Ground Leakage through Snow Melting Mat

[RickLosi]

My company supplied Roof Heating Mats (for snow melting) on a project,

Backgound:
Heaters are constructed of a Thin-Film Polyethylene encapsulating a pair of flat bus-wires. Between the flat bus-wires is a Printed Conductive ink. The heaters produce heat by voltage flowing through the resistive ink and generating heat. The polyethylene is non conductive and terminations are made to the bus-wires then encapsulated in a layer of self-vulcanized insulation. After this the heaters are then sandwiched between two layers of EPDM and the edges are heat sealed to make a waterproof mat with a cold lead sticking out for power.

Heaters are 208 vac rated and installed on an EPDM Roof with adhesive.

Electrical Contractor (EC) noticed that when the mats are wet, and if touched while holding a guard rail, there is definitely a flow of electricity. When a meter is used, it is measuring 110 vac to ground on the outside of the mat (a t-stat is breaking one hot leg of the circuit when not calling for heat). When the T-stat calls for heat, the voltage to ground drops by 50% to 55 Vac.

Manufacturer visited site and felt that there was a manufacturing defect. MFG then sent a whole new set of mats. The originals were taken out and replaced by the new set. The same thing is happening again. This leads me to believe that the mats may not be the cause (These mats are installed in thousands of applications) with zero problems).

The EC did say that they were using the T-stat to break two hot legs from two Circuit breakers. I advised him not to do this.

In the meantime, I have sent a new mat from another manufacturer to the site to see if we get the same leakage, which will rule out the mats and point directly to the building, wiring etc.

Any thoughts on why ground leakage is occurring? I am doubtful it is being caused by the heater mats.

 

[LarryFine]

Welcome to the forum.

Any roofing nails through the mat?

 

[Krusscher]

Could be a screw through a conduit/wire near the roof. Maybe put the mats on a GFPE breaker and see if it trips.

 

[Krusscher]

does the voltage between the two go away when you shut off the breaker to the mat?

 

[RickLosi]

Welcome to the forum.

Any roofing nails through the mat?

No nails, no clamps. Mats are applied with mastic.
Mats are on GFEPD CB's (30 ma trip). No tripping.

Very unlikely that two sets of heaters are having the same exact problem. There are multiple heaters in the system (29). For the problem to be mechanical abuse (Nails, punctures, etc) it would have to occur on each mat installed, (58) total.

EC has isolated each mat and problem remains, new mats and old mats.

Somewhat puzzling

 

[RickLosi]

Welcome to the forum.

Any roofing nails through the mat?

No nails, no clamps. Mats are applied with mastic.
Mats are on GFEPD CB's (30 ma trip). No tripping.

Very unlikely that two sets of heaters are having the same exact problem. There are multiple heaters in the system (29). For the problem to be mechanical abuse (Nails, punctures, etc) it would have to occur on each mat installed, (58) total.

EC has isolated each mat and problem remains, new mats and old mats.

Somewhat puzzling

 

[RickLosi]

does the voltage between the two go away when you shut off the breaker to the mat?

EC said that there was no leakage prior to energizing the CB's. Although they did not test, EC referenced other workers in the same location and no one complained about getting shocked.

 

[RickLosi]

One of my tech's will be onsite when new mat is tested.
I will forward recommendations to test/isolate system to locate problem origin.

 

[RickLosi]

Why would voltage to ground drop 50% when circuit is energized? This is consistent on all mats,

 

[Krusscher]

If there was any recent roofing work done I would suspect a nail or screw went through a wire or conduit below the decking.

 

[synchro]

Electrical Contractor (EC) noticed that when the mats are wet, and if touched while holding a guard rail, there is definitely a flow of electricity. When a meter is used, it is measuring 110 vac to ground on the outside of the mat (a t-stat is breaking one hot leg of the circuit when not calling for heat). When the T-stat calls for heat, the voltage to ground drops by 50% to 55 Vac.

