ATX PSU with no "power good" -- thoughts please to help diagnose.

[fudmuffin]

Hi,

I have 2 identically faulty ATX computer power supplies (model FSP FSP220-60LE), and would like some preliminary advice before I begin any deeper diagnosis. The PSUs are from QNAP branded NAS systems which were "dumpster dives" so unfortunately I don't know their history. Both power supplies exhibit the same issue and I'd like to get them working.

When connected to the (working) NAS, the motherboard standby light comes on but the computer will not power on when the power button is pressed. When testing the power supplies while running (with a power supply tester or with a DMM), all voltages are within the ATX voltage spec however the PWR_OK ("power good") signal does not come on, remaining at 0v.

The components inside the PSUs look brand-new, with no obvious problems, no dust, no bulging caps or burn marks, all looks as though the PSUs have never been running. It looks to me perhaps there is either a problem with the "power good" detection circuitry (i.e. broken or too sensitive), or there may be some other issue like ripple, causing the "power good" circuitry not to ever activate the "power good" signal.

There is a trimmer mounted on a daughter-board near what I would guess are the brains of the PSU, and it is labelled ISENSE (see photos). Does anyone know what that trimpot is likely to do? My best case is that it would adjust the sensitivity of the "power good" detection, but I have no idea if that may be its purpose. I'd love to know what this trimmer's function is.

More generally I'd also appreciate any thoughts/guesses as to what might be the problem or what initial steps I could take to diagnose the issue.

Unless there is a better idea my next steps would be to measure ripple on the outputs again (I have already tried this and not found anything odd), then try to see what effect the ISENSE trimpot has, before measuring individual components. If there is nothing obviously wrong with any individual componets then I would probably not try to troubleshoot the circuit while the PSU is live (and of course I must still take extreme care to discharge the large capacitor prior to handling the circuit!).

I have included pictures showing the PSU label, insides, and voltages, as well as a closeup of the "ISENSE" trimmer...

The specification for the PSU can be found at https://produktinfo.conrad.com/datenblaetter/1300000-1399999/001396911-da-01-en-FSP_FORTRON_FSP220_60LE_220W_PC_NETZTEIL.pdf.

Cheers!

[PKTKS]

Those kind of thing must be tracked down by the wires.

POWER OK  comes from the "FULL" 5V branch.

Usually it is just a BJT switch triggered by some sort of delay
network which starts when the supervisor starts.

Rare chipsets emit this signal - which is fine as ATX specifications
are different and the timing is also different

So just the bare bones  logic should be implemented

Track the  "GRAY"  signal (PIN8 if I recall)  down to the BJT
switch and why the 5V line is not putting it out.

Should be a good start point
Paul

[PKTKS]

** BE CAREFULL **   as those metal dissipators are almost never isolated...

so the  mains dissipator should have a roughly 400V live  peak AC
ready to put you in damn it mode..

2 cents.

[PKTKS]

as a matter of fact SOME CHIPSETS DO OUTPUT PG.

If I recall ..  one is SG6105 which outputs PG
and generally is tied to 5VSB not 5V_ALW

Some others may as well - take note of your controller
as you may be lucky to have this pin

as mundane chips like TL494 will not have PG pin..
it is common to have a 339 instead a BJT as well in those circuits

Paul

[magic]

I is current so ISENSE is current sense. Probably something to do with overcurrent protection, I wouldn't touch that stuff.

Be sure that there is no ripple on any output, trace where PG signal is coming from.

Note that a big capacitor may still give you a jolt even after discharge because of dielectric absorption. With really high voltage and high power capacitors, people put permanent shorts across them during handling.

[fudmuffin]

Just want to quickly say thank you for all the very helpful replies. I've had to put the repair of these PSUs on my back-burner and will reply here again when I hopefully can grab a chance to get them back on the workbench.

I do still wonder if the ISENSE trimpot may have something to do with the sensitivity of the "Power Good" detection, but I will take the advice above not to touch it until I see how it's connected in the circuit, and have tested properly for ripple and any other reasons as to why the PSUs seem to be fully operational yet aren't giving out a PG signal.

In preparation I have constructed a capacitor discharger with LEDs (based on this and this circuit) so that I can discharge the big capacitors with a bit more visual assurance than just with the 2k ohm ceramic resistor I have used in the past.

Cheers!

[PKTKS]

I do still wonder if the ISENSE trimpot may have something to do with the sensitivity of the "Power Good" detection, but I will take the advice above not to touch it until I see how it's connected in the circuit, and have tested properly for ripple and any other reasons as to why the PSUs seem to be fully operational yet aren't giving out a PG signal.
(..)

