Autonomous Vehicle Networks: Is Ethernet Enough?

Article By : Junko Yoshida, EE Times

PARIS — A cross-industry group promoting a standard called HDBaseT announced last week that Denso, a big tier one in Japan, has joined the group as a “contributor member.”

HDBaseT, originally developed as a consumer electronics and commercial connectivity standard for transmission of uncompressed HD video, audio, power, home networking, Ethernet, USB, and some control signals over a CAT 5 cable, now has an automotive variant called HDBaseT Automotive.

The new automotive connectivity standard, touted as a high-throughput pipe at zero latency for ADAS, will enable the transmission of “up to 6 gigabit per second tunneling of video & data, with native networking capabilities over 150m/50ft of a single unshielded twisted pair cable,” according to the group.

Because HDBaseT Automotive is still in its early days, the recruitment of Denso is significant. Daimler is the only car OEM to publicly embrace the technology for in-vehicle connectivity for advanced infotainment, ADAS, and telematics systems. Valens Semiconductor, an Israeli fabless chip company, is thus far the only chip vendor committed to offering HDBasT Automotive compliant silicon.

Proponents of HDBaseT believe that the new automotive variant is a missing link in the discussion of future connected and autonomous vehicles. Given the vastly expanded volume of data passing through vehicles, HDBaseT seems to make sense.

But here’s the thing. Are we now saying that the Automotive Ethernet falls short of handling multi-gigabit bandwidth at minimum latency?

In-vehicle connectivity fatigue
When it comes to internal communications network technologies, carmakers today have plenty of choices. For example, Ethernet is already being considered as a replacement for legacy bus protocols such as MOST, a high-speed multimedia interface, and FlexRay, a general purpose high-speed protocol with safety-critical features.

Bluetooth Low Energy and High Speed 802.11ac Wi-Fi are popping up inside vehicles to connect smartphones and wearables. Wired broadband MHL is being used for rear-seat infotainment.

Meanwhile, legacy buses such as CAN and LIN continue to exist as vehicle control network protocols.

Connectivity technologies originally developed for non-automotive applications are also creeping into the mix. They include Low-Voltage Differential Signaling (LVDS) which operates at low power and can run at very high speeds using inexpensive twisted-pair copper cables, and Serializer/Deserializer (SerDes), a pair of functional blocks commonly used in high speed.

Given this plethora of options already available, how eagerly might automakers warm up to HDBaseT Automotive? Ian Riches, director for the Automotive Electronics service at Strategy Analytics, is already sensing the automotive industry’s “nervousness and reluctance” as carmakers see HDBaseT Automotive as “yet another standard.” Riches said, “Many OEMs are just beginning to get their heads around Ethernet…”

However, there are several unique and redeeming factors about HDBaseT.

Danny Kim, partner and director of advisory services at VSI Labs, pointed out that HDBaseT allows for both point-to-point and daisy-chain connectivity. This daisy chain connection, he said, “enables much less total connections (cost, better user experience).” Further, “It’s bi-directional, so once there’s a connection failure or cable failure, all the devices connected to the cable can still operate/stay connected (redundancy).”

Granted, a host of sensors in autonomous vehicles would generate a lot of data. AVs will demand a fat pipe. But how much fatter compared to that of Automotive Ethernet, and why would the new connectivity standard need to cover so long a distance — as much as 50 feet?   

Micha Risling, senior vice president and head of the automotive business unit at Valens Semiconductor, told us, “It comes down to three things.”

First, “as more and more devices are introduced (e.g., rear and front cameras, mirror replacement cameras, displays including those for passenger, driver, rear seat, sensors such as radars and lidars, smart antennas), the amount of data is increased exponentially,” Risling said. “But more important, this data needs to be transmitted and processed in a short amount of time,” stressing that “any delay can be life-threatening for ADAS data.”

Second, he noted, “Video performance is a major issue in autonomous cars, both in terms of resolution (quality) and latency, as any problems in video delivery can also have fatal consequences. Higher video resolution demands more bandwidth, with near-zero latency in the case of cars, which demands fat pipes.”

Third, he added, “Although it seems that 15m/50ft is a long distance in cars, it really isn’t when you take into consideration that the cables are not going straight from point A to B, but rather, under the seats, over the doors, around the wheels, not to mention the engine bay and other elements in the car.”

HDBaseT vs. BroadR-Reach
So, is HDBaseT meant to leapfrog BroadR-Reach? Or, more specifically, which in-vehicle cables will HDbaseT replace?

HDBaseT is focused on multigigabit transmissions of different types of data. BroadR-Reach is based on Automotive Ethernet, at 100Mbps symmetrical speed. Risling said, “HDBaseT can now deliver 2 gigabit per second bandwidth over unshielded twisted pair — a cheaper, more flexible and easier to install cable (with future versions enabling 4/8/12/16Gbps).”

He added, “Another major advantage of HDBaseT over Automotive Ethernet is that the technology is both symmetric and asymmetric, which can be optimized for camera/sensor links (a symmetric link is not optimized for video delivery, which usually needs higher speed one way only).” In other words, “HDBaseT will be integrated wherever higher speeds and convergence are needed.”

He explained that some basic controls, that today use CAN or MOST technologies, will not necessarily be replaced, as they do not need higher speed connectivity, but they can all be tunneled together through an HDBaseT link toward the ECUs (shortening and reducing the number of cables). “As we move towards more connectivity devices — displays, modems, cameras, and sensors — we will need a technology like HDBaseT to enable the connectivity among the devices, between the devices and the high-performance computers in the car, and among the computing elements themselves.”

