caps in parallel

[volumetricsteve]

I have a weird situation where I've recapped a board - replacing electrolytics with solids or tantalums. So far, great success. My only hang up is one cap gets uncomfortably hot, so my thought was to use a few lesser capacity caps in parallel to distribute the workload, thus distributing the heat.

The problem is 470uf caps are awkward to replace, at least on digikey I couldn't find too many easy combinations that add up to 470. I did eventually find these for my first configuration 100uf+370uf:

http://www.digikey.com/product-detai...313-ND/4552809

http://www.digikey.com/product-detai...539-ND/3462162

I see one of those is aluminum and one is tantalum, also one of them is rated at 63V and i'm not sure if that's ok in a 16v application.

for my second configuration, 220uf x 2 + 15uf x 2 which I'm much more fond of:

http://www.digikey.com/product-detai...3-1-ND/1973911

http://www.digikey.com/product-detai...9-1-ND/2781413

My concern here is I'm just not sure if there's some weird interaction of using different types of caps in parallel. I'm assuming if you use a 10% tolerance and 20% tolerance cap in parallel, your effective tolerance is 10%. On a dielectric level, do any kinds of caps go insane when put in parallel with others?

[eccerr0r]

Re: caps in parallel

I'm not sure why the cap is getting hot. Can you use a lower esr version or is the heat coming from elsewhere? Is the working voltage too high for the capacitor?

470uF is a standard value. Likely it will be difficult finding pairs of caps to get to that value. Connecting caps in parallel should work fine as long as they all can be put close together - additional wire adds inductance and resistance. Effective tolerance will be the weighted average of the caps you have. Voltage limit is the lowest of the set.

[mariushm]

Re: caps in parallel

It doesn't work like that.

Capacitors heat up due to ESR of the capacitor itself and due to heat coming from heatsinks near it and the pcb itself.

If you replaced the original capacitor with capacitors that have higher esr, the capacitor will heat up from within and fail eventually. Lower capacitance capacitors usually have higher ESR.
If you have heat coming from pcb or heatsinks, both new capacitors will heat up just as much.

The 20% tolerance means just that, the capacitor actual capacitance could be within -20% .. + 20% of the value printed on it. It also varies from capacitor to capacitor and it varies on how the capacitor is used in the circuit (the capacitance and esr and other specs change with frequency)
In reality, manufacturing process these days makes it possible to have electrolytic capacitors much closer to the value printed on them, so you don't really have -20%..+20%

These tolerances are not additive or subtractive.. you'll still have -20%..+20%.
Think about it : let's say you have two 100uF +/- 20% and you simply have bad luck and both are 80 uF when you measure them ... right there where they'd be outside -20%.
If you put them in parallel, you will have 80+80 = 160uF, which is still 20% lower than 200uF.
In reality, one may be -5%, the other may be +10%...

IMHO, forget about tantalum capacitors, they have a tendency to explode or burn (with real flames) if they're subjected to overvoltages or you simply have bad luck (few out of hundreds and in time explode with no reason).

Determine the voltage going though that section of the circuit, determine the specifications of the original capacitor (ripple current, esr/impedance). Then replace the capacitor with something more modern and with similar specifications. Where possible, you could use polymer capacitors (ex if circuit uses 470uF 10v capacitor but voltage in circuit is 1.8-3.3v then you may be able to use a 470uF 4v polymer there)

[volumetricsteve]

Re: caps in parallel

Thanks for the replies!



the cap I'm currently using is :

http://www.digikey.com/product-searc...ds=565-3617-ND


the original cap was:

http://www.digikey.com/product-detai...1785-ND/589526

there really isn't anything else on the board producing heat near the cap, so I'm thinking it just has to be running really close to 16v.

The caps I wanted to replace it with were rated at 20v and 25v to see if that'd help.

From what you're saying about tolerance...it sounds like a lower tolerance means I'm more likely to get a cap with the value I'm ordering....i'm confused on that point, I thought higher tolerances were better, but only in regards to voltage, if it's a tolerance on what the actual capacitance is, that sounds like a bad thing.

My application is..pretty unlikely to have over-voltages I think so I'm not too worried about the issue with tantalums. It's an old pentium 3 motherboard where the hardware never changes and it won't enter sleep states or idle modes, load should be continuously about the same while it's powered on.

[mariushm]

Re: caps in parallel

The voltage rating on the capacitor simply tells you the highest voltage the capacitor can tolerate before it gets damaged.

