I usually think of decoupling caps in relation to digital devices. Can you relate wavelength to the rise time of a digital signal? That will help in understanding both the value of a decoupling cap and the distance from the IC that needs power supplied by the cap.
@Sanjib, that's the general thinking, redce loops wherever possible and you reduce EMI emissions and susceptibility problems. There's also the inductance of the capacitor itself, depending on the frequencies in your circuit.
Is there an advantage of using smaller decoupling capacitor packages e.g. 0402 over the bigger packages such as 0805 from signal/power integrity, EMI stand point? Obviously the important electrical parameters such as voltage, capacitance shall meet requirement first. But I was thinking if smaller sizes would reducing the loop & inductance even further, and hence should be the first priority from EMI point of view, apart from the fact that these take lesser space.
I try to put pairs of via's either side of the pad where possible, this lowers the effective loop area and makes the "inductor width" wider, both of which reduce inductance.
And I use ferrite beads wherever possible, a good place for these is on rapidly switching outputs (particularly capacitive loaded ones) this not only prevents ringing , but rounds off the edegs of the circulating currents through VCC and gnd.
Any thoughts on the effectiveness of a few ohms resistance between each power pin and its Vdd plane? This does force the PCB designer into placing a cap close to each pin, but does this confining of the shoot-through current loop to the local capacitor work to keep the planes quieter and reduce radiated EMI?