GFCI *breaker* failure modes?

[Don Y]

Hi Bud,

GFCIs will trip on a N-G short with no load current.

There is a second
current transformer that tries to put a small common mode current in
both the H and N wire.

What are they trying to protect against -- a high resistance neutral?
Or, just being pedantic? This seems like a significant additional
requirement so one would assume there is a significant practical risk...

If there is a N-G short the resulting current
will trip the GFCI. The CT current is obviously not needed for a H-G
trip, but if the GFCI is reverse H-N wired the common mode current will
trip it if there is a actual N-to-ground short, which now goes through
the GFCI on the H terminals.

Ah, OK.

Other trivia - for a GFCI receptacle the load terminals used to be
connected to the receptacle. If the GFCI was reverse line-load wired the
test button would appear to work but the GFCI receptacle was not
protected. Wiring downstream was protected. The recommended test is to
plug in a lamp and make sure it goes out when the test button is pushed.
Recent GFCIs disconnect the internal receptacle from both line and load
on a trip. If reverse line-load wired the receptacle is still not
protected, but if tripped the GFCI can not be reset. For that reason
they are shipped tripped.

As someone wrote, the internal test for GFCIs connects around the CT,
and GFCIs make no connection to the ground wire. But a plug-in GFCI
tester will not work if there is no ground.

Makes sense -- a 2 wire tester would just look like a "load".

I've still not found my HiPot (though DVM doesn't turn up anything
obvious with NO load). Ah, well... I've still got a few months
before I'll *need* that circuit! :-/

 

[bud--]

On 9/19/2013 10:23 AM, Charlie E. wrote:

If it is like my christmas light circuits, it has outdoor recepticals.
These seem to have a tendency to build up dirt and moisture that makes
them trip the breaker pretty easily...

Why? Note the receptacles are "normal" (three wire) receptacles
(the GFCI function is in the breaker).

If this is the case, then I should (possibly) be able to find
some *other* receptacle on that branch circuit that doesn't
exhibit the problem? E.g., one of those on the (covered) porch?

OK, this (Charlie's comment) has to be at least part of the problem.

I went to one of the porch outlets (porch is shaded; other outlets
are out in the afternoon sun -- 2PM being the hotest time of day;
electrical panel similarly "exposed").

Plugged in a single 100W bulb in a fixture that is always used
indoors. Fine. Plugged in a second 100W bulb in another, similar
fixture. Also fine. Now have exhausted both receptacles in that
duplex outlet so introduce a short extension cord that terminates
in a 2G box (with a pair of duplex receptacles).

Transfer both fixtures to the that "extension". All is *still* OK.
(you would *expect* it to be, but... see below)

Figuring the 100W-at-a-time approach is going to quickly deplete me
of fixtures, I went looking for a larger load. Electric hair dryer!
Heck, it works fine on the indoor circuits so that should attest to
it's "quality" (as a load).

Adding this to the 200W load blew the breaker -- before the
hairdryer even got started.

[Hmmm... technically, there's a reactive component to this load
for the fan motor so it's not *purely* resistive]

Repeat with *just* the hair dryer. Works OK.

Add one 100W lamp. OK.

Add second 100W lamp. NFG.

Great! I now need to quantify the load represented by the hair
dryer and I know what the trip point is! (ASSUMING it is an
overload trip).

Hair dryer has four settings (not counting two fan speeds).
So, just switch among those settings and see at which point
the breaker trips (hair dryer had been on lowest heat setting
previously).

Low/cool -- OK.
Warm -- OK.
Hot -- OK.
VERY HOT -- OK!

WTF? Surely the difference between Low and VERY HOT is more than
the 200W incandescent load (that, when added to the Low setting,
had caused the breaker to trip!).

Repeat the previous exercise (with the lamps). Just to be sure
I didn't miss some important detail. Exactly the same results.

OK, stop "overlooking the obvious". SOMETHING is different in these
two tests. What is it?

Well, to plug in the three loads (two lamps plus hair dryer), I have
to resort to the receptacle multiplier represented by the "extension
cord". That particular cord is often outdoors. Just like most
of the duplex receptacles on that circuit!

So, while I can plug the hair dryer into the duplex receptacle ON THE
PORCH and have no problems, this isn't true if the load is fed from
the (three wire!) extension cord.

