Computer PSU Schematic ATX-250GTA

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

Threaded View
This is an SPI power supply (powerman branded), I need the values to
two resistors which have burned out.  They are located in the right
front corner near the corner mounting screw.  One is flesh pink, the
other green.  They are 1 watt carbon or metal film resistors.  The only
bands I can make out on the pink are black and red (brown?) located on
either end.  On the green one, the bands are red (brown)/black/red
(brown)/gold (Difficult to tell difference between the red and browns).
 Is there a place where the schematic can be found, or someone
otherwise advise as to the values?  I can get the identifiers, but
don't have them handy right now.

Re: Computer PSU Schematic ATX-250GTA

On 3 May 2006 06:18:58 -0700, wrote:

Quoted text here. Click to load it

What's nearby these resistors?

Did you identify the part that caused them to burn out, too?
Resistors don't generally just burn out, another component
failure puts too much current through them.  Result is the
resistor is the visual failure but something else caused it.

I don't think I have any of those PSU around here at the
moment but if you can manage to get a good hi-res picture of
the thing with the cover off (linked here, not posted to the
group), we might be able to do some kind of speculating
about possible culprits.

Are they just dark and you assumed they're dead, or are you
sure they are?  Resistors can look pretty charred and still
be OK in some cases... you might pull them out, ohm them
out, and see if what your meter reads could correspond to
what you can make out of the bands.

In a generalize sort of "anything is suspect" way, we can't
be sure what has failed but without visual cues it might be
a diode or transistor... on both of those heatsinks.  If you
hadn't pulled out the circuit board yet you might look for
discoloration of the traces on the bottom.

Re: Computer PSU Schematic ATX-250GTA

The power supply, Powerman ATX-250GTA, failed.  It is now only showing
power on the 3.3V rail, no 5V or 12V.  This power supply is apparently
made by SPI (Sparkle) and was sold under a number of different brands
(ie AOpen), usually with the same model number.  Upon inspection, three
3300uf/10V capacitors are leaking topside.  R47 (green resistor
currently registering 148.5ohm) and R48 (pink resistor currently
registering 6.8ohm) each had one leg of their leads corroded off.  You
can see the caps and resistors here:

or here:

The burn hole is at R48 (pink resistor).  The trace has lifted for
about .25" on the bottomside.  The only remaining bands on this
resistor are what appear to be red (but could be brown) and black at
the extreme ends.  Inner bands are faded. The green resistor has red
(again, maybe brown), black, red, and gold.  I'm not familiar enough
with resistors to know whether they are carbon or metal film (I'm
leaning towards carbon) and their dimensions seem to indicate they are

Re: Computer PSU Schematic ATX-250GTA

On 3 May 2006 17:13:58 -0700, wrote:

Quoted text here. Click to load it

The links are not working for me.  I get
the page then the timer then nothing... they probably expect
my browser to be able to do something insecure which it

The yahoo image loads, I see at the bottom left corner one
resistor near what looks to be a burnt spot.  The green
resistor I see canted at a funny angle looks like a 2W
carbon.  Not being able to see whole supply top-down it's a
bit hard to say but from the location it might simply be a
load resistor.  Before worrying about the resistors you
should replace the caps though, and you might check them
with a meter to see if any had shorted out.

Is this supply worth fixing though??  I don't know about
today but a year or two ago there were sellers on clearing out 250W SPI ATX for about $15
delivered, maybe even $10, I'm not sure about this latter
price.  Anyway, figure the cost of good low-esr caps, and
shipping if you dont' have them, and the time, and that the
rest of the supply has worn over time too, and it's value if
working might have been $5, worth less than time and expense
to repair.

Anyway, since you can see the trace on the other side of the
PCB, you could always trace that to find out more about the
subcircuit it's in.  The resistor values you measured are
probably right, but again resistors don't generally go out
without an external cause.  I wouldn't have thought the caps
shorting out would cause that but I am not sure of this.

Re: Computer PSU Schematic ATX-250GTA

The resistors were 4mm in width and 12mm long, so I identified them as
1W based on the information here:

Is that not correct?

I do understand the fact that a PSU can be had for around $20, this is
just a "project."

The resistor readings don't jibe with the band markings in anyway I can
see, but I'm not sure about the colors.

