By: Bob Siller – Senior Marketing Manager, Achronix

For anyone who has researched a new car lately, it’s hard not to notice how rapidly automotive electronics are evolving. Just compare car safety technology from three years ago with today’s technology and you’ll see that the number of cameras has increased significantly to support things like panoramic vision, driver distraction monitors, stereo cameras, forward-facing cameras and Applications such as multiple rear view cameras. In addition to the cameras, system features have also been enhanced, including automatic emergency braking, lane departure warning, rear blind spot detection and traffic sign recognition. This trend shows that the automotive electronics category continues to innovate at a rapid pace, but it also presents new challenges for automotive original equipment manufacturers (OEMs), including:

When the average time to develop a new car is reduced from 48 months to around 24 months (see the impact of shortening the development life cycle and reducing design complexity on the automotive industry), while also supporting life cycles of up to 10+ years How to quickly develop new features and security functions?

The platform architecture is determined several years ahead of the vehicle launch. How to predict the required hardware and software requirements?

How to improve the performance and quality of automotive electronics to meet user expectations when car users compare consumer electronics?

Application-Specific Integrated Circuit (ASIC) Solutions

To address these issues, automotive OEMs need to rethink their system architectures and add flexibility to their designs to be able to make changes later in the design process and add features and capabilities that are continually being developed. This redesigned system architecture differentiates OEMs from other competitors and provides solutions to technical challenges that cannot be solved using standard hardware components.

Traditionally, device selection for advanced driver assistance systems (ADAS) architectures has relied on off-the-shelf devices designed to support specific automotive safety features and environmental requirements. This approach is ideal for non-differentiated features such as anti-lock braking systems (ABS). However, if you are designing innovative security features, many OEMs choose to build their own custom ASICs. Custom ASICs offer the lowest overall cost and highest performance compared to any off-the-shelf device: however, ASICs come with risks. If you don’t identify all the required features, you won’t be able to make changes without doing a costly and time-consuming redesign of the ASIC.

Despite the inherent risks, one innovative car company, Tesla, believes that the ASIC route is their best bet for delivering advanced security features. Tesla has developed its own Full Self-Driving (FSD) ASIC for this purpose (see “Tesla Full Self-Driving ASIC”). The research and development started in February 2016 and lasted 29 months, and it was not until July 2018 that ASIC was certified for production. For nearly three years before selling cars to customers, Tesla needed to determine the ASIC capabilities it needed. They decided to adopt an architecture that included:

Integrated camera interface running at 2.5G pixels/sec

128-bit wide LPDDR4 memory

Camera Image Signal Processor (ISP)

·H.265

· Graphics processing unit (GPU) that supports FP32 and FP16 precision

12 ARM A72 central processing units (CPUs)

Figure: Tesla’s fully autonomous driving architecture display

future-proof hardware

But what if they think the camera they will be using is discontinued, needs a new camera ISP for better low light performance, or needs a GPU to support block floating point instead of FP32? These things could lead to ASICs being redesigned or major architectural changes that could delay their future vehicle production. Wouldn’t it be a better idea to build flexibility into the ASIC to support adding these features later in the design cycle to avoid ASIC redesign and delays in production?

Fast forward to 2020, and we are seeing automotive ASIC manufacturers address this challenge by developing new ASIC architectures including embedded FPGA (eFPGA) semiconductor intellectual property (IP). Although eFPGA was not a mainstream technology when Tesla developed its full self-driving chip in 2016, it is now mainstream. Achronix eFPGA IP is now in production in several high-volume applications, proving Achronix has the capability to achieve mainstream production. Now, we are seeing eFPGA IP being adopted by automotive OEMs to embed hardware programmability to address these new challenges.

Figure: ADAS ASIC architecture based on eFPGA IP

HD Cameras: HD cameras

5G Wireless: 5G Wireless

77GHz Radar: 77GHz Radar

Custom ASIC: Custom ASIC

Analog Front End: Analog Front End

eFPGA: Embedded Field Programmable Logic Gate Array​

Sensor Interface: Sensor Interface

Pixel Pre-processing: Pixel Pre-processing

Radar Data FormatTIng: Radar data format finishing

V2 Crypto: V2 encryption

Video/Radar Fusion: Video/Radar Fusion

Network Processing: Network Processing

Custom ASIC Accelerators: Custom ASIC Accelerators

CPU ApplicaTIon Layer: CPU application layer Display Interface/Format Conversion: Display interface/format conversion

Vehicle Network Interface (CAN/Ethernet): Vehicle Network Interface (CAN/Ethernet)

Infotainment Display: Infotainment Display

This architecture is possible using Achronix’s Speedcore eFPGA IP technology. Speedcore eFPGA IP provides a competitive advantage over traditional ASICs by helping manufacturers reduce development time while allowing the addition of new and innovative capabilities that can be defined after the ASIC hardware architecture is finalized. eFPGA IP is truly a game changer for the automotive industry. Achronix is ​​excited to see that eFPGAs can bring tremendous value to future automotive ASIC designs.

Leave a Reply

Your email address will not be published.