I have ask that same question and was never able to get a response so in looking for the answer myself took me to AWS B4.0M:2000 (R2010).Figure A4 – Bend Test Nomograph on page 8
note 2 Required accuracy of measurements is as follows:
(1) Specimen thickness: + 0.5mm
(2) Elongation: + 1percent
(3) Bend radius: + 1.6mm
I don't know how applicable this may be for your situation since B4.0 isn't referenced in D1.1:2010. this is what I have been using since 2001 and no one has ever challenged or corrected me on it.
You will notice that some cuts are indicated to be made 1/8" oversized and then machined, ground, sanded down to the required dimension.
When it comes to thickness, at one point you are sent to a section of Clause 5 (5.24.3.1) that indicates that base material thickness cannot be removed over 1/32". So when grinding, sanding, or machining down the face and root you best not take off too much material.
My personal two tin pennies opinion, when the codes give you a dimension you should be within 1/32" and preferably on the positive side. Undersized welds, unacceptable. Material too thin, unacceptable. Standard practice if a size indicated by engineer is not available, next size LARGER (get it approved of course). Machinists are generally within single digit thousandths of an inch, not 30 thou. Taking more material out suggests trying to get rid of the discontinuity. It is a welding test, not a grinding test.
Now, end of personal opinion, what does the code say beyond 5.24.3.1? Look at note 3 on Figure 4.9. Look at note 3 on Figure 4.10. And 4.11. Maybe it's just me, but I would think that "all dimensions are minimum." (emphasis mine)
He Is In Control, Have a Great Day, Brent
Page xvi in AWS D1.1-2010 refer to AWS B4.0 by stating ... "AWS B4.0, Standard Methods for Mechanical Testing of Welds, provides additional details of test specimen preparation and details of test fixture construction.
Have a great day.
Gerald Austin