I have to apologize for a silly mistake in my post of Sunday.  In my
eagerness to give a simple example, I divided an odd number by two and
got an extra digit of resolution :-).  What I got, of course, was an
extra _bit_ of resolution -- you have to average 10 readings to get
another decimal digit, and I believe the _useful_ resolution scales as
the log(?) of the number of readings. Or is it the square root?

It's been too long since I read all those papers.


john perry
jperry@norfolk.infi.net