Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!uunet!zephyr.ens.tek.com!wrgate!midas!jeffw
From: jeffw@midas.WR.TEK.COM (Jeff Winslow)
Newsgroups: sci.electronics
Subject: Re: -How do you pick the right capacitor?-
Keywords: filter, capacitor, parallel
Message-ID: <1246@wrgate.WR.TEK.COM>
Date: 3 Jan 90 20:39:19 GMT
References: <7152@chaph.usc.edu>
Sender: nobody@wrgate.WR.TEK.COM
Reply-To: jeffw@midas.WR.TEK.COM (Jeff Winslow)
Organization: Tektronix, Inc., Beaverton,  OR.
Lines: 27

In article <7152@chaph.usc.edu> cyamamot@girtab.usc.edu (Cliff Yamamoto) writes:

>Again, I'd like to ask those analog and power supply gurus out there how
>one selects the proper values capacitor(s) for filtering ripples that have
>frequencies in the audio as well as RF ranges.  How are these parallel
>combinations of capacitors actually chosen?  Is there any kind of formula
>to use?  Are there any pros and cons?

For every capacitor you can draw an impedance curve with respect to frequency.
Real capacitors have series inductance (leads, and plates wound in a cylinder)
and series resistance (leads, plates, some dielectric losses?), so, above a
certain frequency they don't act like a capacitor. Electrolytic caps have high
ESR's relative to capacitive reactance near the series resonant point, so they
have broad minima in their impedance curves. Film and ceramic capacitors tend
to have much smaller ESR's relative to capacitive reactance at the resonant
point, and so can have quite a sharp minimum. The thing to do is probably to
get the impedance-frequency curves of a whole lot of parts you think might be
useful, and try to overlap selected ones to get an impedance that stays below
what you need to filter out your ripple. (Of course, this is going to depend
on the output impedance of your alternator and connection to it.) Watch out
that you don't start getting parallel resonances when you do this - say, from
the ESL of the electrolytic combined with the C of a film capacitor. You could
play around with SPICE to see what your impedance curve might look like, but
since no SPICE I know of models frequency-dependent resistance, the simulation
will be pretty approximate.

						Jeff Winslow