You made the statement:
> The power/ground planes are your best high frequency bypass
> capacitor
This isn't always true. Power and ground planes are, in fact,
inductive at frequencies between the first "system" resonant frequency
and the first "modal" resonant frequency. The first modal resonant
frequency is at f=1/(2*L*sqrt(u*e)) where L is the largest (x or y)
dimension of the power plane, and u and e are the permeability and
permittivity of the dielectric, respectively. The system resonant
frequencies are more difficult to locate, but they can be computed by
summing a wide range of modal impedances.
As long as "high frequency" is lower than than the power/ground
system's resonant frequencies, then L*di/dt and all your basic circuit
assumptions will work. Once you get above that, then the modeling
problems and solutions get a lot harder. Just for example, when our
group built several 450 MHz digital signal processor prototypes in the
early 1980s, it was necessary to include decoupling capacitors on top
of (and in some cases even inside of) the IC packages.
-- Bob Techentin techentin.robert@mayo.edu Mayo Foundation http://www.mayo.edu/sppdg/ Rochester MN, 55905 USA (507) 284-2702