Hi all,
I just put together my Athlon 64 X2 5200+ system (overclocked to around
3Ghz) and it appears the load was more than my Wintech 350W could handle (as
it died very suddenly).. It still supplies 5V standby, but it refuses to
turn on nomatter what.. The MB/CPU are up and running again with a spare
PSU.. but I underclocked it down to like 1.5Ghz so I didn't fry this one as
well.. (don't have a newer PSU so will need to order one).
I pulled the cover off and it gave off a weird smell, but it was only faint
and nothing appears burned.. (usually when a PSU goes there's remanents of
charcoaled components).. Where should I start in diagnosing the problem ?
(and no I am not going to just throw it out)
Regards,
Chris
So how much work are you willing to do on a PSU that, once
repaired, may have the same limitation as before and other
weaknesses that can similarly cause failure?
How comfortable and competent are you to take readings of a
disassembled, live, partially high voltage (340V or so) PSU?
I'm not saying "don't" but to a certain extent,
troubleshooting it requires a fundamental knowledge of how
it works and along with that - the suspicions of what may
have failed - weaknesses, or parts subject to power line
surges, and how to go about checking it already.
Usual suspects include the switching transistors and
rectifiers, those are on both heatsinks. Random failures
from other parts could be just that - random, anything
"could" fail, or could be due to poor design, poor layout,
cost-cutting, etc. Each PSU will have to be considered if
the most common things aren't wrong, how to try to most time
effectively check the rest.
Consider that Fry's has sold Wintech made Ultra V 500W, and
(don't recall if it was Fry's in this case) 400W, for free
after rebate. Radio Shack has also sold them for somewhere
between $10-20(?) after rebate. If a new higher wattage
version is available (to be clear, I would much rather
have/use other brands of PSU, but for the purposes of this
topic...) for such low cost and no wear on it yet, how much
time is it worth to repair, and pay $ for parts (unless you
happened to have a bunch of replacement parts on hand, but
if they're decent, that is also a monetary investment in the
long run to maintain discrete parts stock), just to end up
with the same thing that failed already?
Perhaps we can try to counter this failure mode with an
improved replacement (discrete) part, or combat the cause if
environmental in some way (like fan RPM too low), but is the
result a PSU reliable enough you would deploy it, and if so,
how many times would you repair it before deeming it
unworthwhile?
Sorry for drifting a bit, to be comprehensive you would need
a scope, and alternate parts to plug in, and to pull a bunch
of parts to test them. That's not always so easy with cheap
circuitboards that have thin copper reinforced by laying
down a wire on top of the copper or flooding the area with
solder. Removing the solder and part(s) can start lifting
up the copper, and it can then require multiple additional
steps to rectify this as you have flux residue on the copper
and board requiring an extra step to get rid of, so your
glue or epoxy can adhere well to hold the copper down again,
so there's a wash cycle and drying time too.
From what I've seen of some Wintechs, they used a fairly
active soldering flux and didn't get a lot of it off. Your
wash cycle needs to do this as getting it wet will make it
far more active and corrosive (to the PSU, it won't melt
your hand in any minor exposure). Such minor little issues
become more and more significant on certain PSU - it is
usually best to take a close look at what you would
troubleshoot and repair before investing the time. Maybe it
is a learning exercise or hobby and worth the time to you,
so I mention a few factors instead of that I would probably
throw it out or just pull the circuit board, crank up a heat
gun and shake the board, keeping any parts of value that
happened to fall out. By keeping some of these parts
without excessive labor to do so, you have alternates for
testing/repairing other poor PSU so you don't incurr so high
a cost to maintain stock of parts as mentioned above.
Buying all individual parts to make such a poor PSU, it
would probably cost over $50 just to populate the circuit
board.
