hello... llhas anybody got the lastest buzz on what to dowith computer circuits? dedicated? GFCI? Arc fault? surge? what is the best recomended these days for office type situations ( 10+ computers )

Depends on how many under the desk space heaters you expect to have? 3 of 5 women will have one...

GFCI? Arc fault? You don't need any of these unless you are in the bedroom or bathroom. Surge Suppressor is always good to have. The ckt depends on how many you will have. You should be able to get the name plate data from the PC.

A lot depends on where this is happening. In office buildings with 3p wye feeders you worry about harmonics and you supersize your neutrals. If this is strip mall or residential on single phase, power quality may be more of an issue so surge may be the biggest issue. I would use surge protection at point of use and service entry in either place but that may just be a Florida thing. In real life anyone can take a lightning hit but it is a certainty here.
Be very careful if you get in a strip mall or other occupancy where they are joining a couple of separate units into one company. When they start stringing LANs around, you just bonded your panels via the LAN cards. Make sure all of the services you might be plugging interconnected machines into get their grounding systems solidly bonded together.

Seperate neutrals won`t hurt.

In my last job, we installed dedicated circuits specifically for PCs. 3.5Amps reserved for each PC (PIII-class w/ 17" monitor and speakers, and determined by actual measurement and not UL plates, I insisted on that!) 4 to a 20A circuit, period, no chance of overload. Separate convenience receptacles were provided for misc stuff; this isolated nuissance trips to a circuit that the computers were NOT on. All circuits were isolated receptacle circuits- there were dedicated panels just to the PCs (and other IT equipment), and all were either on isolation transformers or powered by UPSs with integrated isolation transformers, but this was for upstream protection reasons unrelated to the quality of power going to the PCs, as we never had any issues there.

My present office, we just load 'em up until they bitch about the circuit breakers tripping. (I didn't have any hand in designing this building, btw.) It's not too bad until they start plugging in space heaters, coffee pots and laser printers on the same circuit as 8 PCs...

Remember: not only do laptops draw very little current, they also have their own integrated UPS! Laptops and telecommuting should be encouraged in the office If the cost of power is factored in, the laptops actually cost far less. IT people never think about that until it's too late

[This message has been edited by SteveFehr (edited 10-26-2006).]

As mentioned, you're main concern really isn't supply to the PCs, as long as the line voltage quality is acceptably consistant. Power supplies for both PCs and audio equipment are extreemly well designed these days, and the engineers who develop this equipment take into account reasonable fluctuations in voltage, harmonics, and line spikes.

Server rooms will require specialized supply, isolation, distribution, backup systems, HVAC work, etc., but that's a whole diffent thread.

As Steve and e57 mentioned, your main concern is what ELSE people will tend to plug into the PC circuits. If possible, dedicate circuits specifically to the PCs and label them as such (then sit back and see how long this is actually observed by the user). This will be difficult in a partition farm with integrated outlets, but it can (and should) be done if the business owner finds the cost acceptable.

I assume they may be fixing the problem but there used to be manufactured office partition systems that used multiwire circuits in the cube wiring. Those were particularly troubling when this was hooked to 3pY. Usually it was the interframe connectors that burned up. In the IBM offices they ended up abandoning these built in outlets and running new circuits with single circuit MC cable. It takes away the portability but the reliability was worth it.

Bonding panels via the LAN cards?

As a matter of fact, twisted pair LAN cards are isolated (with isolation transformers) from ground to a minimum of 1500V per IEEE 802.3 speficiations.

Even on old thinnet cards with the BNC connector, the shield of the BNC is not connected to ground on the thinnet card, nor anywhere else (unless someone decided to ground it--this was never done in any thinnet installation I've ever seen). There were even plastic shields that could be placed over the BNC connectors and the "T" to ensure that it didn't touch the chassis of the machine.

I believe the isolation specs for thinnet are also tested out at 1500V.

