My new Seasonic has questionable caps.

[gonzo0815]

Re: My new Seasonic has questionable caps.

At an ATX PSu probably not, but i haven't done the maths to proof this. The only circuit i know, where the Simple switchers LM2576 and similar SMPS controller or switchers (which are pretty old today).

In such an circuit, the effective ESR of the o/p caps must have to be not to low, or the voltage regulation weren't that accurate any more.
But those caveats are not true for other controllers or other circuit with different topology.

From my point of view, the only problem could be an self oscillating phi filter (which would introduce a lot of noise into the o/p) or a lack of voltage regulation.

[PCBONEZ]

Re: My new Seasonic has questionable caps.

I'm not saying go crazy and slap solid polymer in there.
And I'm not saying do it for the heck of it.
I'm I really doubt going up to something like MBZ is gonna hurt if that's the only possible way to get a cap in there.

The 'supposed' ESR/Ripple ratings on those crap 10mm 4700uF OST and Fuhjyyu output caps -ARE- roughly around a 3300 MBZ (which can be had in 10mm).

I tore apart two L&C PSU's I had sitting in a box last night to compare and look-up the regulators used. One is a 350w and the other a 450w.
Turns out one is built cheap and the other is built EXTRA cheap.
- But they are built on the same PCB.
On output the 450w has:
3.3v - Single 1000uF/10v
5v - Single 1000uF/10v
12 - Single 1000uF/16v
The 350w has:
3.3v - Pair 2200uF/10v (filter added a 0.8uH inductor)
5v - Pair 2200uF/10v (filter added a 0.8uH inductor)
12 - Single 2200uF/16v
(All the caps in both are the same series of a logo I don't know yet.)

.

There's other little tricks people don't think about often.
Lets say you have a pair of paralleled 2200uF 10mm and 12.5mm won't go.
Do the math for ESR/Ripple in parallel to find the whole.
There might be a SINGLE 4700uF that is close enough ESR/Ripple and will fit in.
Two 1000 -> one 2200
Two 1500 -> one 3300
Like that.
If it works it works.

...

And DAMN IT!!
Now I need some 2200uF 10mm too!

.

[shadow]

Re: My new Seasonic has questionable caps.

And DAMN IT!!
Now I need some 2200uF 10mm too!

.

Well 2200uF capacitors in a 10mm package are not that hard to come by at least. Around where I live, 3300uF in a 10mm package is like impossible to get unless you order it from somewhere else and get them to send it over.

[PCBONEZ]

Re: My new Seasonic has questionable caps.

In 10 or 16 volt?
Tell me where.

[starfury1]

Re: My new Seasonic has questionable caps.

thanks bonez for the feedback

yeah it has bee discussed a fair bit the aqueous and non aqueous side of things
and true electrolyte formulas to vary

Guess heats the real enemy of any cap at the end of the day

"Generally speaking" was really cause I couldn't recall the finer details accurately, just the possible out comes so I was keeping my big mouth shut

There's other little tricks people don't think about often.

Thats an interesting point on the use of caps above, as you say one not often brought up
does require you know somewhat what your doing with the psu thought.
(so for the noobs , post before whacking anything in place)

Yeah me too whats the scoop Shadow?

Cheers all

[linuxguru]

Re: My new Seasonic has questionable caps.

Spectre - Just got here and took a look at the pcbonez/spectre sub-thread. Just a few cents to contribute:

1) The cost of going with good non-aqueous Japanese caps at the time of original build is *miniscule* - only the secondary caps are critical. Pcbonez's estimate of $1-$2 extra to the BOM is actually widely extravagant - it could be as low as $0.25 to $0.50.

2) There are other components where skimping doesn't pay much - the 1A FR rectifier on +5Vsb could be replaced with a 3A unit for pennies. The pair of 3A FR302 rectifiers on +12v can be replaced with a single Philips BYV32E-200 rated at 2x15A for less than $0.50. Yet we still see numerous examples of similar penny-pinching leading to quick failures.

