The way in which Roundness has always been defined is that it applies to the line element created in any plane perpendicular to the axis of the feature.
Thus roundness is measured perpendicular to the axis of the feature, and not perpendicular to the surface of the feature. This is inconsistent with other geometrical tolerances (e.g. run-out, profile, etc) which are all measured perpendicular to the surface.
While this works out just fine for cylinders (where perpendicular to the axis is pretty much the same as perpendicular to the surface), it does not always work out well for cones. The greater the angle of the cone, the greater the cosine error in any measurement.
For these reasons, a proposal was put forward in the ISO technical committee TC213, several years ago, to change the definition of roundness, so that it would be defined as a tolerance which applied perpendicular to the surface.
ADVANTAGES:-
- it makes more sense from a design point of view, when defining cones, as it is more likely to correspond to the type of variation which affects the functionality of the part.
- it makes more sense from a measurement point of view, as it will reduce measurement uncertainty.
- it is more consistent with the definitions of other geometrical tolerances.
DISADVANTAGES:-
- it is a change
- it introduces another difference between ISO and ASME
- the transition from the old definition to the new would require managing in industry (changes to working practice, changes to measurement procedures, changes to software, etc).
ISO have flip-flopped over this for a while. At first the change was going to be introduced, then it wasn't. Every time it was discussed, it caused a great deal of debate. Eventually, a new definition for roundness was to be introduced in the 2012 revision of ISO 1101, but at the last minute there was another change of heart.
At the time of writing, the plan is to leave the definition of Roundness unchanged. As a way around this, an alternative proposal has been put forward, which is actually a rather elegant way of tackling the issue.
The idea is that if you need to control the 'roundness' in a direction normal to the surface, you can specify this by using a Circular Run-out tolerance, with the feature used as its own datum. Although not an 'illegal' drawing indication under current rules, most experts in this subject would normally discourage this kind of practice, as it creates a 'circular reference'. With just about any other tolerance characteristic (position, coaxiality, parallelism, etc), this would create a meaningless and uncheckable requirement.
Why is it different with run-out? Run-out tolerances have always covered a combination of geometrical characteristics. Applying Circular Run-out to a cylinder controls the combined effect of any roundness errors, and any coaxiality errors. By using the cylindrical feature as its own datum, the coaxiality factors are eliminated, leaving you with a roundness tolerance. Apply the same tolerance arrangement to a cone, and you have a kind of roundness, but applied normal to the surface of the cone.
Some kind of illustration covering this is likely to find its way into a future edition of ISO 1101 (unless people change their minds again).
