ground loops

[ryster9]

we lost a job of upgrading and increasing the existing service of a commercial office building to supply power & lighting to a new addition .the contractor that won the job installed a new sevice for the addition and tied circuits to the existing offices for power ,emergency etc .new service was not of different voltage or additional phases . all computers and associated wiring are interconnected between the offices , as well as power .
we beleive this is not code compliant , nor safe . constant computer problems are also common complaints . we value youre opinion . thanks :-?:

 

[brian john]

I have read your post 4 times and I can not figure out what you are asking....Please repost with more details (or someone else tackle this)

 

[roger]

Welcome to the forum. Don't take this wrong but, your post makes no sense.

Roger

 

[roger]

Brian, I'm glad it wasn't just me.

Roger

 

[zog]

Zog make three. Me scratch head too.

 

[kjless]

sounds kinda like hes saying that theres allot of noise when he says there having problems with there computers

 

[zbang]

I think I understand...... first off, sounds like there are now two services for a single structure. Not so good. However if all the grounding electrodes are properly tied together, and there is only one N-G bond for each service, then there shouldn't be any additonal noise or ground loop issues (over any existing loops).

Does that make sense?

 

[gar]

090125-2132 EST

I have tried to read thru the question and guess what might be important.

Title is ground loops. I think this should be separated into two parts.

If I create a big loop of wire and connect the two ends to the input of a meter and/or a scope, then I can detect an induced voltage or current. This induced signal would come a changing magnetic field coupling to this coil. I can make a single turn coil, I actually have done this, about 5 ft by 40 ft and place it on the ground below the primary lines in my front or side yard and monitor the effective magnetic field from these lines.

This type of loop and magnetic coupling to it can produce data errors. This is why communication cables use twisted pairs.

The second ground loop consideration is where two separate pieces of equipment are connected together with an unbalanced signal path. This is where one lead of the signal path is common (ground). Now the problem is if the ground voltage is not the same at both ends of the communication path errors in data or damage can occur. For example: Consider two computers connected together with RS232. This connection method uses common as one of the two leads in the signal path. Whether or not a common wire is included in the RS232 cable is of no importance other than it could get burned up because of a large ground current.

Thru the EGC to each computer and however these EGCs get connected together there are times when current flows thru the EGC and this produces a voltage difference between the chassis of each computer and thus an added signal into the RS232 signal path. If this ground current is sufficiently low there are no errors or damage. Somewhat higher and you get data errors. Higher yet and you damage RS232 components. Even higher and you may destroy parts of the computers.

The best solution to both of these noise sources is electrical isolation in the communication path.

It is very difficult to solve the ground path noise problem with larger EGCs, and may never solve the problem. Why run 0000 or heavier copper around for your EGC when #12 is adequate for other reasons.

If the only connections are via Ethernet, then I believe the isolation is inherently provided by transformer isolation in the Ethernet interface.


Other types of problems: Excessive voltage drop because of undersized power wires. Most computers will probably work down to 95 V, and maybe can tolerate complete loss of power for 1 cycle. Large printers may cause substantial voltage drop when their heater is turned on. High source impedance in the supply lines could be a problem here.


One needs to determine how equipment is interconnected. Does this type of interconnection present a possible problem?

Voltages at the equipment while operating need to be monitored. If these are normally on the low side, then corrective action is needed.

The types of problems need to be cataloged to see what areas may need investigation.


zbang:

Not correct relative to your comment on grounding. You can do everything according to NEC and have problems. I like to illustrate this by shorting a hot wire to the EGC at one computer. If we are dealing with 120 V circuits, then the peak voltage between the EGCs of each computer could be 120*1.414/2 = about 85 V.

.

 

[zbang]

zbang:
Not correct relative to your comment on grounding. You can do everything according to NEC and have problems. I like to illustrate this by shorting a hot wire to the EGC at one computer. If we are dealing with 120 V circuits, then the peak voltage between the EGCs of each computer could be 120*1.414/2 = about 85 V.