It's possible that the measured voltage is due to the capacitance between each of the flat bus wires and the water on the surface of the mats. If a high impedance meter is used, it would draw a miniscule current and therefore it would only contribute a very small voltage drop across this capacitance. If you open circuit one of the hot legs, then the capacitance from its flat bus wire to the water on the surface will open be circuited (i.e, floating). But the other hot leg will still have 120V to ground, and so it can capacitively couple and cause the 110 Vac that you are measuring. There will be some amount of current leakage off the surface of the wet mat to its surroundings, and so that could account for the 10V drop from 120V down to the 110V you observed.

If the mats are powered by two legs of a 208/120V system, then L1-N and L2-N will be 120° apart. Each flat bus wire is connected to one leg and has its own capacitance to the water on the surface, and I assume that these two capacitances would be about equal from what was described. Therefore these two capacitances will form a capacitive divider when both legs are connected. The portions (i.e., components) of the L1-N and L2-N voltages that are in-phase with each other are cos(60°) = 1/2 times the magnitude of the L-N voltage. The out-of-phase components would cancel. And so we'd expect 1/2 x 120V = 60V to neutral (or ground) on the output "terminal"of the capacitive divider, which is the wet surface in this case. And an observed drop down to 50V to 55V could be due to some current leakage over to the mat's surroundings as mentioned in the paragraph above.

If the above analysis is correct, it would seem that if the mats were powered L-L from 120/240 single phase (or dropped down from 240 to 208 with a buck-boost), and a contactor was used to open both legs together, then the voltage coupled to the mat surface might be minimized. That's because the L1-N and L2-N are 180° apart and have no in-phase component, and so the currents they each capacitively couple to the wet surface would cancel. But you'd still need to open both legs to turn the heaters off, otherwise one leg would still be coupling 120V to the wet surface through its capacitance.

Although it would still be objectionable, it's possible that what you are observing would not be life threatening from an electrical shock if the current is sufficiently low. I could envision some experiments, for example, putting a small patch of aluminum foil over the wet mat and measuring the current from it to neutral or ground through a suitable meter. That would at least give you an idea about what you're up against. Just measuring current by placing a meter probe on the wet surface would not be adequate because of its small contact area when compared to, say, a human hand. I would first measure the voltage with a low impedance meter to make sure the voltage drops close to zero, in case a nail or similar connection to one of the hot conductors might be present.

Are these mats powered through a GFPE or GFCI?

 

[ptonsparky]

Post 6 indicates 30ma protection.
Definitely break both legs to mats.

 

[TwoBlocked]

Could whatever is being used as a ground reference, actually be energized?

 

[zbang]

if touched while holding a guard rail,

Is the guard rail bonded? Assuming it is, I'd go for capacitive coupling if a metal test lead touching a damp but otherwise non-conductive surface.
Also, is this being tested with a high-impedance meter? I'm guessing it is.

 

[GoldDigger]

It's possible that the measured voltage is due to the capacitance between each of the flat bus wires and the water on the surface of the mats. If a high impedance meter is used, it would draw a miniscule current and therefore it would only contribute a very small voltage drop across this capacitance. If you open circuit one of the hot legs, then the capacitance from its flat bus wire to the water on the surface will open be circuited (i.e, floating). But the other hot leg will still have 120V to ground, and so it can capacitively couple and cause the 110 Vac that you are measuring. There will be some amount of current leakage off the surface of the wet mat to its surroundings, and so that could account for the 10V drop from 120V down to the 110V you observed.

If the mats are powered by two legs of a 208/120V system, then L1-N and L2-N will be 120° apart. Each flat bus wire is connected to one leg and has its own capacitance to the water on the surface, and I assume that these two capacitances would be about equal from what was described. Therefore these two capacitances will form a capacitive divider when both legs are connected. The portions (i.e., components) of the L1-N and L2-N voltages that are in-phase with each other are cos(60°) = 1/2 times the magnitude of the L-N voltage. The out-of-phase components would cancel. And so we'd expect 1/2 x 120V = 60V to neutral (or ground) on the output "terminal"of the capacitive divider, which is the wet surface in this case. And an observed drop down to 50V to 55V could be due to some current leakage over to the mat's surroundings as mentioned in the paragraph above.

