Next-generation Central and Zonal Communication Network Topology and Chip Industry Research Report, 2025
The automotive E/E architecture is evolving towards a "central computing + zonal control" architecture, where the central computing platform is responsible for high-computing-power tasks, and zonal controllers are responsible for executing specific control functions.
"Domain-centralized" architecture communication framework:
Various domains form a backbone network through gateways, and data intercommunication and operations are realized through communication protocols such as SOME/IP and DDS, and communication middleware.
A backbone network such as CAN-FD and 100M/1G Ethernet has been formed.
"Central + Zonal" architecture communication framework:
Communication bandwidth improvement: Transitioning from domain controllers to central computing units, which physically concentrate various important computing units, including intelligent gateways, cockpit domain controllers, ADAS domain controllers, and the central computing part of some zonal controllers. This physical concentration directly shortens communication distances and optimizes communication bandwidth by an order of magnitude.
Communication interface upgrade: From CAN-FD and 1G Ethernet to various advanced interfaces such as D2D, 10G Ethernet, fiber optic communication, PCIe5.0, CXL, NVLink, and UCIe.
Integration of high-speed communication and MCU control capabilities: With the rise of advanced functions such as ADAS and autonomous driving, even the most powerful MCU cannot quickly acquire and share data without a high-speed network; conversely, without a powerful real-time MCU, mere communication channels cannot precisely control vehicle behavior.
According to the connection range, automotive communication networks can be divided into in-vehicle networks and out-of-vehicle networks. The in-car network architecture is mainly evolving towards a central ring network architecture, and the application of fiber optic Ethernet in vehicles is advancing; the out-of-car network is divided into short-range and long-range networks, with diverse application scenarios that cannot be supported by a single technology, requiring the collaborative development of multiple technologies such as V2X and satellite Internet.
Application scenarios and trends of next-generation high-speed communication links
Next-generation Central + Zonal architecture passenger cars exchange massive amounts of data in real-time between sensors such as cameras, radars, and LiDARs, high-definition display units, and high-performance central computing units. They also support full-vehicle OTA software updates, remote diagnostics, and functional safety requirements, placing unprecedented composite demands on in-vehicle networks for high bandwidth, low latency, and security.
Such huge data volumes pose unprecedented challenges to data transmission speed and stability. Traditional communication transmission architectures struggle to meet the real-time and smooth data transmission requirements of new-generation automotive intelligence, creating an urgent need for faster and more reliable communication technologies.
(1) Surge in data volume due to improved camera resolution
As the level of autonomous driving increases, the precision requirements for environmental perception become more stringent. In-vehicle cameras, as important visual sensors, are inevitably upgrading in resolution.
1-5MP cameras: Mainly used in surround-view and side-view scenarios, transitioning from 1.3MP to 3MP/5MP.
8MP cameras: Core growth driver in the next 5 years, promoted by upgrades from L2 front-view integrated systems to 8MP, highway (L2.5)/urban NOA (L2.9), and camera mirror system (CMS); 8MP will account for over 35% of total shipments by 2030.
New technologies such as 10+ MP front-view cameras, 4D imaging radar fusion, and light field lenses (commercialization in 2027) will reshape the perception architecture to provide better image quality and more detailed information for advanced ADAS/AD algorithms. Sony has launched a 17MP product with a detection range of 250 meters. High-resolution cameras capture richer environmental details, crucial for autonomous vehicles to accurately identify traffic signs, pedestrians, and other vehicles.
With the increasing proportion of high-level autonomous vehicles and high hardware redundancy among automakers, the average number of cameras per vehicle will grow from 4 in 2024 to 8.3 in 2030, according to ResearchInChina. ADAS camera transmission requires 1 serializer chip per camera, while deserializer chips typically support multiple channels (e.g., 4-in-1), with an average of 4 cameras sharing 1 deserializer.
VelinkTech's self-developed high-speed in-vehicle SerDes chip was successfully mass-produced and installed in the 2026 Lynk & Co 06, marking the world's first large-scale mass production of automotive-grade SerDes chips based on the MIPI A-PHY protocol.
