Having read the items to which you refer, Kai, I have to say I believe the assertions about the size of the larger of the cap's in each case, fail to consider the effect on Vcc rise time.

Using large-value (>4.7 uF) cap's is not a bad idea if there are only a few of them and if there are not so many of them that the Vcc risetime doesn't become objectionably long. Many people seem to think that adding capcitance will compensate for an underdesigned power supply.

In recent months, Ive built a number of small 805x-based circuits that have short Vcc rise and fall times as power is switched on and off, and find them to be very well-behaved. By contrast, I've got at least two, the performance of which has settled down quite a bit subsequent to the removal of several hundred microfarads of electrolytic cap's.

I use the 220 pf + 2.2 uF combination for LSI's such as MCU's and even more for CPLD's and FPGA's with rapidly switching synchronous elements. I do this because I've been happy with the result for several decades. I try to keep the lead length down to about 1mm. The use of ferrite cores in the Vcc distribution chain is a really good idea, as it mitigates the tendency of one circuit segment to impact the supply noise at another one. Imagination and use of accumulated knowledge about the components we use is a significant factor in how we design our circuits. It's certainly clear, however, that Vcc-Gnd capacitance will need to be different for a small (3-4 IC's) circuit, as opposed to a large (>12 IC's) one.

There's a mix of cap's that is "correct" for any given circuit, under whatever conditions it operates, and there are others that are only "close." Some, by extension, are quite wrong, though they may still "seem" to work fine. A lot depends on what, quantitatively and qualitatively, one expects those cap's to do.

RE