What should be Maximum Cpk value?

The ONLY reason to change a specification limit is that the specification limit doesn't accurately correlate with failure. Specifications are about what makes the product work as intended. Cpk is about how the process performs in relationship to the specification. If your Customer doesn't understand this - they are wrong. Pure and simple.

 

Tell them that any efforts your organization does to achieve this will be reflected in a price increase - that will halt their ideas...

 

Here's the physics answer: Tightening the specification doesn't improve the process. Reducing the actual variation of actual parts improves the variation of the process. You already have a very capable process: the actual variation is small compared to the specifications. Tightening the limits will only result in throwing parts away that might exceed the tighter limits at some point in the future (out of control conditions do occur from time to time). If the original specifications are set properly then tightening them only increases cost but not quality. Andy is correct: tell them that any tightening will result in a price increase.

This is the kind of stupidity that gives the quality profession a bad name...

 

Unless it's not the machine. It could be the mold, the material, the feature, the fixturing, the program if CNC, the machine parameters, the tooling, etc...or all of these

 

Good luck with that! My experience of customer auditors is that they often have no clue and are not open to the idea that they might actually learn something by keeping quiet for once and hearing what the supplier has to say. As much as people rail on CB auditors, there's no oversight of customer auditors and therein lies a potentially huge problem...

 

I’m not really sure what you’re asking…. Are you concerned a bout how the ‘specifications’ are set? Or are you concerned about the extreme value? We really don’t have enough information to determine if the extreme value is OK or not. Physics is impervious to opinion, ego and stupidity.

I would with the minimal information we have at least strongly suggest either:
(1) Investigate why the extreme value occurred and if it is real does it matter?
(2) Investigate whether it matters.

THEN we can discuss the larger issues….

 

Its looking like a sales pitch, Bev…

 

This could also be an artifact of some of the practices of electronic component manufacturing that are not seen in mechanical manufacturing. I speak of the practice of sorting components into different levels of precision. You do not get a normal distribution. If you order a resistor with precision tolerance, you get a slice from the middle of the distribution. If you order a resistor with the standard tolerance, you get what started as a normal distribution, but with a slice missing from the middle. Not knowing the detailed practices, it is possible that the upper portion of this distribution is separated from the lower portion into different lots. What you show might have been the result of one component from one lot mixed with the components from the other lot. Yes, there is a lot of speculation here, but enough basis to launch an investigation. If it turns out to be the cause, then it would explain how the outlier occurred and why it would be low risk.

 

My response hasn’t changed. And although I am an electronic engineer with a major and considerable experience in semiconductor manufacturing and design as well as substantial experience in medical device design and manufacturing I can say that you have not provided enough information to make an informed decision regarding your exact question. We will not be baited…

I will say that Stats training in general doesn't’ really go beyond enumerative theoretical homogenous situations. There is nothing special about electronics in that regard. Electronic devices tend to have a lot of non-homogeneity (as does any batch processing and other disciplines.). “Normality” is a concept that has no magical properties.

I suggest you read Deming’s “On Probability as a Basis for Action” before you make any more “statistical” conclusions…It’s FREE - just ‘Google it’ and down load the PDF and read it. Then read it again. Then come back here…

 

In addition to Miner’s response I have an example of how all engineers must be knowledgeable about variation and the realities of the components used in their designs.
I was the Quality Director for a division of a large electronics designs and manufacturing organization. An engineer used a small ‘prototype’ device to help design his circuit. He set the design space according to the distribution of the samples he got early in the design phase. He didn’t use the data sheet specification. Of course you know the rest of the story. When we began production the boards didn’t work. The new batch had shifted (safely within the data sheet specifications but with only a small amount of overlap with the original samples. A simple case of non-homogeneity