EMRFD Message Archive 9979

Message Date From Subject
9979 2014-04-29 03:22:38 Bob Decoupling capacitors
Hello savants-

I am putting together a transmitter (an Ultimate 3 QRSS, from Hans Summer).  I am a mere mortal putting together a kit.  There are comments to add decoupling capacitors close to the equipment.  I often see two capacitors at these locations; one small (~0.1 uf) and one larger (electrolytic of some arbitrarily larger size).  Here, they are to be placed across the power leads to a GPS power by a regulated 3.3v line powered source.

Question: Is there a science to determine the number and size of these decoupling capacitors or is it just "1-2 of from Vcc to gnd, whatever you got"?

9980 2014-04-29 03:32:19 ashhar_farhan Re: Decoupling capacitors
There's science to it. Until recently we thought that two capacitors of different values are needed to prevent self-resonance of the bypass capacitors. However, subsequent simulation and measurements undertaken by W7ZOI have proved that two capacitors of the SAME value are better at cancelling self-resonance.
However, at times you need to bypass at RF and AF frequerncies at once (low noise oscillators, audio amps, modulators, etc.). The electrolytic capacitors do not have a fast enough reaction to pass the RF. You could use a tantalum or do what the impoverished folks do : parallel a disc ceramic with an electrolytic.
- f
Sent from BlackBerry® on Airtel

9981 2014-04-29 03:36:59 David Re: Decoupling capacitors
The lead and trace length between the capacitor and load are short to minimize
the series resistance and inductance which would otherwise compromise the
performance of the capacitor for decoupling.

The best or at least minimum capacitor values can be calculated for well defined
loads and decoupling requirements but it is often easier and adequate to just
determine them empirically.

On Tue, 29 Apr 2014 06:22:37 -0400, you wrote:

>Hello savants-
>I am putting together a transmitter (an Ultimate 3 QRSS, from Hans Summer).
> I am a mere mortal putting together a kit. There are comments to add
>decoupling capacitors close to the equipment. I often see two capacitors
>at these locations; one small (~0.1 uf) and one larger (electrolytic of
>some arbitrarily larger size). Here, they are to be placed across the
>power leads to a GPS power by a regulated 3.3v line powered source.
>Question: Is there a science to determine the number and size of these
>decoupling capacitors or is it just "1-2 of from Vcc to gnd, whatever you
9992 2014-04-29 19:29:10 w7zoi Re: Decoupling capacitors
Hello Farhan, Bob, and gang,

Glad to see folks looking to find the science in this question rather than just the lore.    The measurements are not really all that difficult.

The scheme I use to evaluate the bypass caps is a 50 Ohm signal generator and a 50 Ohm detector.   The detector can be a power meter or spectrum analyzer, or even a 50 Ohm terminated oscilloscope.   Whatever will provide the sensitivity should do the job.  A test fixture is also built, consisting of nothing more than a piece of metal and a couple of coax connectors.   The connectors are mounted in/on the metal and a piece of wire runs between the center pins.   The generator is attached to one port and the detector to the other.    The capacitor is then attached from the center pins of the coax connectors to ground.      As you tune the generator you will notice that the attenuation varies considerably.     It will be very large at the self resonant frequency of the cap where the C resonates with the inductance of the part.   The L will depend upon the wire lead length, but will still be there even when the wire length goes to zero.    The cap has a physical length and that's enough to lead to some L.      A typical 0.1 uF leaded part will resonate at 7 MHz, which is certainly handy for those of us build rigs there.

If you put anther cap in parallel with the first, you have two inductors and two capacitors.    Both will have the expected series resonance that leads to a lot of attenuation at the self resonant frequencies.    What is not initially expected is a parallel resonance.     This is where the L of one of the caps  resonates with the C of the other.    The parallel resonant frequent will be between the two series resonances.    This parallel resonance is a high impedance where there is virtually no bypassing at all.  

This is all too easy to measure.   I urge you give it a try.   Alternatively, it is very easy to simulate in a program such as SPICE.     Some "natural" combinations such as a parallel 0.1 uF and a .001 uF can be especially bad.   

Sometimes we want some extra bypassing at really low frequency, so a parallel electrolytic is added.   This may not be so bad, for the Q of the electrolytic is so low that it does not cause any parallel resonance to occur.    In one experiment I did, I started with a 0.1 uF, measuring about 40 or 50 dB attenuation at the self resonance.   The depth of this notch is a measure of the Q of the LC.   A 10 uF Aluminum electrolytic cap was then paralleled with the 0.1.    The bypassing at low frequency got much better, no parallel resonance was noted, but he depth of the attenuation at the 0.1 uF self resonance degraded by 1 dB.   This was a case where the parallel combination worked, but it was a measured case.    

The case where you have two essentially identical caps in parallel is an interesting one.    The two series resonances are on top of each other in frequency.   The Q of each resonance is low, so the dips are wide.    There is just no place where the parallel resonance can lead to a high impedance.   The bypassing improves without the problems that arise with two unequal capacitors.

OK, that's a review of this subject.    There is a lot that one can do if he or she elects to do some measurements or calculations and this is just a brief look.

Farhan, you attributed this to me.    I've been standing on a figurative soap box to talk about the effect in recent years.    However, that's all I've done.    This work is not a discovery of mine.     We started seeing the paralleling of capacitors, often of unequal value, back in the early solid state power amplifiers of the 1960s, when the subject was young.     It still shows up these days.  But the folks looking for wide band bypassing discovered the errors of earlier times and the folks doing microwave circuits were