[I suspect if I tried plugging the hair dryer into one of the more
"troublesome" outlets, it wouldn't work on the VERY HOT setting,
either!]

And, this then appears NOT to be a case of overload but, rather,
ground fault/leakage! But, only manifesting at larger loads?
(or, perhaps only when certain outlets are utilized??)

So, what is it (?) about the receptacle's mechanical structure
that can cause this sort of failure ONLY WHEN IN USE?

Receptacle
hot -------------------+---| |----+
| o |
| |
+---| |----+
o |
|
neutral ----------------------------+
|
[R]
|
Gnd

R represents a partial short to ground via
spider web/crud/dirt/whatever.

With no load connected, there is no current in either hot
or neutral. Therefore, there is no imbalance between them,
and the GFI doesn't trip. But, when the receptacle is in use,
then there is current in both wires and current in the partial
short to ground. If the current through the partial short to
ground is ~5mA or more, the GFI trips.

If there was a partial short to ground on the hot wire,
then there would be current there without a load, and
the GFI would trip, provided the current was ~5mA or more.

That answers the "ONLY WHEN IN USE" portion of your question.
The failure is an environmental issue - crud getting in to
the junction box - not a receptacle mechanical structure
issue. The crud could just as easily have created a partial
short to ground on the hot side as the neutral side.

Ed

GFCIs will trip on a N-G short with no load current. There is a second
current transformer that tries to put a small common mode current in
both the H and N wire. If there is a N-G short the resulting current
will trip the GFCI. The CT current is obviously not needed for a H-G
trip, but if the GFCI is reverse H-N wired the common mode current will
trip it if there is a actual N-to-ground short, which now goes through
the GFCI on the H terminals.

Other trivia - for a GFCI receptacle the load terminals used to be
connected to the receptacle. If the GFCI was reverse line-load wired the
test button would appear to work but the GFCI receptacle was not
protected. Wiring downstream was protected. The recommended test is to
plug in a lamp and make sure it goes out when the test button is pushed.
Recent GFCIs disconnect the internal receptacle from both line and load
on a trip. If reverse line-load wired the receptacle is still not
protected, but if tripped the GFCI can not be reset. For that reason
they are shipped tripped.

As someone wrote, the internal test for GFCIs connects around the CT,
and GFCIs make no connection to the ground wire. But a plug-in GFCI
tester will not work if there is no ground.

 

[Don Y]

Hi Bud,

On 9/21/2013 9:40 AM, bud-- wrote:

[attributions elided]

30mA sure seems high to me.

I'd always used 30mA and 30-48V as the red flags. I think even
*tiny* currents can cause the heart to be stopped, though.

AFCIs include ground fault protection, typically at the 30mA level used
in the UK for shock protection. In an AFCI the ground fault trip is for
fire protection. The idea is that arcing may involve leakage to ground
(if present), particularly in 3-wire cords. If ground fault protection
was not included AFCIs probably wouldn't need the neutral wired through
them (in the same way that GFCIs do).

Being a cynic, I had assumed AFCI's included GF protection so the
user wouldn't consider the AFCI vs. GFCI choice to be "one or the
other" (type of protection) but, rather, see AFCI's as "GFCI *plus*
arc protection"?

Sort of like smoke *plus* CO detection, etc.

 

[bud--]

I am. They work on un-grounded (two prong - line and neutral, no
earth/safety ground connection) devices.


Well, the "magic numbers" I learned were that 10-100ma is the range of
current (through the heart) most likely to kill a person - below that
the heart does not go into fibrillation and above that the heart is
stopped (and will usually self-start when current is removed.) I can't
say I've independently tested the veracity of that, but it was what I
was taught from reasonably reliable/authoritative sources.

30mA sure seems high to me.

AFCIs include ground fault protection, typically at the 30mA level used
in the UK for shock protection. In an AFCI the ground fault trip is for
fire protection. The idea is that arcing may involve leakage to ground
(if present), particularly in 3-wire cords. If ground fault protection
was not included AFCIs probably wouldn't need the neutral wired through
them (in the same way that GFCIs do).

 

[Anthony Stewart]

GFI's trip the breaker based on a differential current between ground and neutral.

I have seen this trip from 2 conditions.

1. Excess Capacitance from line to ground. >1uF

2. Excess resistive leakage current. <10k from line to ground.

Your results may vary.

 

[Phil Allison]

"Anthony Stewart"

GFI's trip the breaker based on a differential current between ground and
neutral.