If the above analysis is correct, it would seem that if the mats were powered L-L from 120/240 single phase (or dropped down from 240 to 208 with a buck-boost), and a contactor was used to open both legs together, then the voltage coupled to the mat surface might be minimized. That's because the L1-N and L2-N are 180° apart and have no in-phase component, and so the currents they each capacitively couple to the wet surface would cancel. But you'd still need to open both legs to turn the heaters off, otherwise one leg would still be coupling 120V to the wet surface through its capacitance.

Although it would still be objectionable, it's possible that what you are observing would not be life threatening from an electrical shock if the current is sufficiently low. I could envision some experiments, for example, putting a small patch of aluminum foil over the wet mat and measuring the current from it to neutral or ground through a suitable meter. That would at least give you an idea about what you're up against. Just measuring current by placing a meter probe on the wet surface would not be adequate because of its small contact area when compared to, say, a human hand. I would first measure the voltage with a low impedance meter to make sure the voltage drops close to zero, in case a nail or similar connection to one of the hot conductors might be present.

Are these mats powered through a GFPE or GFCI?

Based on the physical layout described by the OP, feeding the mat from a balanced source should indeed cancel out the capacitive voltage/current on the outside surface of the mat. But the effectiveness of that would be less as the spacing between the rails is increased. OP did not state the spacing.

 

[RickLosi]

It's possible that the measured voltage is due to the capacitance between each of the flat bus wires and the water on the surface of the mats. If a high impedance meter is used, it would draw a miniscule current and therefore it would only contribute a very small voltage drop across this capacitance. If you open circuit one of the hot legs, then the capacitance from its flat bus wire to the water on the surface will open be circuited (i.e, floating). But the other hot leg will still have 120V to ground, and so it can capacitively couple and cause the 110 Vac that you are measuring. There will be some amount of current leakage off the surface of the wet mat to its surroundings, and so that could account for the 10V drop from 120V down to the 110V you observed.

If the mats are powered by two legs of a 208/120V system, then L1-N and L2-N will be 120° apart. Each flat bus wire is connected to one leg and has its own capacitance to the water on the surface, and I assume that these two capacitances would be about equal from what was described. Therefore these two capacitances will form a capacitive divider when both legs are connected. The portions (i.e., components) of the L1-N and L2-N voltages that are in-phase with each other are cos(60°) = 1/2 times the magnitude of the L-N voltage. The out-of-phase components would cancel. And so we'd expect 1/2 x 120V = 60V to neutral (or ground) on the output "terminal"of the capacitive divider, which is the wet surface in this case. And an observed drop down to 50V to 55V could be due to some current leakage over to the mat's surroundings as mentioned in the paragraph above.

If the above analysis is correct, it would seem that if the mats were powered L-L from 120/240 single phase (or dropped down from 240 to 208 with a buck-boost), and a contactor was used to open both legs together, then the voltage coupled to the mat surface might be minimized. That's because the L1-N and L2-N are 180° apart and have no in-phase component, and so the currents they each capacitively couple to the wet surface would cancel. But you'd still need to open both legs to turn the heaters off, otherwise one leg would still be coupling 120V to the wet surface through its capacitance.

Although it would still be objectionable, it's possible that what you are observing would not be life threatening from an electrical shock if the current is sufficiently low. I could envision some experiments, for example, putting a small patch of aluminum foil over the wet mat and measuring the current from it to neutral or ground through a suitable meter. That would at least give you an idea about what you're up against. Just measuring current by placing a meter probe on the wet surface would not be adequate because of its small contact area when compared to, say, a human hand. I would first measure the voltage with a low impedance meter to make sure the voltage drops close to zero, in case a nail or similar connection to one of the hot conductors might be present.

Are these mats powered through a GFPE or GFCI?

2-pole 208 vac GFEPD (30 ma GF Trip) Circuit Breaker.

Would the current be below the 30 ma GF trip level? I was on the phone with EC and he said when mats are energized they all eventually trip the GFEPD.

I have been discussing problem with mfg, and they produce hundreds of these mats monthly. From first installation to replacement mats being installed, they sold over 150 mats. The mfg reported that not one customer called with this problem. If the above hypothesis describes what is happening, wouldn't this be the problem for all of the equipment from this mfg assuming his manufacturing is consistent from heater to heater?

