GFCI Ground Fault Circuit Interrupter Inline C13 to C14?

| Has anyone seen a ground fault circuit interrupter (GFCI) that would work | inline with a computer server plug? These typically terminate with an IEC | C13 female plug end, so the GFCI adapter would need to be C14 to C13. | | We are finding when we aggregate a large number of servers with a single | GFCI that we get false positives and end up shutting down whole banks of | servers. So we want to try to isolate each GFCI to a single plug on each | server and see whether that improves reliability.

Do you have a reason to use GFCI on computer servers in the first place?

How many computers are involved? What is the required circuit capacity for serving all these computers through the one GFCI you are serving them with right now?

What is the leakage current rating of the GFCI you are using now?

Is this GFCI an integrated outlet type, or a circuit breaker type?

You are probably having problems because of the asymmetrical loads from switching power supplies. For grins, try plugging in a noise filter between the GFCI and the power strip for the servers. Tripp-Lite Isobar strips have noise filters built in.

[8~{} Uncle Monster

Doesn't the power cord have a C14 on one end and a normal AC plug on the other end? Wouldn't it be easier to put the GFCI on the AC plug end?

The purpose of a GFIC is primarily to protect people not property. The way you describe your installation, the same result should be achievable if you can be certain that all conductive surfaces people can come in contact with are grounded.

Any leakage to ground can increase probability of trip. This leakage can arise from capacitance to ground, electrical leakage to pumped fluids, etc. The more devices in parallel, the greater the problem. I have had situations with pumps where leakage to the protective ground caused trips, probably from inductive kicks, when the pump was switched off.

Bill

The risk we wanted to protect against was someone opening the door to a computer rack, accidentally allowing a power cord to slip into the door seam, and then closing the door onto the cord, partly severing the power cord and electrifying the rack fixture. Our hope would be that the GFCI would trigger immediately upon severing of the power cord, thereby (hopefully) preventing a risk of electrocution. At the same time, we would want to localize the fault to a specific device and not take out the entire server farm.

Probably 5 to 10 servers per circuit. We seem to run about 10 to 20 amps at 125V for those servers.

Unfortunately the GFCI we have today is integrated into the outlet, and I don't know the ratings. I was hoping to just localize a GFCI to each power cord to avoid the need to engineer a single common GFCI at the main outlet to the rack.

I think you are right. It would make more sense to have a GFCI at the base of the power cord, presenting 5-20R to the 5-15P or 5-20P plug of the power cord.

|> | Has anyone seen a ground fault circuit interrupter (GFCI) that would |> work |> | inline with a computer server plug? These typically terminate with an |> IEC |> | C13 female plug end, so the GFCI adapter would need to be C14 to C13. |> | |> | We are finding when we aggregate a large number of servers with a single |> | GFCI that we get false positives and end up shutting down whole banks of |> | servers. So we want to try to isolate each GFCI to a single plug on |> each |> | server and see whether that improves reliability. |>

|> Do you have a reason to use GFCI on computer servers in the first place? | | The risk we wanted to protect against was someone opening the door to a | computer rack, accidentally allowing a power cord to slip into the door | seam, and then closing the door onto the cord, partly severing the power | cord and electrifying the rack fixture. Our hope would be that the GFCI | would trigger immediately upon severing of the power cord, thereby | (hopefully) preventing a risk of electrocution. At the same time, we | would want to localize the fault to a specific device and not take out the | entire server farm.

You should tie up the cords better so they don't get near the door.

Get a big roll of those Velcro loop ties. Use regular nylon wire ties to wrap around metal construction to make some loops ... one tie around the metal bar, and another tie looped through the first tie. This creates a hanging loop of the 2nd tie. Now loop the Velcro ties through this loop to hold onto the cords.

|> How many computers are involved? What is the required circuit capacity |> for |> serving all these computers through the one GFCI you are serving them with |> right now? | | Probably 5 to 10 servers per circuit. We seem to run about 10 to 20 amps | at 125V for those servers. | | |> What is the leakage current rating of the GFCI you are using now? |>

|> Is this GFCI an integrated outlet type, or a circuit breaker type? | | Unfortunately the GFCI we have today is integrated into the outlet, and I | don't know the ratings. I was hoping to just localize a GFCI to each power | cord to avoid the need to engineer a single common GFCI at the main outlet | to the rack.

GFCIs integrated into the outlet are the 5ma type. Maybe you need the 30ma type. Those are only available as circuit breaker types. The 5ma types are best suited for cases where the conductive path includes water or earth. The

30ma types are more suited for what you describe.

How many circuits do you have serving computers in that room, where a circuit is defined as having its own separate breaker/fuse in the panel?