** Nope - the differential is between active and neutral.

If there is an imbalance, current must be flowing to supply or actual
ground.

Deliberate capacitance ( in an appliance) connected from active to supply
ground is strictly limited in value so the current flow is only a few mA at
most and will not trip a GFCI.



.... Phil

 

[Phil Allison]

"bud--"

GFCIs will trip on a N-G short with no load current.

** Neutral conductors are normally carrying some current, creating voltage
drops along cables leading to outlets.

The N to E voltage difference at an outlet may vary from a couple of volts
down to a few mV.

So a short from neutral to safety ground at an outlet causes neutral current
to flow with no matching active - so trips the breaker.

There is a second current transformer that tries to put a small common
mode current in both the H and N wire.

** Hogwash.

As someone wrote, the internal test for GFCIs connects around the CT, and
GFCIs make no connection to the ground wire. But a plug-in GFCI tester
will not work if there is no ground.

** Correct.

But the self test button is proof of functionality.

.... Phil

 

[ehsjr]

On 9/19/2013 12:42 PM, Don Y wrote:
On 9/19/2013 10:23 AM, Charlie E. wrote:

If it is like my christmas light circuits, it has outdoor recepticals.
These seem to have a tendency to build up dirt and moisture that makes
them trip the breaker pretty easily...

Why? Note the receptacles are "normal" (three wire) receptacles
(the GFCI function is in the breaker).

If this is the case, then I should (possibly) be able to find
some *other* receptacle on that branch circuit that doesn't
exhibit the problem? E.g., one of those on the (covered) porch?

OK, this (Charlie's comment) has to be at least part of the problem.

I went to one of the porch outlets (porch is shaded; other outlets
are out in the afternoon sun -- 2PM being the hotest time of day;
electrical panel similarly "exposed").

Plugged in a single 100W bulb in a fixture that is always used
indoors. Fine. Plugged in a second 100W bulb in another, similar
fixture. Also fine. Now have exhausted both receptacles in that
duplex outlet so introduce a short extension cord that terminates
in a 2G box (with a pair of duplex receptacles).

Transfer both fixtures to the that "extension". All is *still* OK.
(you would *expect* it to be, but... see below)

Figuring the 100W-at-a-time approach is going to quickly deplete me
of fixtures, I went looking for a larger load. Electric hair dryer!
Heck, it works fine on the indoor circuits so that should attest to
it's "quality" (as a load).

Adding this to the 200W load blew the breaker -- before the
hairdryer even got started.

[Hmmm... technically, there's a reactive component to this load
for the fan motor so it's not *purely* resistive]

Repeat with *just* the hair dryer. Works OK.

Add one 100W lamp. OK.

Add second 100W lamp. NFG.

Great! I now need to quantify the load represented by the hair
dryer and I know what the trip point is! (ASSUMING it is an
overload trip).

Hair dryer has four settings (not counting two fan speeds).
So, just switch among those settings and see at which point
the breaker trips (hair dryer had been on lowest heat setting
previously).

Low/cool -- OK.
Warm -- OK.
Hot -- OK.
VERY HOT -- OK!

WTF? Surely the difference between Low and VERY HOT is more than
the 200W incandescent load (that, when added to the Low setting,
had caused the breaker to trip!).

Repeat the previous exercise (with the lamps). Just to be sure
I didn't miss some important detail. Exactly the same results.

OK, stop "overlooking the obvious". SOMETHING is different in these
two tests. What is it?

Well, to plug in the three loads (two lamps plus hair dryer), I have
to resort to the receptacle multiplier represented by the "extension
cord". That particular cord is often outdoors. Just like most
of the duplex receptacles on that circuit!

So, while I can plug the hair dryer into the duplex receptacle ON THE
PORCH and have no problems, this isn't true if the load is fed from
the (three wire!) extension cord.

[I suspect if I tried plugging the hair dryer into one of the more
"troublesome" outlets, it wouldn't work on the VERY HOT setting,
either!]

And, this then appears NOT to be a case of overload but, rather,
ground fault/leakage! But, only manifesting at larger loads?
(or, perhaps only when certain outlets are utilized??)

So, what is it (?) about the receptacle's mechanical structure
that can cause this sort of failure ONLY WHEN IN USE?

