CAN bus over COAX

Today a client asked me if we could run our CAN bus over a 75 Ohm Coax, but i couldn't find a yes/no answer on the spot. So before buying some coax:

I thought remembering that a coax (or other transmision line) works because the return path is 180 degrees out of phase, so the coax thus not radiat when properly terminated (Please correct me if this is over simplified)

A CAN or other differential bus cancels out any common mode signals, but if the coax shield of is not grounded, this would work also..

Any pittfals here ? Would coax impedance cause a problem ? Or maybe any other suggestions on going from twisted pair to coax..

Stan

Reply to
stijnvanorbeek
Loading thread data ...

Why do you want to run over coax when a simple twisted pair is going to be much cheaper? Sounds like someones putting the cart before the horse here, designing the system before the requirement is understood!

Dave.

Reply to
Dave

CAN bus is OR-wired, but appart from supporting that, the exact media does not matter. I'd say it is doable, but the industry standard CAN transciever chips do not target CoAx drive - so you will need to roll something custom.

-jg

Reply to
Jim Granville

CAN is not truly differential. The CAN-L and CAN-H outputs are tied to the ground potential, not floating.

You can connect the CAN-L and the shield to the ground. That will work however you can have problems with the common mode currents flowing through the shield.

If you don't connect the shield to the ground, there is going to be a hell of EMI coupling.

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

formatting link

Reply to
Vladimir Vassilevsky

Why do you want to use 75 ohm coax, are there some existing cable-TV etc. coaxial cables you want to use ?

For proper operation of the CAN protocol, you must be able to send and receive the dominant and recessive state and during transmission be able to detect that when sending the recessive state, the line remains in the recessive state and is not pulled to the dominant state by a collision due to an other transceiver sending the dominant state. This could be implemented with quite different methods in twisted pair, coaxial cable, optical fiber or even radio waves.

In an infinite long transmission line, the forward wave will propagate forever and nothing will come back. If you cut the transmission line at some point and insert a termination resistance (with proper resistance matching the transmission line impedance), all the forward wave will be dissipated in the termination resistance. However, if there is a impedance mismatch somewhere along the transmission line (such as an active 10base2/5 ethernet transceiver or an other CAN transceiver sending the dominant state), there will be a more or less a short circuit and part of the original forward wave is reflected at the impedance mismatch and it starts to travel in the opposite direction (reflected wave) traveling towards the original transmitter.

The original transceiver can then detect this reflected wave and determine if there is something wrong with the bus (e.g. a collision). In CAN bus, the allowed total forward and reflected wave propagation time must be much less than the bit time, in 10base2/5 ethernet, this dictates the minimum frame size (46 bytes net, 64 bytes total) of the ethernet frame, in order to reliably detect the collision. This also dictates the maximum distance between stations for a specific transmission speed and transmission line velocity factor.

I very much doubt that you cold use ordinary CAN bus transceivers for the coax configuration.

On a coaxial cable, it might be more natural to use a voltage level for the dominant state and 0 V for the recessive state, but you would have to rely on the voltage levels only, without any help from the reflected wave from an impedance mismatch.

It should also be noted that (especially thick cable TV) coaxial cables do not have the nominal 50/75 ohm impedance at very low frequencies, so the termination resistance might not work properly and the shielding effect might be worse than expected.

For thick coaxial cables, it might be a better idea of using a higher carrier frequency (say 5-50 MHz) and only send bursts of this carrier into the coaxial cable, when the dominant mode is to be sent.

Paul

Reply to
Paul Keinanen

We are using CAN Bus on one pair of Twisted Pair Ethernet cable: truely differential and easy to obtain. Devices get two connectors, wired in parallel. To terminate, cut a normal cable and solder a termination resistor to the appropriate pair and plug that part into one of the connectors. Or daisy chain devices to build the CAN bus.

