We will start seeing automotive Ethernet replacing CAN in eight to 10 years? I am not so optimistic. There is not yet an automotive Ethernet standardized, there is still considerable resistence against Ethernet from OEMs and suppliers, and it will take much time and cost much money to develop new ECUs with Ethernet support. But I am convinced, that sooner or later any kind of automotive Ethernet will replace CAN and the other buses of a car and will make the development costs in the end cheaper.
Currently, there are two efforts underway within the IEEE 802.3 Working Group to standardize Ethernet technologies. The first project is the IEEE 802.3bp Reduced Twisted Pair Gigabit Ethernet project. The second effort was just started, and is referred to as Power over Data Line. So standardized Ethernet is coming to the vehicular market!
@Junko: I suggest you take a look at the history of the introduction of CAN as a reference. It was developed by Bosch a long time before any automobile manufacturer actually considered making use of it, and even then it wasn't until it had gone through revisions and standardization. It even includes a side-trip with GM trying to alter it into a single-wire version.
In my opinion , there is no need to phase out CAN in favor of Ethernet. CAN can and should exist and handle the control related electronics as an isolated network, whereas the Ethernet should handle all the add-on systems -built-in as well as user added .
From the Car's safety point of view such kind of isolation will keep the car systems safe from the hackers who could get entry through the Ethernet ports.
I looked on the Broadcom website and the BroadR reach PHY for automotive is listed as a 100 Mb/s device. The standard being developed in the IEEE is targeting Gigabit Ethernet. Obviously rate of operation can have an impact on what applications would be supported. So it is very easy to envision both being used in the future, just a matter of how far into the future one is looking.
Also, i can't comment on what automotive manufacturers will or won't do. Instead i would suggest looking at the IEEE 802.3bp website (URL - http://www.ieee802.org/3/bp/index.html) for overview of industry support.
Junko, what "security issues the Ethernet community has learned over the decades"? Security issues are handled at higher layers than "Ethernet," for example, by the IP stack riding on top of Ethernet. And those same security solutions can be applied elsewhere too, doesn't have to be just over Ethernet. Ethernet per se, is very simple and straightforward. Remember what Robert Metcalfe said when EE Times recently interviewed him on this subject? Ethernet is successful specifically because it does not attemp to solve problems that higher layer protocols address.
In time, more and more controls are migrating to Ethernet, whether the press reports on this or not. Sometimes, or rather often, hybrid systems where Ethernet ties together other more specialized networks, through media gateways. And absolutely yes, as we have discussed in similar articles, isoltating different networks is definitely important for security. Sometimes the isolation is total, i.e. physical separation, and other times it can be through intervening firewalls.
I've little doubt that Ethernet will start being installed in cars, and I've even less doubt that when this happens, no one will even notice. Unless they're geeks and into these things. Just like CAN buses started being installed, without making a huge fuss.
@Bert, I guess I didn't explain it well in my earlier comment. What I meant was that just because moving from a proprietary bus to Ethernet doesn't mean there will be more security risks. After all, Ethernet has been used in so many data centers running mission critical applications. Whatever strong security people in the networking community have learned can be leveraged in the automotive industry -- that's what I meant, as it was explained to me by the Broadcom executive during the interivew.
Yes, I totally agree with that, Junko. For that matter, you can even have separate Ethernet networks in cars, say one for critical controls and one for infotainment, which are physically separate. I'm wouldn't be surprised if the CAN bus is gradually replaced by Ethernet, for example, one client system at a time.
Single-pair Automotive Ethernet, which uses un-shielded twisted pair (UTP) cable to deliver data at a rate of 100Mbps, is a relatively new Ethernet technology designed to meet automotive requirements for reliability, electromagnetic emissions and minimal power consumption. A single-pair unshielded cable along with smaller and more compact connectors can reduce connectivity cost up to 80 percent and cabling weight up to 30 percent.
At least such isolation will prevent the car hackers from taking control of the Car . A CAN network will control the vital systems of the Car such as ECU, braking etc and these systems can remain out of reach for the hackers.
I think low-voltage digital signaling (LVDS), media-oriented systems transport (MOST), and the controller area network (CAN) operate at different levels of the protocol stack so they really have nothing to do with each other...LVDS is a PHY standard and MOST sounds like a MAC or Protocol layer for example...nothing prevents MOST running over LVDS, the same way as TCP/IP rides over Ethernet etc...Kris
Good catch @rbv...D in LVDS stands for Differential...LVDS is very popular PHY signaling physicallayer protocol (essentially a paif of high speed differential signals with a typical common voltage of 1V or so and small wing of 200mV or so)
I guess I'm not sure why people think there's a pressing need to have wired ethernet replace working technologies in cars. The idea that consumers are bringing more and more devices into their cars, all cars, not just luxury models, is fine, but how many of those devices use wired ethernet?
