Ground Fault, Yea!!!!

Status
Not open for further replies.
N

nhfire77

Senior Member
Location
NH
I do a lot of troubleshooting. A LOT. But, I haven't had to deal this specific problem in a while. I was hoping someone would have a better mouse trap as it were.

Situation:

300,000 square foot steel building. Silent Knight IFP-100 (a.k.a Silent knight 5808)

About 110 devices on the SLC, mostly Addressable Input modules, some Mini input Modules and Relay modules.

SLC class B and T tapped at the panel, in other words there are 4 legs of the SLC home run to the FACP.

Almost half of the devices are on the "bad" leg.

All FPL above the bar joist. FPL in EMT down to pull points and devices. Some of the monitored contact wiring are underground outside to monitor a fire suppression water tank. The Addressable input module is IN the building. This is in PVC, with a UG rated FA cable of some sort.(I'm not sure is 100% code compliant, and we'll deal with that later.)

When metering the one side of the leg that I believe is the problem, no ground fault is detected (about 26 Meg Ohm actually, the other three legs are similar in readings not causing a GF, the threshold for GF in the panel is 40K Ohm)

When it is attached to the FACP, the GF condition returns. However, it randomly goes away, so right now its normal.

When this occurs and the GF condition is present on the panel but not on my meter, typically the GF is on the monitored side of a addressable input module (kinda like the load side of a circuit, but not.)

No exact drawings exist.

Since the GF is transient, do I just resign myself to opening every box that has a device in it on the bad SLC leg?

Or would you have a better idea?
 
Last edited:
nhfire77 said:
I do a lot of troubleshooting. A LOT. But, I haven't had to deal this specific problem in a while. I was hoping someone would have a better mouse trap as it were.

Situation:

300,000 square foot steel building. Silent Knight IFP-100 (a.k.a Silent knight 5808)

About 110 devices on the SLC, mostly Addressable Input modules, some Mini input Modules and Relay modules.

SLC class B and T tapped at the panel, in other words there are 4 legs of the SLC home run to the FACP.

Almost half of the devices are on the "bad" leg.

All FPL above the bar joist. FPL in EMT down to pull points and devices. Some of the monitored contact wiring are underground outside to monitor a fire suppression water tank. The Addressable input module is IN the building. This is in PVC, with a UG rated FA cable of some sort.(I'm not sure is 100% code compliant, and we'll deal with that later.)

When metering the one side of the leg that I believe is the problem, no ground fault is detected (about 26 Meg Ohm actually, the other three legs are similar in readings not causing a GF, the threshold for GF in the panel is 40K Ohm)

When it is attached to the FACP, the GF condition returns. However, it randomly goes away, so right now its normal.

When this occurs and the GF condition is present on the panel but not on my meter, typically the GF is on the monitored side of a addressable input module (kinda like the load side of a circuit, but not.)

No exact drawings exist.

Since the GF is transient, do I just resign myself to opening every box that has a device in it on the bad SLC leg?

Or would you have a better idea?
Click to expand...

If the entire loop goes down when the fault appears, you might consider breaking up the loop with line isolation modules to narrow the fault location. See which devices still report in when the ground fault shows up. More than once I was sure I knew the GF location, only to find the true source elswhere. I also have a very vague recollection that the impedance (high vs. low) of your meter can make the GF "disappear", but I could be wrong.
 
gadfly56 said:
If the entire loop goes down when the fault appears, you might consider breaking up the loop with line isolation modules to narrow the fault location. See which devices still report in when the ground fault shows up. More than once I was sure I knew the GF location, only to find the true source elswhere. I also have a very vague recollection that the impedance (high vs. low) of your meter can make the GF "disappear", but I could be wrong.
Click to expand...

the loop is not going down in GF. Its just being recognized by the panel, No other troubles exist. Good point with the LIM if I ever had that happen.

BTW, I realize that I could have a bad panel, but in this case its most likely the wiring as I have already corrected 4 GF at the ceiling level (zip ties too tight, way too tight)
 
It might pay to pull the mother board out of the panel and inspect for arcing damage. I found one this way in a church that took a lightning strike, worked fine with the loop disconnected, but under load it would show up as a GF. Free wired a new device at the panel, GF returned. Pulled the mother board, could see traces of a flash over.
 
First, a thumbs-up to Hillbilly, but I wouldn't pull a panel's board out as my first step unless it failed the sniff test. I prefer to troubleshoot grounds using an analog meter. I've been able to read continuity to ground on analog meters that were invisible to digital meters.

