O/T: CMM error when less than 90 degrees

[Philcott]

Hi all,

I have recently had a run-in with QA rejecting a part (later accepted ) because of an arc being out of tolerance. This was a turned part with a large arc of less than 90 degrees. I have lots of confidence in both the machine and the program and all tools were set correctly(height, rad etc.) It was later proved the part was right and the CMM wrong.

It seems there is something called lobing (sp?) which makes CMM's have trouble when measuring arcs of less than 90 degrees. Has anyone heard of this before and are there any solutions to this dilema??

We have a Mitutoyo Bright-A504m cmm.

Thanks,

Phil

[MetalMarvels]

We had a similar problem with our Starrett CMM. One of the temporary "fixes" was to do a multiple-hit measurement of the arc (eg. more than 5 points along the arcs). The permanent fix in our case was a new software load that used a different algorithm to calculate the arc fit, coupled with 5 or more measurement points. As I understand the explaination from Starrett, all CMMs are subject to this type of error. The major differences between various CMMs are how the software compensates for it, with some algorithms being "better" at compensating than others. The Starrett rep indicated that the problem was particularly noticeable when at the limits of the measurement capability (5 decimal places or greater).

 

One difficulty is that with fairly shallow arcs of less than 90 degrees, there is a strong coupling between the "measured" arc diameter and the "uncertainty" of the exact position of the "hit" along the arc. If you measure several more points than the minimum required 3 points, the uncertainty is averaged out and the "measured" arc diameter is closer to the actual diameter.

 

Another issue with almost all CMMs is that the touch probe is MOST accurate when taking measurements in the plane of CMM movement (X or Y or Z) and is LEAST accurate when taking measurements at an angle to the X, Y, Z planes of movement (eg. 45 degrees from the X/Y/Z planes). This is due entirely to the way probes are constructed. Most CMMs compensate for this error (more or less successfully).

 

You can easily demonstrate the problem by taking a diameter of 5-inches or larger and performing several 3-point measurements. Do an overall diameter measurement using 0/90/180 degree touch points around the diameter; a second overall diameter measurement using 45/135/225 degree touch points; and a third measurement all along an arc less than 90 degrees total. You will be surprised (dismayed) at the resulting "measured" diameters. How close all three are will tell you how well your software compensates! Originally our Starrett had "measured" diameter differences in excess of 0.00035 inches for the same component (we had a part tolerance of +/- 0.0002). The same test run now will yield differences of less than 0.00004.

 

IMHO this is probably one reason why a lot of folks are looking towards non-contact measurements as a "cure" for this problem. As an aside, two difference operators can come up with radically different measurements due to differences in technique..............

 

Sorry for the epistle, but we went through this while trying to measure that 2 tenths part (on a 5.5000-inch diameter).

 

Cheers

[Scott Bond]

Good afternoon

quote:

---

are there any solutions to this dilemma??

---

This is what do with an inspector that I work with when anything similar comes up. Go to the MC work station and analyze the part on the screen, until we both agree it's a good part on the screen. Project the points onto the part,analyze them and print them out. Then go to the cmm with the inspector and check the points. I know this maynotbe the answer but it might help ship the part.

[Lucky]

Lobing occurs in all Renishaw probe heads. The probe is held in place by three springs (rubber bands) located 120 degrees apart. Depending upon the probe direction, pressure is applied by one or two springs. This direction dependent pressure results in lobing error. A master ring gage will not check round. Renishaw probes are common and available on most CMMs. The lobing errors are usually small.

 

Lobing errors can be compensated for by the software. It is similar to the pitch error compensation used on CNC machines to correct for leadscrew errors.

 

Zeiss CMMs are available with Zeiss manufactured probes that are of a different design and exhibit zero lobing error. They are very expensive.

 

[ 09-16-2002, 08:48 PM: Message edited by: Lucky ]

[Tony]

I wonder which CMM inspected the

Hubble Telescope lens?

[Finecut]

Tony,

It wasnt a CMM that checked the part.

 

Finecut

[chris m]

quote:

---

A master ring gage will not check round

---

I have found that this is definitely the case with both of our CMMs although our computer controlled machine is better than the manually operated one [both Brown & Sharpe with Renishaw].

As far as I'm concerned, CMMs aren't built to check diameter, concentricity, or roundness to fine tolerances anyway so you just need to take those measurements with a grain of salt when you get down to the .0002" range.

 

Just keep measuring until it gives you the number that you want!

[Keith Graydon]

Tony,

I my memory serves. THe Hubble mirror was built to spec. Problem was design. Hehe go figure.

 

Once the media got hold of it they made the fabricators the whipping boy. Some things never change.

[Philcott]

Gary,

 

Thanks for the reply. We will give your test a try and see how dismayed we can get.

