> It seems to me that any digital IC had better make the positive rail an
> effective AC return, otherwise digital switching noise (which is
> responsible for as much if not more current than the oscillator itself AND
> has more harmonics,since the oscillator is a good sine approximation,as
> opposed to the square,fast edges of the digital switching) will radiate
> very well.

That would be nice, but it is usually less than perfect in practice. Given
that the osc caps carry what is likely the highest sustained frequency on
the board, I would suggest that they are rather critical.  Also, experience
with systems failed in EMI testing (not my design) bears me out.


> ALSO, without external components, I think Vss looks the same to AC as
Vdd.

Agreed, but you can't run the chip that way :) (If a signal radiates in the
forest...)

> Internally on the chip, they are pretty symetrical (only asymetry for a
> pure CMOS digital IC should the the difference between PMOS and NMOS,which
> is pretty small). It is the fact that we always guard the integrity of
> ground very well by using ground planes,star points,and short runs,etc.
> that makes GND usually look different to AC than the positive supply. IF
> you were to use an equally large power plane, etc. OR have bypass caps
> located very close to the IC, I think Vss and Vdd look almost identical to
AC.

Well, if you can afford an the trackage to make them equal, then I suspect
you could return the caps to the ground pin for less cost.

While I'll agree that this is theoretically possible, and that a system
constructed this way will function (ie the osc will run), it will always be
more difficult to achieve the same low EMI as a system returning the osc
caps to the uP ground pin. (Ignoring any other EMI in the system, all other
things being equal)