Capacitor choice for switching PSU?

My favorite choices for low voltage use at the moment (for high ripple current, low ESR and low cost) are Panasonic FM series electrolytics and the Panasonic ECJ-3YB1E106M, 10 uF 25 volt X5R 1206 size MLCC (with something like .002 ohms ESR), both available from Digikey.

Do you care about size, lifetime at your operating temperature, sticker price, ...

Depends on what the supply is, and more crucially, what the loop response, output ripple and step response has to be.

The issues are input and output operating voltage (above about 10V, there aren't many MLCCs of high capacitance that readily available anyway), output capacitance requirements, output ESR, input ripple and surge, and the controller you are using.

The tradeoffs are:

Electrolytics have high capacity, reasonable ripple RMS ratings and relatively high ESR (although the panasonic special polymer series are quite low). Good choice especially at the input of a power supply for input voltages above about 8VDC. The special polymer series are quite effective on outputs too.

Tantalums have low esr (in their latest incarnation - see the AVX TPS series for instance), but have a habit of spectacularly failing for high surge currents into the device, even well below their rated voltage. For this reason, they are not usually recommended at the input of a regulator. They are fine at the output, where there are large discharge currents, not [usually] high surge currents.

MLCC have the lowest esr available (below 2milliohm on certain devices) at up to 100uF for 6.3V, but this is not a panacea. Low output ESR can play havoc with loop compensation.

My usual approach is to use some tantalums with the bulk of the capacitance at the output with a couple of MLCC parts (about 1/5 or less total Cout) to lower the esr for small signal ripple performance. At the input, it depends on Vin - I'll use tantalums, but overrate them by at least a factor of 2 for both rated voltage and rated RMS ripple current for safety. I would use electrolytics, but they just tend to be too big for my applications. If the input voltage is low enough, I'll use MLCCs.

For Cout, using a mixture of tants and MLCCs gives me an acceptable Cout/ESRout and superior ripple performance. It won't work for everyone, obviously, and much depends on Vout (and range of load currents too).

MLCCs aren't that expensive, incidentally, especially given their life and temperature stability (Winfield Hill posts here, a co-author of 'The Art of Electronics', which has an amusing table of the various characteristics of capacitor types - look under temperature stability)

There really is no simple answer - every supply has to be designed with it's specific requirements in mind.

Cheers

PeteS

Yeah. I like it too.

Geez! Google Groups is an INCREDIBLE piece of CRAP!

I've even selected "Fixed Font", and it only puts the headers and some quoted text in courier.

Didn't they say "Beta?" Doesn't that mean that they're testing it?

How do we inform them that their test is an abysmal failure, and that they're losing customers except for k3wL IM D3wdZ?

Thanks, Rich

While not arguing with your point about the rendering, clicking **show options** then **Show original** will get you a completely monospaced page--with headers, no less.

To bitch about things (or report a bug)