Trying to repair an ATX power supply

[dev_null_h]

Backstory: On Monday morning my desktop cut off. I tried to power it back on but no signs of life. Was lucky to discover it was just the power supply that went without taking out any other components.  I did order a replacement power supply. It arrived this morning and got it installed.

I'd figure I'd try to take a crack at repairing the old one.  Only the standby 5V rail is functioning. It briefly appears to power up with the fan starting to spin but then goes into shutdown.  Upon cracking the old power supply open, I checked for any signs of any components releasing the magic smoke (or any leaking/bulging caps), and saw nothing.  I inspected the solder joints but didn't find any that were cracked.  I noticed what appeared to be a thermal fuse stuck onto a heat sink. I de-soldered a lead and checked it with my multimeter. I got a reading in the kiloohm range, so apparently it was a thermister. Darn, was hoping that would be it.  I did check to see if any of the rails were shorted. The lowest reading I got was around 20 ohms.

Are there any other easy thinks to check, or am I just going down the rabbit hole to hell at this point?

[wn1fju]

You are partially down the rabbit hole already!

Could be a variety of things, such as a bad cap on the power rail of the IC controller chip or a bad voltage/current sense component. 

Take a look at    https://www.repairfaq.org/sam/smpsfaq.htm   for many other ideas.

And be careful when debugging with the supply powered on.  Don't go hooking your scope ground clip on the primary side, for instance.  If you don't know why, then you shouldn't be attempting to fix a switching power supply.

[dev_null_h]

Thanks for the response, I'll check out the link.  Don't have to worry about me misplacing the scope ground clip, as I don't have one (maybe some day.)

[Whales]

Random thought: Hitting over-current protection when you turn it on might suggest shorted mosfets, shorted rectifiers or shorted transformer.  You can try probing these out with a multimeter (inc on diode mode) whilst the unit is off.  The larger transformer and the parts on the heatsinks around it the targets here, the smaller transformer on the board will be for your (working) 5VSB.

I completely misread your last comment.  "You probes are missing their ground leads?"   Personally I rock an old 2-channel HP scope from the 90's and a CRO with super-nasty cheap probes off eBay   If you're after a scope then I hope you strike luck.

[dev_null_h]

I got around into checking things out again today.  I checked the large TO-220 diodes (had to de solder them to get a good reading) and they all checked fine. I did find a little bit of corrosion underneath some of the goo(that has since gone brittle) used to hold things down that also has turned slightly conductive, so I removed some of the components (the clinched leads were a bit of a pain to deal with) in that particular area, removed what I could and re-mounted those components. I did find a TO-92ish component that was mounted in the opposite orientation of what the silkscreen indicated (glad I took photos beforehand.) I suppose it was a component change.  I did get what appears to be a dead short across two terminals on the primary side of the transformer but that was measured in-circuit. Still need to check the mosfets on the primary side (likely need to de-solder those to take a reading.) 

Although the board is single-sided, the large tinned traces and a lack of thermal reliefs make soldering/de-soldering a little challenging with using just a pencil iron (and a cheap Chinese special at that.) Probably should have gotten out the FX-888D.  Might be a good idea to pickup something like a Weller soldering gun for dealing with the particularly high thermal-mass joints. It did at least gave me some practice with de-soldering by hand, particularly on how to handle clinched leads (I found using the manual solder sucker followed by wick worked well.)

After doing what I did, still same symptoms.

[Whales]

You might be able to check the mosfets without desoldering.  If they're TO-220 then their pinout is almost guaranteed to be GDS.  The gate-to-other-pin resistances should be above ~1kohm or so (ideally megaohms, but other in-circuit elements will bring this down).  If you read a very low resistance (less than a few hundred ohms or so) then the mosfet is probably blown.  EDIT: Oh and the DS resistance should probably be high, but you have to desolder to properly check that when near things like transformers.

Cheap irons: I've been using a cheap ($40AUD?) 900M-tip style unit for a decade now.  I've had to replace the iron and lots of tips, but it has done me well.  The key seems to be using a really big tip with lots of thermal mass.  My latest tip has somehow lasted forever (a year?), I think it was a more expensive one I bought on a whim to see if it was worth it.

