NXP Readying ‘Whole Vehicle’ Platform

　　NXP Semiconductors announced Monday (Oct. 16) a new automotive processing platform, called NXP S32, designed for OEMs and tier ones to maximize re-use of codes, software and common capabilities across vehicle domains, applications and SoCs.

　　One might call the announcement big on concept, but short on detail, since NXP isn’t rolling out any new chips based on the new platform until the middle of 2018.

　　However, NXP, as the world’s leading automotive chip supplier, hopes to stake its claim in the ADAS and autonomous vehicle market by addressing the thorniest issue of today’s vehicle development: bloated software and a variety of applications too complex to integrate and too disparate to be portable.

　　Referring to a host of automotive MCUs used as ECUs located variously within a vehicle, Matt Johnson, senior vice president and general manager of product lines and software, auto MCU and processors at NXP Semiconductors, told EE Times, “Let’s face it. They are not compatible, they are not scalable and they are not reusable.”

　　With the new S32 platform, NXP hopes to change that. Johnson said, “This new platform [S32] spans the entire vehicle. It allows carmakers to move faster, and develop new models more efficiently. As a result, consumers will have new vehicles on the road sooner.”

　　According to Johnson, “Eight out of top 15 OEMs today are already adopting” the NXP S32 platform, and “more will adopt it soon.”

　　Luca De Ambroggi, senior principal analyst for automotive electronics at IHS Markit, told us that it makes sense for other automotive chip suppliers — like Texas Instruments and Renesas — to take a similar common platform approach. “Nonetheless, others have not yet announced something similar, although they might do it in the future,” said De Ambroggi. “I do not think that [NXP’s rivals] are having such a widespread approach yet, in terms of cross scalability from hardware to software.”

　　Safety, Security, OTA

　　One of NXP’s big promises for S32 is the “common capability” it hopes to provide to every SoC [on the platform], across vehicle domains and between apps.

　　In particular, NXP is focused on safety, security and over-the-air (OTA) update. The plan is to make this capability common on every MCU or SoC – regardless where in a vehicle the chip works.

　　“We are making ASIL D performance everywhere — across multiple spaces,” noted Johnson. “As a developer of safety cores for decades, this is what we do. We’ve figured out on our own how to make ARM cores ASIL D certified.” NXP is applying ASIL D to ARM cores across the board, including Cortex-A, Cortex-R and Cortex-M cores.

　　Defined within ISO 26262, ASIL D — or automotive safety integrity level D — represents “likely potential for severely life-threatening or fatal injury in the event of a malfunction.” It requires “the highest level of assurance that the dependent safety goals are sufficient and have been achieved.”

　　Johnson said NXP isn’t stopping at ASIL D. “We are adding fail operational capability” to every ECU, he said. “When a car recognizes a problem, it not only notifies a driver or a system [in a driverless car] that it isn’t properly working any more, but it also lets its fail operational mode kick in and self-corrects.”

　　How it self-corrects — whether it stops, keeps driving or reroutes — is up to each car OEM to decide. “They have different ways to program that,” said Johnson.

　　But because the silicon is already taking care of the fail-operational capability, car designers face less software programming intensity, added Johnson.

　　Security is also a “common capability” NXP is adding to chips on its S32 platform. Johnson said, “Several years ago, people were still debating if hardware security is necessary for every MCU or MPU to be used in a vehicle. There is zero debate on that today.”

　　By leveraging the company’s bank-card security experience, NXP is “pushing the latest security in automotive” across the platform, he noted. “We prevent [MCUs] from side channel attacks. We are capable of updating ciphers with the newest, higher cipher.”

　　OTA is another area where the auto industry didn’t see an absolute necessity until several years ago. No vehicles, except for Tesla, have been launched with built-in OTA capabilities. But now, every carmaker considers it a must for every connected car — both for future security and safety reasons.

　　While most chips on the market today don’t support OTA, Johnson said NXP is adding the capability across its portfolio. An over-the-air-update can come in quickly, or happen in the background in a drip, drip fashion over hours, he explained. “There will be no need for you to take your car to a garage,” he said.

　　NXP claims that all these common capabilities will be available across products (from small low power ARM Cortex-M, real time optimized Cortex-R and highest performance Cortex-A class performance classes) and across different process technologies ranging from 42 nm to 7nm.

　　Further, a common IP set provides “a consistent development environment via the S32 SDK,” NXP noted. “This allows development effort to be shared across domains and eliminates duplication of multiple software modules.”

　　Is this just ‘positioning’?

　　Until NXP rolls out some actual chips based on the S32 platform, it is hard to judge what tangibles NXP is exactly offering.  Mike Demler, senior analyst at the Linley Group questioned “the true commonality” of chips on the S32 platform.

　　Certainly, NXP is solidifying its MCUs on ARM architecture. But Demler noted, “Even within the same CPU vendor’s offering (Cortex-A/M/R) you have different architectures running different software. If you consider that automotive platforms are evolving toward autonomous, which requires deep learning, it requires specialized architectures running specialized software (including the Cognivue accelerators in NXP’s S32V processors).”

　　Consider “debugging” for deep learning, he noted. “It is completely different from that of other software, and debugging any processor is architecture-specific. It might be nice to use the same tools, but sometimes it’s better to have specialized tools,” Demler said.

　　In short, Demler sees NXP’s announcement mainly as a “positioning” statement.

　　Future proof?

　　Another concern is how extensible and scalable the S32 platform will be when, for example, a completely new processing architecture emerges in the future to handle autonomous driving.

　　It remains unclear what exactly NXP plans to offer in the robo-car space, and whether it has something equivalent to Nvidia’s Pegasus or Xavier SoCs. As Qualcomm’s proposed NXP acquisition still awaits all the regulatory approvals, neither Qualcomm nor NXP is discussing what’s in store in the future of the combined companies’ roadmap in the automotive market.

　　Sidestepping the question about fully autonomous vehicles, Johnson stressed, “Nvidia is offering a ‘point solution’ for one application — L5.” He noted, “You still need 100 ECUs running inside the autonomous car. We hope they will be using some of our ECUs.”

　　When asked how the platform may have to change as the need of the future processing arises, Johnson told us, “We are not stuck [on this platform].” He noted that NXP plans to offer “unique solutions” to “unique requirements,” as the company’s platform continues to adapt.

　　Demler summed up: “Yes, NXP offers the broadest catalog of automotive processors. Yes, they’re all ARM based. Yes, customers no doubt use those processors in a variety of different systems. So it makes sense for NXP to ensure customers they’ll be able to reuse much of the software across multiple product lines, but it doesn’t mean those products address all the processing requirements of future vehicles.”

　　Time to invest

　　At this stage, NXP made it clear that it has completely redesigned its IP across microcontroller families, enabling NXP to forge common functionality across technology nodes with consistent hardware and software behavior.

　　De Ambroggi observed that due to the complexity in handling so many devices in a single platform, from MCU up to MPU, including different “portable” software features and hardware functions (eg.: embedded memory), it was “not at all easy” until now to develop a common platform.

　　He believes NXP investment in this common platform approach is timely. It will, he said, “offer customers a flexible portfolio, while maintaining development cost at a reasonable level.”

　　NXP’s customers already committed to the S32 environment are at a different pre-silicon development phase, according to Johnson. NXP will announce multiple SoCs based on S32 in mid-2018. They will be in vehicles launched in 2020, ramping up in 2021.