REGISTER | LOGIN
Breaking News
News & Analysis

Marvell Jumps Gun on Gigabit Ethernet for Cars

10/19/2015 08:00 AM EDT
6 comments
NO RATINGS
Page 1 / 2 Next >
More Related Links
View Comments: Newest First | Oldest First | Threaded View
MarkSinger
User Rank
Freelancer
Re: 1000BASE-T1
MarkSinger   11/1/2016 1:56:39 PM
NO RATINGS
There are several advantages to ethernet in the automotive field.

 - Substantially lower cost for cables and connectors than MOST

 - As an asynchronous network, upgrading one node's performance does not require replacing all nodes on the network

 - Star network topology allows aggregation of bandwidth

The last advantage is greatly enhanced by a 1Gb backbone.  This allows several 100Mb nodes to be aggregated, without requiring 1Gb on the whole network (as a ring configuration would).

To this point the availability of the new Marvell PHY could indeed help drive Ethernet into the automotive environment.  But ... networking is driven by standards. A network is only valuable if there are sufficient connections/devices/nodes on the network. This particular orchestra does not gain benefit from soloists. So perhaps the primary benefit of the new Marvell PHY will be the acceleration it causes to the process of standardization. 

As noted by a prior commentor, AVB/TSN will also be critical. A 2 second network delay of your streaming video from YouTube or Netflix may well be acceptable. But a 2 second network delay of your streaming video from a back-up camera is NOT!  

The penetration of Enthernet into the automotive market will come, but it needs standardization and guaranteed delivery of time-sensitive content before it can become the industry's standard network configuration. 

Bert22306
User Rank
Author
Re: 1000BASE-T1
Bert22306   10/19/2015 5:38:45 PM
NO RATINGS
Much of that depends on switch design, and also on traffic engineering. For example, a switch can be designed to forward Ethernet frames at line speed, between any two ports, through the switch fabric. Unless traffic merges, you won't see any delays. Store and forward delays are minimal - through a 1G Ethernet switch, a 1518 byte max length Ethernet frame takes only 12 microseconds to store, and then 12 microseconds to forward. Typical control signal frames will be much smaller than 1518 bytes.

So here's a pretty fool proof example. You can take whatever loads are currently being sent over a CAN or a MOST bus, send over a 1G Ethernet network instead, and you should see a decrease in latency. If an Ethernet net is used for vehicle controls only, the traffic load should be well-defined. Use an Ethernet that is much faster than that aggregate load, and you'll get excellent performance. Try this enough times in a lab setting, and eventually it will become second nature.

TonyTib
User Rank
Author
Re: 1000BASE-T1
TonyTib   10/19/2015 5:26:01 PM
NO RATINGS
The issue with current Ethernet switches is still real time performance -- with typical store and forward, latency is variable, and if the switch becomes overloaded, that can add delays, too.


Supposedly, the newer industrial/automative standards (there may be more than one) will provide better real time performance, although not at the level of dedicated real time industrial Ethernet protocols such as EtherCAT, Ethernet PowerLink, Sercos III, and Profibus IRT.

Bert22306
User Rank
Author
Re: 1000BASE-T1
Bert22306   10/19/2015 4:54:21 PM
NO RATINGS
I think 1G Ethernet is needed, and beyond. What will "seal the deal" is experience with Ethernet, switched Ethernet that is, where much of the angst about the CSMA/CD protocol simply does not apply. It will take some time to get this engrained in the automotive design culture, but I'm pretty sure it will happen soon enough.

I'm also fairly certain that there will be more than one Ethernet involved in a vehicle. No reason to assume the video signals will be on the same net as engine controls, for example.

As to coexistence of the 100 and 1000 Mb/s variants, this should be a no brainer. Let's say that some subsystems are already designed to use 100 Mb/s Ethernet ports. These subsystems can remain unchanged when the 1G switches are introduced. A 1G Ethernet switch can support multiple 100 Mb/s Ethernet ports, and/or a 100 Mb/s Ethernet client system can auto-negotiate a 1G Ethernet port to operate at 100 Mb/s.

The norm these days, in twisted pair Ethernet, is that switch ports can operate at 1, 100, or 1000 Mb/s compatibly. So increasing the switch speed from 100 to 1000 Mb/s should not create any problems, assuming this T1 standard does what all other copper twisted pair Ethernet standards have been doing for all time.

d-allred
User Rank
Rookie
Re: 1000BASE-T1
d-allred   10/19/2015 3:05:07 PM
NO RATINGS
Until TSN (AVB 2.0) arrives with its promises of guaranteed delivery deadlines, Ethernet in autos won't take off. But 1000BASE-T1 is a necessary, but not sufficient, component of that transition. So kudos to Marvell for pushing forward, but they have jumped the gun - or maybe not. Who can accurately predict the standardization timelines for both of these?  

junko.yoshida
User Rank
Author
1000BASE-T1
junko.yoshida   10/19/2015 12:46:36 PM
NO RATINGS
We have just begun BroadR-Reach-based 100Mbit Ethernet coming into a vehicle.

With Gigabit Ethernet designed for automotive (1000BASE-T1), so much more could be done as a backbone bus inside a car. At least that's the premise being promised by Marvell.

Do you think this will seal the deal for deeper penetration of Ethernet inside cars? 

Like Us on Facebook
EE Times on Twitter
EE Times Twitter Feed