Samxon question

[jonnyGURU]

Re: Samxon question

corsair 550vx is $95 or $80 with rebate. chemicon/hitachi

availability concerns is a valid point though.

With rebate, which takes into consideration about a 50% redemption rate.

Also, Corsair is used to selling memory at next to nothing margins. Making 15 points on a power supply is like a gold mine to those guys.

[gdement]

Re: Samxon question

I can't start with Rubycon, have Rubycon run out of a particular cap and then re-certify and re-test with a Teapo half way through the month.

I'm no expert, but I've had the impression it's standard practice to multi-source commodity parts like this, isn't it? You'd always be taking a chance if you depend on only 1 supplier. Is it particularly more expensive to prequalify say 3 different caps? For example, Rubycon ZL and UCC KZE have identical specs, it's obvious they're intended to be substitutes for each other.

Teapos are cheaper and more available than the Jap caps, but are expensive as far as Chinese caps go and are still not available in every size, type, etc.

That pricing sounds sensible, given they're more reputable than the typical Chinese brand. Teapo seems to do fine in power supplies, but I wouldn't trust anything below Teapo though. I doubt the really cheap manufacturers even know what they're doing - and that's why they ended up buying bad electrolyte in that famous incident years ago. Surely they also use the cheapest aluminum they can find. If they ever make reliable caps, it's probably by accident.

I've noticed that Supermicro uses YEC caps at low stress areas on some P4 boards, so maybe that brand also has some acceptable applications. But I saw YEC fail in about 3 of my 4 InWin/FSP 250W power supplies. I also saw them fail on a Quanta P3 motherboard, but Quanta was using them everywhere whereas Supermicro only uses them at what appear to be low stress areas.

[jonnyGURU]

Re: Samxon question

I'm no expert, but I've had the impression it's standard practice to multi-source commodity parts like this, isn't it? You'd always be taking a chance if you depend on only 1 supplier. Is it particularly more expensive to prequalify say 3 different caps? For example, Rubycon ZL and UCC KZE have identical specs, it's obvious they're intended to be substitutes for each other.

Of course. For example, I'm looking at Samxon and will likely use Jun Fu as a back up. As it is Samxon seems to be able to do better volume than a lot of suppliers, so there may never be a need to use the Jun Fu's.

That pricing sounds sensible, given they're more reputable than the typical Chinese brand. Teapo seems to do fine in power supplies, but I wouldn't trust anything below Teapo though. I doubt the really cheap manufacturers even know what they're doing - and that's why they ended up buying bad electrolyte in that famous incident years ago. Surely they also use the cheapest aluminum they can find. If they ever make reliable caps, it's probably by accident.

I've noticed that Supermicro uses YEC caps at low stress areas on some P4 boards, so maybe that brand also has some acceptable applications. But I saw YEC fail in about 3 of my 4 InWin/FSP 250W power supplies. I also saw them fail on a Quanta P3 motherboard, but Quanta was using them everywhere whereas Supermicro only uses them at what appear to be low stress areas.

Yeah. The behaviour of caps on motherboards is entirely different than in PSU's. Linuxguru explained it rather eloquantly somewhere here in the forums, but I can't find it right now.

2000hrs isn't much. if you don't need ultra low ESR, there are e.g. Panasonic FM
that are rated for up to 7000hrs depending on size and Rubycon YXM with even 10.000hrs.

Actually, most caps used in PSU's are either 2000 hours @ 85°C or 1000 hours @ 105°C. It's good to use 2000 hours at 105°C because you're going to get 10 times the life at < 85°C and rarely do capacitors in power supplies get anywhere near 85°C even under full load at 40°C operating temperatures.

[kikkoman]

Re: Samxon question

Actually, most caps used in PSU's are either 2000 hours @ 85°C or 1000 hours @ 105°C. It's good to use 2000 hours at 105°C because you're going to get 10 times the life at < 85°C and rarely do capacitors in power supplies get anywhere near 85°C even under full load at 40°C operating temperatures.

most caps used in PSUs are crap caps, and that's why most PSUs fail.
not every design runs that cool, but let's just assume one of those caps does 8000hrs according to the mfr's (probably a bit optimistic) calculations at those lower temps.

that's less than a year if you run the unit 24/7.
you see... it is NOT GOOD to use these caps in PSUs. at least not in my world.

[i4004]

Re: Samxon question

kikkoman, caps fail gradually.
and psus are less picky than mobos.

in the end this means psu with mid-grade caps can last a long time.
offcourse, psu with top-grade can last longer, but you'll probably buy new system sooner than mid-grade fails.

heck, my enermax eg365 lasted over 22k hours, and it's still good. pce-tur galore inside.