Re: Computer PSU Schematic ATX-250GTA

On 4 May 2006 04:17:17 -0700, wrote:

Quoted text here. Click to load it

Not all resistors of the same wattage are the exact same
size, you should consider the lead pitch (space between
holes on the PCB to determine what will fit if they were
horizontally mounted instead of tombstoned (standing up).

To give you an example, in my hand right now I have a 1W and
a 2W resistor.  The 1W measures roughly (body only, not
leads) 10.5mm L x 3.8mm W.  The 2W measures 14.5mm L x 5.8mm

"IF" the 2W resistors you find will fit, I would go ahead
and use a 2W, though your does measure more like a 1W.  The
resistor nearer the rear of the supply, next to one of the
failed caps is likely 2W, compare to it.

Quoted text here. Click to load it

It is curious that the leads are burnt, I'd wonder if they
shorted against each other somehow.  Seems like
Brown-Green-Brown is 150 Ohm, could this be one?

Re: Computer PSU Schematic ATX-250GTA

kony wrote:
Quoted text here. Click to load it

FWIW, the green resistor goes brown,black,brown,black (or is that

The bands I can make out on the pink one are violet (outer 2) There
might be 2 inner bands kinda visable, in that picture, but there's not
enough left to say for sure, Not gold, or green. More along the lines
of yellow or brown.

Re: Computer PSU Schematic ATX-250GTA

I spent some time carefully evaluating the colors on the resistor.  I
am certain the green resistor is brown/black/brown/gold which would
give 100 ohm.  It measures 148.5 with a meter.  The pink resistor's
first band is faded and could be either brown, red or violet.  Violet
is a good choice, as I measure 6.8 ohm and violet could give me 7 ohm,
but there is at least one other band that is completely faded,
definitely a very light color, but indeterminate.  The outer band is
black.  What an imperfect system, I see the value in schematics now.  I
was hoping someone had the same model PSU and could get the values off

How do resistor values change when damaged?  Do they usually go up or
down when overcurrent is applied?  I can see how they can go open and
leave infinite resistance, but what is the mechanism whereas the value
changes?  Why would one of the leads corrode or burn out like that?

I will probably order resistors in the different values and see how
they perform.  Would it be too simplistic to assume that if I get the
proper rail voltage, that the correct value is in place or maybe the
resistor's value is not critical?  I do not have a scope to evaluate
the quality of the signal...

Re: Computer PSU Schematic ATX-250GTA

On 5 May 2006 05:42:34 -0700, wrote:

Quoted text here. Click to load it

They usually go up by huge amounts if not going completely
open.   These types of resistors will not fail to a lower
Ohmic value.

I still suspect the values you measured are exactly what
they're supposed to be (within the tolerance for the parts,
a few percent).  It is not likely (common?) for a 100Ohm
resistor to fail and just go to 150 Ohm.  "Maybe" there is
some residue on the leads and you weren't making a good
electrical contact to take the reading, but then the reading
wouldn't be consistent if taken multiple times.

Quoted text here. Click to load it

It would be a separating of the metal end caps (to which the
leads are connected, under the outer covering) from the
element inbetween the two.

Quoted text here. Click to load it

Good question and is why I'd wondered if they shorted
together.  "Maybe" the solder melted, they were tombstoned
and one fell over and shorted against the other.  Did you
ever trace the circuit to see what their function was?
That's a very strong hint at possible failure modes.

Quoted text here. Click to load it

Find out what their function was... I still think load
resistor, and if so it is a fairly simple thing to calculate
the current they dump to ground based on the rail voltage,
and thus, also the necessary wattage rating (but still,
given enough space a 2W is not a problem being higher than
1W, is just more margin).

Quoted text here. Click to load it

Yes that would be too simplistic.  Again it is useful to
trace the circuit to see what their function was.  For
example it's likely they are just load resistors, and if so,
their purpose is to stabilize rails without much (or any)
load.  You can't properly evaluate output voltage without an
external load on the PSU anyway though, since the
manufacturers tend to always spec an external load is
needed... commonly it's about 2A.

