hexagon logo
Register Sign in
  • Replies 20 replies
  • Subscribers 222 subscribers
  • Views 295 views
  • Users 0 members are here
Options
Related

CMM consistently reports .0001" higher than micrometer

xbigfatgoat
Here is the cmm set up:

I'm using a Global S 7.10.7 Green

Probe Setup:
HH-AS8-T2.5
HP-S-X1H_T
HP-S-X1H_26_SH
HP-S-X1h_M3_5WAY
Connection #1 tip3x10mm
Connection #5 convert to m2thrd tip3x50mm

The room is temperature controlled to 68 degrees. Everything being measured has had time to temperature adjust to the room.
I have temperature compensation enabled.

I'm measuring a diameter that is 1.7720" +/- .0002"

I calibrate the angles with 25 hits each angle answering "no" to has the sphere moved. The sphere was set based off a 5x50 master.
I've tried calibrating the sphere with and without temperature compensation enabled.
I've noticed when I don't use temperature compensation my stddev can range from .0002 to .0003, I even noticed .001 one time. With temperature compensation the stddev is always less than a tenth.

I'm scanning outer diameter's. I scan one half T1A0B90 90 to -90 degrees with no filters and the other half -90 to 90 degrees T1A0T-90 with no filters.
I then construct a best fit circle out of the two halves from every single scan point using no filter.
I create a constructed filter out of the constructed circle and use Gaussian filter type with UPR of 177 and remove outliers set to 2. For the UPR I found an equation online that gives you a UPR based on the diameter: (D(mm)*PI)/.8
I construct a final best fit circle with no filters from the constructed filter and report this diameter.

Initially, I was consistently getting one to one results between the cmm measurements and a micrometer.

Recently however, I'm finding that the cmm is consistently reporting .0001" higher than what I measure using two different micrometers.

I know .0001" is extremely small and I'm trusting that the micrometers are giving me the more accurate results.

Am I doing something wrong?
Is it reasonable to assume that I can get the cmm and a micrometer to match to one 10 thousandths of an inch?
There has been a couple crashes (not huge crashes), could this have thrown off the probe by .0001"?

Thanks for any help in advance.
​​
  • .0001" is pretty darn good! Size to size is achievable but hard to come by in a production setting. Perhaps the CMM is taking into account the shape/roundness into account where your micrometer is only measuring two points... I don't know. A tenth is tiny. Do you have multi-point micrometers, I don't know if it's even worth the trouble... are you doing this for funzies or on a production part? If the latter what tolerance are you playing with?
  • in reply to Kp61dude!
    I'll have to check the roundness tomorrow, but i'm measuring a cylinder made up of 5 circles. Looking at each individual circle i measure in relatively the same spot with a mic and across the board its pretty much always .0001 higher at each circle. Im doing this for a production part and the tolerance is +/- .0002" so being .0001" off is half the tolerance. Whats driving me crazy is that for a while it was actually 1:1 between the mic and the cmm so i dont know what happened. At first it was the diameter was spot on but the runout was off and of course when i finally get runouts to match a dial indicator by 1-2 tenths for the most part, now the diameter's arent matching perfectly, go figure.
  • Try measuring a min circ circle instead of least squares one.
  • Using outlier filter set to 2 seems to be large., I would use 3.
    I wouldn't use UPR filter, in a first time.
    I would use also scan calibration to get the right ball radius in scanning.
    What is the roundness value ?
    Did you try to use "size" dimension ?
    You could try to measure with the micrometer the diameter between the start point of the scan and it's opposite.
    Then, search in the scan the hit corresponding (create a feature set SCN1 with both scans, CE1 is the constructed circle, assuming it in Z+plane)
    ASSIGN/P1=SCN1.HIT[1].XYZ*MPOINT(1,1,0)
    ASSIGN/P2=SCN1.HIT[SCN1.NUMHITS/2].XYZ*MPOINT(1,1,0)
    ASSIGN/C_CE1=CE1.XYZ*MPOINT(1,1,0)
    ASSIGN/DIAMETER=SQRT(DOT(P1-C_CE1,P1-C_CE1))+SQRT(DOT(P2-C_CE1,P2-C_CE1))...........................sum of polar radii if the points are not exactly aligned on the diameter.
  • I was calibrating gauge pins with a mic while cmm was doing other stuff and was having discrepancies. I put the gauge pin on the cmm and the micro-vu and found those to all correlate. I deemed the mics were off .0001" on 1 set and .0002" on another. These mics were also just calibrated. (not saying the calibration company is that good). Anyway, usually when I see these small discrepancies, it ends of the roundness is off .0001" Let us know what you find out.
  • Every time I crash my Leitz, I conduct a new Lower Matrix calibration.
    Not only does this help correct for changes to the sensor, but it also is a means of determining if the sensor sustained enough damage to need replacement.
    chasing a tenth is a challenge, and if you allude that you've nailed correlation in the past, and changed nothing other than the fact that the machine crashed, it's probably best to start with assessing and calibrating the probe system entirely after the crash.
  • in reply to louisd
    Ok, reading about this I see it recommends using a 5x20 probe. Now I'm confused because I thought a 5x50 was supposed to be used as a master.
    Should I use a 5x20 for the lower matrix and locating the sphere?
  • Every time I crash my Leitz, I conduct a new Lower Matrix calibration.
    Not only does this help correct for changes to the sensor, but it also is a means of determining if the sensor sustained enough damage to need replacement.
    chasing a tenth is a challenge, and if you allude that you've nailed correlation in the past, and changed nothing other than the fact that the machine crashed, it's probably best to start with assessing and calibrating the probe system entirely after the crash.

    To add to what said here... probe and qual sphere cleanliness is really important here. Inspect ruby tip under a scope and use lint free wipes and/or make sure to use the watchmaker's putty after you've wiped everything down.
  • 'fighting' 0.0001" is insane! What machine do you have that you can even afford to worry about something that small?
  • Here's what I'm getting for circularity.

    ​
© 2024 Hexagon AB and/or its subsidiaries. Privacy Policy | Cloud Services Agreement