In-Car Ethernet: Untangling the Wads of Wires to Enable Ever More Data

If you've ever had a chance to look at the tangled mess that is an automotive wiring harness, you know those things can be pretty huge; hundreds of wires all bundled together by zip-ties that you'll hopefully never have to snip lest you nick the sensitive shielding within.

Those bundles take up precious space within the chassis of cars, adding unwanted weight and creating many more potential failure points. But, thanks to some new, high-bandwidth cabling developed with enterprise-standard networking protocols, that wad of wires can be replaced by a single, slender thread offering more bandwidth than all the rest of them combined.

That's part of the promise of in-car Ethernet, but the benefits go much deeper than that.

See All 8 Photos8

Ethernet and the Software-Defined Vehicle

The era of the software-defined vehicle (SDV) is upon us. With more code running more parts of more cars, it's increasingly possible to radically change the way they act and even drive with code pulled down from the cloud.

But, more software running in the car means more data, too. Lots and lots of data. All those high-definition sensors that help modern cars see the world around them generate massive amounts of information every second. All that data must be shuttled through the car in endless streams that are rapidly growing into rivers.

If that weren't enough, you have modern infotainment systems cranking 64 channels' worth of Dolby Atmos audio or streaming high-definition video to multiple rear-seat infotainment displays.

In other words, there's a lot going on under the skin of your next car, and all of it needs ever more bandwidth. Ethernet is the solution.

See All 8 Photos8

What Is Ethernet?

If you're familiar with the term Ethernet, you probably know it as the kind of cable (most commonly known as a Cat 6 Ethernet cable) used to connect your internet provider's modem to your wireless router. Or, if you're a serious gamer, it's the cable you use to connect your PC or console directly to the network to get the lowest latency.

That's a pretty common thing, but it's not exactly what we're talking about with in-car Ethernet. Formally, Ethernet isn't actually a single type of cable at all; it's a networking standard that defines a way for multiple devices to interact securely and reliably.

(By the way, Ethernet is capitalized because it refers to a technology trademarked by Xerox in the 1970s. Xerox released the trademark so that the technology could ultimately become a cross-industry standard, but it's still a proper name.)

The thing you typically call "an Ethernet cable" at home internally contains pairs of smaller cables, twisted together, which is why they're often called "twisted-pair." At each end, they typically have an RJ45 connector, which looks like an oversized phone jack. The formal name for this kind of cable is 100BASE-T or 1000BASE-T, referring to its speed of either 100 megabits or 1,000 megabits (a gigabit) of data per second. If you're on gigabit Ethernet, that should provide enough bandwidth to transfer a 4K movie in about four minutes.

When it comes to modern in-car data, even that is far too slow. Greater speed requires a different sort of cable altogether. "We are talking a technology that's called T1," Amir Bar-Niv said. He's vice-president of marketing at Marvell Semiconductor, a California-based major international provider of semiconductors and networking equipment. "Everything we are doing, 100 [megabit], gigabit, multi-gig, and 25 gig, all were designed for T1."

When Bar-Niv says "25 gig" he means 25 gigabits, enough speed to transfer that same 4K movie in eight seconds, with plenty of bandwidth to spare for all the making-of featurettes you could ever want.

See All 8 Photos8

The type of cable here is optimized not only for speed, but also for automotive use. It's thinner and lighter than the cables you have jumbled up behind your desk and less expensive, too—a pair of wires wound into a single cable, or even a single fiber-optic thread, offering all the bandwidth needed for next-generation SDV applications, using a newer type of standard connector called T1

It wasn't always that way. In fact, in-car Ethernet goes back 15 years. "It was first introduced by BMW on the F01 7 Series model in 2008," Ian Riches said. He's the VP of  automotive practice at TechInsights, an international firm with a focus on semiconductor analysis. "It was initially used for the diagnostics port, and also for a 20-Mbps front-seat/rear-seat link on the infotainment system." That first application used 100BASE-T cabling, like you'd use at home.

Those evolving cable standards are a key part of the system, of course, but to really make a network you need a place to plug them. On one end, you have the sensors and systems that generate the data and do the work. On the other end are the car's various processors that handle the data. Keeping everything moving in the right direction requires a series of network interchanges, devices called switches and routers.

Optimally locating those switches within the car minimizes the amount of cabling required while still enabling all the devices in the car to speak to each other. That configuration, often referred to as the network architecture or topology, is where things start to get interesting. (Honest.)

Exiting the Zone

Traditionally, as cars gained features and options through their lifecycle, the bundle of wires collectively referred to as the wiring harness just grew and grew, sprouting more strands with every new feature.

This, Bar-Niv said, is a legacy of the old way of connecting things, what he calls a domain architecture.

Think of each system of the car, like the lighting or infotainment system, as a separate domain. Each domain is controlled directly to one or more of possibly dozens of separate microprocessing units, all scattered throughout the car.

If a car picked up a few new features in a midcycle refresh, like maybe heated washer nozzles or a blind-spot monitoring system, yet more wires would be added into the bundle to control that feature.

"In a car where there is no sharing between domains almost at all, every domain needs to have all its cables running around the car," Bar-Niv said. "And the result is, you know, tons of cable, high cost, highways of cables."

