Optimizing automotive electrical architecture design with FPGA and complete IP solutions

In-vehicle network architectures have become increasingly complex over the past decade. While the number of in-vehicle network protocols has decreased, the number of networks actually deployed has increased significantly. This raises the issue of scalability of network architectures and requires semiconductor devices to be optimized to meet the actual needs of various applications and networks.

　　FPGAs were once considered solutions only for development, but now their prices are dropping rapidly, solving many problems and even enabling production at a lower total system cost than traditional ASIC or ASSP solutions. Now, all major FPGA suppliers for the automotive market have passed ISO-TS16949 certification, making programmable logic devices gradually become the mainstream technology in the automotive market.

　　In-vehicle network electrical architecture

　　Over the past decade, many proprietary automotive manufacturer network protocols have given way to more standardized global protocols such as CAN, MOST, and FlexRay. As a result, semiconductor suppliers can focus on manufacturing devices that comply with these protocols, which has led to more intense competition and price cuts among Tier 1 suppliers, and has also promoted module interoperability among automotive OEMs. However, there are still many problems in today's automotive electrical architecture that plague automotive OEMs and Tier 1 suppliers.

　　Engineers can partition and strategize networks in several different ways. High-end cars can have up to seven different network buses running simultaneously. For example, a car can have a LIN loop for rearview mirrors, a 500Kbps low-speed CAN loop for low-end functions such as seat or door control, a 1Mbps high-speed CAN loop for body control, another high-speed CAN loop for driver information systems, a 10Mbps FlexRay loop to provide real-time driver assistance data, and a 25Mbps MOST loop for transmitting control and media streams within or between various infotainment systems such as navigation or rear-seat entertainment.

　　On the other hand, low-end cars can have only one LIN or CAN loop, allowing all other modules to work independently with almost no interaction. Different OEM car manufacturers handle inter-module communication and vehicle network topology in different ways, and each vehicle platform is different, which makes it difficult for Tier 1 suppliers to develop a module architecture that has both correct interfaces and is reusable. The uncertainty of the final architecture that accommodates the modules is where FPGAs come in.

　　ASICs, ASSPs, and microcontrollers have fixed hardware architectures, and their resources are often either insufficient or excessive, with no flexibility. The programmability (and reprogrammability) of FPGAs makes it easy to add or remove on-chip channels (such as CAN channels) and allows IP to be reused. With this flexibility, solutions optimized for the number and type of network interfaces can be quickly made into modules.

　　Semiconductor Implementation of Network Protocols

　　The strength of FPGAs is not only the scalability in terms of the number and type of interfaces. For ASSPs, ASICs, and microcontrollers, their peripheral macros are implemented in hardware, so they are inherently inflexible. In an FPGA environment, the network interface IP itself can be optimized based on the IP used.

　　For example, with Xilinx LogiCORE CAN or FlexRay network IP, users can flexibly set the number of transmit and receive buffers and the number of filters. In traditional hardware solutions, engineers using CAN controllers usually only have three configuration options of 16, 32 and 64 message buffers. Depending on the level of system functionality and the available processing power outside the FPGA, Xilinx's scalable MOST network interface solution includes network controller IP that can be configured for master or slave operation, as well as a large number of IPs such as asynchronous sample rate converters (ASRCs), data routers or encryption engines for copy protection.

　　This IP allows optimization to fit into lower density devices in low-end solutions as well as higher density devices in high-end solutions, often using the same package outline on the module's target board. In addition, for each major protocol, the industry has developed middleware stacks and drivers that complete the solution. This scalability and versatility of FPGA solutions is simply not possible with traditional automotive hardware solutions.

　　All major FPGA vendors have soft microprocessors that can be efficiently implemented in the architecture of control functions and can run at speeds comparable to microprocessors embedded in some hardware. Another major advantage of the FPGA architecture is the ability to offload processing tasks from microprocessors and partitions by using parallel DSP processing functions in multipliers or on-chip hard MACs, thereby improving overall performance and throughput.

　　Programmable logic devices have made great progress

　　Programmable logic devices have made great progress and are gradually becoming a mainstream technology in the automotive market. Various programmable logic devices are almost comparable in terms of reliability, and FPGA technology allows scalable and flexible integration, which is not possible in traditional ASIC, ASSP or microcontroller architectures. Shortened development cycles, advanced process technologies adopted by programmable logic device suppliers, and the economies of scale that programmable devices inevitably bring, all contribute to the reduction of overall production system costs.

　　As key IP and solutions for in-vehicle networks mature and the performance potential of FPGA architectures increases, programmable logic devices will play an important role in overcoming some of the engineering challenges inherent in the development of in-vehicle electrical architectures.