Cylinder Alignment troubles

[[Gr...]]

I am having some difficulty establishing the best alignment for the part sketched in the attachment and was hoping I could get some advise.
The primary datums are the OD's at 4 places. ID's are not relevant and cannot be used. The ends of the tube are very thin and not very good for rotation because it is a small area and not very flat.
My first trial worked but I feel uncertain about it. I used the top thin surface as my rotation and then a 3D line from C to D for my planer (X-)
I used a 3D line from A to B for my X and Y and I also tried using the C to D 3d line for Y at one point. Z origin is the 3D line form C to D.
All of the iterations worked, but with various results. Any ideas on the best approach would be greatly appreciated.
to put this in perspective the OD's are about 21mm Vertical is 120mm and horizontal is 80. The part sits in the CMM as shown.

Scanned from a Xerox Multifunction Printer (00C).pdf

[[Ow...]]

Using the thin face of Datum A for "rotation in space" will not work because it's too small and any deformation will project error throughout the length.
Use the 3D line from A to B as the Rotation in space datum and the X,Y, and then use C to D as planar rotation.

[[Ow...]]

General practices for Rotation in space datum (used to be called Spatial in older versions of calypso):

The print datums and tolerances and probe access may force otherwise but generally, for stability, you want to use the largest / longest surface for your Rotation in Space axis or you will end up causing projection errors.

Using a shaft as example, if you took a 1.0 inch dia rod that you knew was straight, and set the spatial axis off the end face of the rod, and the origins off the diameter by the face, when you measure the rod 2" or 3" or even more down the length of the rod, it will show it not being straight, even though you know it is.

Another example for planes/faces is if you used small planes for Rotation in Space then you may get location errors when projecting the features farther away than the size of the plane. i.e. if your "Rotation in Space" surface is 3x3" wide, then you wouldn't want to trust the location accuracy of features farther than 3" away from that surface

[[SH...]]

Please sign in to view this quote.

Looping the BA can't eliminate this error?. it is not possible to work with big feature for spatial rotation for all time.

[[Gr...]]

Thank you Owen. I think that was one of the scenarios I tried but that was before I put the loop in 4x.
I see your reasoning and it seems sound. The part is plastic and so I'm fighting that as well. The Looping, though time consuming does help considerably.
Thanks again, I'll keep you posted.