Cpk, Cmk in Electronic Assembly (Process Capabiity)



we design and manifacture electronic engine management system for automotive sector.

Our Client want we define Cpk...
But this seems to be realized for a mechanical process, where i can measure dimension and tolerance, and so i can see either the process is in control or out of control.

But in electronic, where the critical elements are hundreds, what i must measure?
Can I avoid to classify every single defects, and focus only on the Printed circuit which fails?

P.S.: Do you know where i can find standards related to statistic (such as ISO/TR 13425:"Guidelines for the Selection of Statistical Methods in Standardization and Specification" or similar)?

Thanks all



I am not from the automotive sector, but if they want a Cpk, they don't care about all the properties of your product, but your product it self.

for example: How much of the product could be defective? Cpk can be correlated to the ppm defective (if ..., you take the distribution as a gaussian ones.). In the net there are many articles that show such table.

I tink that i'ts right to focus only on the parts that could fail, but keep on mind that they just want a value a Cpk (only one) of your product.


Rick Goodson


As Darius implied, Cpk is a measurement of a process not a product. The parameters you mention are not related to the process but to the product. Cpk could be determined for individual processes in your facility. Refernce the AIAG SPC guide for more details. Go to www.aiag.com and look under publications for the SPC manual. It costs $30 for non-members.


thanks all.
You are right: these are product's item, not process'ones.
but to measure process capability i have to measure how good is its output; and process'output is the product.
Don Winton said:" It is ultimately desiderable to establish control over the process by controlling the process inputs. On the other hand, process capability indices are always performed using the critical parameters of the product" ("process capability studies, 1999, elsmar pdf section downloaded)
So, i think i can list all tipologies of defect, and analyze that with statistical techniques (such as Pareto etc...), but at the end i'll measure how many electronic engine system are fully functional and i extract the Cpk.
1) what's the process mean in such a case?
and what is the data value?
'cause Xbar= Xi/n for i=1 to n
(i think Xi is only Pass/Failure)
2) how can i define LSL and USL ?
remind i have no specifications limits... but the product has to funcion!!

thanks again :)

Rick Goodson


You are switching back and forth from attribute and variables during your discussion. An X-bar chart is a variables control chart where pass/fail data is attribute data. You cannot plot attribute data on a variables control chart.

From the AIAG SPC 1995 publication:

Process capability is the total range of a stable process's inherent variation: 6 sigma cap sub R/d2.

For attribute data:

Process capability is usually defined as the average proportion or rate of defects or defectives. From control charts (p,c,u), for example, capability is defined as p-bar, c-bar, or u-bar where capability refers directly to the average proportion or rate of output that does not meet the specification (or as 1- p-bar, etc. the percentage of product within specification.

Ford use to use 99.994% as the minimum process capability for an attribute control chart based process capability calculation. That is, 99.994% of product had to meet the specification. I do not know if that is still a good number or not.

In essence, the capability is the first pass yield on the complete process.

Hope this helps.
Top Bottom