Software-defined vehicles (SDVs) are driving the evolution of automative electronics. Emerging over-the-air (OTA) capabilities and advanced infotainment systems are leading to more centralized data storage and placing increased demands on internal connectivity.
While a great deal of technology innovation with modern vehicles has been around autonomy and electrification, research firm IDTechEx notes that increasingly cars are becoming not just a means to travel but a base that uses smart features to integrate work and lifestyle into the vehicle. One example the firm cites is third-party apps that will enable more convenient drive through purchases that can be charged directly to the car rather than the driver having to stretch out to tap with a bank card or smartphone.
Other innovations include OTA capabilities that will enable the car’s software to be upgraded overnight and more detailed, up-to-date navigation information, while onboard entertainment might include support for Xbox gaming controllers, IDTechEx predicts.
Every SDV features requires more memory and networking, which are driving a more zonal architecture within the vehicle that feeds into central compute.
Networking is the nervous system of an SDV
Chris Mash, vice president of business development at Ethernovia, said the networking within the vehicle can be viewed as the “nervous system” of the car, moving any data generated that is necessary for supporting vehicle functions from ECUs to SoCs and other data storage systems.
A SDV leverages a zonal architecture approach to electronics. It groups functions and centralizes compute even as a great deal of data is collected from sensors and actuators by ECUs to enable features, including hardware enabled by software.
In the past, automotive systems have been siloed with their own domain controller and separate ECUs hanging off, Mash said. As more services and features were added, ECUs proliferated, as did the many sensors for collecting data used by the vehicle. He said an SDV requires an array of compute capabilities that can sit in a very complicated network and have access to all data types in the vehicle. “These old domain architectures don't facilitate that, so most of these networks are isolated.”
Convergence to support OTA subscription services that generate new revenues for carmakers is what’s driving SDV advances, Mash said. Rather than having the advanced driver-assistance system (ADAS) run its own wires and compute, it is a domain with a set of ECUs. Data comes into the ECU and is aggregated in the ADAS domain, with all domains for infotainment, body or chassis supported by an Ethernet backbone to move data around from wherever it's generated or created.
For an SDV, Ethernet must be more than “best effort,” Mash said. “That's not particularly good in the vehicle.” Over time, there’s been a range of features added on top of the Ethernet protocol which make it time bound and deterministic to guarantee delivery of data, he said. “That’s a fundamental requirement to be able to use Ethernet in a vehicle for mission critical services.”
Software complexity drives hardware requirements in SDVs
Ethernovia is focused on moving the data wherever it needs for both modern and legacy ECUs, Mash said. “We're building that nervous system of the car,” he said. “There will be this pain point of transition between these older architectures and the new architectures.”
SDVs are accompanied by increased software complexity, Mash aid. “The OEMs are finding that quite difficult, and because they come from a world where they build cars, they're not software companies.” He said there is a huge Ethernet talent pool that can be tapped to provide resources for OEMs, but the EV companies that are technology companies like Tesla are miles ahead of everyone else.
As automakers look to decouple hardware development from software and lean on APIs to enable the hardware and software to talk to each other, companies like NXP are looking to simplify SDV design.
NXP recently launched its S32 CoreRide platform that combines combine processing, vehicle networking and system power management with integrated software. The platform uses the company’s new S32N55 processor for central compute to provide what it calls “super-integration” of essential real-time vehicle functions, while zonal control is provided by four S32K344 MCUs and edge nodes based on NXP S32K146 and S32M244 microcontrollers.
In a recent online presentation, Janna Garafolo, NXP’s director of automotive application engineering, outlined a few use cases that demonstrate the platform's capabilities to simplify the SDV development and how it can support the delivery of modern automotive subscription services.
ECU consolidation reduces materials costs
The NXP platform provides hardware isolation by running virtual ECUs on a single processor so that safety critical and non-safety critical applications can run on the same processor together without interfering with each other, Garafolo said, which allows for the consolidation of what typically are individual ECUs or boxes inside the vehicle into a single processor. This saves hardware and Bill of Materials (BoM) costs, she said.
Isolation is also beneficial if there is fault with one system in the vehicle, such as an HVAC fan that stops running – the S32 CoreRide architecture ensures that the car continues to run, Garafolo said. “This is a non-safety critical application. It's completely isolated inside the processor, so nothing else is interfered with in terms of performance and operation of the other applications.”
Garafolo said the S32 CoreRide platform enables an architecture that supports cloud-based data collection and subscription services to drivers after the car is purchased, including vehicle personalization features based on driver preferences.
The SDV evolution is changing how semiconductor companies like Infineon Technologies are engaging with automotive OEMs. Ray Notarantonio, senior director of vehicle user experience and EE architectures, said even as software becomes more complex, Infineon is looking to bring it in-house, which means it is designing the ECU that contains Infineon components. “It's changing a lot of our dynamics.”
Although these dynamics are more complicated for Infineon, Notarantonio said it creates opportunities to provide electronics content for SDVs, which are essentially vehicles that can transform themselves. “Now people want to have a software defined vehicle where a vehicle can be upgradeable, which means that the vehicle actually has a lot of hardware that may not be used on day one,” he said. “You get a subscription if you want to implement it.”
Computing remains distributed
That’s where a lot of semiconductor content comes into play to support a computing stack that includes software, Notarantonio said. “Now the OEMs now are trying to centralize computing, but they still have all the distributed computing they had before,” he said. “You're not going to control a seat motor from some central processor.”
Notarantonio said that SDVs are using an architecture that has central data storge, but there is still going to be local storage. “That central storage system still needs to send that software to each of the ECUs that's performing its function, he said “With a software defined vehicle, you're moving out of this mode of how the OEMs used to work, which is a function out of box.”
Breaking down the SDV into zones allows for content consolidation while harmonizing software development, Notarantonio said. Owning the software means OEMs don’t have to pay for changes every time that want to add or update features, he said.
A zonal architecture makes the vehicle is more programmable, but it is also driving more Ethernet, hardware and semiconductor content. “Anytime computing goes up, memory's going to go up,” Notarantonio said, but it’s also driving the need for more power electronics, too. “There's a big demand for increased power semiconductors in these vehicles outside of an inverter for an EV or an onboard charger.”
An SDV is also a connected vehicle, Notarantonio said. “A connected vehicle now is a vehicle that can be under attack.”
Longer term, computing capabilities are going to go up, he said, even as the number of ECUs and other computing devices go down. The compute in systems like the ADAS are going to be more powerful, but it will rely on a lot of microcontrollers supporting the functions that make a car an SDV. “Everything that occurs in the vehicle at low voltage has to be smarter,” Notarantonio said.
There will be some memory consolidation as SDVs mature, he said, but it will follow computing needs. “There's going to be some hubs of computing for sure, but there will still be distributed computing at least for the next 10 years.”