Where in a vehicle will HDBaseT go first?
While declining to name its automotive customers, Risling told us that Valens is already working with OEMs and tier ones in specific use cases for HDBaseT Automotive.

Where in a vehicle will HDBaseT go first?

“It will most likely be adopted first for infotainment use cases, such as USB connectivity,” he said. There’s also “door connectivity,” he added. “HDBaseT can reduce the 20+ cables going from the door hinge for mirrors, sensors and audio, to just one cable.”

Another use is smart antennas. Risling said, “As HDBaseT has the ability to extend PCIe connection for up to 15m/50ft, enabling split telematic control units (TCUs). In this case, the modem units can stay on the roof, while the gateway can be placed inside of the car, closest to the infotainment and ADAS ECU, without having to handle the high temperatures and tight space on a car’s roof.  This reduces the number of components, cables and connectors necessary.”

Arguing for HDBaseT, VSI Labs’ Kim summarized: There is “no need for close grounding; high-speed, future-proofed PHY; interoperability between chipset generations; faster everything (Internet, car maintenance updates, USB, smartphone connectivity), better ADAS (centralized video analytics with higher accuracy); less cables, lighter cables; lower fuel consumption.” Further, “HDBaseT network enables Ethernet switching thanks to its native Ethernet support,” in addition to “high-speed firmware update (via USB or 1GbE).”

What about cost?
Valens insists that HDBaseT brings a slew of cost advantages. Strategy Analytics’ Riches isn’t so sure.

Valens’ Risling pointed out, “Ability to transmit multigigabit data over unshielded twisted pair cables, a cheaper alternative to shielded cables and coax.” He stressed, “Alternative solutions — such as Automotive Ethernet and Serdes — are not able to transmit high bandwidth over unshielded cables.”

Further, “With daisy-chaining connectivity, HDBaseT can reduce the number and length of cables in the car, leading to lower weight and therefore better fuel efficiency,” said Risling. “Daisy-chain connectivity also simplifies connectivity architectures, reducing the number of chipsets necessary in a system.”

Finally, he pointed out, “HDBaseT can tunnel different native interfaces over the same cable and connectors — audio/video, Ethernet, controls, data (USB and PCIe, for example) and power, eliminating the need for different cables/connectors for each interface.” He added, “Again, this reduces the number of cables and other electronics and corresponding weight, simplifying the overall cable harness in the car. Alternative technologies usually require additional chipsets to tunnel different interfaces.”

Riches, on the other hand, believes price will be a hurdle for HDBaseT. He said, “Total system implementation cost is hugely important, but so is silicon cost. Saying, “the silicon is more expensive, but you’ll save in the long run…” can be a hard sell.”

VSI Labs’ Kim agreed. “You need a HDBaseT switch which could be costly and in each node of devices (sensor, display, etc) would need HDBaseT compatible chip components (e.g., PHY in case of Ethernet), resulting in added cost for the vehicle manufacturer,” he noted.

Indeed, asked about price, Valens Risling steered clear. “It is very difficult to come up with a specific number regarding cost, as each case is different from the next, and rely on a variety of factors, such as interfaces and bandwidth supported, volume etc.”

Challenges: Uncertainty in the market
Beyond the cost and the auto industry’s weariness with standards, there are a few other problems.

Strategy Analytics’ Riches cited the issue of “maturity and longevity.” He observed that HDBaseT is a relatively new approach initially centered on the consumer industry. “Consumer standards can come and go, but automotive needs something that will last.  At one stage IEEE 1394 Firewire was touted for automotive usage, when it was increasingly being used in consumer products. I’m guessing the auto industry is glad it didn’t go down that route now.”

It’s also tough to “make the case for high-bandwidth links and networks now,” said Riches. “Future use cases based on highly autonomous driving are easy to see — but will not emerge in high volume in the next five years. Many carmakers are still effectively designing around the need for high-speed links and thus putting in significant processing at each sensor to process the data.”

For HDBaseT to gain momentum, it needs multiple silicon vendors. Riches said, “Car makers do not like single-source.”

Beyond Valens, though, Riches said, “I have not seen any confirmation of automotive HDBaseT chips from others.” However, he added that NXP, the largest automotive semiconductor vendor and a leading player in automotive networking, is a contributor member of the HDBaseT Alliance.  STMicroelectronics is also an “adopter member.”

Valens also sees challenges.

Risling, for example, cited “compartmentalized organizations” as a big hurdle when Valens goes out and tries to pitch HDBaseT. “OEMs have separate departments for each area — ADAS, infotainment, wire harness, etc. HDBaseT is a technology that is optimized not only to each of these connectivity needs, but also can optimize the integration of each of these resources in the car, leading to shared ECUs and storage, for example.” He said, “Being able to convey this message can be difficult when each department is detached from others.”

And then there’s “uncertainty in the market,” said Risling. “Although the connected car is the ultimate buzz in the market, there is still great uncertainty as to what and when things will happen,” he said. “OEMs and tier ones are still defining the most important needs, and we are working closely with them to be able to deliver what is needed when it is expected.”

Valens has become increasingly aware of the significance of “partnerships and alliances.” Risling explained, “As we further penetrate the market, the value of partnerships — with tier ones and tier twos, different standardization alliances, other chipset manufacturers — becomes increasingly clear.”

He said, “To succeed in this sector, it is important to identify the companies and organizations that can make a positive impact in our efforts, and work with them to achieve our goals.”

Risling said Valens, for example, is now working with the MIPI Alliance’s Automotive Work Group.

— Junko Yoshida, Chief International Correspondent, EE Times