Can anyone explain why I am frying this cap?

[rgrayton]

http://www.panasonic.com/industrial/...series_dne.pdf
(pana TSUP 50v 8200uf, ecos1hp822da)

Hello,
Usage is output of rectifier 25v xform, 38 vdc no load. This is a 50v cap, 63v surge. Polarity has been checked dozens of times and recitifier is working fine.
I am in the process of decreasing the ripple of a Brian lm4780 kit rectifier board, but am running out of explainations for this. Ripple for this cap is 4.4A, but this is during charge, on turn on and with no load, so I would think no ripple, but obviously I am missing something. I do not have really any experience with PS design.
Thanks,
-BG

[macboy]

There are only 4 basic ways to fry a cap, and one of them is undoubtedly causing your failures:

1. Excess voltage. You seem fine in this respect. A 50 VDC cap on a 38 V rail is just fine.

2. External heat sources causing excessive temperatures. This isn't it.

3. Reversed polarity. The dielectric breaks down at a relatively low reverse voltage, causing high current to flow through the cap, heating it up, and causing it to fail. Internal arcing can also be created, permanently damaging the cap even if it doesn't catastrophically fail. Make sure that for your negative rail, the positive lead of the cap is connected to ground and the negative lead is connected to the supply, since its voltage is lower than ground. This is a common mistake.

4. Excess ripple current. This is caused by an excessive AC voltage on the cap. The more current the load draws, the more ripple current will be present on the power supply storage caps. Also, if one of your recitifier diodes is short-circuited, it can pass the full AC voltage of the transformer to the cap, causing excessive current. Ripple current causes failure via power dissipation in the capacitor's ESR (equiv. series resistance; P=I^2*R). The heat causes the cap to fail. Since you have no load, you can eliminate actual ripple current, but a busted recitifier could possibly be the problem.


I appreciate that you stated you believe that the polarity and the recitifer are both OK, but I urge you to triple-check both. Ensure that you have everything right before connecting another one of those big, undoubtedly expensive, caps. Since you have no load connected, you can use a small cap to verifiy that the PS is basically working. Even 10 uf (rated 50 V or higher!) will be fine, and is much cheaper to replace if it pops. Ensure that you measure 38 VDC on the positive rail and -38 VDC on the negative rail, and that there is very little AC voltage on both rails.

[rgrayton]

macboy, thanks for your response. I have to say, I was hoping for something I never heard of, like initial charging too fast. I have checked 1 - 4 of your comments, again and again.
Here is why I am confident (but maybe should not be) -- the kit, PS and all worked with the included components, caps installed properly, of course. I could not generate the kind of power I was expecting(>100W 2-3886 in parallel), with the kit 1500uf cap. I installed 8x8200uf, 2/rail/ch and blew a 12A sloblow at input of xform. Took all the 8200s off, built another rectifier (new diodes) and fired up +/-38Vdc (the kit has something like 10uf on the board). Checked the voltage with dvm (a few times) and installed 1 (one) cap on one rail, + rail, cap + to +38v. It would seem like my problem is reverse polarity, cap popped and was warm.
Is there a way to check a polarity of a cap (forget about the marking)? I know this is a far out and wishful explaination and not really the way I do business. I must be an idiot.
THX,
-BG

[peranders]

The stripe at the capacitor is minus. Are you really sure that you have connected right? Wrong polarity kills the capacitor in the matter seconds.

RED wire in the RED hole in the DVM and BLACK wire in the BLACK hole!

You may take a photo and show us how you have done.

[demogorgon]

well, offcource you havn't done like i do regularily, and put both minus sides to GND, forgetting that the minus tab, when using the cap on negative voltages, goes on the rail, not to gnd?

it's one of those faults you do without really thinking of it, and don't see it before someone points it out to you

woth another check?

-Marius

[rgrayton]

The stripe at the capacitor is minus. Are you really sure that you have connected right? Wrong polarity kills the capacitor in the matter seconds.

Whoa, maybe at least one thing in my post was true, I am an idiot

I was going by a '+' marking as plus, it just dawned on me that may have been for the degree c stuff. Not sure right now as I am in work and am also not sure why I would not have used the '-' marking as I have done this properly on several others before now (brain fart comes to mind as a reason ). But, this is shaping up to be the most likely candidate and makes the most sense at this point.

$50 later...


Thanks for all your inputs.
-BG

[rgrayton]

I am using these cans upside down.
The only visiable marking on the bottom of the can is a '+', plus sign, right next to a terminal.
I offer these comments to one, maybe get off the idiot list, and two, just mention to anyone else, cap user beware of random plus signs. Who really needs to know about a degrees c spec on the can anyway?