Is it worth trying to repair a 800W PC power supply?

[RoGeorge]

Main desktop stopped working, and so far the power supply seems to be the cause - not starting with green wire to GND and only a 12V car light bulb as load.

Power supply model is SF-800P14HE
https://www.clearesult.com/80plus/sites/80plus/files/manufacturer-certificate/super-flower-sf800p14he-9013.pdf

It was about $200 when new, capacitors still look good, fuse is good.  Most probably some semiconductor died.  A similar specs new supply seems to cost slightly more now, about $300.

Is it worth trying to repair?  Asking because in my rare attempts to repair switching power supplies (of any kind) they are usually hard to debug, and hard to desolder/remove the power components.  With the component shortage might not even find the right power parts.  I'm not into service, so not sure how big is the success rate when it comes to PC power supplies.

- 1.  Is it worth to repair a PC power supply, when the defect is something else than dried capacitors?
- 2.  Anybody knows a schematic for this 800W Super Flower SF-800P14HE, please?

[Haenk]

 

No point in repairing such an old PSU, which only has a Bronze rating (which means this is more or less a room heater) and uncertain parts-required-to-repair status.

You could get a "Gold" one for less than 100,- EUR; cheap, unefficient ones even for less than 50,- EUR.
For about 200,- EUR, you can get a really nice "Platinum" rated top-level branded one.

[nightfire]

Super Flower is no brand I have heard of in Europe with a good reputation and/or recognition as a manufacturer.
If money is tight, AND knowledge and tools are available, some fault analysis might be ok, and be it only for learning purpose.

Otherwise good quality power supplies for PCs are better value in the long run.

Question here would be, what load does need to have a 800W Power Supply?

My rig also used for some gaming is happy with a 400 something Watts Power Supply from Seasonic, where I have some trust that that PSU is capable of regulating even demanding load spikes from the GPU.
(Got that used for 25 € from ebay)

[shapirus]

Does it not start at all, or starts sometimes? For example, after staying disconnected from mains for, say, a day or more, or after the input capacitors are discharged?

[RoGeorge]

Maybe I should start looking for a new one next week.

Meanwhile I've added one more lightbulb as load.  With the motherboard disconnected, one 45W/12V incandescent lightbulb on the +12V and another one the same type on the +5V it starts for about 2-3 seconds, even the Power Good wire goes up, though the 3.3V only goes to 2.8V or so, then it all shut off itself after 2-3 seconds.

With a 3rd lightbulb as load on the 3.3V, the PowerGood drops almost immediately and all the other voltages shut down much faster with an additional lightbulb on the 3.3V

+5Vstb, +12V and +5V seem to work OK, but -12V is missing entirely, and 3.3V never reaches 3.3V without any load on it.  When loaded, 3.3V stays almost 0.

Good news is that it's not completely dead, has something wrong with the -12V and +3.3V power bars only.  I think I'll open it to look closer what happened, and how come that 3.3V is affected while the +12V and +5V (which are also high current) seem OK.

[wraper]

Super Flower is no brand I have heard of in Europe with a good reputation and/or recognition as a manufacturer.
If money is tight, AND knowledge and tools are available, some fault analysis might be ok, and be it only for learning purpose.

It's one of the best OEM for many "reputable" brands which are not actual manufacturers.

It was about $200 when new, capacitors still look good, fuse is good.  Most probably some semiconductor died.  A similar specs new supply seems to cost slightly more now, about $300.

Not true. You can easily buy 800W 80+ Platinum rated PSU for $200. While this only qualifies as Bronze. While it's a decently built PSU, it's not efficient by todays standards.

[nightfire]

Super Flower is no brand I have heard of in Europe with a good reputation and/or recognition as a manufacturer.
If money is tight, AND knowledge and tools are available, some fault analysis might be ok, and be it only for learning purpose.

It's one of the best OEM for many "reputable" brands which are not actual manufacturers.

Ah ok, good to know- last time I had a look at some tables for which brand and OEM are connected, is some time ago, and since then I decided that Seasonic would be my "go-to" brand for most builds I did/do.

Regarding the defective PSU: When something runs for a few seconds and then stops, thermal issues could be some accompanying symptom. Like some defective semiconductor quickly overheating, so some hunt for them could yield some results.
Otherwise some lengthy de-soldering session would be needed to get them out of the circuit to measure them.

[Messtechniker]

For many of my projects I recycle quite a few items
from dead ATX power supplies. Also saves me having
to do dreaded metal work a saves money on those
nowadays very expensive metal housings.
Here an example of my high voltage amplifier.
Does +33 dBu into 1 kOhm

[exe]

A friend of mine told me repairing computer power supplies is kinda dangerous (probably due to high internal voltage and large electrolytics). Please be careful.

