GFCIs and their performance w/ and w/o EGCs

[MechEdetour]

I got into a little back-and-forth recently and it would help me to better understand some of the nuances with GFCIs. It shouldn't come as a surprise that the whole conversation kicked off when I said a GFCI does NOT need an EGC to operate. I've become obsessed with grounding the last few years. Nothing excites me more than when I here someone say that installing a GFCI in an ungrounded (no EGC) outlet solves the issue of not having an EGC. Mainly DIY'ers that take that stance, but there are professionals that agree.

GFCI= Ground-fault circuit interrupter
EGC = Equipment ground conductor

I am going to make some general comments (feel free to correct me if I'm wrong) and then I'll ask my questions about the nuances afterwards.

• GFCIs do not need an EGC to operate - they measure current imbalance between L and N and trip when the disparity >= 6mA
• EGCs provide a low-impedance path to ground to protect user's from shock when conductive "stuff" is inadvertently energized (ie. ground-faults)

This is where I am going --> Is there a situation where the presence of an EGC in a receptacle would make the operation of a GFCI "better" (ie. a situation where a GFCI w/ EGC would trip but a GFCI w/o an EGC would not)?

To help with the thought experiment:

Scenario 1: GFCI WITH EGC wired to ground screw. The case of a device (let the device be a fridge throughout) becomes energized @ 120V. The device has a L, N, G plug.

• EGC will provide a low-impedance path, breaker will trip.
• Also, current is flowing from L to ground, and the GFCI may also trip.
  • If breaker trips first, fault cleared. If GFCI trips first, fault is also cleared. If GFCI trips and breaker doesn't... Who cares? We're safe.

Scenario 2: GFCI WITHOUT EGC wired to ground screw (or present for that matter). The case of the device becomes energized @ 120V. The device has a L, N, G plug.

• EGC is absent, therefore no return path. Shock hazard - case is @ 120V.
• But, a GFCI is in the circuit! Is it possible a GFCI would ever trip in this situation? I would like to think that there is a possibility of leakage current from the case to ground (to ground, not the EGC specifically) to lead to a GFCI trip. @120V, we're talking 20,000ohm of impedance for 6mA. I know the impedance of the human body varies by A LOT, but if a person can have enough current flow through them to trip a GFCI, why couldn't some other arbitrary device? The physics are the same... If there is enough leakage to ground, the device will trip. Right?

For these two scenarios, it is obvious to me that a ground-fault with a properly installed EGC protects 100% of the time, whereas a GFCI protects ONLY if the leakage current is high enough to trip it. It does also seem that installing a GFCI in an ungrounded outlet may result in a situation where there may be enough leakage current from an unintentionally energized device. So use it where you can?

 

[don_resqcapt19]

It does also seem that installing a GFCI in an ungrounded outlet may result in a situation where there may be enough leakage current from an unintentionally energized device.

The studies done related to the ~5mA tripping point, is based on the protection of 95% of population...that is 95% of the population will not be harmed by ~5mA of current flowing through their body.

EGC will provide a low-impedance path, breaker will trip.

In the case of low level leakage faults, the breaker does not trip, but the only voltage available to drive a shock is the voltage drop on the EGC caused by the leakage current. This will be so low that there is not a hazard.

 

[suemarkp]

• EGCs provide a low-impedance path to ground to protect user's from shock when conductive "stuff" is inadvertently energized (ie. ground-faults)

I would change "ground" to "the source". You don't usually have low impedance paths to earth, but you have a low impedance path to the power source via the EGC.

In general, grounding is better than not having it. Some items need a path to earth to dissipate static electricity. Ungrounded wiring doesn't have this and doesn't get it via a GFCI. NEC 250.114 requires some things have an EGC, so you need to add one if you don't have it. GFCI isn't sufficient.

