How can I determine what ripple current rating I need for my bulk caps?

[Starlord]

Let's say I've got a board with a 5V switching supply, with the recommended ceramic cap on the output.

In my case, it's the TPS62133 and a 22uF ceramic cap:
http://www.ti.com/general/docs/lit/getliterature.tsp?genericPartNumber=tps62133&fileType=pdf

And let's say I have a bunch of LEDs being driven with PWM from said supply, say 16, with an average duty cycle of 25-50% and a max current of 20-25mA, let's say average current 10mA per LED.

In my case it would be the TLC59116:
http://www.ti.com/lit/ds/symlink/tlc59116.pdf

And then throw a couple servos or a DC motor drawing 250-500mA, and some other misc circuitry drawing around 100mA, and its associated noise.

Given all that, if I stick say, a 100uF bulk cap on my 5V rail after the regulator and it's output ceramic, on my LED driver, or on my 12V power input to the regulator before it's input ceramic cap... do I need to worry about the electrolytic bulk cap's ripple current ratings when selecting them?  Or is that something I only really need to worry about if I'm designing a power supply and I know the cap is going to be filled and drained at khz / mhz frequencies? 

Like if I just look at the least expensive 6.3mm and 8mm SMT electrolytics rated for >6V on Digikey:
http://www.digikey.com/short/799p74
http://www.digikey.com/short/799p7q

The smaller ones are typically rated for at least 70mA and the larger ones for at least 110mA.  The least expensive ones I mean.

So what I'm wondering is, is that going to be sufficient?  Is it more than sufficient?  Am I worrying about nothing at all, or would I be operating near the cap's maximum rating?

I haven't finished designing the board yet, so I can't run any tests.  I'm just looking for a best guess here as to whether I may have a problem, or if there's nothing for me to worry about and there's very little chance of there being an issue. 

I'm guessing that I'm worrying about nothing.  And if I knew a little less I'd just throw the caps in there and I probably wouldn't have any issues and I'd live in ignorant bliss, but I know just enough now to worry about these things that I don't have enough experience to know if I should worry about.

[woodchips]

Look at it this way. The PSU is a fast feedback system so that if the load suddenly increases then the dynamic performance of the PSU will detect the voltage drop and counteract it. In this scenario you don't need output capacitors, but they are always fitted. But making them too large can ruin the transient performance of the PSU. Have a look at the circuits of bench PSUs, they have very small output caps.

What the ripple rating is for is the smoothing of the rectified AC. The smoothing caps can only charge when the input voltage is greater than the voltage currently on the caps. This results in a very spikey current flow into the caps, this is the ripple rating.

[tszaboo]

Very (very) simple explanation:
You put the ceramic and the elco in parallel. Both of them has an ESR. The switcher has a maximum input current, that has to come from somewhere. Some of the current comes from the elco (ohms law with ESR apply). If the current coming from the elco is bigger than the rating that is bad, use more capacitors or different/bigger. Otherwise OK.

[The Chump]

IC s like the TPS62133 you are using are intended for use without the need for large value electrolytics. The datasheet says to use 10uF X7R or X5R on the input (12V) and 22uF X7R or X5R on the output. Chances are those ceramic caps will carry almost all of the 2.5MHz ripple and any additional electrolytics will not make much of a difference.

[rx8pilot]

You may see some benefit in using a few 10uf ceramics in parallel to get lower ESR instead of on 22uf. At 2.5Mhz, electrolytics won't do much.

[Starlord]

But what about the LEDs and servos?  Shouldn't I have bulk capacitance for those, by the point where they're drawing power from the board, to cut down on noise?

Also, I figured if my power source is at the end of some long leads, like a wall wart, that I ought to have something more than 10uF by the input to the regulator.

[The Chump]

But what about the LEDs and servos?  Shouldn't I have bulk capacitance for those, by the point where they're drawing power from the board, to cut down on noise?

Maybe some extra 10 or 22uF ceramics if they are a very long way away. You would need to factor the extra capacitance in to the feedback stability calculations. The datasheet has quite a lot of stuff about that.

Also, I figured if my power source is at the end of some long leads, like a wall wart, that I ought to have something more than 10uF by the input to the regulator.

Again, I would use 10uF ceramics with a high frequency converter such as this.

It is possible to prototype this kind of thing, rats-nested above a copper ground plane. Not very easy, but possible.

have fun!