RS485 Ground isolation

[hsn93]

hello,


I've Rs485 <=> usb which i connect to the PC.

and power supply that is powering my board.

Im running my signal (A+B) + power (24V + GND) through this cable (< 5meters)

http://www.cnctech.us/pdfs/2000-2422-1-BL-0025-TS.pdf


the shield is not connected to anything from both sides 


- the lab power supply is isolated than the PC obviously.


if you notice the RS485 has Gnd wire

if i common ground the lab-psu and the (usb-485) .. everything works good
however if i keep them isolated the signal is unrelaiable and noisy.


i dont use (isolated RS485) but still
according to my understanding RS485 measures the potential between A and B (ie Return path is A or B)
so why its affecting the communication here?

[mariush]

You're supposed to use twisted pair for data. That cable doesn't seem to be twisted pair.

See https://www.lammertbies.nl/comm/info/RS-485.html

[capt bullshot]

Non-isolated RS485 needs the common GND to work reliably. Even with isolated transceivers it's recommended to have a common GND for the isolated side. The receiver side does indeed receive the difference voltage between A and B, but isn't capable to deal with larger common mode swings that may easily occur if their GND isn't connected.

[hsn93]

Non-isolated RS485 needs the common GND to work reliably. Even with isolated transceivers it's recommended to have a common GND for the isolated side. The receiver side does indeed receive the difference voltage between A and B, but isn't capable to deal with larger common mode swings that may easily occur if their GND isn't connected.

im thinking,

so from my understanding the return path is in A,B but still the IC chips work with some common ground (basically because of "low common mode noise rejection" ?

i just noticed that usb ground has around (100VAC @ 50Hz) with the lab power supply ground , + , earth.

this means the PSU of my PC is not isolated ?
or it could be isolated but some capacitor is there for some reason .. which reason??

You're supposed to use twisted pair for data. That cable doesn't seem to be twisted pair.

yeah i cant find on the datasheet i have to check physically because i remember that i picked it up specifically for RS485 with the impedance 120 Ohm

[Paul Moir]

so from my understanding the return path is in A,B but still the IC chips work with some common ground (basically because of "low common mode noise rejection" ?

i just noticed that usb ground has around (100VAC @ 50Hz) with the lab power supply ground , + , earth.

this means the PSU of my PC is not isolated ?
or it could be isolated but some capacitor is there for some reason .. which reason??

Inside the laptop brick is a capacitor across the transformer which is used to reduce EMI.  It also leaks some AC into your laptop's ground. 
Your understanding's correct.  You don't say how fast you're running but twisted pair isn't going to matter over that short distance.  However do make sure you have your termination resistors in place; you'll have to guess at the impedance of the cable but if you are a little off it won't matter much.

[capt bullshot]

Non-isolated RS485 needs the common GND to work reliably. Even with isolated transceivers it's recommended to have a common GND for the isolated side. The receiver side does indeed receive the difference voltage between A and B, but isn't capable to deal with larger common mode swings that may easily occur if their GND isn't connected.

im thinking,

so from my understanding the return path is in A,B but still the IC chips work with some common ground (basically because of "low common mode noise rejection" ?

Yes, the signal return path is from A through the termination resistor through B (and vice versa). The receiver provides a (relatively) high input impedance to the signal, so there's not a significant return path through the IC (receiver side). This allows the GND of the receiver side to float in respect to the GND of the transmitter side - so if there's a interference signal (like your 100VAC from the power brick), the receiver IC will see this as a common mode signal (attenuated by its source impedance and the ICs input impedance). The common mode range of the IC can be easily exceeded this way, and voila - your signal gets disturbed. A common GND  line for all the RS485 transceiver provides a low impedance path to attenuate this kind of common mode interference. There are some (edge) cases when RS485 works without common GND, in environments where little common mode interference exists. In general, it's better and more reliable to provide the additional GND wire.

Termination resistor and twisted pair will be necessary at larger distances and higher speed (ballpark figures > 10m, >500kbps), otherwise it just works fine without.

i just noticed that usb ground has around (100VAC @ 50Hz) with the lab power supply ground , + , earth.

this means the PSU of my PC is not isolated ?
or it could be isolated but some capacitor is there for some reason .. which reason??