For your ungrounded GFCI scenario, leakage happens when someone touches it while standing on the earth. The impedance may still be too high to trigger 5ma of current depending on shoes and floor type, so the GFCI may not trip. But I would expect more leakage via touch than by a metal frame sitting on a finished floor (unless it is a bare concrete slab on grade).

 

[tortuga]

I got into a little back-and-forth recently .../...

Nothing excites me more than when I here someone say that installing a GFCI in an ungrounded (no EGC) outlet solves the issue of not having an EGC. Mainly DIY'ers that take that stance, but there are professionals that agree.

Years back a I was working for a realtor on a home sale of a 1950's home, all original two prong recepts, the new home owner wanted all the outlets 'grounded' so I quoted putting in GFCI's.
Well not so fast the new home owner was a savvy techie type and wanted his home office outlets actually grounded. Thats when I learned 250.114(3)(b) 'information technology equipment' covers lots of stuff like a typical home computer with a 3-prong cord. Even a laptop charger with a 3-prong cord is 'information technology equipment'

Scenario 2: GFCI WITHOUT EGC wired to ground screw (or present for that matter). The case of the device becomes energized @ 120V. The device has a L, N, G plug.

• EGC is absent, therefore no return path. Shock hazard - case is @ 120V.
• But, a GFCI is in the circuit! Is it possible a GFCI would ever trip in this situation? I would like to think that there is a possibility of leakage current from the case to ground (to ground, not the EGC specifically) to lead to a GFCI trip. @120V, we're talking 20,000ohm of impedance for 6mA. I know the impedance of the human body varies by A LOT, but if a person can have enough current flow through them to trip a GFCI, why couldn't some other arbitrary device? The physics are the same... If there is enough leakage to ground, the device will trip. Right?

For these two scenarios, it is obvious to me that a ground-fault with a properly installed EGC protects 100% of the time, whereas a GFCI protects ONLY if the leakage current is high enough to trip it. It does also seem that installing a GFCI in an ungrounded outlet may result in a situation where there may be enough leakage current from an unintentionally energized device. So use it where you can?

GFCI protection with no EGC is permitted if 250.114 and 406.4(D) or 410.44 Ex.1 (2023 NEC references) can be met, if something is UL listed with double insulation like the exceptions in 250.114 mention they have to be clearly marked and use a two prong cord not a 3-prong. There are probably zero three prong cords you can legally plug into a Scenario 2 receptacle in a non-residential occupancy.
Its kinda odd that 406.4(D) would let you replace a 2-prong receptacle for say a clothes washer with a 3-prong and not add a EGC but 250.114(3)(b) would not allow the homeowner to plug in the washer.
Most typical electrical inspectors that are going to inspect the work can't or don't enforce 250.114 since they stop enforcing the NEC at the receptacle, but if a accident were to occur and one of those TV accident attorneys gets involved the liability would be pinned on the electrician or electricians insurance for creating the hazard.

 

[don_resqcapt19]

Its kinda odd that 406.4(D) would let you replace a 2-prong receptacle for say a clothes washer with a 3-prong and not add a EGC but 250.114(3)(b) would not allow the homeowner to plug in the washer.

I tried to add and exception to 250.114, but CMP 5 rejected that with a panel statement that the UL product standards for any listed equipment that is provided with a cord having a ground pin on the plug requires an actual connection to earth ground.

 

[tortuga]

I'd change the wording to something like this:

406.4(D)(2) Non–Grounding-Type Receptacles.
Where attachment to an equipment grounding conductor does not exist in a 15 or 20 amp 250 volt or less receptacle enclosure, or a non-grounding type receptacle is installed the installation shall be GFCI protected at the origin of the branch circuit, receptacles or their cover plates shall be marked “No Equipment Ground”, and comply with (D)(2)(a), (D)(2)(b), or (D)(2)(c).
A wire type equipment grounding conductor shall not be connected to any outlet supplied from a Non–Grounding-Type circuit.
Metallic raceways, cables and enclosures used to extend non-grounded circuits shall be grounded in accordance with 250.130(C).

