Suggestions for replacement capacitors

[velinion]

TL;DR: What caps would be recommended to replace a bunch 2200uF 16v electrolytic caps with long-term reliability as a primary goal? Details below.

Hi all,
I'm a traditionally software guy that has recently been getting into hardware and tinkering. While looking into an issue on a NAS that had been up and running for months (and suddenly wouldn't turn on), it turned out that the 5v rail was overloading the power supply due to 16 2200uF capacitors on the hot-swappable drive backplane (manufacturer suggests minimum 15 amp rail.)

Now, what interested me is that this HAD worked for about ~6 months... could the caps have started drawing more power after 6 months of 24/7 use?
They are ChangX caps, which I understand do not have the most... savory... reputation.

 Even though I'm swapping out the power supply for one with a 25amp 5v rail, I am still considering swapping out the caps (Since drive reliability is pretty much the core idea of a NAS)

Figured that 2200uF and 16v would be an easy set of specs to find, and it is, but... there seems to be a lot more to caps than I thought. ESD, leakage current, and endurance specs that I do not know how to read (What does ?5 & 6.3 : 4,000hours   mean?) Would there be an advantage to switching to a polymer cap? Am I making this harder than it has to be?

And here I always thought caps were "so simple"....
Realize I've asked a lot of questions in this post... if anyone wants to just answer "I'd recommend X, and here is why" that's all I really need. I do love to learn though, so... hence the large batch of Q's.

PS. Configuation of current caps is : 8 HDD bays 2 caps each, one on the 5v rail, one on the 12v rail. All caps are 2200uF 16v ChangX electrolytic, picture attached.

[John B]

Ill be interested to hear experiences with ChengX caps. Ive got a few bags of them, and have used them successfully, albeit in not very temperature stressful environments. Though reading your post, I'm not sure how you determined the caps are bad.

[velinion]

Well, they are not exactly bad. A high enough amp input will still power the whole thing, but there aren't really any other components involved that would explain an increase in power draw. Could leave them in with the new power supply, but just thinking that while I'm in here, I might as well replace them with something from Nipon Chemi-con, or a similar respected brand. Plus, this case is known to run hot, with drives idling 40C and hitting 46C, so not the coolest environment for what I've heard are off-brand caps, especially powered 24/7. I'm also planning on improving airflow while the system is apart.

Still, that they are not worth replacing is something I'm also quite prepared to hear.

[Ian.M]

If you are certain the excessive current draw has been traced to the caps, then they *MUST* be replaced.  If they are the culprit, they will be running warm - soak test the backplane, powered but with no drives installed, preferably out of the case on a current limited bench supply.   Note the initial current draw, and see if its gone up a few hours later Whatever is responsible will be running warm or hot so if you've got a thermal imaging camera, use it, otherwise run your fingertips cautiously over them. 

However electrolytic caps rarely fail shorted unless they are grossly abused so, before repopulating, check with a continuity tester that the PCB markings for cap polarity are correct - its possible the caps on the 5V rail been stuffed backwards!

You'll want 105 deg C rated caps for reliability and the 12V rail should be using 25V rated caps as running at 75% of max rated voltage is not good for long term reliability.  They don't have to be 2200uF - if you can get more capacilance in the same form factor you can use it so I'd also look at 3300uF ones.

[velinion]

Would backwards caps work for a few months before tripping power supply protection? I'll check it out though. There are certainly signs of "lowest bidder" assembly on this board, so it wouldn't really surprise me. Shame, because the rest of the case is well made. I guess they decided that power distribution was the best place to make up a few pennies for their budget,

[Ian.M]

In normal use, the applied voltage maintains the oxide layer on the anode foil, but reverse voltage attacks it and attempts to form a replacememt oxide layer on the cathode foil.  Its purely a matter of luck and the available supply current whether this actually happens or the cap shorts as both foils have insufficient oxide for the rail voltage, or the cap bursts or vents due to hydrogen gas buildup, or steam pressure from gross overheating.  The Wikipedia article will give you a basic understanding of this failure mechanism.
https://en.wikipedia.org/wiki/Aluminum_electrolytic_capacitor#Reverse_voltage

If the voltage rating is much higher than the applied voltage, the chances of a reversed Aluminium electrolytic fault not being noticed during soak test increase significantly.  Repair shops do occasionally see this sort of fault, often with incorrect board markings for polarity, and you also sometimes see incorrect board markings with the cap correctly stuffed as if the design fault was picked up in early production they'll often simply reprogram the pick & stuff machine to reverse the offending cap(s) rather than respinning the board for a silk screen error.   TLDR: when replacing bad caps, test polarity, don't assume the board is correctly marked!

[Nusa]

If the original power supply won't start connected to an unloaded backplane, I'd be more suspicious that the immediate fault is in the power supply. Are you able to load-test it at all?

Alternately, how much current does your replacement power supply actually deliver for the unloaded backplane? If it's as high as you claim, there has to be significant heat being generated somewhere! Enough to locate.

Don't let that stop you from replacing the caps anyway.

[velinion]

The original power supply indeed will not start connected only to mainboard and the unloaded backplane... but WILL start with 6 HDDs and an SSD connected without the backplane. So, the backplane is indeed the issue (but a 25AMP 5V power supply will start with the backplane connected. Manufacturer (Of both the case and power supply) said this is a "temporary draw to charge the capacitors" but I have not gone looking for hot spots. I'll try that in the morning.

[DimitriP]

What caps would be recommended to replace a bunch 2200uF 16v electrolytic caps with long-term reliability as a primary goal? Details below.

Panasonic FR series. Low ESR, 105C, ~ 10000 Hours.    Check the dimensions to make sure they will fit, especially if they are right next to each other.

 http://www.mouser.com/ProductDetail/Panasonic/EEU-FR1V222L/?qs=sGAEpiMZZMukHu%252bjC5l7YfvTmG9VJYZ8XKlmn3OKv5Q%3d

[james_s]

I usually buy Nichicon, Rubycon or Panasonic capacitors, the exact series depends on the application but most/all of those are made in Japan, quality parts with trustworthy specifications.

[wraper]

I don't think there is an issue with those capacitors. Yes they are cheap, but not stressed, so does not matter that much.

[Ian.M]

If you can soak test them on a bench supply and they draw negligible current, they are almost certainly still good.   If there are no power on indicator LEDs, fully charge thenm on the bench PSU, disconnect power and time their discharge into a 10K resistor to 1/3 the initial voltage - it should be a little under 3 minutes for eight in parallel if they are still 2200uF each.  If they aren't leaky and aren't low in value, they are almost certainly still good.  They probably aren't even really needed - after all the SATA drives are designed with enough local decoupling on board to work properly plugged into a SATA power cable without any extra caps.

I must agree with Nusa that its probably the PSU gong bad.