Let me try this one step at a time.
Originally from pminmo:
QUOTE: Now try to make a part that has geometry defined to .0005, that perfect repeatability doesn't do you an ounce of good because the machine lacks the precision to do it. UNQUOTE.
Sorry, but the machine you described lacks the RESOLUTION to mill to 0.0005. As good as it gets is defined by the limitations you built in: the stepper and the drive. You can only make assumptions regarding the machine's REPEATABILITY -- until you actually make the part, many parts, and see how close they are to each other. it won't be worth a hoot if its not the size you need for the job or customer - and a lot of scrap lying around too. Milling within our limitations or making adjustments is what we'd be forced to do with such a machine. But clearly that machine lacks the RESOLUTION to do the requested job. If all those scrap parts are the same dimensions, then we'd have good repeatability, within our system resolution - but only when compared to the desired output can we make statements regarding precision and accuracy.
ACCURACY is the ability to get the expected result. To hit what you aim at, to measure the gap with as many significant digits as the ruler has, to set or measure the resistance to the --CORRECT - expected value. Any error, deviation, scatter, bias, or miss of the measured result from the EXPECTED (read - correct) value describes the precision. NOT the number of digits in the meter. More digits increase the ability to make greater descrimination in our measurement. Measuring to 8 significant digits often implies PRECISION (and sometimes ACCURACY). But the additional digits provide only some confidence - if we've calibrate as Phil suggests (thats what we always do righto?).