I don't have an image to post but the description is simple: I have a cuboid part (a block) with dimensions shown against XYZ that match the cuboid planes. The DRF, alternatively, is the top plane (+Z = Datum A), a hole on the top plane (Origin = Datum B) and another hole on the top plane (Rotation = Datum C). The holes are not aligned with the XYZ (Datum B is on one side of the top surface and Datum C on the other). The angle between them with respect to the XYZ of the cuboid dimensionals is nominally 17.6558 degrees.
So, to measure True Positions of other holes against A/B/C, the DRF is rotated that number of degrees in relation to the dimensions to the FCF of the holes to be checked.
Am I correct in surmising that I should rotate the X/Y dimensional components of the measured holes to be checked to match the angle between the rectilinear orthonormal CSY of the cuboid dimensions and the DRF CSY?
Sorry if this seems fundamental but every single example online is the most simple alignment of DRF with no such misalignment between the dimensional alignment and datum alignment.
Thanks!
So, to measure True Positions of other holes against A/B/C, the DRF is rotated that number of degrees in relation to the dimensions to the FCF of the holes to be checked.
Am I correct in surmising that I should rotate the X/Y dimensional components of the measured holes to be checked to match the angle between the rectilinear orthonormal CSY of the cuboid dimensions and the DRF CSY?
Sorry if this seems fundamental but every single example online is the most simple alignment of DRF with no such misalignment between the dimensional alignment and datum alignment.
Thanks!
Comment