as it no longer represents the speed of the vehicle in that moment.
Industrial processes are exactly the same. Most of the data gathered is time-sensitive to some degree and is often processed as soon as it is generated. For example, a paper mill will use a radioactive beam to measure the thickness of paper as it is being rolled, but this information is time-sensitive, as it needs to be processed as soon as it is read. Any delay in processing this thickness reading would result in production mistakes (i.e., too thick or too thin paper).
To fix this issue, IEEE has developed a new Ethernet protocol that integrates time sensitivity into it so that data packets being sent over a network that are marked as time-sensitive will be prioritized over other traffic. This ensures that time-sensitive packets are not stored in a buffer before being sent out to their destination.
Furthermore, the non-deterministic nature of Ethernet means that when time-sensitive packets do arrive, it is difficult to tell exactly how long ago those packets were sent and what delays those packets faced. As such, the new time-sensitive networking protocols developed by IEEE provide deterministic networking whereby the path between a sender and the receiver are well-established and understood. The length of the network is also understood along with the time delays between nodes, and this creates a message whose latency is precisely understood at the receiver.
The three IEEE SPE TSN standards
Like traditional Ethernet (which has different categories depending on the speed), SPE also falls under several categories depending on the speed. To better understand the difference, the table below shows the three main categories for SPE with TSN.
The various categories clearly show that increasing the bandwidth results in a smaller cable length but comes with the added advantage of increased power delivery to devices. As previously discussed, many real-time systems will not be producing vast quantities of data; only the data that is produced needs to arrive in a timely manner. This means that lower speeds (such as 10 Mbps) can provide network coverage over extreme distances while still providing time-sensitive capabilities. Shorter runs of cables that interconnect PLCs and machinery sensors could very easily utilize higher-bandwidth cables while simultaneously taking advantage of Power over Ethernet to remove the need for additional power cables.
Conclusion
Many network solutions exist for the industrial sector, but these are often proprietary and rarely work with each other. Furthermore, the vast number of network technologies also means that different sensors and devices for industrial applications may find it difficult to communicate with each other (often requiring network bridges).
Considering that Industry 4.0 will see large-scale integration of sensors along with real-time data monitoring, a unified network technology is needed. Ethernet is already in widespread use in offices thanks to its large bandwidths and adaptability, but its large cable size and non-deterministic nature makes it hard to integrate into industrial environments.
As such, the combination of SPE and TSN as described by IEEE could be the industry’s standard network solution. Cables that consist of only two wires and a shield would be significantly smaller than Ethernet cables while simultaneously being easier to route. The standardization of networking protocols also allows for various equipment from multiple companies to all operate on the same network with no need for converters or bridges.
Furthermore, the ability for SPE to work with time-sensitive networks means that vast quantities of data being generated in future industrial sites can be processed in real time. Such real-time data can then be combined with AI and other advanced technologies to help tailor industrial process to be as optimal as possible while simultaneously giving site operators more power than they have ever had.
