What are the uncertain digits of a 1/64" calibrated ruler?

[potato]

What are the uncertain digits of a 1/64" calibrated ruler?

There are two schools, I think:

1) person measuring can estimate between markings. Since 1/64"=.015625, and as a rule of thumb (?) the error on a ruler is half of the markings is .0078125, the 3 digit is uncertain. So, a reading of 1/64" would equal .016".

2) Since 1/64"=.015625", and the error is .0078125, then if you add and subtract the error to 1/64", you could get readings of .0078125 to .0234375. In this case, the first digit is certain, and the second is uncertain. So, a reading of 1/64"=.02".

So, which is it?

 

[Norml1]

Re: uncertain digits of a 1/64" ruler?

When recording observed values, the number of digits to be recorded are all exactly known digits plus one additional digit that can be reasonably estimated, otherwise the reading is recorded to the nearest graduation.

 

[normzone]

The measurement should be reported as follows:

Let's say the observed measurement is in between 7/64 and 8/64.

The measurement should be reported as 7/64 +

All else is interpolation and estimation.

 

[Norml1]

Interpolation and estimation should be fine when using a ruler, in my estimation! It certainly depends on how inspectors have been trained on the use of measuring equipment and recording of data.

 

[normzone]

I prefer to call them as I can defend them. I've met too many wannabe metrologists willing to demonstrate that they can hold a ruler better than anybody else, contact error and parallax error be damned.

 

[Hershal]

For line width as a Type B, I agree with potato. However, parallax has also been brought in, a Type B, plus the uncertainty from any previous calibration it may have had. Additionally, temp and RH must be considered as Type B contributions as rulers will change. For a metal ruler, one has the thermal coefficient of expansion/contraction for the metal. Thus, the contributions can get complex.

However, the key is whether the change is significant enough to see.

 

[bbrewington]

Shouldn't you do a Gage R&R study with some samples to determine the actual error inherent in the user-ruler-sample system? Maybe you could estimate the error, but this would allow you to experimentally verify the error.