However, Ian Riches, director of global automotive practice at Strategy Analysis, told EE Times, "Our current estimates are that in 2020 around 120 million Ethernet nodes will be fitted to light vehicles produced that year."
When asked how many Ethernet nodes are expected in a high-end car in 2020, Riches said, "I would be surprised if it hit 100 Ethernet nodes."
He counted up to five domain controllers on the backbone (powertrain, safety, chassis, body, infotainment), 10 to 20 infotainment nodes, 10 to 15 nodes for advanced safety, 1 node for diagnostics, and 10 to 20 additional nodes for other high-end features.
"That makes around 35 to 60 nodes by my reckoning," Riches noted.
In-Car Networking Scenario
Ethernet/IP will coexist with low-bandwidth standards like CAN.
(Source: Frost & Sullivan)
The big gap in forecasts could be partly explained by differences in expectations of how fast the ECU in the vehicle goes Ethernet.
Frost & Sullivan's Chandrasekar believes in the idea of a backbone network supporting other networks. But he made it very clear that "the backbone would be Ethernet-driven. The CAN, MOST, LIN and others will continue to exist on a small scale basis, but Ethernet will drive the majority of the work."
Strategy Analysis's Riches doesn't necessarily agree with that premise.
In Riches's opinion, "Not every function in the vehicle requires the 100Mbps-plus" offered by automotive Ethernet. A seat control motor, for example, won't gain extra functionality from being on an Ethernet network rather than the low-speed CAN or local interconnect network (LIN) that runs it today, explained Riches. "Unless Ethernet is cheaper than LIN, then LIN at the very least will remain for many low-end functions." He added that part of the backbone/domain controller architecture trend leans toward reducing the number of ECUs in the vehicle -- which means fewer Ethernet nodes than in Frost & Sullivan's forecast.
Luca De Ambroggi, senior analyst, IHS's Automotive Component and Device Electronics and Media, is more cautious about the rate of Ethernet adoption. He thinks Ethernet is still looking for a place to break through. "I believe this breakthrough point is infotainment, followed by basic ADAS application like Camera based functions," he said.
Obviously, for automotive Ethernet to succeed, "The target of Ethernet must be the entire auto wiring system from Infotainment down to the most safety and security critical segments like body." Nonetheless, De Ambroggi believes "Ethernet needs a small step approach to be mature and bullet-proof for automotive."
As Broadcom likes to point out, there is a definite upside in using Ethernet in a car -- in terms of savings in cost and weight.
Single-pair Automotive Ethernet, which uses un-shielded twisted pair (UTP) cable to deliver data at a rate of 100Mbps, along with smaller and more compact connectors “can reduce connectivity cost up to 80 percent and cabling weight up to 30 percent,” according to Broadcom.
While acknowledging potential savings, Strategy Analysis's Riches pointed out, "There are also significant risks when adopting a new technology."
For one, there is always a cost-to-change. At this stage, the additional costs are arguably easier to quantify than any potential savings, he said. While an Ethernet backbone would be an enabler for greater functionality in the vehicle, he also noted, "Whether that functionality will be required (or affordable) on a mass-market 2020 vehicle is far from certain."
In summary, Riches noted that there's a theoretical tipping point where the number of features and bandwidth required will be cheaper to "start again" with an Ethernet backbone-type architecture. "But I don't see us as being there before 2020 at the earliest for volume car makers,” he said.
Remaining technical issues
Broadcom insists that its BroadR-Reach Ethernet technology has already met rigorous automotive industry standards. Its Phy layer was specifically designed to meet the automotive industry's EMI and EMC requirements, noted Abaye.
Also, Automotive Ethernet uses a signaling scheme with higher spectral efficiency compared to the signaling scheme used in 100BASE-TX, according to Broadcom. This limits the signal bandwidth of Automotive Ethernet to 33.3MHz, which is about half the bandwidth of 100BASE-TX. As a result, "a lower signal bandwidth improves return loss, reduces crosstalk, and ensures that Automotive Ethernet passes the stringent automotive electromagnetic emission requirement," the company said.
Strategy Analysis's Riches, however, believes that some technical questions still remain. "I still hear concerns over EMC performance for Ethernet over UTP." Riches, however, quickly added that "how much of this is real as opposed to just fear-of-the-unknown is hard for me to judge."
Other criticism questions whether the current 100 Mbps of BroadR-Reach might be too slow, Riches added. "It essentially means that camera images will need to be compressed -- which some see as a big problem for safety systems."
There are players that seem to be waiting for Gbps in automotive, said Riches. "Moves are underway at IEEE level to standardize this -- but these things take time."