Intersecting plane on sphere for diameter

[[Do...]]

Hi guys, I am working on a part that requires the diameter of a flat surface machined off of the top of a sphere. When I create the intersection of the two features I am seeing a result that is ~.040" larger than what the operator and I approximately measure via calipers. If I adjust the points on the sphere to be more closely positioned near the machined flat, I get a much more favorable result. Should I be calculating this dimension in a different way, and what could be the contributing causes to the variation in size I am seeing between strategies?

Additional info:
The sphere is ~ 19.00" in diameter and has a form of ~.004" via Strategy One
Strategy One: Twelve levels of circle paths at four points to measure larger overall area
Strategy Two: Five levels of circle paths at four points near the flat section
Should I be looking into different evaluation methods for the features? (ie LSQ, Min Circ, Max Inscr) or are those generally used more for hole and cylindrical features?

Thanks in advance!

[[Ow...]]

I'm assuming you're taking points on top of the flat and using that feature in an intersection with the sphere to get the diameter.
Are you using the flat plane at top in the alignment for planar rotation? If not, it could be skewed and cause different results.

Do the coordinates of the sphere location change between the two methods?
For example, If sphere was x,y,z zero with one method, it should also be the same with the 2nd method measured.
If they're not, then the probing is bad, maybe the probe is shanking?

[[Do...]]

Hi Owen,
Correct, although the planar rotation is established off another feature which is parallel to the top within.0002". I can try adjusting that alignment to see if anything changes. I'll double check the sphere locations between strategies and see if there is a discrepancy on that end as well since it doesn't look like the probe is shanking anywhere. I'll report back with what I find out. Thanks for the response!

[[Ow...]]

Measuring the diameter of a flat on a sphere with a pair of calipers can be an optical illusion and repeatability between 3 or more people would probably not be good.
Also if you are having to use single points (scanning is better if you have a scanning head) , it's better to choose the circle path strategy ( you probably are) and setting the start height/target/number of sections and clicking single points (if you have to) than just developing the sphere with feature recognition by taking points on the sphere until it recognizes it because, the circle path settings will ensure each vector angle of each point is correct. If you have to take single points, you'd probably get better repeatability taking at least 8 single points on each section instead of four.
Another thing to think about is probe and fixture rigidness and probing pressure. What's holding the sphere, I know it's large so it shouldn't be an issue unless it's setting on a small area, magnet or v-block and it's moving.
Also and maybe most important, is when defining the sphere location and size, make sure you're covering at least 180° of the sphere, anything less would require evaluation constraints of the location and then the radius. much like measuring a small radius.

[[Ma...]]

It might be interesting to scan over the intersections of the sphere and flat and evalulate the kink points.

[[Do...]]

I'll try bumping the circle path points up to 8 from 4 and see if that helps. Fixturing isn't much of a concern because I failed to clarify that this is a large *hemisphere* so it sits flat on the machine. I am currently taking single points as scanning is not an ideal option for this material and size. That being said, I've not used kink points before. It looks like I may need the curve add-on to do this (which I don't have) can someone confirm? Hypothetically, would I be able to get away with minimal scanning along the top/flat area to achieve a result using that method or would I still be looking at measuring up to 180º of the sphere?

[[Ro...]]

Adding my 2 cents...

More points is always better.

If you're going to use single points on each level, stagger them on each level so that you're not just measuring the quadrants/same sections of the sphere. For example, if your using 4 points at the quadrants, rotate the next level 45 degrees. You'll capture more of the form of the sphere.

You need to measure more than 180 degrees of the sphere if you're wanting accurate data in the Z axis. At only 180 degrees, being the top half of the sphere only, you're not getting any opposing points in the Z axis. The fitting algorithms need opposing points to fit accurately.

Take points as close as possible to the edge of each area you want to intersect with each other.

[[Ma...]]

It had been a while since I used kink point. Sorry, that is not what you want. Perhaps just use some circle/line intersections from short scans as a reality check?

[[Da...]]

the "sphere" is probably more of a torus... assuming the flat is on top of the part. Try checking a circle perpendicular to the flat (x axis). run from the flat to the equator. (Less than 90 degrees) Use single points, 10 or more. Compare the radius of the new circle with the reported diameter of the sphere. If it is greatly different, then your sphere is not a sphere. Using the sphere feature will force a result that is sometimes a gross average of the points taken. (You can put a couple points on a cube and if you recall them into a sphere...IT Will give you a SPHERE ! ) The handling of 3D features is pretty complex.