Corsair CX400W 5vsb goes into safety loop

[Per Hansson]

Hi, I have a Corsair CX400W that has been in use for a few years.
Recently I had to replace it because the owner complained that his PC would shut off for no apparent reason (It has a UPS too...)

To make a long story short the PSU is finally here at me now and disassembled.
There seems to be no problem to rectify the AC, I have a constant 330VDC on the main filter capacitor.
The 5vsb is generated by a Fairchild FSQ0165 which has just Q0165R printed on top of it.

If I plug it in I get a nice and stable 5v output, but after a couple of minutes with a small load it goes into a safety loop: shutting down the output and then trying to restart it.
It does this if I plug in a PC too, it runs for a minute or two and then shuts down, at which stage 5vsb pulses all lights on the mainboard as it tries to restart...

I replaced all electrolytic capacitors in the 5vsb section but it made no difference.
I also replaced diode D101 for a UF1007 even though it measured good, with no difference, I replaced it because the soldering on the back looked bad.
All the other diodes and zener diodes in the area measure good in circuit.

Between VCC and GND sits a 100µF 50v Suncon WX cap now, it was a 100µF 25v OST RLS before.
The voltage on this cap starts off at 16v but rises slowly to 18v at which stage the IC goes into it's safety loop.
The optocoupler has around 2 > 3v on the primary side depending on load.
On the secondary side was a 3300µF 10v OST RLX that I replaced with a 6.3v Rubycon MBZ
And finally a 16v 220µF OST RLS right on the 5vsb output cable that got replaced with a Rubycon ZLH 220µF 16v.

Once the chip is in the safety loop it's enough to remove the load for it to be able to stabilize the 5vsb again, but if I don't it seems to be stuck there forever.
I also noticed that if I'm probing the feedback signal on the optocpupler it seems to not enter the safety loop as quickly.
But if it "should have" then it's enough that I remove my probes from the feedback pins and it will enter the safety loop instantly.
It's worth noting that with no load I see around 720vdc between drain and ground, but with a load this goes down to a more sensible 600vdc.
Is not this too high, the datasheet says max 650v for the drain?
Any ideas what more I can check?

[stj]

is the output rectifier D121?
and what's that stain on the ceramic disc behind the chip?

[Per Hansson]

Yes, D121 is a Fairchild SB540 Schottky Rectifier.
It takes the "AC" from the transformer and turns it into 5vdc
My meter manages to read ca 0.5Mhz switching frequency and 3.3vac on the transformer, no idea if it's a true reading or it's just getting fooled.
It goes bananas if I don't have a load, probably frequency out of range

As for the ceramic capacitor I think it's just some strange artifact in the photo, I can't see it myself if I look at it with a magnifying glass...

[ReeceyBurger123]

Think this is what Stj meant Per ???

[Per Hansson]

Yes and what I replied was that I think it's just some strange artifact in the photo.
I took a new picture of it, and also the optocoupler which had a small mark.
But that on the other hand looks to be just a "QC" mark done with a marker pencil...

[stj]

the reason i mentioned d121 is maybe it's breaking down under load - i'v had that on a psu once.
had to find it with freezer spray!

[Per Hansson]

Thanks for the idea stj, I replaced the 5AH Schottky with a 3AH stolen from this PSU but unfortunately it made no difference.
That is 5vsb worked fine for a few minutes and then went into safe mode again.
It's a little annoying how they have grouped the safety conditions into the same trip point.
It's not exactly easy to figure out what is causing it to trip that way

[stj]

maybe swap the opto-isolator next.

[Per Hansson]

Just did, found an exact match in this disaster of a PSU.
No go though, 5vsb worked fine for a few minutes and then went into safe mode again just like before.

Originally Posted by Per Hansson

It's worth noting that with no load I see around 720vdc between drain and ground, but with a load this goes down to a more sensible 600vdc.
Is not this too high, the datasheet says max 650v for the drain?

I'm stuck on still thinking this is wrong, could it be that Zener diode ZD102 is not doing it's job?
From what I can see it seems to have E150A written on it, so 150v Zener?
Don't think I have anything to replace it with though.

[Per Hansson]

I replaced the Fairchild FSQ0165 5vsb chip and the PSU is now able to handle a load a little longer than before, but it's not close to fixed.
Probably the new FSQ0165 just has a bit different tolerances for when it goes into it's safe mode...

So I would like to continue my original plan, I took out ZD102 which I suspected in my previous posts, it's marked: P6KE150A so it's actually a TVS diode...
But on the backside are more Zener diodes which I have no idea what their ratings are: ZD103 & ZD101
And while I'm at it it might also be a good idea to check D102 & D103
So does anyone have a schematic for the 5vsb in this PSU?
It's a Seasonic S12-II platform...

[Behemot]

Corsair CX Seasonic? No way, that's CWT. It's a long shot, but I can only think about asking Jon, maybe he'll get at least the information about what parts those are from some enginner. Don't think they'd send you full schematic…

[Per Hansson]

CWT? No way, this looks very much like a Seasonic to me.
I'll give Jon a ping and see if he is able to help, thanks for the idea.

http://www.hardwaresecrets.com/corsa...supply-review/
"Like most power supplies from Corsair, CX400W is manufactured by Seasonic"

http://www.bit-tech.net/hardware/psu...w-psu-review/4
"Underneath the soldering quality is also excellent - it's very neat with no large blobs (just as we'd expect from Seasonic)."

http://www.anandtech.com/show/2697/8
"The inside reveals a typical Seasonic design"

http://www.x86-secret.com/dossier-36...400_Watts.html
La majorité des alimentations Corsair sont conçues autour d’une base Seasonic. La première chose à faire était donc de vérifier de quelle plateforme est issue la CX400. Un simple coup d’œil à l’architecture interne ne laisse planer aucun doute: Cette alimentation repose donc elle-aussi sur la même base que les autres modèles haut de gamme de Corsair, c'est-à-dire la Seasonic S12-II.

