Modbus RS485 Ground vs Drain

[TwoBlocked]

Was looking at this drawing, doing some research, and considering how we are actually wiring devices. We use shielded, twisted pairs with a drain wire for DC power, analog signals, and RS485 communications. Often in a Teck cable bundle. At the field end, we snip and tape off the shield and drain wire as is recommended for analog signals. But am now understanding that for RS-485 it is recommended (required?) to provide a ground conductor, yet still ground the shield at only one end. To do this, you cannot use the drain as a ground. Still, we haven't been experiencing comm problems (that I know of...) by not wiring a dedicated, insulated ground. The devices themselves are grounded through the cable armor and fittings.

Any thoughts? Should we be using twisted 3 wire? (Probably, but that doesn't seem to be the typical way here...)

 

[Birken Vogt]

I don't know, but...

I work on a lot of RS485 stuff where they haven't even provided a spot to land the ground/drain, on the circuit boards. They instruct you to ground it to the chassis at one end and carry it through, only on the cable at the drops, and of course do not term at the far end.

I suspect that transceiver mfrs. are expecting people to do it this way, and testing so that these non-standard installations will still work OK.

 

[TwoBlocked]

Thanks, I am going to guess that as long as there is some kind of ground present, the circuit has a reference to the data signal.

Speaking of which, just today I was inspecting the terminals on a device and they were marked A+ and B-. This is what I am used to seeing, yet the attached drawing shows A- and B+. This same drawing seems to be used anywhere the wiring is explained and I have not found another as clear.

Anyone know which polarity is the true standard?

 

[ModbusMan]

On all our pre-Ethernet Caterpillar generators, ASCO transfer switches, and Liebert FPCs, we've only ever connected the two data lines. Adding the drain, regardless of whether we connected it to signal ground on the JACE masters or a chassis ground somewhere "in the field" has never made things better, only "just the same" or worse.

 

[TwoBlocked]

On all our pre-Ethernet Caterpillar generators, ASCO transfer switches, and Liebert FPCs, we've only ever connected the two data lines. Adding the drain, regardless of whether we connected it to signal ground on the JACE masters or a chassis ground somewhere "in the field" has never made things better, only "just the same" or worse.

Thanks, that is what I've seen also. Although, I was speaking of using a ground wire, not the shield. But, hmmm... even with a ground wire, if it is within the shield, could cause a ground loop current.

 

[Birken Vogt]

I've seen many an installation where some trunk slammer just hooked up the A and B on 14 THHN and ran it with all the other wires, working fine. It really must be a robust system.

 

[ModbusMan]

I've seen many an installation where some trunk slammer just hooked up the A and B on 14 THHN and ran it with all the other wires, working fine. It really must be a robust system.

More like invented in the 70s when you couldn't trust anything that wasn't loud and honkin' obvious (which, come to think of it, describes the vast majority of rock music from that era as well)

 

[hbiss]

But, hmmm... even with a ground wire, if it is within the shield, could cause a ground loop current.

That's why you connect only one end.

-Hal

 

[brycenesbitt]

What are your cable run lengths? Number of devices? Noise exposure?
How fast you trying to run? Do you have access to equipment to measure your noise margin or just packet loss?

And importantly: are your RS-485 transceivers optically coupled or not?

How much money do you have to get to the right solution? You just want it to work over
a short distance with slow data rate: do about anything. You want to go faster than the spec of your transceivers go ahead but be prepared to put in some extra work. Start at about double the specified rate: don't worry you can make it work with careful attention to detail.

=====================================================
The "ground is the braid" approach does work, but is not the very best you can do.
The very best is A+B on a twisted pair: with the more twists the better, thinner wire is better.
Then the ground wire (go ahead and use a triad cable or a two pair cable with one pair dedicated to ground).
Then the shields/braids terminated on only one end. DO NOT CONNECT BOTH ENDS OF THE BRAID.
DO NOT MIX A+B on different pairs. DO NOT USE straight cable.

This gives the shield the best chance of keeping noise out of the A+B pair. Return current from the A+B pair will
go on the ground wire, separate from the shield/braid wire. No ground loops or unwanted currents.

Optically coupled makes a difference, though could also reduce your max data rate.

 

[LarryFine]

In audio, the single-ended shield is usually grounded at the destination end of the line.

 

[petersonra]

If I recall correctly, the RIA spec for RS485 requires that the common terminal be kept at no more than 15 volts from ground. A fair number of people that make RS485 devices abide by this rule and others just ignore it. In most cases it doesn't really matter all that much whether you run three wires including a shield or just the two signal wires. In fact in most cases it doesn't matter if you include the terminating resistors or not.

 

[ModbusMan]

In fact in most cases it doesn't matter if you include the terminating resistors or not.