[starfury1]

Re: Samxon question

Gonzo

I think they would be more reliable then lower voltage and especially low ESR caps (if there is no over voltage present).
But certainly, i am not an expert too

Excuse the slight "brain fart"(as pc bonze so well puts it)
I should have clarified it better.

I was thinking in regards to older TV sets were they are on the more likely to fail list.
160V 10uf up etc
they are Probably GP junk anyway.
it was a quick thought when I seen the voltage

As you say, its probably a case of horses for courses here.
I am no expert on it either

Cheers

[starfury1]

Re: Samxon question

my enermax eg365 lasted over 22k hours, and it's still good. pce-tur galore inside.

Mine too, 5 years plus
Gonzo seems to think the ability for them to keep caps cool has probably helped.
although he repaired 2 he mentioned in a thread that were blown
(cant recall the details now)

This may have been from abuse of PSU's
and thats something that has to be considered I suppose
its not the manufactures fault if someone overly abuses a supply
it will happen, so probably better to make sure it dies and nothing else

I'd still rather see good caps used thought

anyway hope something works to your favor JG
I know your trying to do the right thing here

Cheers

[kikkoman]

Re: Samxon question

kikkoman, caps fail gradually.
and psus are less picky than mobos.

in the end this means psu with mid-grade caps can last a long time.
offcourse, psu with top-grade can last longer, but you'll probably buy new system sooner than mid-grade fails.

heck, my enermax eg365 lasted over 22k hours, and it's still good. pce-tur galore inside.

sorry, i didn't quite make that clear. failing caps cause higher ripple which in turn kills other components over time. this is no better.
sure, i've had good experiences with cheap hardware too, but in most cases one or more components act up and have to be replaced or die prematurely - if you're lucky, it's only the PSU.
DUE TO BAD CAPS.

i agree that a $100 PSU is no high-end unit. but it ain't budget either.

[starfury1]

Re: Samxon question

yeah I got what you mean Kikkoman,
its sort of a positive feedback loop in a way
As they start to go they get hotter from the work they are doing
which drys them out more and so on in a loop
Which most probably follows an exponential curve...the more they die the quicker they die.

Ok my not be totally accurate and putting it simply

Yeah the extra ripple cant be good for other components and stability
so the longer you can keep that low (Ripple) the better

All caps do have a finite working life the longer that is and the longer you can maintain it by not overly stressing them the better I think

yeah the use of caps in a psu may not be as brutally treated as a VRM circuit but its still a stressful environment and a cap has to be up to the task over the "long term",
even crap caps could keep a supply humming along for a 1~3 years but what is there true state at about 3 years
how far in or out of their usable life are they?

The working environment has to be considered in cap selection

Here we are talking the technical aspects
but from a manufacturing point of view (which is to build and sell psu's)
if Joe and Jane Blogg get 3 years out of it they would probably be happy to fork out for another psu
(or computer)
That keeps the manufacture happy and JG in a job.

Also the pace of computing, what the super duper rig today in 3 years would be bottom rung if its lucky so its high turn over in this respect.

So the "whole" question does have many aspects to it

at lest JonnyG is taking the time to try and get a psu out the door with at least a half decent cap in it, so I'll thank him for that and wish a few more manufactures would take a leaf out of his book

I think I do remember that post JG and yeah I don't know what thread it was in either but I'll have a scratch around.

Cheers

[jonnyGURU]

Re: Samxon question

I'm going to disagree with kikkoman on this one.

Your statements are full of generalizations. Most PSU's are full of crap caps? That's why PSU's fail? Wrong.

I used to do RMA for TCWO (3 years) and ASI (two years) and most PSU's did not come back with bad caps.

Actually, most PSU's die due to fan failure (and then typically whatever fails due to the fan failure, which typically isn't cap related either). Of course, the fan is the part of a PSU with the shortest MTBF and it's MTBF is never calculated into the PSU's total MTBF (go figure). Most PSU's come back with burnt rectifiers with nice big burn marks on the insides of the housings. This happens for different reasons that are hard to isolate once the units already blown up: bad input power? Overloaded? Overheated? Who really knows once it's done.

PSU's that come back with bad caps tend to consistantly come back with bad caps. For example: Antec RMA? Probably bad caps. But an Enermax RMA? Probably popped a switching rectifier on the primary side, but hardly ever bad caps.

