CAN over Ethernet via splitter?

Does anyone know if it is possible to run CAN over the unused pairs of an Ethernet cable? We need to run CAN and Ethernet to our turret through a snaking line, and it would be nice to have fewer wires going back and forth.
I know 100mbs ethernet requires only two of the four pairs in the cable, so I was thinking of using a splitter and running CAN over the other two pairs. I don’t know enough about CAN’s wiring requirements to know if this is a good or bad idea…
(For the splitter, I am looking at something like this:


I think it should work. Try it.

1 Like

Make sure to use one of the sets of twisted pair.


I would need to use both - out and back. (Assuming CAN termination remains at the rRIO and PDP).

Nothing saying you can’t put a CAN terminator out there and treat this like a branch, no?

I wouldn’t say it’s a recommended practice to do that, but it’s an option.


Keep in mind that if you are using POE (for a limelight, gloworm, etc) to power the device on the other end you will not be able to do this as passive Poe utilizes two of the pairs.

Yea, twisted pair in both directions.
If it’s a long run then you can use only a single pair and terminate like Eric suggested, but start it from the Power Distribution and turn off the terminator there and it’ll be a standard configuration.

Your CAN bus doesn’t have to end at the PDP/PDH. Turn off the termination resistor there, and add a 120Ω resistor to the end of your bus. Now, you only have to run the CAN there, not there and back again.

1 Like

Adding onto this, don’t terminate on this branch and also at the PDP. Adding extra termination resistors could cause issues. Remove the PDP termination resistor if you want to terminate on a branch.
You can also probably just leave it as a long branch with no termination and it’ll Just Work, as long as the branch isn’t more than a few feet long.

1 Like

Keeping termination at the rRIO and PDP makes it a little easier to bypass components if necessary for testing or troubleshooting. We may change our mind, but I think we will be fine on cable length.

What about the wire gauge? All the CAN cable I see for sale is 22AWG. The Ethernet cables I see are all 24AWG or thinner. Is that too thin? It makes me question the RJ45 connection also. I am seeing posts recommending care what you use to connect CAN segments.
I don’t know enough about CAN bus requirements to know what’s safe. I’m starting to think we should just run the CAN wire and not get too clever.

Should not be an issue. I believe that 2495 has used Ethernet wires in the past as long CAN wires (we cut the end of the cables).

See also CAN BUS with RJ45? | Microchip

Technically the CAN Bus wants to see a very specific type of wire, with so many turns per inch, having a certain capacitance, and so on. If you want ideal conditions for your network, use the right wire.

That being said, UTP will work, but over a shorter distance than with the correct wire. You should try it and test carefully to verify this. My opinion is that it should work fine for the distances we’re concerned with (i.e., less than 10 feet).

BUT this is not a blanket endorsement for teams to use UTP wiring for ALL of the CAN Bus: It is a false economy, and you’ll be asking for trouble.

Good luck, tell us what you learn!!

Good info also at Using CAT5 Ethernet cables for CAN - Interface forum - Interface - TI E2E support forums

1 Like

This sounds reasonable, provided you have two twisted pairs that were otherwise unused (e.g. no POE), and you use one twisted pair out, and the other back.

While I’m not sure of the requirements CAN was designed for, it’s most common usage appears to be automobile control systems. This means that all but the most complex and inefficiently designed FRC robots will have significantly shorter bus lengths than CAN’s most common use.

CAT5 cable has 52-72 twists per meter (actually four different numbers to reduce crosstalk among the four pairs). The SPARK MAX CAN cable that happens to be here on my home desk has three twists per two inches, which works out to 59 twists per meter, practically in the middle of the range. CAT5E, CAT6, and up have more twists - and that will never be a bad thing; more twists implies smaller antenna loop area. I have seen much sketchier CAN configurations (e.g. star rather than bus, and using power poles for CAN connections) which teams have reported using with no issues. I can’t make any guarantees, but this seems perfectly sound.

1 Like

@gartaud, I remember 2495 using CAT5 - it’s what gave me this idea. However, the differences between Ethernet and CAN make me a little nervous about introducing uncertainty. It may be over-cautious, but I’m inclined go with the CAN wire unless I find a compelling reason not to. Right now the cable won’t be that long, and it should wrap up reasonably well.

Ethernet and CAN expect slightly different characteristic impedances. This is probably not an issue, in practical terms in most FRC scenarios, but it’s one reason to be hesitant. Cross-talk could be another. Given that some CAN issues can hide pretty well and be tough to track down, not doing this is the safe option.