If you don't have an analog meter, use the voltage scale on your digital multimeter:

Power down your system, and remove 1 battery entirely from any battery connection.
Connect a jumper between the negative terminal of this battery and ground.
Separate the t-taps.
Look for voltage between the positive terminal of the battery and each conductor in question. No voltage reading = no ground fault OR the fault is intermittent. HTH.
 
The method I described in my last reply has been useful in finding water related grounds on addressable systems especially. I mention this after having found water getting through the jackets of FPL cabling over long periods of time (in conduit buried in slabs). Unless it's jacket has been cut or abraded, your UG cable should be ok.
 
Last edited:
MichaelGP3 said:
The method I described in my last reply has been useful in finding water related grounds on addressable systems especially. I mention this after having found water getting through the jackets of FPL cabling over long periods of time (in conduit buried in slabs). Unless it's jacket has been cut or abraded, your UG cable should be ok.
Click to expand...

Well,

I just spent 5 hours on it. When they pulled in the UG cable they left the classic burn in marks on the threads of the stub ups in to boxes, Obviously it was not pulled in safely, and no bushings. But still not GF on the panel. I found a couple of other problems, gaskets missing from T bodies and a special room for the water tank that was 160F at the ceiling!!!(at 9am) There were two resistance plate heaters bolted to the floor, about 4" above that was a steel plate guard. I touched it, not knowing the heater was under it. Now I have burns on my fingers, becuase it was so hot. There are Tyco butterfly sprinkler valves in there. The valves oil was so over heated it was leaking out into the J box for the tamper wiring, but it didn't appear to be the cause. Also, turns out there are leaks in the roof and walls in a few key areas that have devices mounted there. Well, not wet, but it did rain, then the GF occured Now its 100F out side at 1pm and im sure its dried up if it was the cause.

And Still no definitive answer. I will check the board closer, but since its a transient GF that comes and goes when ever I am not there, its unlikely.
 
nhfire77 said:
Well,

I just spent 5 hours on it. When they pulled in the UG cable they left the classic burn in marks on the threads of the stub ups in to boxes, Obviously it was not pulled in safely, and no bushings. But still not GF on the panel. I found a couple of other problems, gaskets missing from T bodies and a special room for the water tank that was 160F at the ceiling!!!(at 9am) There were two resistance plate heaters bolted to the floor, about 4" above that was a steel plate guard. I touched it, not knowing the heater was under it. Now I have burns on my fingers, becuase it was so hot. There are Tyco butterfly sprinkler valves in there. The valves oil was so over heated it was leaking out into the J box for the tamper wiring, but it didn't appear to be the cause. Also, turns out there are leaks in the roof and walls in a few key areas that have devices mounted there. Well, not wet, but it did rain, then the GF occured Now its 100F out side at 1pm and im sure its dried up if it was the cause.

And Still no definitive answer. I will check the board closer, but since its a transient GF that comes and goes when ever I am not there, its unlikely.
Click to expand...

If you think it's the switch side of the monitor module, have you tried taking a module you know is good and creating a deliberate fault to ground on the switch legs when the usual ground fault is absent? This would tell you if you're getting the same type fault report as with your "gremlin".
 
100901-0811 EST

I do not know anything specific about your system. So the following is a guess on the system design.

Assuming the normal circuit consists of a balanced two wire communication path with both wires floated off of ground, then to detect a ground fault it would be necessary to apply a common mode voltage to the pair of wires relative to ground, and measure the current from this voltage to detect a ground fault.

Assume this is a RS422 or RS485 system, then a DC bias of maybe +5 referenced to ground is applied thru a pair of 56 ohm resistors to the two signal wires. A current sensor monitors the current to the midpoint of the two 56 ohm resistors. A 40 K ground fault will cause a current of 5/40,000 = 0.000125 A to flow. This is 125 microamps. The average voltage of the signal wires will not change much. 125 microamps thru 56 ohms is a drop of 7 millivolts. So the change would be from 5.000 V to 4.993 V.

Change the 56 ohm resistors to 80,000 ohms and assume there is a 120 ohm terminating resistor at the far end of the transmission line. Now there is a series resistor of 40,000 from the +5 V source to the wire pair. A 40,000 ohm short to ground will cause the average voltage of the transmission line to drop from 5 V to 2.5 volts. So a voltage threshold detector instead of a current sensor could be used to detect a short to ground. This is probably the simpler and lower cost method.

