Simple ID meas using different methods

[PowerJunkie]

I'm measuring an ID... It's deep and my probe selection is limited. Everything is at A0B0.
I use a standard everyday probe 3mm x whatever and probe a circle. I trust this circle.
I then take all the hit targets and vectors from that circle and create one by one vector points then create a circle from those points.

The standard probed circle gives different (enough) results than the point to point / constructed circle to draw much concern and my question is why??? Has anyone run in to this? We're talking fifteen thousandths of an inch.

I have looked at the point to point hits and I see that it's out of round. If I construct the circle from the points leaving out the far off ones I get good results, but then why not with a standard circle with these same results? If I included the far off points the circle is way off.

Again, why the huge difference. Tried everything to get an average. Using LSBF. My boss says that it should average them and I agree.

Sound familiar to anyone?

I have one hour to figure this out. SIGH>

[PowerJunkie]

Update. I constructed a disk probe and my program runs 5 levels that it can reach at A0B0 and the results are that it's all good. The 1st level that the 3mm probe can reach the results of it and from the disk probe are the same.

It's the constructed points. The XY's and vectors are spot on. I don't get it. None of us do.

[JEFMAN]

Try BFRE...

[Schrocknroll]

^ This ^
If your probing isn't right on the target, this should help.

On a side note, instead of recreated all the points one by one with the coordinates and vectors, you can just create the first one. Then paste with pattern to create the rest.

[PowerJunkie]

The problem is like this... (full calibration for the single tip performed prior to starting. Results came out all .0001". Good to go). I probed an ID circle A0B0 it came out as expected. Based on the numbers (XYZ IJK) for each hit I generated vector points, for each. All theo perfect (verified). I took those points and generated circle. The results of the generated circle differed greatly from the standard probed in one shot circle. I set the constructed circle to LSBF, BFRE, and every other option it allows me and the data didn't change. Looking at the points taken, yes, some were out, but it should be able to calculate an average, it didn't appear to be doing so.
UPDATE from that... we decided to turn the controller off and then back on and reboot everything and I'll be ***** it seemed to have fixed the problem (w/o recal as noted above).
So now I guess my only question is why did this happen like this? AND IT'S IMPORTANT that I get some kind of info because my boss keeps asking the question as to why this happened and I have nothing!! Anything?

[JEFMAN]

Loosing the compensation ?

[PowerJunkie]

Loosing the compensation ?

Perhaps. ? Can you elaborate just a little?

[JEFMAN]

It depends on your cmm / controller.
Sometimes, the compensation matrix is in the PC, not in the controller.
If you loose the link to the matrix, then the cmm works in a wrong mode.
I'm not sure that it can be the issue, because I don't see why the cmm could measure a circle, from hits and not from the same points.

[PowerJunkie]

With "matrix" are you referring to the lower level matrix and in used with analog scanning heads? Sorry for all the questions, my boss knows his **** and if I talk any kind of BS at all he'll sniff it out quick!

[JEFMAN]

No, I talk about the compensation matrix.
A cmm is built from machined parts, so you can't have a cmm geometry allowing the measurement of µm !!!!!
To perform it, there's a compensation matrix which corrects some defects (linear, straightness of displacements, roll, pitch, yaw) + perpendicularuty between each axes.
It's often called the 21 prameters matrix.

[JEFMAN]

A little more about the matrix (just read the 1, it should be enough )

https://www.researchgate.net/profile/Jean_Francois_Manlay/publication/301450721_Fast_checking_of_CMM_geometry_with_a_patented_tool/links/59db164baca272e6096e1104/Fast-checking-of-CMM-geometry-with-a-patented-tool.pdf

[PowerJunkie]

Thanks Jefman !!

[JEFMAN]

You're welcome !