Why would you need the wired comm's in a car (which often need to be bullet-proof for safety) to be ethernet? It's unnecessary and could be dangerous.
Now the need or desire for Bluetooth, or even Wi-fi built in to some car utility processor if you're going to argue that you're going to want data from your higher speed cameras brought into a portable device, or even a USB port is one matter, and it makes a lot of sense, but replace the working, tried and true, and perfectly sufficient existing infrastructure with a wired ethernet and all it's issues? I have to ask why?
My tablet doesn't have a wired ethernet connector... nor does my phone... and they don't need one... my laptop doesn't even have an native ethernet port anymore and needs an adapter to go to wired ethernet now. You're going to give me a hardwired ethernet plug in my car in 8 to 10 years? For what exactly?
Al: This is not about the devices people are brining into cars, but about the devices being built into cars. The cars, in essence, are evolving into complex machines that require a more sophisticated network.
Now, as exciting as I think that is, I think it's fair to ask another question. There is likely an evolving and strong market for this, but isn't there also a growing market for a simpler car? Wouldn't a car be a lot more affordable if it didn't have all this technology built in? I know this is EETimes and this is a bit heretical, but, candidly, I'd like to know how much technology is adding to the cost of cars, making them unaffordable for many people.
@amagnani, your points are well taken. Actually such efforts as brining in Wi-Fi to wirelessly connect consumer devices (brought into a car) with a backseat display, for example, are already happening.
Obviously, Ethernet won't replace all the tried and true networking technologies inside a car overnight.
But carmakers do have immediate needs for a bigger bandwidth networking technology, for example, to accomodate more than several cameras already installed inside a car for driver assistance. These embedded vision cameras will be streaming video in parallel in real time. LVDS won't cut it.
When you look at designs of future cars, you need to consider networking technologies that can scale. That's I think where Ethernet comes in.
The comments to this article thus far point towards two areas which are important: the physical layer (PHY, cables, etc.) and the higher level protocol (software) layers that ride on top of Ethernet. The part in the middle ("Ethernet") is not being affected.
The automotive industry identified several challenges that need to be overcome to make the whole system be suitable for automotive applications. On the physical layer are discussions ongoing on the required bandwidth and speed, resulting in discussions about both 100 Mbit/s and 1 Gbit/s. And which existing or new PHY solutions could support these communication speeds and meet the automotive EMC, power consumption and cost requirements at the same time. OPEN SIG, IEEE RTPGE and JASPAR are just three examples of consortia that focus on this aspect. For the higher level software is also discussion ongoing about topics like AVB, TSN (Time Sensitive Networks) and profiles. Involved consortia are for example AVnu, AUTOSAR, JASPAR.
All in all enough hurdles that need to be overcome to justify the careful remarks of some of the people who were interviewed in the article.
There are a couple questions we need to ask ourselves prior to putting Ethernet in a car.
1) How many devices needs to be connected?
2) Do the devices need to talk to each other?
3) Do the existing techologies serve the purpose?
4) How is the wiring done?
5) Where is the Ethernet Switch going to be?
There is an advantage of direct connectivity - 1 fewer device; 1 less uncertainity. Who wants to see a blinking dashboard when the Ethernet switch goes down?
I agree Infotainment is one of the key drivers to the wired automobile. Nowaday, kids are enjoying their cartoon shows and browsing their social network sites in the backseat with their connected tablet. Does Ethernet wired automobile add value to the Infotainment push?
On the other hands, if Ethernet wired automobile is demanded, I believe BroadR-Reach is a superior technology to the job. I learned about the technology in a couple years ago that it can serve with 2 wires (twisted or parallel). The fact that it has high noise rejection ability will definitely fit the environment of any vehicles.
While you are at it you might want to see what technology they are planning on using. Switched Ethernet is pretty much all that is being used today, but shared backbone media might significantly simplify the wiring. At a minimum I could see multiple localized switches. You are not going to want 100 ports on a single switch!
The obligation of auto manufacturers to maintain spares and repair capabilities for vehicles during their likely lifetime gets much complex with yearly changes in networking systems. By the time security upgrades are included (the dozens that Microsoft feels compelled to provide to my computer), the configuration control for a car will become an absolute nightmare. I think it is time to standardize and simplify. 100 year old "Early American antique" cars grace our highways from time to time; will it be possible to maintain the current generation of cars in a century?
100 year old "Early American antique" cars grace our highways from time to time; will it be possible to maintain the current generation of cars in a century?
DrQuine, I haven't even thought about it before...but it's true. With all these complexity (and not to mention, "islands" of different proprietary nteworking technologies proliferated in cars over time), we may never see cars today being trotted out on highways 100 years from now!
Blog Make a Frequency Plan Tom Burke 17 comments When designing a printed circuit board, you should develop a frequency plan, something that can be easily overlooked. A frequency plan should be one of your first steps ...