Power supply filter cap values

[GGBB]

What is the determining factor for choosing the values for the filtering caps used for a typical 9V power input on a pedal?  The 100uF seems common for the big cap, but I've read some claim bigger is better.  For the small cap, I've seen it range from 1nF to 100nf to nothing at all.  Is there a guideline for this, or is it just a matter of parts availability or cost?

[kurtlives]

http://en.wikipedia.org/wiki/Ripple_%28electrical%29

Decide how much ripple voltage is acceptable.

A small film cap often bypasses the large electrolytic for better ESR handling and RF protection.

[Cliff Schecht]

Modern electrolytics usually have good enough ESR to shunt most noise that is introduced to our pedals. A small value cap isn't pointless mind you, they have very low ESR and will swallow up higher frequency content (where the impedance of the electrolytic falls off the film cap should take over, at least ideally).

Unless you know why you are using a specific value of bypass capacitance for the 9V supply, a 10-100uF cap usually suffices (if any is needed at all). On higher gain pedals which are more susceptible to supply noise (or any effect that is really) you can put in additional capacitance to help hush things up sometimes.

In the industry the standard practice for capacitors on a board is to use a "large" cap (10uF is usually fine for audio) and decouple the supply every 6"-12" depending on how critical the application is. Also every IC should have a 0.1uF decoupling cap as close to the supply pin(s) as possible to prevent any noise from getting in. Realize that VDD traces without decoupling caps are essentially big inductors/resistors that will increase the impedance seen by the circuit(s) the VDD trace is feeding. As the trace becomes longer the capacitors become less effective at decoupling supply noise/transients.

[GGBB]

Thanks.

[PRR]

The big cap should generally be 1uFd/mA (or 1,000uFd per Amp).

More is better but bigger/costlier.

[GGBB]

Thanks, Paul.  I really like answers that involve concrete mathematics - satisfies my analytical side. 

[Paul Marossy]

The big cap should generally be 1uFd/mA (or 1,000uFd per Amp).

That's a good rule of thumb I've never seen anywhere before now.

[PRR]

> answers that involve concrete mathematics

This is lightweight concrete. I know many pedal-builders don't want so much arithmetic.

1uFd/mA actually gives *around* 1V of ripple.

If you are building a 400V supply, 1V is "nothing", or "too good" (expensive). A 400V 200mA supply will probably use, not 200uFd, but 40uFd, because 5V ripple is still fine.

If you are building a 5V supply, 1V ripple is a Big Deal. Back when we had real 5V logic at a couple Amps, 10,000 and 22,000uFd caps got us down sub-Volt ripple.

Filter-caps are a large part of the cost of electronics. When you plan to build a million units, you need to do your sums and shave-out the last cent tolerable.

But for a vast range of low-medium voltage work, in one-off building, 1uFd/mA will give you a workable supply (not too hummy, not too big) without the effort/pain of actually doing a ripple analysis.