Recommended RS485 connector for Cat 5 Cable - Not RJ45

I like RJ45 as it enables me to use standard patch cables - and easily terminate all conductors when making custom stuff... The connectors are crap when lying around (the tab dies) so if one needs a more sturdy construction I'd opt for the XLR-style RJ45 or perhaps the Harwin (or similar) industrial ones.

Not a problem if it's short. And definitely not at 9600 baud. If it's a noisy environment you might miss the shield though....

Always use slew rate limited parts when possible!

I use micromatch and ribbon cable!

/Anders

We have an RS-485 network that uses CAT5 cables, and have yet to have anyone plug a laptop into it (that I know of), and at 1000 plus networks installed so far, that's pretty good. I think the most important thing here is that I don't see that attaching an ethernet device to the net would do anything other than not work and possibly drag down the RS-485 network. We use slew rate limited, failsafe, with short circuit protection drivers and everything has been goin along great. I know it is always a good idea to design the possibility of an idiot doing something dumb out of a system, but in this case I really don't see it as a problem. If you are REALLY worried about it have get some cable tags and mark the CAT5's with RS-485 or maybe some scarely warning about hooking to computers.

Jim

I don't think this would harm anything. It definitely won't harm typical RS-485 drivers, which are robust. I believe it is unlikely to harm any ethernet device either, but I suppose there are no guarantees. Remember, 100 baseT ethernet doesn't use all the wires, and 1000 baseT requires transformer coupling on all pairs, so I would think you could arrange a safe pinout for RJ-45.

If you haven't designed the electronics yet, maybe you would consider using LVDS instead of RS-485 levels. Then you can use RJ45 with no fear of damaging ethernet should someone inadvertently plug-in their laptop.

Not at 9600 baud.

Yes.

Yes.

I like the idea of using Cat5 with RJ-45 or RJ-11. One thing you should know is that you can plug a regular telephone cord into an RJ-45 socket. This is what we do where I work (I mean the telephone jacks are RJ-45, but we plug regular phone cords into them). So you could terminate the CAT5 network in regular RJ-45 wall plates, and then go from the wall to the RJ-11 in your device with ordinary telephone cords. You might want to use the inner two wires for ground, and use the outer two wires for your signal. This will avoid any conflict if someone plugs the cable into a telephone. It would still allow the possibility of someone plugging a laptop into the wall, but only if they have their own patch cord.

You're welcome. I hope I helped!

--Mac

Really? Never tried this... interesting!

Thanks. I will test this myself. But is it reliable? (after all the RJ45 jack is larger than a RJ11 connector, so doesn't the RJ11 connector shift around or can potentially move while in the jack, possibly breaking the connection?) What happens if someone pulls on the RJ11 cable while inside the RJ45 jack? Does it pull out relatively easily or sticks in the jack and the connector rips off the cable first?

Here's another question: are there such things as *surface mounted* RJ45 jacks (instead of wall plates)? By this I don't mean "surface mount" as in "DIP vs SMT" package but RJ45 jacks that come in a similar box like modular telephone boxes that can be screwed down to any surface?

And how about topology? The recommended topology for RS485 is a bus, but in the real world connections can sometimes be more conveniently made if one uses a star topology (which from most RS485 documention/app notes discourages). How do you run RS485 in the real world where things can be moved around somewhat and you need to have some flexibility to rearrange cable as needed?

In my case this is how the network would work. Master node requests data from a slave node. Slaves don't need to talk directly to one another (keeps things simple and reliable).

I understand the long transmission line concept - a long cable (terminated at both ends) with a whole bunch of taps in the middle (where individual node cables connect to), but what is a practical way to actually connect nodes to the main Cat5 cable with this approach?

How do you actually make the drop connections to each slave node without doing something like stripping the main Cat5 cable at various points and soldering another Cat5 cable to make the drop connection? (to make the solder connection I'd probably have to untwist the wire which I guess wouldn't be a good thing if done a couple dozen times along the wire). Actually the more I think about it I don't think it would be realistic to even attempt to strip and solder Cat5 (at least I don't want to do it!).

