Points vs. Scans

[[Ja...]]

I understand that there is greater accuracy, from less noise, when taking individual points opposed to scanning. I also understand that scanning is necessary for profile callouts. Coming into this company, I am finding that many of the old Calypso programs are done predominantly through scanning. I have since adopted this habit of scanning every feature. Is it more accurate to get out of this habit and go back to taking a sufficient number of points per feature?

We have a Vast XXT.

Thank you.

[[Cl...]]

I would agree that scanning is generally the way to go. Although there are applications where it is not appropriate. If you have for example a small radius, or sphere segment and you need to evaluate size, I would almost always use single points with constraints. If you need to evaluate form I would scan.

[[Ri...]]

Scanning will always be more accurate, faster, and repeatable than single-points.

You cannot properly evaluate the true size, form, and location of a feature with single-point measurements - you will never be able to take enough points in a timely fashion.

[[Cl...]]

Richard, I respectfully disagree with part of your response.
I had a part some time ago with a very small ID sphere to
measure for size and position.

Scanning several small circle segments would not repeat well,
or give me correct size. I changed my strategy to single points
and that worked much better.

I would say that each situation might need a slightly different approach.

[[De...]]

Less points will never equal more accuracy.

When people say scanning gives bad results, what they are really saying is that additional data makes it harder to interpret the results.

The thing is, just because scanning a cylinder allows you to see that your cylinder is three point out of round and tapered from top to bottom and your two point check at one end of the diameter does not, does not make the cmm check incorrect or inaccurate.

For example, the standard says form is a function of size, so if your part is out of round more than your allowable tolerance then it is no good, yet people check close tolerance bores with a two point check and call it a good part because they cannot see the form error. This does not make the part good, it makes the two point check bad.

I have yet to find a part that when scanned and then verified 'correctly' on the plate was 'wrong' from the CMM, generally when a comprehensive study is done into the manual inspection methods and the CMM program one of two outcomes is generally found, errors in the cmm inspection, or errors in the manual inspection.

Errors in the CMM check often include:
The CMM program has error which are causing improper results, such as incorrect filters / outliers / and evaluation settings.
The part was loaded incorrectly.
The part or probes were dirty.
The probes were loose or shanking out.

Errors in the manual inspection often include:
Incorrect inspection method.
No ability to properly align part to Datums.
Operator error. Improper understanding of the Standards involved.

[[De...]]

In reply to Clarke's comment I will amend my statement to say that due to the nature of the geometry of parts and or the dimension defined on the blueprint, that it is not always easily possible to get a 'valid' answer from the CMM using scanning.

In these very limited circumstances single points may provide a better method, however in general the idea that scanning is less accurate is predicated on the idea that bad checks give better answers because they agree with poor inspection methods carried out on in manual inspection.

[[Cl...]]

Here is the specific geometry I spoke of.

(cylinder)

Cylinder.PNG

[[Me...]]

How do you guys feel about scanning planes? I typically prefer not to unless the tolerances are tight.

[[Ri...]]

Please sign in to view this quote.

I can agree with that.

I apologize if my statement was over-reaching. I should clarify that 95% of the time scanning is the absolute best measurement.

My statement was more aimed at standard geometry, that is complete. More so to the fit-function aspect. You just cannot accurately and repeatably report max/min sizes with single points.

[[Ri...]]

CMM's only know what they touch. If we were measuring perfect geometries, then it wouldn't matter if we collect 3 or 30,000 pts. The results would be the same. Trouble is, nothing is perfect. The degree that the parts we are measuring depart from that perfect state makes data density more or less important. I like a combination of scanning and single points, especially during the program development stage. I will scan a feature, then take some single points to verify the scanned result. I call these "sanity points". I stole that term from a colleague, Gwyn Carter. He was, and still is, one of the best CMM programmers I ever had the privilege to work with.

[[Ja...]]

Please sign in to view this quote.

I like this idea of "sanity points", it seems like a great verification technique.

[[Cl...]]

Richard, no need to apologize at all, I understood what you were conveying
in the post. My situation I think was a little unique, and I thought I'd share it.

This forum is a better opportunity than any classroom setting to learn!

[[Ri...]]

Clarke, have you ever been in one of my classes? 😃 You might have a different opinion of the classroom setting.

[[Cl...]]

Let me re-phrase that Richard. How about "more entertaining".
I'm sure you have a lot of knowledge to offer. I would love to
attend one of these days.

