Troubleshooting ground indication on ungrounded ac system without isol 4

[electricpete]

We have a ground indication on an ungrounded single-phase 120v system supplied by an inverter, feeding numerous electronic loads.

We measured from each leg to ground and found 85vdc on one leg and 45vdc on other leg. (total voltage 130vdc)

With a single-phase system there is no possibility for unbalance. With no known loads connected to ground there should be balanced voltage from each leg to ground, based on similar physical configuration (capacitance) from each phase to ground.

We don't think our electonic equipment has any caps connected to ground.

The loads cannot be removed for ground isolation. Any suggestions for troubleshooting. Our plan is:

Check another similar system to confirm these measured voltages are in fact abnormal.
Attempt to measure equipment grounding conductors by clamp-on.... or attempt to measure both supply conductors with one clamp-on. This may help us to isolate the suspect load or component.

 

[Shortstub]

Electricpete,
There are numerous threads on ungrounded AC systems. However, most cover three-phase, 3-wire systems. The unbalanced reading is caused by different phase-to-ground capacitance, usually related to both cable size, length, and insulation deterioation due to aging. Perhaps the aggregate L1 and L2 circuit lengths are appreciably different.

A question that begs an answer is "why is the system operated ungrounded?"

 

[busbar]

‘pete — What is the impedance of the tester used to see the imbalance? For plain-vanilla station DC, the internal resistance of a utility-grade ground-detector relay may typically be ±20kΩ. I don’t know if that correlates with an otherwise similar AC circuit. If something like a 10MΩ multimeter is used for the voltage-balance test, the just looking at the circuit crosseyed could give an undesirable reading.

With tiny {charging} currents-to-ground associated with an AC ungrounded circuit, the only positive means of detection may be taking components out {by opening circuit breakers or pulling fuses} one at a time. Else, it may be interesting to see what happens when “forcing” a lower-impedance center-tapped ground with something like two ±50kΩ resistors.

 

[electricpete]

Thx for good comments, guys.

It is ungrounded by design for reliability.

We made a measurement on an identical sister unit and found 60.3vac to ground on one leg and 59.5vac to ground on the other leg. This leads me to believe that if cable capacitances play a significant effect, they are probably similar.

The ground detection is GE NGV relay (12NGV12C2A). Instruction booklet here.

http://www.geindustrial.com/products/manuals/ngv/gei90805.pdf

We use it as an overvoltage relay set to pick up at 88 vac. Left/Right relay elements are connected from L1 to ground and from L2 to ground. Internal to the relay you see a series resistor feeding a full-wave rectifier, feeding a series reverse-biased zener diode, feeding a telephone relay. To my way of thinking the zener diode acts like a switch, when it's avalanche voltage is reached (overvoltage) the switch closes and energizes the telephone relay. When the relay is not picked up there is some leakage current through the reverse-biased zener in series with telephone coil.

I agree with all above comments that combination of cable impedance to ground in parallel with relay impedance to ground establishes the voltage divider which determines how the system floats. I'm not sure whether one or these impedance dominates over the other.

I can start to estimate the impedance of the relay. The manual (page 5) states: "The burden of the voltage unit is at unity power factor and the values are given in table....Rated Voltage: 120; Maximum Burden 4.2 watts. I know that 4 wattts at 120 volts converts roughly to 3600 ohms. I am a little unsure of several aspectss:
#1 - is this the rating of the telephone alone, or the telephone fed by the bridge and reverse-biased zener.
#2 - If fed by the bridge and particularly the zener, isn't this a highly-non-lilnear element.... should excercize caution in applying this to determine voltage divider.
#3 - Where does "unity power factor" come from? I can't imagine that for either the coil or the electronics-plus-coil.

It would be a struggle to try to put together the above in parallel with an estimate of cable capacitance. But I believe there must have been some thought put into this type of calc during establishing of the relay setpoints. Does anyone have a reference for establishing setpoint of an overvoltage relay used as ground detector in this kind of setup?

