Re: [avr-gcc-list] AVR ADC (waaaaaaaaaay off topic)

[Graham Davies]

Clemens Koller wrote:
> For a propably [sic] "better than okay" solution I would recommend a
> 100pF C0G or even a more priced 1nF C0G capacitor in parallel to
> [a 100nF X7R ceramic], close to the pins, to be very sure to get rid
> of any high frequency components.

Personally, I think Benjamin is right to be cautious about "common sense" 
and to know his limits.

What you're recommending used to be "common sense", but doesn't actually 
work.  If you draw out the equivalent circuit of two different valued 
ceramic capacitors connected in parallel you will discover you have a 
parallel reasonant circuit.  For simplicity, draw the series parasitic 
inductance at the top, away from ground, on both capacitors.  Now consider 
frequencies at which the larger capacitor is effectively a short circuit. 
Look at what's left - a parallel (anti-) reasonant circuit with the power 
supply tapped into the middle of the inductor.  Near the reasonant 
frequency, the impedance flies up, which is the opposite of what you want.

These days, "common sense" for high frequency decoupling is to use 
capacitors of the same value everywhere and concentrate on getting the 
lowest possible parasitic inductance between power and ground.  The 
parasitic inductance of a ceramic chip capacitor depends largely on the 
physical dimensions of the chip and isn't much affected by the capacitance 
value.

None of this really has much bearing on decoupling the voltage reference of 
a microcontroller.  The original 100nF ceramic chip plus good circuit layout 
will be fine.

Graham.