[shapirus]

A friend of mine told me repairing computer power supplies is kinda dangerous (probably due to high internal voltage and large electrolytics). Please be careful.

This is a good advice. While the regular high-voltage precautions are pretty much common knowledge, there are less obvious things such as the input capacitors which can hold high voltage charge and be dangerous.

[T3sl4co1l]

If you have the means to troubleshoot, and can spare a handful of hours, sure.

Maybe that's already not worth it, in which case definitely buy.

"Means" includes: a general understanding of typical circuits (flyback for the aux; half-bridge, two-switch forward or resonant for main); a scope to probe signals; an isolation transformer to probe safely; load resistors to test the outputs; and whether you can find datasheets for the controllers (often they're no-name Chinese types) or can identify them well enough to infer pinout and function.

Let alone if you have replacement parts; if it's a dead chip, you'll have to order a replacement anyway, and maybe that doesn't do it, and you're out a cycle until you can try something else, etc.  That'll easily rack up enough labor or down time to cost you.

I chose to repair my last PSU failure, which turned out to be one of the more esoteric possibilities: symptom was, it would click on and immediately turn off.  Interpretation: primary-side power stages are all working, at least somewhat.  Probed around some things, found 3.3V output was overshooting, hmm.  With -- I forget if I checked it under load, or that was sufficient evidence?, I identified that was the problem; the power monitor chip (which fortunately has a datasheet) was faulting out.  Okay, trace 3.3V power path.  The magamp regulator was saturated full on.  Well that's no good.  The BJT driving it had failed somehow.  No visible damage.  Replaced with an in-stock part, good as new.

Whereas if it's just stupid leaky caps everywhere -- you'll probably need a range of new parts (or only have random salvage to replace with -- not much better than what's already there), and it's a big pain swapping out everything.  And maybe that doesn't do it anyway.  Probing a few power rails might be prudent to avoid wasted effort like that -- measure ripple esp. switching ripple, also check with hot air / freeze spray to identify sensitive caps.

Or if it's failed semiconductors, the cascade of damage might not be worth checking and replacing.  Replacing the most visibly failed transistor(s) only exposes other failures; most often the controller is failed as well, and maybe the nearby resistors.  Maybe there are other devices in parallel with the local supply that failed due to overvoltage (current flow from MOSFET drain to gate to controller to VAUX -- remember these are fault conditions, multiple chips may arc over), and of course resistors may be out of tolerance or blown (I've even seen some fail downward).

On that note, I once found a dead PSU that its only failure was the UC3842 controller.  No clue how that happened!

Tim

[jonpaul]

boxes of these in computer, electronic recyclers.

value about 10$

lots of experience, test and safety equipment to troubleshoot dead,SMPS

Jon

[RoGeorge]

Opened it, and seems difficult enough to even measure, let alone replacing parts. 

Has plenty of modules soldered perpendicular to the PCB, with thermal glue all over between the boards and parts, for rigidness and anti vibration.  All the backside is full of SMDs, so hot air to remove the soldered perpendicularly modules would be tricky.  Then, it has something that looks like a lacquer layer, and some IC on the back side look unmarked.

Would clean the dust and look closer to deduce the topology.  I've spotted 2 optocouplers, those are easy to reach and prone to fail.

No visible defects at a first look, no burnt smell, so I guess it's a low power component that failed.  I don't know what are the whitish drops that looks like spill.  Will look under microscope after cleaning.  Certainly never were any liquids inside this PC. 

My best hope is that the missing -12V was also powering something in the +3.3V feedback loop (seems to be having a compensation grey wire coming back from one of the 3.3V pins at the MB connector - there are a grey and an orange wire in the same pin).  So if I fix the -12V, I hope the +3.3V will simply work, or else it will take too much work to debug it further.

[fzabkar]

Measure the resistance between each rail and ground.

[nightfire]

Is it just me or the display of my notebook I am currently using, or are there several areas of the PCB that look like they have gotten too hot and/or dirty?

[abdulbadii]

Imho, depends of course on your tool, equipments and kits resources condition
if those're not available as suggested, simply not worth it, otherwise it is, as it seems working (starting correctly) so seems just few components fault

[amyk]

Clean up the dust, and then resolder all the connections.

That has worked for many electronics in the past where thermal cycling caused a joint to crack just enough not to be entirely visible, but enough to cause a failure.

First thing to check is whether there is 5Vsb and if it's the correct voltage. I'm not sure about that model in particular but in the past a lot of models had a capacitor in the 5Vsb circuit which would go open or high ESR over time.

