Do you have a question? Post it now! No Registration Necessary.  Now with pictures!

Threaded View
I have a 4 or 5 year old Gateway (with Vista) here that a guy brought in
because it wouldn't turn on. Power supply checked out OK and there was an
amber LED lit up on the motherboard. I started taking stuff out and
disconnecting things and it finally took off. So with only half the memory
and no dialup modem in it I spent an hour or so updating stuff and checking
for malware. It was clean. So I shut it off and started hooking stuff back
up. Like the front USB ports, the CD Rom, etc. Then I put the rest of the
memory back in and finally the modem. Mostly all one step at a time. And
each time it restarted just fine. So now everything is back in it and it
still starts up OK.

I called the guy and admitted that I had no idea what had caused it but
that it was working fine. I set it aside and he came and got it later that
day. You guessed it. He took it home and it wouldn't start. He brought it
back and it wouldn't start. Finally he decides that he'll just go buy a new
computer (he's on the way as I speak).

Meanwhile, on a whim, I sat down with a hairdryer, took the side off, and
blew dried it (with heat, of course) for about 5 minutes. Sure enough, it
took off. I'm copying his stuff to an external drive right now.

Solder joint? Hairline crack in a trace somewhere? Who knows. Everything is
so microscopic these days it would be almost impossible to trace.

It seems a shame for him to have to buy a new computer, but what to do?
Obviously it isn't feasible to take the motherboard out spend time trying
to trace it. Then maybe it isn't a crack of a joint. Maybe a component. Too
bad I lose sleep over crap like this... :-)

                    --- Long live Fat32! ---

Re: Frustration

On 4/9/2011 12:23 PM, Menno Hershberger wrote:
Quoted text here. Click to load it

What did you you do to check out the power supply?

Re: Frustration

Quoted text here. Click to load it

I have a power supply tester, albeit a cheap one.
I also tried a known good power supply on it.

                    --- Long live Fat32! ---

Re: Frustration

Menno Hershberger wrote:
Quoted text here. Click to load it

OK, say you've established it's an intermittent. Tell your
customer that, explain the computer might not be stable, and
let him buy another. Customers want equipment that's going
to be reliable (or has good warranty coverage, such as cross-shipped

I could be a motherboard problem. The cheapest repair might be
another motherboard. You could test the components individually,
and determine whether they seemed OK or not.

BGA (ball grid array) chips, used to crack solder joints on the
corner. (I had a few do that.) As a BGA has larger and larger grids,
stress builtup within the array is a problem. Mechanical engineers
sometimes simulate the problem, and they've shown me diagrams of
where the failure is most likely to occur.

Now, after that point in time, I noticed that new grid patterns
had come out. The purpose of the modified patterns, is to remove
balls from perceived high stress areas. As a consequence of the
additional research, I would guess the problem isn't as bad as it
used to be. The grid arrays have become larger (making the problem
worse), but the compensation techniques are a bit better.

At the factory, they can use X-ray analysis to check for cracked
solder joints. The head on the machine, tilts at various angles,
as a single photo isn't sufficient to find problems. By combining
the results of the photos, you can spot cracks. That tells you
whether, say, a repair job was completed satisfactorily. Such
inspection might even be used on brand new boards (but I doubt
the motherboard industry is doing that - the technique is
"for rich guys").

It could be a cracked ball. It could be a cracked surface mount
components. I've had SMT resistors crack in half, and the fracture
only shows under a microscope. That's a factory problem too, and
I never saw any occur, say, a year after a product was manufactured.
The crack happened because of stresses when the component was soldered.
It can be easily detected with the microscope, but I don't know whether
the cameras they sometimes use for automated inspection, can resolve
quite as well as the low-power microscope can.

The tracks within printed circuit boards can also crack. They can
crack without any outside forces. I had 300 tracks fail in a board
which was sitting at room temperature in a storage cabinet. (It was
continuity tested at the factory, before being shipped to me, so
all the tracks were known-good on original delivery. The continuity
check was repeated later, as part of the forensic analysis, and
that's how we determine 300 tracks were cracked.) The PCB manufacturer
in that case, was blacklisted (they had a few other issues as well).

The PCB is seldom completely flat, when it comes off the production line.
Usually, the factory has a max spec for the bending, so if the warpage
is too great, the board has to be tossed. I never bothered to ask for
a breakdown of the consequences, whether clamping a warped board would
lead to premature failure or not, but I presume that was the reasoning.
The spec was tighter for boards with BGAs on it. Your Northbridge and
Southbridge are examples of BGA devices, and fairly large ones at that.

If your motherboard PCB was perfectly flat, then there would not be any
stress applied when the screws are tightened up. (I.e. Not pushing
down on a warped board.)

The various heatsink schemes, are stress points. Some heatsinks apply
extreme stress (to the point that it appears there is a dimensional
tolerance issue with the parts). It's one of the reasons I like screw
mounted cooling assemblies, as I can avoid cranking them down fully,
if I notice something else is "complaining". Being able to apply
variable force, allows an installer to compensate for mechanical
tolerance issues. (It's better than those Intel heatsinks with
the cam and lever arrangement.)

So take it apart, take your time, and either you'll decide to change
out the motherboard and recycle the thing. Or just part it out and
make someone else happy.

We used both heat and cold in the lab, in a similar way to what
you're doing. On some occasions, we'd use a heat gun, to find
things in a system that were sensitive. Or alternately, we'd
use "freeze mist" (a fluorocarbon at one time), to cool individual
components. Freeze mist can drop the component temperature to as
low as -55C. If you apply a generous blast of the stuff to certain
kinds of components (windowed EPROMs), you can ruin them instantly
(first hand experience). A light squirt may be enough to find a bad
solder joint, or a chip that is flaky (or even a timing problem).
The stuff tends to be a waste of money, because employees just waste
it screwing around :-) You can't have nearly as much fun with a
heat gun, so that's cheaper to operate.


Re: Frustration

Quoted text here. Click to load it


My money is on faulty ram.

If it has more than one stick of memory, I would try testing with each
single ram module at a time.


Re: Frustration

Sounds like a trace or track has opened up and the heat expanded it
enough to make connection.

Liked your commentary, Paul.

Re: Frustration

Quoted text here. Click to load it
Well, that doesn't sound feasible, because it should at least *try* to
start. But since I still have the computer here, I went ahead and tried
each stick one at a time, in two different slots.

No go. If I really wanted it to run again I would have went back in the
bathroom and gotten the hair dryer... :-)

                    --- Long live Fat32! ---

Re: Frustration

Quoted text here. Click to load it

That hairdryer would NEVER have gotten hot enough to reflow solder. You need a
heat gun to do so.

Quoted text here. Click to load it
At 5 to 6 years old, he will be much happier with a new computer. I bet that
PC was barely Vista capable!

Re: Frustration (GMAN) wrote in

Quoted text here. Click to load it

Expand and contract, not melt. Duh.

Quoted text here. Click to load it

He *is* much happier with his new computer (bought the same day I made my
original post.)
And as far as I'm concerned, *NO* computer was ever Vista compatible!

                    --- Long live Fat32! ---

Site Timeline