Suppose the second method is used. You can test the system to see what is the likely test method.

First, with no changes to the system measure the DC voltage from each of the communication path wires to ground with a high input impedance meter, a Fluke 27, 87, or similar. If you see some likely voltage, maybe anything from 0.1 V to 24 V, then put a 10,000 ohm resistor across the voltmeter terminals and record the voltage before and after the shunt resistor. If you see a substantial % change in voltage, then you can assume a current limiting resistor or constant current source is part of the detection scheme.

Second, if the current limit (resistor or other) is the detection method, then you could put a high impedance input recording meter on the circuit, and monitor to see if an actual fault is causing the system to indicate a fault. The simplest recorder is a Fluke 27 in min-max mode, but this provides no timing information.

.
 
gar said:
100901-0811 EST

I do not know anything specific about your system. So the following is a guess on the system design.

Assuming the normal circuit consists of a balanced two wire communication path with both wires floated off of ground, then to detect a ground fault it would be necessary to apply a common mode voltage to the pair of wires relative to ground, and measure the current from this voltage to detect a ground fault.



.
Click to expand...

Its this. It is Signaling Line Circuit, Floated off of ground. It is 32 VDC nominal, 150mA.

The voltage fluctuates slightly with data tx/rx,(50 tx-rx/sec +/-) I don't get what you mean.
 
100901-1318 EST

nhfire77:

I need more information on your signaling circuit.
Are there two wires that go to your various detectors or sensors?
Is one of the two wires grounded at the monitor equipment?
Or neither grounded?
At the far end of the two wires is there a terminating resistance?
How is power supplied to any sensors that may have electronics that requires power?

Is there the possibility this is a current loop system?

.
 
gar said:
100901-1318 EST

nhfire77:

I need more information on your signaling circuit.
Are there two wires that go to your various detectors or sensors?
Is one of the two wires grounded at the monitor equipment?
Or neither grounded?
At the far end of the two wires is there a terminating resistance?
How is power supplied to any sensors that may have electronics that requires power?

Is there the possibility this is a current loop system?

.
Click to expand...

The signal line circuit (SLC) is a two-wire loop.

I do not believe that either wire is grounded at the panel since the panel will only clear when the wire causing the fault is lifted and does not need the pair to be lifted to do so.

There is no terminal resistor or other end-of-line device.

For monitor modules the standby current is 0.55mA, first alarm is 23 mA, two alarms is 46mA and three and up are 0.55mA each additional. Smoke detectors (photo) are 0.55mA in standby or alarm.

Any other devices that send power to notification appliances, sounder bases, relays for duct detector housings, etc require an auxilliary power source.
 
I had a similar issue once with an old analog system except mine went in to alarm.

The FPL cable was run above the bar joists in the hollows of the decking.
It was being pinched between the deck and bar joist. The alarm would sound when somebody walked on the roof in that one spot or when the roof heated then cooled.

It was by chance I found it. I just happend to on the ladder in that vicinity when an AC tech walked across the roof and the alarm went off.:roll:
 
100901-1954 EST

Thanks gadfly56.

On the basis of your description I will simplify the circuit to an equivalent.

Basically I do not think we care what is happening on the two signaling wires relative to signaling. Suppose there is a battery for the signaling wires with a small resistor for current sensing. These all float off of ground.

I am going to assume the voltages discussed are DC, but the system could be built with AC voltages.

I do not know if the 32 V mentioned by nhfire77 is between these two wires, or from one of the wires to ground. For the moment assume it is from one of the wires to ground, and maybe there is 5 V between the signaling wires. I do expect that there is a significant voltage between the wires and ground. This is what I want measured. For the moment pick on just one of the wires.

With a high impedance voltmeter between said one signal wire and ground measure the voltage. Then place 10,000 ohms across the voltmeter input. This should be a 1/2 W or 1 W resistor. With the resistor present what is the meter reading. If the voltage with the resistor drops to about 1/5 of the voltage without the resistor, then we can assume that internally a 40,000 ohm resistor is the current limiter for ground fault detection.

To visualize the rough circuit I am describing -- draw a ground wire connected to the negative end of a 32 V battery. Separately draw a 1.5 V battery connected to a pair of signaling wires. From the negative end of the 1.5 V battery draw a 40,000 ohm resistor to the positive end of the 32 V battery. Call the signaling wire connected to the negative end of the 1.5 V battery wire A. The other signaling wire is B.

