[Animats]

Those look much like AMP automotive connectors.

Using one kind of unkeyed connector in a car may be a bad idea. Things can be plugged in wrong during repairs. There's a lot to be said for making connectors not fit where they shouldn't. Especially in automotive, where many connectors are plugged in blind, by feel. This simplifies manufacturing at the cost of repair.

If it can be plugged in wrong, it will be plugged in wrong. AAA put a battery in backwards in my Jeep once, and most of the vehicle electronics had to be replaced.

[elcritch]

Actually there’s two paths: first the traditional path of different unique wiring harnesses and various signals with unique connectors, or second a common power and coms bus design.

If the Cybertrucks electrical ethos is followed by others there’s only 48V and Ethernet. Ethernet doesn’t care or fry if plugged into a wrong port. Any complex wiring can be done inside a part or component as needed, but the interface is one of a few options.

Let’s say the window motor is plugged into the power and not the switched motor plug. As long as it’s a 48V motor it’ll just turn but not fry. You just unplug it and reconnect it.

IMHO industrial everything should become 48V Ethernet. Just like for gadgets usb-c rules the roost.

[michaelt]

> If the Cybertrucks electrical ethos is followed by others there’s only 48V and Ethernet.

I'm looking at the electrical schematic [1] and 'eth' appears 107 times while 'CAN' appears 805 times.

And if you think about it, you probably don't want your brake-by-wire system to share a bus with your sound system and your trunk latch for obvious reasons.

[1] https://service.tesla.com/docs/Cybertruck/ElectricalReferenc...

[elcritch]

Cool, thanks! It looks like the cybertruck does extensively use CAN still. Tesla only talks up Ethernet.

Makes sense, though I’m a bit bummed.

> And if you think about it, you probably don't want your brake-by-wire system to share a bus with your sound system and your trunk latch for obvious reasons.

Sharing a bus for those two wouldn’t make sense. However Ethernet topology wouldn’t preclude having those on separate buses and linked via switches.

Though that’d pose problems with my view above about plugging anything in anywhere. Though it’s really more of a philosophical goal. ;)

[HeyLaughingBoy]

> However Ethernet topology wouldn’t preclude having those on separate buses and linked via switches

The topology might not care, but someone designing a trunk latch that communicates via Ethernet might get a visit from an annoyed representative from Value Engineering.

[elcritch]

Haha touché. However unintuitive as it may seem an ethernet based trunk latch could be overall cheaper from a system level than a straight relay based system or even CAN system.

Especially with the newer two wire ethernet with power via 10BASE-T1S. The connector likely costs more than the ethernet chip, and microprocessors can be had for nickels. So at large manufacturing scale it could easily cost less to share a single ethernet and power line and save on wiring, connector complexity, etc.

[HeyLaughingBoy]

If costs have dropped to the point that Ethernet is less expensive than LIN, it would definitely be a surprise.

[aeonik]

What are the obvious reasons?

[solarkraft]

Different quality of service requirements. The infotainment system is complex and it failing is just annoying while the brake failing could mean death.

[Glawen]

Denial of service of the bus due to a bug or electrical malfunction

[aeonik]

I thought the discussion was on context of CAN vs Ethernet.

These two are still vulnerabilities with Ethernet.

[Aloisius]

Even Cybertrucks aren't purely 48V and ethernet.

There's a high voltage (800V) rail for high current devices like the AC compressor. There's redundant CAN for communication with things like the motors. There's a host of 12V, 16V, 5V components (door locks, lights, seat motors, etc.).

They did switch many components to 48V, but not literally everything.

[black6]

One of my takeaways from my time as a RATELO is the thought put into connector design. Everything in the Army was designed to plug into one thing and one thing only to prevent mistakes.

[viraptor]

> Using one kind of unkeyed connector in a car may be a bad idea. Things can be plugged in wrong during repairs.

I tried to lower the window in a Ford shortly after leaving the dealer. It fried the lock. Somehow they were connected together by accident. I agree dumb wires should be hard to mix up like that - both via orientation and similarity.

[themaninthedark]

These looked keyed to me. Aside from the bulge on the top of the connector where the lock attaches to it's mate there is a middle vane that extends 2/3 way down with a little bit of a J hook on it.

[cduzz]

I think by keyed it isn't just "can't be plugged in upside down" but also "can't be plugged into the other plug that's basically the same but connects to some other subsystem."

