Ethernet Connection to the SDV Endpoints Automotive Ethernet as a Unified SDV Network

Software-defined vehicles rely on deterministic, high-bandwidth communication to support ADAS, centralized computing architectures, and over-the-air updates. In zone-based SDV architectures, Automotive Ethernet serves as the central network connection for this purpose.

Even today, high-data-rate cameras (Figure 1) rely on proprietary point-to-point serialization links based on low-voltage differential signaling (LVDS) technologies such as Gigabit Multimedia Serial Link (GMSL) or Flat Panel Display Link (FPD-Link). These connections terminate directly on system-on-chip (SoC) devices and operate outside the Ethernet infrastructure. As a result, stream sharing, path redundancy, and dynamic bandwidth allocation for high-quality perceptual data are significantly limited.

Extending Ethernet all the way to the sensor level eliminates these limitations and enables a unified, programmable SDV network from the camera to the central processing unit.

Converting Visual Data at the Ethernet Edge

As shown in Figure 2, the introduction of MultiGig-capable single-pair Ethernet PHYs for the automotive sector, with data rates of 2.5, 5, and 10 Gbit/s, closes the bandwidth gap that has previously limited cameras to proprietary serialization connections. At these data rates, Ethernet enables the efficient, low-latency transmission of high-resolution video data directly from the sensor.

An Ethernet camera bridge converts the sensor data into standardized Ethernet frames right at the source. The IEEE 1722 standard defines the Audio Video Transport Protocol (AVTP), which encapsulates time-critical media streams at Layer 2. The bridge acts as an AVTP talker and segments each video frame into protocol data units that are tagged with stream identifiers and sequence numbers. Downstream processors act as listeners and reconstruct the frames deterministically.

Each packet contains a presentation timestamp that is synchronized via the Generalized Precision Time Protocol (gPTP) in accordance with IEEE 802.1AS. This mechanism ensures precise temporal synchronization with radar, lidar, and inertial sensors. An exact temporal correlation between the sensors is crucial for sensor fusion and trajectory prediction.

Deterministic Transport and Resilience over Ethernet

Once video streams are encapsulated as Ethernet traffic, they are transmitted within deterministic networks. Time-Sensitive Networking (TSN) mechanisms—in particular, traffic shaping and time-sensitive scheduling—enable bandwidth guarantees across all traffic classes. Prioritization using Virtual Local Area Networks (VLANs) also ensures predictable data transmission even under high network load.

Unlike point-to-point serialization connections, Ethernet switching enables the multicast transmission of a single video stream to multiple compute nodes without requiring the source to be duplicated. Driver assistance systems, infotainment processors, and data logger subsystems can subscribe to the stream simultaneously. Cameras with native Ethernet output can be connected to central switches or zone switches via short cables. In this architecture, stream distribution is a configurable network parameter rather than a fixed physical connection.

Redundancy mechanisms further enhance reliability. The IEEE 802.1CB standard supports the redundant transmission of frames across different paths, followed by the elimination of duplicates. If a connection or a switch segment fails, the receiving node reconstructs the stream without interruption. This allows functional safety goals to be achieved without the need for parallel proprietary cabling.

Programmability, Control, and Ecosystem Integration

Ethernet camera streams are controlled and configured via IEEE 1722.1. Processors detect camera bridges, establish stream connections, and send commands such as changes to exposure or operating mode. General-purpose input/output (GPIO) events can be stamped with the network clock’s timestamps, enabling synchronized triggering across multiple sensors.

This architecture transforms cameras into programmable network endpoints. The software can reassign streams, add additional subscribers, and adjust bandwidth allocations without the need for rewiring. A camera is added by allocating bandwidth within the switch fabric, rather than redesigning the serialization chains. As capabilities evolve, the network can accommodate new sensor traffic without requiring structural changes.

