SMPS repair

[Rachie5272]

I'm trying to repair an LCD monitor with a bad power supply.  It's a fairly simple SMPS, with 5 V and 12 V outputs.  It uses a UC3842 IC, and a single TO-220 FET.

I get nothing at all on the secondary side.  The primary has a steady voltage across the main capacitor, but the low voltage supply for the IC seems to pulse between 0 V and 16 V.  I don't have an isolation transformer, so I can't put my scope on it.  Clearly this is the problem, but I don't know how to fix it.  I've read some guides which say to test the zener diodes, but how can I test them when I don't know what value they should be?

The power resistors all seem fine, as do the ancillary diodes.  Everything looks fine, too.

It looks like someone had a go at fixing it previously.  There are some scorch marks on the monitor chassis near the secondary, and the filter capacitors have been replaced, so I'm guessing the old ones blew up from the plague.

Any suggestions?

[duskglow]

It's a single sided board, right? Since it looks like repair's been attempted before, the caps might not be the only things that failed.  I'd also check for burnt or corroded tracks - when caps die, sometimes they leak electrolyte all over the place.

[retiredcaps]

but the low voltage supply for the IC seems to pulse between 0 V and 16 V.

Over at badcaps.net, we usually suggest the following may be culprits.

1) bad startup cap.  Many people neglect replacing this cap because it is not "bulged".  It is typically at 47uF 50V cap or similar.  Simple and cheap to replace.

2) bad SMPS IC.  Sometimes VCC is shorted to GND.

3) something on the output side is shorted like diode.  To test a diode, simply put your multimeter on diode test function and put the probes on either side for one reading and then swap.  You should get 2 readings.  One will be 0L and the other should be between 0.3 and 0.6V DC.  In my limited experience, diodes usually fail shorted.  If this is the case, your multimeter should show 0.0 in both directions.

[Rachie5272]

The power supply must be mocking me, because now I get nothing at all on the IC power rail.  I'm reading about 17 ohms between it and the negative rail.  I'll check the stater cap now.  [Edit: I replaced the starter cap, to no avail.]

The board is single sided, with no signs of corrosion.  There are no scorch marks on the PCB itself, so apparently the caps only blew at the top.

[Edit again: I lifted the negative rail on the chip, and this solved the 17 ohm short problem.  Looks like the chip is bad, but I doubt I have a spare.]

[retiredcaps]

Sometimes the UC3842 is found in ATX power supplies.

[Rachie5272]

I checked several old PSUs with no luck.  I ordered a sample from TI.

[kripton2035]

before changing the uc3842, change the small 10uF or so capacitor that is between the pins 7 and 5 of the chip.
I never changed that ic, but these caps are always the faulty thing.

[retiredcaps]

before changing the uc3842, change the small 10uF or so capacitor that is between the pins 7 and 5 of the chip.

That would be the startup cap I that referenced in my earlier post.

[dr_p]

3) something on the output side is shorted like diode.  To test a diode, simply put your multimeter on diode test function and put the probes on either side for one reading and then swap.  You should get 2 readings.  One will be 0L and the other should be between 0.3 and 0.6V DC.  In my limited experience, diodes usually fail shorted.  If this is the case, your multimeter should show 0.0 in both directions.

just my 2c: what I do for a basic switcher for speeding things up is testing diodes in whatever direction. Anything except 0.000 is ok (since the measurement is in circuit it might be 0.100 for a good silicon diode). IMHO a controller failure-to-start will not be caused by a diode failing open circuit anyway, so no reason to look for that. Plus, the only diodes I saw failing open circuit were physically broken in half or had pins exploded off.

[Rachie5272]

The diodes on the secondary side appeared to be good when tested in-circuit.  They're not shorts at least.

Curiously, one little diode on the primary side appeared to be shorted in-circuit, but was fine when I desoldered it.  It had some low value resistors in parallel.

[dr_p]

Curiously, one little diode on the primary side appeared to be shorted in-circuit, but was fine when I desoldered it.  It had some low value resistors in parallel.

Test it in circuit on diode mode. If it's not 0.000 or 0.00x, then it's fine.

[amyk]

If you've had enough of trying to repair it, you can replace the PSU with a length of cable with a Molex socket on the end and just power it off the computer its connected to. 5V and 12V are available inside any PC. (Might even be more efficient...)