Failing caps can cause excess ripple, yes. But even a relatively cheap PSU (not rock bottom cheap) after five years of field use doesn't show enough cap degrade to put ripple out of spec. A $100 unit with Jun Fu, Jee or whatever cheap cap isn't typically showing out of spec ripple and isn't showing any signs of failing caps. That said, I've had customers buy motherboards from me and really, really, really cheap (like $10) power supplies elsewhere and come back with failed motherboards one year later, but it's not even because the caps in the PSU failed causing the ripple to be out of spec... the PSU's didn't have the right caps in them in the first place and the ripple was ALWAYS out of spec.

So maybe based on your experience, all PSU's you've used that you consider "cheap" have had cheap caps and those caps have failed or you have had other components fail because you've had caps in PSU's begin to fail causing excess ripple... but that's certainly the exception.... not the rule. At least for any decent $50+ PSU and not some hunk of crap bundled with a case (or an inadequately cooled Antec).

Now that that's out of the way... It turns out that my lack of ability to speak Chinese has gotten in the way of progress. The factory foreman visited Samson... not Samxon. I just got the spec sheets in my email this AM and they're all Samson branded. The QC guys scoffed at the idea of using Samson caps in the units... not Samxon. So I can see their concern when asking to replace Jun Fu's with Samson. Why? So, Big Pope, I forwarded your info to the factory. Hopefully they'll be able to figure out the difference between Samxon and Samson from here on in.

Cheers!

[starfury1]

Re: Samxon question

Oh close name but not the right stuff...humm yeah now that could be a problem ....

the PSU's didn't have the right caps in them in the first place and the ripple was ALWAYS out of spec.

That is part of what I have stated with Teapo in a thread somewhere else no Idea where now.

I think its silverstone PSU high end job one of JG reviews that has these fitted and that was my statement in the thread (here somewhere)
(not exactly word for word but as follows)

maybe its not the teapo's so much as the choice of series used in the silverstone, (or whatever) where probably the "RIGHT" series has been selected for the job, they will last.

there was a question raised in a thread here somewhere on when was the last time you seen a bad Teapo
(I suppose neglecting the Bad cap era stuff)

I dont know
Cheapie supplies, may just use a the less speced teapo (cheap) that does the job and will live a year...after that they don't really care
(and if you throw away the EMI filter, do you really care about the caps?)

unfortunately I have to be a back seat driver as I don't have the vast experience others have with first hand knowledge of the situation, that a lot of forum members have.

Its a good point on the fan as these things if I remember and understand it correctly NEED air flow it is part of the design as a total.
So if that stop's, it going to run at way too hot temps and sooner or later failure is going to happen.
Unless there is some thermal shut down mechanism and that I don't have a clue on.

with the real cheapies
(and hopefully none of these are still in production)
I suppose even if the design does have it, it gets turfed along with the EMI filter and anything else not relevant to having a working psu.

like I said I have to leave the real world experience to you guys that do it on a day to day basis.

Had a look for that post but couldn't find it JG

Cheers

[gonzo0815]

Re: Samxon question

Well ifyou want to see bad Teapo`s, well i can show you some
I think Jonny is right, it is all about getting the right caps for a given PSU and it should last.

The FSP250MDN PSU where this pictures came from, are equipped with a very slow 120mm fan. It was very very silent, but obviously for such an design it needs something different then Teapos. The Jamicon in this PSU are fine, but i haven`t ESR checked them.
But certainly some Samxon RS or Panasonic FC, FM would be the best suited ones here.

Any way, if i want a new psu today in the range of 100$, i expect silent duty up to a load about 200w. A lifetime of at least 5 years and a efficiency in the 80+range.

If someone can do this with some of the crap capacitors, ok no problem.
But if there is an competitor doing the same and putting in some UCC caps, you bet what i would go out to buy.

[acstech]

Re: Samxon question

Someone brought up Teapo...

I can't tell you the number of Dell PSU's that I've recapped. Only one had visually faling Teapo caps. The rest of them just didn't work right. Either randomly shutting down the computer, or otherwise causing problems. Yet recapping always fixes it. Huh. A Teapo failing without showing it. Imagine that....

[starfury1]

Re: Samxon question

twas me that raised Teapo...you do tend to find them used in a lot of PSU's, also the more up market jobs can sport them
(if this is a good thing or a bad thing I don't really know..I guess time will tell)

Yeah and mostly the cheaper ones that they fail in so are we really dealing with a bad choice of teapo for the psu use?

Astech
that is interesting as it still points to possibly failing caps if a recap sorted it...not conclusive I know
did you do any checking of the teapos removed?

Gonzo

I think Jonny is right, it is all about getting the right caps for a given PSU and it should last.

