Mythical specifications come in many forms. One realm they seem to haunt is that of repetitive features, also known as patterns. Many attempts to shorthand pattern callouts are continuously made. Bad habits die hard as old mistakes are passed down from one generation of engineers to the next. One particularly bad habit is the use of linear dimensions with the term “NON-ACCUMULATIVE TOLERANCES”, or something similar. There is no such thing. Trying to use this shorthand leads to tolerance issues.
In the example above, the dimensional callout attempts to simply dimension a pattern without considering tolerance stack-up. However, this attempt fails since any two non-adjecent holes cannot avoid accumulation of tolerance due to the dimensioning scheme. Tolerance stack-ups on linear dimensions have accumulation. There’s no way to avoid it without dumping linear dimensions.
I had originally planned on a short article about this topic. However, once I started delving into it, I found out that there is a lot of ground to cover. So, this topic will be addressed in detail within a future article (Feb 23, 2011) where examples of different pattern dimensioning schemes will be explored.
Also see How to dimension feature patterns on drawings.
If I were to use a pattern callout like that, I would do so instead of using baseline or ordinate dimensions for every hole in the pattern, which themselves are an attempt to reduce tolerance stack-up between holes. I have only ever used callouts similar to that one in areas where I needed to save space on the drawing sheet.
Nathan, the example in the above image needs to be avoided, as it is impossible. There are effective tolerancing methods that do not take up a lot space on the drawing. You mentioned ordinate dimensions. This normally doesn’t take up a lot of room on a drawing, and is actually an achievable spec. But there’s another superior option I’ll talk about in the next article (tomorrow).
It has been my practice to put the “non-cum” note not on the hole to hole dimension but instead on an additional overall dimension from the first pattern hole to the last.this will force the holes into proper tol. over the entire pattern. Is that acceptable practice
Well, as noted above, there’s no such thing as non-cumulative. However, your method is similar to calling out a pattern, which kinda has the same issues (ambiguity on how to apply tolerances). I cover this in more detail on a later article: http://www.fcsuper.com/swblog/?p=2440
It’s perfectly OK to specify a non cumulative tolerance simply as a simple way of defining what you want with reference to the manufacturing process. If I say I want a +/-0.20 non cumulative tolerance on a hole pattern then it says that no two holes can be in error by more than 0.20mm. Nothing more, nothing less.
Eric, Without establishing common datums for the pattern, your callout is based on hope rather than a mathematically defined set of tolerances. If you analyze your described system, you are double/triple/quadruple (however many holes in the pattern) dimensioning. You are assuming there is a dimension from each hole to every other hole in the pattern. If you lay that out with shown dimensions to represent this for a linear pattern of just 5 holes, each hole will have 4 overdefining dimensions , for a total of about 10 or so linear dimensions (not including the repeats).
The proper way to achieve what you describe is to use Positional GD&T with established datums.