Bypass capacitor theory

[Oli]

Dear All,

Thank you all for your thoughts and ideas.

Thinking in particular about the digital decoupling situation...

Am I right in understanding that the purpose of the decoupling capacitor is not really to address the small amplitude of supply ripple at the fundamental frequency of switching logic; rather to address the significant ground bounce and the harmonic ringing that would otherwise occur after each transition? (much higher frequency)

The use of 0.1uF might make more sense.

 

[rayma]

Am I right in understanding that the purpose of the decoupling capacitor is not really to address the
small supply ripple at the fundamental frequency of switching logic; rather to address the significant
ground bounce the harmonic ringing that would otherwise occur after each transition? Being a
much higher frequency- the use if 0.1uF might make more sense.

Yes, and in some vhf cases a 0.01uF works better.

 

[Oli]

rayma,

Thank you. I had become confused. After researching a lot of ceramic ESR vs frequency curves I discovered that the approx. 2mohm minimum impedance for a 10uF ceramic of interest occurred at 2Mhz. If I changed for a 0.1uF ceramic, the approx. 2mohm minimum occurred at 16Mhz. This led me to wonder if the fundamental frequency was not the issue, rather some harmonic. Then I remembered the pictures of ground bounce... a phenomenon involving higher frequencies and some significant amplitude.

I suppose also that any RF choke or ferrite in the supply line should also 'target' higher frequencies?

0.1uF it is then and higher frequency ferrite?

 

[rayma]

any RF choke or ferrite in the supply line should also 'target' higher frequencies?
0.1uF it is then and higher frequency ferrite?

Yes, or a small resistor instead. An SMT capacitor 0.01uF-0.1uF, with ground and power planes, is best.
Minimizing the loop inductance is key. https://www.speedingedge.com/PDF-Files/emcvccgndbounce.pdf
http://ksoehein.weebly.com/uploads/2/5/9/7/25977945/right_the_first_time_vol_1_09-15-08.pdf

 

[wg_ski]

Yes, and in some vhf cases a 0.01uF works better.

Sometimes a .01uF ceramic with the worst possible dielectric absorption is actually what you want/need. Low inductance and comes with a built in de-Q resistor.

 

[Oli]

I think I've got the right idea about the decoupling capacitor

Going back to the supply RF Choke / Ferrite.

If I have a 1.4Mht would clock, what would be a sensible peak impedance and frequency?

 

[rayma]

Sometimes a .01uF ceramic with the worst possible dielectric absorption is actually
what you want/need. Low inductance and comes with a built in de-Q resistor.

Yes, and with no added inductance.

 

[benb]

...
There is a reason for the customary .047 or .1 *ceramic* bypass cap "as close as possible to chip or even discrete amp feed legs".

Also *where* to bypass to needs careful consideration.

I feel the need to add (because I've seen layouts where this wasn't understood!) that "as close as possible" means positioned so that the connections (as in PCB traces) between the cap and the chip it's bypassing are as short as possible.

 

[gerhard]

I have once made some experiments with decoupling capacitors:

< http://www.hoffmann-hochfrequenz.de/downloads/experiments_with_decoupling_capacitors.pdf >

pdf, 6.8 MB

cheers, Gerhard

 

[phase]

That's interesting to see, thanks for sharing...

 

[Eva]

Post #29 is just in the spot, it shows how to measure test decoupling fixtures, the "double resonance" effect of mixing capacitors without clue, and the inefficiency of some capacitor types. Post #19 shows the basics of how to model it in software for designing decouplings skipping some lab work, but the lab work has to be done a few times in life before the "modelling in advance" method is reached.

There is a third discipline: Checking decoupling in final applications. This in most cases has to be done with oscilloscope and a square wave current source. In switchmode circuits the square wave generator is "built in". For other circuits and low budget, as in DIY, there are simple test oscillators easy to build in breadboard. Starting from a 555 timer, CMOS inverter gates, or SG3525, used as FET drivers, or driving the load resistor directly through a diode for low current applications.

This is my multi-function fixture (primary function is to measure SMPS inductors):
http://www.diyaudio.com/forums/power-supplies/77919-measuring-inductors.html#post895799