Place the meter between ground and wire A. because it is a high input impedance it should read close to 32 V. Next put the 10,000 ohm resistor between wire A and ground. Now what does the meter read? Do the calculation.

.
 
gar said:
100901-1954 EST

Thanks gadfly56.

On the basis of your description I will simplify the circuit to an equivalent.

Basically I do not think we care what is happening on the two signaling wires relative to signaling. Suppose there is a battery for the signaling wires with a small resistor for current sensing. These all float off of ground.

I am going to assume the voltages discussed are DC, but the system could be built with AC voltages.

I do not know if the 32 V mentioned by nhfire77 is between these two wires, or from one of the wires to ground. For the moment assume it is from one of the wires to ground, and maybe there is 5 V between the signaling wires. I do expect that there is a significant voltage between the wires and ground. This is what I want measured. For the moment pick on just one of the wires.



.
Click to expand...


The only AC voltage is the main input to the board. Everything in the system is DC 24-32 VDC.

NO circuits are connected to ground except the AC power EGC, as this is the sole point of reference to ground for the entire system. There are sub panels that have their own AC power feed. They are not directly connected to the main system, as they are monitored by input devices. The sub panels have their own internal GF monitoring. If, they were in GF it would be annunciated separately as a distinct trouble. None of the sub panels have any connection to ground (other than the EGC)

I will now go back a reread over what you posted, let my brain cool down and hope to get it.

I do appreciate learning new things, keep it coming.
 
Last edited:
shepelec said:
I had a similar issue once with an old analog system except mine went in to alarm.

The FPL cable was run above the bar joists in the hollows of the decking.
It was being pinched between the deck and bar joist. The alarm would sound when somebody walked on the roof in that one spot or when the roof heated then cooled.

It was by chance I found it. I just happend to on the ladder in that vicinity when an AC tech walked across the roof and the alarm went off.:roll:
Click to expand...

I have already corrected 3 GF at ceiling level on NAC's, similar to what you are referring. It very well could be at the ceiling level and a wire rubbing on a bar joist, but the problem came after it rained and went away after it dried up.
 
100902-0800 EST

nhfire77:

Your last post would imply that the problem is in the signal wiring or sensing devices external to the monitoring panel. Therefore, you should not be changing any of those monitoring boards at this time.

In your second to last post your following comment needs clarification:
NO circuits are connected to ground except the AC power EGC, as this is the sole point of reference to ground for the entire system. There are sub panels that have their own AC power feed. They are not directly connected to the main system, as they are monitored by input devices. The sub panels have their own internal GF monitoring. If, they were in GF it would be annunciated separately as a distinct trouble. None of the sub panels have any connection to ground (other than the EGC)
Click to expand...
You say no circuits except AC and EGC are connected to ground. If none of the internal circuitry was connected to ground, then you could not detect a grounding of the signaling wiring. To detect a ground there has to be circuitry between ground (that would be the EGC) and the signaling wires.

I am conjecturing that there is a voltage source and a current limiting resistance between the signaling wires and the enclosure and therefore the EGC, and that when the voltage drop across the current limiting resistor (current flowing thru this resistor means there is some degree of ground fault) exceeds a certain value this triggers the ground fault alarm. To prove whether this theory is correct is the reason to measure the voltage from a signal wire to ground and then introduce a known fault to ground to see what happens. The reason for not using a direct short is that more information can be obtained about the system.

If my conjecture is correct on the means of detecting a ground fault, then a recording voltmeter can be used to determine the time of a fault or faults and an estimate of the magnitude of the fault.

Whether having this information will assist troubleshooting or not is not known, but it may help.

Also using two people and just a voltmeter you might be able to mechanically perturb the wiring and narrow down the likely location of the problem.

If there is a high correlation with rain, then look for ways water from rain might get into the wiring or the sensors.

.
 
gar said:
100902-0800 EST

nhfire77:

Your last post would imply that the problem is in the signal wiring or sensing devices external to the monitoring panel. Therefore, you should not be changing any of those monitoring boards at this time.

In your second to last post your following comment needs clarification:You say no circuits except AC and EGC are connected to ground. If none of the internal circuitry was connected to ground, then you could not detect a grounding of the signaling wiring. To detect a ground there has to be circuitry between ground (that would be the EGC) and the signaling wires.