(a) A non–grounding-type receptacle(s) shall be permitted to be replaced with another non–grounding-type receptacle(s).
(b) A non–grounding-type receptacle(s) shall be permitted to be replaced with a ground-fault circuit interrupter-type of receptacle(s).
(c) A non–grounding-type receptacle(s) shall be permitted to be replaced with a grounding-type receptacle(s)

 

[MechEdetour]

I would change "ground" to "the source". You don't usually have low impedance paths to earth, but you have a low impedance path to the power source via the EGC.

In general, grounding is better than not having it. Some items need a path to earth to dissipate static electricity. Ungrounded wiring doesn't have this and doesn't get it via a GFCI. NEC 250.114 requires some things have an EGC, so you need to add one if you don't have it. GFCI isn't sufficient.

For your ungrounded GFCI scenario, leakage happens when someone touches it while standing on the earth. The impedance may still be too high to trigger 5ma of current depending on shoes and floor type, so the GFCI may not trip. But I would expect more leakage via touch than by a metal frame sitting on a finished floor (unless it is a bare concrete slab on grade).

All good points. I do my best to use "source" instead of "ground" when talking in this context but I think I defaulted to ground since the bonding screw and equipment bonding conductor are referenced as "ground" in the NEC (and other places).

I guess that's one of those things I don't have a sense of is how much current would flow in various leakage situations. Maybe nobody does.

 

[tortuga]

I would change "ground" to "the source". You don't usually have low impedance paths to earth, but you have a low impedance path to the power source via the EGC.

Yeah that is a good point in the UK and EU codes I think they even define the limits on what low impedance means a little better that we do.

I guess that's one of those things I don't have a sense of is how much current would flow in various leakage situations. Maybe nobody does.

There are a ton of academic papers on it, recently NASA did some research. Whats overlooked in the US standards is the exposure time in ms to the full fault current instead the NEC focused on the trip current setting, but not trip time.
If you look at a Residual Current Device (RCD) used in Germany for example are required (DIN VDE 0100-410)
they must trip in 300 milliseconds (ms) when tested at its rated residual operating current 30mA, thats General Non-Delayed RCDs.
(They also allow a few more grounding schemes or earthing systems than we do here but I am not up to date on the DIN's)
In part this 300ms limit is due to the likely interference of fault current with the hearts T-phase that occurs at around 400 ms which is more likely to cause fibrillation of the heart, you can find this in all kinds of documents.

When I look at UL 943 they use a formula that states the maximum permitted time to trip in seconds is equal to the quantity (20/fault current in milliamps) raised to the 1.43 power. If I did the math right this formula would permit about 7200 ms (7.2 second ) trip time for a 5 mA ground fault.
UL 943 does not get even close to 400ms until around 30mA of fault current.
At 30mA of fault current a UL 943 GFCI must trip no later than in 560ms or around 33 cycles of AC, slower than a RCD but close enough.
Between 5mA and 30mA there is a 'dead band' where trip time in UL 943 is in excess of 400ms.
Eaton has a paper on RCD's that covers this you can read here they discuss this trip time.

The kind of extremely hi resistance faults you'd see indoors on old ungrounded wiring *might* be in this range, that I dont know of any studies on. Where as a equipment grounding conductor intact would pull the full fault current in a few cycles triggering trip times below the 400ms.

 

[don_resqcapt19]

Where as a equipment grounding conductor intact would pull the full fault current in a few cycles triggering trip times below the 400ms.

Would it if the fault path is through the human to the EGC or other conductive grounded thing?

 

[tortuga]

Would it if the fault path is through the human to the EGC or other conductive grounded thing?

I was not thinking human, just stating the obvious when I commented.
A 'bolted fault' in a 120V appliance with a 3-prong cord and a proper EGC will facilitate a very fast (~100ms) GFCI trip probably 6 cycles or less with up to 1200 ohms.
I am not aware of any academic research into the 406.4(D) scenarios it would be interesting to see.