Bad caps aren't always visibly bad.

[max.wwwang]

Worth or not, that's really too personal a question. 

One woman's trash, another woman's gold!

[DavidKo]

I will start with cleaning the dust. If it is conductive, it can create issues. Than check the capacitors visually. If you see that they are NOK, check their availability - in my case one of the capacitors was not available (thin and tall). I have bought new PSU and than I have tried to repair the old one just for fun. Did not finish that due to the capacitor availability (I will not buy parts which costs more than new PSU).

To discharge the capacitors I use insulated 50 Ohm power resistor.

[RoGeorge]

Thank you all for help and suggestions.  Before anything else I'll have to clean the desk, and cobble an isolation transformer from 2 back to back normal transformers if I want to use the oscilloscope.  Cleaning that workbench might take days!

Meanwhile I've searched more for a schematic.  Could only find other ATX schematics from older models and max 200-300W only.  The surprise was that they all have a post regulation mag-amp stage for the 3.3V (3.3V is one of the defective rails).  I've read that more recent ATX models might use transistors instead of a mag-amp.

On the outside sticker it says max 24A on 3.3V, max 24A on 5V, max 66A on +12V, max 05A for -12V and max 3A for 5Vstb.  Not sure yet what kind of 3.3V stabilization uses mine, but I can not identify which one would be the magamp coil, maybe mine is using transistors instead of mag-amp.

The idea with the magnetic amplifier is that a saturated core shows a much lower permeability, thus a much lower inductance.  For high frequency an AC this will look like a closed switch.  When the core is not saturated, there is a big permeability, so big L, and a big L will stop AC current.  We can saturate the core using a small DC current (passed through many turns) in order to stop tens of amps through another coil with less turns.  Thus the magnetic core can be used as a switch, as an amplifier, even as a memory.

Found a seminar from former Unitrode (now TI) that explains the mag-amp very nicely in the first 2 pages:  Magnetic Amplifier Control for Simple, Low Cost, Secondary Regulation by Bob Mammao
https://www.ti.com/lit/ml/slup129/slup129.pdf

While this idea of taking advantage of the magnetic hysteresis looks very interesting, the history of the magamp is even spicier, going back to WWII, V2 rockets and the operation Paperclip:  The Vacuum Tube’s Forgotten Rival
https://www.nb9m.com/index.cfm?key=view_news&TransKey=D5623FA0-C14B-4F99-AD74-0E38C60CEDE2&Small=1 (the page is missformated in my browser, the rows show truncated, I had to CTRL+A to select all then copy paste into a text editor to read it in full)

The magamp idea seems very clever, and I didn't know about it before, so now I'm tempted more to experiment with magamps than to try fixing the power supply. 


So far only -12V and 3.3V are not working.  All other voltages are OK even when a 45W/12V car light bulb is attached on each of the +12 and +5 rails.  It start for 2 seconds then shuts off.  To answer a few direct questions:

Question here would be, what load does need to have a 800W Power Supply?

I guess half of it would be enough.  The desktop is an i7-4790K with 32GB RAM, a single nVidia GTX760 and 1-3 HDDs.  No overclock, no gaming.  The 800W choice was to make it run colder, thus less noise and less dried capacitors over years.  Can also power the PC without a hitch over mains power surges of about 1-2 seconds.

Does it not start at all, or starts sometimes? For example, after staying disconnected from mains for, say, a day or more, or after the input capacitors are discharged?

I don't suspect the capacitors.  It died at a wake from standby, the speakers made a pop lawder than usual.  Now it starts for about 1-2 seconds, power good comes up for a fraction of a second, than it realize it's missing the +3.3V and the -12V, and it shuts itself off.

symptom was, it would click on and immediately turn off.  Interpretation: primary-side power stages are all working, at least somewhat.  Probed around some things, found 3.3V output was overshooting, hmm.  With -- I forget if I checked it under load, or that was sufficient evidence?, I identified that was the problem; the power monitor chip (which fortunately has a datasheet) was faulting out.  Okay, trace 3.3V power path.  The magamp regulator was saturated full on.  Well that's no good.  The BJT driving it had failed somehow.  No visible damage.  Replaced with an in-stock part, good as new.

Very similar symptoms, thought mine might not have a magamp, but some MOSFETs instead, not sure yet.  It's very hard to even see if on the radiators are MOSFETs or double diodes.  They are covered and only 2mm or so space above the markings.

Measure the resistance between each rail and ground.

It's increasing from 0 to about 3k in 10-30 seconds, except the 3.3V which is increasing much slower - after about 2 minutes was only half of those 3k and raising.  Looks like charging capacitors up to the max voltage supplied by the ohmmeter.