With Ethernet, he said, the car transitions into a collection of interconnected zones, each zone containing sensors or devices, which he calls agents: "Each zone now needs to have its own Ethernet switch that connects to all the agents of this domain at that zone, and then aggregates all this traffic over a single, thin Ethernet cable that is really building the backbone of the car."

In this way, multiple onboard systems all communicate over the same network. By sharing, a car's design can be radically simplified. Rather than dozens of bespoke little chips scattered all around, fewer, more powerful central processors handle all the digital heavy lifting for the entire car.

Beyond simplification, this opens the door to the evolution of a car's behavior that is the hallmark of SDVs. Bar-Niv cites car cameras as an example: "Cameras today are using point-to-point connectivity," he said, transmitting data directly from the imaging sensor to the system where it's used.

Because of that, a rear-view camera can only ever be used as a rear-view camera, because no other systems in the car can access it.

See All 8 Photos8

"So you cannot, for example, share cameras, between infotainment and ADAS," Bar-Niv said.

But, once all these cameras are instead communicating over Ethernet, all on the same network, software can be written to use those cameras in interesting new ways, like turning a rear-view camera into a dash cam that streams footage over the internet. This feature could then be downloaded via an over-the-air (OTA) update.

It's this flexibility that is core to a SDV, and according to Bar-Niv, a recent survey among automotive manufacturers showed that 80 percent of all new cars are expected to shift to this architecture by 2026, becoming a $5.6 billion industry alone.

In other words: This isn't just a trend on high-end luxury cars.

And, as that trend continues, the need for more bandwidth will likewise increase. "100 megabits was good enough for domain infrastructure. Now, nothing below 2.5 gig is acceptable," Bar-Niv said.

Automotive Grade Reliability

Much of the core technologies here are based on common enterprise networking hardware—the same stuff your bank or business uses. Bringing it into an automotive application, however, requires even greater reliability. If a piece of network hardware fails at a bank, that bank might lose money until it's fixed. That's not good, but when a car's active safety systems run through a network, any failure there could potentially result in loss of life in a worst-case scenario.

A key part of preventing that kind of failure is improving the ultimate quality of the components that are installed in cars. "Reliability we measure in DPPM," Bar-Niv said, which stands for defective parts per million. For a consumer network device, Bar-Niv said the DPPM is "in the range of a few hundred to 1,000."

When it comes to automotive, the target is fewer than one DPPM, just one defective part in a million produced.

But even though those parts are manufactured to a higher standard, failures can still happen. That's where redundancy comes in, like network switches that use what's called a dual-core lockstep processor. Here, two processors handle every calculation. The output from each is compared and, if there's a discrepancy, an algorithm can ensure the overall system continues to run, relying on the output of the functional processor.

See All 8 Photos8

Improving Your Car's Security

If having every component in your car, from high-beams to heads-up display, all communicating on the same network sounds like a bit of a security risk, you're not wrong. Previously, if a hacker accessed a given system on a car, one of the domains we talked about above, it was very unlikely that hacker could then take over other aspects of the car's system. Thanks to those inefficient amounts of wiring, every system domain effectively ran on separate, disconnected networks.

In a car optimized to use in-car Ethernet, with everything merged onto a single network, the potential for system-wide access through a single compromised component is greater. Thankfully, the protection is greater, too.

"What we need to prevent is attacks like what's called man in the middle," Bar-Niv said.

This refers to a sort of attack where the hacker effectively pretends to be an authorized system within the car, interfacing with the other systems and using them to gain access to the data within.

One way to lock out an attack like that is by using a hardware security module, or HSM. These generate digital keys within the car, encrypting and digitally signing all data passed around within. This kind of tech is usually only used for high-end enterprise digital networks, its presence in cars ensuring a similar degree of security on the go.

See All 8 Photos8

Cost and Benefits

Between the fancy cabling, the switches, and the HSMs, not to mention all the software to run it all, it might sound like an expensive proposition to run a car on Ethernet instead of the former, more basic but redundant domain model.

"Yes and no," TechInsight's Riches said. "Yes, Ethernet is more costly than slower, existing CAN links in the vehicle. However, it is a fundamental building block of the new vehicle architectures that are needed to bring about functions-on-demand, as well as the high-speed communications required for automated driving."

In other words, there's no way around it. Riches says Ethernet is the "only game in town" if you want to do things requiring lots of data. But, compared to the cost of those other sensors, like the 3D laser scanners installed on the new Lotus Eletre, the cost of networking hardware to connect them will be insignificant.

In fact, in some cases switching from a domain to a zonal architecture, dropping all the redundant components along the way, could actually save money. "There is also the long-term view that, as compute performance is centralized, there are vehicle-level savings that could be made in deleting many of today's existing discrete ECUs," Riches said.

According to Bosch, Tier 1 supplier for many OEMs, switching to this kind of architecture could result in a 10 percent reduction in cost for those onboard networks. Those future cars would also require 20 percent fewer onboard components, according to Bosch, things like microprocessors—the continued shortage of which still plagues the industry.

So, a shift over to in-car Ethernet might mean cars that are not only smarter, cheaper, and lighter, but a little less hard to find, too, in the future.

See All 8 Photos8