Battery powered devices communicating via RS485 and grounds

[e100]

This scenario doesn't seem to be mentioned in any of the application notes I've looked at.
 
If you have two devices that are running on batteries with no connection to protective earth, do you need a 3 wire (A, B, Gnd) connection, or will a 2 wire (A, B) connection between the two devices work exactly the same? As in same interference rejection, same bandwidth etc.

With a 2 wire connection presumably the grounds of the two devices will be capacitively coupled through the air. What does this mean in practical terms with the limited common mode range of a transceiver such as a MAX485 https://datasheets.maximintegrated.com/en/ds/MAX1487-MAX491.pdf

[Ian.M]

https://www.planetanalog.com/signal-chain-basics-84-why-rs485-does-not-need-ground-wires/

There was some debate here about RS-485 grounding and the need to take special measures to avoid excessive ground loop currents a few years ago:
https://www.eevblog.com/forum/beginners/rs485-common-ground-or-not/

[e100]

https://www.planetanalog.com/signal-chain-basics-84-why-rs485-does-not-need-ground-wires/

The planetanalog article is different from my scenario as it shows each node with a power supply connected to protective earth ground.

There was some debate here about RS-485 grounding and the need to take special measures to avoid excessive ground loop currents a few years ago:
https://www.eevblog.com/forum/beginners/rs485-common-ground-or-not/

Again I don't see anything that corresponds to my scenario of two devices without a common ground connection.

[Ian.M]

I thought the PlanetAnalog article was fairly clear - the receiver input biasing network typically tends to keep an isolated receiver's common mode input voltage near local 0V, and the differential signal from the transmitter has a common mode voltage of Vcc/2.   However relying on the receiver's input  biasing network will give poor noise immunity, as capacitive coupling to the device body can cause it to transiently exceed the permitted common mode range.

If you use appropriate termination and failsafe biasing networks at both ends they'll hold the local common mode voltage at Vcc/2 and you can have a robust two node half-duplex link over a single pair, with no other connection.   

[voltsandjolts]

RS-485 should have 3 signal wires (A,B,GND) plus shield, if needed.
If ground loops are an issue use one of the RS-485 isolator ICs, in which case you may also need an extra wire for isolator power.
https://www.analog.com/en/parametricsearch/11064#/

Yes, it can work with less than this but that would be asking for trouble in an industrial application.

[voltsandjolts]

If you have two devices that are running on batteries with no connection to protective earth, do you need a 3 wire (A, B, Gnd) connection, or will a 2 wire (A, B) connection between the two devices work exactly the same?

A is actually voltage on A with respect to Gnd
B is B w.r.t. Gnd
Differential is (A-B) = ((A w.r.t. Gnd) - (B w.r.t Gnd))
So, yes 3 wires required.

On battery power you have the advantage that you avoid ground loops, so isolator not required.
BUT be sure sensors and/or actuators do not have a sneaky connection to earth, e.g. Gnd to body.

[voltsandjolts]

IMHO that planet analog article is written by someone academic, with no actual industrial experience.

[e100]

I thought the PlanetAnalog article was fairly clear - the receiver input biasing network typically tends to keep an isolated receiver's common mode input voltage near local 0V, and the differential signal from the transmitter has a common mode voltage of Vcc/2.   However relying on the receiver's input  biasing network will give poor noise immunity, as capacitive coupling to the device body can cause it to transiently exceed the permitted common mode range.

If you use appropriate termination and failsafe biasing networks at both ends they'll hold the local common mode voltage at Vcc/2 and you can have a robust two node half-duplex link over a single pair, with no other connection.   

If you added another battery operated node presumably it would need its own set of failsafe bias resistors which in turn would require whichever driver is currently transmitting to supply more current? Presumably this would reduce the total number of nodes supported by a single bus?

[e100]

IMHO that planet analog article is written by someone academic, with no actual industrial experience.

At the end of the article it says you need to measure the ground potential difference and then pick a transceiver that suits.
What if the ground potential difference is low on the day you take a measurement but the following it goes up because the electrical load changes?
Is every day a new transceiver day as the last one goes up in smoke and you have to upgrade to a more robust model?

[Ian.M]

You should only use failsafe biasing at the terminations, which limits you to two nodes with no common wire (signal ground).  If you need more nodes, for the intermediate ones you are back to the problem of EMI causing their common mode range to be transiently exceeded with likely data loss.  The easiest way out of that is to bite the bullet and add a commonwire.

[e100]

You should only use failsafe biasing at the terminations, which limits you to two nodes with no common wire (signal ground).  If you need more nodes, for the intermediate ones you are back to the problem of EMI causing their common mode range to be transiently exceeded with likely data loss.  The easiest way out of that is to bite the bullet and add a commonwire.

That is really useful information which doesn't seem to appear on any of the application notes from the manufacturers.
I had no idea that a point to point connection could be treated differently from a multi-drop situation.

[Ian.M]

T.I's AN-1027 may be helpful: https://www.ti.com/lit/pdf/snla049