I've been to basic, advanced, planner, curve, O-inspect. All good, but
depends on the instructor.

[[Ri...]]

I no longer have a classroom to train in but I can show up at your facility and we can go over whatever needs to be gone over. 3 or 5 days,, it's all up to you. Back when I was training in a classroom, I did try and make it as light and entertaining as possible. I found most people are more receptive to new information if they are relaxed and not suffering from "information overload".

[[Al...]]

Please sign in to view this quote.

I attended your Curve class in 2012 and you did make a comfortable environment, especially with the great music before class. And I recall you saying that Space points made good sanity points to verify your Curve results.

Back to the original topic; My Aukom II instructor made a blanket statement that touch points were always more accurate than scanned points; I took that to mean if you needed the data of a single point it was more accurate to measure that point than to extract it from scanned data.

[[Ow...]]

Please sign in to view this quote.

I remember taking one of your classes Richard, I believe it was "Sweatin' to the Oldies".....oh wait a minute...wrong Richard..lol.

Back on topic,
there are instances where points work best, such as when using the unknown contour in curve to reverse engineer an involute and tangent small arc geometry (and those instances mentioned above) but, scanning is always optimal if possible.

In another world I used to work, we had an older touch trigger LK CMM that verified everything before we purchased our 1st scanning CMM. During the transition of moving everything over to the scanning CMM, machine operators would often ask us to check their part on the old CMM to see if it looked better.
To explain the difference, I'd tell them the old CMM was connecting the dots and the new CMM was taking a picture. In the long-run, we improved our product quality and performance.

[[Ri...]]

Alan, I would take exception with anyone saying, "blah blah blah always or never blah blah blah", concerning a Calypso topic. It depends on variables. Material, machine, size, environment, tolerance, stylus size and material. 2012 was during my Bluegrass phase?

Owen, nope, that was me. I am nothing if not versatile.

[[Ja...]]

There have been some great points here, no pun intended. A while back we were having an issue measuring a large (~20in. long) part. The clocking datum was a cylinder where I could only get about 23 degrees of the arc on each side, making a total arc length measurement of 46 degrees (~13 percent). I used scans, and my results (True Position) were seemingly all over the place. When removing the datum from the calculations, everything went fine. Is this a case where I should be taking points instead of scans?

[[Ri...]]

20 inch long cylinder with only 46 degrees of arc as a datum is tough. What all CMM softwares do is to take measured data and create perfect features. With this cylinder, any form error, debris, problems of any kind are magnified in the creation of a 360 degree cylinder from 46 degrees of data. Imagine how difficult it would be to build a physical fixture to locate on this feature as a clocking feature. Is this a customer part or is this your company's design? If it's your company's design, can you get the designers to explain to you how and why this it's necessary to use this feature as a datum? If this is a customer part, can you get information about how they measure this? Material, methods, size, surface finish, are all variables that can effect your approach. Can you provide us that information? If the answers to all the questions above don't allow for any change to your approach then you will have to first determine what the inherent variability of your manufacturing process gives you. If you have the Curve option that would help. Measure this surface as a 3D Curve on 10 parts, and look at the difference. I don't like using constraints on datums but in this case, it might be necessary. Constrain Radius for a more repeatable location.

[[Ja...]]

Please sign in to view this quote.

Thanks Richard. The part itself is roughly 20" long but the datum is only an inch or so long in the center. This is a customer part and I can try to get some more information from their side for the future. We have Curve but I was just constructing a cylinder by using four circular scans on each side. I will look into 3D Curve some more, I am relatively new to using Calypso.

[[To...]]

Please sign in to view this quote.

Whew. I'm glad you cleared that one up. Now, this is going to be really easy.... 🤣

[[Da...]]

Whether to use touch trigger or scanning really depends on several things.

- Is it a known precision surface?
If so, it would save a lot of time taking only the amount of points you need in order to well define the measured feature. The form of precision surfaces are often times at or below what the CMM is capable of measuring.

- What are the tolerance requirements? Are they tight or wide open?
Parts with loose tolerances can usually forgo the added time required to scan a feature. The difference in error you would get between touch and scanning may be insignificant compared to the stated tolerance.

- What manufacturing method or process was used to produce the part?
It may be advantageous to characterize and validate the manufacturing process via design of experiments or by simply performing thorough measurements of the first batch of parts produced. Scan these first few pieces and evaluate to see if changing over to touch trigger is feasible. If it is a long production run this may be beneficial.