How to work with Mil Std 105E OC curves?

Hello,

Our lots are 150 000 - 500 000pcs. Inspection level is II => Code letter "P".
Attached are the tables from MilStd 105E. Please can somebody explain how to work with these tables?
Our business is automotive (microelectronics) and the philosophy is "zero defects". Which AQL should we choose? How to Select the AQL?

Please help!

Re: MilStd 105E - how to work with the OC curves?

Lets assume that the customer expects to receive not more than 30ppm defective parts.
The inspection is by attributes.
Should I use inspection level II or some different? Why?

Re: MilStd 105E - how to work with the OC curves?

I don't think that checking at the 30 PPM level is practical with a MIL-STD-105 plan (also known as ASQ Z1.4). That would be AQL of 0.003!

The smallest number listed on the P table is AQL 0.015, which would be inspect 800 and reject if there is a single defect. However, looking at the other table, a lot with 0.036% defective (more than 10x your desired level!) would be accepted 75% of the time. Even a lot with 0.375% defective (more than 100x worse than 30 PPM) would be accepted a little over 5% of the time.

It would be much better to find some continuous data to check, so that you can do SPC and calculate capability.


Tim F

Re: MilStd 105E - how to work with the OC curves?

Is it applicable MilStd 105E in my case? Should I choose another inaspection level and code letter?
Please advice how to proceed!

Re: MilStd 105E - how to work with the OC curves?

at 30 ppm, mil-std 105 is inadequate. it does not contain a sample plan for 30ppm (.00003 or .003%). If you calculate aplan using the POisson or BInomial you are at roughly 100,000 samples per lot at 95% confidence of detecting a lot that is 30ppm or worse...

Try mil-std 414. it is for continuous (variables) data.

OR you should use 100% mistake proofing to prevent known defect types or a 100% mistake proof inspection method to detect any defect made.

clarification

Mybe I did a mistake:
Actually this PPM is the AOQ (AOQ=30ppm) which is required.
The AQL levels used in our industry are 0.025, 0.04, 0.65... We use MilStd-105.

Is there a connection between AOQ and AQL? How can I link them?

You may want to consider using the sampling plans in Mil-Std-1916. I believe these were designed with the intent of limiting defect rates to less then 100 PPM. Mil-HDBK-1916 contains OC, AOQ and AFi curves and an explantion of the sampling plans.

The case is the following:
We produce microelectronics for the automotive industry.

What is specific in our business:
After the devices are manufactured we make 100% final test electrical inspection under certain conditions. Then we have so called electrical quality control (QCel) for every lot - we take a sample size from the 100% tested lot and check it on another test station. This QCel is similar to the final testing and the goal is to verify that the 100% final test was done correctly. The requirement is always zero defects at QCel.

How to link this to the standard? I cannot set no AQL here.

That's the limitation and misunderstanding of the AQL - and all other categorical data sampling plans. There is NO sampling plan that for 0 defects.

If you work with continuous data there is.

to determine a sample plan you MUST select some defect level to look for. but beware that if you select a very small defect level, the sample size will be very large.

In your case, what is the escape rate from 100% inspection? if not very close to zero, then you can use it as your defect rate for the AQL or other type of plan. You could aslo consider determeining the root cause of the escapes and reduce/eliminate/control those. For example a periodic Kappa test with known good, bad and marginal product to ensure that he test equipment is stil functioning properly. Calibration should be done as well. preventive maintenance on the wear items like fixtures, sockets, etc.

If you have very few escapes then there is nothing wrong with a small sample audit. A catastrophic event will be caught. There is no supreme law that you must use a statisticaly based plan for all inspections.

Bev -

I usually completely agree with your posts, but when you say:
"If you have very few escapes then there is nothing wrong with a small sample audit. A catastrophic event will be caught. There is no supreme law that you must use a statisticaly based plan for all inspections."

than I have to reply that in the world I live in (FDA regulated), there is a law (really a law with penalties up to and including incarceration):mg:

I do agree in less regulated industries a more ad hoc approach would be OK.

[Code of Federal Regulations]
[Title 21, Volume 8]
[Revised as of April 1, 2004]
From the U.S. Government Printing Office via GPO Access
[CITE: 21CFR820.250]

[Page 150]

TITLE 21--FOOD AND DRUGS

CHAPTER I--FOOD AND DRUG ADMINISTRATION, DEPARTMENT OF HEALTH AND HUMAN
SERVICES (CONTINUED)

PART 820_QUALITY SYSTEM REGULATION--Table of Contents

Subpart O_Statistical Techniques

Sec. 820.250 Statistical techniques.


(a) Where appropriate, each manufacturer shall establish and
maintain procedures for identifying valid statistical techniques
required for establishing, controlling, and verifying the acceptability
of process capability and product characteristics.
(b) Sampling plans, when used, shall be written and based on a valid
statistical rationale. Each manufacturer shall establish and maintain
procedures to ensure that sampling methods are adequate for their
intended use and to ensure that when changes occur the sampling plans
are reviewed. These activities shall be documented.