Ethernet can officially be added to the list of automotive networks, such as CAN, LIN, FlexRay, and MOST. But with so many networks already available, why is Ethernet needed? And what specific automotive applications need it? The answers are simple, and they relate to why Ethernet now dominates office networking.
Simplicity and field-proven open standardization have significantly cut the cost of ownership in every application Ethernet has entered, and they are part of the reason that Ethernet has recently become the technology choice for both the factory and the home. Volumes of scale in the office and a voracious consumer market, supported by a magnitude of Ethernet vendors, have driven pricing levels far lower than any “custom” designed protocol can compete with.
Initial applications for automotive Ethernet now routinely include onboard diagnostics (OBD), which has rapidly reduced software download times. There seems to be a general consensus that OBD will move to an IP-based (Internet Protocol) interface, i.e., a physical layer (PHY) of Ethernet rather than the traditional slower CAN.
In the future, “real-time” Ethernet audio-video bridging (AVB) will also offer high-performance infotainment network solutions, but herein lies a challenge that isn’t purely technical. OBD applications have the benefit of being operational when the car is in the service garage and not on the move.
When the car is running, the car manufacturers demand more stringent electromagnetic interference (EMI) requirements. Ethernet was never designed for such applications. So, can this technology rise to the challenge?
Thermal And EMC Performance
The industrial control market has shown that Ethernet networks can deliver robust performance in extreme conditions. Extended temperature ranges, heavy vibrations, high electromagnetic compatibility (EMC) radiation, and dusty or wet surroundings are typical in many of these applications. Raising the ambient temperatures “under the bonnet” over the common 85°C won’t cause thermal issues for Ethernet devices.
For example, Micrel’s KSZ8041NL AM AECQ-100 single-port fast Ethernet PHY solution consumes 175 mW inside a thermally enhanced 5- by 5-mm micro lead-frame (MLF) package. The KSZ8041NL family also offers a military specification variant that supports ambient temperatures of up to 125°C.
Due to the demands of the industrial and automotive markets, many newer Ethernet devices offer improved electrostatic discharge (ESD) performance. This is a major shift in emphasis from original office applications where ESD rating isn’t a major concern.
Continue to next page
For instance, Micrel’s KSZ8041 PHY and KSZ8851 controller families have a human body model (HBM) ESD rating of greater than 6 kV. The evaluation board has also been shown to provide greater than 9-kV contact and greater than 16.5-kV air ESD ratings without the need for any external over voltage protection devices. This surpasses general automotive EMC requirements such as those required by BMW Group Standard GS 95002.
The stringent EMI requirements currently demanded by the car industry represent one of the biggest challenges for any automotive electrical device. As data speeds increase, signal edges become faster. This results in higher energy emissions. The first goal for the industry is agreement on what the emission limits should be for Ethernet technology as well as the bandwidth required.
For video and camera imaging transmission, whether 100-Mbit/s Fast Ethernet is actually sufficient or Gigabit Ethernet data rates are going to be needed is debatable. The choice will probably depend upon whether video compression is acceptable for camera applications.
Figure 1 shows the typical radiated emissions characteristics of an Ethernet board, using a Micrel KSZ9021 Gigabit PHY. Not surprisingly, the peak emissions can be found at the 125-Mbit/s harmonics of the reference clock, exceeding typical OEM limits. FlexRay is used as an example here.
For conformance today, a shielded cable (either twisted pair or coax) or plastic optical fibre (POF) can be used for Ethernet while the car is running. The standard Ethernet RJ45 connector and CAT5 cable have proved very robust and remain extremely popular in other applications, including industrial.
However, existing vendor specific connectors and wiring looms are likely to be adopted in automotive applications, at least initially. Moving forward, there is a drive to adopt a standardized IP diagnostics interface, as specified in ISO 13400. The Ethernet PHY (transceiver) is flexible enough to use such connectors and cabling without any significant degradation in performance. The table shows the typical characteristics of a CAT5 cable.
Standard CAN cable exhibits similar characteristics to unshielded, twisted-pair CAT5. Testing has proven the feasibility of long-term error-free transmission of Ethernet over in excess of 100-m CAN cable. The major difference between the two cables is that a CAN cable is only partially specified and does not provide controlled impedance or twist ratio.
As a consequence, EMC behaviour and signal integrity cannot be guaranteed, making CAN cable generally unsuitable for high-speed data transfer. A CAN cable is currently used for Ethernet OBD and flash updating, where the lines can be disabled during normal driving and can only be activated while in the repair shop or production plant.
Continue to next page
The Leoni cable suits high-speed data transfer such as low-voltage differential signaling (LVDS), USB, and Ethernet in automotive applications. It’s shielded with controlled 100-Ω impedance and qualified up to 1 Gbit/s for performance similar to CAT6 rather than CAT5. It isn’t actually twisted pair but a four-wire twist known as Stern-Vierer, translated as Star Four Wire.
The FlexRay cable from Kroschu, which is the first automotive unshielded cable with controlled impedance, provides better EMC and signal integrity compared to CAN. Standard FlexRay cable is only a single twisted-pair cable, although two-wire twisted-pair and Stern-Vierer versions are available. CAT5 provides four twisted pairs, in which all pairs are used for Gigabit Ethernet but only two pairs are needed for 100-Mbit/s Fast Ethernet.
To enhance reliability, cable diagnostics technology such as Micrel LinkMD goes beyond Ethernet-defined standards to provide a solution to such problems. LinkMD cable diagnostics uses time domain reflectometry (TDR) to analyze the twisted pair for common cable problems, such as open circuits, short circuits, and impedance mismatches.
An alternative to traditional copper CAT5 cabling comes in the form of POF. Car manufacturers are already very familiar with this physical media, as it is deployed in Media Oriented Systems Transport (MOST) networks. The same 1-mm LED POF technology from MOST (including new MOST-150) can also be used for 100-Mbit/s Fast Ethernet transmission with a reach of 100 m. POF is extremely robust and lightweight. And like all fibre, it’s immune to electromagnetic noise, as it emits no radiation. Figure 2 shows an example of an Ethernet-to-POF interface circuit.
However, the drawback of using a shielded cable or POF lies in the cost. Investigations are ongoing, exploring ways of meeting radiated emission levels using unshielded cable, at least for Fast Ethernet. Some proposed approaches involve using a different modulation technique, resulting in a proprietary non-conformant “Ethernet.”
This goes against the grain of Ethernet’s appeal and success, which is its field-proven, interoperable open standardization and low cost. Such an approach not only would need to be made open and freely accessible, it also would take up of a number of silicon vendors. If this can or will happen in the near term remains to be seen.
The ideal solution is to enhance standard Ethernet PHY devices to reduce EMI emissions. These improvements, coupled with other board and device techniques, provide a realistic chance of success, certainly for Fast Ethernet.
Mike Jones is a senior field applications engineer with Micrel Inc.