Weird Concentric circles TP result

Is the plane perpendicular to the diameters?

Tried concenctricity?

Is the plane perpendicular to the diameters?

Tried concenctricity?

Yes it is as basic as it can get. The only complex thing is that the space to probe is limited so I am taking circles instead of cylinders.

Hmm, no I did not try :S I will try testing some more today and if I face the same issue, I will post code and drawing to investigate more. It is really weird to me and correct me if I am wrong, just a leveling and a XY translation as far as alignment is concerned, should be enough for something simple as that, even without the rotation restricted.

In this case, you could just make a fine line with a marker, and take a point on it to rotate the alignment. It's not accurate, but it could help understanding the problem...

In this case, you could just make a fine line with a marker, and take a point on it to rotate the alignment. It's not accurate, but it could help understanding the problem...

Should I though? Logic says that no matter the rotation, the distance of the centers remains the same. For example either 0.03 in Y or 0.03 in X deviations (and everything in between while rotating) should result in a 0.06mm TP.

I am not sure why the machine is failing on this point and if logic prevails and the rotation is not the real issue, then only the surface quality and distortion under stress remains as an explanation. Having it placed in A position, the machine projects 2 circles on the leveling plane with 0 axis deviation and rotating the part (Z axis rotation) 180 degs in position B, causes the circles to be projected with offset due to the "bended" plane. Not really sure though since flatness is not that high at 0.02mm

Possibly look at the roundness of your circles, see if there's a form error there that's part of the problem.

Possibly look at the roundness of your circles, see if there's a form error there that's part of the problem.

It was one of my initial concerns yes. If I am not mistaken, the Datum A is around 5-6 microns while the inner is around 12 microns no matter the placement. Not so high imo.

Maybe take a look at fixturing? Is it consistent? Is there is a wall thickness issue around the part?

Test out the leveling factor by taking hits on the granite / base plate rather than on the part. Also do what Jeff did and "simulate" a rotational clocking somehow.

Maybe take a look at fixturing? Is it consistent? Is there is a wall thickness issue around the part?

Test out the leveling factor by taking hits on the granite / base plate rather than on the part. Also do what Jeff did and "simulate" a rotational clocking somehow.

The rotation is locked since I am recalling an exterior alignment. The 3axis figure is not floating / moving when I am creating new alignments so I guess even if it not needed, the rotation is restricted indeed. The only rotational element eitherway is the fixture, maybe I will be forced to probe it each time and make a full 6 dof restriction alignment as you guys suggest

The only inconsistency is the surface of the plane that has limited area to be probed. Otherwise the part is griped in a 3 point vice gently enough imo. Well, I am off for more testing! I will get back to the topic once I have more data

UPDATE:

So I did more testing... The offset is around 0.015-0.020mm in TP result (meaning around 7-10 microns of radial distance between the centers). I tried everything and I am led to the conclusion that the part itself is to be blamed. I changed probes, calibrated everything, used a full 6 dof alignment and even probed the inner circle as cylinder eventhough it had 1.5mm of length (it was 5 microns perp to the plane I lvled). I made sure all auto features are having perfect round vectors and finally changed the probing points to 27 and 23 (overkill for circles less than 10mm I know).

If I trust the cmm and accept that the part is somehow bended it means that "at the same Z" (Z dependent of the level/translation plane which changes slightly between part rotation due to distortion) the circle I am probing is shifted almost 10 microns radially. The form error is even changing depending the rotation from 13microns best case to 20 microns worst case.

I cannot explain it otherwise, I have to accept that the part has such a bad quality hole that micro Z difference in the same theoretical Z slice, ends up with a 0.01mm offset.

There's a lot going on here.
I would look into a few fundamental thoughts before red-tagging it, especially if it's a sample set of a larger lot or batch.
1: Can you affirm the OOS condition using another means? I know 0.0008" / 0.020mm is too tight to determine with a caliper, but i'm sure if you throw it on a surface plate with a v-block and indicate the ID while spinning the part, you'd be able to affirm concentricity to the OD with ease.
Note: if both cylinders are tri-lobed this could potentially be invisible in a v-block, so look at your form error plots of the circles for patterns/lobing.

2: ​How is it manufactured?
3: What kind of screwup during manufacture, could have caused this condition?
4: How could it have been manufactured and brought to you as presumably acceptable?

Another possible contributor is your probe itself. Since it's such a small diameter, I'm guessing you are using a smaller probe? 1mm? the Probe diameter-to-shank clearance for a 1mm probe is miniscule (the 1mm probe with TC shank gives you just 0.010" between the shank and the edge of the sphere, when it's made perfectly) and something like a glob of glue at the sphere, or a slightly bent tip, no matter the calibration results of that probe can interact when you start probing at a depth into the part.

If you don't assign a rotational datum, the demon simply refers to the rotation as being square to the machine's volume, about your local level plane. The floating tetrahedron animation in the graphic view is a tool to make sure you clearly isolate all 6DOF to the part's constraints. If there's no rotational datum on the print, technically, you don't need to establish one in the alignment, unless you are trying to lock down a possible issue with the part or method.