BMW to launch new network protocol in next X5
The next-generation BMW X5 is said to implement the first application of Flexray in the automotive world, using the relatively new networking protocol for its electronically controlled dampers. This will come approximately two years before the manufacturer rolls out its first platform that uses Flexray as the main vehicle communications backbone. The CAN-C systems currently in place are limited to 500 kbs, which many manufacturers are finding isn't fast enough. Even more problematic, CAN isn't deterministic; that is, it cannot be assured that a message will be successfully transmitted at a particular point in time. There are ways of asserting priority on the network, but this involves playing favorites - not acceptable when it comes to deciding whether the braking or steering is more important, and while those two were arguing, your electromechanical valve system just crashed. Flexray solves both problems by increasing the data rate, and implements a low-cost means of establishing a network clock so that each module is given a discrete slice of time in which it can do its thing. There's a decent introduction to it here for those looking to learn more. Expect to hear a lot more about this in the next few years. In the meantime, those of us heavily invested in CAN development tools will hopefully get our money's worth.
_opt_thumbnail.jpg)
Reader Comments (Page 1 of 1)
Loudog 10:26PM (12/18/2005)
So let me get this straight: Flexray uses a time sliced access method? This thing is just an overclocked T1. Sheesh.
Reply
Loudog 10:26PM (12/18/2005)
So let me get this straight: Flexray uses a time sliced access method? This thing is just an overclocked T1. Sheesh.
Reply
distantbody 10:26PM (12/18/2005)
Am i the only one who doesnt get a word of this? particularly given that general automotive "news" should be understandable to a majority of people!
Reply
Peter W. 10:26PM (12/18/2005)
Here's what there is to get:
BMW, the company that brought us the iDrive, is now going to change how all of the complex computers inside the X5 (ABS, Traction control, emissions, ignition, valves...) are going to talk to each other.
Anyone who got one of those first-gen-iDrive 7-series will tell you to call up your lemon lawyer and start the paperwork about the same time you sign up for a new X5.
Reply
Rando 10:26PM (12/18/2005)
This isn't really anything to panic about... "retooling for US TV" basically means they have to add in the US Dollar amounts on the video, and re-record the in-studio bits. That takes some time, so that explains the lag between the UK and US versions.
But all the interesting bits are still the same (the driving, and the in-car videos), so there's nothing to get too excited about, as long as it does air here in the US. :)
Reply
auto_dev_engineer 10:26PM (12/18/2005)
> Even more problematic, CAN isnt deterministic;
>that is, it cannot be assured that a message
>will be successfully transmitted at a
>particular point in time.
At a particular microsecond? Why is this necessary? If you were to purchase many current production GM vehicles, or just some recently produced Ford vehicles with the J1850 bus, they're dealing with messages transmitted every SECOND or so.
>There are ways of asserting priority on
>the network, but this involves playing
>favorites - not acceptable when it comes
>to deciding whether the braking or steering
>is more important, and while those two were
>arguing, your electromechanical valve system
>just crashed
What in the world are you talking about? The priroity scheme is built into the CAN message ID. That is, the LOWER the message ID, the higher the priority. I'm assuming that by "electromechanical valve system", you're talking about the ABS valve block. In the MICRO seconds that it takes to arbitrate priority on the CAN bus, why are mechanical valves going to "crash"?
500k is NOT the limit of CAN -- both DCX and Volvo are using higher speed variants.
Perhaps BMW has some new technology that they're going to launch with the new X5 -- but I don't think you'll see an immediate change to a new bus architecture. If you look at the diagram in the article, they're just implementing several Flexray/CAN gateways, which hardly appears to be a "stable" solution to me. The Big 3 have only recently started making the jump to CAN! Are all module suppliers going to immediately make these huge hardware and software changes? Not likely.
This article submission is very, VERY misleading. It would've been better by just indicating that there's a new bus architecture out there, it's faster than CAN and has new features... Not by trying to create bogus scenarios like module "arguments" and subsequent hardware failures.
See: http://www.can.bosch.com/ for more accurate information.
Reply
Justin Hollabaugh 10:26PM (12/18/2005)
I for one thank Autoblog for bringing us this kind of news.
I am helping my boss inject a Jaguar V-12 for his XKE with a MegaSquirt/MegaSpark controller. The next generation of MegaSpark controllers are supposed to be CAN enabled so you can create an array of them to drive whatever you need to drive (equential injection, turbo retard, nitrous activation and retard, fuel pumps, anything). I had a suspicion that CAN bandwidth wasn't going to be high enough to live up to it's potential.