I guess this is where a daisy chain config would make life easier - cable plugs into first piece of equipment, then another cable plugs into the first and then into the second etc. That way, I would only have to worry about making cables and not interfering with the main cable (in order to make drop connections). The only thing with the daisy chain approach is a bad or unplugged cable can bring down the whole network since a fault effectively breaks the terminated transmission line.

I need assistance with the realities of running Cat5 cable and making connections - what works and what doesn't for RS485?

I didn't believe it at first, either. My company recently moved to a brand-new building, and the guy who installed all of the telephone and Ethernet cable and jacks said that he was going to use RJ-45 for the telephone. I immediately tried it out, and sure enough, it works just fine. If you inspect the connector and socket, you'll see that grooves kind of hold the connector in the socket laterally, and ridges separate the individual conductors. I don't think there are any mechanical or electrical problems with doing this.

Only the width dimension differs. The retention tab in the center insures that the plug will be lined-up correctly, and the individual ridges keep the connector from wiggling around or coming out any more easily than it would from an RJ-11 socket. Hold an Ethernet patch cable up right next to a phone cord. You'll see what I'm talking about.

I don't know.

I honestly don't know. My gut feeling is that with 9600 baud, and if the wires are not too long, you MIGHT get away with star routing. It would be nice to simulate the whole network. For your purposes, the cat5 cable could be simulated as two 50-Ohm transmission lines with a common ground. The phone cable might be more like two 40-Ohm lines with a common ground, but that is just a guess.

If you don't want to simulate it, and it has to work the first time, it would be safer to either have one long transmission line with short pigtails for each device, and termination only at the two ends, or a star topology with a dedicated transceiver for each node. That would only work if there is some central resource who can control the transmit receive direction, and if the devices don't need to talk to each other, but only to the central resource.

Oh, actually, if you don't want to use pigtails, then each device could have an "in" and an "out." This would make it a "daisy-chain" type of configuration. This would mean that every device would have to be on all the time, unless you work out some kind of automatic bypass scheme.

--Mac

How many total devices do you have? Would using a dedicated transceiver for each device be feasible?

No, I wouldn't use any solder. There are junction box type things out there for dealing with cat 5 cable (I'm not a cable guy, so I can't give you better guidance than that). The idea is that at each drop, you cut the cable, and use screw terminals or something to join the cable together. If you can't find RJ-45 versions, you can certainly find RJ-11 versions of these types of junction boxes. I don't think putting breaks in the wire and short (a few inches) untwisted portions will significantly effect the transmission line impedance in your application.

Yeah. That is the downside.

I don't have any practical experience which can help you, but my gut feeling is that for 9600 baud, you can get away with pigtails of at least 10 feet or so.

One way to look at it is that the time a signal takes to go down the pigtail and reflect and rejoin the main line is around 4 times the pigtail length/C. So for 10 feet, it is maybe 40 ns. If your rise time is much greater than 40 ns, then you should be good. At 9600 Baud, you have around 100 microseconds between edges, and you talked about using slew-rate limited devices, so I would think you would be OK. If your rise time was, say, 1 (or even 10) uS, and your total length of cable is only a few hundred feet, it almost doesn't matter how you arrange it as long as you have some kind of a termination scheme that doesn't overload the driver at DC.

The above sketch analysis is more from the perspective of the main device. If you look at it from the perspective of one of the peripheral devices, then you have a short transmission line, which then forks to create two approximately equal and terminated transmission lines. That is not too bad a situation either, because the reflection comes back from the fork pretty fast, and after that, it just looks like Rterm/2.

The bottom line is that at 9600 baud, the system should be pretty forgiving of less than perfect cabling. ;-)

--Mac

Yes. Mouser carries Leviton jacks like that: 503-41018-Q or

503-41038-IDA for about $5 quantity 1, or a DGS Pro-Audio 154-3009 for about $3.50 quantity 1. These are like the common telephone jacks, in that the connector is molded into the base plate, and there are terminals for the wires. You can also get small surface-mount housings that accept the "keystone" type jacks that normally snap into a wall plate with several rectangular openings - see for example .

Buy a few samples before you place a big order. Some of the cheaper phone jacks use sheet metal screws directly into the plastic for the wire terminations. These tend to strip out when tightened. Better ones have metal inserts in the plastic for machine screws to thread into.

Standard disclaimers apply; I don't get money from any of the companies mentioned above.

Matt Roberds