
Can Automotive Ethernet contribute to the vehicle weight-loss plan?
Today, high-strength aluminium alloy is one of the alternatives employed but merely focusing on reducing the weight of the existing cable harness is only kicking this problem further down the road. The harsh reality is that a lot of the data being transferred around the vehicle could be transported over a simple high-speed twisted-pair connection. Features such as reverse parking cameras often require dedicated cabling for a point-to-point connection to the appropriate electronics control unit (ECU). However, such point-to-point solutions merely serve to enable a desirable consumer feature in a manner that meets a suitable price point, rather than delivering a technology solution that could be expanded to support a wide range of in-vehicle data transport needs.
Is Ethernet the answer?
Ethernet as a technology has been in consideration as an alternative for a long time. Its ubiquity is such that one questions why Ethernet has not been integrated into the vehicle, rather than why it should be. It is well understood, it is integrated into a wide range of microcontrollers (MCU) and system-on-chip (SoC) devices, and there is ample software available, both commercial and open source. Furthermore, the engineering community already has ample understanding with regard to its implementation and fulfilling regulatory compliance.
As more and more data are generated in the vehicle to support advanced driver-assistance systems (ADAS) and autonomous driving, Ethernet seems to be the obvious choice for transporting data from radar, LiDAR and the multitude of cameras such systems require. Certainly, the bandwidth requirement is fulfilled. Unfortunately, classic Ethernet falls down when it comes to supporting applications with time-critical or safety-critical needs, since there are no mechanisms for time-sensitive networking, traffic shaping, or time synchronisation across the network. To support such requirements, changes in the lower layers of the OSI model need to be made.
Further issues arise when reviewing the cabling and signalling. Ethernet as implemented in homes and commercial buildings, if using CAT5e cabling, uses two of four pairs of wires for uni-directional data transfer inside an unshielded cable. This will obviously not improve the weight of the cable harness. If the existing technology were simply reduced to a single twisted-pair without shielding, the resultant solution would not fulfil existing electromagnetic interference (EMI) requirements.
Automotive Ethernet, AVB and TSN
To make Ethernet automotive-ready, several working groups have been developing standards to meet the automotive industry’s needs. At the physical level, a 100 Mbits/s full-duplex physical interface has been defined that can operate across unshielded twisted-pair cabling. Named 100BASE-T1, differentiating it from 100BASE-T, it is described in the IEEE 802.3bw standard. It makes use of PAM-3, 3-level signalling with a data rate of 66.67 Msymbols/s and can reach lengths of 15 m, or 40 m if shielded. A 1000 Mbits/s interface, known as 1000BASE-T1, is covered by IEEE 802.3bp. The result is two speed grades enabling an ultra-high-speed data backbone between key ECUs, and a more cost-effective, but still high-speed, interface to end nodes (Figure 1).
For simpler functions and comfort options, such as window openers, ambient lighting, and motorised seats, traditional in-vehicle networks, such as LIN, CAN and CAN-FD, will continue to play a vital role beyond the 100BASE-T1 enabled end nodes.
When it comes to transporting audio or video data for in-vehicle entertainment, it is essential to be able to operate within defined latencies and reserve bandwidth across the network. The Audio Video Bridging (AVB) working group developed a set of standards that deliver these functions. Many of these changes impact layer 2. IEEE 802.1Qav is responsible for defining the rules that ensure Audio/Video (AV) Bridges allow AV streams to pass through the network within a defined time constraint. It handles traffic shaping ensuring that the network is not overwhelmed by bursts of traffic. IEEE 802.1Qat guarantees the end-to-end resource necessary to support the transfer of a data stream and provide Quality of Service (QoS).
Further improvements are provided that ensures time synchronisation between network nodes (IEEE 802.1AS) and procedures for ensuring that multiple node present their data at the same moment in time (IEEE 1722). This is important in an application where a head unit is distributing audio data to be output at two or more loudspeaker nodes.
Although the AVB standards cover bandwidth reservation and fixed latency, there are other use cases where even shorter latencies are required. This is the case when closed-loop control is being implemented over the Ethernet interface. The Time Sensitive Networking (TSN) standards address these issues. These include IEEE 802.1Qbv-2015 that provides defined time windows to ensure end-to-end latencies by blocking low-priority traffic.
Finally, with consideration for the all-electric powertrains of the future, the Energy Efficient Ethernet (EEE) standard IEEE 802.3az2010 provides mechanisms for putting nodes into standby, leaving receive circuitry in a mode that enables reception of a wake-up message.
Automotive Ethernet as a peripheral, or as a stand-alone solution
Automotive Ethernet will find itself in one of three clear application spaces; the large central ECU on the central data backbone, in a domain controller, or in an end node application. In the first two cases, Automotive Ethernet is ideally implemented using a suitable peripheral device to complement the selected high-performance SoC. The third case is typically the domain of a smaller 32-bit MCU that can bridge the balance between cost and sufficient performance.
With the introduction of the TC9562 there is a single-chip solution that can support both peripheral functionality and operation as a stand-alone Automotive Ethernet node (Figure 2). Large SoCs are supported via the PCI express (PCIe) interface and it includes L1 low-power mode support when required. 6 channels of DMA are also provided that can be used to automatically filter incoming data according to IP address and transfer them into the DRAM of the host SoC (Figure 3). The Ethernet MAC supports all common media-independent interfaces (MII) including serial gigabit (SGMII).
Alternatively, the device can also be used as a stand-alone Ethernet node solution by making use of the integrated Arm® Cortex®-M3 processor. Operating at up to 187.5 MHz, and with access to up to 320 kB of memory, this makes it ideal for implementing audio amplifiers, audio systems, and interfacing with cellular network modems (Figure 4).
In addition to common serial interfaces, such as the two UARTs, I2C and SPI, a QSPI peripheral provides an interface for booting the device. The TDM/I2S interface provides a full-duplex TDM port with support for multi stream operating as a clock master (TDM/I2S) or clock slave (TDM only) supporting 24-bit audio at up to 192 kHz.
Summary
After all the hard work of advisory boards and committees, Ethernet has been extended to fulfil the harsh demands of in-vehicle networking. With plenty of bandwidth, guaranteed latencies, and support for specifying presentation time, this veteran technology is again revitalised and ready for the emerging demands of ADAS and autonomous driving. The TC9562 tackles the challenges of in-vehicle networking by offering developers a customisable solution for simpler cost-optimised nodes that are more limited in their functionality, such as audio amplifiers, as well as a compact solution that fulfils the requirements of larger ECUs, requiring Automotive Ethernet support alongside the chosen SoC.
About the author:
Klaus Neuenhüskes is Senior Manager, Solution & Standardization in Semiconductor Marketing at Toshiba Electronics Europe. Klaus holds a degree in Electrical Engineering and previously held positions at OKI Electric Europe and NEC Electronics.
