Will Ethernet cancel CAN?

10BASE-T1S Ethernet is the perfect protocol for implementing zonal electronic and electrical architectures in automobiles. 10BASE-T1S runs 2x to 3x faster than a controller area network (CAN) topology and eliminates the need for CAN to Ethernet protocol translation gateways, saving cost and weight. Compared to CAN’s 2 Mbps and CAN-FD’s 5 Mbps, 10BASE-T1S delivers 10 Mbps. Also, by architecting Ethernet throughout the vehicle, 10BASE-T1S streamlines firmware over the air (FOTA) updates, a requirement that is of paramount importance for the growing software defined vehicle (SDV) trend.
Ethernet evolutions
Ethernet has been the preferred wired connectivity method for desktop computing and data center connectivity for over forty years. Over that time, the IEEE Ethernet 802.3 standard has evolved and been adapted for several different use cases, such as Industrial Ethernet, which features a low latency and determinism. Automotive manufacturers have also been keen to embrace Ethernet's resilience and robust characteristics.
There are very few downsides to multi-gigabit throughput Ethernet. However, using four twisted wire pairs in a Cat 5e/Cat 6a cable makes for sizeable cables and connectors. Cable and connector size has never been an issue for most commercial applications. However, they are limiting for industrial and automotive applications, as are the cable weight and the maximum bending radius. Compared to a typical four-pair Cat 6A cable, a single-pair twisted pair cable weighs 60 % less.
IEEE 802.3cg 10BASE-T1S specification
Single-pair Ethernet is already specified by 10/100/1000BASE-T1 for full-duplex point-to-point applications, having been ratified years before IEEE 802.3cg. The IEEE 802.3cg specification defines two 10 Mbps variants - 10Base-T1L (full duplex, < 1 km) and 10BASE-T1S (< 25m). 10BASE-T1S is then further divided into two sub-variants – multi-drop, half duplex, up to a minimum of 25 meters (minimum of 8 nodes) cable distance, and point to point, full-duplex, up to a minimum of 15 meters cable distance. onsemi’s focus is with the multi-drop variant of 10BASE-T1S, because within the industrial and automotive domains, the multi-drop variant is particularly attractive for a wide range of medium-speed applications for several reasons.
Multiple legacy bus protocols are typically encountered within these application domains, adding communications complexity, and requiring additional protocol gateways and converters that introduce latency, add cost, and consume power.
Industrial and automotive applications for 802.3cg
Within legacy automotive in-vehicle networking (IVN), for example, controller area network (CAN), FlexRay, and local interconnect network (LIN) have evolved to create a complex networking architecture. As the need for sophisticated, high-bandwidth, and interdependent vehicle functionality grows, automotive vendors are increasingly turning to Ethernet to provide a gigabit backbone. Adding single-pair Ethernet will considerably simplify IVN, facilitating a single "all Ethernet" protocol throughout the vehicle, from edge sensor nodes to the central computing platform.
The above image illustrates the fundamental differences between standard 10/100/1000Base-T (IEEE 802.3/u/ab) Ethernet and 10BASE-T1S multi-drop". Rather than requiring network switches to identify point-to-point nodes and arbitrate network traffic to avoid collisions, 10BASE-T1S has a multi-drop topology with a physical level collision avoidance mechanism (PLCA). This approach reduces the PHY requirements, reducing the component cost and negating the need to route cables from a switch to individual nodes. Instead, lightweight SPE cables are "daisy-chained" between nodes, reducing cable length, weight, and cost.
Automotive shifts to E/E zonal architectures
In the last decade, the automotive industry has undergone a significant transformation. The move towards battery electric and hybrid vehicles, coupled with an increasing emphasis on advanced driver assistance systems (ADAS), has presented several challenges. With electric vehicles, their overall weight directly impacts range. Reducing cabling is paramount, with several miles of copper wire weighing up to 132 lbs (60 kg) (1) in an average car. Also, many ADAS functions, for example, adaptive cruise control, emergency braking, and lane departure assistance, are interdependent software applications. The rise of the software-defined vehicle decouples the hardware from the software. It is becoming a service-oriented architecture, a long way from the "signal-oriented” approach of the past. Rather than the domain-based electronic/electrical (E/E) architecture that connects sensors and actuators by function, a zonal architecture determines connectivity by physical location.