Looks like I was right and and BMW is fixing it. One can only hope that the specification is open and well documented.
Reply
Eric Bryant 10:26PM (12/18/2005)
auto_dev_engineer,
Let me clarify - CAN has some shortcomings that does not make it suitable for X-by-wire applications, which is the context in which I wrote the above entry. I will not apologize for it being "misleading" as you allege; instead, I simply did not provide sufficent background information for the average reader, and for that I must apologize.
By "electromechanical valve system", I'm refering to the next-gen systems of engine intake and exhaust valve acuation. If one wishes to control the exact position of the valves via electromechanical means, it will indeed take some bandwidth to do so.
"Microsecond" arbitration becomes critical with X-by-wire systems - there simply will be no tolerance for bus collisions when the vehicle is trying to drive itself away from an actual collision.
Reply
auto_dev_engineer 10:26PM (12/18/2005)
>I will not apologize for it being "misleading"
>as you allege; instead, I simply did not
>provide sufficent background information for
>the average reader, and for that I must
>apologize.
It's quite misleading IMO. It indicates that modules can "argue back and forth", and in this period of time, catastrophic failures can happen. This is not the case. Modules do not argue. They all attempt to send their messages. Put simply, the high numbers must yield to the lower numbers. That's it.
You also indicate that "there are ways of asserting priority" -- which again leads the reader to believe that there is some advanced coding that needs to take place. Not true. The message ID *is* the priority. Steering wheel angle gets a low ID, "door state" messages get a high ID.
>By "electromechanical valve system", I'm
>refering to the next-gen systems of engine
>intake and exhaust valve acuation. If one
>wishes to control the exact position of the
>valves via electromechanical means, it will
>indeed take some bandwidth to do so.
It would seem that the module that needs this precision would be the module that's directly interfaced to the appropriate sensors and actuators. I can't imagine wanting this precision over an actuator and trying to get it by sending messages over the bus. The ABS module is directly coupled to the valve block because *it* is precisely controlling the valves.
>"Microsecond" arbitration becomes critical with
>X-by-wire systems - there simply will be no
>tolerance for bus collisions when the vehicle
>is trying to drive itself away from an actual
>collision.
Current production Adaptive Cruise Control systems are CAN-based. Stability control, rollover protection, traction control, ABS -- all CAN based. (even J1850!)
As for brake by wire: The linked article indicates that we are moving in that direction. Mercedes-Benz is moving in exactly the opposite direction, moving away from Brake by wire in the SL and the E-Class. The only other brake by wire application that I can think of at the moment is the hybrid Escape.
Steer by wire and so forth: there are some interesting ratio-changing and assist systems being developed, but all that I've seen are still mechanically connected in order to fail "safe". The jury is still out as to whether or not these systems will deliver performance great enough to equal their cost.
Of course, I don't think that CAN is the last answer... I could see a new bus architecture to carry multimedia or other communications that CAN was really not designed to handle. I do disagree with your statements intending to lead the reader into thinking that the CAN protocol is unstable. CAN is now the standard automotive communication protocol because of its excellent error handling and stability.
Reply
Eric Bryant 10:26PM (12/18/2005)
auto_dev_engineer,
But what happens when you've got multiple modules on the same bus that all have equally-high priority? What happens when a flooded-bus condition occurs? How much more room, speed-wise, does CAN-C have to grow?
I'm certainly not implying that CAN is unstable, but it has its limits, and I remain convinced that those limits will soon be exceeded. I was discussing this topic with my wife (an aerospace engineer who works on, shall we say, "mission-critical equipment"), and she certainly feels the same way. The MIL-STD-1553 bus that's been used for years is suitable for X-by-wire applications largely through the virtue of being deterministic.
Regardless, don't argue with me that something else is required - go argue with the members of the Flexray consortium. You may think you have a better grasp on this topic than I do (and you may be right), but I doubt that GM, BMW, DCX, VW, Ford, Honda, Hyundai, PSA, Renault, and Toyota have invested a lot of time and money in Flexray just for the heck of it.
By the way - your theoretical fiber-optic high-bandwidth bus for entertainment? That's been around for a while in the form of MOST. I've never worked with it (I do powertrain stuff, not entertainment), but it seems quite impressive.
Reply