Can you have a subpanel installed in, or very near, the computer room? One problem with using breaker panel GFCI is that the wiring run from the panel to the outlet contributes to the leakage a small amount.

Someone suggested using filtered power strips. Maybe a UPS on each could help, as well.

Believe it or not, Wal-Mart will sell power strips with GFCI plugs.

Just power each PC (or small cluster) with it's own power strip and you are home free.

The ones I have seen only have #16 wire so the continuous load should be kept around 10 amps although it's rated for 15 amps.

** Posted from **

| Believe it or not, Wal-Mart will sell power strips with GFCI plugs. | | Just power each PC (or small cluster) with it's own power strip and you are | home free. | | The ones I have seen only have #16 wire so the continuous load should be | kept around 10 amps although it's rated for 15 amps.

He seems to have a rack cabinet cluttering problem with just his cords (if they are getting in the way where closing a door could pinch one). Adding a power strip for each 10 amps worth of computers could make this worse.

It sounds like he needs a bunch of short cords. There are places that sell

1 foot long cords. These are good for rack cabinets with outlet strips that run vertical in the back on both sides.

And he needs to ground all the cabinets together solidly and connect them to a ground feed. Then if someone does cause damage to a cord, he can catch the culprit red-handed.

I agree that if you are just protecting equipment 30ma is more appropriate. Another possibility is an AFCI breaker - required to trip at 50ma ground fault but commonly made as 30ma. They might be available as outlets - I haven't seen any.

|> |> Do you have a reason to use GFCI on computer servers in the first place? |> | |> | The risk we wanted to protect against was someone opening the door to a |> | computer rack, accidentally allowing a power cord to slip into the door |> | seam, and then closing the door onto the cord, partly severing the power |> | cord and electrifying the rack fixture. Our hope would be that the GFCI |> | would trigger immediately upon severing of the power cord, thereby |> | (hopefully) preventing a risk of electrocution. At the same time, we |> | would want to localize the fault to a specific device and not take out the |> | entire server farm. |> | |> |> What is the leakage current rating of the GFCI you are using now? |> |>

|> |> Is this GFCI an integrated outlet type, or a circuit breaker type? |> | |> | Unfortunately the GFCI we have today is integrated into the outlet, and I |> | don't know the ratings. I was hoping to just localize a GFCI to each power |> | cord to avoid the need to engineer a single common GFCI at the main outlet |> | to the rack. |> |> GFCIs integrated into the outlet are the 5ma type. Maybe you need the 30ma |> type. Those are only available as circuit breaker types. The 5ma types are |> best suited for cases where the conductive path includes water or earth. The |> 30ma types are more suited for what you describe. | | I agree that if you are just protecting equipment 30ma is more | appropriate. Another possibility is an AFCI breaker - required to trip | at 50ma ground fault but commonly made as 30ma. They might be available | as outlets - I haven't seen any.

Beware, Cutler-Hammer makes AFCI with a 5ma GFCI integrated.

"Seems to me" that your racks should be well bonded to each other, other equipment in the room, and GROUND.

Back when I was working when we had "racks" of electronic equipment, they were routinely connected with copper braid that was 3/4" or wider.

Stuff like that can easily conduct enough current to trip a CB but in any case, in any fault condition they would hardly permit a rack to be "electrified."

That's quite reasonable.

Again, the power strips with the GFCI built into the plug will take care of your needs and trip quickly.

** Posted from **

It seems to me that your first priority ought to be to ensure that the racks are very well grounded. There are all kinds of failures (not just pinched power cords) that can apply line voltage to the metal case of a device in the rack, and thus to the rack itself, and you want to ensure that the rack remains at ground potential no matter what.

Once you do that, when you *do* get a metallic short to ground, it will blow a fuse or trip a circuit breaker quickly - and the rack will remain at ground potential. So you don't need a GFCI to protect against this particular risk, you need good ground connections to all the racks.

Once you have that, the only thing GFCIs are good for is to detect relatively low-current faults to ground through a human path. So what needs protection? Do people stick their hands into power supplies with power applied? Perhaps you can apply individual GFCIs in places where there is some plausible risk of human contact with line voltage.

I've seen small plug-in GFCIs. They plug into an outlet, and a cord plugs into them. Add a power bar with relatively widely-spaced outlets, and you should be able to run quite a few loads (e.g. 4-6) each with their own GFCI from a single building circuit. Here is an example:

Dave

I always wonder if it is safer for large metalic object to be grounded or to float.

If it is grounded, it could provide a lethal return path to someone who touches a hot lead somewhere and is also in contact with the grounded metalic object.

