Bizarre assertion!

[Pierre]

Someone I know has a com problem with some equipments.

They are connected in a token ring pathern.

PC, PLCs and devices are sharing this RS485 network.

On on drop, they constantly loose communication.

The existing cable they use is a 25 pair CAT3 cable.

There supplyer has told them that the problem they had was cause by the fact that "Both SEND and RECEIVE wires are in the same cable."

That if they where to separate them the com problem would go away...

Does that make any sense to you guys?

 

[rsdoran]

No. RS485 is designed to be 2 or 4 wire and work with either a single or dual twisted pair.

The cable could require shielding, some environments its needed.

Cat3 should be fine, it should have different lay length to prevent cross-talk. Cat5 is definitely a good choice.
Another thing to check is termination, cable length can factor into it.

If that one ring is using a master that communicates to the devices outside the ring I would check it.

 

[Pierre]

Thanks Ron,

What puzzles me is the SEND and RECEIVE thing.

Don't they always are in the same cable anyway?

When I setup this kind of network ... I don't have 2 seperate cable for them.

I have not had a chance to see the network but when he told me this thing I looked at him and said

 

[rsdoran]

The specification states single twisted pair or dual twisted pair...yes the wires are in the same cable normally. Send and recieve is what cross-talk is about, CAT3 should easily handle cross-talk as would any cable speced for RS485.

I would check termination, shielding issues and if there is a master that may be dropping out.

 

[CaseyK]

I switched to CAT-5 several years ago, but previously had CAT-3 furnished, and on occaision, still have some left over rolls handed to me. I have never had problems with either, as a result of the cable itself, of having TX/RX in the same cable. I can guess how much I've strung, but probably more than 500,000 feet.

Problems were always something else, like poor termination, damage after installation, strange ground loops, 20,000k paths to ground, etc.

In a high electrical noise situation,shielded cable can certainly help, BUT GROUND ONE END ONLY!

regards.....casey

 

[Ken Roach]

What sort of signals are running in the other 23 pairs in the same cable ?

Focus on connection and grounding at the one drop where the problem occurs. A scope couldn't hurt.

This fall I unearthed a Tektronix 2350 scope and have applied it to quite a few badly installed serial networks. There's nothing quite like visual comparison to try out hypotheses on termination and grounding.

 

[RussB]

The thing that you want to remember with RS-485 is that you must have the proper signals paired together in the cable. i.e.
Pair 1 = TX+ and TX-
Pair 2 = RX+ and RX-
If you pair a TX and RX you will get Cross talk.

 

[JesperMP]

Pierre,
can you describe the whole thing a bit better.
What types of networking equipment ?
What do the "call" the network ? (a name like "Profibus", or "DH-485" or..).

They are connected in a token ring pathern

Really ? Can you describe that a bit more ?
A regular RS485 network has "line" topology.
To achive "star" or "tree" topology repeaters must be added.
(I think that Profibus for example has to move to fiber in order to achieve ring topology. In other words not RS485).
Maybe they are talking about the "token passing" that is a part of most protocols that are used with RS485.

edit:
I remember vaguely that early versions of INTERBUS had many wires, but not how many.

 

[Jiri Toman]

No one has mentioned an end-of-line terminating resistors.
I have had problems with RS-485 on several occasions and in each case
it was either missing or wrong terminating resistor that caused it.

 

[panic mode]

is it supposed to be "on one drop they loose communication"?
does the rest of the network work ok? what is the length of drop?
are the nodes configured correctly (no duplicates...)?

 

[akreel]

There's nothing quite like visual comparison to try out hypotheses on termination and grounding.

No kidding! If you're not seeing a clean square wave, you know right away that you're asking for trouble. It's a shame most scopes are so costly.

Fluke as a moderately priced unit ($150-200). Has anyone tried it for monitoring serial data?

AK