2 replies to this topic

[GI68]

We use two types of monitors : a) individual chart where 5 samples from the lot are taken and go to the same monitor and lot average chart

I am interested to get recommendations: what should be rule for right sigma estimation ? Sigma is needed to Cp/Cpk calculation.
When to use overall sigma (standard deviation ) and when to use user defined sigma (estimated from moving range)

[Adriaan]

To calculate the standard deviation one would normally use 20 values. A (partial) sigma can be calculated with at least 5 values but obviously any single (extreme) value then has a very significant influence.

Rather than using the st.d. to define the chart in your case I would use a chart with the target value at the center and an upper and lower limit chosen in such a way as to yield significant results; setting limits narrower than the actual specification limits for the product if the process capability allows, narrowing the limits if possible. If you have to broaden the limits you obviously need to investigate WHY.

Do your quality control methods include rules on re-sampling / retesting? If your testing is done by different people, is it known what the reproducability of tests is ? What SPC rules do you use ? These questions obviously influence the quality control / test data processing strategy needed.

What (software) do you use for the chart(s) ?

[GI68]

Dear Adriaan.

Thank you for your replay but it was not exactly for my question
I did not ask how to set control limits and nothing about the strategy.

See my question again :
" am interested to get recommendations: what should be rule for right sigma estimation ? Sigma is needed to Cp/Cpk calculation.
When to use overall sigma (standard deviation ) and when to use user defined sigma (estimated from moving range)
"