Entering the Ethernet Era: The Difference Between CAN and EthernetMay 11, 2021
CTO & Co-Founder
Cars are quickly transforming into computers-on-wheels designed to optimize the driver’s experience. Advanced software capabilities require a high-performance hardware foundation, and the current CAN bus networking backbone is stretched to its limit.
The biggest problem with CAN is the limited bandwidth it provides. At best, it can provide 5 Mbps payload throughput (i.e., goodput) with the newer CAN-FD, which is pretty meager by today’s standards. However, because CAN is reliable and cheap, automakers are not keen on changing it and they have been adding more and more CAN buses to provide additional bandwidth, but this solution is far from ideal.
Cars today require stronger underlying tech capabilities to support new functionality, and Ethernet, proven in avionics and industrial applications, is perfectly suited to the task with high line rates, higher payloads and higher resulting goodput.
Let’s take a look at the difference between CAN and automotive Ethernet and see what the future holds for the automotive industry.
CAN is at the Heart of the Vehicle…
The CAN revolution in automotive technology took place in 1991 when CAN buses were first deployed. The CAN bus quickly became the norm, acting as the heart of the car’s networking system. It was for good reason that CAN’s popularity took off, as some of its key technical features made it the perfect match for cars’ needs as it:
- Replaces multiple discrete I/O lines with one 2-wire message bus
- Runs in realtime
- Has a physical layer and protocol resulting in high tolerance for noise
- Supports native multicast and broadcast
- Offers built-in frame priorities
- Supports non-destructive collision resolution
- Has 100% distributed in operation
- Supports long single buses
- Is low-cost to build
Despite all of the upside to the CAN bus, it does have its limitations, especially as demands on it grow. The number of devices needed to connect to a single CAN bus is constantly increasing and with its low data-transmission rate of 5 Mb/s (CAN-FD) or 1 Mb/s (CAN), each new device decreases performance significantly.
As a first step towards addressing the new demands for software-defined vehicles, car manufacturers have simply added more and more CAN buses to the cars. While this can work by adding significant amounts of wiring and complex gateways, it isn’t a scalable nor cost effective approach, and it leaves connected cars more vulnerable to threats and cybersecurity attacks.
While newer versions of CAN that resolve some of the challenges are in development, many automakers find themselves in the midst of the automotive Ethernet vs. CAN debate.
...But Ethernet Provides a Stronger Backbone
When comparing CAN vs. automotive Ethernet, automakers will find that Ethernet offers a strong solution to the problems raised by the traditional CAN bus architecture. Cars need a high-performance backbone that can support the increased communication from all the new software that allows for the driver-centric experience. Trends are showing carmakers moving away from the more traditional complex wiring to a Zonal Architecture that relies on Ethernet to run smoothly and securely.
The key differences between CAN and automotive Ethernet and the advantages of Ethernet include:
- Supports significantly higher throughput rates (up to 10Gbps and working on more), allowing to aggregate multiple CAN buses into a single Ethernet link. This results in smaller wiring harnesses than CAN. This means lower installation and maintenance costs.
- Supports quality of service and time-sensitive networking (TSN) for allowing realtime communication to be multiplexed with lower priority data.
- Advanced security features and transport layer services
- Plug and play capabilities mean components can be connected and disconnected as needed with automatic detection and configuration
There is also a downside to Ethernet which includes a more costly controller and physical-layer interface, complicated electromagnetic compatibility issues, as well as overhead for allowing for realtime communication (e.g. TSN). But it doesn’t seem that the bad outweighs the good in the question of automotive Ethernet vs. CAN.
Overall, the flexibility inherent in Ethernet is what gives this type of architecture the edge. Especially as more and more lines of code are added to car models, the choice will sway even more to Ethernet as that system will better support the multitude of infotainment, safety, and autonomous driving features along with over-the-air updates.
So Who Wins? Ethernet or CAN?
While all indications point to Ethernet taking over with predictions of the Ethernet market growing from $1.6 billion in 2019 to $4.4 billion in 2024, we haven’t seen the last of CAN buses either. In fact, in all likelihood, we will first see a combination of CAN and Ethernet complementing each other in the same vehicles.
As automakers move to Zonal E/E Architecture, there will still be instances of newer ECUs that need to (and are able to) operate alongside legacy ECUs that are still perfectly good and do not necessarily need to be replaced. In such cases, it may be more cost effective to continue to use CAN for the legacy parts but add Ethernet for the newer components. One attempt to replace CAN altogether comes in the form of 10Base-T1S Ethernet, which reduces Ethernet’s costs significantly.
Nonetheless, additional technical issues (EMI/EMC, costs, complexity, wire length limitations, etc.) remain to be solved prior to Ethernet becoming a mainstream and widely accepted replacement for CAN. That being said, the change is coming, the need for high-bandwidth communication and low latency is just too much for any CAN bus to handle.