Recapping Article is Up

[Oklahoma Wolf]

Just thought I would post a quick message saying that the recapping article is live - we're still working on getting it to the front page, but it's up and available for all to see now. Went over it a couple more times today and there are no mistakes that I can find.

Enjoy!

http://www.jonnyguru.com/article_details.php?id=109

[davidhammock200]

Just thought I would post a quick message saying that the recapping article is live - we're still working on getting it to the front page, but it's up and available for all to see now. Went over it a couple more times today and there are no mistakes that I can find.

Enjoy!

http://www.jonnyguru.com/article_details.php?id=109Recapping! I though you posted Kneecapping! :D

[Kab]

Eggscellent! :D

Haha I can see you were trying to scare everyone away from recapping ther :lol:

Seriously WTH :eek:

[CAD4466HK]

:wtf:
I was not expecting my NEO to be worse when I got it back{ripple anyhow}:p
I hope the ripple will not effect the 8+ HD's it will soon be powering:confused:

At least I won't have to worry about the "Fuhjyyu Complex" :D
To tell you the truth OW, I think your article makes me not want to do a recap anytime soon :o
Excellent job, Oh and by the way, that NEO is more then a couple of months old, more like years :p
Thank you very much OW, I feel honored to have my PSU violated by the master himself, Oklahoma Wolf :D :D :beer: :beer:

[jonnyGURU]

OKW: You need to FTP into the site and delete the pictures that are not used in the article.

I woke up this morning to another overage email. But this time it was file space and not traffic. :D I don't think there are a total of 106 photos in that article. :D

It'll get on the front page once I figure out how to categorize the articles within the article category, "DIY, troubleshooting, etc."

Until then, you all are welcome to share the link with friends. :D

[Oklahoma Wolf]

Will do... forgot about the unused pics ;)

40mV ripple on the 12v and 25mV on the 5v is not going to kill hard drives, CAD :p

[jonnyGURU]

CAD: In case it wasn't already made clear, the V/Div on those Oscope screenshots is .01V and not the .05 I usually use in the PSU reviews.

[GalvanizedYankee]

Thanks for the effort OKW & jonny. I was suprised at the results.
I PMed the article to willawake and he had the same thing happen when he recapped as PSU with Rubycon MCZs.

[Oklahoma Wolf]

I was surprised too, but figured there had to be some reason they went with the likes of Fuhjyyu or OST. The engineers at both OEM's do know what they're doing after all... it's just the heat they're not always taking into account.

[GalvanizedYankee]

Yeah..With Fuhjyyu caps it's not a matter of IF but WHEN are they going to fail. Anything at or above 45C and they are headed for their knees.

[Oklahoma Wolf]

Well, it's a matter of when for any capacitor... my ancient 235W CWT is still plugging along with its all Fuhjyyu cap compliment going on 9 years now. It's all about the cooling ;)

[CAD4466HK]

CAD: In case it wasn't already made clear, the V/Div on those Oscope screenshots is .01V and not the .05 I usually use in the PSU reviews.

Ahhh so :D
Nevermind my paranoia OW :crazy: :D

[Kab]

CAD - make note of this PSU. I wanna know how long the CWT->Fuhjjyu based units last after recapping and if there can be "unforseen incompatibility" issues, as in making it worse for the end user overall. Maybe some EC's do better, as in reactions per second than others? I dunno, not enough time to get into the nitty gritty.

[GalvanizedYankee]

Kab, my two cents. :D

The increase and or change in ripple is negligble. It means little if anything.

The Fuhjyyu caps are probably rated for 2k hours at 85C but in real life suffer badly at temps over 45C. Nearly everyday I post in a thread where the issue is the Antec SP or TP PSU over at AnandTech.

Rubycon (very good Nipponese cap) states that for every 10C drop in operating temp cap life doubles. So a Ruby rated at 2k hours at 85C will last 32k hours at 45C.

The Nipponese caps OKW installed are rated at 105C, so you might have a very long wait. I am not a cap expert but the needed reaction might be better suited to the 85C cap vs the one rated for 105C.

[jonnyGURU]

The increase and or change in ripple is negligble. It means little if anything.

Exactly. The small increase is likely just from ambient RF interference.

[Aristide1]

Mr Guru,

I think the author of the article was on the right track, but his testing may have been a tad premature.

Capacitors should be given time to settle down. His should burn them in for at least 100 hours and then test his updated units again for noise.

Your power supply reviews are very helpful.

Thanks
Aris

[Oklahoma Wolf]

Capacitors should be given time to settle down. His should burn them in for at least 100 hours and then test his updated units again for noise.

Unfortunately, that wasn't possible due to logistics factors.

Thanks for the comments :)

[cypherpunks]

Capacitors to filter switching power supply noise often need low ESR,
and definitely need high ripple current ratings. (Which amounts to the same thing, because high ESR when the ripple current is high gives I-squared-R losses which overheat the cap.)

