AQL Difference in Checking LCD Pins

terranceyu · May 20, 2010

Hello to covers,

Right now, I have a tough situation, and supplier is not going to give in our findings.

Details are explained in the below statement:-
Production found LCD pins misaligned and not able to fit into PCB holes. ( pin is misaligned by 0.2 mm measured from the centre of pin to the edge of glass). Almost 40% of the LCD can not fit to the PCB. All the parts were purged out and return to reject warehouse. IQC checked the PCB hole, it is ok. Then, IQC took 3 pcs LCD from each three lots, some lot found: 2 pcs reject, 1 ok, and some is 3 pcs reject. We would like to charge the supplier on the sorting cost to segregate the LCDs, and supplier is not agreed.

Instead, they wanted us to measured according to AQL 1.0 (in other means 80 pcs for each lot they delivered), as they use AQL1.0 sampling for their minor defect quality control. They even said, if you like to strictly tighten the quality standards ( we are applying the AQL=0.25), they can either increase the price or we need to find another LCD vendor, as they can't meet our requirement.Now they request us to measure 80 pcs and tell them the result.

I am wondering is it appropriate for measuring 80 pcs, since we know there is a problem in their process, causing 40% of failures in the end of line. If their process average is very capable and stable, why should they want us to check until 80 pcs, and feedback them the result? And in AQL, is there a clause to state that measurement must be according to AQL, as we normally take 5 pcs for mechanical measurement.

Please comment and advise.

Tim Folkerts · May 20, 2010

terranceyu said:
IQC took 3 pcs LCD from each three lots, some lot found: 2 pcs reject, 1 ok, and some is 3 pcs reject. ...

Instead, they wanted us to measured according to AQL 1.0 ...

Checking the tables, for AQL 1.0 and a sample size of 80, the chart says to reject any sample with 3 or more defectives. Once you have found 3 defectives, you could reject the lot, so even if you test only the first 3 items in your sample of 80 and find them bad, you could reject the lot. If you collect a few more items from the lots that only had 2 defectives and you found a third defective before reaching 80 for the sample size, you could legitimately reject the lot.

You might also consider a multiple sampling plan if you have such bad lots. For AQL 1 and Sample Code J, you would test 20 pieces and reject if you got two or more bad items (so under this plan, you could already reject all the lots where your found 2 of 3 bad). If you got less than 2 bad items out of 20, you would have to test more. This would be a bit more complicated, but it would allow you to more quickly reject the lots that are much worse than 1% defective. (the down side is you might have to test up to 140 items before deciding what to do with the lot). Furthermore, this follows the supplier's request to use AQL 1.0 (although they may not have expected multiple sampling).

(Also, it is not legitimate to choose the sampling plan AFTER starting the testing.If you are using multiple sampling, then with 80 items, you accept with 1 bad and reject with 4 bad. With single sampling you would accept 2 and reject 3 with 80 items.)

Tim F

terranceyu · May 20, 2010

Tim Folkerts said:
Checking the tables, for AQL 1.0 and a sample size of 80, the chart says to reject any sample with 3 or more defectives. Once you have found 3 defectives, you could reject the lot, so even if you test only the first 3 items in your sample of 80 and find them bad, you could reject the lot. If you collect a few more items from the lots that only had 2 defectives and you found a third defective before reaching 80 for the sample size, you could legitimately reject the lot.

You might also consider a multiple sampling plan if you have such bad lots. For AQL 1 and Sample Code J, you would test 20 pieces and reject if you got two or more bad items (so under this plan, you could already reject all the lots where your found 2 of 3 bad). If you got less than 2 bad items out of 20, you would have to test more. This would be a bit more complicated, but it would allow you to more quickly reject the lots that are much worse than 1% defective. (the down side is you might have to test up to 140 items before deciding what to do with the lot). Furthermore, this follows the supplier's request to use AQL 1.0 (although they may not have expected multiple sampling).

(Also, it is not legitimate to choose the sampling plan AFTER starting the testing.If you are using multiple sampling, then with 80 items, you accept with 1 bad and reject with 4 bad. With single sampling you would accept 2 and reject 3 with 80 items.)

Tim F

Tim,

The fact is, we are not able to measure 80 pcs, as it took longer time to complete the measurement. FYI, we are using the semi-auto OGP scope to measure the part. It will take approx 2 minutes to measure 1 piece part. 80 pcs will take approx. 2.67 hours just to measure the part. Is that making any senses, to provide 80 pcs solid data to prove their lot having bad quality, in the fact we can measure 5 ~ 10 pcs and know they already failed AQL?

I also have the question regarding the difference AQL standard between supplier and us. Should we drop our AQL standard (0.25), and adapt the supplier's (AQL 1)? Sincerely, I do not think this is the right way.

harry · May 21, 2010

It appears that your supplier is very confident and hence the take it or leave it stance.

1. First thing you need to check is your agreement or contract with this supplier - what is the agreed AQL level and other details.

2. Carry out an analysis of the problem. If the LCD end cannot be changed or modified, how about the connectors on the PCB side?

3. If someone from your end had not done a good job during the 'review of requirements/contract' stage, it is your problem. Choose a cheaper option and proceed. If my information of the Malaysian market is still current, I don't think you have much choice.

Bev D · May 21, 2010

terranceyu said:
Tim,

The fact is, we are not able to measure 80 pcs, as it took longer time to complete the measurement.

Tim is saying that you don't have to measure all 80 pieces. Once a sample from any given lot has 3 defects you have reached the reject number, completing the testing of the 80 piece sampel will not reduce teh number of defects. reject the lot.

But Harry is also giving good advice: your supplier has a take it or leave it attitude and dependign on the contract and your Customer rep's attitide you may be stuck with the cost of screening the parts...

AQL Difference in Checking LCD Pins

terranceyu

Tim Folkerts

terranceyu

harry

Bev D

Heretical Statistician

Similar threads