3) There are some of us here who have taken throwaway Deer/Allied/Powmax/Leadman or similar poorly engineered power supplies and have actually rebuilt them with a few carefully selected higher spec capacitors, rectifiers and fans. After some experimentation, it turned out that these are over 10 times as durable as they were when they left the factory. Conclusion: some of these designs are just fine, especially Deer, but they've been built to die quickly, either deliberately or inadvertently by bean-counters taking liberties with component selection.

4) There are very well-engineered designs that have also been run into the ground by bad capacitors - Hipro and Acbel come to mind, some might also claim Channel-Well/Antec/Fortron etc. Just replacing the bad caps on a Hipro could lead to a 50x improvement in durability.

5) Fans are critical, but not the number. A single NMB/Pabst/Panaflo/Nidec ball-bearing fan with a 10k-100k hour endurance trumps a dozen sleeve bearing fans. If you have to use a sleeve bearing fan, at least use a drop of high-quality bearing oil/grease on the bearing - I've revived utterly stuck sleeve-bearing fans with a single drop of hobby/model-railroading oil, which then continue to run like they're new.

The cost of good engineering is not high, it doesn't add more than 5% to cost of a PSU.

[i4004]

Re: My new Seasonic has questionable caps.

my 2(or more) cents:
i dunno if such
"10 times as durable" and "50x improvement in durability."
generalizations are possible here.

if we (for example) presume even crappiest psus will last a year, you need 50+ years for a claim like the latter one. otherwise it's just not true.
(and cheapest psus usually will last a year)

about recapping:
http://www.jonnyguru.com/modules.php...ory4&ndar_id=8
so results of recap aren't guaranteed at all. infact there are chances you'll get worse psu out of it. this probably has to do with the fact that normal caps were replaced by low esr caps, for no good reason. who says low esr is always better?
i've also seen willawake doubting psu recapping(with low esr caps?) etc.
jonnyguru guy didn't check the differences between caps he put in and put out(esr and capacitance measuring come to mind), and that was important thing, it turned out.

so while i'm against "blind" recapping with low esr caps, i also don't see a purpose for sites like the one spectre does.
none of his tests tells us anything about durability and reliability; let me use car analogy: when you get a new car, do you keep your foot on the gas all the way down all the time? how will engine react?
now, fine, electronics are not car engines and can take more beating, but overall the simillarity is that accelerated test doesn't represent real life data that well, as pcbonez said.

i also must strongly support pcbonez and linuxguru when it comes to pricing issue; the costly psus are indeed costly for no good reason. component prices don't justify such jump in price(they essentially use same caps as cheaper psus, as witnessed by OP and many others..i think i'll post some pix of enermax psu; gold plated(they say anyway) fan grille, but pce-tur inside....). that is the aspect of hardocp tests i mind the most: it's essentially nothing but pimping of expensive psus for no good reason(this is something pcbonez mentioned, i believe).
so i find it weird how spectre says he needs to be aware of his site, and that he has a name etc. we all have a name and i think most here will give you their names publicly without problems, as i don't think this is a kidde forum where hypocrisy and hiding is everything.
i'll sure give my personal data without problems anytime, anywhere.
i'm actually proud of my name and who i am.
site talking almost exclusively about expensive psus is the site with suspicious moral.
if you were to pimp more expensive psus you need to tell us and show us exactly why we should pay 100$ more.
ie you need to tell us why we shouldn't get cheaper one. that data hardocp lacks.
(as said voltage/current measuring and accelerated(fully loaded) tests are not it).

also, let's take that site for what it is; it has niche audience, as majority of pc users don't push their machines that much, and as such it's rather irrelevant.
at least for me, as my main interests are durability and reliability in the field, not some lab tests.
(one more suggestion for that site; white text on black background? how's that wise? who's eyes can take it?)

[acstech]

Re: My new Seasonic has questionable caps.

I like to match the cap series to their replacements, I think it makes for a better recap. Or maybe I'm paranoid again.

Apparently I wasn't paranoid? I would like to see results after a proper recap.

[i4004]

Re: My new Seasonic has questionable caps.

i think we all would, but they didn't do that test.
ie it's a half-assed test. they did exactly half the job.
but caps were the only thing they changed, which means it's caused by caps.

[linuxguru]

Re: My new Seasonic has questionable caps.

