>Double-precision numbers are stored internally in a binary format, and with the format of a number in scientific notation; one bit for the sign, so many bits for the exponent (base 2), so many bits for the mantissa (base 2). There are 15 significant digits, but there can be some rounding errors in the last digits.

Thanks

>I have no personal experience with this, but I believe the rounding errors will cause you trouble, but only occasionaly.

Iguess I won't, not soon, be using for scientific app, just business apps, where mostly it will be for value and qty. calcs.

Just a passing thought. Is the precision we define for a double in the table, say 3, can we change this programatically to say 6?!? <g>