> I did find a TO-92ish component that was mounted in the opposite orientation of what the silkscreen indicated (glad I took photos beforehand.) I suppose it was a component change.

Yep.  TO-92 pinouts for things like BJTs are either CBE or EBC, often changing between makes of even the same part number!  Evil.  Before I bought a cheap transistor tester (LCR-T4) I was often making mistakes.  Fun fact: BJTs do actually work in reverse, just not very well.

[dev_null_h]

Thanks for the response, and thanks to your help the problem may have been found. I'm getting around 10.6 ohms across gate and source.  The part number of the mosfet is mdp18n50.  I'm going to have to search my stash to see if I have anything suitable.

EDIT: Got nothing suitable in the stash.

[james_s]

Yeah the gate should be effectively an open circuit to both the source and drain with a mosfet. ATX power supplies are some of the more challenging devices to repair. They tend to be highly cost-engineered, as you noted by the single sided PCB and they are usually very dense and any sort of schematic or other documentation is rare. I've made a token attempt at fixing a few of them but I don't think I ever actually succeeded. I'm sure I could have if I'd been sufficiently motivated but for how cheap they are it just wasn't worth investing a lot of time. The old AT and PC/XT power supplies on the other hand, I've fixed a number of those. They're usually larger and a lot less dense being typically much lower power.

[amyk]

What's the make and model? At least show us what the board looks like, front and back.

If the gate oxide blew out on a MOSFET, the driver circuitry may also have been damaged, and if that includes the controller IC, then it's likely beyond repair.

ATX PSUs, especially the cheaper ones, are pretty simple. Many schematics of them can be found here: http://danyk.cz/s_atx_en.html

[Vincenzo]

You are partially down the rabbit hole already!

Could be a variety of things, such as a bad cap on the power rail of the IC controller chip or a bad voltage/current sense component. 

Take a look at    https://www.repairfaq.org/sam/smpsfaq.htm   for many other ideas.

And be careful when debugging with the supply powered on.  Don't go hooking your scope ground clip on the primary side, for instance.  If you don't know why, then you shouldn't be attempting to fix a switching power supply.

If the original poster studies the material in this link (I recommend printing it in a booklet format and think of it as a final test) and succeed in repairing a few of those supplies, he may gain an experience that is close to a whole college degree worth of lab work.

Remember, two necessary tools, a series (incandescent) bulb, and a variac, and an optional tool that will make your life a whole lot easier is a hakko desoldering vacuum iron (don't fall for cheap chinese fakes, it's worth the price).

[dev_null_h]

Someone asked earlier the model of the power supply. It's a Thermaltake TR2-600W.  I'm looking for a replacement mosfet that I can get in a reasonable amount of time. The original part seems to be of limited availability and a long wait to get a hold of. I did some quick searching and found a possible replacement, FDP18N50. A quick glance at both datasheets suggests that it might be a drop in replacement. EDIT: That part doesn't look too terribly available either on Fleabay (there was 1 US seller, but I'd rather not spend close to $10 USD), will keep searching.

http://semiown.com/pdf/MDP18N50.pdf
https://www.onsemi.com/pdf/datasheet/fdpf18n50t-d.pdf

[dev_null_h]

Cheap irons: I've been using a cheap ($40AUD?) 900M-tip style unit for a decade now.  I've had to replace the iron and lots of tips, but it has done me well.  The key seems to be using a really big tip with lots of thermal mass.  My latest tip has somehow lasted forever (a year?), I think it was a more expensive one I bought on a whim to see if it was worth it.

I've found that even the cheap Chinese tips last better than the tips for the Weller red series irons. Fun story: Years back I complained to someone at Weller about the tip plating failing after a few uses and the copper core starting to be dissolved.  He said he was going to check the failed tip at the lab to find out why it failed.  He sent replacement tips, but he never told me why the tip failed even when I persisted. He likely knew they were crap, but didn't want to get into any hot water for admitting it.

[dev_null_h]

Got trolled. I de-soldered the mosfet, and checked it. It tested fine.  That 10.6 ohm or so reading was from something else on the board. EDIT: Looks like it could be the mosfet on the other side of the heatsink, will see if I get trolled again. EDIT: Got trolled again, not it either.

