OCZ Fatality 700w PSU Repair Help

[Quagmire]

Looking to re-cap the above flaky PSU. I pulled the caps that were visibly bulged and disgustingly, they're Teapos (SC series). The damaged ones seem to be 3300uf/10V . Looking around at Digikey, I see that Panasonic have some 3300uf ones but they're all 12.5mm in diameter; the Teapos were 10mm. Doubt I'll have the room to get anything bigger in there. The largest 10mm caps that Panasonic has are 2700uf (FS Series).

Another fly in the ointment, I pulled 5 caps (all from the same location) that appeared physically identical without realizing one was a 2200uf/16V. Now I'm not sure which position on the board it sat in. How big of an issue is that?

[Per Hansson]

Re: OCZ Fatality 700w PSU Repair Help

Member Behemot here has caps of the right type afaik.
As for the 16v cap that is easy: just measure to the 12v leads from where the caps are.
The 16v cap will have a low ohm path to 12v output...

https://www.badcaps.net/forum/showthread.php?t=39123

[Quagmire]

Re: OCZ Fatality 700w PSU Repair Help

Thanks Per Hansson, I tested the resistance between the cap locations to the 12V rails and I'm not seeing low ohm readings; all 5 measure roughly the same (4 measured 105 Ohms and one was 110 Ohms). To compare, I measured from one of the adjacent 25V caps and it measured 0.11 Ohms. Maybe I just don't know what I'm looking for.

[momaka]

Re: OCZ Fatality 700w PSU Repair Help

It's possible that the 16V cap was used on the 5V rail as well. I know Enermax did that in some of their units (they used 16V/2200 uF caps for everything, as it was probably cheaper to buy the same caps in bulk than mess with different sizes.) That said, still make sure to double and tripple check that none of the caps you removed have their (+) leads connected to the 12V rail. Also check if that 16V cap wasn't for the -12V rail - if it is, its positive (+) lead will be connected to ground (i.e. low resistance to ground on + side.)

Regarding the caps... you don't necessarily have to use Panasonic only. Chemicon KY and KYB should be readily available as well. I actually prefer those, as they are very close to Teapo SC spec-wise. You can also use 6.3V caps in place of the 10V caps, if those make it easier to find 19 m diameter caps. The lower voltage won't be an issue because both 6.3V and 10V caps can only be used on the 3.3V and 5V rails.

And if you really are deadset on using Panasonic, then most likely those 2700 uF FS series caps will work fine too. IME, 2700 uF caps have never caused an issue when I used them to replace 3300 uF caps. After all, electrolytic caps have a +/-20% tolerance on capacitance. -20% corresponds to 2640 uF. So technically, good 2700 uF caps will work too (and they often do.)

[Quagmire]

Re: OCZ Fatality 700w PSU Repair Help

That said, still make sure to double and tripple check that none of the caps you removed have their (+) leads connected to the 12V rail.

I'm fairly new at all this but I did a continuity check between the positive hole of each capacitor location to the 12V rail and got no response. Is that what I was supposed to do?

Also check if that 16V cap wasn't for the -12V rail - if it is, its positive (+) lead will be connected to ground (i.e. low resistance to ground on + side.)

Did a continuity check to GND from the positive legs for all the locations and again, got nothing.

Pain in the ass that I misplaced the sheet of paper that I had recorded the locations but I was able to figure out 4 of the 6 caps by matching up the white silicone adhesive still attached to the sides of the caps and the components they were glued to inside the PSU. The result is I'm left with one location that has continuity to a lead labelled 5VS+ and the other one connected to a lead labelled 5VSB. Interestingly, this latter spot has a neighbor that is also connected to this lead and houses a 16V 1000uf cap. Could that be a hint as to where the 2200uf 16V cap belongs?

Regarding the caps... you don't necessarily have to use Panasonic only. Chemicon KY and KYB should be readily available as well. I actually prefer those, as they are very close to Teapo SC spec-wise. You can also use 6.3V caps in place of the 10V caps, if those make it easier to find 19 m diameter caps. The lower voltage won't be an issue because both 6.3V and 10V caps can only be used on the 3.3V and 5V rails.

