Meeting the connectivity challenge in next-generation automobiles

This article originally appeared in the Oct'22 magazine issue of Electronic Specifier Design – see ES's Magazine Archives for more featured publications.

Connected, shared, autonomous and electrified are the major themes driving the automotive industry today. Enabling cars to make complex decisions requires huge volumes of data to be gathered from sensors positioned outside the vehicle to be processed or displayed inside. For decisions to be made and actioned in real-time will therefore require ultra-high-speed connectivity solutions.

In this article, we consider the potential of a range of legacy and current automotive networking technologies for use in this safety-critical application by examining their ability to meet some of the key technical requirements.

Legacy connectivity solutions

Advanced driver assistance systems (ADAS) work by simultaneously processing images from multiple cameras and combining it with other data provided by LiDAR, RADAR, and ultrasonic sensors.

This information must be transmitted to several electronic control units (ECU) distributed around the vehicle, over cables up to fifteen metres in length. Key requirements for transporting this data include:

Latency: As vehicles become more autonomous, latency and data errors become intolerable. Signals from multiple sensors and cameras are combined to create a real-time three-dimensional representation of the external environment. Delays and/or errors in the data stream from even a single sensor can make it difficult for an ECU to fully resolve the surrounding environment which could lead to potentially disastrous consequences for the vehicle, its occupants and those in its immediate proximity.

Reliability and robustness: The confined physical space within a vehicle makes electromagnetic interference (EMI) a real problem. As the number of electronic components increase, space becomes an issue and as components and cables are placed closer together, they become vulnerable to crosstalk.

Redundancy: Like aeroplanes, highly automated and autonomous vehicles require fail-safe systems that can automatically reroute data around points of failure to allow a vehicle to continue to function normally or come to a controlled stop.

Bandwidth: ADAS data must be transported at data rates in excess of ten gigabits per second.

Some automotive networking technologies have been around for decades and continue to be deployed in present-day vehicles because they operate simply, reliably, and cheaply. Typically deployed in low-speed control applications, these include Local Area Interconnect Network (LIN) which operates at tens of kilobits per second, Controller Area Network (CAN) at up to 1Mb/s and its successor CAN_FD (up to 12Mb/s).

FlexRay was adopted by some high-end automobile OEMs for safety-critical applications and supports data rates up to 10Mb/s. However, none of these come remotely close to providing the multi-gigabit bandwidth requirements of current ADAS applications thereby eliminating these legacy automotive networking solutions as potential candidates to meet this challenge.

Automotive Ethernet

Ethernet was developed in 1973 and standardised by the Institute of Electrical and Electronics Engineers in 1985 as IEEE802.3 before later becoming the near-universal Local Area Networking data communications protocol. Ethernet signals can be transmitted over coaxial, fibre optic, and unshielded twisted-pair cables, with speeds increasing from 10Mb/ sec initially, to more than 1000Gb/sec currently. As computing technology began to creep into automotive applications, the industry began to investigate the trusted Ethernet protocol as a data connectivity solution. In 2016, the IEEE published the first Automotive Ethernet standard, 100Base-T1 in IEE802.3bw. While there are similarities (both versions use unshielded twisted pair cables with two copper wires twisted together along the length of the cable to reduce electromagnetic radiation and crosstalk) there are some differences. 

100Base-TX uses two pairs of wires – one pair to carry the transmitted signal in one direction with the other pair carrying the received signal in the opposite direction. Automotive Ethernet uses only a single pair of wires (SPE) for transmission and reception, making the cable lighter and less expensive. The 100Base-TX standard is specified for a maximum cable length of 100 metres while automotive Ethernet is specified for a maximum of only fifteen metres, a distance more suited to the size and scale of an automobile.

Another difference is the encoding scheme used to reduce EMI and crosstalk. The 100Mb/s IEEE802.3bw version of the Ethernet standard has been widely adopted in automotive applications. However, this speed is not sufficient to simultaneously carry the data from multiple sensors and higher-definition camera streams to an ECU and onwards to a display. IEEE 802.3bp, or 1000Base-T1, allows for gigabit speeds over twisted-pair wiring but it operates at 600MHz which has the disadvantage of making cables more susceptible to crosstalk, further complicating the task of managing electromagnetic noise.

In 2020, the IEEE produced 802.3ch, which enables multigigabit Ethernet at standard rates of 2.5G bit/sec, 5G bit/sec and 10G bit/sec over cables up to fifteen metres in length. While future versions of Ethernet may have higher data rates, for now, automotive ethernet is an excellent replacement for the functions performed by legacy connectivity solutions, but it falls short of the bandwidth required by ADAS systems and high-resolution displays.

Serial links

Connecting a high-resolution camera to a display does not require a fully symmetrical data connection like Ethernet. Asymmetrical ‘SerDes’ systems, which use a serialiser IC on the transmitter side and a deserialiser IC at the receiver are now commonly used by automobile OEMs to carry high-speed video and sensor data. Early solutions included APIX III (Inova), GMSL (Maxim Integrated), and FPD III-Link (Texas Instruments) which provided data rates of up to 3Gbps over a single coaxial or differential cable. Second generations of this technology increased data rates of up to 6Gbps on one channel (or 12Gbps using two combined channels).

By comparison with automotive Ethernet, a SerDes system use an asymmetrical link, meaning that the data rates in one direction (the downstream link) are much higher than for the other (upstream). For video and sensor applications, this is sufficient as cameras are primarily a source of high-speed data, only receiving control signals at much lower rates. Display units also receive high-speed data but only occasionally send control data to the ECU e.g., when a finger is placed on a touchscreen. This asymmetric approach simplifies physical complexity and reduces channel requirements, allowing automotive OEMs to tailor a more cost-effective system than could be done using a full duplex, symmetrical Ethernet-based implementation at the same speed. 

To address the need for a single unified high-speed physical layer interface suitable for use with automotive serial links, members of the Mobile Industry Processor Interface (MIPII) alliance set about developing the MIPI Automotive SerDes Solution (MASS), culminating with the release in September 2020, of A-PHY v1.0, the first supplier-agnostic, high-speed, long-reach, physical layer SerDes solution for automotive applications and this was also adopted by the IEEE. This standard has a roadmap which will eventually achieve data rates up to 32Gbps, making it well placed to meet the increase in bandwidth requirements as more electronic systems are added to automobiles in the future.

Conclusion

As automobiles move towards full autonomy, the number of ADAS systems and the speeds at which they transport, and process data will increase dramatically. Legacy automotive networking solutions are far too slow to be considered viable solutions to meet the arising connectivity challenge. While automotive Ethernet is inching towards providing the required data rates, it just falls short in terms of the bandwidth required for use with high-resolution displays. For now, asymmetrical serial links offer the best solution for multi-gigabit data communication and the recent establishment of the APHY v1.0 industry standard for this technology offers a roadmap that future proofs its application.