What am I suppose to look for while measuring with an ohmmeter?

Is it just me or the display of my notebook I am currently using, or are there several areas of the PCB that look like they have gotten too hot and/or dirty?

The pics were with a flash will try better ones in daylight, but it was indeed something splashed there, I suspect those are some flux residues from the heavy soldering.  It even has a metal plate solder on top of the PCB traces.

[djsb]

Get yourself a nce Corsair or Seasonic supply. My Corsair HX650W has outlived my old PC from 2007 and is now being used in my new build. It replaced an original HX620W replaced by Corsair free of charge due to a switch problem early on. You could repurpose your power supply as a bench supply once you've fixed it, but be careful.

[wraper]

Get yourself a nce Corsair or Seasonic supply. My Corsair HX650W has outlived my old PC from 2007 and is now being used in my new build. It replaced an original HX620W replaced by Corsair free of charge due to a switch problem early on. You could repurpose your power supply as a bench supply once you've fixed it, but be careful.

Suggestion with no good basis. Corsair sold PSUs from different OEMs and for different price. Some were/are good and some were/are garbage. Seasonic is actual higher end manufacturer, so more stable quality should be expected.

[shapirus]

Seasonic is actual higher end manufacturer, so more stable quality should be expected.

I had to repair my Seasonic X-660 (which wasn't very old by that time, about 7-8 years) once when it began to fail to turn on what looked like randomly. Fortunately it turned out to be a known issue which could be solved by replacing some of resistors, apparently the original ones degraded over time, because they were 0805 used in a high voltage circuit. As far as I remember, the article about this repair said that they were used in the input voltage detection circuit which disables turning it on. I can't find that article now.

It's been working fine since that repair, which happened somewhere around 2019 I guess.

[RoGeorge]

For many of my projects I recycle quite a few items
from dead ATX power supplies. Also saves me having
to do dreaded metal work a saves money on those
nowadays very expensive metal housings.
Here an example of my high voltage amplifier.
Does +33 dBu into 1 kOhm (Attachment Link)

(Attachment Link)

He, he, don't encourage an EE hoarder with such habits please. 

My childhood was under the communist regime of Ceausescu, which was happy childhood from a family standpoint, but an atrocious scarce life from an EE/western components supply standpoint.  Even the most common parts were close to unobtainium back then, under the Iron Curtain in the Eastern Europe of Romania.  Not any longer, but any childhood trauma leaves long lasting damage into adulthood.  As a results, nowadays I'm tempted to throw away NOTHING, and even more, sometimes I collect EE waste laying on the sidewalk.  I'm a halfway professional hoarder, but in EE scrap only. 

Have used many times before parts from power supply, just like you did (i.e. my battery charger with diodes recovered from defective power supply, or my mains dimmer and my DMX dimer, too, wich are reusing the enclosure of a power supply each, etc).

Looking through my power supplies scrap box (because I was wanting to experiment with a mag-amp coil), I've found a working 430W supply brand 'PowerLine', of max 430W.    It has most of the connectors cut out, but the 24pins ATX connector for the MB is still OK.  Easy-peasy to solder a few more MOLEX on it, from other defective supplies, and power the desktop again.

This 32 bits original WinXP I'm typing now is driving me crazy! It's not the eevblog com forum's fault, it's the laptop or the WinXP that buggers me, it's the very smal 1024x768 LCD screen and the very demanding web browsers these days.  EEVblog is still great, but other places (like for example PC parts shops to search for a new power supply in my case) are so bloated that they are unusable from a 32 bits/1.7GHz/2GB RAM laptop. 


Apart from that, meanwhile I've de-dusted the defective 800W supply and took a new photo session, this time in full daylight.  It really looks like something was spilled on it, maybe at the factory or during transportaintion.  Since it was sitted in my PC, would have been impossible for any liquid to go into the power supply. 

Will post more pics, but pics editing is more difficult than I thought on a 32bits laptop with a tiny LCD, a full and slow HDD, and terrible color rendering. 

[Messtechniker]

slow hdd? Speed things up with an SSD
and there are IDE to SATA adapters. Will give your slow
puter a lot more umpf. And there are free cloning progs
so that you can avoid a new installation of whatever
operating system.

Addendum
Oops, just seen that this will be for a laptop having not enough
space for an IDE to Sata adaptor and a 2.5-in. Sata SSD.

You will need instead:
1 NGFF B/M-key SSD to 2.5 inch IDE 44Pin Hard Disk Case Enclosure
1 M.2 Sata SSD. Not too large as there are BIOS restrictions as to the size of hdds.