I took a peek at the KY and KYB series and it appears in order to get a cap with a 10mm diameter, I'll need to go with 2700uf or 6.3V. The 3300uf 10V ones are all 12.5mm. Makes me wonder if the Teapos were actually 3300uf at all.

[momaka]

Re: OCZ Fatality 700w PSU Repair Help

I'm fairly new at all this but I did a continuity check between the positive hole of each capacitor location to the 12V rail and got no response. Is that what I was supposed to do?

Yes.

Furthermore, you can also check continuity (or preferably the resistance, actually) between each positive hole and each PSU rail. So as an example, if you're trying to figure out, say capacitor C12345, you check resistance to 12V rail, then to 5V rail, then to 3.3V rail, then 5VSB, and then -12V rail. Whichever one of these readings shows a resistance of 1 Ohm or less (i.e. same as when you have multimeter probes shorted together in resistance mode on lowest resistance scale), that's likely the rail to which the cap is connected to.

Which leads me to this:

The result is I'm left with one location that has continuity to a lead labelled 5VS+ and the other one connected to a lead labelled 5VSB.

There you go, you figured this out for yourself. This is exactly what I meant with the above example.

That means one cap is for the 5VSB rail. As for the cap connected to 5VS+.... not sure if that's part of the 5V rail, or some internal IC/chip supply rail. But if it's for small cap (i.e. less than 470 uF), it's probably an internal rail of some sort.

Interestingly, this latter spot has a neighbor that is also connected to this lead and houses a 16V 1000uf cap. Could that be a hint as to where the 2200uf 16V cap belongs?

Yes.
Sometimes manufacturers use 16V caps on the 5VSB rail, because the 5VSB pretty much always uses flyback topology. By using 16V caps, they are essentially getting the same capacitance, but with lower ESR specs (and higher RC) This is preferable, because flyback topologies tend to be noisy and do benefit (in terms of noise suppression) from very low ESR caps.

If this is the case, and that "lost" 16V cap really is connected to the 5VSB rail, then you can safely use a 6.3V or 10V cap in its place. Just match or exceed the ESR specs of the original, and all should be good. Or you could use a 16V cap again. No big deal either way.

I took a peek at the KY and KYB series and it appears in order to get a cap with a 10mm diameter, I'll need to go with 2700uf or 6.3V. The 3300uf 10V ones are all 12.5mm. Makes me wonder if the Teapos were actually 3300uf at all.

They probably were. But clearly that spec didn't last for too long.
That said, it shouldn't really matter if you go with 10V/2700 uF or 6.3V/3300uF caps. Both will work fine, since these are for the 3.3V and/or 5V rails (which modern PCs barely use). Also, I just checked some KY 6.3V/2700 uF caps in my stock, and they read closer to 3000 uF (and no, they weren't leaky). So that's very much close enough. In my case, I used one to replace an OST RLX 10V, 3300 uF cap for the 5VSB of an Antec Earthwatts EA-430 PSU. Worked fine.

[Quagmire]

Re: OCZ Fatality 700w PSU Repair Help

As for the cap connected to 5VS+.... not sure if that's part of the 5V rail, or some internal IC/chip supply rail. But if it's for small cap (i.e. less than 470 uF), it's probably an internal rail of some sort.

Took a look today in better lighting and the label 5VS+ is a small red wire grouped together with a bunch of other red wires. Parting those a bit, I was able to see another label: +5V which of course go to the molex, SATA and 24 pin motherboard header. Mystery solved, all 5V related. I'm a learnin'

5VSB is a small purple wire and goes exclusively to the 24 pin header connector. Googling that, it's 5V standby.

Sometimes manufacturers use 16V caps on the 5VSB rail, because the 5VSB pretty much always uses flyback topology. By using 16V caps, they are essentially getting the same capacitance, but with lower ESR specs (and higher RC) This is preferable, because flyback topologies tend to be noisy and do benefit (in terms of noise suppression) from very low ESR caps.

