Today’s advances toward connected and autonomous cars bring challenges that must be addressed by the vehicle infrastructure. Infotainment, ADAS, sensor fusion, and the autonomous car all call for in-vehicle networking and ultra-high-bandwidth connections. But traditional automotive networks such as CAN, CAN-FD, LIN, MOST, and FlexRay don’t provide the necessary bandwidth to support all of the devices and applications of connected—and eventually autonomous—cars.
The many sensors and electronic control units (ECUs) that must be connected, along with the demands for ever-higher throughput, require high-bandwidth connections and scalability to allow for more flexible architectures. Automotive Ethernet, one of the options in the market today, was introduced in 2008. 100Base-TX-Ethernet was deployed into the automotive sector to connect on-board diagnostics (OBD) interfaces, enabling a unified diagnostics port running Diagnostics over Internet Protocol (DoIP) and a fast software download (with greatly improved data rates).
However, as the demands increase for higher bandwidth and more applications, Automotive Ethernet may not be the ultimate solution.
These are the three key challenges facing Automotive Ethernet:
- Multi-gigabit transmission of data within the car in the near term, while aiming for much higher numbers (over 100G) long-term.
- Electromagnetic-compatibility robustness and reliability (including “fail operational” modes) in the noisy automotive environment.
- Low-cost, low-weight wire harness to support all of the devices in the “data center on wheels” with the most optimized, cost-effective architecture.
Automotive 100Base-T1 and 1000Base-T1 Ethernet utilize IEEE Audio Video Bridging (AVB) for real-time streaming applications and time-sensitive networking (TSN) for deterministic behavior. Of course, AVB and TSN protocols add overhead (and latency) and require processing power. There’s a price to pay (development complexity and cost of solution) whenever one wants to sample and packetize audio or high-bandwidth video data and push this streaming data through software stacks.
100Base-T1 and even 1000Base-T1 Ethernet doesn’t bring enough bandwidth to connect the cameras and high-resolution graphic displays for autonomous driving, as bandwidth requirements in some cases are significant. Although higher-bandwidth Automotive Ethernet is in development, it will likely require more robust cabling to handle electromagnetic interference (EMI). EMI is a major issue as electronic systems become more dependent on safety-critical considerations. This is compounded by the car’s increasing bandwidth needs—faster data rates create higher risks of data errors caused by interference.
Automotive Ethernet relies on pulse-amplitude modulation (PAM) for its physical-layer (PHY) modulation, such as PAM-3 and eventually PAM-4. PAM is an analog pulse-modulation scheme in which the information being transferred is encoded in the amplitude of the signal pulses themselves. The number of possible amplitudes directly affects the number of data bits conveyed by each amplitude.
Automotive Ethernet relies on forward error correction (FEC) to ensure the validity of data being transmitted. However, adding the redundant data bits increases overhead to the link and reduces the effective bandwidth—a key consideration when budgeting the system throughput. Furthermore, FEC-based solutions can only respond to gradual noise profiles, making them unsuitable for handling instant noises from EMC events.
Another option to address the needs of high-speed in-vehicle connectivity is HDBaseT, a technology developed by Valens. HDBaseT Automotive handles these issues by utilizing PAM-16 modulation in conjunction with fast adaptive noise cancellation, which can be automatically activated only when needed (“just in time cancellation”). In addition, local packet retransmission is employed to handle any instant noise attacks. As such, the link operates error-free at the optimal operating point, with zero packet-loss.
HDBaseT Automotive can transmit native Ethernet data (100 Mb/1 GbE/2.5 GbE) over a single unshielded-twisted-pair (UTP) wire. And it can do so while providing exceptional EMC robustness. As the wire harness is unshielded, there are no grounding complexities associated, and there’s more flexibility regarding reuse of multi-pin legacy connectors (such as MQS), even for 1 GbE. As the chipsets transmit standard Ethernet RGMII/SGMII MAC PHY interface on both link ends, no changes occur for Ethernet-based systems above the physical layer, simplifying the overall architecture and lowering area and power needs.
Lastly, HDBaseT Automotive enables aggregation of additional interfaces (such as video, audio, PCIe, USB, etc.) over the same link, reducing the number of necessary wires in the car. For PCIe and other interfaces, HDBaseT Automotive already supports 16-Gb/s bandwidth over UTP, with additional interfaces in the roadmap, such as CSI-2.
The connected car is a reality. With the right technology, we can surpass the challenges of today’s connectivity.
Daniel Schwartzberg is Director of Technical Pre-sales at Valens.