We weren't, I think, talking about fault conditions. In a no-fault situation, if all the EGCs are properly tied together and assuming that the data reference is tied to the EGC, then as I said, there shouldn't be any problems data-wise. OTOH, running a ground-referenced signal outside of the immediate area is usually asking for trouble. Five racks down ought to be OK, 500' across a factory floor might work for a while.... I did have an installation that ran rs-423 between the 1st and 10th floors of a building, and we didn't have problems, but we had pretty solid grounds everywhere and were only running 9600bps for terminals and printers.

(And yes, I am using 'ground' to mean both protective ground and signal reference.)

 

[quogueelectric]

we lost a job of upgrading and increasing the existing service of a commercial office building to supply power & lighting to a new addition .the contractor that won the job installed a new sevice for the addition and tied circuits to the existing offices for power ,emergency etc .new service was not of different voltage or additional phases . all computers and associated wiring are interconnected between the offices , as well as power .
we beleive this is not code compliant , nor safe . constant computer problems are also common complaints . we value youre opinion . thanks :-?:

I understand and what he is saying is the other contractor just put annother service in rather than upsizing the existing service and what is happening is circuits from different services are being mixed through the data connections causing ground loops.

 

[iwire]

Not correct relative to your comment on grounding. You can do everything according to NEC and have problems. I like to illustrate this by shorting a hot wire to the EGC at one computer. If we are dealing with 120 V circuits, then the peak voltage between the EGCs of each computer could be 120*1.414/2 = about 85 V.

.

What problem would this cause with computers connected with modern methods?

 

[iwire]

I understand and what he is saying is the other contractor just put annother service in rather than upsizing the existing service and what is happening is circuits from different services are being mixed through the data connections causing ground loops.

You can have 'ground loops' with one service just as easily.

 

[quogueelectric]

You can have 'ground loops' with one service just as easily.

Yes but the issue is he lost a job because rather than increasing the service size the competing ec just installed a second smaller much cheaper and not allowed by poco in my area. He fears the two grounds are being mixed by data links. As far as I know ethernet will not use ground as a reference for its timing so I dont think this will cause comm problems. He seems to be looking for ammo to discredit the other ecs dirty pool trick.

 

[gar]

090126-0910

zbang:

Why shouldn't we talk about fault conditions? They occur and most likely cause physical damage rather than just data errors.

Disregarding faults there are still ground path problems relative to noise.

In a no-fault situation, if all the EGCs are properly tied together and assuming that the data reference is tied to the EGC, then as I said, there shouldn't be any problems data-wise. OTOH, running a ground-referenced signal outside of the immediate area is usually asking for trouble. Five racks down ought to be OK, 500' across a factory floor might work for a while.... I did have an installation that ran rs-423 between the 1st and 10th floors of a building, and we didn't have problems, but we had pretty solid grounds everywhere and were only running 9600bps for terminals and printers.

Here is a real world example:

Two HAAS VF-3 CNC machines located side by side, physically 4 feet between them. There was about 1.5 years age difference. One machine used DC servos and the other brushless servos. Both were fed from the same bus duct, 240 V delta, and both were grounded by the EGC ground bus. Additionally each machine had a supplemental ground rod. A separate RS232 cable was run from each machine to a common computer. Computer to CNC distance across the floor was about 50 ft. Cable lengths were about 120 ft, high ceilings.

With servos off communication to both machines was OK, but limited to about 9600 baud. This was cable length dependent. In DNC mode the machine with brushless servos had an error rate so high that no effective communication was possible in either direction. Switching the cables between the machines made no difference. The measured voltage with a Fluke 87 in AC position between the enclosures of the two machines was about 4.5 V with the servos on. The noise probably approximates a Gaussian distribution and therefore peaks would be substantially higher than 4.5 V.

Another observation was by the HAAS serviceman that with his laptop operating on batteries and no AC power cord connection there was no communication problem.

Installation of our I232 isolator system solved the noise problem and allowed the user to work at 115.2 kbaud. With this system you could add a 1000 V 60 Hz common mode signal to the computer and while present communicate without error.

To depend upon a low impedance ground path to avoid ground path noise problems still leaves you with a marginal system compared with using some form of isolation.


iwire:

What is the definition of a modern method?

As I had pointed out copper based Ethernet is most likely to include transformer isolation. This should not experience any problem from normal ground path noise or normal faults.

RS232, 422, and 485 are still common communication interfaces and all are limited to moderately low common mode voltages.

.