If this is the case, and that "lost" 16V cap really is connected to the 5VSB rail, then you can safely use a 6.3V or 10V cap in its place.

Thanks, so I guess we've determined where that pesky 16V cap belongs.

Just match or exceed the ESR specs of the original, and all should be good. Or you could use a 16V cap again. No big deal either way.

Okay, so capacitance and voltage I get but ESR and ripple current, I'm way out of my league.

[Quagmire]

Re: OCZ Fatality 700w PSU Repair Help

In my last post, I indicated that ESR and ripple have me stumped but I guess you can add impedance to that list as well. So for example, if I want to replace a Teapo SC cap with the following specs:

- 2200uf 25V (1.810A ripple @ 100KHz, 29 mOhms impedance)

Looking at DigiKey and grabbing a few samples, I found one somewhat close:

Rubycon YXJ: 2.23A @ 100kHz, 32 mOhms

Then a bunch more with specs similar to these that seem way off:

Panasonic FR: 3.63A @ 100kHz, 13 mOhms
Nichicon UHW: 3.2A @ 100kHz, 15 mOhms

[momaka]

Re: OCZ Fatality 700w PSU Repair Help

Okay, so capacitance and voltage I get but ESR and ripple current, I'm way out of my league.

Don't worry, it's still pretty simple. You just need to find the datasheet of the original capacitors (looking for "Teapo SC" should produce some results on just about any sane search engine) and then compare to the datasheet of the new caps.

The way to do it:

- ESR of new capacitors should be the same or lower value than the original capacitors. In general, typically anywhere from 25% higher to 150% lower ESR should be OK. In some circuits, even 200-300% lower will be OK... but try to avoid that, if possible, unless you really have no other choice (after all, even a cap with "inappropriate" specs will usually be better than a failed cap that is completely out of spec.)

- RC (ripple current) of new capacitors should be the same or higher than the originals. Again, this figure can go from 20% lower than the original value to any value that is higher than the originals (i.e. there is no upper limit on how high you go for RC.) RC simply indicates how much "AC stress" the cap can handle before it starts to overheat internally. So the higher the RC, the less the cap will heat up internally when performing in the circuit.

In my last post, I indicated that ESR and ripple have me stumped but I guess you can add impedance to that list as well.

Ah, forgot to mention this.

When you see "capacitor ESR" mentioned for a switching circuit (like ATX PSUs, for example), this typically implies impedance in the datasheet.
While impedance is not exactly equal to ESR... at high frequency around 100 KHz that the datasheets use for comparison, ESR and impedance are very nearly equal. So for all practical purposes when comparing datasheets, we say that ESR = impedance (@ 100 KHz).

So for example, if I want to replace a Teapo SC cap with the following specs:

- 2200uf 25V (1.810A ripple @ 100KHz, 29 mOhms impedance)

Looking at DigiKey and grabbing a few samples, I found one somewhat close:

Rubycon YXJ: 2.23A @ 100kHz, 32 mOhms

There you go, you got this!
That Rubycon YXJ cap is a very close match for the Teapo SC, despite its ESR (impedance) being rated slightly higher at 32 mOhms. (remember, ESR should be the same or lower, if possible... though in this case, it's within 20% of the original, so this cap will work just fine.)

Then a bunch more with specs similar to these that seem way off:

Panasonic FR: 3.63A @ 100kHz, 13 mOhms
Nichicon UHW: 3.2A @ 100kHz, 15 mOhms

Both of these caps have ESR (or, again, impedance if you like ) that is a bit lower than suggested. So they are not really ideal replacements, technically speaking. However, even these caps are better than the original failed Teapo SC and in all likelyhood will work fine. So don't completely rule them out as possible replacements... but if you can find something a bit closer to the ESR, like you did with the above Rubycon YXJ caps, that would be better.