Agree with that Gonzo...
right horse for the course

So Samxon RS is the recommended for PSU use ?


cheers

[i4004]

Re: Samxon question

That said, I've had customers buy motherboards from me and really, really, really cheap (like $10) power supplies elsewhere and come back with failed motherboards one year later, but it's not even because the caps in the PSU failed causing the ripple to be out of spec... the PSU's didn't have the right caps in them in the first place and the ripple was ALWAYS out of spec.

checking output ripple is not a way to establish if a cap in psu caused fried mobo, because output caps (those that produce this ripple) can't cause such disastrous failures.
places to inspect are +5vstby supply and caps around the pwm-ic, and there you'll probably find failed caps on these "mobo-killer" psus.

so the "oh, too high ripple before, too high ripple now=not badcaps caused" is not a way to reach conclusions at all.

----
i also had the idea that chinese didn't uderstand what you said too well, but i've found that idea too funny to be true....hehe...
i thought you used email to communicate, not the phone.

[jonnyGURU]

Re: Samxon question

checking output ripple is not a way to establish if a cap in psu caused fried mobo, because output caps (those that produce this ripple) can't cause such disastrous failures.
places to inspect are +5vstby supply and caps around the pwm-ic, and there you'll probably find failed caps on these "mobo-killer" psus.

so the "oh, too high ripple before, too high ripple now=not badcaps caused" is not a way to reach conclusions at all.

True. It may very well be that "bad caps" didn't even kill these boards. But you don't know the whole story either. There is far more anecdotal evidence beyond "the PSU has excessive ripple output". I'll give you more background....

At TCWO we had barebones systems that had components in them that we already qualified. Motherboard, PSU, etc. Sold these units for years. Never had a problem. RMA rate on either motherboard or PSU was under 1%.. EASILY.

In comes local PC builder guy. Buys just the motherboard, case w/o PSU, opticals, etc. Ten of each to be precise. He goes down the street to Kung Fu computers to buy the PSU because "it's a better deal". One year later, the man comes back with ALL TEN motherboards. Six have swollen caps, two have burnt VR's, two more have swolled caps AND burnt VR's. The gentleman was under the impression that the boards were just crap. But how do you explain to someone that has a 100% RMA rate that your typical RMA rate on the same board is usually less than 1%?

I hooked up his cheap-o PSU's to the load tester since it was the only part in the build not bought from us, applied a 200W load and flipped on the scope. I measured no less than 200mV on the +5V and the +12V.

I told him I would only replace his motherboards "under warranty" if he would buy replacement PSU's too.

So sure, the problem could have been too high voltage, too low voltage, what have you. But any PSU that comes in that just wreaks of cheap and the guy wants to warranty something of mine that the cheap hunk of crap is powering, it's time to AT LEAST load test it and scope it and in RMA you often play the odds to save time. I couldn't open the PSU's to inspect them internally because they were not mine to open. I would've hoped that the customer return all of them for a refund. But I doubt that there was any cap degration or anything else unusual inside the PSU outside of just over all crappy design and build quality since every unit was under 1 year old and every motherboard hooked up to every unit experienced a similar problem.

As for the Teapos: It's PC Power & Cooling Turbo-Cool units that uses them most often. But so do a number of other higher end units like the Enhance built Cooler Master Real Power Pro, Antec True Power Quattro, Silver Power SP-1000E, etc. as well as some Andysons (the Ultra X3's are all Teapo), some SilverStone (the Impervio and Etasis built units) and Etasis.

Another interesting use of Teapo caps in PSU's: Typically low end Andyson units use Fuhjyyu. You'll find them in Hiper and Sunbeam units. You don't typically hear about cap failures despite the use of Fuhjyyu caps because the PCB layout allows for good airflow and they use a decent fan (high CFM while still low noise). Aerocool wanted to use the same platform for a "low-noise" unit. The fan's CFM's would essentially be cut 50% to 75% at the lower RPM's. All of the Fuhjyyu's are switched out for Teapos for "longer life at higher temps."

[i4004]

Re: Samxon question

yes, you have a point there.
i'm more interested in cases where psu goes with a bang and destroys everything in the case, while your example show effects of slow frying the (otherwise) good mobo.

bulging caps and destroyed vrms, but other components mostly fine.

and yes, buying the cheapest of psus..always a bad idea.

it can happen they work for awfully long time (just inspected one jnc (famous mobo killer down our way) that's 6-7years old, has bulging caps on outputs....but powering intel mobo with chemicons..and system still quite fine) but i would never place that bet.
and if that intel mobo can survive 7 years with such crappy power...uh...how long can it last with clean power....