> even crappiest psus will last a year, you need 50+ years for a claim like the latter one. otherwise it's just not true.
(and cheapest psus usually will last a year)

I've seen Deers go under in less than 3 months, and it's often the 3A rectifiers on +12v that fail first on P4 systems. At this point, I've recapped (and upgraded rectifiers, fans, etc.) on something like 20 Deer/L&C PSUs, and the earliest of those have now happily completed 2 years, with no failures. 10x improvement in durability, with careful component selection is not an exaggeration as far as Deers are concerned.

OTOH, with Hipros you are correct - I'll have to wait maybe 5 years with 10 Hipros to be able to substantiate my claim. At this point, I'd be very confident of my claim if I had also upgraded to fans to NMB ball-bearings - as things stand, the fans on the Hipros are likely to fail first.

> jonnyguru guy didn't check the differences between caps he put in and put out

Seems to have been Oklahoma Wolf, who posts here from time to time. I'd interpret the
results to mean that the original Fuhjyyus were ultra-low ESR aqueous, leading to the
low ripple measurements when new. When replaced with KZE and KY, the ripple increased
because of slightly higher ESR. No big deal, in my view. The long-term durability of the
KZE and KY will trump the Fuhjyyus, regardless of the initial ESR. A properly implemented
pi-filter on each rail would have squashed the ripple, in any event (many of the el-cheapo
designs omit the pi-filter and use cheap ultra-low ESR caps to meet the ripple specs.
Those ultra-low ESR caps usually happen to be aqueous, so it's a double-whammy as
far as durability is concerned).

[PCBONEZ]

Re: My new Seasonic has questionable caps.

What linuxguru said makes perfect sense to me.

You don't need 50x years to show 50x better durability, you need 50x less units to fail in the same time period.
-
MTBF has to do with sample size and math. If it's 100,000 hours they didn't do a 100,000 hours of testing. They tested maybe 100 units for 1000 hours.
-
I have 2 Antec TP2's in the 'to do' box and the longer lasting of the two lived a whole 4 months.
-
My $1-$2 was based on $1.60(something) to replace all the critical caps in a TP2.
The assumption was the crap caps originally installed were FREE.
[You'd have to read the other thread.]
-
The Fuhjyyu TM and TN series caps usually found in PSU's are ultra-low ESR.
TN is aqueous, TM is non-aqueous.
Ultra Low ESR at the start doesn't matter much when it only stays that way for a few months.

I do think the importance of using non-aqueous is blown way out of proportion.
There are a number of aqueous caps with better endurance ratings than non-aqueous.
I think the other additives in the electrolyte counter the effects of water content.
Which wins (water or additives) depends on the specific cap.
-
In other words there is no direct correlation to water content and endurance.
There may have been years ago, but not today.

.

[gdement]

Re: My new Seasonic has questionable caps.

What linuxguru said makes perfect sense to me.

You don't need 50x years to show 50x better durability, you need 50x less units to fail in the same time period.
-
MTBF has to do with sample size and math. If it's 100,000 hours they didn't do a 100,000 hours of testing. They tested maybe 100 units for 1000 hours.

For that reason though, I don't find MTBF very useful. It's really just a measure of the infant mortality rate, expressed in a way that deceptively sounds like a long term reliability spec. It's a practical way to test, but it's meaning is limited unless you can assume that the failure rate will be linear over time.

Using the obvious example of say, a GSC brand low-ESR capacitor, I'd expect they could produce reasonable MTBF ratings even though they all went to crap after 6-12 months of real use. A rubycon is (at least!) 50x more reliable, but the MTBF wouldn't show that.

[i4004]

Re: My new Seasonic has questionable caps.

You don't need 50x years to show 50x better durability, you need 50x less units to fail in the same time period.

i would say that's 50x better reliability within determined timespan.
strictly speaking, 50x better durability means device lasts 50x longer, and that's hard to prove in these circumstances.
to me, it would mean (for example) that 90% of some psu type lasted one year prior to enhancements, and now 90% of same type lasts 50 years after replacements.

that would make sense. calculating in a way you describe seems illogical to me. or, it is logical, but from a perspective of a seller or mfr, not the user.
such calculus essentialy swaps horizontal and vertical planes, so to speak.
that's why i have huge logical problem with it.

durability is how long something lasts, in my book. has nothing to do with amount of devices put to test that lasts few days or a month.