[Quagmire]

Re: OCZ Fatality 700w PSU Repair Help

Thanks again momaka; learnin' lots. From the bit of prior searching I did, I had an inkling that ESR being too low could be a possibility but the PDFs that hinted at that were way over my head and I had no idea that impedance was a good metric to use.

As for ripple current, nothing I found really explained that the value in the spec sheet is the maximum spec'd for the cap so you cleared that up. Makes sense now.

So based on my newly minted knowledge, I spent way too much time (actually I kinda enjoy perusing specs) on DigiKey and this is what I've come up with so far: (pic attached, yellow are my prelim picks). I did manage to find one 3300uf 10V cap but it's impedance is a bit low at 15 mOhm.

[momaka]

Re: OCZ Fatality 700w PSU Repair Help

^ Looking good. You did your homework correctly.

And just a further note about PSU 5VSB circuits: it's usually not a big deal if you go even 2x lower on the ESR on those... though sometimes that can make the transformer produce a light audible "whine" or "squeal" at low-load. But it's not an issue. Did many monitor PSUs and 5VSB circuits with Panasonic FR/FM/FS (or equivalent-spec caps). All still work to this day.

Also another note regarding caps and ESR: general purpose (GP) caps typically won't list ESR/impedance in the datasheet, and ripple current rating will be @ 120 Hz. So what do you do when you encounter caps like that in a PSU? - You can replace those either with GP caps again or go with "entry-level" low ESR series. That would be Panasonic FC, Nichicon PS/PM/PJ/PW, United Chemicon LXZ/LXV/LXY, and Rubycon YXJ.

[Quagmire]

Re: OCZ Fatality 700w PSU Repair Help

I finally got around to placing the order with Digi-Key.

Essentially I'll be replacing all the output caps but decided not to touch the large input caps since they appear fine, aren't cheap and look to be a bitch to de-solder. Also, from what I read, they rarely go bad.

Once I finally get the new caps installed and the PSU sufficiently tested, I report back.

[momaka]

Re: OCZ Fatality 700w PSU Repair Help

Sounds good! Yes, please do. Always good to hear a success story and also for the sake of thread completeness.

[Quagmire]

Re: OCZ Fatality 700w PSU Repair Help

So this weekend, I managed to set aside some time to get the new capacitors soldered in and the PSU tested. I'm pretty much a soldering but I've been watching a ton of YouTube videos over the past week. One through-hole gave me a fair bit of trouble (it was a corroded from a leaky cap). Used flux but even with that, the solder joint isn't great. Seems to hold fine nonetheless.

For testing, I really gave it a workout. Used a bunch of benchmarks, games, encoding etc. By far the biggest stress was the OCCT PSU test. Man, that applies a load like I've never witnessed before; essentially a worst case scenario. Regardless, the unit blew cold air throughout all of it. Here's the results I got using that benchmark with my multi-meter "back-probing" the 24 pin ATX connector:

3.3V

• 3.360 idle
• 3.325 load

5V

• 5.061 idle
• 5.027 load

+12V

• 12.117 idle
• 12.017 load

-12V

• 11.986 idle
• 12.028 load

+5VSB

• 5.036 idle
• 5.005 load
• 5.050 sleep

Anyway, thanks for all the help. It was a great learning experience which was the primary reason for undertaking this project. Of course, saving $140 for a new power supply is a bonus.

[sam_sam_sam]

Re: OCZ Fatality 700w PSU Repair Help

Another one saved from the land fill

Good work

[momaka]

Re: OCZ Fatality 700w PSU Repair Help

Awesome! Great to hear it's working properly again.

Yeah, OCCT PSU test is basically a full-blown CPU + GPU test. Thus, if you have a high-end GPU, you probably really will see the absolute worst power draw. I just tried that OCCT PSU test on a PC with a PSU on which I performed some "repair experiments" and... PSU kept shutting down. It was a nicely-designed Enermax unit, though, so it shut down gracefully each time I tried the test. No fireworks!

[Quagmire]

Re: OCZ Fatality 700w PSU Repair Help

Another one saved from the land fill

Good work

Thanks. You're right, another great reason to learn electronics repair.

