Step Points

[[Mi...]]

Hi All,

Does anybody have some tips on Step Points? Better to scan up the step or down? XXT vs XT?

I have a task where they would be very useful, but I just tried it on a similar geometry and the results were lackluster, when it calculated the results are fairly close to other methods, but it doesn't like to reliably calculate and it takes FOREVER to calculate, even on a very powerful PC, for some reason It's struggling to iterate.

 

I did the scan with 4 different ruby sizes, 1.5mm, 1.8mm, 3mm, and 5mm. If I run repeat attempts different ones actually calculate.

I've used a step width between 0.005mm to 0.04mmm, speed as low as 1.25mm/s. No filtering/Outlier. I've scanned longer and shorter paths, No obvious difference between anything.

 

 

 

[[Cl...]]

Doing this, I only ever tried scanning up (XXT). Worked well for me.

Edited March 5

[[Ke...]]

Unless I'm scanning an incredibly short/long feature, I usually usa a 0.05mm to 0.10mm step width; I try to make sure that I don't collect "too much" data... maybe 5-7k points for "larger" features

[[Cl...]]

I've also found that scanning slower is better.

[[Ma...]]

Always try and scan up. Pulling instead of pushing, less tension on the stylus and reduces surface roughness inflection when not applying the filters. I generally use 3 point per every nominal point used in the definition, for example, nominal defined with 100 points, I’ll use 300 points to scan. This will allow for triangulation to set the correct point vector as it lies on the surface.

[[Mi...]]

Please sign in to view this quote.

Mark,

 

Just to clarify, I assume you're talking about using a 2D Curve? I understand how the nominal definition vs actual points scanned applies to curves... but I'm using a 2D line for a step point, not a curve for a Kink Point.

 

Does that change the answer? 

[[Ma...]]

In manual there is written strategy, so there is some help
Manual Calypso 2023 - page 4-190 ->

Please sign in to view this quote.

 

[[Mi...]]

Thanks Martin,

Those are essentially the steps I was taking where I was having issues.

However I just experimented again and found that it was much better.

I scanned up

Using 5mm, 3mm, and 1.5mm rubies.

2mm/s, 0.04 step width. 25mm total length.

RFP the upper section to use as the datum reference in the step point.

Results were very consistent and not significantly different with different ruby sizes, they were however more different from the Contour Tracer than I hoped (about 50 microns).

I think some of the error is from the calculated circle I was using in the Caliper Distance, as that circle had about 0.075mm of roundness. Next test I'm going to take a point or scan another very short line to measure a 2D Polar Distance from that to the Step Point.

Or recall only that section of the circle to save taking another point.

[[Ma...]]

Please sign in to view this username.

 I would normaly try curve and bestfit that. Then place there theoretical point with alignment from bestfit curve. Since i have no idea how small is this part, then this "Step point" is somewhat more accurate.

I thought that if you have there radius instead of corner, then this is what makes it slower for calculations. ( at least i see radius on that model )

[[Mi...]]

Please sign in to view this username.

 Unfortunately I must not use curve. Only 1 of our CMMs has it so if I do this it needs to work well enough on 2 of them to get through the job when it comes. This is at least calculating repeatedly now so that's a big step in the right direction. Not sure if the problem before was because of scanning direction or just something straight up wrong with the features. I deleted and started over.

Now the feature I'm going to measure has a .5 micron tolerance, and isn't critical, I just don't want to have to trace 150 parts if the CMM can do it for me.

I RFP 5° of the ID circle into a point and used 2D Polar distance between that and the Step Point, now I'm only 0.03mm from what I get on a tracer. Getting closer lol.

There is a slight corner break on this real part, and I bet that's part of the issue with correlation. The Prototype part I will actually be using this for won't have a corner break, I just don't have it in hand so I'm testing the method ahead of time, hoping it works when the time comes as there will be a time limit on the results.