There is one other anomaly. The readings recorded as above 85+45=130. But later I looked at the installed voltmeter and saw 120. Then on the sister unit we saw roughly 60+60=120. Maybe just an inaccurate voltmeter... or maybe something else

Tomorrow we will recheck the two voltage measurements as well as the sum. Then proceed to measure the currents with what we believe are high accuracy current probes (1 millivolt per milliamp).

 

[DanDel]

'pete, if you had an actual ground fault, I believe you would have a much wider imbalance in voltages, something close to zero volts to ground on one leg, and close to full voltage on the other. Is it possible that you simply are seeing an uneven capacitive coupling due to different loads or cabling?
Did the problem just pop up, or has anything in the system been changed recently?

 

[busbar]

Looks like the NGV12C relay has two {assumed 120V} sections at ~3400 ohms/3.5mA/1.0PF each. That probably precludes hair-trigger operation, even though the relay has no intentional time delay.

A 2-miiliampere fault {with the current probe} may not be the easiest thing to track down in a 120V circuit.} It’s probably safe to assume that the published numbers would almost have to be ‘external’ values—at the test paddles.

 

[electricpete]

Dan - The problem just popped up i.e. we have had no ground alarm (relay-actuated) in recent past. The alarm came in and lasted for 30 minutes, during which time no measurements were taken.

After alarm cleared we took measurements approx 85/45vac volts identified above. More precisely the measurments were was 82/49. This is somewhat close to the setpoint ~ 88A.

I believe it is abnormal because a (sister) system measured 60/60 volts and we have had no other recent ground alarms.

I agree based upon the relatively large impedances through which the system is grounded (cable capacitance and relay) that a very "mild" (high-R) ground will cause this unbalance, and a hard ground will cause much more.

Still it is part of the philosophy of ungrounded system that we aggressively pursue grounds when they occur. The reliability of this particular system at this particular time is critically-important, so we don't want to walk away from it if we suspect possible problem.

busbar - good comments. I agree with your analysis assuming the relay impedance dominates. If I add a 3600 ohm in parallel with one of existing 3600ohm relay, I get 1800ohm to ground on one phase and 3600 to ground on the other phase giving the voltage split of 40 volts/80 volts.

120v/3600ohms ~ 33 milliamps. I am hopeful to detect one load that stands apart but time will tell.

If possible I will repeat the voltage measurements with and without the ground relay connected. This may help to establish the relative magnitude of relay vs cable impedances to ground.

By the way the relay tested sat.

 

[busbar]

'pete — You're right. My 3.5mA is incorrect. 33mA is much better, and hopefully easier to chase.

And you raise a critical point about ungrounded/high-resistance-grounded systems—you have to act on the first fault. The second one will not be so cooperative.

 

[electricpete]

You're right that 33mA is not 100% accurate. When I drew out the circuit I realized there only 40vac (not 120vac) across the ground resistance, so current would actually be 40vac/3600ohms ~ 11 milliamps.

Maybe one piece of good news is that IF I assume that a [balanced] cable capacitance makes the normal impedance to ground even lower (than 3600ohms), that means I will need an even lower fault resistance (less than 3600 ohms) to drag it down to 40/80vac. So if cable capacitance plays a role I should have more than 11 milliamps to work with.

But I'm still a little wary of the consequences of the assumption that this highly non-linear circuit (particularly the zener) can be effectively represented an impedance.

Any comments? I realize it is getting a little carried away with calcs in absence of firm (measured) data... but it will be very beneficial for me to understand this particular issue as fully as I can.

 

[electricpete]

Whoops, I'm wrong again. The 4 watts at 120v is listed as max power draw. It likely applies to the relay in the pulled-in position with sensed high-voltage to ground and relay alarming. Current during operation or soft-grounded operation below setpoint (our current condition) will be much lower.

 

[electricpete]

Will measure the relay current directly.

 

[busbar]

A change in relay burden is not discssused in the IB, but may be a valid concern for sensitve ground detection.