I suspect you had a 2W 150 Ohm resistor on the 12V rail (to
Gnd) and the 6.8 Ohm 2W on the 3.3V rail.  What are the
ratings (stripes at least) on any other power (larger)
resistors on that side of the board (around general area of
the larger output filter caps and wiring harness-PCB

If all else failed, if you have the board out of the PSU
case you could flip it over, make a good scan of it and post
a link to that.  It would have to be higher resolution than
the other yahoo pictures though, high enough to discern and
follow the traces on the back and a corresponding top-down
picture of the top of the board.  We should be able to trace
that if the pictures are  high enough quality.  If yahoo or
whoever limits picture resolution too much, instead of
linking to an image display oriented site just post it (to
that other place, not the group) as a generic file, ie- on a
file storage supportive 'site if you have no personal
webspace to do that.

Re: Computer PSU Schematic ATX-250GTA

For some reason the pics are not showing up, I'll have
to re-post elsewhere for the record.  I will take a full board shot
also, as well as an underside and a close up of the trace lifting.

I agree that 148.5 reading is suspiciously close to 150 and would be
the proper value of that resistor, but the bands say otherwise.  What
if the band is not correct?  The green on other resistors is distinctly
different from the black and this IS a black band.  This resistor is
labeled as a 100 and is most likely really 150.  Now, what if the
circuit required 100 and a 150 was placed in it, could that possibly be
a reason why this thing went haywire?  Intuitively, a 150 ohm resistor
would draw LESS current, right?  This would seem to be the opposite of
what happened, but heck, I'm still in the early learning stages...

Re: Computer PSU Schematic ATX-250GTA

On 5 May 2006 09:32:22 -0700, wrote:

Quoted text here. Click to load it

Load resistors do by nature (within budgetized and
space-constrained designs) have the potential to get darker.
Even so, the surest thing to do is trace the circuit, see if
it isn't connected between one of the rails and ground.  If
that is the case, the value you use to replace it could be
off a fair amount and still be workable, providing at least
two conditions exist:

1)  That rail has a load from the system, you aren't just
leaving the PSU disconnected from everything and expecting
all the rail voltages to be *perfect*.  The PSU could even
shut down from voltage being out of spec if the value were
too high, or low but low is point #2;

2)  if the value were too low, it could cause a great deal
of wasted power, but as importantly could exceed the wattage
rating of the resistor.  For example if the 12V rail had a
10 Ohm 2W resistor replacing a 150Ohm, you'd soon have a

Quoted text here. Click to load it

Ok, but you can't necessarily compare colors from other
resistors, we don't even know if the same (resistor)
manufacturer made them, let alone used same lot of dye or
whatever other variables were involved.

Quoted text here. Click to load it

Of all the possible situations, I doubt this.  It is quite
uncommon to find a resistor mislabeled like that, very low
odds that it both happened, AND that the resistor happened
to fail as well (being a higher value than marked) such that
you would notice it.

Quoted text here. Click to load it

Not at all likey, particularly if it's a load resistor which
the odds favor.

We are not really going to be productive considering these
things though, not nearly so much as simply finding out
what/where in which subcircuit it was.

I had suggested a scanner instead of a camera because it
tends to provide a higher resolution and uniform lighting.
A high quality picture would probably be about 10MB JPEG,
25MB PNG, or too large for most to want to download as BMP.

Quoted text here. Click to load it

Pass less to ground, yes, but this is not a problem when the
PSU is in a system that makes use of the rail.  150 Ohm is a
rather low load if it were 3.3V rail, and the system itself
(just about ANY system) would be putting enough of a load on
the 5V or 12V that it wouldn't stop working or cause a
problem even if that resistor wasn't in the PSU at all...
but it's role becomes far more important when there is no
load at all, like when the PSU is out of system, unloaded
and then expected to produce near-spec voltages and pass
some current.

Quoted text here. Click to load it

The best way to do it is straightforward, consider the
general topology of the subcircuit and what that
subcircuit's purpose is.  From there, you would consider the
technical requirements of the resistor, what value would be
useful, what the wattage requirements would be for a
particular value, the amount of space to fit an alternate

It's conceivable that you could just throw in any very
roughly same ballpark ohm value and try it.  120Ohm, 240Ohm,
etc... but I wouldn't hook it up to any valuable equipment
till the resultant voltage is know.  Personally I have a few
very ancient hard drives I use as a load (among other
things), I just plug a couple in.  I took an old ATX
motherboard that was due to be thrown away, used a fine
pencil torch to melt the solder on the PSU connector and
pull it off the board (not esthetically perfect way to do
it, but the quick and lazy way can work too if you're
proficient at it).  So then I took a 1/2W 22 (?) Ohm
resistor and  tied it between the PS-On (pin 14) and a Gnd,
such that I no longer have to fiddle with a jumper wire or
paperclip, etc, to turn on the PSU... can just plug in that
salvaged motherboard connector.  I've done the same with
other motherboard connectors to make a plug with a (wired to
it) load for other tests but it is beyond necessary or
useful to go into any detail about it as it's simply enough
power resistors on each rail to draw a little under a dozen
amps from each, which also needs appropriate heat-mounting
and cooling provisions.

