How to guage noise of 3 wire serial line (RS485)

[cpscdave]

How should I gauge noise on a RS485 line? (2 equal and opposite data lines and a shield line)

I tried connecting a Oscope. Persumably however the same noise is appearing on all 3 lines. So if I attach my oscope probe to 1 of the data lines, and the ground to shield all the noise will be removed.
But if I leave the ground of the scope unconnected doesn't the result become undefined?

I'm trying to trace down problems with 1 serial line, noise on the line makes the most sense but it could of course be a multitude of other issues.

Thanks for any insights :)

 

[tech99]

How should I gauge noise on a RS485 line? (2 equal and opposite data lines and a shield line)

I tried connecting a Oscope. Persumably however the same noise is appearing on all 3 lines. So if I attach my oscope probe to 1 of the data lines, and the ground to shield all the noise will be removed.
But if I leave the ground of the scope unconnected doesn't the result become undefined?

I'm trying to trace down problems with 1 serial line, noise on the line makes the most sense but it could of course be a multitude of other issues.

Thanks for any insights :)

Not quite clWith CRO ground connected to the shield wire, and using the "DC" CRO input, do you see data on line 1 and inverted data on line 2? Data pulses should be based on zero volts.

How should I gauge noise on a RS485 line? (2 equal and opposite data lines and a shield line)

I tried connecting a Oscope. Persumably however the same noise is appearing on all 3 lines. So if I attach my oscope probe to 1 of the data lines, and the ground to shield all the noise will be removed.
But if I leave the ground of the scope unconnected doesn't the result become undefined?

I'm trying to trace down problems with 1 serial line, noise on the line makes the most sense but it could of course be a multitude of other issues.

Thanks for any insights :)

I notice from the Wiki article that the network should at some point have bias resistors fitted, so that one wire is pulled up and the other wire is pulled down in the absence of a signal. Can you see this bias on the CRO?

 

[Jeff Rosenbury]

By hooking the ground wire on your probe to the opposite data line (instead of ground) you should get some common mode rejection. I don't know how much.

Does anyone know how to find/figure that out?

 

[the_emi_guy]

Use two scope probes, one on each data line. Both grounded to shield (good short ground connection, not long wire).

On scope:
Add the two channels to see common mode signal.
Subtract the two channels to see differential signal (on some older scopes you do this by selecting INVERT on second channel the selecting ADD).

 

[meBigGuy]

DO NOT GROUND ONE OF THE SIGNAL LINES WITH THE SCOPE PROBE! BAD ADVICE! Great way to fry the circuit if there are any ground issues in your lab.
That works alright if you isolate the scope chassis from AC ground (with a 3 - to - 2 AC adapter), but it is not advised. Never do this unless there is no alternative.

The correct way is as the_emi_guy advised. Use two probes and subtract the channels. Reference (connect the probe grounds) to true ground, which in your case is the shield.

If the scope ground is unattached, the signal is essentially undefined.

You can use one probe grounded to the shield to look at 1 data line at a time and combine them in your mind.

If all the "noise" goes away (and you see a clean logic signal) when you look at 1 line with ground on the shield, then that is the real result relative to your shield.

I'm assuming the shield eventually goes to the power system ground.

 

[Jeff Rosenbury]

DO NOT GROUND ONE OF THE SIGNAL LINES WITH THE SCOPE PROBE! BAD ADVICE! Great way to fry the circuit if there are any ground issues in your lab.
That works alright if you isolate the scope chassis from AC ground (with a 3 - to - 2 AC adapter), but it is not advised. Never do this unless there is no alternative.

The correct way is as the_emi_guy advised. Use two probes and subtract the channels. Reference (connect the probe grounds) to true ground, which in your case is the shield.

If the scope ground is unattached, the signal is essentially undefined.

You can use one probe grounded to the shield to look at 1 data line at a time and combine them in your mind.

If all the "noise" goes away (and you see a clean logic signal) when you look at 1 line with ground on the shield, then that is the real result relative to your shield.

I'm assuming the shield eventually goes to the power system ground.

You are right. That's not the best advice. It will work, but there are lots of things that can go wrong if one is not careful. That is particularly true if working with higher currents.

 

[cpscdave]

The working theory is that when they ran the line out into the field, they ran it adjacent to some 600V 3PH power lines.
This then goes back to my confusion.

If something external is generating enough noise to cut through the data lines shielding, would that noise not be present on all 3 lines? Then using the shield for the scope ground would cause the scope to ignore the noise?

Would attaching the Oscope ground lead to the power ground resolve this? then I can use the 2 channels + or 2 channels - to figure out the common mode?

I'm also using a fairly inexpensive (cheap) hand held battery powered portable OScope. Not sure if that would make a difference or not.

Thanks for the insights :)

 

[Jeff Rosenbury]

The working theory is that when they ran the line out into the field, they ran it adjacent to some 600V 3PH power lines.
This then goes back to my confusion.

If something external is generating enough noise to cut through the data lines shielding, would that noise not be present on all 3 lines? Then using the shield for the scope ground would cause the scope to ignore the noise?

Would attaching the Oscope ground lead to the power ground resolve this? then I can use the 2 channels + or 2 channels - to figure out the common mode?