[Behemot]

Oh yeah, it's the 400W version, the only one made by Seasonic…I can see that now.

Than it's classic OEM version of S12II…those are quite common but I don't know if anybody made or got any drawing of that…

[fzabkar]

Is the 5VSB output stable and ripple free all the time? If so, then ISTM that the IC, optocoupler, TL431, etc are all doing their job. However, the fact that the auxiliary supply is slowly climbing from 16V to 18V suggests that the IC is gradually driving harder and harder to compensate for an overload on the transformer windings.

Is there an 18V zener on the VCC pin? Is the IC shutting down because its OVP sync pin is being triggered?

To me the problem looks like a leaky rectifier or filter capacitor on the 5VSB output, but both components have been replaced. ??? Ideally the 5VSB and VCC outputs should track each other according to the turns ratio. Perhaps it would be an idea to interpose an 0.1 ohm sense resistor in the output and monitor the current.

[Per Hansson]

Thanks for your post fzabkar, yes this PSU has turned out difficult to troubleshoot.
Unfortunately I don't have a scope, but according to my multimeter the 5vsb is stable as a rock.
The testing equipment I run from it before it goes into the safety reboot loop also has no problems, so it would imply that it infact is stable.

Yea I think that Zener diode ZD103 is a 18v Zener, it sits between VCC & GND and would be a sensible rating since the FSQ0165 is rated for max 20v on the VCC pin.
It's also sensible from the perspective that this is the voltage I read before the thing goes into the safety loop...
Actually with the new FSQ0165 VCC can go up to 18.2v now before the unit shuts down.
There is also a 100ohm resistor here so the voltage on the FSQ0165 is slightly lower at 16.9v, the Zener diode sits on this side of the resistor too so that's what it sees as well.

The other SMD Zener diode ZD101 I'm not sure about, but it sits on the feedback signal which is rated for up to 9v so I guess it's sensible that's what it's rated for.
As for the TVS diode marked ZD102 on the topside I left that out now for a test, since a big leakage here would case problems since it sits between drain and B+ from the main bulk capacitor, but it made no difference at all.
(The unit also did not blow up and continued working as before so the problem was not here )

I have not connected a sense resistor but I do have a quite accurate power meter to measure the load of the whole PSU.
With nothing connected it shows 0.0w draw.
With a small ~0.4w lightbulb for load it shows 0.4w.
With the lightbulb and a small fiberswitch it shows 3.0w draw.
All of this is sensible because the devices by themselves do draw that amount of power if measured using a DMM in series...
The 5vsb is rated for up to 2.5Ah or 12.5W according to the power rating plate provided on the PSU by Corsair so we are not even close to actually overloading anything...

[stj]

so the power for the chip passes through a resistor, then has a zener across it to ground?

that sounds like a shit attempt at a voltage regulator to me!

[CapLeaker]

Had a PSU a while back that would work for a while, then shut down. One of the voltage rails was jumping up and down after running a while, then shut off. Traced it to a couple of inductors that I had to desolder, take out, clean the through hole and resolder it. Works fine as kinds now. I finally found the problem by watching the voltage rails and using my hot air soldering station, heating up the inductors.

[Per Hansson]

stj: no the zener is a protection device.
The Zener will only conduct if the voltage crosses the breakdown voltage, which most likely is around 18v.
Thus protecting the FSQ0165 from too high a voltage (it's rated for max 20v on the VCC pin)

I have attached an annoted version of my picture above, it's the closest thing to a schematic so far
The two diodes in the top: D102 & D103 are marked "A1 W73" on their casings.

[stj]

if there is a resistor between the source and zener it wont protect the chip.
the zener will conduct, but because of the resistor it wont produce enough of a load to shut the source down.
so it will just keep conducting till it burns out.

if it was a safety device it would drive the base of an scr to short the supply - a crowbar circuit.

when you find one after a resistor it's often intended as a dirt cheap voltage regulation circuit.

so what is the source of the original supply voltage?

[Per Hansson]

Ok, it serves another purpose too, during startup the FSQ0165 has an internal switch that takes power from the bulk capacitor and charges the VCC voltage up to 12v.
If this where to fail internally I guess we would be conducting 325VDC here so then it works as a safety device for that as well, though there is a resistor on this side too.

In any case when that is done the switch closes and the device is powered from the transformer.
But seeing as this is high frequency pulsed DC I can't tell you what the voltage is on the other side of the diode (D103) because my meters can't get a reliable reading at this frequency.
But obviously since we have a stable supply after the diode it's somewhere in the range of 20V
My UNI-T meter says it's around 0.2 > 0.5Mhz and the voltage varies depending on load, but as I said it's not a reliable reading.
E.g it can sometimes show 5.7v which is completely impossible since we have a stable ~17VDC output after the diode depending on load.
And my Fluke meter is not able to show anything on either AC or DC so it's similar to my experience in this thread with these meters.