Signal Reference Ground Grids

[joan271273]

Can someone help me with information in how to size the signal reference ground grids usually installed in raised floor areas. IEEE and TIA conflict with each other in recommendations, but I would like to understand for example what factors do you use to size the perimeter ground for example.

 

[peebee]

Here's my understanding of the situation:

Frequency x wavelength = speed of light
or vl = c (sorry, no greek letters on my keyboard)
c = 3x18^8 m/s

Current microprocessors operate at about 1 GHz or faster. That translates to a wavelength of 3*10^8/1*10^9 = 0.3 meters or shorter (about 12 inches or shorter). If we are trying to damp out noise around that frequency, we need a SRG on the order of 1/3 wavelength maximum, meaning 4 inches or smaller, which implies using something like chicken wire or foil. The 2' SRG made sense twenty years ago, but became obsolete as soon as we invented the 200 MHz microprocessor.

Fortunately, we don't usually need to resort to foil or chicken wire, because the raised floor tiles are steel, and they work pretty darn well as an SRG Actually, the concrete and rebar also do a pretty good job, as does the real "ground", that is, the dirt. Data center networks are not that much different from most office networks, which work just fine with no consideration given to ground noise (usually). Capacitive coupling to the rest of the planet works pretty well as an SRG.

Putting some copper grounding beneath the data center floor can be nice for a few reasons, including that it makes for a convenient place to bond electronics racks. Also, the copper will help correct any high resistance connections in the floor grid. If you use this as your criteria for sizing, then whatever is convenient for you, 2' grid, 8' grid, perimeter conductor, busbar, whatever, is fine with me.

This is all very much a voodoo science, so I long ago gave up arguing for or against SRG's (the same as I also gave up arguing against isolated ground receptacles, but that's another story). At least you know the SRG won't hurt anything, and it doesn't cost too much so long as you don't go overboard with it (meaning, don't pony up for the copper foil unless you've got a great VC), and it will make the telephone company guys happy since they have a nice convenient grounding point for their racks. If you're trying to save a few bucks, leave it out, that's OK with me too.

 

[peebee]

oops, c=3x10^8 m/s, not 3x18^8m/s.

 

[alehman]

I agree with most of the above. There is little consensus on what is the best solution. Common practice (not necessarily right or wrong) for raised floor installations is to used the support grid as the SRG if the stringer connections are bolted at the joints. Then install a perimeter grounding conductor (#2 to #4/0 AWG depending on length) which bonds the support posts every 8 to 12 feet. It is also common in larger rooms to install a grid of conductors, similarly bonded to the floor posts, spaced 10 to 20 feet, each direction.

It's preferred for the perimeter/grid conductors to be bonded above the floor post leveling screws. If multiple rooms are involved, they should obviously be interconnected.

Another good reference (although officially withdrawn) is Federal Information Processing Standard 94.

 

[peebee]

So far as conductor sizes go,

The SRG can be about whatever size conductor you want. There's no NEC requirement to install the SRG and therefore no minimum code-mandated size. If you're building a grid out of building wire, then #8 is a nice size to use based solely on mechanical strength. Obviously, if you're using chicken wire, the conductors will be much smaller, and that's OK too.

Don't forget, you still need to bond your floor, building steel, electrical equipment, piping, etc. to your building ground system. NEC 250 is a good starting point for sizing those conductors.

 

[joan271273]

a) In the Color book series and the TIA stds there is reference for the perimeter grouding to be #1/0 or larger and the vertical/horizontal runs in the grid should be #6 or higher. How these requirements are calculated is what I am interested in understanding as the Telecomm std ( #3/0)contradicts the IEEE ( #2/0).

b) The wavelenght calculation is used to calculate the spacing of the vertical/horizontal runs.

 

[peebee]

a) Most likely, someone pulled these sizes out of their butt 40 years ago. Keep in mind that the currents handled by these conductors are very very low (unless there's a major design problem), so sizing has very little to do with ampacity and everything to do with impedance and imposed voltages at high frequency. That's all very specific to each installation, and very much more an art than a science.

b) Um, yes, uh, I think so. I'd consider them all to be horizontal, though. . . . Could you please clarify your question (is that a question or a statement?)?

 

[joan271273]

Thanks for the follow up peebe.

a) I found some mention on a book about SRG and the perimeter size as related to wavelenght and voltage drop between two points therefore requiring a size larger than #1/0 , go figure why? someone does not provide the calc for reference.
b) This was a statement.

 

[peebee]

Again: Most likely, someone pulled these sizes out of their butt 40 years ago. The diameter of the cable should have precious little to do with its high-frequency impedance, outside of some relatively inconsequential skin-effects. The current levels we are talking about here are very low, typically mA levels, certainly less than tens of amps.

Resonances due to length, inductance due to bend radii, and magnetic coupling have a far greater impact on high-frequency impedance.

By the way, here's another description of how SRG's work. I believe this is more consistant with what's stated in FIPS 94. By putting in a grid, you are ensuring that there are lots of different pathways to ground, each one being a slightly different length. Therefore, you are increasing the number of frequencies at which the SRG is resonant and therefore at its lowest impedance. I believe this is really the same analysis I first presented, just stated in a somewhat different manner.