[Rachie5272]

I got the chip, and replaced it.  No more internal short, but it's back to its old routine of pulsing on and off.

Measuring the output of the secondary, the 12 V rail jumps up to about 1.4 V, drifts back down over the course of two seconds, then jumps back up again.  A similar event happens with the driver chip's power rail as well.  Evidently it's shutting itself down in some sort of protection mode.

Any more suggestions?  I currently have a one ohm resistor in series with the mains to prevent any more damage.

[dfmischler]

I think a protection circuit should shut it down a lot faster than that.  My guess would be the feedback circuit.  Usually this involves some kind of voltage reference (zener or something) on the secondary side and an opto-coupler to provide feedback to the primary.  I have had a bad opto cause the kind of slow shutdown-restart cycle you're seeing.  Freeze spray or jumping in with a scope might help you catch the culprit.

[Rachie5272]

I'm getting closer.  Freezing the optocoupler did nothing, but freezing stuff on the primary certainly did.  When I chilled a little SMD rectifier and transistor, the 12 V rail started pulsing all the way up to 12 V.  Sometimes it would even stay there for a few seconds before dropping again.

I'm not quite sure which component is bad.  They're too close together to freeze separately.  I'm guessing the rectifier is jellybean, and can be replaced with pretty much anything, but what should I do about the transistor?  It's not as if anyone labels them, and trying to find a schematic would be hopeless.

Or maybe I should use a mini cryocooler to keep them cold.  That would be kind of fun...

[dfmischler]

I've used pieces of paper or a vinyl tube to make sure I know which component likes to be cooled.

[dfmischler]

... trying to find a schematic would be hopeless.

You might be right, but don't be too sure.  Maybe one of us has been there already.  What monitor is it?  What part number is on the power supply board?

[Rachie5272]

I got it fixed, somehow.  I kept freezing different components, but couldn't isolate the problem.  It only worked when I froze a whole area, but it would also stay on after that.

Out of desperation I decided to reflow a bunch of joints, and it seemed to work on powerup--sometimes.  Then I replaced a couple more little electrolytics on the primary side, and it seems to power up just fine repeatably now.  Or at least it better, because I'm closing it back up.

The monitor is rather old and analogue only, but I felt like getting in the practice.  I might have a whole bunch more monitors coming in soon, too.  I suspect they'll mostly be either bad caps or a cracked LCD.  Either way, fun to play with.

Thanks everyone.

[hpapagaj]

I am trying to diagnose a small 12V power supply. I have an X2 cap rated ".1uFK"-250V and another ".1K" (near to the input 230V). Is that 0.1uFK/.1K = 100nF? When I  test the cap with my meter (MG328 tester), it reads around "60/70nF" (desoldered).

Is the cap bad? Am I interpreting the rating on the cap case right?

[Armadillo]

I am trying to diagnose a small 12V power supply. I have an X2 cap rated ".1uFK"-250V and another ".1K" (near to the input 230V). Is that 0.1uFK/.1K = 100nF? When I  test the cap with my meter (MG328 tester), it reads around "60/70nF" (desoldered).

Is the cap bad? Am I interpreting the rating on the cap case right?

MG328 tester is an ESR meter, not a capacitance meter, though it also show the capacitance. 100nf = 0.1uf yes. I don't think the capacitors are faulty. Such low capacitance cannot be accurately measured by this meter even for mid range meter.

[kripton2035]

X2 caps are rarely faulty.
what is the topology of your 12v power supply ? SMPS or linear ?
what is the failure ?
post close-up pictures of the board if possible

[hpapagaj]

It's a simple PSU from ebay, lasted few months and I am trying to repair it. Output voltage is 0V.

I checked:

-capacitors: OK
-diodes: OK (it's showing value 0.5-0.6V on my multimeter)
-fuse in the black heatsink: OK

[SeanB]

Flip board over and you will see some SMD resistors that are a start up supply for the controller. Probably one or more of them are open circuit, a common failure. typical values are 33k to 100k for the resistors. Change the small electrolytic capacitor near the controller as well, that also is a common failure if the chip itself and the mosfet are not faulty.

[kripton2035]

how did you check the capacitors ? with your multimeter ? it's not enought. you must check the esr also.
try to replace the small green capacitor in the middle of the board. it's often that one that is dry.
notice that the capacitors hate too much heat and that this one is just beside ... a heatsink.
and is critical for the smps controller to start oscillating.