MTBF has to do with sample size and math. If it's 100,000 hours they didn't do a 100,000 hours of testing. They tested maybe 100 units for 1000 hours.

mtbf seems like a misleading advertising gimmick to me.
i first heard it in context of hdds, where you hear all those silly numbers for devices that usually last up to 5 years, if you're lucky...
one might say that mtbf doesn't describe durability anyway, but..well..what exactly does it describe, then?
what is the purpose of it?

purpose of any accelerated test; advertising gimmick.
also, everybody in hdd industry seems to use same numbers, so i dunno if they're doing tests at all, but instead just writing numbers.

I have 2 Antec TP2's in the 'to do' box and the longer lasting of the two lived a whole 4 months.

i would be interested to hear how many psus of that type die sooner than 1 year of usage, but on 1000 pieces sampling rate. or at least 100.

[starfury1]

Re: My new Seasonic has questionable caps.

The whole crux of the matter is you want a supply to last "durability"

well as soon as you change something with different characteristic's it can and does alter circuit performance.

Staying close to the original design spec part with a better quality part
isn't such a bad idea

The other alternative is you "know" how the circuit works and what the pro and cons of changing some parameters is going to result in.

This would then put you in the position to improve what you got

mostly you find the advice here is to change the caps to a low ESR non-aqueous quality caps
(not super low as in VRM type)

The non-aqueous/aqueous debate has been going on for awhile here
I think either should give a reasonable account of its self provided they are not crap caps.

I guess its based on the fact a psu is a High heat area and that aqueous caps cook better
Now if this is still the case with present electrolytic formulas I don't know

so yeah I think I'll leave that one alone

linuxguru raises 2 good points;
that a crap supply maker will seek the cheapest way of doing things which in the long run wont last.
(their hope I suppose is the warrenty period)

The fact that sub powered parts are used with respect to what they claim it can do wattage wise. (probably)
(over rated then over loaded I guess)

It was the OW recap article at JG's that started me pondering this

I think manufactures are selling us short but in their eyes they are probably think 2~3 years the box will be history anyway

I think you guys are far more into the realty of whats happening and have much more experience then me on the subject so I'll leave it to you guys

Cheers

[gonzo0815]

Re: My new Seasonic has questionable caps.

Personally, i have no explanation, why oklahomaWolf`s recapping has failed, but may bee the Chemikon KY caps are not the best type for this reason.
Furthermore, he ´has not stated, if he left them burn in, which may be required.

I am still inclined to believe, that his PSU may have far better performed with some Samxon RS, GD or Panasonic FM capacitors.

Regarding Spectre, i think i know what he means, if he is talking about responsibility.
It is the same with al mainstream media, if you depend on advertisers money, you just can`t do things they way they should be done.

This is especially true, if the people you are depend one are not interested in education and knowledge, rather than keep up consuming simple advertiser claims, which are more gullible.

Edit:
i think that PSU`s aren't that hot, compared with other device, thus may be it does not hurt to use ultra low esr capcitors in PSU`s.
I have recaped a few FSP 250w Medion OEM units with MCZ series, which are running a lot hotter then similar psu`s .
I haven`t get one back this day, and they i am felling confident they never will.

[starfury1]

Re: My new Seasonic has questionable caps.

This is especially true, if the people you are depend one are not interested in education and knowledge, rather than keep up consuming simple advertiser claims, which are more gullible.

Yeah the rock and the hard place.

like Active PFC, what psu manufactures are pushing..well the spin doctors for them is not really why they had to do it...but they got to make it seem like some real advantaged to the consumer....when in fact its not really..well very small if any.

Dont know if you could say failed on OW recap it archive I suppose what he was after but the by product was slightly worse noise or ripple (whatever it was)...which I go the impression was not expected
(but as I said above you change something it has to have some other effects happen)
possibly your right...maybe a bit of a burn was needed then re evaluate.
(as Ive said before I wonder if there is some fine tuning done with the caps selected and the circuit components...other words optimized for the lower grade)

anyway you guys work a hell of a lot more with this stuff then me and understand it much better.

