Can the Standard Deviation (SD) value be zero?

C

Cognizant

I am actually qualifying my CMM probe, I remember in my training Standard deviation(SD) for qualifying the probe must be less than .003" for Zeiss Contura using calypso. Now one of my probes got bent and i replaced with a new one and I qualified the New probe but the SD value is zero ...Is that value sounds reasonable.....

though the sd=0 my qualification window says probe is good to run for measurement plan....

Please advice
 

Tim Folkerts

Trusted Information Resource
Mathematically, the only way for SD to be zero is for all the measurements to be identical. If the measurements are truly perfectly repeatable (and the instrument had sufficient resolution for the task at hand) then that would be a great thing. :yes:

You might at least think about whether that sounds reasonable. It is always possible that some thing else is wrong - perhaps mechanically or in software - that causes the numbers to always be reported the same. That would obviously be a bad thing. :nope:


Tim F
 
Q

qualeety

go with the requirement

ralphsulser said:
SD=0 at how many decimal places? If you take it out to 5 or 6 places is it still zero?

if the requirement calls for less than o.003" and you are gettting identical results @ three decimal place...then you got SD of zero........there is no need to measure upto 5 or 6 places...
 

Jim Wynne

Leader
Admin
qualeety said:
if the requirement calls for less than o.003" and you are gettting identical results @ three decimal place...then you got SD of zero........there is no need to measure upto 5 or 6 places...

It's not a question of measuring out that far; what's probably happening is that the SD field is formatted for only 3 decimal places and everything to the right of that is getting truncated (not rounded) when displayed. Calculations will still include the full precision, but it's good to also be able to see some significant numbers.
 
B

Brian Myers

qualeety said:
if the requirement calls for less than o.003" and you are gettting identical results @ three decimal place...then you got SD of zero........there is no need to measure upto 5 or 6 places...

Not exactly true. Yes, the requirement is to 3 places, but an SD of 0 is NEARLY impossible. Checking to 5 or 6 places is a very quick way to discover if there is something else interfereing with correct reporting of the measured values. It is cheap and fast, the computer has the data and does the computation, use those extra decimal places as the "debugging" tool.

This is just a case of needing to look deeper into your data to discover the REAL truth. The answer is buried in there somewhere, don't stop at the "letter of the law", execute it for it's "spirit".:D

Brian
 

CarolX

Trusted Information Resource
Cognizant said:
I am actually qualifying my CMM probe, I remember in my training Standard deviation(SD) for qualifying the probe must be less than .003" for Zeiss Contura using calypso. Now one of my probes got bent and i replaced with a new one and I qualified the New probe but the SD value is zero ...Is that value sounds reasonable.....

though the sd=0 my qualification window says probe is good to run for measurement plan....

Please advice

FWIW,

I just returned form CMM training for my brown and sharpe cmm. Our instructed taught us that when qualifying your probe, your SD should not exceed 0.0003". A SD of 0.0000 is normal, and means that nothing has changed with your machine since your last calibration of your probe.
 
Q

qualitytrec

Speaking from experience (now long past) you should have some deviation at some point. You have a new probe tip it will not be exactly the same as the previous tip. I agree with the others above who are encouraging you to take your decimal place out further to qualify your program settings. I had a time once where my B&S with PC-DMIS said my qualification was good and I ended up almost shipping bad product because I did not know that there was a setting set wrong in the program until the next day when I decided to double check something else that I was being told was good and knew it wasn't. There were other times that I almost rejected good product for the same reason.

FWIW
Mark
 

cbearden

Involved In Discussions
I agree.....take your results out to 5 or 6 decimal places......I have mine at 5 decimal places......Also, your criteria should be < .0003"...not .003"

Zeissuser
 
P

peterj

Hi I have the same issue where all of the parts in my 32 piece sample are identical. I have to work with the data I have, and I have to produce a Cpk! Since my stdev is zero, I end up dividing by zero (which is a no no) when I use any Cpk formula. Since I'm stuck with this sample, how can I show that the process is capable. Also, technically, is this in control?
Thanks
 
Top Bottom