At first, let's shoot down Microsoft Excel :)


---- Microsoft Excel ----



It's suprised that 1.77636e-15 is large. Besides, 1.9984e-15 at C7 is not multiple of 1.77636e-15. And there is no criterion about 2^-49.
Under epsilon = 1.6e-15 results is zero, and so 2^-49 seems to be the epsilon. But, considering differences of C-columns are 2^-52, thet's none other than the epsilon??? ... I wonder how the software calculates in inner handling.

By searching about epsilon in the Microsoft Knowledge Base Article, an enigmatic article was found as follow: Microsoft Knowledge Base Article 214118. Simply to say, according to that article, the epsilon is comforming to IEEE(754) but the expression is sometimes dissimilar. Actually, do calculate formulas in the article, as follow:
P.S. There is no explanation of its algorithm.


By nature, it shold be expressed -2.77556e-17 at A4 as same as B4. It seems to reluctant to be known the error of computing and handling in ad hoc. That's characteristic oftypical of the Microsoft. To put it plainly,


It's just a household accounts software. Its inner manupulating is so irresponsibility. But there are many macro libralies of calculating for the software... I think consumer have to be more smart and also think the Microsoft is overmatch.
Whatever Bill says anything, that's the expression to be economic supremacism even if that destroys the culture of computing.

Above results are same on Excel 97 and 2000 in my hand, and accoding the article Excel 2002 outputs as same results.
On my mind, I wish to know results of Lotus and OpenOffice in particular. But both of them are not installed into my computers.


---- Psion Sheet ----





That means it's available on scientific calculating.