Ideally a system like this would let you select some per-subsystem physical lockout mechanism.

[themaninthedark]

That level of keying would be done at the harness/electrical engineer level. You shouldn't put two same plug ends at one termination site.

You really couldn't control it at plug or part level.

[theamk]

It has to be both.

The plug manufacturer has to provide many, many keying options (say for a 2-pin plug, one might have variants 2A, 2B, etc... which explicitly cannot mate with each other)

An engineer designing the car has to ensure all plugs with same keying options are interchangeable (2A is for power supply; 2B for door switch; etc...)

If the plug manufacturer does not provide enough keying options, this will be pretty hard to during design time.

[themaninthedark]

You are generally not using just one plug style on a harness so most of the control is at the engineer level.

Unless you are a big player, what plugs you are using boils down to what receptacle the component that you want to attach to use.

[taneq]

Have they said there isn’t a mechanism for individually keying connectors? There might be something optional, even if it’s just blanking pins.

[numpad0]

Connectors with distinct shapes and sizes are usually used. If it feels like it could go in then it's the right one, and if it goes in it easily seats fully and never comes out without a tool. Wrong connectors don't even feel like they could go into wrong locations.

Cars using whole bunch of different connectors relying on whole bunch of suppliers is feature-not-bug situation. It is optimal for large scale volume production; work will be more distributed, SPoF will be more localized, etc. Standardized connectors with trivial visual differences and/or field configurable keying is a suboptimal solution for car problem. Usually.

...is it a local minima for small scale production? Are they having issues with scale outs, and therefore seeking downward scalability?

[taneq]

I work in this space and 100% agree, in fact I have been on a bit of a crusade to get the mechanical guys to stop designing parts that are rotationally symmetric for this exact reason. Letting a set of three parts be installed 17 different ways isn’t “keeping our options open” it’s “fucking up maintainers.” Each part should fit exactly one obvious place. Each connector should plug in at exactly one obvious place. The only exception is when it doesn’t matter where the part goes, which is approximately never.

That said, sometimes there are cases where an entirely new connector style isn’t warranted, and that’s where you use blanking pins or adjustable keyways or whatever.

[Animats]

That's the latch, not a key. Keying prevents putting the wrong plug in a receptacle. Usually that's done with notches on the receptacle and ridges on the plug.

[themaninthedark]

You can use the ridge that the latch attaches to as part of the key structure.

There is also a J-shaped structure inside the plug that would function as a key as well, prevents the plug from starting when rotated 180 degrees.

    ==
  ------
  |  | |
  |  / |
  ------

[maxerickson]

At least reverse battery protection is pretty standard these days.

Doesn't make things work when they are plugged into the wrong connector, but they should still work once the connections are straightened out.

[Animats]

There's no excuse for that any more, especially since the invention of the ideal diode.[1]

[1] https://www.ti.com/lit/an/slvae57b/slvae57b.pdf

[jiveturkey]

Nah, these look like keyed Deutsch DTM connectors. They are very obviously keyed, so not sure why you say unkeyed.

This will never take hold. These are extremely expensive vs weatherpack and other cheap connectors. Auto manufacturers care about literal pennies.

[inferiorhuman]

Well that's a lot to unpack. Saying AMP connectors doesn't really mean much of anything because, for starters, AMP (TE Connectivity) owns Deutsch.

In the retail space, at least, DTM's largely been superseded by the lower-cost ATM line. Weatherpack and Metripack are pretty common in the US because GM developed them and GM is a huge company. The only company that comes to mind for using DTM connectors is Caterpillar, so you'll definitely see medium duty trucks with DTM assemblies.

The two automotive companies I'm most familiar with (80s Volvos and 00s BMWs) use keyed connectors all over the place. The Volvos used off-the-shelf washing machine connectors (AMP /(junior )+(power )+timer/) that can be had with keyed connectors. BMW saved pennies by going with high density connectors and small gauge wire. Most of that stuff is off-the-shelf as well, but often with proprietary keying.

Both DTM and Weather/MetriPack are pretty bulky compared to what's available now though, and when you're talking about 100 or 200 pin connectors size probably matters more than a few pennies. And, of course, once you start adding the retention doodads to MetriPack assemblies you start getting closer in price to DTM style stuff. The simplicity of the wedgelock design means fewer parts to stock and potentially faster assembly which could easily negate the more expensive housings.