Give me a pair of wires, I’ll give you Ethernet in cars

A very astute gentleman said to me a few years ago that he’d seen a lot of networking technology come and go – Token Ring, FDDI, Fibre Channel, InfiniBand – but the only one that held up over time was Ethernet.

His premise was simple. The ubiquity of Ethernet ports and the TCP/IP protocol forces the standard to evolve. If it isn’t fast enough today, a future revision gets faster. If connectors are too expensive and bulky, a future revision comes up with a smaller one. As it turns out, that evolution plus the interoperability factor has displaced many other network technologies into specific niches, like high performance computing or storage, where costs per port aren’t the main criteria.

In a car, the spin on costs per port is a bit different. Wires once carried only power, connecting batteries to starters and alternators and headlights and taillights and (I can see paradise by the) dashboard lights. With power from front to rear bumper, engineers started adding more convenience items, like radios, power windows, and power seats.

But then, computational elements started to enter the car. Anti-lock braking, fuel injection, traction control, cruise control, air bags, climate control, and other subsystems required data to be passed between control units. CAN bus was invented, to standardize the interface and reduce the number of wires between systems, which in turn got rid of some weight and made installing aftermarket devices easier.

That was all good until infotainment and telematics systems showed up, with greater data throughput requirements than CAN could possibly handle. Today, we see a commercial being run by Cadillac, suggesting a car should work more like a tablet.

The answer is yes, of course, but it goes deeper than the user interface. The first attempt at a multimedia bus for cars was FlexRay, but it didn’t last long. Why? The IEEE and the OPEN Alliance SIG have created a new standard for a one pair, unshielded Ethernet interface specifically targeting cars. My wise mentor is right, again.

This suddenly makes Ethernet a winner for cars in three dimensions: bandwidth, cabling cost, and software. That last one is important. The “cloud” is TCP/IP down to the car, and there are huge gains from adding smart devices with only a change in the media access controller (MAC), the physical interface (PHY), and the connector. Other initiatives like the AVnu Alliance for middleware and GENIVI Alliance for infotainment applications are even better than expected with software running over an Ethernet interface.

Cadence is first out with merchant MAC design IP for Automotive Ethernet, also announcing verification IP. Broadcom, Freescale, NXP, and Renesas are the notable SoC and MCU names likely to have chips out first since they are OPEN promoters.

Will Ethernet take over in cars completely? Not right away. As with most technology, improvements find their way into luxury vehicles with high end systems first. CAN will still be around in cars for the foreseeable future, but for the infotainment and telematics systems with cloud connections, Automotive Ethernet is likely to be adopted rapidly.