Please do not separately post the same question to different newsgroups.

Oh, sorry for this. Never more I will do it.

As respondants in both threads have pointed out, this is a well known
property of finite precision arithmetic exacerbated by binary representation
of numbes. It is not a bug and is not unique to Excel.

Well, it's not a bug, but a "feature".
Where in the hell is this "feature" explicited at help system ? Not so
easy to find...

And of course, for a common user, a number is a number, and nothing
more. If it's typed at 2 decimal precision, it could be good that sum
and multiplycation of it could be also at those 2 single decimal
precision. No use of thinking about IEEE, 15 limit precision, binary
internal representation, and so on... Unless the user ask for it.

I tried the solutions and as trunc() worked well, I put it everywhere.
Now the calculations take some time (few miliseconds more). Also the
parens () works well, but I can not trust 100%.

Well, thanks anyway.

yours, Daniel
from Brazil