[rescue] Prices... Amazing

Scott Newell newell at cei.net
Mon May 2 19:37:37 CDT 2005

At 05:09 PM 5/2/2005 -0400, Nathan Raymond wrote:
>On Mon, 2 May 2005, Rox wrote:
>> ... or "A reminder that sometimes it's easy to overlook the importance
>> of fundamental physical laws when you are dealing with the subtle
>> details of a complex problem."  ;)
>But nobody has made a reasoned case why all electrical energy going into a 
>given computer system is necessarily converted entirely to only infrared 
>radiation.  Spouting a mindless phrase like "laws of thermodynamics" does 
>not equal thoughtful reasoning or make a case for anything other than 
>intellectual laziness.

I doubt that I can, but as I feel that I started this little dustup...

Power supply efficiency: A real world non-perfect power supply is less than
100% efficient.  The loss is dissipated as heat.  You've got losses from
the resistance of the switching elements (FETs, IGBTs, bipolar transistors,
whatever), junction loses if you use diodes instead of sync rectification,
losses due to reverse recovery time in switching diodes, copper losses in
the magnetics, and so on.  (For more info, scan through some of the
*excellent* Linear Technology application notes...app note 35 is a good
place to start.)

Capacitance: You've got to add or remove charge on the gate capacitors to
flip a CMOS gate.  The faster you need that gate to switch, the more
current you must provide to get the cap charged or discharged.  Now, where
does the current come from?  A transistor.  Connect the gate cap to the
rail through the transistor and the cap charges, but the current also heats
the transistor.  Power dissipated is I^2R, so increasing the current to
charge the cap faster also generates more power loss in the transistor,
which is of course, heat.  (Same situation when you need to remove charge
from the cap to ground.)

Mechanical stuff: Sure, you do a little work when you spin up a disk.  But
once it's at speed, any energy to keep it spinning (at a constant speed) is
due to friction.  Got friction?  Then you've got heat.  What if you need to
slow down a spinning object?  Use something mechanical (like a brake) and
again, you're just converting energy to heat through friction.  Use the
spindle motor as a generator to load and slow the spinning disk and you
could recover some energy.

Did anyone else bother to run the numbers Nadine posted through their handy
HP-48 calculator and see that watts and BTU/hr are both measures of power?
(Break watts and BTU/hr down into the base units, kg*m^2/s^3.)

>From the Sun web site
"Servers and related equipment generate a considerable amount of heat in a
relatively small area. This is because every watt of power used by a system
is dissipated into the air as heat."

As to why two systems can pull the same measured current yet put off
different amounts of heat--I don't know.  I suspect measurement artifacts
due to the nasty non-sine nature of the input current waveform of most
switch mode power supplies.


More information about the rescue mailing list