You can probably trace the circuit without going to the
trouble of taking the pictures.  Set the multimeter to
continuity test or just skip that and set to resistance and
measure between the burnt trace and the ground... any known
common ground point.  Measure between the other end (PCB
hole) trace, resistance to each of the power rails... even
to the ATX plug if it makes it easier for you, probably
going to be 3.3V, 5V or 12V.  (orange, red or yellow
respectively are the most common color codes and all SPI
I've seen did use them in a std. ATX PSU).

Re: Computer PSU Schematic ATX-250GTA wrote:
Quoted text here. Click to load it

Does nobody remember "Bad boys rape only young girls but violet
gives willingly"?

"If you want to post a followup via, don't use
 the broken "Reply" link at the bottom of the article.  Click on
 "show options" at the top of the article, then click on the
 "Reply" at the bottom of the article headers." - Keith Thompson
More details at: <
Also see <

Re: Computer PSU Schematic ATX-250GTA

CBFalconer wrote:
Quoted text here. Click to load it

They do, there's a flash ad that pops over the location you have to
click a few seconds before the timer ends. Once closed, you can open
the pic. I agree that that site is about 99% spam, though.

I will take a full board shot
Quoted text here. Click to load it

Yep. Politial correctness overpowers mnemonics, though....

Re: Computer PSU Schematic ATX-250GTA

Here are the relevant photos:

Shot of the PSU from the Front Topside

Shot of the PSU from the Filter Cap Side

Shot of the PSU from the Back

Shot of the PSU from the Back At R048

Spec Sheet for KA3511

Both resistors have one end grounded.  R47 (green resistor that reads
150ohm value) is traced back to PIN 16 on KA3511 (PIN 16: PT an Extra
Over Protection Input).  It also wanders out through a very small
surface mounted resistor (orange-orange-orange-gold, 3300ohm), turns
into a red wire that goes to the small board that is mounted on the top
of the right side heat sink (it is labeled "fan speed control").  R48
(pink resistor that reads 6.8ohm) is traced back to PIN 15  on KA3511
(PIN 15: Overvoltage Protection for Output 12V).  It too also wanders
around, goes through a jumper and comes back to KA3511 on PIN 14 (PIN
14: Overvoltage Protection for Output 5V).  Both the 12V and 5V power
are dead on the motherboard connector, by the way (the 3.3V is still
good).  I was thinking that these two resistors were shown in Section
6.5 OVP Block on the KA3511 spec sheet, with R101 being the 3300 ohm
resistor and R102 being the 150 ohm, but I don't think there is enough
information to verify the values are correct.

By the way, can alcohol be used to clean up the backside of the board
that has the traces?  If not, what is normally used?

Re: Computer PSU Schematic ATX-250GTA

On 6 May 2006 14:44:06 -0700, wrote:

Quoted text here. Click to load it

It is a single-sided (copper) board, there are no
surface-mounted parts at all, AFAIK.

Since it appears there was a part soldered on at the end of
the chip (it has 22 pins but there are 24 shown), it would
have been even more helpful to have indicated the pin
numbering, since there is no shot of the top of the area to

Do you think what you are calling pin 16, might actually be
pin 15??  That would make more sense.  150 Ohm load resistor
on 12V rail.

Quoted text here. Click to load it

You forgot to mention the thing that was important, that
besides wandering around and the other stuff, that trace
leads back to the main wire harness connection.  It is a
power resistor directly between one of the output power
rails and ground, that is the sole purpose of that resistor.

What I suspect happened, is poor design and cost-cutting
resulted in long-term heat accumulation on these load
resistors, separated the copper trace from the board (part
of why those who "can" choose to spec a build, will use
plated through holes to mount such things... but this was a
budgetized PSU, not to point fingers though as practically
all of them put in PCs, are.  Their second mistake was one
of layout, too little (and/or too thin) copper on the board
in that area to spread the heat, and/or cling to the board
so the entire trace didn't lift off.