[amyk]

Don't forget that the transformer being in circuit may look like something has shorted. If you test at a higher frequency, its impedance will rise but that of other shorted components won't. Here's a video showing that trick:

[RoGeorge]

Have you attached any load when testing the power supply?

Last week I almost start desoldering 800W power supply parts for testing them, but decided before doing that I should test with a proper load first.  TLDR, with a 12V auto lightbulb (and the green wire tied to ground) all the voltages were good.

After replacing the CR2032 battery and resetting the CMOS, it all started to run again.   

[dev_null_h]

I tried that (used a power resistor vs a 12v auto bulb), no effect.

I did notice that there is a high pitch sound that appears be be coming from the main transformer (might just be the transformer for the 5V standby power though.)  When I try to start it up, I hear a tick sound. I tried it in a dark room to see if there were any sparks, but nothing.

[RoGeorge]

To distinguish which transformer is whining, use "bone conduction" hearing:  take a solid wood or plastic stick, and touch one end to the ear's external cartilage, and with the other end of the stick touch each suspected coil's body.  Direct sound conduction is incredibly effective.

Another thing about transformers, when they have internal short circuits they usually have a very specific and strong burnt smell, easy to detect if sniffing the coil (unpowered, of course), not kidding.

Another generic thing to do is to check the electrolytic capacitors.  I have had a Thermaltake (model W0008R Purepower 420W, dual fan, active PFC, some many moons ago), and after only 2-3 years I had to replace all the electrolytic capacitors, they dried out and developed high ESR (the low voltage ones only, the high voltage ones were OK).  It's been probably 10 years since then and still working, but since it was the only branded power supply I have ever had to re-cap it, I never used Thermaltake again.  Expensive and very hyped brand considering its quality, IMO.

[dev_null_h]

Thanks for that, the high pitch noise was just from the standby power transformer.

[huehu3gg]

Got trolled. I de-soldered the mosfet, and checked it. It tested fine.  That 10.6 ohm or so reading was from something else on the board. EDIT: Looks like it could be the mosfet on the other side of the heatsink, will see if I get trolled again. EDIT: Got trolled again, not it either.

hi! im getting troled too by this Mosfet, outside the measuring its fine, but in board G S give 10 ohms ih the Q3 and Q2 . found anything for this? my psu start but shutdown after few seconds. its thermaltake smart 750w modular

[CaptDon]

Power mosfets require a low impedance drive circuit to get sharp on/off transitions and stay out of the linear range. When you said you measured 10 ohms G to S I wasn't surprised. I have seen some of those supplies drive the gate directly with a transformer and then the gate looks like it is dead shorted.

[Labrat101]

The other thing to check there are normally 2 very small caps very easy to over look . they are  50v @ 2.2 uf   &  50v @ 10uf .
( may very slightly depending on design )
 these go bad . If these fail will also cause start up problem as they are part of the power good circuit .  I have found these 2 caps to be about 50%
of problems on startup or pulsing .

[MathWizard]

OP what brand/model PSU is it? And there should be 2-3 big IC's that are the controller IC's, 1 of the most common is the CM6800, which runs the PFC and foward converter on the primary side (small/old/cheap PSU's might not have PFC)

The general layout is there's a mini-SMPS that runs off the rectified mains AC, it creates the Aux voltage on the primary side, around 12V. And on the secondary side, that also creates the 5V stby voltage. When the PSU is plugged in and on, both the Aux and Stby V should always be there.

Then if it gets the turn-on from the computer, the PFC would start up, and pump the main capactior's voltage up to about 2x rectified line voltage ~340V (in 120V countries anyways). Then the forward converter makes the secondary voltages.


Then a big chip on the secondary side looks at secondary voltages and decides if they are too high/low, and if any other warnings like from a temp sensor circuit, are ok, and it tells the main chip on the primary side what to do, through opto-couplers.

Ok it's a Thermaltake TR2-600W, and u found a bad mosfet, was it the PFC/boost converter or the forward converter ?

I'm working on a TR-750W, it's synchronous rectifier, was on an undersized heatsink, and it burned out big time. I got replacements mosfets, but I need to $$$ up and get the proper driver transistors, they worked but I don't trust them after that or me cooking them with the soldering iron