Receptacle
hot -------------------+---| |----+
| o |
| |
+---| |----+
o |
|
neutral ----------------------------+
|
[R]
|
Gnd

R represents a partial short to ground via
spider web/crud/dirt/whatever.

With no load connected, there is no current in either hot
or neutral. Therefore, there is no imbalance between them,
and the GFI doesn't trip. But, when the receptacle is in use,
then there is current in both wires and current in the partial
short to ground. If the current through the partial short to
ground is ~5mA or more, the GFI trips.

If there was a partial short to ground on the hot wire,
then there would be current there without a load, and
the GFI would trip, provided the current was ~5mA or more.

That answers the "ONLY WHEN IN USE" portion of your question.
The failure is an environmental issue - crud getting in to
the junction box - not a receptacle mechanical structure
issue. The crud could just as easily have created a partial
short to ground on the hot side as the neutral side.

Ed

GFCIs will trip on a N-G short with no load current.

Yes, if the short's resistance is 1.6 ohms or less, with
the neutral's resistance at .4 ohms, per UL 943.
But we're talking about a partial short from neutral to ground,
whose resistance can be more than that 1.6 ohms. Let's say
it is 100 ohms. At 100 ohms partial short the GFI won't
trip with no load.

With a 100 ohm partial short and a .4 ohm neutral, say you
connect a 5 amp load. The .4 ohm neutral will carry 99.6% of
the current and the 100 ohm load will carry the rest. 99.6 %
of 5 amps is 4.98 amps - leaving 20 mA carried in the partial
short. Bingo - the GFI trips. Obviously, YMMV - except make
that your mileage Will vary. Not every circuit has a .4 ohm
neutral or a 100 ohm partial short, or every load 5 amps.

So lets take another approach:

Say the neutral is 50 feet of #12 at 2 ohms per 1000 feet,
yielding a .1 ohm neutral, assuming all connections
are 0 ohms. 4 amps through .1 ohms is a voltage drop of
..4 volts across the 50 feet of neutral wire. Lets say the
partial short is 50 ohms. .4 volts across 50 ohms causes
an 8 mA current. Bingo, the GFI trips, but that same 50
ohms won't cause the GFI to trip with no load connected.

How neutral to ground short circuit trips a GFI with no load:

The neutral to ground transformer forms part of an oscillator
circuit that runs at 8 kHz per the RV4141A datasheet. That
transformer, and the sense transformer are mutually coupled
through Rg (the short) Rn (resistance of the neutral) and the
neutral wire ground loop (the neutral is grounded at the panel,
so with a short or partial short at the receptacle end, a
ground loop is formed). No load current need be drawn for
the oscillator to function. With the oscillator running, a
small current is induced on both the hot and neutral wires, but
the current on the neutral wire differs from the current on the
hot wire, because the neutral wire has a path to ground through
the short. The sense transformer "sees" a difference between hot
and neutral and the sense amplifier amplifies it. If the
difference is large enough, the GFI trips.

Ed

 

[ehsjr]

Hi Bud,



Really? <frown> Then that kills that (trip-but-only-when-loaded)
possible expnation...

No it doesn't. See my reply to bud for the details.
He's talking about a short, but what was discussed was
a partial short. The short would be 1.6 ohms or less
per UL 943, while the partial short would be higher,
perhaps as much as 100 ohms.

Ed

 

[Phil Allison]

"Mike Perkins"

( snip utter garbage from some troll )

I have known a low resistance path from Neutral to Ground to also cause
random trips.

** Neutral conductors are normally carrying some current, creating voltage
drops along cables leading to outlets.

The N to E voltage difference at an outlet may vary from a couple of volts
down to a few mV.

So a short from neutral to safety ground at an outlet causes neutral current
to flow with no matching active - which trips the GFCI breaker.

The possible random nature of this is obvious - if the "short" has some
resistance.



.... Phil

 

[Phil Allison]

"bud--"
Phil Allison

Which everyone agrees with.

** Show me where that has been said here ??

1. It is comforting to know that people outside the US are so familiar
with the operation of US GFCIs.

2. This feature is rather well know to technical people working with power
wiring.

** What a pile of smug, pompous crap.

3. If you look at the datasheet for an IC for GFCIs
http://www.ti.com/lit/ds/symlink/lm1851.pdf
the "hogwash" is clearly in the "Typical Application" circuit as
"GND/NEUTRAL COIL", which requires only an added capacitor to function.