Anyway, I suspect the two power resistors did short out
against each other.  It's also possible this strain has
damaged something else but for the time being you might as
well throw a couple 10 cent resistors in if you have them...
but again, I wouldn't go ordering parts online to fix this
psu, it becomes uneconomical to do it.

If you can fit a larger wattage resistor, it will run cooler
due to the larger surface area.  Since it is tombstoned, you
might also consider slipping some heatshrink tubing over the
exposed lead that is bend back towards the board from the
end of the resistor that sticks up.  Even better would be to
enamel paint the whole thing but for a couple mm at the end
to allow soldering and then put the tubing on too, but what
lengths you go to at this point are your call to make...
again this PSU doesn't have much value at this point.

Quoted text here. Click to load it

I also suspect this is one pin off, that it's pin 14.  It
too appears to trace back to another of the output wiring
harness bundles.  I think it was a 6.8Ohm resistor as

Quoted text here. Click to load it

It's not going to run till you connect back that torn trace.
I don't know how good you are (or what tools you have)
towards semi-precision work, so it's a bit hard to suggest
the fix that would be best for you there.  Given PCB drill
bits and a drill press, it'd be possible to drill out new
holes into the existing trace, scrape off some of the laquer
and solder a jumper wire in instead of trying to rebuild the
trace.  It does need trimmed off though, all of the loose
areas either glued down or cut off so there is no chance of
them later shorting out against anything adjacent.

It also looks like it "might" be helpful to move the ground
pad some... there is a lot of copper show to the left of the
board mounting hole in one of your pictures.  "IF" you can
move them over that direction some (hard to tell without
good picture of the top and the resistors there to compare),
drill a new hole and scratch off the laquer to solder in the
new spot, you could put the body of the resistor towards the
board there, 'sinking away more heat to that large copper
area but it might be too close to the lid... I can't tell.

The other end could be taken just about anywhere (again I
can't tell where the space is on the top), or put down
through the original hole... and bare wire used in lieu of
the damaged trace, and finally epoxied over as mentioned
elsewhere in my post.  It will have to be mechanically
secure and semi-heat-resistant though, I would not trust
multi-stranded wire for any mechanical fixing since there
will be thermal cycling (not that it would be the ideal
regardless of this but again, I dont' know what you *can*
do, feel comfortable doing, have tools to do, etc, etc.

Quoted text here. Click to load it

Doubtful, there is no call for a 1+W 3300 Ohm resistor in
these subcircuits.  Maybe to drain the main HV cap(s) but
that's about all, and this was not used for that purpose.

Quoted text here. Click to load it

Do you "need" to clean it?  Personally, I would not clean it
now, rather wait until after you have done the repair, as
then you will have more flux residue.

It depends on what the flux was, how to clean it.  Some
dissolve in water, some in water but have a petroleum base
that needs detergent, and some will need alcohol.  I
wouldn't worry about getting the hardened flux off though,
if it were conductive they would not have left it on the
board as they did, and if it were clean enough to eat off
of, that won't  help it run any better.

The trace is entirely lifted off, and the one where it
appears in place may have also seperated from the board as
well (hard to tell but it's not uncommon), so if that has
happened it is not a good foundation for soldering.
"Sometimes" the trace can be glued down again, taking a glue
applicator bottle with a fine tip and putting down a spot at
the edge of the trace (away from the exposed solder pad so
no glue gets on top of it), the glue will wick under the
entire trace.  Obviously the glue has to be thin enough to
wick and the area tacked down till it dries.

I wouldn't generally trust that with a somewhat larger
tombstoned part that gets hot, like a power resistor.  I'm
sure there's a technically "best" way to fix it but I would
probably tack down a piece of bare copper wire where the
trace was, looped where the holes for the resistors are,
solder in the resistors then apply non-conductive epoxy over
the entire area... and in that case, I would at least clean
off the flux and any residue at least on the area the epoxy
was applied to.  Epoxy (the type I'd use) is quite permanent
though, you would need be sure the repair is good before
covering it.

