[Techtalk] hardware issues

[Alvin Goats]

Since this question has a lot of people interested, I'll take you
through the issues from several points.


Regardless how much cooling you have, there is a thing called the
"temperature coefficient of expansion" (TCE or CTE) that still has a
major effect on the motherboard. Your CPU gets violently hot, heating
the board locally, making it expand. The rest of the motherboard is
cooler, and doesn't expand. The mechanical stress from the expanding PC
(printed circuit) board is very high, so the board will bend. The
bending or flexing of the board will eventually crack solder joints
(creating open circuits) or crack the copper traces or "wires" in the PC
board. Keeping everything in the box (computer case) at nearly the same
temperature is best.

Since most printed circuit boards in your computer are "multi-layered"
boards, you can't get at the broken circuits. The only thing you can do
is replace the board. Design workarounds use exotic ceramic circuit
boards that cost a lot of money (about 100x to 1,000x what your
motherboard does). 

Moisture will corrode aluminum in the presence of many resins. The
resins are in plastics. Plastics are resistant to water, but moisture
will eventually get through the plastic (time frame is in years). Most
IC's are plastic encapsulated. Most integrated circuits (cpu, memory,
flip-flops, bus drivers, etc) have aluminum bond wires welded to
aluminum bond pads on the IC and to the leads or solder bumps on the IC
package.  General time frame for failure in normal home use is 5 to 8
years; in a humidity and temperature controlled computer room, 10 to 15
years. Military and high reliability (commercial talk for a military
equivalent) designs take all the issues in this text into consideration
and the issues are minimized to a very great extent. This is part of why
they cost so much more than your home computer. 

ESD, electrostatic discharge, is a major issue for electronic failure
(as opposed to the mechanical failures of moisture and TCE above). Dust
collects and builds up a static charge. When the charge is high enough
to break down the dielectric force of air or the dust particles, a
"thunderbolt" of electricity will surge through the shortest, lowest
resistance path to ground. If you have a cpu rated at 3V power, a dust
particle discharge can be upwards of 1000V. If the discharge time is
long enough, it will stress the device, possibly even blowing the
device. Since most cpu's have fans, they don't fail as much, but RAM is
another matter. The discharge can kill a single memory cell, giving you
inconsistent values for the byte that contains that cell.

Dust has another issue: it acts like an insulator. So the heat building
up on your memory chips, cpu, power supply transistors and other hot
running devices can't get rid of the heat fast enough. So they get too
hot and are stressed. Think in terms of a car: dusty parts = no water in
radiator; no dust = water and coolant in radiator.

Power supplies for computers are typically "switching power supplies".
Depending on how the power supply fails, you either have an AC current
at the same voltage flowing through a DC path (bad news), no power at
all for that voltage, or you can have the FULL WALL SUPPLY flowing
through and killing everything. Fortunately, the major failure mode is a
lost voltage which really doesn't harm anything, followed by the AC
current at the same voltage which might blow a few things, but not
everthing. 

Power supplies get stressed from the AC power going to your house or
business. The power spikes on US 120V power lines can be as high as
100,000V (I have no idea what the source was, but it fried quite a few
machines that cost US$60k each). One of the things a UPS (uniterruptable
power supply) does is called "power line buffering". They filter the
spikes and brown outs before they get to your computer, reducing the
stress on the computer power supply. 

There are more reasons for failures, but these are the major ones (about
80% - 90%). 


Best suggestions: 

Blow out the dust in the box frequently. Twice a year if you don't have
a lot of dust, more often if there is. In areas where dust storms
happen, blow out the dust more often.

Use a UPS and, tie a very tight knot in your power cable. The complex
coils of the knot create magnetic fields that interact with itself and
will help reduce the strength of the spikes in the power line (it
approximates something called a "bucking transformer" that will kill the
spikes). 

Keep your computer in a cool, even temperature, dry, clean area. This
reduces the moisture in the air, keeps everything in the computer at a
lower, even temperature, and reduces most of the static discharge
issues. And I'm talking "dry" for the majority of homes. If there is NO
humidity at all, static discharge is worse. 

Technically, 45% to 65% relative humidity is good for static discharge
issues, 65 +/-5 degrees F (18 +/- 2 C) for the room. Room design
approximately a class 10,000 to class 1,000 clean room for dust and
particles. Now you know why the computer rooms are so cold and clean...
:)  And I'm sure everyone has a room in their house that meets this
criteria! <heh-heh> 


As for web sites talking about this.....I don't think you'll find one.
You just have to ask like you did. There are specifiations (IPC, ISO,
ANSI, Military, SAE, API, NASA, ...) that go into gory detail about
design, but not always the "why" of the design. This is the
responsibilty of the process engineer, product engineer, design
engineer, electronic/electrical engineer and mechanical engineer. And
each engineer has his own area of concern and generally knows little or
nothing about the other issues (EE's generally know nothing about TCE,
ME's nothing about static discharge, neither about moisture corrosion in
plastic encapsulated parts, ...). 


Alvin