Accecptalbe variation between cmm's

[Paul Sarrach]

Is there a industry standard or unwritten accecptance a variation from machine to machine?

[kbotta]

Not sure about that one. I've heard "people" say as much as .1mm from lab to lab (seems extreme huh?!).
I guess depending on the controlled enviroment, machine, machine specs, etc...it could add up.
I would think same machine type, same lab location - it should be within your Volumteric acc.
Anybody else??
Kev

[ag162]

Hmmmmm,
I would think that this is a little hard to calculate. If you were to do an accurate rating, the temp, humidity, machine size, make, model and mfg date would have to be the same. Then the material being checked would have to stay at the same temp, to prevent any changes between facilities. But as for a real life experience, I measured a part at the prev company I was at, then we checked it at the company that was purchased and the same prg has within .05 and .07mm of the same results. So, .1mm is a generous but effective guess. It definately would not be worse.

A.Gore

[Jeffrey Lovely]

are these machines in the same lab?

[Paul Sarrach]

Yes .1 is what I heard also. But if your tolerance is .1 on a fixture then What the gage company says is correct, So if my readings are out I still have to accecpt?

[ag162]

No, they were in different cities.
This is workable in stamping. When you have a tol range of 1.0. But if you are comparing results of a gauge, which is 10% of the tol, or a casting with milled surfaces and you are in a tight tol situation. This could be a slight obstacle. Who's right? Who's wrong?

A.Gore

[Matthew D. Hoedeman]

In reality, there should NOT be any difference from one machine to another, no matter where the second one is. We all know, however, that it will never be so. If ALL CMM's were calibrated correctly, then there should be very very small difference from one to another, BUT......

Some of you may want to cover you eyes as you may not want to read this, it is a subject I have harped on for many years. The ball-bar calibration method is not a 100% calibration method. It can NOT be used to anything other than electronically square your machine. It can be used (if calibrated) to see if the linear of the machine is close, ie does 13.457 on the ball bar measure 13.457 on the machine? If it does NOT, it can not be used to accurately adjust the linearity of the machine. Also, when you get down to the absolute bottom, it can not really check the linearity of the machine since it is only good for one increment, not multiple increments. So, just because your machine measures the 13.457" spread of the balls on the ball bar does not mean that the machine will also measure something 9.321" as being 9.321" The ball-bar is a place to start, but it isn't the end, not by a long shot.

OK, you can un-cover your eyes now.

[kbotta]

(j/k)

[Jeffrey Lovely]

I read in CMM Monthly magazine
if you "pay extra" at calibration time
you can have the machines mapped and calibrated to match the same
how exact I dont know.
I assume with lasers

[Paul Sarrach]

So no standard. What do you do when you are within .1 but out?

[ag162]

My opinion, would be to report as out. Let the higher ups make the judgement call. Don't fudge or sweep it under the rug. Use these posts for your defense. That amount may just be dev between machines and companies. You can also check the temp that it was checked at before and try to duplicate the atmosphere and try it again.

A.Gore

[Matthew D. Hoedeman]

But the calibration of the CMM’s is supposed to be the standard, guaranteeing that they will all measure the same thing the same way. As for what you do now, I would say you have to figure out which one is measuring correctly AND giving the correct values. Basically, you are between a rock and a hard place. Each ‘company’ will say that their CMM is correct and the other is wrong. However, if YOU are the customer, then YOU are the one who will decide what is acceptable. This is where third party certifications come in. If you find a good source for the third party certification, I would have to assume (dangerous word and idea that it is) that since THEY are required by some of the BIG companies to have an ‘extra’ certification that they would be a good source for ‘master’ results. However, some of them will do no more than run the program supplied by the fixture build source. And, we all know this happens, if the build source fudged something in the program to make it check good when it isn’t, then the third party source will not see it as bad either, unless they write their own program. It could be something as simple as changing a nominal so that no one would see the fudge unless they compared EVERY point back to the cad nominal. OR, it could be something as insidious as using an ACTUAL stock thickness value so that a comparison of the nominals to the cad data will show a match. Using an actual thickness value does NOT change the nominals of the point, it forces Pcdmis to do extra compensating of the point by the amount and in the direction of the amount used (- would make it comp MORE for the probe, while + would make in comp less). Also, they could slip in an extra alignment in the program to offset an axis, then go back for the next point.

Just remember, for each idea I came up with, someone, somewhere, will come up with something else entirely, so what do you do to catch all the cheats and liars? Who will guard the guardians?

[Schrocknroll]

When you are doing a CMM correlation study the standard is usually 10% of your tolerance, so I'd take the tightest tolerance you work to and use 10% of that as the standard.

As far as Matt is concerned, I agree 100%. The guy who does our calibration ALWAYS uses a uncalibrated ball bar and then a step gage (he's currently looking into getting a laser). He then trys to get the accurcy a little better than it was before he got there, even if it is within the machine specs.

Did I mention he's alot cheaper than Brown and Sharpe?

[Matthew D. Hoedeman]

Sounds like my guy, except he does the ball-bar LAST and only to prove that the machine will pass it, he does NOT use it to set anything on the machine, only to prove to people who 'don't get it' that what he does is right. In fact, until last year (or 2 years ago) they never even had a ball-bar. They only added it because some (some, not all) of their customers were asking WHY they didn't use one when 'everybody' else used them for calibrations.

[Jeffrey Lovely]

look for data manipulation
if most of the readings are matching yours
and the points that are out dont come close to yours
you be the judge
if its a very large gage, may be many reasons for variation