So far as the pictures go, I was suggesting a top-down shot.
You provided an angle that wasn't as useful.  The purpose
was to be able to look at two pictures side-by-side and
have them oriented the same way, with both practically 90'
from the board surface, a straight top-down view.  By having
this, the components on the top of the board can easily be
compared to the traces below.  Regardless, it seems these
pictures are not necessary now, it is just a suggestion for
easier troubleshooting should you ever have another similar
situation of sharing board & trace layouts with others.

I'm sure I've left out a few ideas, my post was interrupted
a few times by real life and besides that it was getting a
bit long already... but in short, yes they are load
resistors and their values determine how much load the PSU
has when there is no system connected.

In theory (probably in reality too), if you repaired that
torn trace, removed both of the resistors entirely and
connected a couple hard drives before turning it on (and
always had this load connected, or a system, whatever on
both 5 & 12V rails), the PSU would run ok without the
resistors.  You should not run it without them when there is
no external load (ie- the  2 drives/etc I just mentioned).

Don't take what I wrote for granted, double-check it to be
sure the pin #s were off by one.

Re: Computer PSU Schematic ATX-250GTA

Thanks for all of the analysis.  I am using this project as a learning
experience and following the comments has enlightened me on quite a bit
of the background I am going to need to further progress in this hobby.

On the closeup of the burnt resistor

the KA3511 chip is to the upper left hand of the burnt out holes.  What
you have pointed out, I completely missed.  This KA3511 chip has 24
pins and the spec sheet I posted has 22.  I was assuming Pin 12 is the
lower pin on the left row and this is incorrect, it is 13.  The chip
must be the KA3511BS, spec sheet here

Strangely, though, the chip on my board only says KA3511, but has 24
pins as you can see here

In this picture you can also see R45, the orange resistor I mentioned,
it is the bottom one on the right side of KA3511, orange banded. I
trace the green resistor around the hole for the pink one and then goes
through R45.

Regardless, I plan on replacing the three leaking caps and these two
resistors to practice soldering, better on a throwaway PSU than on
something of value.  Would 450F be the proper temp?  I am using some
pretty thin 60/40 with flux, I also have some silver solder, but will
have to apply the flux seperately.  I am using a Weller digital
station, WESD51.  My plan was to solder a jumper directly to Pin 16,
through the hole originally for the green resistor (pad missing),
attach the resistor to that and secure the assembly with epoxy, I think
this is what you are describing, but wouldn't the epoxy insulate the
resistor?  This doesn't sound desirable.  This is the 150 ohm resistor
and is "tombstoned," as you say, but the other pink one was mounted
flat to the board, its pads look okay.

As far as the caps, I was going to use Nichicon PWs, but those come in
12.5mm width, these caps are 10mm.  The tight fitting area will require
them to sit up some.  Is the proper technique to fill the underside
with epoxy?

I am thinking it is going to be very difficult to spot the locations of
the caps for de-soldering--that area is very congested.  What are the
techniques available for finding the correct pads? You mention the use
of a topside shot and backside shot.  I guess one can flip a side
(mirror image), print on a transparency, then overlay to get a rough
idea of the location?

Enamel paint for covering exposed leads, this is a specialized paint
for electrical work, I am assuming.

Re: Computer PSU Schematic ATX-250GTA

On 7 May 2006 14:48:04 -0700, wrote:

Quoted text here. Click to load it

Even when solder has flux in it, I often add flux,
particularly on aged parts or repair.  

IMO, you should avoid the silver solder since it has a
higher melting point, which is worse for even new/good
condition single-sided boards but in this case possibly
worse.  Upon looking again at the board you may need to
clean it a little, previously I'd suggested waiting till
after it was soldered, but you should not need to get the
old flux off, mainly the dust/soot/etc accumulation.
Alcohol should do well enough for that, just make sure it
stays on the effected area and doesn't leak into other
parts, especially if it is household grade with a lot of
water in it too.

60/40 melts at roughly 185C, you should aim a little higher
than that but whether 450F (232C is *ideal* or not can also
depend on the amount of heat the traces 'sink away.  In
other words the resistor lead to the traces themselves won't
need much more than 425F (only an estimation), but soldering
the other end to the ground pad will probably need a little
higher temp to be sure there's a good bond.

Quoted text here. Click to load it

I was thinking of securing anything loose or glued that you
put on the bottom of the board, and whatever is on the top
depends on what it is (of course).  Epoxy is bound to
conduct heat better than air so it wouldn't insulate but
rather help cool the resistor... but that is not it's
primary function and it also means a heat-resistant epoxy
must be used.  It need not be exotic, even JBWeld available
in local discount stores would probably do fine (but check
it's rated temp, I dont' recall it).

