Hacker News Reader

Comment by tw04 | Hacker News Reader

Just, no. A taillight short doesn’t cause the radio not to work because of canbus. I have no idea what this nonsense is doing here but a guy advocating for carbureted ICE engines because “computers and electronics bad” in 2023 shouldn’t be take seriously on this site of all places.

View on HN

It can and does. I had a 2021 Ford Transit with a non working front passenger turn signal. I figured at best maybe a bad assembly, at worst, maybe something was up with lighting module.

Nope. Turns out the light shorted and killed my cluster while it was at it. It was under warranty so no cost there, but it would’ve easily set me back almost $2k for parts and labor.

I work on cars a lot, and if I wasn’t a software engineer, I’d be a mechanic (I’m actually going to be opening a restoration shop in the next few years). Basically every system on the CAN bus is integrated with every other system and even if they aren’t, there is a ton of cross pollination of systems you’d never expect.

That same van had a feature that would turn down the a/c blower if you received a phone call so as to not be too loud. Neat feature but i was just counting the days before that somehow cause my radio, a/c or car control module to go belly up.

I can easily imagine how that particular fault could occur. So I wouldn't rule out this particular complaint without having a thorough look at the schematic of the vehicle involved to see through what pathways the CAN bus or the radio itself (usually on a separate bus, but not always) could be drawn high or low. That's the problem with a bus: if you force it low or high that's effectively a denial-of-service.

I've had a case recently where a radio locked up hard because of a CAN bus issue that had nothing to do with the radio at all... (2005 MB CLS). That was quite the headscratcher and it took multiple people with a lot of gear to debug the problem to the point that we figured out what the culprit was (a little box on the main CAN bus embedded quite far away from the radio that apparently controlled the power to the console portion).

That's the problem with a bus: if you force it low or high that's effectively a denial-of-service.

This obviously doesn't always happen in reality but CAN is supposed detect/deal with that - pegged bus should be detectable, bad nodes are supposed to float, etc.

In the presence of moisture and a pulsed LED driver all bets are off.

CANbus is on the way out. The future of networking in cars is Ethernet. A proper switched network with TSN [1] can dramatically cut down on the amount of copper in the car while increasing bandwidth and reliability.

[1] https://en.wikipedia.org/wiki/Time-Sensitive_Networking

My car has a regular good old wiring hardness to the rear lights and there isn't any electronics in there. For increased reliability.

The problem with Ethernet is that it's (usually) not a bus. CAN is still plenty used in newer projects and they're getting even a newer higher bandwidth version.

Ethernet in cars is based on the idea of putting a switch right on the die of all your microcontrollers. Then every device has multiple Ethernet links and helps to route traffic on the network. This way there’s no bulky wiring harness (bus) and everything connects with a mesh topology with single links between devices.

From what I’ve heard, it can save about 100lbs of wiring per car!

Sounds like a routing nightmare. Do you have a more detailed reference about this? Full disclosure, I work in automotive SW, but maybe I'm not working in the right ecu to see this. Timelines for these projects are quite long.

I worked on the networking side at NXP Semiconductors who are building the chips to do all of this stuff. Here's an example of one of their recent devices built with these sorts of applications (automotive and industrial) in mind [1]. They've partnered with Bosch and TSMC on this [2]. Here's Bosch's description of the tech [3].

[1] https://www.nxp.com/products/processors-and-microcontrollers...

[2] https://www.nxp.com/company/about-nxp/tsmc-bosch-infineon-an...

[3] https://www.bosch-mobility.com/en/mobility-topics/ee-archite...

Ooh, that's related to the whole "domain ecu" buzz. I see a lot of push in this direction, but I'm skeptical of the direction they say they're taking. It does not make sense to me, but maybe it's just my failure to wrap my head around it all.

Networking as in the SW stack or some silicon magic?

Would love to pick your brain further about this if you can talk about it because so far, no one has managed to clear to me how they intend to do this magic transition to "SDV", all I saw were buzzwords and hand wavy explanations.

I worked on a pretty narrow feature so I’m not sure how it all fits together. I do know they’re really excited about TSN [1] and its potential to let you mix bulk traffic (in-vehicle entertainment stuff) with safety-critical signals and never drop the latter.

I think it’s fair to say that a lot of this stuff is being pushed from the hardware side and it’ll be up to software devs to figure out how to use it all and make everything work together properly. A lot of this networking tech is being pushed down into silicon though, so on the software side you’ll just be tagging Ethernet frames and letting the hardware do the rest.

[1] https://youtu.be/O5krQUbqpcY?si=QDqfoJDkczo-PwN7

...but that is what happened. Ford designed a car that bricked itself if that failure mode occurred. Whether it was born out of malice or incompetence does not matter.

Reply on news.ycombinator.com