- I forgot it!
As @robertferanec said, the biger caps are intended to filter low frecuencies, medium caps medium frecuencies and small caps high frecuencies, this last caps (smaller caps for filtering HF) should be closer to the pin.
- Sometimes you will see that some reference designs include a lot of small equal capacitances, they can be intended for create a large capacitance with very low inductance and ESR or for serve as individual pin decoupling, you most observe if the number of small caps equals the pin count, if yes probably they are intended to place near the pin, and the medium caps more away from the pin to act as bulk capacitance for the local plane area.
-If not, the best way it´s to act by "common-sense" analyce the network and try to find a logical approach.
- Agood example about all of this, it´s the OpenRex Ethernet Phy decoupling and placement scheme, try to observe the schematic, and the component placement, look how they placed small capacitances near the pins and medium caps to filter the local plane that provides power to the small caps.
Its imposible to imagine how electrons and holes will flow but acting by common sense all will be more or less right
I´m altera lover but this is one of the most complete guidelines i ever seen