"Bonding panels via the LAN cards?
As a matter of fact, twisted pair LAN cards are isolated (with isolation transformers) from ground to a minimum of 1500V per IEEE 802.3 speficiations."


The LAN output is an isolation transformer but if you have a high frequency transient it will go right through that transformer and the baseplate of the PC is tied to the EGC via the power supply case and DC ground via the grounding pads on the system board. That will not show continuity from machine to machine via the LAN but it will still couple a transient.
We had so much trouble in one of our installations (Holiday Inn Ft Myers) that we actually ended up running a bond wire from PC frame to PC frame, just to prove what was happening.

If it's high-frequency enough, it's RF and that can do all sort of strange things. What was the frequency of the transient? I wonder if a good choke on the power supply would have stopped it.

I should point out that the better computer power supplies usually include such chokes and filtering since they're usually required for the computer to pass FCC Class A or B testing.

The cheap power supplies, which also aren't UL or CSA listed, do not. Of course what do you expect when a 300 watt power supply costs $15--that should be a clue as to what you're getting, which is a piece of crap.

The RF filter is on the line inputs but the EGC is direct connected to the frame of the supply, which is connected to the baseplate and the system board grounds to the baseplate so any transients on the EGC will get reflected on all the signals. As long as all the EGCs are going up and down together you will be OK but if you span 2 services these transients get reconciled across the signal cables. There are a number of things you can do to minimize this, like the ferrites you see on parallel printer cables but when you get too far away or on different grounding systems the only answer is bonding around the signal cable. Start with bonding all the services and if that doesn't fix it bond the machines.
To avoid NEC questions we call that a "drain" wire and make it anything but green. (Black THHN was the usual choice)
The "drain" wire we had at Holiday Inn had some very ugly stuff on it on a sunny day. In a thunder storm it was really seeing a lot of nasty stuff. This was observed with a current probe connected to a scope. Try that sometime to see what is on your grounds.

I think I've seen power supply filters that pass the grounding conductor through a choke.

That would make the problem worse. The ground should shunt out the transient, not provide impedance that would allow it to propagate. Look again, I bet that line filter only uses the EGC for a reference

NJI ---

My practice is as follows:

Office computers draw 100-200 watts. Supply power with a 1500va UPS.

LCD monitors draw 20-30 watts.

There is no need for any other consideration.

Much computer equipment these days no longer even has a ground connection, which may solve the problem at least as far as that equipment is concerned.


[This message has been edited by brianl703 (edited 10-27-2006).]

Incidentally, it would seem to me that connecting all of the computer chassis together would cause a ground loop. But apparently it fixed the problem anyway?

It certainly creates a ground loop but that doesn't really seem to be as bad as coupling inter ground transients through the signal lines.

I wonder if ScTP (screened twisted pair) would've helped too.

As I recall the shield/screen on ScTP is connected to ground at both the NIC and the hub/switch end (if indeed the hub/switch is grounded--many are not).

The original DEC ThinNet specs called for the shield to be grounded at one point only, which was provided at the DEMPR (hub). You rarely saw grounding implemented on ad-hoc thinnet segments.

The ThickNet specs called for the shield to be grounded at one point in each building the cable passed thru.

We are not talking about the same thing as the shield on a transmission line. You are right, shielded wiring methods should only be connected on one end but I am talking about bonding that does not run with the signal. IBM went full circle on loops in the grounding to the chassis of machines sometime around 1980. We stopped talking about IG circuits on the line side and they stopped doing "baseplate grounding checks" on the machine side to insure a "star" ground. By 1980 you had to remove about 50 intentional grounds before you could get to that single point ground you were supposed to check. Somebody pulled their head out of the anal orifice and decided we were going for the "ground plane" model and everything gets bonded to everything. That ended up leaking out to the network world when we decided the same concept would stop us from buying $10,000 worth of cards everytime there was a thunderstorm. A lot of that development happened right here in southwest Florida.