I am on site Tuesday and will run check to see if guard rail is the cause. EC never tested a wet roof to the guard rail. For a GFEPD to trip the imbalance must be in the circuit, not sure if an improperly grounded guard rail can cause a GFEPD to trip in a 208/120 3 ph 4 wire system, unless it is somehow back-feeding the ground bus in the panel. Even then that would be a stretch. I have seen it happen on 480/277 3 ph 4 wire systems where a bad ground was causing GFEPD trips (Neutral bonded to ground) but that is because the current was back-feeding the neutral running through the Current Transformer.

There is a considerable amount of rooftop HVAC units in the vicinity. May investigate if they are not properly grounded

I feel that we need to test mats in a controlled environment to prove that it is not the mat but something electrically wrong with the power/feeds.

More to come, I appreciate your input
Thanks

 

[RickLosi]

Could whatever is being used as a ground reference, actually be energized?

I am on site Tuesday and will run check to see if guard rail is the cause. EC never tested a wet roof to the guard rail.

 

[RickLosi]

Based on the physical layout described by the OP, feeding the mat from a balanced source should indeed cancel out the capacitive voltage/current on the outside surface of the mat. But the effectiveness of that would be less as the spacing between the rails is increased. OP did not state the spacing.

Bus wire/strips are 18" apart in the heater.

 

[ruxton.stanislaw]

It's possible that the measured voltage is due to the capacitance between each of the flat bus wires and the water on the surface of the mats. If a high impedance meter is used, it would draw a miniscule current and therefore it would only contribute a very small voltage drop across this capacitance. If you open circuit one of the hot legs, then the capacitance from its flat bus wire to the water on the surface will open be circuited (i.e, floating). But the other hot leg will still have 120V to ground, and so it can capacitively couple and cause the 110 Vac that you are measuring. There will be some amount of current leakage off the surface of the wet mat to its surroundings, and so that could account for the 10V drop from 120V down to the 110V you observed.

If the mats are powered by two legs of a 208/120V system, then L1-N and L2-N will be 120° apart. Each flat bus wire is connected to one leg and has its own capacitance to the water on the surface, and I assume that these two capacitances would be about equal from what was described. Therefore these two capacitances will form a capacitive divider when both legs are connected. The portions (i.e., components) of the L1-N and L2-N voltages that are in-phase with each other are cos(60°) = 1/2 times the magnitude of the L-N voltage. The out-of-phase components would cancel. And so we'd expect 1/2 x 120V = 60V to neutral (or ground) on the output "terminal"of the capacitive divider, which is the wet surface in this case. And an observed drop down to 50V to 55V could be due to some current leakage over to the mat's surroundings as mentioned in the paragraph above.

If the above analysis is correct, it would seem that if the mats were powered L-L from 120/240 single phase (or dropped down from 240 to 208 with a buck-boost), and a contactor was used to open both legs together, then the voltage coupled to the mat surface might be minimized. That's because the L1-N and L2-N are 180° apart and have no in-phase component, and so the currents they each capacitively couple to the wet surface would cancel. But you'd still need to open both legs to turn the heaters off, otherwise one leg would still be coupling 120V to the wet surface through its capacitance.

Although it would still be objectionable, it's possible that what you are observing would not be life threatening from an electrical shock if the current is sufficiently low. I could envision some experiments, for example, putting a small patch of aluminum foil over the wet mat and measuring the current from it to neutral or ground through a suitable meter. That would at least give you an idea about what you're up against. Just measuring current by placing a meter probe on the wet surface would not be adequate because of its small contact area when compared to, say, a human hand. I would first measure the voltage with a low impedance meter to make sure the voltage drops close to zero, in case a nail or similar connection to one of the hot conductors might be present.

Are these mats powered through a GFPE or GFCI?

You may also observe this type of phenomenon on commercial boilers with one of the three phase elements burnt out or 120/240V single phase units running on two legs of 208/120V.

 

[RickLosi]

Is the guard rail bonded? Assuming it is, I'd go for capacitive coupling if a metal test lead touching a damp but otherwise non-conductive surface.
Also, is this being tested with a high-impedance meter? I'm guessing it is.

I do not know what meter is being used, I will find out.
I am on site Tuesday and will run check to see if guard rail is the cause. EC never tested a wet roof to the guard rail.