 

Ps I was just in Denver for the Fly Tackle Dealers show. I was impressed with the downtown core. Very tourist friendly. Even at night (mostly)

 

Phil

[Philcott]

Scott,

 

I have a good repore with QA and we usually come to agreement on discrepencies. I pretty much do as you said to show the part is drawn correctly and I have the proper radius in the control and cutter.

 

Phil

[Tri-Tech]

quote:

---

Lobing occurs in all Renishaw probe heads

---

Sure blame it on Renishaw

 

Change the blueprint radius +/- callout to a profile tolerance and the problem goes away.

 

Ernie

[MetalMarvels]

quote:

---

Just keep measuring until it gives you the number that you want!

---

The sad part is that you CAN do exactly this......

 

[ 09-17-2002, 05:43 PM: Message edited by: MetalMarvels ]

[chris m]

Gary - this can be helpful sometimes, especially if you know its right!

 

Ernie - I don't understand; if the CMM thinks that the radius is wrong, it will also think that the profile [which is unmeasureable in true form anyway] is wrong too, won't it?

[Thad]

Having trouble with Starrett CMMs? Check out this article

 

Thad

[Philcott]

Thad,

Thanks for the link to the Starrett article. Very interesting. I guess we aren't the only ones who expected the CMM to give us reliable measurements.

 

Phil

[Guest CNC Apps Guy 1]

quote:

---

THe Hubble mirror was built to spec. Problem was design. Hehe go figure.

 

Once the media got hold of it they made the fabricators the whipping boy. Some things never change.

---

LOL!!! Isn't that the truth. Seems as though many inspectors seem to think that CMM's are the panacea(sp?) of the inspection world. Truth is it's a tool, prone to error just like everything else. I believe non-contact measurement is a better deal. Lasers are FAR more precise than a mechanical device. I used to work for Seagate (the HDD people) and the plant I worked in manufactured the HDD platters. We checked those with Lasers. Laser is the only way to measure 17 Angstroms (our flatness callout if memory serves me corrrectly). 1 Angstrom is like .00000001 mm. Besides there is less human intervention in the measurement process with lasers.

 

JM2C

[Verisurf - Ernie Husted]

Chris,

 

Let me try to explain why profile is the way to go,

 

Measure Profile:

Measuring points and fitting and comparing them to a nominal radius will give you a true deviation of profile (1:1 uncertainty from CMM accuracy).

 

OR

 

Measure Radius:

Measuring points and fitting a 90 deg. radius is fine, but you must have a measuring device 6 times more accuracy then the radius tolerance (6:1 uncertainty), 45 deg radius is 25:1 and so on…… (180> deg is 1:1)

 

Here is an example to help understand why there is so much uncertainty.

90 deg. 1.000" R [.002| profile] or (+/- .001)

If you draw a 90 deg. 1.000" Radius then offset it .001 both ways you will find that you can create a 0.99421" Radius and a 1.00587 Radius between the offset arcs. That is a 0.01166" variance of possible radius fits that can occur between the +/- .001 profile.

 

Do you agree?

 

GD&T all the way

 

Ernie Husted

Verisurf Software, Inc.

www.verisurf.com

 

[ 09-19-2002, 10:02 AM: Message edited by: tri-tech5-axis ]

[McRae]

Ernie,

 

Nice example, but the second frame has a true position on the arc center. Nice to use the bonus tolerance until the designer figures out how to use GDT properly. Now we have used up all the standard excuses that we tried to jam scrap at the customer with from the my old Job Shop Days.

 

Oh Well - Remember that QC staff are not permitted to think. There job is to measure and report deviation only, not to manipulate figures and solve problems.

[chris m]

Whoa, Ernie!

 

Too much thinking for me

 

And to think I got into this trade because I wasn't good at math...

 

D'oh

 

C

 

[ 09-19-2002, 09:24 AM: Message edited by: chris m ]

[Verisurf - Ernie Husted]

MfgEng (Andrew McRae)

 

No! that was a Radius measurement example and the true position on the arc center would be an even higher uncertainty for a 90 deg fitted radius.

 

You need to Remember that QC staff is permitted to think.

 

Ernie

[McRae]

Ernie

 

RFS - (Regardless of Feature Size) No Fitting Permitted

[Tri-Tech]

Andrew,

 

quote:

---

RFS - (Regardless of Feature Size) No Fitting Permitted

---

Fitting is Permitted and has to be done if there is no datum callouts associated with the profile or position tolerance.

 

Ernie

[McRae]

True position call outs by definition have to be referenced to datums. We are taking this thread outside of the scope of the original subject matter. Are you an inspector by trade/Job Description?

[Guest CNC Apps Guy 1]

No, he (Ernie) writes inspection software that runs inside Mastercam.

[McRae]

Ernie -

 

Please email me directly. I need information on the services that you offer ie- compairison with model to machined part.