--
Uwe Bonnes                bon@elektron.ikp.physik.tu-darmstadt.de

Institut fuer Kernphysik  Schlossgartenstrasse 9  64289 Darmstadt
 Click to see the full signature
Reply to
Uwe Bonnes

NO, CAN BUS IS NOT OR - WIRED. ITS AND-WIRED. YOU ARE WRONG MR. JG

Reply to
dk

Firstly, a CAN bus is only AND-wired if you are referring to the actual electrical levels using standard CAN drivers. If you are thinking about the logical levels, "dominant" and "recessive", you think of it as OR-wired.

Secondly, it doesn't matter in this situation - what is important is that any node is able to drive the bus and all nodes can see it. Wired-OR and Wired-AND is nothing more than getting your termination resistors right - it's not the issue at hand.

Thirdly, every keyboard for the last forty years has had small letters as well as capitals. You might think about upgrading sometime.

Reply to
David Brown

He doesn't want to. A client asked him if it could be done. That's how some of us earn money to buy things: we answer questions from clients.

That might not be true.

The client might already have coax strung around a plant and wants to know if it could be use for CAN communications. If the client doesn't have twisted-pair installed but does have coax installed, then coax is going to be way, way cheaper than twisted pair.

--
Grant Edwards                   grante             Yow! Being a BALD HERO
                                  at               is almost as FESTIVE as a
 Click to see the full signature
Reply to
Grant Edwards

All fair points, but I wasn't the only one that thought this. Maybe a clarification in the OP as to why it's wanted would have prevented this. Additionally coax may already be installed, but as another poster pointed out there needs to be a ground reference so still not necessarily cheaper. I think he could run a "single wire" CAN but only if low speeds are OK.

Dave.

Reply to
Dave

Thanks on this one. Indeed. This is what we do, Try to make a living by supplying solutions that are more cost effect for the client.

To clarify, we are indeed in an enviroment where there is a lot of 75 Ohm coax already supplied.

I also forget to mention that there will be no other nodes on the coax, only at the beginning and end of the 'long' line.

Tu summarize the most important opinions:

- It could be doable with a custom driver circuit

- Trying to connect the CAN-L and the shield to the ground. This could work however problems with the common mode currents flowing through the shield are possible.

- Low speeds could be OK.

Reply to
stijnvanorbeek

How low speed ? LIN BUS would go into coax relatively easily, would that be fast enough ?

You also may be able to cobble a CAN driver, into use. Unlike RS485 CAN devices only drive one polarity, and have an offset receive threshold.

the CAN-L node sinks only, and typically has a Series Diode. The CAN-H can only source, and the RX threshold is between 0.5 and 0.9V So, you could GND CAN-L, and drive CAN-H into the cable.

LT1796 data shows drive of 2.2V into 75 ohms

-jg

Reply to
Jim Granville

Since when have "dominant" and "recessive" been logic levels? Where I went to school logical levels are true and false, or 1 and 0. And in those terms CAN is most obviously _AND_-wired: the only way to get a 1 is for all nodes to be sending a 1 (recessive state).

Electrically (assuming any of the usual PHY layers) CAN is neither AND- nor OR-wired --- the distinction simply makes no sense for a differential signal.

Reply to
Hans-Bernhard Bröker

It becomes an exercise in semantics.

Google search for CAN BUS and Wired-OR easily outnumbers CAN BUS and Wired-AND, but both are used.

CAN bus is a two wire bus, but without push-pull drive - unlike RS485, CAN drivers only drive with one polarity of current. (CANH can only source, and CANL can only sink.)

So you can say if Node A _OR_ Node B is driving current, then the Receive sees one state, or you can say Node A _AND_ Node B must NOT be driving current, {or NEITHER (Node A _OR_ Node B) Driving } for the receive to see the other state.

Common usage for OR-Wired simply means that multiple-drivers are possible, and in i2c and CAN that is used for arbitration in Multi master collision instances.

-jg

Reply to
Jim Granville

Jim pretty much wrote what I meant. I wrote "logical levels" dominant and recessive, but in retrospect it was a confusing choice of words - I was referring to the two abstract states that CAN uses. It's the same sort of thing as referring to "activated" and "deactivated" for a signal at an abstract, or logical level, which may actually be implemented as active low or active high.

Reply to
David Brown

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.