I have played around a awful lot with getting the renishaw as precise and accurate as possible. There is one important thing to be taken into consideration: the precision of the gauge ring and its accuracy. Its not round. The printed dimension is not 100% correct. Reference the inspection protocol for their measuring error. The best way I have found: fix the ring as flat as possible to the machine. make sure it can´t move. make sure its not distorted. Take work offset from ring. zero over ring and validate center as perfectly as possible (even now, with micron indicator it will be quite impossible to find a "perfect zero"). Zero machine on this spot. Now calibrate from exactly this zero point, in exactly same height where the indicator was zeroed. This method improved my measuring error a lot. After calibration it´s also advisable to measure several gauge rings right away to validate the calibration went correct. I even found small differences in precisely repeated calibration processes. And of course: never chase the last micron. :)
Great video, just found your channel excellent work! One question, in this experiment the uncertainty is attributed to the probe, wouldn't it be more accurate to say this is the uncertainty of a probe in a Mini mill/system? Since errors in backlash, servo couplings, axis alignment etc all contribute to the recorded values?
What would be really awesome. Is a tutorial on how to program a probe routine for inspecting a part for position of 4 holes or something. Keep up the awesome videos
We're working on a video on the GD&T position tolerance and I think this might fit well there. We have just the part to use as an example too. Thanks for the suggestion!
If you don't have a perfect machine or vice , two things are important: Checking the ring for flatness after clamping. Checking the spindle for eccentricity.
Another awesome video, thank You. I don't fully understand the proces but I assume You check widest dimension with that probe for calibration. But in order to work, it should start from exact middle of the circle. Any 0.00000x offset changes the length of x and y reading right ?
Yes, 1.99996 was the actual value on the certificate of calibration. Per the 10:1 rule, that number would have been obtained with a measuring instrument that could resolve at least to +/- .000001.
Anyways, how do you know or how can you be certian you are actually calculating the Machines repeatability? I use a probe on a machine with a whole factor of magnatiude more, precision and accuracy; cylindrical grinder. We routinely hold tenths on postion using a probe during production. repeatability of machine is .00001. (Swiss made Studer CNC). Production would be scrap if our probe and machine couldnt hold tenths. You might have better results and infact anedocatly like I say, it is possible using a different method and more accurate tooling. edit:Hass say it can hold only 0001 on position, renishaw claims in the millionths. Should have started there.
Your machines are on a whole different level. Most swiss machines I have seen have glass scales on the axes. This machine definitely does not. Somebody else made a good point in another comment about thermal expansion. If the ball screw heats up 10 deg F, it'll expand by .00012 over that 2" range.
@@tarkka Well thats what i am saying, you arent testing the Renishaw probes accuracy or precision anymore then you are the positioning accuracy and repeat-ability of the HASS machines,
There was one measurement that seemed to be an outlier. Any idea what might have caused it? Also, what would be the uncertainty if that data point were ignored?
Sorry it took so long to get to this. You're correct that the 1.999773 data point appears to be an outlier according to a Grubb's test (P=0.002). Throwing out the outlier, the accuracy error becomes .00012 and the 2*sigma value becomes 0.000019. This gives a new total uncertainty of 0.00014 (compared to 0.00017 without throwing outliers out).