Quoted text here. Click to load it

If there really isn't room, not even if slightly shifting
some adjacent parts away slightly (sometimes possible,
sometimes not), it could be better to just use a 10mm
diameter part again.  If there are other adjacent load
resistors, be sure they are not in contact with the caps.  A
certain manufacturer (who will remain unnamed) made that
mistake in some of their units too but it wasn't SPI AFAIK).

I don't know that I'd call filling in the space with epoxy
proper since it's not really a common situation, tring to
fit one where it won't go.  How far out off the board is it
going to sit?  I wouldn't put it up very high, and IF I
decided to use something I'd probably take a piece of
plastic and drill a couple of correspondingly spaced holes
for the leads and use that as a spacer.  Remember that caps
are ideally supposed to vent out their tops but sometimes do
out the bottoms instead... it is best not to completely seal
either end with epoxy as would probably happen if it were
used here.

I can't see the minor  details of this board though, you are
the best judge of most things.

Quoted text here. Click to load it

It's usually pretty easy to spot the caps' leads.  While
there are some other misc. things in the area, there has to
be a fair amount of space for the wide current-carrying
traces and some things are obviously inductors, not caps,
from how large their leads are in diameter.

Since the caps are damaged anyway, one way you could do it
is to locate what you think is the correct lead, put your
finger on top of the cap and while heating the suspect lead,
try to gently rock the cap sideways with your finger.  With
one of the two leads melted, it should be obviously moving
sideways a little.  That usually isn't necessary though, and
if you want to make it more educational, again you can trace
back the circuit.  Note where the wiring harness main rails
are (since these are almost always the caps that fail), and
begin tracing it backwards.  There will be a cap between
each rail and gnd, usually an inductor then another cap.

You'll need your iron cranked up pretty high to get those
caps out, especially the ground on one of them looks like
it's buried in a pond of solder (on the ground, where the
wire is tacked down to beef up the trace).

It would be easier to locate the caps if the board were here
but those I've marked in red look like the right locations
at first glance, ~ 300K d/l;

Quoted text here. Click to load it

I don't recall seeing any for sale but I've never really
looked...  it looks like that's what it is  on some
resistors (those that are painted, but are rarer these
days).  If you have no enamel paint around for any other
purpose, don't bother with it.  A piece of heat-shrink
tubing would help still, if the tombstoned one were ever to
fall over which we hope won't happen.  Some manufactuers
have even taken a larger piece of tubing and slipped it over
the entire resistor.  Mainly it's just a matter of figuring
out why it failed the first time and trying to make sure the
repair is preventative of same/similar thing happening

Re: Computer PSU Schematic ATX-250GTA wrote:

Quoted text here. Click to load it

I use a solder suction bulb to slurp up large amounts of solder, such
as found around the power wires, and narrow copper desoldering braid
(Solda-Wik is one brand) for pins on components because it tends to
damage the circuit board less and is more thorough.  For single-sided
boards I normally use 25-30W, but for heavy wires I use 40W.  I have no
idea what temperatures I use because my irons are cheapos.

Quoted text here. Click to load it

I don't think so.  I'd rather use silicone rubber because it sticks
better to other plastic components, plus it's so soft that you can cut
it off if you have to remove something later on.  If you have to use
long leads on the capacitors, consider bending an "L" or "Z" here
they'll sit on the circuit board, for extra more mechanical strength.

For fixing circuit boards, I've had bad luckwith regular epoxy because
it melts when the soldering iron is used (apply it after soldering, but
JB Weld supposedly can take much higher temperatures, like 200F hotter.

Have you tried Digi-Key, Mouser, and B&D Enterprises for caps?  You can
usually find 10mm in brands like Nichicon, Rubycon, Chemicon, Sanyo, or
Panasonic.  OTOH 8mm caps in larger sizes seem to be rare.  Also sells caps as well as provides lots of advice.

Re: Computer PSU Schematic ATX-250GTA

I am going to go with mouser because they do not have a minimum.  I
have other supplies to get and they still don't add up to $25.  I did
not find a 10mm cap 3300uf/10V there, but I'll look again.