(2) Massive data transmission pressure from improved display resolution
Increased communication transmission requirements in intelligent cockpits stem primarily from improved display resolution, advancing from 720P and 1080P to 2K, 4K, and even 8K. 4K single-screen resolution reaches 3840×2160; 8K is even higher, with exponentially growing data volumes. 4K screens require tens of Gbps transmission rates, with multi-screen setups exacerbating demands. High-resolution content transmission between screens in multi-screen interactions must maintain quality while synchronizing additional data, with dynamic switching increasing load. High-resolution multimedia processing and cloud interactions, such as 4K/8K video, AR functions, and AI features, all consume significant bandwidth.
Rsemi launched a 32Gbps high-performance SerDes chip for in-vehicle displays at the 2025 Qualcomm Automotive Technology and Cooperation Summit. This chip adopts an advanced technical architecture, supports full-rate lossless DP interface solutions, is compatible with speeds from 32Gbps to 3.2Gbps, supports 2 to 4 R-LinC outputs, can directly drive 4×4K displays with DSC (Display Stream Compression) technology, and up to 8 displays with daisy-chain technology, providing rich and detailed display effects and flexible, efficient display system solutions for smart cars. Additionally, the deserializer chip integrates Bridge and OSD functions to further enhance system integration.
Norelsys is gradually building a product matrix covering full-scenario in-vehicle transmission needs through step-by-step iterations: "2G → 3.2G → 6.4G → 12.8G → 25.6G". Norelsys has currently mass-produced over 20 HSMT standard in-vehicle SerDes chips, with product lines covering transmission rates from 2Gbps to 12.8Gbps. These chips can adapt to diverse needs such as different specifications of in-vehicle cameras (supporting up to 17MP), 4D radars, LiDARs, and 4K displays.
(3) "Central computing radar" is an important evolution direction for in-vehicle millimeter-wave radars, with raw ADC data transmitted to central computers via high-speed SerDes
With the evolution of vehicle central computing architectures, central computing radar represents an important development direction for in-vehicle millimeter-wave radars. A "central computing radar" refers to a "simplified radar" in which only RF front-end and minimal preprocessing are implemented. The radar transmits raw data to domain controllers via high-speed buses (e.g., high-speed Ethernet or SerDes) for subsequent post-processing. Its advantages include:
Satellite radars adopt centralized processing and power supply: Centralized processing transmits radar data to a central processing unit, reducing processing requirements around sensors; centralized power supply simplifies system power management, improving energy efficiency, reducing energy consumption, and enhancing radar system reliability and performance.
RF front-end technology will gradually mature: Triggering standardization of communication interfaces for "central computing radars," evolving radars into standard perception sensor components (similar to "cameras," where sensors are no longer coupled with domain control software). This will enable more flexible adaptation and replacement of "central computing radars" in vehicles.
Transmission of raw ADC data: Under end-to-end algorithm architectures, using more raw radar signals (with less information loss) may yield better overall perception performance.
MMICs for central computing radars require higher RF front-end performance but lower processor performance. Currently, TI and NXP have launched chip solutions for central computing radars.
XretinAl Technology launched a 4D radar central computing system based on Black Sesame Technologies' Huashan-2 A1000 chip, which uniformly processes raw radar data in domain controllers via high-speed Ethernet or SerDes.
Application trends of fiber optic Ethernet high-speed communication
In the automotive sector, the rapid increase in sensor number and higher real-time requirements have gradually strained traditional electrical communication methods. From sensors to ECUs and from central computing platforms to display systems, numerous devices require high-speed, stable interconnection. The complex electromagnetic environment inside vehicles further subjects electrical communication to signal interference and reduced reliability.
In 2023, the IEEE Standards Association released the in-vehicle fiber optic Ethernet technical standard IEEE 802.3cz-2023, adding physical layer specifications and management parameters for 2.5 Gb/s, 5 Gb/s, 10 Gb/s, 25 Gb/s, and 50 Gb/s operations over glass fiber in automotive environments.
