Paralleling caps to meet ripple current requirements?

What value cap are you looking for ?

Paralleling caps for increased ripple current is quite normal btw.

Graham

Yes, you can. But since ripple current is what heats the caps and determines their life, 4 might be better. I would like to leave a fat margin (perhaps 50%) between the ratings and the actual current.

If you want to keep almost all of that ripple current isolated between those caps and the converter, add a little inductance between the upstream supply and the capacitors. This will pass almost all the ripple current through the caps, instead of sharing it between the source and the caps, but that is where it belongs.

It's a 12V-to-0.9V converter. Need 270uF for the input and 2000uF (or so) for the output. Looking at the Vishay OS-CON right now...nice ESR specs. They're pricey but this is just a one-off converter I'll be using "in-house".

It seemed that idea would work just fine. Just wasn't sure if there was a "gotcha" somewhere. Thanks!

John

-- remove SPAMMENOT for e-mail responses --

Good tip, thanks.

Interesting...the test-rig schematic for the converter (an Artesyn SIL40C) showed a 270uF cap, 1.5uH series inductor, and then another

270uF cap at the Vin to the converter. Wasn't quite sure of the inductor's function...now I know. :-)

Do I still need the full 10A ripple requirement for both sets of caps, before the inductor and after the inductor?

John

-- remove SPAMMENOT for e-mail responses --

Really !

If you're worried about lifetime just use higher temperature parts.

To the OP. Electrolytic caps are usually rated as having x thousand hours life based on operation at their max rating (ripple current) in a specified ambient temperature, the 'wear out' mechanism basically being heat which slowly evaporates the electrolyte).

Lifetime can be increased by reducing the ambient temp OR using a cap with a higher temperature rating. Lifetime doubles for every 10C reduction in temp typically.

For example a cap rated for 2000 hours at 85C would last 8000 hours at an ambient temp of 65C and a '2000 hour 105C' cap would last 32000 hours under the same conditions.

Graham

Ummm, it starts getting more complicated when you add inductance. The inductor actually reduces the ripple current in the caps since it stores some energy too.

You won't ever go wrong by being generous with ripple current specs though. There's also no reason you can't simply increase the capacitance of the input cap (rather than fit several smaller caps - provided you still meet any required ESR specs) and get extra ripple current capability that way.

Graham

I seriously doubt it, if the design does not involve any resonances. The cap between inductors would normally see very little ripple current. I also doubt that having two equal capacitors is optimal. The one between inductors is essentially there just to capacitively divide the RF current that jumps through the inductor via its stray capacitance. I good, low equivalent series inductance (i.e. stacked film or ceramic multilayer) might well do better than the 270 uF even if only a fraction of its capacitance. But you may need a parallel larger capacitor to suppress any resonances between those inductors and the small capacitor that the converter generates.

After you get past the ripple current requirement of the final cap, you have to deal with where that current goes. The loop area this current surrounds needs to be minimized to get the most use from those capacitors and minimize magnetic fields caused by the ripple.

Similar sized parts from the same maker with different temperature ratings usually have shorted life ratings for the higher temperature units, and lower ripple current ratings. In other words, the main difference between the lower and higher temperature rated units is a derating of the ripple current rating and a shortened life prediction. I'm sure they tweak other things but this is the man difference.

For this reason, if you derate the ripple current rating, as long as you don't exceed the ambient temperature rating, the lifetime can safely be assumed to improve.

The test rig schematic only had one inductor.

+-------+ 1.5uH | | 12V---+------- ___ -------+-------+Vin | | -UUU- | | | --- --- +-------+ ---270uF 270uF--- Converter | | GND GND

Draws about 4.4A at 12V and delivers 40A at 0.9V.

Good tip. I was going to try to minimize it as much as possible thinking that I really didn't want to sling whatever EMI was coming off of that 10A of ripple. :-)

Thanks! John

-- remove SPAMMENOT for e-mail responses --

the

Very good point. I'm not sure how many options I have with the low ESR I'm looking for but I guess it can be a tradeoff between ESR and temp rating...especially if the price difference is large. :-)

John

-- remove SPAMMENOT for e-mail responses --

I don't know the physical shape of the converter and whether it is a package or a bunch of components, but since you are paralleling capacitors for the final storage, you may be able to make two groups and route their ground sides to the common point on the converter by two equal and opposite loops (from a current standpoint) so that the bulk of the external field is canceled.

too.

required

I noticed that the ripple rating went up as the capacitance went up. Since my load for the converter is pretty constant (other than ON/OFF), does that ease the requirement for ultra-low ESR caps?

The data sheet and app notes for the converter emphasize ESR but I'm thinking that if I have only a slowly varying load (if varying at all) that low ESR caps aren't nearly as important for voltage regulation but might only needed to satisfy any input requirements for the converter?

I guess the lower ESR caps would make it easier to keep the caps cool though, taking the 10A ripple current into account.

John

-- remove SPAMMENOT for e-mail responses --

Great idea!

It's a vertically mounted SIP with 24 pins (single row) spread out over 2.4".

There are three Vin pins in the center with the Vout and GND pins alternating for pins 15-25 and scattered GND in pins 1-8 to go with the POWER GOOD, TRIM, and CURRENT SHARE signal pins (pins 1-8).

I guess I have either side of that row of pins to play with.

John

-- remove SPAMMENOT for e-mail responses --

the

You can end up juggling these things forever.

Graham

LOL, you read my mind on that one! I guess I just gotta' dive in and start buying caps. It's only a one-off device. OK, maybe 2 or 3 if it works well. :-)

John

-- remove SPAMMENOT for e-mail responses --

That is just the equivalent of paralleling several smaller caps, inside the can.

Not at all. The ripple current is related to the value of the output current, not how fast it changes. It is the load current being drawn from the input capacitors, but only during part of the converter cycle, and not at the other part of the cycle. Remove the average from that pulse train and you get an alternating direction current of about half the DC output current.

It lowers the actual watts dumped into the caps, but also lowers the peak spike voltage that appears across them and that propagates upstream through the filter inductor.

That is what I was picturing.

In that case, please include a few 1uF, 50 or 63 volt stacked film caps (Panasonic type V) to experiment with as high frequency bypass for the electrolytics, if you can get them. These would go as close to the SIP pins as possible, to keep the highest frequencies out of the larger loops.

Ahhhh...OK.

So the ESR times the ripple current = the voltage spike? I can see how reducing the ESR would be valuable here.

Thanks! John

-- remove SPAMMENOT for e-mail responses --