How do YOU use Master Scan?

[Mark Kenny]

I’m trying to scan the turned profile of two parts next to each other in order to compare one to the other using a Linear Open scan. I know that the “MASTER” nominal method is correct for this purpose, but I’m trying to determine how to set up the program. Currently, I’m probing the datum features, setting the alignments, and performing the scan, all inside of a loop. The loop repeats twice, once for each part. I then dimension each part outside of the loop, using a “[2]” to designate the second scan. For example, the scan is titled “SCN1” so dimensioning the scans looks like “SURFACE PROFILE: SCN1” and “SURFACE PROFILE: SCN1[2]”. This seems to be working because the first profile reads a measured profile result of zero (which I think it should since it is the master scan), and the second is reading about .0009”. I’ve tried several methods for this project and this seems to be the one that gives me the most reliable results (although still not quite what I am expecting).

Let me say that I spoke with Tech support earlier this week and the gentleman there said that inside of the loop was not the way he would do it since each time it repeats the loop, it’s going to run the scan as a master scan. He said to copy the scan and create a pattern, then to set the first scan to master and the second scan to find nominal. I tried this and the results were way off from what I was expecting, with the second part out of tolerance by up to .010”. If this is the correct way of doing this then I definitely have something wrong.

My question for those of you that have successfully scanned a profile using master scan, how do you do it? If you’re doing it inside of a loop, how are you managing your alignments? If you’re not using a loop, how are you doing it?


Any experienced input is welcome.

[Ego Murphy]

I have seen the Master function perform an unwanted and uncontrollable amount of what can only be described as 'smoothing' of the scan data from the master part as it sets it as nominal, and then showing profile deviations of more than .010" when tested on the master part.

That was using v4.3 back in the day. Haven't tested it since... but from what you are seeing the smoothing may still be in place.

I'm not confident at all in PC-DMIS's ability to employ anything other than a CAD model for nominal when scanning contours. Simple round or flat geometry it can do fine without CAD, but it can't use raw XYZ-IJK text for nominal so how could it use raw scan data (which is just a bag of XYZ data only) as nominal?

[Mark Kenny]

I have seen the Master function perform an unwanted and uncontrollable amount of what can only be described as 'smoothing' of the scan data from the master part as it sets it as nominal, and then showing profile deviations of more than .010" when tested on the master part.

That was using v4.3 back in the day. Haven't tested it since... but from what you are seeing the smoothing may still be in place.

I'm not confident at all in PC-DMIS's ability to employ anything other than a CAD model for nominal when scanning contours. Simple round or flat geometry it can do fine without CAD, but it can't use raw XYZ-IJK text for nominal so how could it use raw scan data (which is just a bag of XYZ data only) as nominal?

As disappointing as that may be, it would allow me to move on from this issue and try to find another solution.

Thank you very much for your input.

[NinjaBadger]

Hmmm, interesting.


If indeed when looping it, it is run as MASTER then you should see no deviation on the second scan as this will be MASTER.


Personally I'g go with the second method, BUT surely the second scan should be set to NOMINAL (comparing to nominal data) instead of FIND NOMINAL which is looking for the nearest CAD data.


A few tests should tell you if it's working.


i.e. Scan MASTER part twice rather than switching parts should give very small deviations.

Run master part and check part, then run gain but reverse the order and you should see similar deviations but in the opposite direction.

[Mark Kenny]

Hmmm, interesting.


If indeed when looping it, it is run as MASTER then you should see no deviation on the second scan as this will be MASTER.


Personally I'g go with the second method, BUT surely the second scan should be set to NOMINAL (comparing to nominal data) instead of FIND NOMINAL which is looking for the nearest CAD data.


A few tests should tell you if it's working.


i.e. Scan MASTER part twice rather than switching parts should give very small deviations.

Run master part and check part, then run gain but reverse the order and you should see similar deviations but in the opposite direction.

Thank you for your response.

There is definitely deviation using the looped method. Not a lot but more than there should be considering that when the parts are scanned individually and compared to the CAD model they are both good, and when looped the second part is not.

I will try your suggestion (running a second scan patterned after the first scan, with the second scan using "nominal" for the nominal method) later this morning after I've had the chance to put out a few fires.