As efforts to develop software-defined vehicles (SDVs) intensify, the technology that powers vehicles loaded with increasingly sophisticated computing, networks, connectivity, sensing and electric propulsion needs to keep pace. This is true not just for SDVs being designed and produced over the next few years, but also for even more-complex architectures that will follow, while also allowing existing vehicles to be maintained and upgraded with the latest features and services.

At its recent Connects developer event in Silicon Valley, semiconductor supplier NXP revealed the latest details and updates for its S32 Vehicle Compute Platform. NXP states that this is a scalable, flexible and optimized platform for the requirements of SDVs both now and into the future. “Our platform needs to be able to do things we’re not even thinking about today,” said Brian Carlson, NXP’s director, global product and solutions marketing. “The auto industry is going through a huge transformation and it’s all about delivering new services to drivers.”

These services will include cloud-connected services that can be added throughout vehicle lifecycle during development, testing, production and post-sale, Carlson said. Features will now be delivered via software not on just ECUs onboard the vehicle but also on virtual cloud-based ECUs, he added. As a result, vehicle compute platforms will need to deliver new forms of vehicle-data intelligence to provide continual and contextual improvements, decouple vehicle hardware and software and allow differentiation for competitive OEMs.

It also will require OEMs and suppliers to work together and remove existing silos and work in real time, Carlson says. “This is a new approach for the industry.”

Evolving to Fit SDV Needs
The NXP S32 platform, which initially debuted in 2017, is designed for future vehicle architectures with distributed electrical and communication systems that require complex traffic engineering, as well as to fit the needs of increasingly connected and electrified vehicles. The S32 has been continually updated with innovations such as new cross-domain integration, vehicle network and radar and imaging processors, as well as MCUs and SoCs for zonal electrification and central vehicle computer SoCs. It also features new semiconductors that can handle both critical and noncritical vehicle systems and applications and bridge edge and embedded functions.

The March 2023 introduction of the next-generation S32G3 series vehicle network processors combines ASIL D safety, hardware security, high-performance real-time application processing and network acceleration. The S32G3 series also supports new vehicle architectures such as service-oriented gateways, domain controllers and zonal and safety processors. The S32G3 offers up to 2.5x the performance, memory and networking bandwidth as the previous S32G2 series.

The new S32G3 processors are package pinout- and software-compatible with the S32G2 processors, allowing customers to scale approximately10x in performance to support multiple product tiers and increase product performance over time. OEMs are increasingly requesting performance overhead in electrical architecture platforms, Carlson noted.

In May, NXP also announced a collaboration with semiconductor contract manufacturer Taiwan Semiconductor Manufacturing Company Limited (TSMC) to produce the industry’s first automotive embedded Magnetic Random Access Memory (MRAM) in 16 nm FinFET technology to speed the transition to SDVs and support multiple generations of software upgrades on a single hardware platform.

According to NXP, this allows updating 20MB of software code in about 3 seconds, compared to flash memories that take up to 1 minute, reducing the downtime for software updates and eliminating chokepoints triggered by long module-programming times. MRAM also provides highly reliable automotive-grade technology by offering up to 1 million update cycles, a level of endurance that’s 10x greater than flash and other emerging memory technologies, according to NXP.

Virtual silicon and digital twins
A longstanding issue in automotive is the disparity in development time between technology and new vehicles. New vehicle platforms can take several years to go from concept to predevelopment, series development and the early stages of production. NXP’s latest version of S32 can reduce this lead time by leveraging virtual silicon development and running it on digital twin of a future vehicle, Carlson said. This allows OEMs to virtually develop software in parallel during predevelopment or even prior to hardware design and continually refine it throughout vehicle design and production.

A virtual approach not only helps automakers reduce vehicle development time and consequently vehicle deployment, but also represents a “vehicle lifecycle transformation,” Carlson said. “And without this, we don’t have software-defined vehicles.” Software deployed throughout the lifecycle of the vehicle can create better user experiences, he added.

“Vehicle health apps, for example, will better assist with recalls, maintenance and look for aging components, and we have the hardware that’s going to do it,” Carlson said. He predicts a doubling of value for drivers and revenue for OEMs, and that AI and machine learning will accelerate this. “Machine learning picks the needle out of the haystack of data.”

At 2023 Auto Shanghai, Chinese automaker GAC’s Aion division unveiled its Hyper GT premium EV sedan and started presales as the first car in the world to adopt the NXP S32G3 as a central vehicle-computer processor. “In China they can move much faster and deploy within 18 months because they don’t have the legacy issues [of incumbent OEMs],” Carlson said.

Ron Martino, NXP executive vice president, global sales, predicts an increase in the use of NXP processors and machine learning as electric vehicles proliferate. “Machine learning is deployed to better understand the condition of the battery and across different subsystems so that we can truly monitor the vehicle of the future and optimize it with more intelligence,” he said. “Whether that’s for performance, to offset a future need such as mechanism a wear or for over-the-air updates, intelligence and upgradeability is becoming more and more important.”

Machine learning can be used on all S32 processors across a vehicle, Carlson added. “We haven’t really made a big noise in the market yet and this is a preview,” he explained. “We announced product a year and half ago for ADAS and vision and we’ve extended it across the whole range of S32 processors – from the control processor and to the gateways.”

“We can optimize the platform for every target so [our customers] can start to do machine-learning exploration,” Carlson said. “It’s a whole dev-op environment that gives the ability to target and optimize the model performance and latency. If you take that model and run it on different cores or DSPs you can see the difference in performance. There will be dozens of machine-learning models and this is how we can move them across the vehicle.”

This also allows differentiation and competitive advantages among NXP’s various OEM customers, Carlson added. “We can leverage machine learning to provide the tools that allow [OEMs] to optimize the machine learning on our platform and create new use cases And do it efficiently so they don’t have to write their own automotive-grade inference engines that can port across all their platforms.”

According to Carlson, NXP’s latest technology is being implemented by its customers now. “Depending on what version it is, it will go into production at the end of the year or the first quarter of next year”