Switching PSU trouble.

[John]

I'm servicing an old PC switching PSU.
It's odd form factor and very small dimension prevents me to substitute the
inner board and forces me to repair...
It gave 23V (!) on the +12V, +3.81V on the +5V, -5 and -12 were OK.
At first glance, I noticed a resistor with a dark brown colour, 1/4W, R1
across the +12V (red-red-burnt-gold-white).
Not considering the fifth stripe (white) it could be 22/220/2200.
But 22 Ohm will dissipate about 6.55W, too much for 1/4W.
Near R1, there's R2, (brown-black-brown-gold-white), 100 Ohm 1/4W
I replaced all capacitors. Almost all of them were leaking (ESR>5), and
one was severely burnt on the bottom.
Now the oddities.
Initially I thought that R1/R2 were 5 band resistors, so I put R1=22 Ohm,
R2=10 Ohm.
The PSU came up with +5V and 11.8V, but the PCB near R1,R2 was very hot.
Then I put the correct (?) value for R2, 100 Ohm.
The voltages were correct again, and R1 suddenly smoked (6.55W on 1/4W).
Increasing R1 (on the +12V) leads to Vcc of the 3844 (PWM) under startup
value.
I put back the original R2, measured as 97.7 Ohm, leaving R1 at 22 Ohm.
Vcc again under minimum value.
R2=100 Ohm, holy smoke from R1.

Odd, because R1=22 Ohm, R2=10 Ohm worked...

So, I'm missing something.
The original resistors were 1/4 W, across 5V and 12V...
I don't think R1 can be 22 Ohm...

Initial condition was this
http://www.supervinx.com/temp/01.JPG
----
http://www.supervinx.com/temp/02.JPG
http://www.supervinx.com/temp/03.JPG
http://www.supervinx.com/temp/04.JPG
http://www.supervinx.com/temp/05.JPG
http://www.supervinx.com/temp/06.JPG

 

[[email protected]]

I'm servicing an old PC switching PSU.

From the pictures, it looks like R2 (100 ohms) is across one of the
outputs - probably the +5 V output (red wires) - and 0 V / chassis
ground. This gives 0.05 A through it and 0.25 W dissipated. If it
really is a 1/4 W resistor, this is right at the limit, but
believable.

R1 looks like it is across the +12 V output and 0 V / chassis ground.
My guess is that R1 and R2 are there to provide a bare minimum load
on the +5 V and +12 V outputs.

Since R2 seems to have been OK originally, and it measured close to
the 100 ohm value you would get from reading the brown-black-brown-gold
bands and ignoring the white one, it seems reasonable that R1 would work
the same way: red-red-?-gold. You already found out that 22 ohms for R1
doesn't work. 220 ohms for R1 would give 0.055 A through it and 0.65 W
dissipated - too much for a 1/4 W, but maybe believable. 2200 ohms for
R1 would give 0.0055 A through it and 0.065 W dissipated - fine for a
1/4 W.

Have you tried measuring the original R1?

At a guess, I'd put in a 100 ohm, 0.5 watt resistor for R2, and maybe a
220 ohm, 1 watt resistor for R1. If you don't like how hot R1 gets,
maybe 2200 ohm, 0.25 watt for R1. I guess that a 22,000 ohm resistor
for R1 won't draw enough current to really do any good.

Matt Roberds

 

[[email protected]]

First of all I would bet that PS was built or at least designed by Phillips, or Philips depending on if you are a NAFTA ASC or not. It has that look. But to the point...

"It gave 23V (!) on the +12V, +3.81V on the +5V, -5 and -12 were OK."

That is alot more onformation than you might think. In this case obviously the 5V line is the sense line, the one used for referenc, not unusual. The one 12V line is high becasue the 5V line is low and not giving proper feedback to the regulation loop. Most likely in that 5V supply you will find thefirst lytic in a pi filter with high ESR, or a recifier with low efficiency. I have had that happen, use the scope, do not trust the DVM. The only thing I can figure is that some high speed rectifiers are dual diodes, one ishigh speed but then for the rest of the (usually) square wave it rectifies, a diode with a bnigger, slower die takes over. One that can dissipate better. This happened to some of the MUR (I think) series. However my money ison that first filter.