** Says nothing like the pile of crap YOU posted.




.... Phil

 

[bud--]

"bud--"

** Neutral conductors are normally carrying some current, creating voltage
drops along cables leading to outlets.

The N to E voltage difference at an outlet may vary from a couple of volts
down to a few mV.

So a short from neutral to safety ground at an outlet causes neutral current
to flow with no matching active - so trips the breaker.

Which everyone agrees with.

** Hogwash.

1. It is comforting to know that people outside the US are so familiar
with the operation of US GFCIs.

2. This feature is rather well know to technical people working with
power wiring.

3. If you look at the datasheet for an IC for GFCIs
http://www.ti.com/lit/ds/symlink/lm1851.pdf
the "hogwash" is clearly in the "Typical Application" circuit as
"GND/NEUTRAL COIL", which requires only an added capacitor to function.

4. I could provide a picture of a GFCI with 2 CTs (sense and "hogwash")
if the datasheet is not enough.

 

[bud--]

Yes, if the short's resistance is 1.6 ohms or less, with
the neutral's resistance at .4 ohms, per UL 943.
But we're talking about a partial short from neutral to ground,
whose resistance can be more than that 1.6 ohms. Let's say
it is 100 ohms. At 100 ohms partial short the GFI won't
trip with no load.

Testing a GFCI, it tripped on 10 ohms N-G and did not trip on 15 ohms.

With a 100 ohm partial short and a .4 ohm neutral, say you
connect a 5 amp load. The .4 ohm neutral will carry 99.6% of
the current and the 100 ohm load will carry the rest. 99.6 %
of 5 amps is 4.98 amps - leaving 20 mA carried in the partial
short. Bingo - the GFI trips. Obviously, YMMV - except make
that your mileage Will vary. Not every circuit has a .4 ohm
neutral or a 100 ohm partial short, or every load 5 amps.

So lets take another approach:

Say the neutral is 50 feet of #12 at 2 ohms per 1000 feet,
yielding a .1 ohm neutral, assuming all connections
are 0 ohms. 4 amps through .1 ohms is a voltage drop of
.4 volts across the 50 feet of neutral wire. Lets say the
partial short is 50 ohms. .4 volts across 50 ohms causes
an 8 mA current. Bingo, the GFI trips, but that same 50
ohms won't cause the GFI to trip with no load connected.

I agree that with relatively high N-G resistance, load current may trip
a GFCI.

How neutral to ground short circuit trips a GFI with no load:

The neutral to ground transformer forms part of an oscillator
circuit that runs at 8 kHz per the RV4141A datasheet. That
transformer, and the sense transformer are mutually coupled
through Rg (the short) Rn (resistance of the neutral) and the
neutral wire ground loop (the neutral is grounded at the panel,
so with a short or partial short at the receptacle end, a
ground loop is formed). No load current need be drawn for
the oscillator to function. With the oscillator running, a
small current is induced on both the hot and neutral wires, but
the current on the neutral wire differs from the current on the
hot wire, because the neutral wire has a path to ground through
the short. The sense transformer "sees" a difference between hot
and neutral and the sense amplifier amplifies it. If the
difference is large enough, the GFI trips.

Ed

The datasheet for an LM1851 shows the N-G sense coil more simply powered
by 120Hz from the bridge power supply.

Your datasheet is more recent:
http://www.fairchildsemi.com/ds/RV/RV4141A.pdf

What a coincidence - your datasheet also shows a second CT to detect a
N-G short.

 

[bud--]

"bud--"
Phil Allison


** Show me where that has been said here ??



** What a pile of smug, pompous crap.

I have seen it referred to by several people who are more technically
competent than you about GFCIs.

But you just post smug, pompous crap.

** Says nothing like the pile of crap YOU posted.

The pile of crap I posted, which you replied to, was:
"There is a second current transformer that tries to put a small common
mode current in both the H and N wire."

The second current transformer is plainly seen in both the data sheet I
posted and the data sheet Ed referred to, that I posted a link to.

But some people have a problem admitting they are wrong.

 

[Don Y]

Hi Mike,

I have know a low resistance path from Neutral to Ground to also cause
random trips.

"Low resistance" would show up readily -- with just a DVM.
I don't see that in a "static" probe of the items in question.

I.e., if it was "obvious", I wouldn't be asking! :>

 

[ehsjr]

Testing a GFCI, it tripped on 10 ohms N-G and did not trip on 15 ohms.