Neither the KZE or KY series are particularly low impedance. See http://www.chemi-con.com/u7002/applications_ms.php for advice on what you should have used.

I've replaced bad caps on motherboards with great success, although never in a PSU.

[Oklahoma Wolf]

Neither the KZE or KY series are particularly low impedance. See http://www.chemi-con.com/u7002/applications_ms.php for advice on what you should have used.

I used what I could get. I wanted to go with Panasonic FM, but that didn't work out.

[jonnyGURU]

Something you have to realize about power supply capping and motherboard capping is not all things are equal. Caps that fail on motherboards fail for different reasons than caps that fail in power supplies and the caps used in power supplies are completely different values, sizes, etc. than caps used on motherboards.

When I bought the caps for Wolf's recaps, Matsushitas simply weren't available in the sizes needed for the secondary. And you'll note that some of the caps had to be replaced with different value caps when caps of the particular size required to fit into a tight space were not available. I can't find the thread over at Bad Caps right now, but the SamXon rep once pointed out that some caps in certain sizes may only be manufactured by one particular company and none other.

That said, I will say that it is an EXCELLENT point about not all caps enjoying the same impedance. Unfortunately, the Chemi-Con KZE and KY were what we could find even after hunting through several electronics shops in Taipei and a few PSU factories in Shezhen. It actually took about a week to round up all of the caps (not including shipping time, etc.)

[Aristide1]

Not sure how switching PS handle the rectifier diodes, but generally bridge rectifier diodes make a fair amount of noise themselves, unless one uses fast recovery or Shottky diodes. These normally are found in high end audio circuitry when such noise is not desirable. Its also better to prevent it than to try to correct it later on.

Compared to generic bridge rectifiers these may seem costly, but it's still not that bad for playing around in a circuit like this.

Have a look around here
http://www.partsconnexion.com/catalog/semiconductors.html

Try this link for various high quality electrolytic caps.
http://www.partsconnexion.com/catalog/CapacitorsElectrolytic.html

There are tabs at the bottom of the page for different brands. Since you would use these strictly in a power supply I would tell you to look at the cheapest ones they have. There are no bad caps on this link anywhere.

[Oklahoma Wolf]

Not sure how switching PS handle the rectifier diodes

Usually Shottky.

[cypherpunks]

Not sure how switching PS handle the rectifier diodes, but generally bridge rectifier diodes make a fair amount of noise themselves, unless one uses fast recovery or Shottky diodes.

Which is exactly why Schottky diodes are used universally in power supplies.

For those who care, a Shottky diode (http://en.wikipedia.org/wiki/Schottky_diode) is basically "half a diode". Rather than a P-N semiconductor junction, it is metal to N semiconductor. This gives it half the forward voltage drop (saving power), and speeds up reverse recovery due to the lack of minority carriers (http://en.wikipedia.org/wiki/Minority_carrier) (holes).

The thing is, all diodes require a bit of reverse current to turn off. When in forward conduction, the junction region is full of electrons to carry the current. If you just set the voltage to zero, the electrons eventually drift out of the region, but it's not a fast operation. To turn it off quickly, you have to reverse the current and suck the electrons out of the junction regions.

And then, very abruptly, the reverse current goes to zero and you have a bit of a noise spike. This is actually used on some test equipment; there are specially constructed "step recovery diodes (http://en.wikipedia.org/wiki/Step_recovery_diode)" that switch off with a bang like a ruler hitting a desk. It turns out to be useful for generating microwaves.

Schottky diodes have a somewhat "soft" turn-off, which makes the bang less loud. And they turn off very quickly, before the reverse voltage can build up.

The downside is that they have high reverse leakage. In a power supply, a milli-amp of leakage (it's actually a few dozen micro-amps for most power Schottkies) is nothing. In a sensitive analog circuit, it would be a problem.

The one limitation on Schottky diodes in power applications is that reverse leakage increases with temperature. At high voltage, a bit of leakage heats the diode, which makes it leak more, which heats it more, and you get thermal runaway. Your diode decides it's tired of being a rectifier and is going to try playing the role of a fuse for a few milliseconds.

The alternative, which is apparently used in the OCZ ProXStream 1000W reviewed at Hardware Secrets (http://www.hardwaresecrets.com/article/491/5) is synchronous rectification (http://en.wikipedia.org/wiki/Synchronous_rectification). That's where you don't use any diodes, but turn a transistor on at just the right time. The timing is critical, but you can get higher efficiency that way, especially at low voltages where even a Schottky diode's 0.3V forward voltage drop is a significant fraction of the power consumed.

[jooch]

Mr Guru,

I think the author of the article was on the right track, but his testing may have been a tad premature.

Capacitors should be given time to settle down. His should burn them in for at least 100 hours and then test his updated units again for noise.

Your power supply reviews are very helpful.

Thanks
Aris

Aris, you're right on the money. Have a look at this discussion on DIYaudio.com (http://www.diyaudio.com/forums/showthread.php?threadid=81355), it confirms you're theory.
Here are the pictures for non DIYaudio users: