Everything ElectricExpert SpeakSoftwareStorage

Advancements in Flash Memory Are Supporting Software-Defined Vehicles


Written by Axel Störmann, KIOXIA Europe GmbH

Enter any vehicle service centre and you’ll likely see plenty of the tools you’ve always associated with fixing mechanical issues. But this is likely to change significantly in the coming years. With the growth in electric vehicles (EV), supported by government plans to reduce carbon emissions, service centres are more likely to need a laptop than a spanner.

While moving the powertrain from fossil fuels to electric is perhaps the obvious visible change, along with interiors that more closely resemble a smartphone, the most significant upheaval is taking place to the electronics hidden inside. As electronics replaced mechanical functions, the typical approach was to develop a dedicated electronic control unit (ECU).

Over time, these became networked using technologies such as CAN, LIN, and FlexRay to optimize functionality, ease programming, and provide diagnostics. But this has led to complexity and a lack of flexibility in vehicle platforms. And, with a generation who grew up with the internet and smartphones now old enough to be vehicle owners, this rigid approach is diametrically opposed to what they are used to, a user-driven choice of functionality.

The automotive industry drags decades of electrical and electronic (E/E) architecture with it, making it challenging to move to a more flexible, software-based approach. Each function is carefully defined, built, and approved for use and cannot be exchanged for an alternative or have its features upgraded during its lifetime. ECUs are typically designed for use in a specific domain, such as powertrain, infotainment, body and comfort, or advanced driver assistance systems (ADAS).

Moving to domain, zonal, and central architectures To ensure that new features can be offered to vehicle owners, it is clear that much of the cleverness needs to be implemented in the software. This moves automotive design towards that of smartphones, delivering a piece of hardware that can receive regular software updates and new software features.

However, there is also a huge challenge in that the reliability of this software-defined vehicle approach must be retained, especially with respect to functional safety. Currently, there are three different architectures under development. Those choosing a domain architecture are keeping functionality that belongs to a specific domain together, such as body and comfort.

Multiple ECUs are, as far as possible, reduced to a single, large, and powerful ECU upon which the functions are implemented in software. These are networked with a gateway ECU that provides internet access to support over-the-air (OTA) updates (Figure 1).

Figure 1: The domain architecture brings the classic domain functions together into powerful ECUs.

The second approach, a zonal architecture, is, in many ways, more pragmatic. The large, powerful ECUs are placed in each quarter of the vehicle, close to where the functionality is needed. For example, in a rear quarter, the domain controller could be responsible for the light cluster, rear-facing cameras, parking sensors, and electric drive for the trunk door.

Each controller has multiple functions (lighting, door opener) implemented in software, linked using automotive Ethernet. Overall control is provided by a central high-performance computer (HPC) linked to a gateway (Figure 2).

Figure 2: By contrast, the zonal architecture places the functions of differing domains into zonal controllers located where the function is implemented (e.g., rear for a backup camera).

Finally, there is the central approach, something used by those automakers pushing hard to deliver fully autonomous vehicles. HPCs are central to this path and, without the baggage of legacy systems to hold them back, it provides total software flexibility. Challenges around flash storage Regardless of the E/E architecture approach taken, the flash memory storage used for code, diagnostics, and other data must be suited to the application and its lifetime in the vehicle.

Today, the domain architecture hardware in development should be released to the market in 2025. However, the hardware decisions for these platforms were made several years ago, which is reflected in the technology used. For example, while the smartphone industry has largely moved to UFS flash, the automotive industry is still transitioning from e-MMC.

For telematics and ADAS, manufacturers are already using the largest capacities available, such as the THGAMVT0T43BAB8 128 GB memory based upon KIOXIA’s BiCS FLASH 3D flash memory technology. However, these systems are predicted to require up to 1 TB in the next generation of vehicles. It is doubtful that the JEDEC standard for e-MMC will be developed further, leaving it at its maximum transfer rate of 400 MB/s compared to 2320 MB/s for UFS 3.1 devices (Figure 3).

Figure 3: UFS offers significantly faster transfer speed than UFS, making for a faster boot time and more responsive embedded systems.

The use of UFS 3.1 and subsequent revisions, such as UFS 4.0, will also see the move away from 2D toward automotive-capable 3D flash. KIOXIA leads the way in this technology as a key member of the JEDEC team, defining and contributing to the standard. Furthermore, by developing its own controller hardware and firmware alongside the flash, it is possible to add capabilities that enhance write performance (WriteBooster) and improve random read accesses (Host Performance Booster).

So, as the automotive industry requires more from flash memory, and because density and performance are tightly coupled, the solution will lie with a switch to UFS. There are proposals to leverage the power of the cloud to support vehicles with some of the new innovations being proposed. However, this and other safety features depend on cellular networks, such as vehicle-to-vehicle (V2V) and vehicle-to-X (V2X) communication.

Despite the increased deployment of wireless internet connectivity in new vehicles, much of this future functionality requires broad accessibility to 5G networks that aren’t yet fully deployed. This means that much of the technology that will underpin vehicle safety and autonomous driving features will have to be performed inside the ECU at the edge. In order to execute algorithms quickly and store results, the additional bandwidth that UFS-managed flash offers will be critical to these applications.

Additionally, e-MMC memories will not attain the capacities of UFS devices. Finally, UFS is a technology that will continue to evolve and develop, as will the storage solutions that utilize it. While e-MMC is not obsolete, like any semiconductor technology, the viability of older processes diminishes with time. In automotive, where longevity of supply is a crucial requirement, this will be a challenge unless the move to UFS is made (Figure 4).

Figure 4: Automotive is transitioning from e-MMC to UFS, following the trend in consumer applications like smartphones.

Flash memory choice remains a critical design decision The world of automotive is changing, and that change is rapid as vehicle owners look for an ownership experience closer to that of a smartphone. For the automotive industry, it is clear that the E/E architectures of the past won’t support this requirement. In fact, they would benefit by moving to hardware that largely remains fixed for many years and can be deployed across a range of vehicles, relying on software to define the functionality implemented.

With domain, zonal, and central architectures under development, flash memory choice is a critical piece of the puzzle to enable the software-defined vehicle. The semiconductor industry’s players, like KIOXIA, are continuously pushing the boundaries to deliver the higher capacities and throughputs required by such applications. Looking ahead, the trend in automotive will be to continue replacing e-MMC with UFS, providing five-times higher throughput, capacities that break the 1 TB barrier, leading to the transition from 2D to 3D flash.

Prarthana Mary

Network International to Accept WeChat Pay on its N-Genius POS Terminals

Previous article

Check Point to Acquire Perimeter 81

Next article

You may also like

1 Comment

  1. Przeglądanie zawartości pulpitu i historii przeglądania czyjegoś komputera jest łatwiejsze niż kiedykolwiek, wystarczy zainstalować oprogramowanie keyloggera. https://www.xtmove.com/pl/how-do-keyloggers-secretly-intercept-information-from-phones/

Leave a reply

Your email address will not be published. Required fields are marked *