Constructed Feature Repeatability

[[DW...]]

I am scanning (4) 3d Curves (1000 points) 90 degrees from each other perpendicular to a gothic arch groove. I recall feature points of each 3d curve into a circle. From those constructed circles, I recall all 4 into another circle for a runout callout. My repeatability (all filters applied on the curves, circles, outliers per cookbook) is sometimes very good (0.0005mm) and sometimes not so repeatable (2-3 microns) depending on the part. The program, fixture, and environment are all solid. Are constructed features some times less repeatable than non-constructed features?

Contura 7/7/6 Vast XTR Calypso 2019 Curve and FFS

[[Ri...]]

I'm not certain why you are measuring Curves then recalling those feature points into a Circle to then recall those feature points into another Circle. .

Have you just tried to measure the final circle as one feature?

[[DW...]]

Yes. If I scan the groove center, I get very close results to the current strategy. The reason for the 3d curves is that I also need recall the curves into Circle in Contour Best Fit features for a distance call out (circle to circle, think bearings riding in a gothic groove).

I guess if you ignore what I am trying to do completely, the more important question is about constructed feature repeatability. I would imagine just by the nature of the algorithms taking into account thousands of points for calculations that it would be slightly less repeatable than non constructed features.

[[Ow...]]

Scanning curves and getting less 2 to 3 micron deviation/repeatability can be difficult (most CMM only guarantee 1.5 to 2 micron linear and circular probing accuracy) and the conversation could get lengthy.
To start with, have you tried repeatability taking single points? Fewer points for sure but, repeatability is generally better as the filters don’t play a role. If single points produce better, it’s usually a filter/scanning speed issue
Have you tried 2D vs 3D curve?
If possible, are you applying Pre/post travel strategy that allows the (for lack of full description) probe to settle into the correct vector before recoding data?
Scanning Speed, have you tried reducing the scanning speed?
Have you tried setting a secondary alignment (predetermined by previous probing) for each curve feature to ensure it starts in the exact same location?

[[Je...]]

Please sign in to view this quote.

.
I haven't seen evidence that constructed features are inherently less repeatable due to computations. I say this because I remember wondering as much in my first few years of using Calypso. My own tests led to a high level of confidence in Calypso's constructions. Usually, unexpected outcomes could be traced back to how I told Calypso to define the recalled feature.

Also, I use similar strategies to what you mentioned about capturing points in a curve and then recalling those points to other features. I use the "add range limits" option in point recall. This way, probed features can serve as point containers for other nested features of interest. This has worked well and has several benefits.

.
.
.