Ethernet was invented at the legendary Xerox PARC (Palo Alto Research Center) by Robert Metcalfe and David Boggs, based on earlier ideas from ALOHAnet developed at the University of Hawaii (memo to my boss: I think I need to make an onsite visit to check this out). It was designed to connect up all the Alto computers, which were the personal computers that pioneered everything we have come to expect such as a mouse, windows, popup menus, and more. In fact, these were the computers that Steve Jobs and others at Apple saw that eventually led to the Mac, and to Microsoft Windows. The original Ethernet ran at 3Mbps. A standard, known as the blue book due to its blue covers, was created in 1980 driven by Digital Equipment Corporation, Intel, and Xerox. This was for a 10Mbps version, which was about the limit of semiconductor technology of that era. It was also known as the DIX Ethernet from the three company's initials, and eventually became IEEE 802.3 standard. Both the original and the DIX Ethernet ran on coaxial cable. The data rates have increased every few years and new standards have been created, and it has conquered the world as the way computers are connected. All those colorful patch cables are carrying Ethernet packets. The coaxial cables went and Ethernet ran over standard telephony twisted pair cables, and fiber optic for the highest data rates. But one place Ethernet has not conquered is automotive. Until now. There are a lot of different standards for networks in cars. LIN, Flexray, MOST, FPD-Link LVDS. But the most common is CAN bus. This has a datarate of 1Mbps, and there is a faster version called CAN FD with a data-rate of 2.5Mbps (the FD stands for flexible data-rate since it is compatible with devices that only support the slower standard, too). CAN is ubiquitous and is in every vehicle. However, cars are changing from mechanical horses to smartphones on wheels. A modern car contains 70-100 electronic control units (ECUs), all intercommunicating. Advanced driver assist systems (ADAS) are the stepping stone betweeen where most cars are today, to fully autonomous vehicles in the future. But ADAS requires cameras and radar and thus has video performance-level requirements for the in-car network. CAN, and even CAN FD, are not good enough. The in-car network environment is very fragmented even at the physical level with CAN bus for slow data, Flexray over optical fiber, MOST over coax. Another big challenge is that wiring harnesses in cars are getting very large, over 100lbs (50kg) with obvious disadvantages to cost and fuel efficiency. For a number of reasons it is looking like there is going to be one winner: Ethernet. One reason is that Ethernet has a rich ecosystem already, including software stacks, test equipment, security, power standards, and more. Networks in a car have some differences from other environments. One is that there is no requirement for building-sized distances, cables can be limited to 15m. Good EMI is important. They also have strict reliability and lifetime requirements. Automotive electronics needs to operate over a huge temperature range for 20 years or more. Perhaps, the biggest issue, though, is safety. The buzzword here is ISO 26262, the title of which says it all: Road Vehicles: Functional Safety . Even before the 26262 standard was published in 2011, safety was paramount and this impacts the design of networks. You do not want your ABS brakes to fail because you were turning up the volume on the radio. New PHY specifications are being defined by the IEEE for automotive (802.3bw for 100Mbps and 802.3bp for 1Gbps) with a 15m distance, low-cost single balanced twisted pair, full duplex, PAM3 encoding, over 100BASE-T1 (1000BASE-T1 for 1Gbps). Another area of importance where there has been a lot of standardization effort is the VLAN support capability that has seen industry-wide usage for several years. This is one feature that ensures segregation of real-time and non-real-time data, to ensure functional safety and security (so that ABS data is segregated from turning up the volume on your radio, for example). There are also upcoming time-sensitive network (TSN) standards developed specifically with real-time functional safety automotive applications in mind, including time-aware transfer, packet pre-emption, and redundancy forwarding. It looks likely that all the networks in the car will be replaced with Ethernet, except keeping CAN for engine control only. However, for now it is being overlaid over the existing networks in the ADAS and infotainment areas, to add 360° panoramic view, backup cameras (which will be required in the US by 2018), blindspot detection, and so on. Automotive moves fairly slowly since they have a very long design cycle. It used to be four years but with better tools that is being driven down. Adoption seems to be fastest in Europe, starting with the BMW 7 series, which has about ten Ethernet devices, followed by Mercedes, Audi, Jaguar, and more. The whole automotive electronics market is seeing new companies since the knowledge required is not the traditional expertise that the so-called Tier-1s such as Delphi, Bosch, and Denso have in-house. Bosch just acquired a team from ST to do vision ADAS products, Delphi just acquired Ottomatika (basically the CMU team , see Ten Years Ago Self-driving Cars Couldn't Go Ten Miles ), and it seems safe to guess that there will be more such acquisitions since the Tier-1s have to acquire existing teams, they don't really have the capability or the time to build it up in-house. It turns out that some of the companies that eventually withdrew from the mobile chip market moved into automotive to leverage their technology. Freescale, NXP, and Texas Instruments, for example, are strong players. It is not as large a market as mobile but still attractive. There are only 100M vehicles produced a year (compared to billions of smartphones) but every vehicle contains a lot of different systems. Cadence has an automotive Ethernet MAC, available as IP for integration into ECU SoCs, that supports both the current 100Mbps and the 1Gbps datarates, along with a selection of PHYs: Supports IEEE 802.1Qav quality of service (QoS) requirements, with multiple hardware priority queues and credit-based traffic shaping Supports IEEE 1588/802.1AS time sync with one-step or two-step clock adjustment and high-resolution time-stamping Power is kept low with energy-efficient Ethernet state control and wake-on-LAN packet detection For the safety and reliability area, there is FCS generation and checking, overrun/underrun error detection, and maskable interrupt on error conditions. Full details on Cadence Automotive Ethernet solutions .![]()
![]()
↧