Single-pair 10BASE-T1S Ethernet is the ideal choice for a zonal E/E architecture. Its credentials provide connectivity for simple sensors and actuators using a single PHY per node - everything from seat comfort controls to an ADAS blind-spot detection mirror. 10BASE-T1S offers faster communication speeds than CAN-FD, for example, and the opportunity to remove legacy network protocols such as CAN and FlexRay.10BASE-T1S enables an 'all Ethernet' IVN approach to support point-to-point and multi-drop topologies and a cost-optimized bill-of-materials. One major benefit to this approach is that it negates the need for gateways to translate CAN or CAN-FD to Ethernet, saving weight and reducing material costs. Also, it benefits from using the same software stack as 100/1000Base-T1 Ethernet. Apart from a different PHY, no additional gateways between the two are required. With the advent of implementing a software-defined vehicle (SDV) architecture, another advantage of Ethernet throughout is that it removes the need for firmware-over-the-air updates to navigate through gateways, streamlining and simplifying software management.
10BASE-T1S Physical layer collision avoidance
With the multi-drop topology employed by 10BASE-T1S, all nodes are connected using the same single twisted pair.
Instead, 10BASE-T1S uses a physical layer collision avoidance (PLCA) technique, assigning each node a unique identification number unrelated to the MAC address. Node #0 is defined as the PLCA coordinator. It is the only node that knows the total number of connected nodes. Collisions are avoided by node #0 allocating transmission slots to itself and all other connected nodes on a round-robin basis. The number of nodes will determine the maximum PLCA cycle time and the maximum communications latency. As a consequence, the length of each PLCA cycle is very elastic.
A PLCA cycle commences with node #0 issuing a 20-bit beacon, causing all nodes to reset their "transmit opportunity timers" as defined in the IEEE 802.3cg specification. Nodes can only transmit when the transmit opportunity matches its node ID. In each cycle, nodes take turns sending a "commit" signal if they have data to transmit. A node remains silent if it does not have data to send. A commit signal from a node allows it to transmit Ethernet frames onto the bus, during which period all other nodes wait.
The above figure illustrates two examples of PLCA cycles. The top diagram highlights a fast cycle where no nodes have any pending data to send, staying silent during their opportunity. The lower diagram is a scenario where every node has pending data, with each node placing a commit signal on the bus and its data packaged into Ethernet frames.
Implementing 10BASE-T1S
Examples of 10BASE-T1S MAC and PHY transceivers suitable for industrial applications are the onsemi NCN26010 and onsemi NCN26000 (to be released in Q2 2024) devices. The NCN26010 integrates a media access controller (MAC) and a PLCA reconciliation sublayer. Capable of half-duplex 10 Mb/s multi-drop operation of up to a minimum of 25 m over a shielded or unshielded single-twisted pair (STP/UTP), the NCN26010 connects to a host microcontroller via a slave serial peripheral interface (SPI). It requires a single +3.3 VDC supply to operate and is available in a 32-pin QFN, 4 mm x 4 mm package.
The onsemi NCN26000 PHY provides similar functionality to the NCN26010 but has a media-independent interface (MII) in place of the MAC and SPI connections. The NCN26000 can use its MII interface to communicate with a companion microcontroller equipped with a MAC.
Both devices have enhanced noise immunity (ENI) characteristics that extend the PHY's noise immunity specifications above 500mVpp compared to other devices available on the market. ENI increases the maximum distance and number of nodes the transceivers can support in electrically noisy environments. For example, lab results show with ENI enabled, both devices can support successful communication across a 25-meter (SPE) cable distance with 40 nodes, or 16 nodes across a 50-meter cable (SPE) distance.
The NCN26010 and NCN26000 suit various industrial applications, including process automation, sensor, and actuator nodes, and building automation use cases.
For automotive applications, the onsemi NCV7311 and NVN7410 will launch in 2024.
10BASE-T1S for industrial and zonal automotive architectures - a winning combination
Ethernet has become the backbone of our connected world. First ratified in 1980, Ethernet continues to show no signs of stopping and has found widespread adoption outside of traditional computing applications. The rise of the industrial internet of things and the dawn of the software-defined vehicle with its zonal E/E architecture present new opportunities for wired network infrastructure. The availability of IEEE 802.3cg single-pair Ethernet topologies enables an "all Ethernet" capability from edge to cloud.
10BASE-T1S runs 2x to 3x fast than a controller area network (CAN) topology and eliminates the need for CAN to Ethernet protocol translation gateways, saving cost and weight. Compared to CAN’s 2 Mbps and CAN-FD’s 5 Mbps, 10BASE-T1S delivers 10 Mbps.
Thank you to Bob Card, marketing manager at onsemi, for input and images provided for this article.



