# SMT (Surface Mount Technology) Sampling Plan in a SMT & COB Assembly Company

#### Marc

##### Fully vaccinated are you?
Dear Marc,

I am working as quality system engineer at SMT & COB assembly company. I need your advice about the sampling plan.

Regarding QS9000 manual third edition,
Inspection and Testing - Element 4.10.
Acceptance Criteria for Attribute Characteristics - 4.10.1.1
Acceptance criteria for attribute data sampling plans shall be zero defect. Appropriate acceptance criteria for all other situations (eg. visual standards) shall be documented by the supplier and approved by the customer.

I'd like to settle the sampling plan with my customer. Pls. advise what is the correct method to find out sampling plan for zero defect between a) or b).

Example : Lot size 21,000 pcs. with 0.10% AQL, level II.

a) Refer to C=0 sampling plan table (as attached "C=0 sampling.jpg") from Nicholas L. Squelia, 4th edition : Zero Acceptance Number Sampling Plans, the sampling size is 294 pcs. The lot is rejected if found any defect.

b) Refer to single sampling plan table (as attached "ANSI ASQC Z1.4-1993 table 2a.jpg") from ANSI ASQC Z1.4-1993, the sampling size is 500 pcs. with defect allowance 1/2 (acc,/rej.). But according to mentioned QS9000, adapt it to sampling size 500 pcs. with defect allowance 0/1 (acc./rej.).

Thank you and Best regards,
Sarunya P.
<center>
<br>
</center><hr>Any takers on helping this person?

D

#### Dave Strouse

What is wanted ?

Would love to help, but don't think the right question is asked.

As to which is correct, both are!

That is; there is no "RIGHT" sampling plan. All are correct based on the desires of the participants, i.e. the producer and consumer. There is the questions of statistical validity and the operating characterstics. Statistical validity is largely a function of the sampling method. Is it random? Is it properly carried out?

A better question might be, "What are the differences in operating characteristics between the plan Nick Squelia has published and the ANSI standard plan as modified to c=0 by the poster.

At AQL probability of acceptance is 74.5 % for the Squelia plan and 60.6% for the other .

IQL is 0.002 and 0.0013 for the plans in the same order.

The RQL or LTPD is at 0.007 and 0.0046.

Draw the curves and decide what you want . As always tradeoffs are needed. Is the additional sampling and testing worth the additional protection.

RANT/RAVE mode on : I've seen a thread or two here recently about acceptance sampling.

It has it's place.

Ed Schilling was mentioned one or two times. Ed was my graduate advisor and a finer human being would not be found. He wrote what is considered the definative work on acceptance sampling, was on the comittee for the ANSI standards and wrote the section in Jurans on acceptance sampling. Yet he says " acceptance sampling is at best a stop gap measure to be used until process contol can be implemented."

It seems many still believe the answer is in these c=0 plans. That way we aren't "shipping defects to the customer" is the battle cry. What a crock! Look at the OC curve points above. We can say with some 90% confidence that the percent defective shipped is less than .7 percent for the Squelia plan and .46 for the other.

But no acceptance sampling plan or process control plan will assure no defects. The first will limit the exposure to a agreed on level and the other will limit the likelihood of production and allow the producer to stop and fix the problems sooner.

RANT/RAVE mode off

R

#### Rick Goodson

Dave S. has said it as well as can be said. Given you are required to use an acceptance sampling plan, you need to look at the OC curves for both plans and choose the one that best serves your neds.

Rick