Partial Arc/Circle Segments

[[To...]]

I'm checking a 90 degree segment of a 44" diameter ring. Within the 90 degree arc, the circle is broken up into 4 segments, each roughly 20 degrees length of arc. I am measuring 4 circles and re-calling the feature points into 1 circle. Each of the circle features is reporting .00006" to .00017" of LSQ form. but not reporting the correct diameter. The constructed circle is reporting .00032" of LSQ form and also reporting an under-sized condition. Roundness of constructed circle is .00029". (I am super impressed with the form results on apart this size.) If I constrain either the X/Y or the radius, the form numbers jump to .0015" and .0048", respectively. The size does not improve much with X/Y constrained. I don't have a lot of faith in the location values anyway because the location is established with another full 90 arc segment.

My question. Since the form is reporting quite good, should I trust the size results?

[[Ch...]]

Hey Tom,

My name is Chris Rotolo. CMM Programmer for Moog Aircraft in Torrance, CA. I've been reading the forums for a while now, haven't interacted much. There's a ton of good info here though, and it's great to be able to share with people that understand.

You're probably well aware the uncertainty magnification with 90° or less of a circle.

Are you able to combine the other 90° full arc with the 4x20° segments? Perhaps manually probe them all (180° total?) to get a ballpark.

Here's what I would probably do, which was recommended by Zeiss, and has worked well for me in the past:
1. Measure Circle 1 (4 paths 20° each) no constraints - no filter/outlier
2. Copy this Circle 1 to Circle 2, recall feature points and constrain radius to get "accurate" location. Apply filter/outlier
3. Copy Circle 1 to Circle 3, recall feature points. For the nominal XY use formula to getActuals from Circle 2 XY,
Apply filter/outlier. Finally constrain the XYZ to find size.

This has worked well for me in the past to verify radii w/in .0001" compared to video comparator with 200x.

Hope this helps some, Go Pats! I was born in Mass (Worcester), went to school in Boston. Lived in CT 10 years before moving to California in 2014.

My specialty is gears, however I can handle just about anything.

Happy measuring !

[[Kl...]]

Found this Handout at Google

http://publications.npl.co.uk/NPL_WEB/PDF/MGPG41.PDF

Chapter 6 / Page 48 - 50 describes the question you ask for and also a possible solution.

My experience
I had (have) to evaluate a lot of gothic profiles and the calculated radius has have had always a big variation, therefore not stable.
We switched the evaluation to profile form, now the customer produce the arc segment into +/- 2µm.

[[Br...]]

It sounds like (If I'm understanding you correctly) the problem is actually in your drawing Tom. If the feature in question does not have opposed points and cannot contain or be contained by an actual mating envelope, it is NOT a feature of size, and therefore cannot have a size tolerance applied to it. To locate the surfaces of a non feature of size like yours, a radius tolerance or even better, a profile tolerance should be used. That is why you are seeing good form results but poor diameter results because calypso has to fill in the gaps where the surfaces of the feature are missing. Usually though if this is the case, constraining the axial location of the feature will make the diameter/radius look much better. If this is not the case, you may actually have an issue with where the surfaces are located. Throw a profile tolerance on it and see where the actual surfaces land vs. the nominal surfaces. This should tell you a lot. And first and foremost, make sure the nominals are actually what the print specifies. Models can be notoriously fallible.

Another thing, when you put those constraints on and the form jumps up, try toggling the "Connect Segments" checkbox in the filter options on and off to see if you see any difference.

[[To...]]

Thank you Chris, Klaus and Brett for the feedback. I know I can count on the forum to help me out. I haven't had time to try everyone's suggestions but I quickly created a Free Form Surface from the data collected on my Datum B cylinder and am showing .010" profile (no DRF) and can clearly see the out-of-round condition. That gives me something to work with. I will review everything else provided. Again thank you.

[[An...]]

Please sign in to view this quote.

Using Profile and simultaneously taking serious the plus-minus-tolerance means the following:
The tolerance value of profile is by a fraction of the diameter tolerance.
for example:
arc diameter 10±0.2 (90 degrees)
means
profile 0.04 (estimation)

[[Br...]]

Please sign in to view this quote.

It can have a similar meaning depending on how its shown on the drawing. Just remember that pesky crescent shaped tolerance zone that you so lothe Andreas. You may prefer a profile tolerance in Tom's case as to achieve a very clear and reproducible meaning on the drawing.

[[An...]]

Please sign in to view this quote.

Good catch!
Let me think it over.

[[An...]]

Regarding the real points but not the derived element by using Profile.
There is still no crescent shaped tolerance zone.
Watch the reduced tolerance value and the factor in order
to come back to the drawing.

See attached!

Contribution_12_09_2019.pdf

[[Br...]]

Only if the center line is explicitly dimensioned. Otherwise there is an implied tangency rule which will make the crescent shaped TZ. We don't actually know what Toms drawing looks like. Youre probably right though that it has its center line dimensioned. But if this is something so critical, why not use Profile. You have full control over a datum reference frame to develop your tolerance zone then. You also have full support of the Y14.5 standard. The radius tolerance is less than optimal. Like which surface to I set off of to get the orientation for this tolerance zone? A Radius tolerance is fine for fillets and corner rounding, but for critical surfaces, its in your best interest to use profile.

[[An...]]

Please sign in to view this quote.

I agree.
The problem for the designer is,to "translate" the plus-minus-tolerance
into a appropriate profile-tolerance.
±0.5000 yields profile 0.2266 U 0.0979.

±0.5000 does not yield profile 1.000

Also see attached.

Contribution_12_09_2019_a.pdf

[[Br...]]

Please sign in to view this quote.

Well that's a difficult and backwards way of toleranceing. Calculating where your clearance needs to be and putting that into a profile tolerance is much easier. Who cares what the "old tolerance" used to say. Time to start fresh and do it right. There's way less work involved.

[[An...]]

Please sign in to view this quote.

So what can we do,if the designer refuses to alter the dimensioning?
(My experience over lots of years)

Translation and retranslation and ready to do some math.
Just showing a way in the attached file.

Contribution_13_09_2019.pdf

[[An...]]

Some further thoughts about arcs and polylines and dimensioning.
See attached.

Contribution_16_09_2019__.pdf