Awesome! Great to hear it's working properly again.

Yeah, OCCT PSU test is basically a full-blown CPU + GPU test. Thus, if you have a high-end GPU, you probably really will see the absolute worst power draw.

In my case, my old PC has an i5 overclocked to 4GHz and two GTX-285's in SLI so the power draw has got to be pretty insane. The whole machine needs replacing but at least now I know the power supply can be re-used in a new build; would never do SLI again and new GPUs are much more power efficient.

I just tried that OCCT PSU test on a PC with a PSU on which I performed some "repair experiments" and... PSU kept shutting down. It was a nicely-designed Enermax unit, though, so it shut down gracefully each time I tried the test. No fireworks!

Yea but fireworks are so pretty. Seriously though, what "experiments" did you perform?

[momaka]

Re: OCZ Fatality 700w PSU Repair Help

In my case, my old PC has an i5 overclocked to 4GHz and two GTX-285's in SLI so the power draw has got to be pretty insane. The whole machine needs replacing but at least now I know the power supply can be re-used in a new build; would never do SLI again and new GPUs are much more power efficient.

Ha! Yeah, two GTX 285 cards will indeed double as room heaters. 204 Watts TDP per GPU x 2 + the CPU TDP (around 100W?)... and that's indeed 1/2 KW of heat to get rid of (in the worst case, of course).

I would say that new GPUs are more "power efficient". An RTX 2080 Ti is rated for 250 Watts TDP. So two of them in SLI will draw even more power than the GTX 285 cards. Of course, newer cards are able to do "more work" with the same amount of power, so in all likelyhood, if you have a game or piece of software that pushes your pair of GTX 285 cards to their limits, chances are, a single RTX 2080 will hardly go past 33% utilization (and TDP probably).

So in the end... yeah, you will get a lot less heat output.
But you being located in Canada, I don't think you find heat "unwelcome". )

Yea but fireworks are so pretty. Seriously though, what "experiments" did you perform?

Eh. Nothing fancy. Just removed the APFC circuitry out of a PSU due to not having the right input caps to use. Doing so lowers the voltage that the primary side has to work with, which apparently reduces the power output capability of the PSU... and by that, quite a bit more than I expected. It was a 525W Enermax PSU, and with this temporary "repair", I couldn't go higher than 260-280 Watts before the PSU shut down. Basically a GTX560 + Q6600 CPU + OCCT PSU test = shut down.
But hey, nothing blew up!
Here is the thread for that PSU, btw:
https://www.badcaps.net/forum/showthread.php?t=72147
(And note, I haven't updated it yet with the results of testing this mod... but will do that sometimes this week. )

[Quagmire]

Re: OCZ Fatality 700w PSU Repair Help

I would say that new GPUs are more "power efficient". An RTX 2080 Ti is rated for 250 Watts TDP. So two of them in SLI will draw even more power than the GTX 285 cards. Of course, newer cards are able to do "more work" with the same amount of power, so in all likelyhood, if you have a game or piece of software that pushes your pair of GTX 285 cards to their limits, chances are, a single RTX 2080 will hardly go past 33% utilization (and TDP probably).

RTX-2080Ti? Yea, won't be buying one of those, let alone two.

Eh. Nothing fancy. Just removed the APFC circuitry out of a PSU due to not having the right input caps to use. Doing so lowers the voltage that the primary side has to work with, which apparently reduces the power output capability of the PSU... and by that, quite a bit more than I expected. It was a 525W Enermax PSU, and with this temporary "repair", I couldn't go higher than 260-280 Watts before the PSU shut down.

Huh, interesting, never would have known that.

[RJARRRPCGP]

Re: OCZ Fatality 700w PSU Repair Help

250W or the like? Sounds like the Radeon R9 290X. Those are one of the AMD GPUs that have GP-GPU computing as their forte.

I wonder if even my R9 280X would draw less than that, even with 2x 6-pin plugs required, IIRC. (or similar)