Cheers

[starfury1]

Re: My new Seasonic has questionable caps.

well yeah if you think about it with the VRM you got one big heat sink that needs a fan pumping right on it and mostly the caps are right be side it...
So they have got to be getting realativly well cooked there.

OK GSC go volcanic (maybe that was the bad formula) but FM's seem to survive ok

[PCBONEZ]

Re: My new Seasonic has questionable caps.

~50x better durability ~
The word 'Durability' in the language doesn't mean the same as 'Durability' in the specs of a cap. I think everyone (including me) tends to blend the two meanings together and that doesn't quite work in a discussion.

In caps specs Durability and Lifetime are not the same thing at all.
Durability is toughness. Heavy duty vs light duty. How much abuse it can handle.
Lifetime is how long it lasts is it's operated in the correct 'duty' range.
-
Many of the factors that go into determining Durability and Lifetime are the same but that doesn't make them the same thing. Unfortunatley most people don't 'get' the difference and habit is to use the terms interchangably which confuses things.

-

MTBF and Durability (Endurance, Load Life) tests are mostly approximations and math.
They do a short test and estimate the end results if it were continued.
-Why:
Take a hard drive.
It is not pracatical to test them for 100,000 hours.
That would be 11.4 years.
By the time the test was done they would be obsolete anyway.
And an 11.4 year test would cost how much?
-
So they test say 100 or 1000 drives for x many hours and do the math.

Endurance tests for caps are similar.

NOT ACCURATE tests at all.
But the only practcal way to compare similar parts.

~~~~~
~~~~~

One of the reasons I don't hold any stock in the value of Endurance ratings on caps is that there inherent flaw in the way they are tested.

Ripple tears up the oxide layer.
DC voltage rebuilds the oxide layer.

The caps are tested at max rated ripple and max rated DC volts.
- That means the rebilding takes place at the maximum rate.

Lets say it's a 16v cap.

At 16v the rebuilding keeps up with the damage from max ripple.....
Now we put this cap on a 5v in a circuit that has max ripple.
How do we know the rebuilding at the 5v rate will keep up with the ripple damage?
- We don't.
- So the endurance rating number is useless information at any voltage less than the rated DC voltage.

.

[PCBONEZ]

Re: My new Seasonic has questionable caps.

OW recap article

I didn't reread but as I recall they changed from a crap brand to a good brand but in doing so they went from a high grade (by ESR/Ripple Ratings) to a much lower grade.
Worse ripple from the PSU would be expected.
I see this mistake all the time.
It is why you should pay attention to grade when choosing replacements.
-
Swapping GSC with Rubycon isn't going to help if the Rubycons choosen only handle 1/2 as much ripple as the GSC's did before they went bad.

.

[gdement]

Re: My new Seasonic has questionable caps.

MTBF and Durability (Endurance, Load Life) tests are mostly approximations and math.
They do a short test and estimate the end results if it were continued.
-Why:
Take a hard drive.
It is not pracatical to test them for 100,000 hours.
That would be 11.4 years.
By the time the test was done they would be obsolete anyway.
And an 11.4 year test would cost how much?
-
So they test say 100 or 1000 drives for x many hours and do the math.

Endurance tests for caps are similar.

NOT ACCURATE tests at all.
But the only practcal way to compare similar parts.

I just wish the standard expression for the results was more honest about what they actually tested. The spec should be reported as "X% failures in Y days/months" not "700,000 hours MTBF!!!" They didn't measure "mean time before failure", nor is the spec an accurate estimate for this. They measured infant mortality, so that's what it should be called.

Durability is toughness. Heavy duty vs light duty. How much abuse it can handle.
Lifetime is how long it lasts is it's operated in the correct 'duty' range.

Ripple tears up the oxide layer.
DC voltage rebuilds the oxide layer.

The caps are tested at max rated ripple and max rated DC volts.
- That means the rebilding takes place at the maximum rate.

- So the endurance rating number is useless information at any voltage less than the rated DC voltage.

Interesting info, thanks for the post.