I agree that with relatively high N-G resistance, load current may trip
a GFCI.


The datasheet for an LM1851 shows the N-G sense coil more simply powered
by 120Hz from the bridge power supply.

Your datasheet is more recent:
http://www.fairchildsemi.com/ds/RV/RV4141A.pdf

What a coincidence - your datasheet also shows a second CT to detect a
N-G short.

AFAIK that's the only way N-G short detection is done
in GFI's.

Maybe you remember and can comment on this:
I seem to recall that a while back, when GFCI's were first
introduced, they were not designed to detect N-G shorts.
I'm not sure.

BTW - you can make a nice AC load sensing switch using a
modified GFCI receptacle. I've done it with both the 1851
and 4141 IC GFI's.

Ed

 

[Don Y]

Hi Mike,

I presume in the US the Neutral and Earth are linked in the house or
building.

In the UK the Neutral enters the property and is generally kept separate
from the house or building Earth, so there is always a small voltage
difference between Neutral and Earth. Here, I've not seen the need to
introduce a small common mode voltage into the load Live-Neutral to
check for Neutral-Earth conduction paths.

Utility supplies line & neutral to customer. *Customer* provides
earth. In the past, a cold water pipe (as long as it was metal
for at least 10 ft through soil) was used for earth. Code was
revised to *add* the need for a separate "earth" (metal stake driven
into the soil near the service entrance -- can you spell "thermite"?).

Neutral is bonded to earth IN the electrical panel.

Neutral and earth (avoiding the use of "ground", temporarily)
are supposed to be at (roughly) the same potential. The key
difference is earth is never *intended* to carry current (in
normal/non-fault circumstances).

E.g., you can EARTH the case of a device but can't tie the
case of a device to "neutral" (notable exception is the main
electric panel and some large appliances). I.e., the earth
conductor on a three-wire outlet has to be connected to *earth*,
not "neutral". Ditto for metal Jboxes, etc. (this is one of the
reasons I *prefer* metal Jboxes over plastic -- miswired fixtures
in plastic boxes aren't obvious and can have potential on surfaces
that are exposed to users!)

As many rely on water pipes (possibly mistakenly!) for "earth",
anything that can interrupt that continuity must include provisions
to ensure earth's integrity even when it would otherwise be
interrupted. E.g., our outdoor plumbing (irrigation, bibbs, etc.)
are all immediately fed from copper pipe -- but, just upstream from
that is *PVC* so you can't rely on the hose bibb to be "earthed".

E.g., if you install a water softener in your supply line, you add
a jumper bonding the "inlet" pipe to the "outlet" pipe so that
continuity is maintained even if the water softener is removed.
(ditto for water heater, filtration unit, etc.).

Perhaps the biggest single section/subject in the Code book is
concerned with grounding! :-/

 

[Phil Allison]

"ehsjr"

Maybe you remember and can comment on this:
I seem to recall that a while back, when GFCI's were first
introduced, they were not designed to detect N-G shorts.
I'm not sure.

** Any GFCI will detect the neutral current that flows when you short N and
G at an outlet. It is often several amps.




..... Phil

 

[Phil Allison]

"Mike Perkins"

I presume in the US the Neutral and Earth are linked in the house or
building.

** This is the case here in Australia too.

In the UK the Neutral enters the property and is generally kept separate
from the house or building Earth, so there is always a small voltage
difference between Neutral and Earth. Here, I've not seen the need to
introduce a small common mode voltage into the load Live-Neutral to check
for Neutral-Earth conduction paths.

** Seems OTT to me in any case.

The issue that GFCIs may simply fail to work when you need them is way more
important.



..... Phil

 

[Phil Allison]

"Mike Perkins"

They are for us, but the separation of Neutral and Earth mean that under
Live-Neutral fault conditions, any fault Neutral potential doesn't find
its way to the Earth. Under normal conditions they are up to a few volts
different.

Our Neutral is Earthed at the local substation which may be quite a
distance from the property.

** The hazard with linking neutral and earth at the supply entrance to each
premises arises if ever the incoming neutral is lost. Load current in the
premises then returns via the local grounding system - which can include
plumbing.

Joe plumber then comes along, digs a hole, hops in and disconnects a pipe.

Might be the last thing Joe ever does.



..... Phil