Because the cameras operate within the Ethernet framework, they use standardized mechanisms for security, monitoring, and power supply. IEEE 802.1AE MACsec secures data during transmission. OAM (Operations, Administration, and Maintenance) functions provide transparency at the link layer, and Power over Data Line (PoDL) or Power over Coax (PoC) supply power to the end nodes. Established programs for Ethernet conformance testing, electromagnetic compatibility (EMC) testing, and manufacturer-verified interoperability also reduce reliance on proprietary validation processes.

End-to-End Ethernet with Native High-Speed Ethernet in SoCs

SoCs for the automotive industry are increasingly integrating native high-speed Ethernet interfaces that operate at 10 Gbit/s and are scalable up to 25 Gbit/s. This enables end-to-end Ethernet connectivity within the vehicle network. These interfaces reduce the reliance on external deserializers while also reducing the number of interface pins required on the SoC.

Integrated hardware engines decompress IEEE 1722 streams and transfer video directly to processors via Direct Memory Access (DMA) for image processing or graphics acceleration. Offloading these functions from the CPU reduces latency and frees up computational resources for perception algorithms.

High-speed Ethernet connections bundle bandwidth into fewer channels than parallel camera interfaces such as the Mobile Industry Processor Interface Camera Serial Interface-2 (MIPI CSI-2). A single 25 Gbit/s Ethernet interface can bundle multiple camera or radar streams, thereby reducing the number of pins, simplifying package routing, and lowering the complexity of the printed circuit board (PCB).

Because Ethernet supports switching and multicast functions, a single Ethernet switch can distribute a camera stream to multiple SoCs. Streams are added or redistributed via software configuration within the network fabric rather than by reconfiguring the hardware. The SoC thus functions as a time-synchronized, high-throughput node within a scalable communications infrastructure.

Optimization of Camera Connections with Asymmetric PHYs According to IEEE 802.3dm

Although the symmetric MultiGig PHYs defined in IEEE 802.3ch enable high-bandwidth data transmission, camera data traffic is fundamentally asymmetric. The video stream flows primarily from the sensor to the processor, while data traffic in the opposite direction consists mainly of control and status information.

IEEE 802.3dm addresses this imbalance by defining asymmetric single-pair Ethernet PHYs for automotive applications that provide high downstream data rates for uncompressed video while supporting low upstream bandwidth for control data. The design targets specify downstream rates of 2.5, 5, and 10 Gbit/s, as well as upstream rates of approximately 100 Mbit/s.

By adapting the link capacity to actual traffic patterns, it is possible to reduce chip area, lower power consumption, and alleviate thermal constraints in compact camera modules. Smaller PHY implementations enable more efficient sensor integration while maintaining deterministic transmission characteristics.

IEEE 802.3dm builds on existing automotive Ethernet frameworks and maintains Layer 2 compatibility with TSN. The standard is designed for cable lengths of approximately 15 meters and supports both shielded twisted-pair and coaxial media. Optional features such as PoDL and PoC remain available for edge devices as well.

From Isolated Connections to a Programmable Vehicle Network

Replacing proprietary serialization chains with a unified Ethernet fabric enables deterministic transport, precise synchronization, configurable redundancy, and standardized control along the entire communication path. The native integration of high-speed Ethernet into SoCs, combined with new asymmetric PHY specifications, improves efficiency at both the architectural and physical levels. These developments support scalable bandwidths, software-driven reconfiguration, and adaptive lifecycle management across the entire vehicle platform. Because SDV architectures consolidate computing power into zonal and central domains, environmental sensing functions are increasingly being shifted or adapted through software updates rather than by redesigning the hardware. A standards-based Ethernet fabric makes it possible to reassign, duplicate, prioritize, and secure video streams through configuration changes within the network infrastructure. This flexibility reduces structural dependencies between sensors and processors and supports long-term functional evolution. Instead of tying perception data to fixed serialization chains, Ethernet functions as a programmable communication infrastructure that flexibly supports new algorithms, updated security requirements, and increasing data rates. 

*Amir Bar-Nivi is Vice President of Automotive Ethernet Strategic Marketing at Infineon Technologies

IP Down to the Field Level

Single Pair Ethernet in System Integration