 

[electricpete]

Went in yesterday and got some interesting results.

L1-G: 56
L2-G: 67.5
L1-L2: 119.6
L1-L2 was higher than expected since it exceeds (L1/G+L2/G) by about 4 volts. We repeated several times at several measurement points and got the same thing. (Must be either phase angle difference in the voltages or possibly effect of harmonics.)

L1 relay current - 0.188 amps
L2 relay current 0.190 amps
(the above readings seemed steady and repeatable).

Based on above, we expected possibly 0.002 amps or a little more as indication of ground. Our readings on the individual loads were too unsteady and non-repeatable to identify this.

Based on the decrease in voltage unbalance from previous readings, it looks like the condition is getting much better. However we are still very concerned about the problem since we suspect it may come back. We have been asked to do a thorough investigation and make preparations for troubleshooting if it returns. We are also limited in the number of measurements allowed.... it makes the operations crew uneasy when we go near this equipment (that is a reflection of the importance of the equipment, not their confidence in the troubleshooting team... I hope).

The fact that the apparent ground is getting better may suggest a problem like a ground due to moisture. Some of the loads are in semi-outdoor areas. By the way the weather was wet at the time of the original alarm, has dried up since then.

I tend to think this rules out cable capacitance. Also would seem to rule out change in relay impedance.

I was very surprised to see the relay current readings so high. I conclude it must likely be the capacitor connected in series with the relays which is shown in the drawings, although no capacitance value is provided.

The capacitive reactance must be 120vac/0.2A = 600ohms. Using similar logic as before, the ground that we had when we measured 80/40 must have been even less that 600 ohms (parallel combination of the 600-ohm capacitive-reactance and the unknown ground resistance must be 300).

But now one fly in the ointment.... if the relay can be represented as a simple capacitance, then why didn't we see a current imbalance roughly proportional to voltage-to-ground at the time of the measurement (56/67v)?

I'm very interested to hear any and all thoughts along the lines of troubleshooting (upon recurrence) and analysing the events.... bearing in mind that isolating loads is not an option we are prepared for yet.

 

[busbar]

‘pete — In the linked GE instruction booklet page 14 figure 12 seems to show the 12NGV12C not having capacitors. Am I missing something? [The IB list a unity-PF burden, presumably for all NGVs, but shows no capacitors in the -12C, unlike the -12A and -12B variations.]

Agreed that in this case the “two-hots-to-ground” relay burden is likely a significantly large portion of the circuit monitored by the relay itself.

 

[electricpete]

Very good observation, busbar. The panel vendor provided us with drawings showing the internal of the relay which shows caps installed. I have checked the part/number above and it is what we have installed and as you say the manual shows it shouldn't have caps. Raises more questions.

 

[huntercon]

electricpete: I used to work in utility relay engineering, and recognize this GE relay as a very old EN type and most likely obsolete.

If you intend to keep your system isolated (ungrounded) for reliability, you may want to consider 'monitoring' for grounds vs. 'detecting' them. If so, I recommend you toss the old GE fossil and install an Insulation Monitoring Device (IMD).

You can do an Internet search for 'ground fault' and find a few serious suppliers of IMDs. My preferrence is Bender, Inc. - they have a nice little unit for 120VAC ungrounded systems:

http://www.bender.org

 

[electricpete]

huntercon - when you say "monitor" vs "detect".... are you talking about displaying a value of insulation resistance in addition to go/no-go alarm contacts?

That would make pretty good sense to me at , given the amount of time we spend chasing grounds. For one thing we will learn what is normal and possibly identify degrading conditions before they get to the alarm point (perhaps we'll learn our alarm point is not appropriate). And we'll be able to distinguish between the severity of the grounds.

Do you know off-hand if it uses the same principle as the old relay (passively sensing voltage between each phase and ground) or new principle (applying test dc voltage)?

(I downloaded a data-sheet but having a problem opening the pdf at the moment.)