[Phono-L] American Scientist Online Article

Ron L lherault at bu.edu
Wed Feb 13 08:20:41 PST 2008


Ron L thought you might be interested in this article from American Scienti=
st Online.=20

Found it.  Our Edison fans will like this.


---------------------Article Starts--------------------------
Published in American Scientist:=20

Vist http://www.americanscientist.org/template/AssetDetail/assetid/56694 to=
 view the article with illustrations


Edison's Final Revenge

David Schneider

The story of how our nation did away with gas lamps and adopted
electrification has been told many times. And why not? It's a dramatic
tale, with the larger-than-life Thomas Edison fighting for the
direct-current (DC) system he had built to power his light bulbs and
electric motors, while George Westinghouse championed the more
sophisticated alternating-current (AC) approach that Nikola Tesla had
devised. That Westinghouse's forces won this "War of the Currents" very
early in the 20th century is no surprise. The voltage of AC could be
easily transformed, allowing long-distance power transmission by virtue
of the fact that electricity sent at high voltage (and correspondingly
low current) suffers very little loss in the wires. Edison's DC system,
by contrast, required that the generating station be located within a
mile or so of where the electricity was to be used.

Though far less practical than the AC distribution system that soon
supplanted it, Edison's DC system did not die immediately. The power
utility that serves Manhattan, Consolidated Edison, continued for
decades to offer DC power to those who needed it--say, to operate ancient
DC motors in old elevator machine rooms. But Con Ed had been urging such
customers to switch to AC and, as of last November, it ceased supplying
DC power altogether. So Edison's brainchild, a system of distributing
electrical power as DC to equipment located just a short distance away
from the generator, is now completely dead--or is it?

In fact, Edison's concept is alive and well, particularly among people
who manage data centers. These facilities, which might belong for
example to an Internet service provider, typically contain racks of
furiously cooled file servers, which are set up to operate through short
power outages. These computers can continue to run because they are not
directly connected to the grid. Rather, they are fed by uninterruptible
power supplies (UPS), which contain batteries that are continuously
being charged off the grid. When the lights go out elsewhere, the file
servers draw their power from the center's many UPS batteries.

But batteries are DC devices. And file servers, like the computer that
sits on your desk, normally run on AC. So a number of conversions have
to take place: from the AC that the grid provides to DC to charge the
UPS batteries and then back to AC for the various servers. Actually, the
situation is even worse than that, because the output of the kinds of
UPS systems found in data centers is typically transformed to a lower
voltage before it is sent to the many computers. And within those
computers, that AC is converted to DC, and that DC is converted yet
again to low-voltage DC, at least once if not twice. So there can easily
be five or six power conversions between the grid and the circuitry
that's actually doing the computing work.

The inefficiencies of each of these conversions are small, but they add
up. A recent study of this issue sponsored by the California Energy
Commission found that for each watt used to process data, another 0.9
watt was required to support the upstream power conversions. And those
losses generate heat, so they exacerbate the problem of trying to keep
equipment cool.

William Tschudi, who heads Lawrence Berkeley National Laboratory's
High-Tech Buildings program, helped lead that study. Its roots go back
to work he and others had done earlier for the California Energy
Commission to ascertain the efficiencies of the various power
conversions carried out within the UPS units and power supplies of file
servers. "Somebody on the team said, ?What if we just eliminate some of
those conversions?'," Tschudi recalls.

Eventually, after much testing and experimentation, Tschudi and his
colleagues found the answer: By converting to DC just once, distributing
the DC power around a data center and stepping the voltage down as
necessary, the overall efficiency could be improved by 5 percent
compared with the very best AC equipment available. And compared with
more typical equipment found in data centers, the gain was 28 percent.

As this study was going on, a very similar exercise was taking place in
Sweden. The municipality of Gnesta, located near Stockholm, wanted to
provide high-speed Internet access to residents and local businesses as
a public service. The information-technology managers there thus needed
to upgrade their equipment. The five existing UPS units never worked
very well anyway, so they took the bold action of replacing them with a
UPS system that provides 350 volts DC, which is then fed to standard
server equipment. John =C5kerlund's company, Netpower Labs, provided the
necessary electronics. He characterizes the manager who committed the
town to this then-untested scheme "a brave man."

What's remarkable is that Gnesta's off-the-shelf computer equipment ran
just fine on 350 volts DC--or almost fine. The European standard is 230
volts AC, a number that refers to the root-mean-square value of the
sinusoidally varying voltage. The peak level is considerably higher, so
feeding a piece of European electronic equipment 350 volts DC will not
damage it. This strategy requires only that the connectors and fuses be
changed over to ones rated for DC; the built-in switching mode power
supplies typically work just fine. The worst hitch that =C5kerlund and his
coworkers discovered was that special protection circuits in some
equipment may detect that something is amiss with the power and either
prevent the unit from starting or perhaps allow it to run but report a
fault.  Before coming up with more sophisticated solutions, =C5kerlund and
his colleagues got around such difficulties by just ignoring the
bad-power alarms or, in the case of stalled gear, by merely pulling the
plug and plugging it in again. All of Gnesta's servers have been running
on DC now for nearly a year.

Similar change-overs are being tested in France and Japan. And the
telcom industry has a long history of running switching centers on 48
volts DC. Tschudi points out that one advantage of this approach, above
and beyond the energy savings, is that it allows a facility to run more
easily off of various DC sources, such as photovoltaic panels. And for
power-hungry installations, the notion of generating the power locally
is growing in popularity. Part of the reason is that waste heat from the
generator (be it a fuel cell, diesel engine or conventional steam
turbine) can be used to warm nearby buildings.

By combining the production of heat and power, facility managers can
squeeze much more useful energy out of the fossil fuels they use, so
this approach will certainly become more widespread as time goes on. At
the moment, this kind of locally generated power is AC, but perhaps soon
the advantages of DC will pave the way for its reintroduction in more
places than just data centers. If so, George Westinghouse and Nikola
Tesla will no doubt turn over in their graves .--David Schneider



---------------------Article Ends----------------------------

Copyright (c) 2003, Sigma Xi


More information about the Phono-L mailing list