After the visual examination of the top you might examine
the bottom of the circuit board as well. Solder balls,
lifted traces, cold joints, may all manifest problems later
in life. You may also see discoloration indicating thermal
stress on a part that led to it's failure. It could be
something very minor like a 3.3V load resistor had ran hot,
pulled off the trace, and now PSU shuts down because that
rail isn't within proper voltage range due to the system not
putting a load on it. We dont' have a description of any
basic voltage reading tests while it is hooked up to various
loads on 3V, 5V, 12V rails, only that it didn't turn on
(Hooked up to a system?). If you are going to continually
troubleshoot and repair PSU, a bit of initial work to build
suitable minimal loads to plug a PSU into, keep it running
knowing all rails the manufacturer specs as *needing* a
load, have one, and more significant loads to simulate
moderate system... could be useful.
A schematic would also be useful, so you can better
visualize what parts of circuits should have what kinds of
voltage levels. Otherwise you check what you can, trying to
trace back some of these oddly-routed and cramped budget PSU
can be headache inducing but if you can't find the problem
yet you will need to proceed *somewhere*, otherwise you end
up pulling off half the board components to check them.
Another alternative is to trace back the sensing rails on
the PSU - this I mean, is after checking other major things
like the transistors and diodes, and disconnect them from
their respective voltage/current sensing shutdown
subcircuits. By disconnecting them, the faulty one, the PSU
may remain running still and you have found the subcircuit
to more carefully examine- either back through the supply
rail like that dual diode pack, or forward through the
sensing circuit (less likely to fail, but parts can and do
just fail for no obvious reason in some cases, you will have
to decide how important it really is to fix before taking a
fix-it-no-matter-how-long-it-takes, stance). Obviously when
disconnecting these shutdown circuits, you don't want
anything of value hooked up as a load - so you use the minor
load jig/whatever you had set up for the purpose.
Also beware of multiple cascaded failure, one part shorting
puts more current into subsequent parts and it may be a
problem that kills more than just the first part that
failed. You will have to examine the particular circuit to
see if it is a factor, I can't directly predict what a
super-budgetized PSU may have changed to cut production
costs. If you are powering this PSU while testing, be sure
you do so safely. A low current isolation transformer is
not a bad idea.
Other things you can try if you plan on doing this a lot is
building separate supplies to generate signals or voltage
levels you can "plug into" the problematic PSU to see if the
subcircuits you are troubleshooting, respond as
expected/required. IE - if the PSU won't turn on to
generate the voltages, what if you put 5V on the sensing
line generated from another source- then you can see if the
zeners, transistors, etc are working, if the input at the
comparitor is as expected, and more. For these you will
need spend a lot of time tracing and reverse engineering the
subcircuit if you don't have, I mean can't get, the
datasheets for the PWM controller chip or see a subcircuit
topology you recognize- and recognition will come after a
lot more attempts to repair, it is slow starting out if not
one of the more common failure points.
I seem to have drifted around a lot in this reply, a lot of
things are more straightforward when you have the PSU in
front of you and can just do it, rather than trying to
speculate without being able to rule anything out yet except
the most obvious things like that the fuse can't be blown
when it's still providing 5VSB. For example you write that
nothing appears burnt but maybe with very strong light and
circuit board out, different angles, you might see something
on a component.
There are also some guides found by googling for various
applicable terms, for example "PC Power Supply Repair" turns
up this,
http://www.nutsvolts.com/PDF_Files/PSRepair.pdf
Googling for ATX schematics will find some using the typical
TL494 controller yours probably uses, and while it may
differ from the schematics, a lot of it will be pretty
similar. For example,
http://www.saunalahti.fi/elepal/vahvistin/Atx_power.gif
As expected you can to to TI's website and get a datasheet
for TL494,
http://focus.ti.com/docs/prod/folders/print/tl494.html
I think I've already typed too much and haven't covered
everything so in closing I suggest Googling for each thing
you want to do if you don't know how - like checking BJT
transistors or diodes with a multimeter, or whatever, and
remember that a failed $10 PSU is not worth risking
electrocution or a jolt that sends your finger colliding
with a sharp metal edge or your whole arm against something
that causes harm. Safety comes first.
