How Two Low-Power Versions of Automotive Ethernet are Cutting CO2 Emissions
Road vehicles are responsible for more than 17% of global CO2 emissions. Two types of automotive Ethernet designs may trim that percentage down.
Road vehicles are responsible for a considerable amount of global CO2 emissions—about 17% in 2013. Mandatory goals are set for car manufacturers to reduce these harmful emissions.
To lower carbon emissions (as well as the costs associated with fuel consumption), we have to minimize the power consumption of different electronic components of a car, including the in-vehicle network (INV). That’s why low power consumption can be a differentiating feature when choosing a networking technology for automotive applications.
In this article, we’ll briefly discuss two low-power versions of Ethernet. We’ll also look at Microchip’s LAN8770—a recently-released Ethernet PHY that, according to Microchip, features a sleep current four times lower than the sleep current of other products on the market.
The Advantages of Automotive Ethernet
Automotive electrical systems continue to get more and more complicated. A modern car can have more than 100 electronic control units (ECUs).
To support fully autonomous driving, we’ll probably need up to twenty radars, six cameras, and vehicle-to-everything (V2X) communications. The daily data generated and processed by such a system can be as large as 20 TB. A high-bandwidth network is required to transfer all this information in a timely manner. The demand for high bandwidth has made Ethernet an appealing solution for modern INV applications.
Another advantage of Ethernet is that it can reduce the weight of the required wiring harness and, consequently, the fuel consumption. The wiring harness is usually the third heaviest and highest cost component in a car (behind the chassis and engine).
This is due to the fact that several different communication standards, such as CAN, FlexRay, MOST, and LVDS, are usually employed in a car and each electronic component can have different wiring and communication requirements. A typical wiring harness is shown below:
Typical wiring harness. Image used courtesy of Ixia
The high bandwidth of Ethernet allows us to have communication from different components that are physically close to each other on the same Ethernet network.
The 100BASE-T1 automotive Ethernet, which is specified in the IEEE 802.3bw standard, offers a communication speed of 100 Mbps over an unshielded single twisted-pair (UTP) cable. A unified communication standard based on UTP cables can significantly reduce the weight and cost of the required wiring harness.
The high bandwidth of Ethernet can make autonomous driving and ADAS a reality. However, the power consumption of this technology, which was originally developed as a networking solution for local area networks, is not optimized for automotive applications.
Below, we’ll take a look at two low-power versions of Ethernet: 1) energy-efficient Ethernet and 2) a version based on the OPEN Alliance sleep standard.
The IEEE 802.3az specifies a low-power version of Ethernet called energy-efficient Ethernet (EEE) that attempts to save power on inactive Ethernet links.
As long as a node has data packets for transmission, it remains active burning full power. However, when there is no more data to transmit, the node transmits idle packets over the link. Observing these idle packets, the nodes connected to the link can agree to go into a low-power idle (LPI) mode where the unit consumes only about 10% of its full power.
In this mode, the link remains quiet for Tq seconds and goes through a refresh cycle for Tr seconds as shown in the following figure.
The approximate transitions times and power consumption of various energy-efficient Ethernet states. Image used courtesy of Linköping University
Typical numbers for these time durations are Tq =20 ms and Tr=200 µs. During the refresh cycles, the link is used to share some synchronization data between the devices. This keeps the sending and receiving nodes aligned and allows us to reactivate the link quickly when required.
As shown in the diagram, the transition from active to LPI and from LPI to active mode requires Ts and Tw seconds, respectively.
The EEE can lead to great energy savings for Ethernet networks with low utilization. However, as researchers at Linköping University point out, the transition intervals (Ts and Tw) increases the latency, which can be a serious challenge in real-time applications.
The left graph depicts link utilization vs. power use in EEE and no EEE while the right graph shows link utilization vs. packet delay. Image (modified) used courtesy of IEEE Communications Magazine
Global Wake-Up Through the Ethernet Network
The OPEN Alliance has developed a wake-up mechanism where a wake-up signal can distribute throughout the network and activate all ECUs that were previously in the sleep mode. In this model, a wake-up (WUP) signal is triggered by the microcontroller. This wake-up signal spreads very quickly to all ECUs in the network.
After a global wake-up, the units that are not needed go back to the sleep mode as quickly as possible. The advantage of a global wake-up rather than a selective wake-up is that the knowledge of what device is going to turn on remains in the relevant layer of the communication stack.
Note that the wake-up signal is distributed through the Ethernet link itself rather than a separate wake-up line. Hence, we need to keep a portion of the PHYs always active to detect any imminent wake-up signals.
Microchip’s Single-Port Ethernet PHY
Microchip's new single-port 100BASE-T1 Ethernet PHY, the LAN8770, demonstrates the efficiency possible in global wake-ups through the Ethernet network; specifically, this device supports the OPEN Alliance TC10 sleep standard. It has a sleep current of less than 15 µA that, according to Microchip, is about four times lower than the sleep current of other products on the market.
LAN8770. Image (modified) used courtesy of Microchip
It is designed to meet the automotive EMI requirements and is Grade 1 (-40°C to +125°C) automotive AEC-Q100 qualified. The supported MAC communication interfaces include MII/RMII (LAN8770M) and MII/RMII/RGMII (LAN8770R) interfaces.
The device includes an optional 125 MHz or 50 MHz reference clock output that can be used for RGMII and RMII interfaces. The LAN8770M and LAN8770R are available in 5 mm by 5 mm and 6 mm by 6 mm QFN packages, respectively.
The new Ethernet PHY is suitable for infotainment head unit, telematics, and ADAS applications.
Do you have experience designing with one or both low-power versions of Ethernet? Share your experiences in the comments below.