Unprecedented in-car data generation straining current interfaces to their limits
PARIS — Not long ago, the mobile industry was the master who led chip vendors around on a leash, commanding them to sit, heel, and roll over. Lately, the leash has been passed on to the automotive industry.
Level 4 and Level 5 autonomous vehicle designers are asking for in-vehicle connectivity at multi-gigabit speed, to which a host of industry alliances and standards bodies — ranging from Networking for Autonomous Vehicles (NAV) Alliance to MIPI and HDBaseT group — are all saying that they are ready to roll over.
Fully automated vehicles are, of course, not here yet. But their heavy reliance on high-resolution sensors including vision, lidars, and radars is destined to vastly expand the volume of data that must pass through vehicles. This prospect compels every industry group that has ever developed high-speed connectivity technologies for data centers, enterprise infrastructure, or professional studios to see golden opportunity in the automotive market.
Last week, MIPI Alliance, a group who started by developing interface specifications for the mobile industry, revealed preparations to address high-speed data interface specs in the automotive industry. It has already initiated development of an MIPI A-PHY specification to meet 12 to 24 gigabits per second (Gbps) of a physical layer up to 15 meters. MIPI A-PHY v1.0 is expected to be available to developers in late 2019.
Just last month, HDBaseT group announced that Denso joined as a “contributor member.”
HDBaseT — originally 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.
In a recent interview with EE Times, Micha Risling, senior vice president and head of the automotive unit at Valens Semiconductor, said, “HDBaseT can now deliver 2-Gbps bandwidth over unshielded twisted pair — a cheaper, more flexible, and easier to install cable” — while it readies for “future versions enabling 4/8/12/16 Gbps.”
Also, there is NAV, an industry alliance consisting of chip vendors Aquantia, Nvidia and Bosch, Continental, and Volkswagen. The group’s big advantage is said to be that it’s moving in lockstep with IEEE Ethernet group to develop IEEE Multi-Gig Automotive Ethernet PHY. The force behind this movement is Aquantia, a chip vendor that set itself apart from others by pioneering high-speed, multi-gigabit Ethernet technologies now used by data centers (10/25/100 GbE) and enterprise infrastructure (2.5/5/10 GbE).
In an interview with EE Times last week, Amir Bar-Niv, vice president of marketing for automotive products at Aquantia, acknowledged, “There had been a vacuum” in the automotive industry — where no multi-gig automotive Ethernet existed. That, in turn, created an opening for other players to come in and pitch proprietary solutions, he explained. But now that the IEEE Multi-Gig Automotive Ethernet PHY standard is well on its way, “I no longer see a vacuum,” he said.
In Aquantia’s mind, the adoption of the Multi-Gig Automotive Ethernet by OEMs and Tier Ones is almost a forgone conclusion.
After all, with the latest addition of Automotive Multi-Gig Ethernet PHY, the automotive industry is aware that it can take advantage of many existing Ethernet attributes, according to Aquantia. Such blessings include “multiple vendors and low cost” and features such as “switching, synchronous, AVB/TSN, security, and topologies,” said the company.
Are they complementary?
It’s important to note, however, that not all long-reach automotive interfaces are created equal. MIPI’s automotive efforts, for example, differs from HDBaseT, observed Ian Riches, director for the Automotive Electronics service at Strategy Analytics. MIPI A-PHY “is a point-to-point asynchronous link. Huge bandwidth in one direction, limited in the other. It’s a simple data pipe and doesn’t natively support USB, Ethernet, and power like HDBaseT.”
Nonetheless, MIPI will compete with other multi-gig connectivity like Automotive Ethernet “for connecting a high-bandwidth source (e.g., a camera) to a central controller,” said Riches.
Asked to compare these competing connectivity solutions, Riches said: “Ethernet is great if you want a network. HDBaseT allows support for things such as power transmission, USB, etc. MIPI A-PHY is a more direct competitor for proprietary LVDS solutions such as APIX, FPD-Link, GMSL, etc. It’s ideal for simply moving data from one place to another, with a limited back channel for control.”
Reuse of MIPI mobile protocols?
Rick Wietfeldt, MIPI Alliance board member and senior director, technology at Qualcomm, described Ethernet, MIPI, and HDBaseT as “all complementary technologies that have a place in the automotive ecosystem.”
He explained to us that MIPI brings three specific benefits to the automotive industry — namely, “performance, reuse, and ecosystem.”
More specifically, MIPI’s A-PHY offers “the benefit of asymmetric links tailored for camera, sensor, and display connectivity — an efficient approach that also differentiates it from other technologies.” Wietfeldt argued, “While Ethernet is a widely adopted, mature solution, it’s also relatively new to the automotive market and will have to be based on new cables, connectors, Ethernet PHY, and switch products.”