The other supplies that read normal are obviously subregulated. That is there are other regulators on the lines. This MUST be true because there is only one transformewr and the windings do not change. The turns ratio is the same as it ever was.

Given this, your first task is to find out why the 5V line is low. Of all the different branches of a PS, almost always, only one is sensed. That is why one is high and the other is low. Right now the subregulators are holding up on the negative supplies.

Any other damage is subsequent to the overvoltage condition and you probably should not have removed any resistors. When you see a burnt resistor, treat it like a fuse. It did not burn itself up, something else did. Once you get the rest of the circuit running properly THEN you deal with that. It means nothing now. The resistance value in that range might just burn with overvoltage, especially if it happens to be feeding a Zener, but not when theunit is in regulation.

In short, FIRST THING, fix the 5V line and see the 12V line go down. If it is for a PC, that is not subregulated usually. Just hope it didn't fry the thing it supplies.......Double the voltage makes quadruple the power, it isnot good for anything.

It really is hard to tell certain things without being there.

Get that much done and get back here and then deal with the rest. And also,consider options like Jeff L said, if it has a standard AT or ATX connector, hell something can be duct taped to the back if it is that important, unless the environment precludes that.

Either way, don't be afraid of that standup black thing, all it is likely to be is the voltage reference and current limiting reference and drive circuitry. It is not likely to be at fault. If it is, it can also be dealt with, but you have to prove it first. If ALL voltages were low or high, maybe, but not with these conditions. Don't worry about it for now.

 

[supervinx]

Il Mon, 17 Jun 2013 03:13:08 +0000, mroberds ha scritto:

From the pictures, it looks like R2 (100 ohms) is across one of the
outputs - probably the +5 V output (red wires) - and 0 V / chassis
ground. This gives 0.05 A through it and 0.25 W dissipated. If it
really is a 1/4 W resistor, this is right at the limit, but believable.

R1 looks like it is across the +12 V output and 0 V / chassis ground. My
guess is that R1 and R2 are there to provide a bare minimum load on the
+5 V and +12 V outputs.

Since R2 seems to have been OK originally, and it measured close to the
100 ohm value you would get from reading the brown-black-brown-gold
bands and ignoring the white one, it seems reasonable that R1 would work
the same way: red-red-?-gold. You already found out that 22 ohms for R1
doesn't work. 220 ohms for R1 would give 0.055 A through it and 0.65 W
dissipated - too much for a 1/4 W, but maybe believable. 2200 ohms for
R1 would give 0.0055 A through it and 0.065 W dissipated - fine for a
1/4 W.

Have you tried measuring the original R1?

At a guess, I'd put in a 100 ohm, 0.5 watt resistor for R2, and maybe a
220 ohm, 1 watt resistor for R1. If you don't like how hot R1 gets,
maybe 2200 ohm, 0.25 watt for R1. I guess that a 22,000 ohm resistor
for R1 won't draw enough current to really do any good.

Matt Roberds

R1 measures 22.1 Ohm .

 

[supervinx]

Il Sun, 16 Jun 2013 20:38:21 -0700, jurb6006 ha scritto:

First of all I would bet that PS was built or at least designed by
Phillips, or Philips depending on if you are a NAFTA ASC or not. It has
that look. But to the point...

It's labeled SeaSonic.
It had been connected with the P8 & P9 connector reversed.
Yes, the black wires were centered, but connectors were flipped
upside down. The MB works fine with another PSU: that's unbelievable...

That is alot more onformation than you might think. In this case
obviously the 5V line is the sense line, the one used for referenc, not
unusual. The one 12V line is high becasue the 5V line is low and not
giving proper feedback to the regulation loop. Most likely in that 5V
supply you will find the first lytic in a pi filter with high ESR, or a
recifier with low efficiency. I have had that happen, use the scope, do
not trust the DVM. The only thing I can figure is that some high speed
rectifiers are dual diodes, one is high speed but then for the rest of
the (usually) square wave it rectifies, a diode with a bnigger, slower
die takes over. One that can dissipate better. This happened to some of
the MUR (I think) series. However my money is on that first filter.

The other supplies that read normal are obviously subregulated. That is
there are other regulators on the lines. This MUST be true because there
is only one transformewr and the windings do not change. The turns ratio
is the same as it ever was.