That depends only in the supervisor chipset used.

Vast majority ATX are "cheapos" and supervisor does not even
count with a power good pin.

2 or 3 BJTs delayed as switch do the job - no relation with ISENSE
ISENSE will just entering "OVERCURRENT" protection

But that depends really on the supervisor. Being all cheapos
don't expect nothing beyond bare bones .. really

Paul

[fudmuffin]

I finally got around to dismantling and investigating these PSUs further and found that both units had a blown diode and resistor in exactly the same place, and sure enough this area is connected to the chip responsible for the "power good" signal, the PS223H.

I tried replacing the burnt components on one board by canibilising the other but it doesn't seem to have helped, so perhaps the PS223H chip has been damaged.

Here are photos of both relevant PCB's showing burnt components and also a pic of the reverse side showing the ICs.

Out of interest, does anyone know what kind of diodes these would be? I can't see any markings...

I've chosen not to persist with this repair attempt because two units failed in this way doesn't leave me with much confidence that simply replacing the blown components will be sufficient/safe, and didn't seem to fix the unit anyway.

I'd love ideas about what I could have done next, but but I think I'm out of luck on this repair and will move on to the next one

As you can see from the pictures there is a "missing" diode (HD4) next to the one that has overheated (HD1) and I couldn't help wonder if it would helped to add a diode there...

The fact that both PSUs have failed in this same way makes me think perhaps there was a fault in the original design.

[PKTKS]

Well 2 cents of opinion... 

I deal with PSUs  a long time....  one thing that I think
will never be "solved" is that the more power you cram
into a smaller space... the more dangerous it become.

This board is a classic cheapo with high power density....

A single rail (+5V_ALW) handles the PG output.

**IF** the soft start fails.. the +5V will go wacko
and so does all things referred to that  "rail".

A small SMD board concentrates heat and components
age very fast... things tend to fail shortly. More and more.

In this case ATX below say 400W are still cheap enough
to not bother... and replace - confidence result

But newer  ATX beyond 800W are hard to just replace
as prices went above the line. So does the 1000W class.

And unfortunately  the power density has become insanely
dangerous and impossible to fix without the schematics.

Too much to guess.
 Replacing caps ad hoc just will not fit a safe list.

Just replace the thing for safety reasons
don't waste time drawing schema for this thing...
Paul

[magic]

Try to figure out what they are connected to and what's their role.

The circuit is probably similar to the typical application schematic from the datasheet.

[PKTKS]

(..)

The circuit is probably similar to the typical application schematic from the datasheet.

yes very close to .. including the PFC companion chip.

But also another way. The board is marked with the maker and serial
there is a chance to obtain a new board if price is worthy

otherwise time and cost may be not worthy...

[fzabkar]

Do the diode and resistor connect to the PGI pin (#1) of the PS223H IC?

[fudmuffin]

Do the diode and resistor connect to the PGI pin (#1) of the PS223H IC?

No, interestingly the annode of HD1 has continuity with both pins 8 and 12 of the PS223H. There is only 0.1R between pins 8 and 12. These are 12V under/over voltage protection input pins A+B. I'm not sure if that's expected, but doesn't seem to match the reference circuit. On one of the units there is also continuity between pins 2 & 9.

But also another way. The board is marked with the maker and serial
there is a chance to obtain a new board if price is worthy

otherwise time and cost may be not worthy...

I'm not able to find any source for a replacement PCB, is there somewhere I should look?

[magic]

There is only 0.1R between pins 8 and 12. These are 12V under/over voltage protection input pins A+B. I'm not sure if that's expected, but doesn't seem to match the reference circuit.

It matches the reference circuit if there is only one 12V rail and both RS12(1) and RS12(2) are connected to the same point. Then you see their series combination between 12V(1) and 12V(2).
This is common on those "dual rail" PSUs. It allows separate OCP on both 12V rails without actually using two transformer windings, diodes, capacitors and so on.

On one of the units there is also continuity between pins 2 & 9.

That's wrong. Perhaps something is shorted on the PCB, perhaps the IC is blown.

[PKTKS]

(..)
I'm not able to find any source for a replacement PCB, is there somewhere I should look?

There are just a few real makers on the BUZ.

Vast majority of "BRANDS" are cosmetic "man-in-the-middle" sellers.
(like "CORSAIR" "THERMALTAKE" and 99% others)
(these folks do not "make" their PSUs)

Your board is no exception and the real maker is FSP

https://www.fsp-group.com/en/product/PCPSU.html   

Which may have parts for sell or inquiry ..
if worthy - best and safest method for repair

Paul