Sigma survey

[[Ja...]]

Just out of curiosity, I’m wondering what the smallest sigma value people use before you consider it to be to the point where you’re eliminating too much data.

[[Lu...]]

My filters are set for plus/minus 3 sigma

[[Pa...]]

I've never considered using sigma for this.  I'm not sure there's an answer.  I mean scan a circle with a gazillion points with any attenuation settings, then come back in and take 5 discrete points at the same location.

They will both have sigma values.  They will both report different results.  I can't use S to determine which is correct.  I know the 5-point method threw away no points.  Maybe use the S value of this as your target?

 

[[Ke...]]

I think I remember someone stating thet a Features Sigma value should be about 25% of it's Form... but IDK where one would go from there.

I'm more of a visual person, so I always "show masked actual points" when I'm programming, and I plot roundness & flatness for important features so that I can see if there is too much filtering

[[Ch...]]

This is a question that is dependent on too many factors.   A default of 3 sigma should be good for most.  Remember, with 3 sigma you're keeping 99.7% of your data.  At one sigma you're only keeping 68% of your data.  Its just a normal distribution.  

There's only been a few times where I knew I wanted to eliminate a ton of data where I lowered the sigma value on my filtering.  

[[Ke...]]

Please sign in to view this quote.

There have been a few times that I have needed to increase the the value to 4... always on a Flatness evaluation on a stamped part, so that I did not discard a large chunk of "good data" on a "bad part"

[[Ja...]]

Far and away I use +/- 3 sigma for the huge majority of the dimensions as 90% of the tolerances here aren’t very tight. But when the tolerance window starts getting below 0.002” I start tightening up the sigma values because I don’t want noise, random freak vibrations, probe chatter and the like affecting the results. The closer the window gets to 0, the closer I get to a sigma value of 2.0. At that value I’m eliminating roughly 4.5% of the data and for me personally, I don’t want to be removing any more than that. I’ve been using this method for quite some time and seems to work fairly well. I was just wondering what other people here consider the stopping point.