Physical questions

Robert G. Brown rgb at phy.duke.edu
Tue Jan 14 13:12:19 PST 2003


On Tue, 14 Jan 2003, Ken Chase wrote:

> What 'absorbs' RF? (other than stealth bomber paint)

Lots of things.  Water, frequently, for one.  Resistive conducting
pathways in general, via induced eddy currents, in particular ones that
are the right size and that have the right orientation relative to the
field.  Hence lots of cages have holes and multiple layers.

All that EMF that doesn't come OUT of your case is eventually absorbed
INSIDE your case and turned into heat.  Just pure reflection (as is
likely enough to happen with sheet metal) can cause reverberation and
potentially positive feedback for resonant cavity modes UNTIL it is
absorbed by the gradual attenuation that occurs as the radiation
traverses matter inside the case and surface heating at the
reflecting/absorbing surfaces.

Microwave ovens work fine if there is water inside the cavity to absorb
the rays.  With no water or other material to absorb the energy, it must
be dissipated SOMEHOW, energy being conserved by natural law and all.
So either the reflective grid absorbs it, heating the grid and maybe
starting a fire, or it feeds back into the magnetron, heating IT and
maybe starting a fire or just breaking your microwave, or it leaks out
through the case and heats up you on the outside, maybe starting a fire
or breaking you.  

The same is true, on a slightly more modest scale, for computers and
other RF-emitting devices.  It's attenuated by nearly any kind of matter
at a rate with all sorts of interesting dependences on frequency and
mode of radiation and kind and orientation of the matter.  It can be at
least partially reflected by most surfaces.  Real RF (as in stuff with a
wavelength of meters to kilometers) tends to diffract around lots of
meter-to-kilometer scale obstacles and hence isn't well blocked by just
line of sight reflectors or absorbers.  Shorter wave stuff -- cm to m
(where a GHz is roughly 30 cm, 3 GHz roughly 10 cm) is relatively easy
to block and can be absorbed by conducting pathways on the scale of a
quarter of a wavelength, especially when aligned with the E-field.

I haven't looked, and it may well be classified (since they classify
silly things) but I'd guess that "stealth bomber paint" is just a
laminated layering of conducting pathways designed to absorb, instead of
reflect, e.g. cm wave radiation, probably not at all unlike the way
layers of polaroid films absorb and block visible radiation without a
lot of reflection.  Something like pressed steel wool, in a composite
matrix.  I'll bet steel wool layers in a ziplock baggie would make a
very good RF absorber, too...

> Our cabinet in the end with its smoked glass door with its futuristic
> rivetting/bolting pattern (no charge) looked *MUCH* nicer than a tower case.
> People see towers and go "hmm regular computer, possibly a low end server".
> When they see the Monolith (as it got dubbed) they're much more impressed ;)

No charge?  I'll take two...;-)

Seriously, one of a kind stuff like this is the MOST expensive.

I like to start any comparison of spending my time vs money with:

  "Hmmm, my time is worth (to me, anyway:-) $100/hour..."

followed by the

  "...and I'd be spending Other People's Money (OPM)."

My time or their money?  My time or their money?  Decisions,
decisions...I wonder.

Of course, I've spent hours just doing list mail today and not gotten
paid a nickel for it, so go figure...;-) Maybe the $100/hour figure is
just what I OUGHT to be making, or what I'd like to be paid to do
something I don't like doing anyway.

> How many clusters have gone up in flames? :) Not that this is a concern,
> it just suprises me that things could get that hot.

Its a Murphy's Law sort of thing.  I've smoked (literally, as in puffs
of smoke) a number of power supplies.  I've also smoked (literally, as
in a charred spot on the motherboard) onboard resistors that blew, IC's
and specific wires.  In a really bad short, one can melt or even
vaporize wires remarkably easily.  A 20 amp 110V socket can deliver
order of thousands of watts if you ask it nicely, BEFORE it blows a fuse
or breaker -- the only question is whether there is an accidental
condition within your system that can short out and draw that kind of
power and turn it into heat without blowing a "fuse" (which might be a
wire itself).  If you run enough systems, long enough, you'll see smoke,
and where there's smoke there's fire, right?

Or at least the danger of fire.  It would worry me to mount PC
components directly onto wood of any sort.  It would worry me to have
paper or other flammable material "on" or near the electrical
components.  My guess is both are prohibited by some sort of code,
although I'd have no idea where to find out.

High density clusters also carry a fire risk of their very own.  One can
easily achieve node densities that consume 2 or 3 thousand watts in a
rack (and with effort, can maybe double that).  All that heat has to
come out through the surface perimeter of the block of systems.  A
catastrophic failure of the cooling system(s) can cause the heat to
build up and be contained (being more or less uniformly produced
throughout the stack).  Systems in the middle are heated from below,
from above, and generate heat themselves.  If that heat isn't actively
moved out from the case, the temperature increases until the case gets
hot enough on the OUTSIDE to convey heat away as fast as it is being
produced and accumulated on the inside (or the system dies and stops
drawing power at all).

