hexagon logo

How to properly measure circular runout on a cylinder?

I have a bore that is Datum B. It has a perpendicularity callout to Datum A, which is a flat surface. So, I have to measure it as a cylinder. I have another bore that has a circular runout callout to AB. So do I have to measure B again as a circle? Because I would have to assign that as another datum to build the feature control frame. But if I just use B as a cylinder, then wouldn't it be measuring total runout and give me a different result than circular runout?
Parents

  • Maybe I missed it, or didn't understand it. I have a FCF with three requirements in it. It calls position of a bore to two datums. It then calls perpendicularity of the same bore to the same plane being used in the positional callout, to 0.08mm. Then it calls runout of that same bore to the same plane to 0.05mm. Is this something that came out of the office of the redundancy office, or do I just not understand it?

    I'm using 2022.2 and attempted to dimension this using Geometric Tolerancing, but it wouldn't allow it, so I went back to Legacy, which is my preference anyway. It "seems" they attempted to make dimensioning easier, or more compliant might be a better word, but there appears to be a lot of stipulations in using it. By that I mean it will only work properly, "if this is that..., and that other thing isn't this type of thing..., with the exception being..., but only if this condition is true...etc". Seems too convoluted to use effectively. Of course it may just be because I'm too used to Legacy. Right or wrong, my reasoning is that I've used Legacy for nearly 20 years. It's dimensioned hundreds of parts that have passed PPAP and are still being used today, so it can't be "wrong", just different.

    In addition to that, I am kind of gun shy after Hexagon came out and said their math was wrong in Xact Measure. At least that's how app support has described the reasoning for the upgrade to Geometric Tolerancing, to me. Does anyone else share that same aversion to Geometric Tolerancing over Legacy?

    Thanks in advance for your comments.​



    , Are you able to share the drawing (or a sketch) showing the datums, the feature and the callout's you are tying to verify? When you say the Geometric Tolerance command "wouldn't allow it", what wouldn't it allow? What error message(s) do you see and for which callout? From your description, I suspect it's this...

    Then it calls runout of that same bore to the same plane to 0.05mm


    The datum reference frame for a runout tolerance needs to define an axis of revolution because runout analyses radial deviation from that axis. If you only had a primary datum plane, there would be no centre of rotation which is why the Geometric Tolerance command would not allow it. Although legacy won't give you any error messages, you would still have problems with the results. Strictly speaking, you would need to recall the initial, machine coordinate alignment (usually called "STARTUP") and then level to you datum plane. Although the plane would set the orientation, your feature would be nowhere near the origin and your results would be huge. The problem with legacy is that it is based on the active alignment and alignments inherit certain information from previous alignments. This can sometimes mask mistakes. For example, in this case, I suspect you have recalled whatever alignment was currently active in your routine (one which was already levelled, rotated and origined to features and more-or-less centred on the cylinder you are reporting) and then levelled to the datum plane before reporting runout. Alternatively, you may have selected the "use datums" checkbox and picked the plane as the datum but this would still be creating a new alignment in the background in the same way I just described. This means that you have actually, without realising, added a secondary datum, not shown on your drawing - you have origined to whichever feature was used in your prior alignment. Although this may give "good" results, it does not mean that they are correct. How do you know that what you have done is what the designer intended?
Reply

  • Maybe I missed it, or didn't understand it. I have a FCF with three requirements in it. It calls position of a bore to two datums. It then calls perpendicularity of the same bore to the same plane being used in the positional callout, to 0.08mm. Then it calls runout of that same bore to the same plane to 0.05mm. Is this something that came out of the office of the redundancy office, or do I just not understand it?

    I'm using 2022.2 and attempted to dimension this using Geometric Tolerancing, but it wouldn't allow it, so I went back to Legacy, which is my preference anyway. It "seems" they attempted to make dimensioning easier, or more compliant might be a better word, but there appears to be a lot of stipulations in using it. By that I mean it will only work properly, "if this is that..., and that other thing isn't this type of thing..., with the exception being..., but only if this condition is true...etc". Seems too convoluted to use effectively. Of course it may just be because I'm too used to Legacy. Right or wrong, my reasoning is that I've used Legacy for nearly 20 years. It's dimensioned hundreds of parts that have passed PPAP and are still being used today, so it can't be "wrong", just different.

    In addition to that, I am kind of gun shy after Hexagon came out and said their math was wrong in Xact Measure. At least that's how app support has described the reasoning for the upgrade to Geometric Tolerancing, to me. Does anyone else share that same aversion to Geometric Tolerancing over Legacy?

    Thanks in advance for your comments.​



    , Are you able to share the drawing (or a sketch) showing the datums, the feature and the callout's you are tying to verify? When you say the Geometric Tolerance command "wouldn't allow it", what wouldn't it allow? What error message(s) do you see and for which callout? From your description, I suspect it's this...

    Then it calls runout of that same bore to the same plane to 0.05mm


    The datum reference frame for a runout tolerance needs to define an axis of revolution because runout analyses radial deviation from that axis. If you only had a primary datum plane, there would be no centre of rotation which is why the Geometric Tolerance command would not allow it. Although legacy won't give you any error messages, you would still have problems with the results. Strictly speaking, you would need to recall the initial, machine coordinate alignment (usually called "STARTUP") and then level to you datum plane. Although the plane would set the orientation, your feature would be nowhere near the origin and your results would be huge. The problem with legacy is that it is based on the active alignment and alignments inherit certain information from previous alignments. This can sometimes mask mistakes. For example, in this case, I suspect you have recalled whatever alignment was currently active in your routine (one which was already levelled, rotated and origined to features and more-or-less centred on the cylinder you are reporting) and then levelled to the datum plane before reporting runout. Alternatively, you may have selected the "use datums" checkbox and picked the plane as the datum but this would still be creating a new alignment in the background in the same way I just described. This means that you have actually, without realising, added a secondary datum, not shown on your drawing - you have origined to whichever feature was used in your prior alignment. Although this may give "good" results, it does not mean that they are correct. How do you know that what you have done is what the designer intended?
Children
No Data