Motherboards work with 5v and 12v and -12v (for some very few things) Older motherboards also used -5v. So even if the power supply of your computer is a bit lousy, the capacitor shouldn't "see" more than 12-13v.
The capacitors won't heat more or not work right if they're closer to 16v, they'd work just the same.
But, you can use capacitors rated for 25v or 35v but it wouldn't do much of a difference, unless the capacitor has a bigger volume (diameter and/or height)

Nichicon PW is a good series.

The ESR of the Nichicon is about 100 mOhm which is a high value if you compare it with modern low esr capacitors, but back then when Pentium 3 was new it was considered very good.

It's not a great idea to replace it with a capacitor that has ESR 10 times lower at around 10mOhm, the circuit may be tweaked or designed to take "advantage" of that higher ESR.

Now...motherboards in the pentium 3 era pretty much used only 5v from the computer power supply, only with Pentium 4 and newer the motherboards started to convert 12v to a lower voltage to power the processor and other things on the motherboard.

So my guess is that capacitor was used for -12v (which was used by ISA slots and serial ports) or for 12v from the power supply.
It shouldn't be a critical capacitor, therefore you would be fine with a regular electrolytic.

Going back to the capacitor, the Nichicon PW capacitor is rated for 5000h @ 105c ... the saying is with every 10c decrease in temperature the lifetime rating doubles, so the capacitor would be rated at 10k hours @ 95c , 20k hours @ 85c and so on.
Don't know how "hot" that capacitor is, but if it's just a bit warm, let it be.

If you still want to replace it, better choices would be Panasonic FR or Nichicon KY:

http://www.digikey.com/product-detai...417-ND/2433551
http://www.digikey.com/product-detai...419-ND/2433553
http://www.digikey.com/product-detai...1549-ND/756065

They're rated for 6-8k hours @ 105c, they have a more reasonable ESR at around 40mOhm, they're high quality parts.

[volumetricsteve]

Re: caps in parallel

Thank you so much for the feedback, sounds like good advice. I was surprised by the 16V rating as well, considering the voltages typically seen in motherboards. For what it's worth, this motherboard uses RIMM, which may do funky things with voltages, I'm not sure. When I say the cap gets hot, it's hot enough that I can't keep my finger on it for more than a few seconds.

I wondered about the ESR issue as well, though the motherboard seems rock solid, I've been stress testing it and so far nothings blown up.

Could the ESR being too low account for the high heat?

Thanks again

[SteveNielsen]

Re: caps in parallel

"Could the ESR being too low account for the high heat?"

I had the same thought. I don't know enough about exactly what happens with the dialectric and plates in them to give a good answer though.

[volumetricsteve]

Re: caps in parallel

I would think...high resistance = more energy being converted to heat, thus....heat. low ESR should be the opposite, but...I have no idea.

It seems like I have a 16V cap running in what should be a 12V environment that gets hot enough I can't touch it. I'm starting to suspect goblins.

[budm]

Re: caps in parallel

Cap getting hot does not make sense, the new cap is only 10mOhm (old one is 90mOhm), the ripple current rating of the new one is 6.1A (old one is 0.76A), so for it to get hot it may be it is installed backward?

[volumetricsteve]

Re: caps in parallel

wouldn't the computer flip out? I mean electron flow going the wrong way...that'd cause serious issues I'd imagine, right?

This computer boots, I've been running all different versions of windows and linux without any apparent issues

[budm]

Re: caps in parallel

If they are installed correctly then I cannot see why they will be too hot to touch. Wrong polarity applied to the cap will heat up the cap, you should verify just to be sure.

[eccerr0r]

Re: caps in parallel

Actually installing electrolytic caps backwards the effective capacitance at first will still be similar to forward. However it will electrolyze the electrolyte causing it to bubble, and usually there is some recombinant chemicals in the capacitor to make a small amount of gassing back to liquids... and convert that energy to heat. The heat will boil the rest of the electrolyte and cause the capacitor to explode.

So double check polarity and make sure voltage is well within the voltage range, and make sure you use quality caps. I've noticed some knock off "fake" caps really having lower voltage rating than printed so I'm sure these will start heating up earlier than the real deal - like if it was rated for 16V it would have significant leakage current at 16V (100 microamps or more) but a quality cap would not. This leakage current will heat the capacitor. But you should be fine with caps from DigiKey - this is only a concern for eBay and other shady places...