Another key [benefit] is that “A-PHY is intended to service high-speed ‘streaming applications’ such as camera and display applications that require strict latency control in point-to-point architectures.”
Wietfeldt observed, on the other hand, “Many view Ethernet as a ‘networking’ (often ‘many-to-many’) architecture interface primarily (although not exclusively) for high-speed control applications.”
He added, “Note also that the current automotive Ethernet is 1 Gbps, with a plan to 10 Gbps in about the same timeframe as the MIPI A-PHY.”
Indeed, while Aquantia is the first to introduce “pre-standard Multi-Gig (2.5G, 5G, 10G) Automotive Ethernet PHY,” the IEEE Multi-Gig Automotive Ethernet PHY standard, isn’t ratified yet. It still faces a normal IEEE standard setting procedure.
Asked about the reuse of MIPI technologies, Qualcomm’s Wietfeldt explained that with MIPI A-PHY, it’s possible to make “direct efficient implementation of MIPI mobile protocols CSI/DSI, compatible with silicon used in billions of smartphones to date and already prevalent in automotive today.”
Will Ethernet have the last word?
Aquantia does not agree with the MIPI Alliance’s characterization of Automotive Ethernet. Bar-Niv claimed, “At the end of the day, only Ethernet has already built-in pretty much of all the features that have been identified by the OEMs to truly deliver the In-Vehicle Network required for future Level 4/5 autonomous vehicles.”
He noted, “MIPI seems to focus on a specific use case/topology of point-to-point connections. Ethernet not only solves that use case already, among others (built-in support for star, mesh, ring, daisy-chain, etc.), but brings to the table all of the other IEEE standards needed for Level 4/5 autonomous driving (reliability, security, etc.) that MIPI and others like LVDS and HDBaseT have to develop from scratch.”
Bar-Niv also added, “With regard to speed, Ethernet already has standards for MAC rates up to 100G, which open the door for much higher speeds in future automotive networking.”
Although MIPI A-PHY is not a competitor for Ethernet as a vehicle network backbone, one area where MIPI and Ethernet will clash is “the links out to sensors and displays,” said Strategy Analytics’ Riches.
Point-to-point connection isn’t the only thing that Ethernet does, but the NAV alliance is keen to support the huge data generated by high-resolution cameras, lidars, and radars.
Aquantia explained to EE Times that camera resolutions used in vehicles “have already shifted from 720P to 1,080P and, later, to 4K resolutions.” This means that network speed should increase to 10 Gbps, noted the company.
More importantly, driving this multi-gigabit connectivity today are AV designers who hope to do sensor fusion by using uncompressed, raw sensory data.
Aquantia’s Bar-Niv told us that today’s sensor modules come with pre-processing devices. “A new trend is to eliminate these pre-processing chips to reduce the cost of the sensor modules and reduce the power dissipation.” After all, these sensors are usually located in the hottest places in the car, where performance can degrade significantly as the temperature rises, he explained. If the industry is going to eliminate these pre-processing chips, “the cable needs to be able to carry the raw data at rates of 2 Gbps to 8 Gbps.”
Fully aware of such needs, the MIPI Alliance’s Wietfeldt stressed, “MIPI A-PHY is a high-speed asymmetric streaming interface for sensors, cameras, and displays as an alternative to Ethernet networking.”
The Alliance is confident that MIPI A-PHY will ensure “data integrity while utilizing the lowest-cost, lowest-weight physical channel in the complex, challenging in-vehicle environment.” Beyond developing the MIPI A-PHY spec meeting 12–24 Gbps, which is “well underway,” the MIPI Alliance has also begun “gathering requirements to support higher speeds, including over 48 Gbps, for display and other use cases.”
Who will supply these chips?
Strategy Analytics’ Riches cautioned: “Future use cases based on highly autonomous driving are easy to see — but will not emerge in high volume in the next five years.”
Alluding to the fact that raw data sensor fusion might remain an aspirational goal, he said, “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.”
When it comes to who, among chip suppliers, are supporting which in-vehicle multi-gigabit connectivity, Valens, so far, is leading HDBaseT efforts. Aquatia is ahead on Automotive Multi-Gig Ethernet PHY.
Besides Aquatia, Broadcom, Marvell, and NXP Semiconductors are listed as active on the IEEE Multi-Gig Automotive Ethernet PHY task force.
In the HDBaseT group, NXP is a contributor member. STMicroelectronics is an “adopter member.”
While it is unclear which companies are specifically committed to MIPI A-PHY, the MIPI Alliance listed chip vendors such as Qualcomm, ST, NXP, Intel/Mobileye, and Valens as members.
In the early phase of standards setting, it might be imperative for most automotive chip companies to keep their hands in multiple standards groups.
— Junko Yoshida, Global Co-Editor-in-Chief for AspenCore Media, Chief International Correspondent of EE Times