Matt Bonner wrote:

> The reason the crystal capacitors can be tied to the 5V rail is
> because they need to be tied to _any_ low impedance point.

  My point was that in discrete/ single-ended transistor design, a
common ground point is used, with all bypassing directly to it.  Since
the power rail has the load currents impressed on it, bypassing signal
points to it rather than direct to ground is likely to introduce
feedback and instability even though the power rail is itself bypassed
to ground.

  Aside: In the earliest days, condensors{sic.} were expensive, and
bypass impedances were only *low enough* to do the job!

  CMOS on the other hand is complementary, using common-source (equiv.
to common-emitter) circuitry with one source of a pair connected to
*each* supply rail.  Its behaviour is such that the threshold varies
proportionately to the rail voltage, compensating for output
fluctutations with rail voltage.  Of course, operating mainly in a
digital mode, this should not affect any part other than the oscillator
for which it can be stated that to whatever extent that supply rail
variations do affect the oscillator, it will be an identical effect for
bypassing to either rail.

  It is of course obvious that ADCs where provided, reference to the
negative rail and bypassing must be fitted accordingly.

  Cheers,
        Paul B.