Currently, fiber optic Ethernet has moved from experimental verification to commercial implementation, building high-bandwidth, low-latency, secure, and controllable in-vehicle communication backbones through CSI packaging, path replication, and multi-interface integration. However, there remain unresolved controversies in in-vehicle fiber optic communication solutions, primarily regarding fiber optic and optical communication components, especially laser selection.
A complete in-vehicle optical communication system consists of fiber optic harnesses, optical modules, and connectors:
Fiber optic harnesses represent the most technically mature component with the highest industry participation, being one of the first key components to evolve from purely electrical to fiber optic.
In-vehicle optical modules operate in harsher environments, requiring stricter specifications including wide temperature range adaptation (-40°C to over 105°C), ultra-long service life (over 15 years), high reliability, and adaptation to various extreme environments.
In-vehicle fiber optic connectors must not only meet conventional performance metrics such as insertion loss and return loss but also maintain stability under high-frequency vibration.
Compared to relatively backward traditional 100M/1G/10G copper automotive Ethernet, China's supply chain has developed competitiveness in fiber optic Ethernet, with automotive-grade solutions available across all links, creating opportunities for leapfrog development. As intelligent vehicles transition to advanced autonomous driving and central centralized architectures, "fiber advancement and copper retreat" has become a viable option.
HingeTech has introduced a communication architecture for automobiles using an all-optical network. Its self-developed high-speed fiber optic TSN centralized gateway architecture enables high-bandwidth, ultra-low latency, low-cost, and highly deterministic transmission of massive in-vehicle network communication data via fiber optics, supporting a maximum transmission rate of 10Gbps with excellent EMC performance. This architecture is primarily applied in systems including ADAS, autonomous driving, 360° surround-view, in-vehicle infotainment, BMS, and centralized computing architectures, with a maximum transmission bandwidth of 25Gbps.
EEA optical communication architectures built on optical modules connect multiple optical modules with multiple zonal gateways, which can be replaced with other controllers such as T-Boxes and domain controllers as needed.
In hardware design, BTB connectors link optical modules and zonal gateways, with data and control signals transmitted via interfaces such as MIPI-CSI, SGMII, I2C/SPI, and GPIO.
Optical modules and zonal gateways are placed in different vehicle zones, with nearby ECUs connected to adjacent optical modules or zonal gateways. If zonal gateways receive traditional CAN or LIN signals, they transmit them to optical modules for conversion to optical signals for processing by the central computing platform. Different zonal gateways can exchange data via optical modules.
Optical modules are primarily responsible for fiber optic signal transceiving, receiving GMSL2 camera signals and fiber optic Ethernet camera signals, receiving fiber optic LiDAR signals, and forwarding fiber optic signals. EEA optical communication architectures built on optical modules enable high-speed, low-latency transmission of large data flows with beneficial EMC performance while remaining compatible with traditional networks.
Li Auto is collaborating with HingeTech to develop an in-vehicle optical communication test bench, which has completed A-sample delivery. This test bench incorporates jointly developed in-vehicle optical communication Ethernet technology, with core components including in-vehicle optical modules, in-vehicle fiber optic connectors, and in-vehicle optical fibers.
In 2024, Dongfeng collaborated with Yangtze Optical Fiber and Cable (YOFC) to complete the first research phase, achieving the transition from industrial-grade to automotive-grade cable assemblies. The research comprehensively validated performance under extreme environments such as high temperatures (125°C) and high vibrations (V3 level), completing a full verification process from assemblies to individual components and from test benches to actual vehicles, ensuring applicability across all vehicle environments including cockpits, chassis, and roofs.
Research focused on designing and optimizing optical fibers, cables, and connectors, ultimately producing automotive-grade cable harness assemblies with complete optical, mechanical, and environmental characteristics. Rigorous verification confirmed stable operation under various complex environments including extreme cold and heat. Component verification included 53 key tests covering optical performance, mechanical strength, and environmental adaptability. Bench testing evaluated over ten indicators including Ethernet communication functionality, robustness, and voltage stability according to national standards (e.g., GB/T 24581, QC/T 2910) and enterprise standards.
For vehicle testing, the Dongfeng Eπ007 model completed 12,000 km of extreme road testing in Xiangyang, including bumpy and high-vibration scenarios, with stable communication and no packet loss.