(from Alex McEachern) In this Power Quality Newsletter:

DC power distribution in data centers

When engineers look at power distribution in a modern data center, we see something silly: AC converted to DC (in the UPS), then converted back to AC (again in the UPS), then converted back to DC (in the equipment power supplies). Why not distribute DC, and avoid the inefficiencies of all these back-and-forth conversions?

Seveal highly respected companies from the U.S. and Japan are jointly considering a single DC voltage for powering data center equipment, probably be in the 350 Vdc ~ 400 Vdc range. There is a huge amount of experience in the world with -48 Vdc power distribution in telecommunications, but the power requirements are much higher in data centers. So a higher voltage, such as 380 Vdc, is needed to keep the currents and copper requirements in a reasonable range.

The arguments in favor of DC are:

There are some other minor advantages, such as simplified wiring (two conductors instead of 3 or 4 in AC delta or star systems) and ease of paralleling redundant backup systems.

But there are problems, too: What kinds of connectors should be used? What about interrupting devices like fuses and circuit breakers? Sustained arcs that would have self-extinguished during an AC zero-crossing? Magnetic fields and forces?

Power quality in DC systems People who are not familiar with power problems seem to think that DC will provide perfect power quality. I have been trying to correct this misunderstanding. There will be faults on any DC system, and loose connections, and capacitive loads turning on and inductive loads turning off. Also, I've pointed out that the data about power quality on -48 Vdc systems, which is very good, is probably not applicable because for the same number of watts, at 400V, we're talking about a source impedance that is probably one or two orders of magnitude higher.

If you would like to be connected with the people who are working on this project, please send me an e-mail at Alex@PowerStandards.com, and I will see what I can do.

Data centers
Will future data centers be powered from DC?

Back to top

Source impedance and power quality problems

For some reason, recently I've been getting a lot of questions from all over the world about power quality and source impedance. Here's a quick tutorial.

Sometimes it's useful to think of any AC power source as a perfect source of voltage that has an impedance inserted between the source and wherever you are. Think of this impedance as a series resistor and a series inductor. Sometimes the impedance can be mostly inductive (especially if the conductors are far apart, as on transmission and overhead distribution lines), or it can be mostly resistive (if the conductors are bundled closely together, as they usually are inside buildings).

When your load draws current, there is a voltage drop across the impedance, so your local voltage becomes lower. You can see this in the RMS voltage -- turn on a big load, and the local RMS voltage goes down. And you can see it in the waveform -- draw a pulse of current, and you'll see a notch in the voltage waveform. The bigger the current pulse, and the bigger the impedance, the deeper the notch (or, for single-phase electronic load currents, the flatter the peak of the voltage waveform).

Now imagine what happens when you try to draw a really large amount of current. Maybe you have a short circuit, or maybe you're placing a discharged capacitor across the conductors. What happens? Plenty of current flows, but how much? For resistive impedances, the amount of current is simply the source voltage divided by the source impedance. I've seen source impedances on a 120-volt 20-amp branch circuit in a building in the 4-ohm range, so you get 30 amps into a short circuit. And I've also seen source impedances on a 200-volt 20-amp branch circuit in the 0.02-ohm range, which gives roughly 1000 amps into a short circuit. It all depends on the source impedance.

So what happens if the impedance is too low? Well, you can get very large current flows - perhaps higher than the designer of the equipment expected (if the designer even thought about the problem). You see this problem when your first turn on a piece of equipment: fuses open, or circuit breakers trip, The initial inrush current in some carelessly-designed equipment is limited primarily by the source impedance, and if the source impedance is too low then the overcurrent devices will operate. The same problem occurs more frequently at the conclusion of a voltage sag, if the source impedance is sufficiently low.

But it's far more common to find a source impedance that is too high.

When you're looking at voltage problems at a load, it's fairly easy to tell if you actually have a source impedance problem.
  1. Turn off the major loads. If the voltage gets better (goes back to nominal RMS voltage, becomes a clean sine wave), then you're dealing with a high source impedance.
  2. Find the high source impedance problem by looking for heat.
    1. A power cable that is too hot, all along its length, has conductors that are too small.
    2. Loose connections almost always smell like burning insulation. Sometimes you can hear them buzzing or hissing.
    3. I've found loose connections inside a power connector by observing that a large power cable was hot just at one end...
  3. An overheating transformer is overloaded. Measure the current on its secondary conductors. (If you're drawing less current than the nameplate allows and it's overheating, you're probable dealing with a current harmonics problem.) Calculate the expected voltage drop, based on the transformer impedance.
If you have a voltage quality problem, and you can trace it to a source impedance problem, the fix will be easy and obvious, and usually cheap as well.

For an interesting, practical paper on source impedance, see J.M. Russell's paper Optimizing Mains Impedance: Real World Examples.

Bad connections and protective devices increase source impedance
Bad connections and protective devices increase source impedance.
(Photo courtesy of Bob Dettore)

Back to top

Upcoming power quality seminars - world-wide

Here's the general outline of most of my power quality seminars: McEachern Seminar Outline. If you would like to arrange a seminar, please let me know.

Alex McEachern's seminars
Alex McEachern's power quality seminars.

Back to top

Organizing power quality seminars in India - 2008

Can you help me organize a series of 1-day Power Quality seminars in India, for 2008?

P.Chandrasekar kindly suggested that the following organizations might be interesting sponsors for these seminars:

If you would like to help arrange a seminar ( seminar outline) sponsored by any of these organizations, please contact me. Thank you!

2008 Power Quality Seminars in India - can you help organize?

With best wishes -

Alex McEachern
Power Standards Lab
1201 Marina Village Drive #101
Alameda, California 94501 USA
TEL ++1-510-522-4400
FAX ++1-510-522-4455

Alex McEachern