Grounding racks in Data Center with isolate ground

Status
Not open for further replies.

jmcr

Member
Somebody know what is the correct grounding method for a Data Center with a raised floor and isolated ground?
Where do you ground the racks?? If you connect the rack to the non-isolated ground, maybe interconnect the grounds, because the racks equipment (servers, switch's, etc) use the isolated ground and this one is connected to the metal enclosure of the equipment.
Furthermore, Where do you ground the metal supports of the raised floor??
 

ron

Senior Member
Re: Grounding racks in Data Center with isolate ground

Every engineer has a different opinion/method. It can be, and is done, in hundreds of different configurations.
 

dereckbc

Moderator
Staff member
Location
Plano, TX
Re: Grounding racks in Data Center with isolate ground

Like Ron who answered, I also design data center power and grounding system for technical equipment. I agree with Ron, there is no paticular standard method, rhyme, or reason. Every telephone/data company has their own way, no plan at all, and for some it just happens via plain chaos.

A lot of equipment manufactures are now using the ideas presented by IEEE Emerald book from chapter 9. The standard that was developed from the Emerald bood is IEEE T1.333-2001. However there are four diffent methods presented in T1.333-2001, and it takes great care in planning and design to pull it off

All depends on budget and expertise. Here is the basic big picture idea of how the largest telco in the world does it. Hold on to your hat.

It starts when the concrete is poured for the foundation. A 2/0 grid is laid in the concrete on 10-foot centers to form an Ufer ground. At each grid intersection under the planned raised floor, a stub is brought to the surface to be used later. Stubs are also brought to the surface at every location where electrical panels, transformers, motors, service switchgear, DC power plants, etc will be installed. .

Once the raised floor is constructed, another grid is constructed (called a raised floor signal reference grid, or SRG for short) of #2 or # 6 placed strategically under the equipment line-ups. This new grid is bonded to the Ufer grid at every stub under the raised floor. All floor pedestals, conduit, metal pipes, raceways, etc is bonded together to form a system call a common bonded mesh network (CBN for short). Any place a service panel, switch gear, transformer, etc is mounted, they are bonded to the dedicated Ufer stub, ground busses, and again to the nearest CBN.

When equipment racks are installed each equipment rack is bonded to the SRG via either a copper strap or a # 6 AWG jumper. The AC outlet installed either inside the equipment rack/cabinet on under the floor are the solid grounded type, not IGR. The outlets are dedicated, solid grounded types fed from a UPS PDU combination. The PDU?s are self contained isolation transformers with built in branch circuit distribution fed from a UPS. The PDU Xo lead is bonded to the SRG and Ufer grids.

I stress we do not use IGR because anyone knows (that?s a joke, more like no one) once you mount the equipment in the rack (metal-to-metal contact), plug the cord in, the IGR is compromised and rendered useless. Compound that by the fact once you plug I/O cables into the equipment with shields or grounded signal conductors like coax or RS-232, and again the IGR is corrupted.

We also use another method called Isolated Ground Planes in which all the equipment frames are electrically isolated from any contact with concrete, building steel, etc and form what is a called a ground window buss to ground all the isolated equipment frames radialy. But that is for special switching equipment and another long story.

So to answer your question directly, there is no answer.

[ September 03, 2005, 10:17 AM: Message edited by: dereckbc ]
 

LarryFine

Master Electrician Electric Contractor Richmond VA
Location
Henrico County, VA
Occupation
Electrical Contractor
Re: Grounding racks in Data Center with isolate ground

Originally posted by dereckbc:
It starts when the concrete is poured for the foundation. A 2/0 grid is laid in the concrete on 10-foot centers to form an Ufer ground. At each grid intersection under the planned raised floor, a stub is brought to the surface to be used later. Stubs are also brought to the surface at every location where electrical panels, transformers, motors, service switchgear, DC power plants, etc will be installed.

Once the raised floor is constructed, another grid is constructed (called a raised floor signal reference grid, or SRG for short) of #2 or # 6 placed strategically under the equipment line-ups. This new grid is bonded to the Ufer grid at every stub under the raised floor. All floor pedestals, conduit, metal pipes, raceways, etc is bonded together to form a system call a common bonded mesh network (CBN for short). Any place a service panel, switch gear, transformer, etc is mounted, they are bonded to the dedicated Ufer stub, ground busses, and again to the nearest CBN.

When equipment racks are installed each equipment rack is bonded to the SRG via either a copper strap or a # 6 AWG jumper. The AC outlet installed either inside the equipment rack/cabinet on under the floor are the solid grounded type, not IGR. The outlets are dedicated, solid grounded types fed from a UPS PDU combination. The PDU?s are self contained isolation transformers with built in branch circuit distribution fed from a UPS. The PDU Xo lead is bonded to the SRG and Ufer grids.
It sounds like you intentionally create so many "ground loopd", that the overall point-to-point impedances are so low (short-circuited, if you will) that no voltage differences can occur. Infinite parallel paths.

Instead of noise induction from potential-induced currents, it's all shunted, or bypassed, directly to earth. There can be no chassis-to-chassis potentials. Sort of like a equipotential ground mat under a sub-station.

[ September 04, 2005, 01:14 PM: Message edited by: LarryFine ]
 

dereckbc

Moderator
Staff member
Location
Plano, TX
Re: Grounding racks in Data Center with isolate ground

Originally posted by LarryFine:
It sounds like you intentionally create so many "ground loopd", that the overall point-to-point impedances are so low (short-circuited, if you will) that no voltage differences can occur. Infinite parallel paths.
Correct, it is called a "mesh ground". The size of the loops determines the frequency. The idea is Point-to-Point impedance, do not really care about the impedance to earth, that is a by-product.

Keep in mind none of this is done to replace the required EGC, as the mesh ground is not intended to clear an AC fault. It is used as a fault path for DC in some designs, but I supply a seperate frame ground for DC.

The other method used is a "single point isolated ground" in which no current flows, but that is another chapter and only used by certain equipment.

[ September 04, 2005, 04:16 PM: Message edited by: dereckbc ]
 

iwire

Moderator
Staff member
Location
Massachusetts
Re: Grounding racks in Data Center with isolate ground

Originally posted by LarryFine:
Instead of noise induction from potential-induced currents, it's all shunted, or bypassed, directly to earth.
You need to forget about the earth, as Dereck stated the connection to earth is incidental in this installation.

The earth is not a sponge for unwanted currents.

Any current has only one place to go, back to the source.
 

jmcr

Member
Re: Grounding racks in Data Center with isolate ground

Thanks "dereckbc". You're answer is very consistent and useful for me.

JMCR
Electrical Engineer
 
Status
Not open for further replies.
Top