I am conjecturing that there is a voltage source and a current limiting resistance between the signaling wires and the enclosure and therefore the EGC, and that when the voltage drop across the current limiting resistor (current flowing thru this resistor means there is some degree of ground fault) exceeds a certain value this triggers the ground fault alarm. To prove whether this theory is correct is the reason to measure the voltage from a signal wire to ground and then introduce a known fault to ground to see what happens. The reason for not using a direct short is that more information can be obtained about the system.

If my conjecture is correct on the means of detecting a ground fault, then a recording voltmeter can be used to determine the time of a fault or faults and an estimate of the magnitude of the fault.

Whether having this information will assist troubleshooting or not is not known, but it may help.

Also using two people and just a voltmeter you might be able to mechanically perturb the wiring and narrow down the likely location of the problem.

If there is a high correlation with rain, then look for ways water from rain might get into the wiring or the sensors.

.
Click to expand...

I believe what you state is the correct way the panel detects GF. What I meant was, none of the field wiring is grounded, and there is no shielded cable in use, nor is it required for the panel.

I have quite a bit of wiring outside but can only inspect the conduit, when above ground and open all the device boxes, which I did. Some of the UG wiring was wet at a pull point 4' AFF.

Shaking wiring does work, but there is 10's of thousands of feet of cabling throughout the building.

Oh and I don't need a data logging DMM for the time component. The panel reports the ground fault to an alarm receiver, that is a device that it calls with a modem and logs the time of when the GF is detected and when it goes away, along with all other activity for the system.

Also, the GF has not returned. However, the Hurricane will be dumping some rain in my area. I plan on going to the site the moment the panel reports a GF on the weekend. We'll see what happens.
 
Last edited:
100902-1313 EST

nhfire77:

The amount of wire you describe presents a real problem.

From your earlier comment did you measure the voltage from a signal wire to ground at about 32 V? If so then was it about the same from both wires to ground or substantially different values?

Do you have access to any resistors? Some standard values are 10 K, 15 K, 22 K, 27 K, 33 K, 39 K, and 47 K. With a 10 K and a 39 K you can pretty well estimate if the ground fault circuit has a 40 K internal resistance. If the internal resistance of the ground fault circuit is known, then a voltage measurement from a signal wire to ground will allow you to estimate the magnitude of a shunt to ground.

If when an actual ground fault occurs the ground to signal line voltage goes to near zero, then you have close to a dead short. If it was about 1/2 of the non-fault condition voltage, then the fault is approximately equal to the internal resistance of the circuit. If wetness is the sole cause, then it probably will not look like a dead short. Hopefully the fault is more than momentary. You may need to break the wiring in a binary fashion to find the region where the short is. This means break the cable near the middle. Is the fault still present? This determines where to make the next break point. Repeat until you find the problem. Two or more different locations might require repeating the process.

.
 
gar said:
100902-1313 EST

nhfire77:

The amount of wire you describe presents a real problem.

From your earlier comment did you measure the voltage from a signal wire to ground at about 32 V? If so then was it about the same from both wires to ground or substantially different values?

Do you have access to any resistors? Some standard values are 10 K, 15 K, 22 K, 27 K, 33 K, 39 K, and 47 K. With a 10 K and a 39 K you can pretty well estimate if the ground fault circuit has a 40 K internal resistance. If the internal resistance of the ground fault circuit is known, then a voltage measurement from a signal wire to ground will allow you to estimate the magnitude of a shunt to ground.

If when an actual ground fault occurs the ground to signal line voltage goes to near zero, then you have close to a dead short. If it was about 1/2 of the non-fault condition voltage, then the fault is approximately equal to the internal resistance of the circuit. If wetness is the sole cause, then it probably will not look like a dead short. Hopefully the fault is more than momentary. You may need to break the wiring in a binary fashion to find the region where the short is. This means break the cable near the middle. Is the fault still present? This determines where to make the next break point. Repeat until you find the problem. Two or more different locations might require repeating the process.

.
Click to expand...

Gar,

Thank you for your ideas. Some of it is new to me and I appreciate it. I have extensive experience of troubleshooting GF. The Actual data loop isn't 10,000 feet, but all cable at the ceiling level is white, including the fire alarm data circuits and all the other. So its a huge pain to trace down.

What I was getting at in my first fews posts was that I believe that I determined the GF was on the negative side of the data loop. That was as far as I got before it disappeared. I didn't think of measuring the voltage difference to ground, that might be helpful when I get closer to it.
 
Status
Not open for further replies.
Top