I think floating is safer if there is NO possibility of it becoming energized from a grounded supply. Otherwise grounded is safer. The problem with floating is that it could assume any potential to ground, even from static electricity buildup. That is why the secondary of step-down distribution transformers are grounded, to prevent them from being energized by leakage from the high primary voltage. Ungrounded isolation transformers are used but they are generally 1/1. Grounded objects and systems can be hazardous under certain circumstances but I think overall grounding is safer.

Don Young

Nothing will ever be totally idiot proof. Hot conductors are supposed to be sequestered away from casual contact. Grounding prevents large surfaces,such as a refrigerator case, from reaching dangerous voltage levels. The next step is the GFCI. But in many situations, you will get nuisance trips or reduced protection by increasing leakage levels before tripping.

Grounding requirements in electrical codes are there because experience has justified the practice. They are not there as a casual whim.

Bill

Suppose, as the OP stated, that a wire gets pinched and contacts the housing. The entire metal surface then becomes the "hot conductor". That is why we ground them!

On Sun, 31 Aug 2008 09:43:26 -0500 Ben Miller wrote: | john wrote: |>> It seems to me that your first priority ought to be to ensure that |>> the racks are very well grounded. There are all kinds of failures |>> (not just |>

|> I always wonder if it is safer for large metalic object to be |> grounded or to float. |>

|> If it is grounded, it could provide a lethal return path to someone |> who touches a hot lead somewhere and is also in contact with the |> grounded metalic object. | | | Suppose, as the OP stated, that a wire gets pinched and contacts the | housing. The entire metal surface then becomes the "hot conductor". That is | why we ground them!

I think most people are aware of the wire getting pinched and insulation breaking enoigh to energize the rack frame. What I think some are not aware of is that grounding the frame(s), beyond what is already done inside devices with a grounding conductor, is a suitable remedy. Previous posts have already suggested that in the case of appliances, connecting a metal frame to the grounding conductor is actually a hazard in itself. These posts say that the correct appliance would be "double framed" so that an inner frame is the one connected to the grounding conductor, while the outer frame that people can touch is isolated. This would protect people from the brief voltage that can be present at a fault event. So the question is, in the case of a device or appliance or equipment that is not "double framed", should the one frame be connected or not. IMHO, it should be. However, I also believe it should be connected through a separate ground path than the ground path used for various other faults, going all the way back to the bonding point. Also, to avoid any inductive effects, including the possibility of reducing fault current to a level below instantaneous trip levels, this additional isolated grounding path must follow the current supply conductors tightly. This would require a cable or conduit feeding power to the room to have a 2nd grounding conductor. This may even need a 3rd grounding conductor to keep direct equipment ground (the grounding pin of the outlets), and housing grounds (those that ground the metallic boxes and conduit) to be separate.

Now, considering all that (my opinion) above, what does your engineering education and experience tell you is really required, and what is just a waste of money providing no protection at all?

Here are my thoughts:

Theoretically, one could argue that an individual grounding conductor from each device back to the main bonding point would be the most reliable, since nothing that happens anywhere else could adversely affect any other device's ground integrity. Obviously, this is impractical in terms of the number of grounding conductors. You grouped things as appliances, raceways, etc, which is a compromise, and I am not sure it accomplishes anything. It reduces the number of conductors, but is still probably impractical. I am of course not addressing systems such as hospital patient care with isolated grounds, etc.

Current industry design practice in the US is to properly bond every exposed surface to a common egc system. Certain requirements, such as using pigtails at receptacles and switches, prevent the most common possible interruptions of that grounding system, but do not eliminate every possibility. Using green wires instead of relying on conduit connections alone is a big reliability improvement. I have not seen any studies or statistics proving that our current industry practice is inadequate based on the number of electric shocks/electrocutions on properly grounded systems. These accidents more often appear to be the result of improper grounding.

The "double framed" appliance concept is an interesting one, but certainly not one that can be universally implemented. I am not sure it is necessary if you prevent exposure to the live conductor to begin with. Remember that typically a fault occurs within the appliance. We already have double insulated appliances which accomplish the same thing except when they get wet! In that case, the "double framed" would offer an advantage of a grounded barrier to prevent a conductive water film on the outside.

Given a choice of using solid copper wire or a GFCI to protect a receptacle or appliance, the solid copper is preferred over an electronic circuit. The GFCI is acceptable practice where there is no copper ground (ie. older two-wire systems), as it is better than no protection at all.

The bottom line is that I believe the current grounding practices are adequate, when PROPERLY executed. I have seen older systems that were well grounded because someone took the care to do it right. The biggest issue I see is when poor workmanship or improper installation compromises the grounding system. Even in industrial environments there are "do-it-yourselfers" who don't have a clue. You won't eliminate this type of problem by designing a "better" more expensive and complex system that won't get done right anyhow.