 

[don_resqcapt19]

A 'bolted fault' in a 120V appliance with a 3-prong cord and a proper EGC will facilitate a very fast (~100ms) GFCI trip probably 6 cycles or less with up to 1200 ohms.

The very reason I see no need for GFCIs on hard wired equipment.

 

[WD40]

One I time was working on a older home with 'old work' boxes in a plaster wall, The boxes were held in with those metal 'battle ship' things.
The GFCI is so wide, one of those battle ships accidentally made contact with hot pin and energized a box and the the ground pin on the GFCI and I did not notice, the hot pin was not on the GFCI side and someone got a shock between a tool and a grounded object. The GFCI obviously did not trip as its not in the circuit. I felt really bad and now I typically use GFCI breakers and avoid messing with those tiny old work boxes when I can.

 

[Jraef]

In general, grounding is better than not having it. Some items need a path to earth to dissipate static electricity. Ungrounded wiring doesn't have this and doesn't get it via a GFCI. NEC 250.114 requires some things have an EGC, so you need to add one if you don't have it. GFCI isn't sufficient.

This is the key point I make to people all the time when they are asking about using GFCIs to “fix” an older home with no grounding. The GFCI only addresses the SAFETY aspect. But modern electronics often rely on a solid ground path to help deal with not just static, but more importantly, common mode current (noise) which is inherent to how anything using PWM for voltage control works, and nearly ALL Switch Mode Power Supplies use a form of PWM. A good rule of thumb on that is that if your electronics has a grounding plug on it, it’s because it needed it for this reason and a GFCI, although safer, is not going to help.

So what can happen is that the CM current that would normally go to ground cannot, so it circulates inside the electronics looking for a path to ground, causing added heating on sensitive components and shortening their lifespan (heat x time = failure). How much shorter is not calculable because there are too many variables, but it is going to be shorter. So if they are talking about an outlet for a lamp of heating appliance, no problem, but if it’s an expensive PC, TV or music system, expect it to not last as long.

Now, considering that lots of people jettison their PCs and TVs every few years for other reasons anyway, nobody may notice.

 

[pipe_bender]

Can someone use 406.4(D) to put in a 4-prong range or dryer receptacle on a GFCI breaker?

 

[MechEdetour]

Yeah that is a good point in the UK and EU codes I think they even define the limits on what low impedance means a little better that we do.

There are a ton of academic papers on it, recently NASA did some research. Whats overlooked in the US standards is the exposure time in ms to the full fault current instead the NEC focused on the trip current setting, but not trip time.
If you look at a Residual Current Device (RCD) used in Germany for example are required (DIN VDE 0100-410)
they must trip in 300 milliseconds (ms) when tested at its rated residual operating current 30mA, thats General Non-Delayed RCDs.
(They also allow a few more grounding schemes or earthing systems than we do here but I am not up to date on the DIN's)
In part this 300ms limit is due to the likely interference of fault current with the hearts T-phase that occurs at around 400 ms which is more likely to cause fibrillation of the heart, you can find this in all kinds of documents.

When I look at UL 943 they use a formula that states the maximum permitted time to trip in seconds is equal to the quantity (20/fault current in milliamps) raised to the 1.43 power. If I did the math right this formula would permit about 7200 ms (7.2 second ) trip time for a 5 mA ground fault.
UL 943 does not get even close to 400ms until around 30mA of fault current.
At 30mA of fault current a UL 943 GFCI must trip no later than in 560ms or around 33 cycles of AC, slower than a RCD but close enough.
Between 5mA and 30mA there is a 'dead band' where trip time in UL 943 is in excess of 400ms.
Eaton has a paper on RCD's that covers this you can read here they discuss this trip time.

The kind of extremely hi resistance faults you'd see indoors on old ungrounded wiring *might* be in this range, that I dont know of any studies on. Where as a equipment grounding conductor intact would pull the full fault current in a few cycles triggering trip times below the 400ms.