No, the PSU is isolated. It's an EMI supression capacitor that shorts the high frequency components of the switcher across the transformer. Alas, this provides a high impedance path for line frequency, and that's the 100VAC you can measure. It's considered harmless (for humans) but as you experienced, it can disturb your circuitry. Shorting that leakage current with a common GND is a proper measure to deal with it.

[cypher007]

back in the 90's we had all sorts of coms problems with some HV controllers using RS485. cant remember if we figured it or were told by the manufacturer, but essentially we were connecting the shield to GND both ends, it turns out this is not good and once only one end was connected coms cleaned up.

I wonder if it was some sort of ground loop causing interference?

[capt bullshot]

back in the 90's we had all sorts of coms problems with some HV controllers using RS485. cant remember if we figured it or were told by the manufacturer, but essentially we were connecting the shield to GND both ends, it turns out this is not good and once only one end was connected coms cleaned up.

I wonder if it was some sort of ground loop causing interference?

Yes, ground loops can happen if your interference source is low impedance. If the open loop voltage is low enough, then the RS485 transceivers can handle this, and that's a use case they're intended to solve. Note that the common GND is still active here, just some other way round. Nasty ground loops often are an issue with power electronics (like VFDs). Today it's often recommended to have an isolated RS485 (or whatever) comm port to break the ground loop so one can have the shield connected on all ends. The correct solution for interference problems heavily depends on the usage scenario.

[T3sl4co1l]

Because there is no such thing as truly differential.*

One-wire connection: a ground reference is required.  Ground noise is indistinguishable from line noise.

Two-wire connection: the ground reference doesn't go away.  You just have two one-wire connections.  Internal logic just happens to be decoding them differentially.

This is a bit more pessimistic than reality, but it does apply to the most gross case.  An RS-485 receiver is good for about 7V of common mode noise; below that, it behaves differentially.  Above, it starts reading gibberish, and you have to account for your two-wire connection being made of two one-wire connections.  In other words, not providing the ground means you're inviting trash.

An isolated receiver is still two wires, but because of the isolation, the two wires can have much more control over what the receiver sees, and a direct ground connection is not needed.

There are still conditions where an isolated receiver will see trash.  Namely, at very high frequencies.  But these are practical to filter out, as well, and it's possible to make it behave in general.

*For the pedants -- in the sense of differential inputs.  An input pin is still an input with respect to the receiver's supply and ground pins.  Not with respect to the other input pin, as a true isolated differential input would be.  The voltage range where each condition applies, is the problem.

Tim

[macboy]

Because there is no such thing as truly differential.*
...

Ethernet solves this with transformers at each end of the differential cable. This provides true isolation and true differential signalling with very good CMRR.

For the OP, a possible solution may be to capacitively couple the grounds, perhaps with a high-ish value resistor in parallel. This maintains isolation at DC, but shunts AC noise to hopefully bring the common mode signals back into range of the receivers' input pins. The amount of capacitance (and optional resistance) required will depend on each situation.

[T3sl4co1l]

Even Ethernet is not perfectly free, as it has a 1.5kV limit and the isolation is galvanic, not RF (there's a 1nF to ground to knock down peak ESD voltage).  Obviously, a far sight better than RS-485, still.   The most truly differential is propagating EM waves, I suppose (confined to a fiber or waveguide, or otherwise)?

Yeah, RF bypass caps can help.

Tim

[Brutte]

The RS-485 signal is differential but only within common mode range. Typically +12V down to -7V with respect to GND pin of transceiver. If you exceed this range then it stops being differential. Just keep A and B within these limits. Of course the idea with third wire is ok.

What about a latched comparator that senses for +12V and -7V violations? That could be helpful in your development.

[mikeselectricstuff]

For one-way comms, you can use passive-input isolators like IL610, which are truly differential and don't care about the common-mode voltage. Loading is quite high so only viable for niche applications, but can be very useful.