API 5CT Rounding error

[avdbos]

Hi,

can anyone help/explain we why in API 5CT sometimes figure like 1.235 are rounded up to 1.24 and in other cases rounded down to 1.23 (which to ISO in incorrect)?

I am talking espically from the Table C.1 page 82 (please see my attached PDF file).
Here the Outside Diameter of 13 3/8 inch in mm is rounded to 339.72 but the entire figure is 339.725 (so correctly 339.73mm)

The same matter with 2 3/8 and 2 7/8 on table C.3.

Other Outside diameteres of sizes like 9 5/8 are rounded correctly: 244.475mm --> in API 244.48mm

Can anyone explain these differences?


Regards Arie

 

[DraftingMan]

avdbos,

You need to read Annex G, page 263 to discover to rounding rules in API 5CT/ISO 11960.

G.2.1 Rounding

The last retained digit in a number was unchanged when the next digit was less than 5 or raised when it was greater than 5.

When the digit following the last retained digit was exactly 5 followed by all zeros, the last retained digit was unchanged if it was even, or was raised if it was odd.

ISO has a preference for even numbers as the last digit when rounding. I have had to deal with this issue at least once a week for the last year or so. As more of these specs are aligned with ISO more a more folks will have the same question.



John H. Dunten, CD
Certified Drafter

 

[redpicker]

DraftingMan is correct. This rounding method is actually an ISO requirement and is a consequence of the ISO adoption of the API specifications.

It is considered to be "better" since in the case where the digit following the retained digit is a 5 followed by zeros, the retained digit is rounded up only half the time, the other half the time it is rounded down. The theory is, that in the long run, the errors introduced by this will cancel out. This may be true, but it sure is annoying that nobody at Micorsoft will add this capability to Excel.

rp

 

[avdbos]

Yes,

you are right excel would be in the need to support this but with a "if-than" clause this should be also possible, I guess.

Nonetheless thanks a lot for your replies. Was really helpful and a bit stupid from me not looking in the annex of API.

Thanks

BR