[volumetricsteve]

Re: caps in parallel

I'll check the polarity for sure, but for what it's worth, I've had the system on running stress tests (fully populated ram, pci and AGP busses) intentionally to stress the caps I put in the board, this is the only cap that heats up, but I'd think if I did it backwards and put it through that kind of stress, wouldn't it have gone nuclear by now? Thanks again

[budm]

Re: caps in parallel

The original single cap with rating of only 3/4A means that is not high power power supply section.

[volumetricsteve]

Re: caps in parallel

I realize I made a mistake when I posted the "original" cap, this is the actual original cap (I took a good long look at original capacitor and found it was a part of the VR series of nichicon caps.

http://www.digikey.com/product-detai...2-3-ND/3438545


the ripple current is really low, but I see it doesn't have an ESR rating at all...I'm guessing that means it just doesn't have any ESR?

That'd explain the heat of the cap I replaced it with, which has 10 mOhm ESR.

Admittedly, I don't fully understand the mOhm nomenclature, I'm used to seeing MOhm, which I thought meant Mega Ohm (one million ohms, yes?) so mOhm I'd assume would be milli Ohm, or one millionth of an ohm. Am I even close?

How bad did I mess up my motherboard? (even though it seems perfectly stable from a software standpoint)

[SteveNielsen]

Re: caps in parallel

It is a general purpose cap, not a low ESR cap, so the ESR it has is not given nor relevant for genral purpose use.

1 mega-ohm is 1 million ohms, 1 milli-ohm is 1/1000th of an ohm.

[mariushm]

Re: caps in parallel

Lower ESR means the capacitor has less internal resistance, which means there's going to be less heat generated by the ESR inside the capacitor.

Nichicon VR is a general purpose series, so since it's not designed for high frequency circuits (switching power supplies and others) the ESR value is not worth mentioning (it's going to be a bit on the high side). If I were to guess, I would say maybe around 0.4 ohm (400 mOhm).

Like I said, it's probably used to filter -12v (for serial ports). In this case the capacitor may normally be inserted the other way around... positive side connected on the ground pad. But the printing on the pcb should be correct.

If it's hot to the touch, something's wrong.

[volumetricsteve]

Re: caps in parallel

huh.....that's all interesting....not to derail, but if a cap has any ESR, that's resistance that augments the voltage of the output..right? I can't envision a situation where the output voltage wouldn't matter...but I'm sure they exist.

I have several of the same motherboard and I've only recapped one of them.
I checked another board and it had the right polarity shown, at least, the same polarity rules I used for all the other caps also appear to apply to this cap (the cathode faced the pci slots) so at least in keeping with the polarity of the other boards, it would appear mine is inserted correctly.

Maybe I just got a bad cap....granted I got it from digikey, and all the others are fine, this one was the only 470uf 16v I ordered...I think it's also the only
United Chemi-Con cap I bought. Do they have a history of producing duds?

I have an especially reckless question though....if my board is live and I leave that capacitor in there, can I measure the voltage across that cap to see if it reads 16v?

[mariushm]

Re: caps in parallel

not to derail, but if a cap has any ESR, that's resistance that augments the voltage of the output..right? I can't envision a situation where the output voltage wouldn't matter...but I'm sure they exist.

All capacitors have ESR and other parameters. There's no electrolytic capacitor without ESR, it's just that for general purpose capacitors, this ESR value is not specified in the datasheet because it's not a big deal for the type of circuits that require general purpose capacitors.

Yes, there's going to be some tiny voltage drop due to the ESR, but as the capacitor can charge super fast, it doesn't matter that much and the voltage drop is very small in the first place.

UCC is a good manufacturer. They had some problems in the past with a few series (KZG is the most known) but the ones you'd buy on Digikey are perfectly fine.

If your board is live yes, you can put the multimeter on voltage and the probes on those two leads (or pads) and measure the voltage.
You would not see 16v, you would see the voltage that's in circuit. If your power supply outputs 12v, then the voltage on the capacitor will go up to 12v, not 16v.

[SteveNielsen]

Re: caps in parallel

Concerning the role that ESR plays in heat development and dissipation of a capacitor it is more complicated than Ohm's Law applies in DC circuits. Have a look:

http://www.radio-electronics.com/inf...ial-basics.php

Even the best manufacturers occaisionly pop out a dud that doesn't get caught by QA, so it's possible the part is bad.