[linuxguru]

Re: Samxon question

It's possible for a reasonably rugged board to survive a marginally bad PSU with high ripple on some rails - as long as the really bad rails are +/- 12v. +12v is used only by the VRM, legacy RS-232c and in the PCI slots; -12v is used only by legacy RS-232c. A well-designed VRM can handle large variations on the input. RS-232c is spec'ed for +/- 15v, so the ripple doesn't matter. Most PCI cards don't use +/- 12v at all. Some boards have onboard linear regulators to generate Vdimm, and in some cases +3.3v, from +5v.

So it's possible for a board to survive bad +/- 12v rails, and in some cases even bad +5v/+3.3v rails if the board is designed very conservatively, which was the case with most P3/early-P4 era Intel mobos (before the bean counters took over and started using dodgy HM, HN and KZGs).

OTOH, a HDD usually only has a linear regulator(s), if at all. High ripple voltage and/or spikes on +12v/+5v will almost certainly reduce HDD MTBF by 90-99% - most premature failures on Deskstars as well as the slim Maxtors were probably caused by bad PSUs and/or poor airflow. If you really want to see the effect of good PSUs vs. bad PSUs in a desktop setting, just count the HDD failure rate. With a good PSU, just about any HDD should last the entire warranty period and beyond: 5 years of continuous operation is the norm; 8-10 years certainly achievable by over 50% of all drives. A cluster of HDD failures within 6-24 months points unerringly to marginal PSUs, even if there's no other visible problem with the systems.

Edit: All 4 of my Deskstar 60GXPs are in their 7th year of operation; and I haven't yet experienced a slim Maxtor Diamondmax failure out of ~6 drives. However, all are sitting in systems with good or pre-emptively recapped PSUs.

[i4004]

Re: Samxon question

i think diamondmax was just a bad series.
got one crap on me after bit more than a year, and than another one with same lifespan. enermax psu.
a lot of other people has simillar problems with them.
no such problems with barracuda 7200.7

it's a pitty google report on hdd reliability didn't speak of this, but i bet maxtor would be the crappiest overall manufacturer.

bean counters?
yeah, some of the hdd bios is actually on the platter. how the hell can you get cheaper than that?
hehe....

[jonnyGURU]

Re: Samxon question

It's possible for a reasonably rugged board to survive a marginally bad PSU with high ripple on some rails - as long as the really bad rails are +/- 12v. +12v is used only by the VRM, legacy RS-232c and in the PCI slots; -12v is used only by legacy RS-232c. A well-designed VRM can handle large variations on the input. RS-232c is spec'ed for +/- 15v, so the ripple doesn't matter. Most PCI cards don't use +/- 12v at all. Some boards have onboard linear regulators to generate Vdimm, and in some cases +3.3v, from +5v.

So it's possible for a board to survive bad +/- 12v rails, and in some cases even bad +5v/+3.3v rails if the board is designed very conservatively, which was the case with most P3/early-P4 era Intel mobos (before the bean counters took over and started using dodgy HM, HN and KZGs).

OTOH, a HDD usually only has a linear regulator(s), if at all. High ripple voltage and/or spikes on +12v/+5v will almost certainly reduce HDD MTBF by 90-99% - most premature failures on Deskstars as well as the slim Maxtors were probably caused by bad PSUs and/or poor airflow. If you really want to see the effect of good PSUs vs. bad PSUs in a desktop setting, just count the HDD failure rate. With a good PSU, just about any HDD should last the entire warranty period and beyond: 5 years of continuous operation is the norm; 8-10 years certainly achievable by over 50% of all drives. A cluster of HDD failures within 6-24 months points unerringly to marginal PSUs, even if there's no other visible problem with the systems.

Edit: All 4 of my Deskstar 60GXPs are in their 7th year of operation; and I haven't yet experienced a slim Maxtor Diamondmax failure out of ~6 drives. However, all are sitting in systems with good or pre-emptively recapped PSUs.

Good thing the guy didn't buy hard drives from us!

For the record: I never had the problems with Deskstars that everyone reports either. Had four with no problems what so ever. About seven years too, I'd have to say. Then again, cheapest PSU I've ever used was an Enermax or Antec.

I hear some people complaining about Seagate SATA drives and high failure rates. Again, no problems here. And I deployed about 300 of them as voice mail servers with RAID1. Only one failed within a year and 2 after a year's time. The PSU in all of the servers? An FSP based (NOT Epsilon. FGN or something like that) Aopen PSU.