I'm also using a fairly inexpensive (cheap) hand held battery powered portable OScope. Not sure if that would make a difference or not.

Thanks for the insights :)

That sounds like it might be dangerous. High voltage lines can couple real power into parallel wires. This happens in proportion to the area of a closed wire loop, so twisted pairs are largely immune. But the ground of the cable could loop through the equipment at both ends and the ground to create a large area. I've heard of cattle dying due to currents in barbed wire fences running parallel to high voltage lines. Still, I don't know how long the wire is, how high it is (the area), and 600V isn't very high for these purposes. It's probably not an issue, but do check on it for safety.

Measure the voltage and current in the ground cable (at various accessible points for the voltage). See if the data lines have the appropriate voltage as well. Perhaps check the data line current as well. (This is a bit of a pain on non-NRZ protocols since you'll need to make sure the number of ones and zeros are the same to get zero current. Plus the termination resistors could be an issue. You might want to make other, less hassle checks first.) The idea here is to protect your more delicate equipment (like your hands) if there is some power coupling.

I was sort of assuming you had a portable one channel scope with an isolated ground when I made my brain dead comment. If you have two channels use that.

Power coupling should be common mode noise at 60 (50?)Hz. Most RS-485 receivers should reject it quite well. However if the signal is over the receivers input rails, it could cause problems. I once heard opto-isolation suggested for that problem. I'm not sure how that would work.

Do you have a ground wire and a shield? They serve different purposes. The shield acts like a Faraday cage, protecting the wire inside. The ground wire provides a return path for common mode noise.

You might want to carefully consider the best grounding scheme.

Are you using 120Ω twisted pair cable? Here's cable guide ap-note. And another.

 

[tech99]

The working theory is that when they ran the line out into the field, they ran it adjacent to some 600V 3PH power lines.
This then goes back to my confusion.

If something external is generating enough noise to cut through the data lines shielding, would that noise not be present on all 3 lines? Then using the shield for the scope ground would cause the scope to ignore the noise?

Would attaching the Oscope ground lead to the power ground resolve this? then I can use the 2 channels + or 2 channels - to figure out the common mode?

I'm also using a fairly inexpensive (cheap) hand held battery powered portable OScope. Not sure if that would make a difference or not.

Thanks for the insights :)

It seems to me that if you 'scope is battery operated, there is no risk making a connection to the ground wire, provided you keep the unit isolated fr5om any other ground. You should be able to see the two data streams. As I mentioned before, RS485 is supposed to have bias applied by resistors. If absent, noise will occur when data is absent.

 

[the_emi_guy]

The working theory is that when they ran the line out into the field, they ran it adjacent to some 600V 3PH power lines.
This then goes back to my confusion.

If something external is generating enough noise to cut through the data lines shielding, would that noise not be present on all 3 lines? Then using the shield for the scope ground would cause the scope to ignore the noise?

You need to keep in mind what signals your RS-485 receiver actually sees: Specifically -
1 - It will see all differential signals and noise across differential pair.
2 - It will see some of the common mode noise from pair-to-shield depending on the common mode rejection of the receiver.

It should not see anything else if the system is designed correctly. If your receiver is sensitive to noise between your cable shield and, say, an Earth ground rod, that means you have not properly bonded your cable shield to your receiver signal ground.

In other words, don't disconnect the cable and probe it. Instead, probe right at the RS-485 receiver with your two scope probes, and scope ground connected to PCB ground plane: This is all that the RS-485 receiver is aware of.

 

[meBigGuy]

You have not described the basic issues you are seeing, whether you tested in the lab with long coiled cables before deploying, what the signals look like relative to the shield at the receiver, what data rates you are using, what signal conditioning, what cable terminations, etc etc etc.

Your problem could be ground fault related. But, you need to verify basic signal integrity before concluding that.

You need to verify that each leg of the 485 signal has a decent rise time and proper peak to peak voltage at the receiver.
If you ground the scope to the shield at the receiver, you can look at the receiver + input and the receiver output at the same time. That will tell you what is going on in a simple signal integrity sense. You should also do this in the lab with a full length cable.

If the basic waveforms look good (decent rise time, no huge over or under shoot, good symmetry), then consider the possibility of grounding between sites. (actually, if there is a grounding problem or induced noise, you will probably see it when you look at the inputs to the receiver relative to ground in the real environment)

Running near 600V 3PH AC can induce current into the cable. (also different ground potentials between the transmitter and the receiver is a common problem) Problem is, that current has to flow somewhere. Say the shield current flows to ground at the receiving end, through the shield. The voltage on the shield at the receiving end is defined as Zero for the receiver. So, current flows into that ground and the node is at zero since there is a low impedance at the shield connection.. But the other two lines of the cable are high impedance (through the receiver), so they show a high common mode voltage. The RS485 receiver is not very common mode voltage tolerant.

You can see this easily (if it is happening) by grounding the scope at the receiving end shield and looking at either data line. If there is induced 60 Hz noise, you will see it easily.

If there are large electrical loads at either site, like switching motors, those can cause ground differences between the two sites, and also cause problems (its common for RS-485 receivers to blow when a large motor switches on or off, or during lightning storms).