14

Since CAN Bus is a differential signalling system, what is the significance of GND/COM signal? Can I get by without it?

I've got a very simple 2-node system that has some weird happenings going on. (see here for that issue, if you're interested) The two nodes are my device and a USB-CAN monitor; both have 120ohm termination resistors, and the wire run between them is <1m.

I just remembered that in my earlier tests I only had the CAN_H and CAN_L signals connected between the nodes; I did not have anything connecting the CAN_COM of my device to the ground pin on the USB-CAN monitor.

A (the?) point of differential signaling is to be much less (not?) effected by common mode noise. So by not having the ground connected, I would guess we lose the absolute reference for CAN_H and CAN_L... but does that make a difference?

Community
  • 1
Adam Head
  • 1,416
  • 2
  • 17
  • 27
  • 2
    even with differential pair communication a common reference is needed. Simply because the floating grounds on both ends can have a *very* different potential, whose difference will just blow your system up. – Eugene Sh. Nov 03 '15 at 19:09
  • 1
    Short answer is yes. Unless you have complete optical or magnetic isolation, you ALWAYS need common ground for basically any circuit. Capacitive coupling can also sort of get around this problem of needing a ground although you still need some reference point or you will be above the voltage rating on the cap. – MadHatter Nov 03 '15 at 19:16
  • Is there any way to quantify what _not_ having a ground will do? – Adam Head Nov 03 '15 at 19:20
  • The problem is that you nearly always have something (capacitance, induction, other connections) which sets up the relative ground levels of your two ends. If you don't have an explicit ground, then you're at the mercy of whatever this might be. Given that it may well involve filter caps connected to mains supplies, there could be quite big voltages involved. You should wire the CANbus up how it's supposed to be wired (proper termination, proper cable, proper grounding), and not try to second-guess which particular cheats will be successful. –  Nov 03 '15 at 19:22
  • Of course. I'm not trying to cheat. I just want to understand so I can trust that I've solved my problem (or not) – Adam Head Nov 03 '15 at 19:31
  • Well, the symptoms of running out of range would first be that the system couldn't reliably tell one state from another, which could give all sorts of odd behaviour with CAN, then if things were worse you'd start to damage the transceivers. –  Nov 03 '15 at 20:16
  • "I never saw a CAN networked device with 3rd wire, only CANH, CANL." What about in a automotive CANBus, gnd is common to all units on the bus. As @Lundin has pointed out. – Tom George Dec 23 '21 at 05:58

4 Answers4

9

A perfect, theoretical, differential system can run with no ground, because the input can always do (A-B) without worrying about the absolute values of A or B.

Unfortunately, CAN is not such a system, and the receiver will have limits to its 'common mode' capability, beyond which it be damaged. So you have to have a ground to keep the common mode voltages within limits. 'Common mode' just refers to stuff (voltage, in this case) which is common to both wires.

There are other ways of building receivers which are far freer of this sort of constraint - for example an opto isolator can do the (A-B) detection with maybe 1000s of volts of common mode offset. Transformers are often used too (e.g. in twisted-pair Ethernet, which has no ground). But for normal CANbus, you need a ground.

6

Yes, a common ground is need.

A can transceiver has a maximum common mode voltage. If the common mode voltage of the differential CAN signals exceeds the maximum, then the transceiver will not be able to recognize the bits. In practice, this means that the grounds of CAN nodes have to be connected (although a small voltage differences between grounds can be tolerated).

Special case. Some CAN devices are galvanically isolated. Then there is not common ground throughout.

Nick Alexeev
  • 37,739
  • 17
  • 97
  • 230
3

It depends what system you are building. If you are building in your house or in your car it is better to use common ground. Because it is differential pair type communication the common ground is not required as what is matter is the potential voltage difference between H and L. Actually on some cases you can blow everything by using common ground - when the nodes are powered from different power sources with different ground references. I am always building large industrial CAN networks without common ground and it is working with no issues.

Krachmarov
  • 31
  • 1
0

I would say no. CAN is differential and withstanding common mode voltage is relatively high. Once you connect them together the potential will be equalized, for short discharge usualy transzorb diodes are mounted at CANH and CANL to prevent damage. The nodes are capacitively coupled to the earth, in case of floating PSU. For non isolated version would be not a bad idea to reference GND to the earth on each node, there will be no such difference in earth potential to damage the transceiver.
I never saw a CAN networked device with 3rd wire, only CANH, CANL.

Marko Buršič
  • 23,562
  • 2
  • 20
  • 33
  • 1
    How exactly will the potential be equalized with no common ground? Suppose for example that the CAN bus of a vehicle, supplied from a 12V battery, gets connected to a PC supplied from the 230V power grid. The difference in potential could be anything. – Lundin Nov 06 '15 at 07:17