I took ESR readings with the Dick Smith meter and found that the
leaking caps have ESR of .24, .54, and .55.  They should be in the
neighborhood of .01-.02 per the Jamicon spec sheet.  I checked all the
other caps on the board and they check out ok.

Here is my explanation of what is going on in this PSU, please correct
or expand as necessary:

Power comes in and is rectified, stripped of its AC nature (might
involve 1N5406 or other general purpose rectifier).  The filter or
power supply caps (large pair of black caps on the left side) charge up
to a high voltage (300V, here I think).  There will be a resistor in
parallel with each to discharge these when the power is switched off.
This is a dangerous voltage.

Current then goes to the larger transformer, the one in the rear (SPI
"Hi-Pot") and this steps it down to the required voltages of
3.3/5/12/-5/-12.  There is always an inductor and a capacitor that goes
to ground for each voltage after the step down.  Also, a "load"
resistor for the no-load condition.  There are two smaller transformers
next to this one.  One of them steps down to supply power to the
control circuit. This is "unstabilized" voltage, not subject to the
PWM.  The other is for a second power supply of some sort.

Are the leaking caps in my PSU the ones that are paired with inductors
(they are 3300uf/10V and there are three of them)?  From the point of
the transformer to the output source, this is referred to as the

Are load resistors always "tombstoned"?  I guess my bad resistors;
at least the green one standing up was a load resistor? The pink one,
by the way, was lying down on the board.

KA3511 is the "brains" of the PSU, a PWM controller switching at
100 KHz, turning the power transistors that are the ones with the heat
sinks attached on and off.  The output voltage regulation is done by
controlling the length of on/off time, that's why it is called Pulse
Width Modulation.  These transistors are located just after the filter
capacitors and before the transformer, continuously adjusting the DC
voltage that the transformer sees.  This is happening only on the power
to the board (external to the PSU)

There is some circuitry  (small caps, resistors, etc) right where the
AC line voltage comes in and before the rectifier to prevent backfeed
of noise into the AC source.

There is some over voltage protection going on as well as a "power
good" circuit.

Re: Computer PSU Schematic ATX-250GTA wrote:
Quoted text here. Click to load it

You don't necessarily need 3300uF or 10V.  Since it's only 10V rating,
obviously it's not on the 12V rail.  Many manufacturers use a 6.3V part
here.  As for the 3300uF, one of the signs of a cap without enough
margin is that it has a suspiciously high mfd value but it's physical
size isn't any larger.  In other words, you should choose a 10mm
diameter part, 6.3V, the tallest one available which would mean the
highest mfd value available in that diameter with at least enough
voltage rating for that point in the circuit.

Quoted text here. Click to load it

After the chopper, transistor(s) or fets on the left-side heatsink that
switch it at high frequency, it is AC now.  Thus, much greater output
possible from the small transformer but also requires the rectification
mentioned below and the inductors.

Quoted text here. Click to load it

The second large heatsink, on the right side, has the rectification
diodes... TO220 or 247 typically, at least ultra fast if not schottky
since it's high frequency.  After rectified, then the inductors/caps.
Often it's not just inductor then cap but a cap before the inductor
too.  That appears to be the case on your supply, IIRC, two of the
failed caps are before and after an inductor on the same rail.

Quoted text here. Click to load it

I don't know about yours but the control circuit is often PWM
controlled too, or a linear regulator afterwards, not just unregulated.
 2nd supply for 5VSB, if they didn't use same transformer for both.

Quoted text here. Click to load it

Yes, but one seemed to be before the inductor too.  That is better but
inevitably real-estate and cost factors may cause it to be omitted in
many supplies, rails.

Quoted text here. Click to load it

Not always but it's quite common.  Putting a power resistor parallel to
the board takes up a lot of board real-estate.

Quoted text here. Click to load it

... continuously adjusting the AC, not DC, to keep enough current
through the transformer such that the average voltage out of the
inductor remains reasonably constant... until the caps fail, then that
"average" deviates too much in high/low values, if the caps didn't
short themselves out too.

Quoted text here. Click to load it

... mostly, but a secondary benefit of keeping that noise in the AC
source reduced too, and at least a little power factor correction
(yours does not have active PFC, at least what I saw of it didn't)
unless the supply were some other bottom-of-barrel generic brand that
skipped all this to save money.

Site Timeline