Given this, your first task is to find out why the 5V line is low. Of
all the different branches of a PS, almost always, only one is sensed.
That is why one is high and the other is low. Right now the
subregulators are holding up on the negative supplies.

There are a LM340T12 and a LM7905CT.

Any other damage is subsequent to the overvoltage condition and you
probably should not have removed any resistors. When you see a burnt
resistor, treat it like a fuse. It did not burn itself up, something
else did.

But with R2=100 Ohm, R1= 22Ohm it seems to work (voltage are OK, under
dummy loads) but R1/22 Ohm across 12V burns quickly, of course...
The pair R1/R2 seems to determine somehow 3844's Vcc startup value.

Once you get the rest of the circuit running properly THEN you
deal with that. It means nothing now. The resistance value in that range
might just burn with overvoltage, especially if it happens to be feeding
a Zener, but not when the unit is in regulation.

In short, FIRST THING, fix the 5V line and see the 12V line go down. If
it is for a PC, that is not subregulated usually. Just hope it didn't
fry the thing it supplies.......Double the voltage makes quadruple the
power, it is not good for anything.

Now, after having replaces all capacitors, voltages oscillate, without
going over default values... i.e. 5V oscillates between 1 and 4V, 12V
between 6 an 8. 3844 Vcc value is under the minimum value.
The 23V on the +12V was initial, when I powered the PSU for the first time.
Just after having replaced all capacitors it went down.

Either way, don't be afraid of that standup black thing, all it is
likely to be is the voltage reference and current limiting reference and
drive circuitry. It is not likely to be at fault. If it is, it can also
be dealt with, but you have to prove it first. If ALL voltages were low
or high, maybe, but not with these conditions. Don't worry about it for
now.

I cleaned a bit the black card: it holds a 3844 PWM, a photocoupler, some
diodes, resistors and capacitors.
Two electrolitic capacitors were high ESR. I had to guess the values,
since the black paint (yes, it had been painted) made a correct
identification impossible.

 

[supervinx]

Well...
I put back the original resistors and, after some testing, found the first
problem.
The dummy load I used was not enough to drive the PSU.
So I connected an old MB, and the PSU came up with the correct voltages...
but...
-5V and -12V were 0 !
Removing the P9 connector, results in the -12V back.
I noticed, this morning, that the LM340T12 had a desoldered pin, and I
soldered it back.
After less than a minute, with the PSU running connected to the MB, the
LM340T12 were volcano-like hot. It's mounted back to back to the
LM7905, so they were both hot. But the LM340T12 was desoldered again.
I replaced the LM7905 with a spare (its pins were rusty, oxydated, corroded
and burnt ) but have only a doubt: can I substitute the LM340T12 with a
MC7812?
Lookin' at the datasheet it seems they're equivalent (BTW, the
LM340T12 has another marking, 7812).
----
The R1 and R2 resistors are dummy load, but not enough to let the PSU
start without any external load. I found a similar behaviour in some IBM
original PSU (5150 and 5155).
R1 is slightly burnt, I think, because of the 23V on the +12V, when the
PSU was malfunctioning...
Now the area around R1, R2 is warm, not hot...
-----

 

[supervinx]

Removing the P9 connector, results in the -12V back. I noticed, this
-12V and -5V disappeared under load...

LM340T12 were volcano-like hot. It's mounted back to back to the LM7905,

LM340T12 *was*

 

[[email protected]]

Let me try to get up to speed here.....

"Now, after having replaces all capacitors, voltages oscillate, without

going over default values... i.e. 5V oscillates between 1 and 4V, 12V
between 6 an 8. 3844 Vcc value is under the minimum value. "

This is not resolved correct ? In that case likely it is the Vcc or Vdd filter at the chopper IC. The big IC by the transformer. If you can't get the pinout you can identify that pin thus; the transformer has two windings on the hot side. One the the output of the chopper and drives the whole shebang. Another is a secondary which supplies the lower source to the IC. This voltage is initially supplieed by a resistor from the rectified AC, or even AC from one side of the bridge, depending, but it still has a resistor and there will also be a diode from that winding going to the same IC pin. The resistor just starts it, but does not supply enough current to run. The capmust hold the voltage until the thing gets started up and supplies itself.The reason it would rune with the secondary filter bad is that there was less inrush current drawn and thus a short on time for the output device.

