Why does ethernet not have common mode/noise problems? 3

[danw2]

Why does ethernet appear to be such an easy plug & play imlementation that avoids the hassles of RS-485 with grounding, biasing, termination, common mode & noise?

I've torn my hair out over the years with local RS-485 networks that aren't any longer than 100m in length, that have had all sorts of 'electrical' problems that just don't seem to appear with ethernet in the same environment. The 485 work was done using good practice: twisted pair signal cable, not running cable in high voltage trays, limited T spur length, etc.

What is it about ethernet that makes it work if the cable connector is wired OK?

Dan

 

[geekEE]

Ethernet is transformer-coupled, which is really good at rejecting common-mode noise. Also, since you use a hub/switch to connect nodes together, each segment is essentially a point-to-point connection and thus, you don't need to worry about the termination that you would on a multi-drop bus. And on top of it all, you're usually running TCP/IP or some other protocol that has error-correction and retries built in. It's all very robust, but each node is much more expensive and complex than an RS-485 connection.

I don't know if you remember when Ethernet was commonly connected with coax (10BASE-2). That was probably 10 or 15 years ago. At that time, you had to daisy-chain the coax between nodes and you had to put termination at the endpoints. It was a huge pain to add more nodes. The current Ethernet implementations are so much easier to work with.

Glenn

 

[itsmoked]

Yes 10Base2 was a huge pain in the butt. I was in charge of the infrastructure at moderate sized engineering firm. Something would go wrong and the whole network would go down. Isolating bits and pieces was torturous.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[LionelHutz]

With RS-485 try using galvanically isolated nodes and run the twisted A and B pair plus a ground for each node. A lot of RS-485 devices use local ground and when they do this the A and B pair voltage shift is limited to about 7 or 8 volts. In an industrial environment it's fairly easy to get local ground shifts more than that.

FYI, in case you were not aware, RS-485 is not a differential pair.

 

[CharlieGill]

All the specifications I have seen for RS-485 indicate that it is a differential transmission. The EIA specification defines ones and zeros in terms of differential voltages.
Charlie

 

[itsmoked]

Da! I was under that impression too Charlie.

BTW: Welcome to Eng-Tips Charlie.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[dpc]

I'm pretty sure RS-485 is differential or at least that was always my understanding. RS-232 is single-ended referenced to local ground.



"An 'expert' is someone who has made every possible mistake in a very narrow field of study." -- Edward Teller

 

[tinfoil]

http://www.maxim-ic.com/appnotes.cfm/an_pk/736/

Rs-485 IS differential

 

[DanEE]

Look up the electrical characteristic called "longitudinal balance".

To me, it is the single most important electrical characteristic that gives ethernet (and telephone lines) their noise immunity.. While transformers were mentioned as a common technique, the complete front end design along with the transformer itself, at both the driver and receiver ends must be designed with the degree of longitudinal balance needed in mind. The relatively short paper below discusses briefly.

http://www.waynekerrtest.com/appnotes/telecom transformer measurement requirements.pdf

Bell Labs and the telephony industry was well aware of how to make this stuff work before the computer industry in general adopted them.

The earlier days of my development engineering work in the computer industry was during the period when data transmission (for the computer industry) began moving away from RS-232, RS 485 and other standards prone to noise and crosstalk, to balanced line designs with characteristics you quote regarding ethernet.

The thing that is always impressed me was that the telephony world (and you have to include T-1) was that you can hang 15 - 18 miles of unshielded pairs on telephone poles, generally 10 - 12 feet below high voltage AC power lines yet, have good signal to noise, with appropriate balance.

 

[LionelHutz]

I posted that RS-485 is not a differential pair. The A and B wires are both positive conductors and both feed to the circuit common. The circuit relies on the common of the different transceivers being connected together. This connection can be via earth ground or via a third conductor. To me, using 3 conductors is not a differential pair.

Often, the common connection is done via earth ground. However, the transceivers can only handle a certain voltage between A or B and the common which I think is specified as -7V to +12V. In an industrial environment it is easy to have a higher voltage difference in earth ground between 2 points some distance away, especially during fault conditions. This ground shift can disrupt communications or destroy equipment.

It's a differential method of transmission in that the A>B = 1 and B>A = 0.

 

[geekEE]

Lionel, I see the point of your distinction in that RS485 can only handle a relatively limited common-mode voltage, but I haven't seen that distinction commonly made in the definition of "differential pairs". Is there a standard that you have seen that does this?

Glenn

 

[LionelHutz]

I make the distinction about it not being a differential pair because it uses a pair of signal wires and a common. 3 wires is not a pair.

 

[itsmoked]

Interesting point.
A differential signal. Not a 'differential' line.

I'll agree to that.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[IRstuff]

I think that you'll find that RS-422 actually does need the ground connection, as well, to prevent common-mode destruction, for the general case.

TTFN

FAQ731-376

 

[geekEE]

That's a very good point. It's hard to argue with the fact that 3 is not a pair. So, that would put RS485, RS422, and USB into this category that is not a differential pair. So real differential pairs would include Ethernet (well, at least the common CATx cable implementation that we've been discussing) and what else? There would be much fewer types of signaling that would fall into this category.

Glenn

 

[IRstuff]

Cat 6(usually) and Cat 7 are shielded, so they technically contain additional connections beyong the 4 twisted-pair signal wires. Cat 7 is supposed to have a shield for each twisted-pair, so, technically, each physical comm link has 3 connections.

I personally don't have a problem with calling all of these twisted-pair, since that's a signalling concept, and not the actual hardware implementation, per se.

TTFN

FAQ731-376

 

[itsmoked]

POTS would be differential too.. Right?

Keith Cress
kcress -

http://www.flaminsystems.com

 

[IRstuff]

I always thought POTS was single-ended?

TTFN

FAQ731-376

 

[itsmoked]

Could be IR, I'm just asking.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[LionelHutz]

[bold]A differential signal. Not a 'differential' line.
[/bold]

I like it. That's a good way to sum it up.

IR - aren't those shields just connected to earth ground though, not part of the circuit path?