Podcast Title: You Down with SDV?
Release Date: December 19, 2023
Parker and Stephen discuss a recent news story about how car manufacturers are increasing their focus on Software-Defined Vehicles (SDVs) which are cars that use more software instead of physical components to control vehicle functions. This prompted a comparison between SDVs and Drive-by-Wire or DbW technology in the automotive industry, which is the use of electronic or electro-mechanical systems in place of mechanical linkages that control driving functions, how DbW relates to Fly-by-Wire tech and the fate of the F-117 Nighthawk, how Marvell Semiconductor is working on technology, particularly Ethernet, to support this change to SDWs, plus we get a brewery project update from Stephen, and much more!
Did you do that illustration using AI? It’s excellent! I’d love to see the prompt.
That was an interesting discussion. Standard Ethernet in cars is quite overkill compared to CAN and I think it’s only used in the infotainment system (to send audio/video), but there’s a big push for adoption of single pair ethernet from few companies. Will be interesting to see if they manage to get that adopted widely enough.
First started with:
what do you think a software defined vehicle / car would look like? provide it in a wide image format. throw in some brewery equipment look to the car.
Then I added:
can you make the image in the style of a patent drawing?
And finally I added:
add engineers writing code at terminals and having that code turn into the car
Which produced the final image:
I then changed the code in the blocks to some random json I thought up
Yeah, I do think its a bit overkill but the idea of using PoE style cabling is appealing for devices that can be in the wattage range. I wonder if we will see maybe a dedicated PoE that can handle more power? Similar to what we have seen with USB Type-C and 3.1 Power Delivery.
USB device all the things in a car instead? Can power your door motors and actuators off of that USB Type-C connector!
As I understand things, one big issue with these different busses is the cost per node. Clearly, at one point in time at least, a CAN node was cheaper than Ethernet. But even CAN was too expensive to use everywhere, so newer and cheaper protocols were used — things like LIN and more recently SENT for things like switches and sensors. For faster links there is Flexray, MOST and CanFD, but fast here means like 10 MB or less.
A lot of focus is on safety and robustness with these buses. Although Ethernet has packet collision detection, CAN’s collision detection is designed with priorities so that the highest priority message wins and everyone else backs off. a concept that doesn’t exist in Ethernet where everyone backs off.
Bottom line IMHO is that a single super bus won’t work because of the economics — you don’t want the same full featured bus / connector that is suitable to for your electric steering module to be used on your window up/down switch.
But one could design a bus / cabling system that merged various busses and had a standard connectors / cabling defined for each level. Not entirely like HDMI carries multiple LVDS channels and I2C and power all together.
Disclaimer — I am not an automotive engineer. These are just tidbits that I’ve picked up here and there.
Regarding the different maintenance needs of electric cars vs ICE powered cars, this EEVblog video (#1339) from a few years ago where Dave tours the maintenance depot of the Sydney transit system. They’ve fully adopted electric busses, and the impact this had on just the maintenance schedule alone was truly impressive (YMMV)