That is IF I read it right. The reason the negative supplies drop like a rock is probably alot less filtering on them.

"but have only a doubt: can I substitute the LM340T12 with a

MC7812? "

On negative supplies they are 79XX series, 78XX are positive. The XX is theoutput voltage.

 

[supervinx]

Now the situation is this.
+5V and +12V stable (12V is 11.80, 5V is 5.01)
-5 and -12 were absent under load.
They are generated by two cascaded 7812 and 7905.
The output of the 7812 is fed reversed to output (becoming -12V), and becomes
the input for the 7905, generating the -5V.
If the 7812 is dead, no -12V, no -5V.
I substituted both.
But the 7812, coming from scrapped boards, shouldn't work correctly.
It receives 14V as input, but generates an instable +8V, enough to
drive the 7905. So, now, I have the -5V but not -12, yet.
The 7812 becomes hot again.
I have no spares, so I must suspend.
But I think we're near the end, since only the -12V is missing, and it's not
sensed and generates autonomously.
The input voltage, is stabilized by a capacitor, replaced. 14.5V seems a
correct value, since it should be at least 1.5V above the +12V.
Since it runs hot (and input voltage is correct) I must assume it's not
working...
value

 

[[email protected]]

Ummmm, to use a 7812 to regulate negative would be illogical.

Which pin is common ground ? On the 78XX it is in the middle. If it is off to the side that is actually a 7912.

I don't care what it says on it. I'd bet it is a 7912. There is no reason for the engineer to **** around and go through what it would take to use a 7812 there. If it is actually a negative regulator with the ground pin in the middle then it is custom, or a different series. It is actually possible.....but.

They love to use custom parts to create a captive market, but I can think of a whole lot of other places in a PS like that to accomplish that, the standup board with the black conformal coating would do it.

Look at the board, I think you are going to find the ground off to one side, which means you need a 7912 no matter what the numbers say.

Take it from me, I sometimes don't even look at part numbers anymore, I just look at where something is in the circuit and usually know what is neededthere. Sometimes I do have to look it up, but not all that often at least when it comes to something like this.

 

[supervinx]

Il Mon, 17 Jun 2013 21:16:24 -0700, jurb6006 ha scritto:

Ummmm, to use a 7812 to regulate negative would be illogical.

Which pin is common ground ? On the 78XX it is in the middle. If it is
off to the side that is actually a 7912.

I don't care what it says on it. I'd bet it is a 7912. There is no
reason for the engineer to **** around and go through what it would take
to use a 7812 there. If it is actually a negative regulator with the
ground pin in the middle then it is custom, or a different series. It is
actually possible....but.

They love to use custom parts to create a captive market, but I can
think of a whole lot of other places in a PS like that to accomplish
that, the standup board with the black conformal coating would do it.

Look at the board, I think you are going to find the ground off to one
side, which means you need a 7912 no matter what the numbers say.

Take it from me, I sometimes don't even look at part numbers anymore, I
just look at where something is in the circuit and usually know what is
needed there. Sometimes I do have to look it up, but not all that often
at least when it comes to something like this.

No, it's a LM340T12 (7812). Datasheet says it's a positive regulator.
The scheme used is this.
+14.5 V between In and Gnd of 7812. Out of the 7812 is connected to
output GND, Gnd of the 7812 to the -12V line. And the -12V line is done...
Gnd of the 7812 is connected to Gnd of the 7905, Out of the 7812 to In
of the 7905.
So the regulated 12V of the 7812 feeds the 7905, Out of the 7905 is connected
to the -5V line.
That's a bit odd...

I tested the spare MC7812C off circuit, maybe 14.5V were not enough.
I used 18.5V as input voltage to the MC7812C, resulting in 8.xx volts
as output. So the spare MC7812C is dead. It became so hot that I saw a bit
of smoke.
I need to buy another 7812, my spares are ended...

 

[supervinx]

Found another 7812. Tested standalone, mounted, added a heat sink,
tested, all voltages OK.
Last test will be a long, controlled, poweron.