If it gets hot enough to melt insulation, or to increase the resistance
of critical conducting pathways, one can potentiate catastrophic and
sudden failures that can very definitely start fires.

A not quite disconnected story.  When I was two I stuck a hairpin
directly into a 110V socket behind a chair in our living room.  I
remember as clearly as day that I was pretending that I was plugging in
a lamp.

I was.  It glowed white hot within my fingers before vaporizing, blowing
the fuse, knocking me a couple of feet back, and giving me third degree
burns on every finger of my right hand (although I only have one small
scar from this, amazingly:-).

Oh yes, there is ALWAYS a chance of sparks, melted or vaporized metal,
and fire where there is a power supply and conductors in close proximity
-- that is, just about everywhere strangely enough.

(Now you can see why I became a physicist:-)

> Arent things generally considered ok if the high voltage (AC 110) is in
> the powersupply box? The boards are using 12 and 5v and lower, which 
> isnt regulated nearly as tightly. (Our cabinet was made of metal so
> im not too worried here, just asking for future 'build a test cluster
> on a shelf' designs that I am *SURE* _MANY_ beowulfers are not
> admitting to having running AS WE SPEAK! :)

Are you talking about the risk of electrocution (lower for low voltage)
or fire (where voltage is nearly irrelevant) Look at how much POWER your
power supply can deliver.  Typically, this will be order of 150-250
watts on certain lines (tens of amperes).

Watts, not volts or amps, is what matters.  Think of the filament of a
100W light bulb.  Pretty hot, right?  WHITE hot.  Would burn through or
melt anything wimpier than tungsten, with its phenomenally high melting
point.  

A simple, not horribly dangerous experiment.  Take an ordinary piece of
thin copper wire (e.g. magnet wire, bell wire) and short out the ends of
any 1.5V battery.  Although they are fairly limited in their ability to
deliver power, even a AAA can usually deliver enough to burn your
fingers holding the bare wire in a matter of seconds, and don't try this
with a 9V battery or even THINK of trying this with a 12V car battery
that you can carry around by its poles (with dry fingers:-) otherwise.

If you did this with two poles of a power supply, you'd get a BIG spark,
a flash of light, and yes, very likely third degree burns on every
finger that was touching the wire before it either burned through or the
power supply blew or a breaker blew.  With a car battery it is even
worse -- you could ignite hydrogen gas produced by electrolysis in the
battery chamber (with older non-sealed batteries) and blow it up,
spattering hot sulphuric acid all over yourself, in addition to creating
a MOST impressive spark as car batteries can typically deliver a LOT of
amps at 12 volts.

Needless to say I've been there, done that messing with jumper cables.
I've even melted and smoked the insulation clean off the ends of a a
pair of jumper cables in spite of the fact that they are a
half-centimeter thick, while trying to start a car with a really bad
(nearly shorted internally) battery.

Low DC voltage is NOT safe from a firestarting point of view -- it is
just less likely to be able to deliver a current (0.1 Amp) that can
defibrillate and stop your heart, like the one delivered to your home
daily at a frequency (60 Hz) that alas is particularly bad from the
point of view of heart-stopping.  Do NOT take it lightly when working
inside or on any computer system.

Note that homebrew mounting motherboards on a common conducting base
provides one with all sorts of interesting possibilities for a short
circuit if a wire, a solder trace, the wrong part of some component
comes in contact with the conducting base.  This is what the motherboard
spacers in a case mount are designed to prevent -- they keep the board
"far enough" away from the conducting case wall so that you don't
accidentally short it when e.g. pushing in a new card.  You can short
and blow a motherboard by just dropping a screw or a screwdriver it if
it lands dead wrong and manages to connect a voltage carrying trace or
connection point and any neighboring ground at the same time.

There is nothing like a UL rating, or rating from any of the various
agencies that rate such things, for electrical hardware.  It may not be
ENOUGH -- UL rated devices still fail and cause fires.  It is, however,
a lot better than nothing.  If an incompetently faulty case starts a
fire and burns down your house, maybe you can sue.  At least you can
blame someone or something other than yourself.  If an incompetently
faulty electrical installation in a homebrew cluster starts a fire and
burns down your house, who can you blame, but yourself?

   rgb

Robert G. Brown	                       http://www.phy.duke.edu/~rgb/
Duke University Dept. of Physics, Box 90305
Durham, N.C. 27708-0305
Phone: 1-919-660-2567  Fax: 919-660-2525     email:rgb at phy.duke.edu






More information about the Beowulf mailing list