I'll QFT (Quote For Truth) this bit. I don't think a single one of our 485 lines is "properly" terminated, and in a couple of cases adding termination during troubleshooting broke the data link entirely. What HAS helped several times was to enable bias resistors somewhere along the way (usually at the JACE since the 600 series had jumpers on the comm cards, while the new 8000 uses a small slide switch)

 

[ModbusMan]

The very best is A+B on a twisted pair: with the more twists the better, thinner wire is better.
Then the ground wire (go ahead and use a triad cable or a two pair cable with one pair dedicated to ground).

Curious if you've ever tried using CAT-x (5/5e/6/6A), and what results you've seen if you have. Our reliability engineer and I have kicked the idea around some, but since we've never needed to go faster than 9600 and none of our lines are longer than ~500ft tops, the cheap, non-shielded, single pair stuff that always seems to be sitting around has never failed to satisfy.

 

[brycenesbitt]

In fact in most cases it doesn't matter if you include the terminating resistors or not.

Um....
The terminating resistors provide an important verifiable function, which is to prevent signals reflecting off the end of the cable, and returning the wrong way. I can show this effect to you on a scope in a heartbeat.
The engineers who designed this stuff knew their stuff, and were schooled old school in analog design really understanding the fundamentals, many of which have since been forgotten and ignored.
If you think removing a termination resistor improved your signal integrity: look elsewhere for another flaw.

Here's a random google hit on the topic which looks well informed:

SSZTB23 Technical article | TI.com

RS-485 Basics: When Termination Is Necessary, and How to Do It Properly

e2e.ti.com

 

[petersonra]

Most rs485 networks are relatively short lengths at low baud rates so termination just does not matter much. You can crank up the baud rate on longer runs and thus need to add termination but people who know what they are doing don't do this because long runs at higher baud rates tend to have other issues beyond needing termination.

Plus a lot of people designing these things cheat a bit so higher baud rates are a bit of a problem anyway, especially when mixing different designs together on the same network.

It rarely makes much difference what the network baud rate is anyway. If one needs high speed data one does not use modbus.

I have seen a fair number of such networks run with regular wire rather than some kind of shielded cable.

 

[ModbusMan]

Um....
The terminating resistors provide an important verifiable function, which is to prevent signals reflecting off the end of the cable, and returning the wrong way. I can show this effect to you on a scope in a heartbeat.
The engineers who designed this stuff knew their stuff, and were schooled old school in analog design really understanding the fundamentals, many of which have since been forgotten and ignored.
If you think removing a termination resistor improved your signal integrity: look elsewhere for another flaw.

I've no doubt that reflection is real (check it sometime on the load side of a long-run 480v VFD without a line reactor... you'll see ringing upwards of 2kV), just for whatever reason it's generally not an issue in the world I (and apparently petersonra) live in. I do, however, see them every time I come across an ARC156 network (even really short ones), so baud rate likely plays a critical role in whether or not they're required. I am curious if you've ever tried CAT-whatever, though... like I said, we're just burning through spools of single not-really-even-twisted pair 18ga that are sitting around from years-ago projects, but eventually that'll run dry, and if CAT-6A can do the job, well, there's a reason Graybar's on speed dial

 

[ModbusMan]

If one needs high speed data one does not use modbus.

Weeeellllll..... yes and no. Caterpillar and ASCO are both using Modbus/TCP as their primary interface these days, 10BaseT for the former (stripping CAT-6 and landing it on terminal blocks was a very strange feeling the first time), and 100BaseT for the latter.

 

[zbang]

I can show this effect to you on a scope in a heartbeat.

(a couple of heartbeats later )

This is aimed at DMX lighting control, but that's RS-485 at a higher speed (250kb/s). The scope photos clearly show what can happen on the data lines.

DMX waveforms

Discussion of DMX signal quality over different network configurations, with lots of detailed waveforms and a necessary modification added to an Enttec signal splitter

techno-fandom.org

 

[Speedskater]

In audio, the single-ended shield is usually grounded at the destination end of the line.

In audio, if the shield is only connected at one end, it would be to the chassis at the send end (not the destination end). Note that shields are attached to metal chassis at the connector. A shield connected at only one end does not have as good high frequency RF interference rejection as a shield connected at both ends.

 

[brycenesbitt]

A shield connected at only one end does not have as good high frequency RF interference rejection as a shield connected at both ends.

Avoidance of ground loops is a big reason for single ended shields for communications purposes.

Slew rate limiting is another important topic for such signals: not for the cabling guy but for the equipment designers. End of line termination is critical. And we should be concerned not just with received noise but transmitted noise: end of line resistors are important to avoid random junk RF interference to other equipment, public safety radio, cellular signals, etc....
--
Cat5 wire is great for RS-485. Or cat3. Or cat6. Or cat5e. Whatever you have available.


But as the link above says with regards to end of line termination:

This should forever lay to rest *any* arguments for not bothering to terminate DMX lines.
"Never had a problem" is not an excuse; you can keep getting lucky for just so long.