Thanks for taking the time to share all of that.

Tying this back to my initial question... if I was touching an object that was properly bonded through an EGC, and then I also touched L on a GFCI equipped with an EGC, I would assume the impedance would be considerably lower than if I wasn't touching the grounded (err, bonded) object - correct? A larger current would flow through me and the bonded object than through me and the earth. If the trip time is current sensitive then this could be a good thing?

With that being said, do GFCIs operate on a time-current curve like typical OCPDs? If they do, then I can see how an EGC could potentially benefit a GFCI receptacle even though an EGC is not explicitly required for a GFCI to operate. Sure, more current doesn't always mean safer, but it can be if the time of exposure is very low. If this is a wrong assumption and GFCI trip timing is independent of the magnitude of current, then it solidifies my initial point that an EGC doesn't do anything for GFCI operation.

 

[don_resqcapt19]

If this is a wrong assumption and GFCI trip timing is independent of the magnitude of current,

The product standard for GFCIs says:

The maximum permitted time to trip in seconds is equal to the quantity (20/fault current in milliamps) raised to the 1.43 power. The application of this formula would permit a 7 second trip time for a 5 mA ground fault.

Note, that in my experience they trip faster than the maximum permitted time...If it was set at the maximum permitted time, there would be a number of seconds between pushing the test button and the device tripping. I have never seen a noticeable delay between pushing the test button and the device tripping.

 

[MechEdetour]

The GFCI only addresses the SAFETY aspect.

Not in it's entirety though right? I am trying to envision all the various situations where a GFCI is beneficial on a system with EGCs throughout, a system with no EGCs, faults where conductive objects are energized, faults where the only object energized is the person (ie. wet hands, etc.)

I think what might help set me straight is if I knew where a GFCI is beneficial and a EGC isn't, and vice versa.

Taking a shot at at some blanket statements - hope the internet does it's thing and proves me wrong:

• Case 1: A metal object faults to L with EGC present; without GFCI = Fault is immediately cleared by OCPD. No shock hazard to person.
• Case 2: A metal object faults to L without EGC present; without GFCI = Shock hazard. No means of clearing the fault during shock (assume impedance through earth is not enough to trip OCPD).
• Case 3: A metal object faults to L with EGC present; with GFCI = Fault is immediately cleared by OCPD. No shock hazard to person. GFCI is "useless"... might as well be a standard receptacle.
• Case 4: A metal object faults to L without EGC present; with GFCI = Shock hazard. Object may or may not have enough leakage through earth to trip GFCI before shock occurs (the latter being the likely scenario). During personnel shock, GFCI may (err, should?) or may not trip. If the requirements of UL943 are met, the current through person+earth will cause GFCI to trip, hopefully limiting the shock exposure and saving the person.
• Case 5: A person with wet hands touches L with EGC present; without GFCI; Person is not making contact with anything but earth = Shock hazard. Presence of EGC is irrelevant. Current through person+earth is insufficient to trip OCPD.
• Case 6: A person with wet hands touches L without EGC present; without GFCI; Person is not making contact with anything but earth = Shock hazard. No different than Case 5.
• Case 7: A person with wet hands touches L with EGC present; with GFCI; Person is not making contact with anything but earth = Shock hazard, but there's hope - current through person+earth may be high enough to trip GFCI and limit shock exposure. EGC does nothing here as well.
• Case 8: A person with wet hands touches L without EGC present; with GFCI; Person is not making contact with anything but earth = Shock hazard, and similar to Case 7 - current through person+earth may be high enough to trip GFCI and limit shock exposure.
• Case 9: A person with wet hands touches L with EGC present; without GFCI; Person is also making contact with an adequately bonded metal object = Shock hazard. Same as Case 5, but now there are parallel paths. Person+earth and person+object+EGC. So current may be higher on the EGC path (assuming earth impedance is much greater than person impedance). I wouldn't want to be in this situation to be honest. I believe the current through the person would be much higher, but likely still not high enough to trip an OCPD.
• Case 10: A person with wet hands touches L with EGC present; with GFCI; Person is also making contact with an adequately bonded metal object = Shock hazard. Referring to my previous post (post#15), if a GFCI is time-current sensitive, an EGC may benefit the operation of a GFCI due to the increased current that may result from being in contact with a lower impedance path to ground (ie. higher leakage current).

In the cases above, I see scenarios where a GFCI does what an EGC can't, and vice versa. Case 4 definitely stands out to me because it could potentially provide some good protection in the absence of an EGC. Case 9 is particularly interesting too because it seems as though the combination of having an EGC and not having a GFCI can lead to an even greater shock hazard (therefore adding more value to the addition of a GFCI - Case 10).

Am I mad?

 

[MechEdetour]

Note, that in my experience they trip faster than the maximum permitted time...If it was set at the maximum permitted time, there would be a number of seconds between pushing the test button and the device tripping. I have never seen a noticeable delay between pushing the test button and the device tripping.

That's exactly what I'm asking. Just because a standard has an equation that governs minimum clearing time (with current as a variable), that doesn't mean they actually operate slower/faster at lower/higher levels of current (like a fuse or breaker). I am not familiar with the nuances of GFCI design, but with the presence of electronics within it I am guessing that if a particular GFCI trips in 200ms @ 6mA, then it will trip in 200ms @ 50mA as well. Hard for me to see why time would change if there is no thermal/mag element (no pun intended) of some sort.

 

[tortuga]

Thanks for taking the time to share all of that.

Tying this back to my initial question... if I was touching an object that was properly bonded through an EGC, and then I also touched L on a GFCI equipped with an EGC, I would assume the impedance would be considerably lower than if I wasn't touching the grounded (err, bonded) object - correct? A larger current would flow through me and the bonded object than through me and the earth. If the trip time is current sensitive then this could be a good thing?

Yes, google says the human body is between 1k and 100k ohms, so in the 406.4(D) scenarios if you have over 24k ohms of total impedance only a few mA will flow.
And you might fee a tingle but the GFCI should not trip at 4mA.
The example I gave before was 1200 ohms total, 100mA fault current, 6 cycles of 120 AC a GFCI will trip in 100ms.

A 15A household inverse time breaker may need about 3x its rating to trip in a similar time frame of 6 cycles, but to trip that fast the impedance would need to be below 2 ohms so adding a human it would clearly never tip, but a well grounded appliance frame should trip it.

With that being said, do GFCIs operate on a time-current curve like typical OCPDs? If they do, then I can see how an EGC could potentially benefit a GFCI receptacle even though an EGC is not explicitly required for a GFCI to operate. Sure, more current doesn't always mean safer, but it can be if the time of exposure is very low.

Yes thats correct, if I had a old house I'd prefer just to use 250.130(C)

 

[don_resqcapt19]

Case 1: A metal object faults to L with EGC present; without GFCI = Fault is immediately cleared by OCPD. No shock hazard to person.

There is a shock hazard for the time it takes to trip the OCPD. That shock hazard is the voltage drop on the EGC.

Case 7: A person with wet hands touches L with EGC present; with GFCI; Person is not making contact with anything but earth = Shock hazard, but there's hope - current through person+earth may be high enough to trip GFCI and limit shock exposure. EGC does nothing here as well.

UL 943 assumes that if the current through the person is not enough to trip the GFCI, the shock is not a serious hazard.

Case 9: A person with wet hands touches L with EGC present; without GFCI; Person is also making contact with an adequately bonded metal object = Shock hazard. Same as Case 5, but now there are parallel paths. Person+earth and person+object+EGC. So current may be higher on the EGC path (assuming earth impedance is much greater than person impedance). I wouldn't want to be in this situation to be honest. I believe the current through the person would be much higher, but likely still not high enough to trip an OCPD.

It is almost impossible to flow enough current through a person to cause an OCPD to open.