AOI False call rate vs defect rate PPM

Subscribe to Thread

I wanted to open a discussion and ask about opinion: What is... - Jun 07, 2024 by DavidK

You would need to be more specific on the AOI in question he... - Jun 13, 2024 by Carl_P

Hi David, In one of my previous jobs I worked as an AOI/AXI... - Jun 13, 2024 by MagyarT

Hello Carl, Thank you for your comment Obviously, we are evo... - Jun 19, 2024 by DavidK

Hi Tom, Thank you for your comment 75% FPY is also our targe... - Jun 19, 2024 by DavidK

DavidK

#91045

AOI False call rate vs defect rate PPM | 7 June, 2024

I wanted to open a discussion and ask about opinion: What is a normal result of the inspection of an AOI? How many falls calls per component /opportunity is bad/normal/good? How many defects per component/opportunity is bad/normal/good? What is good FPY?

We are a high mix low or middle volume production a lot of samples for development. We produce per month around 10 000-15 000 multi-panels 30 000-45 000 subboards 5 000 000 - 10 000 000 components 20 000 000 - 40 000 000 opportunities FPY 35 - 50% False call rate per component 2500 - 3800 PPM per opportunity 1000-1500PPM Defect rate per component 150 - 300PPM per opportunity 80 - 140PPM Escape rate per component 5 - 12 PPM.

So, per one found defect we have around 10 false calls.

reply »

Carl_P

#91061

AOI False call rate vs defect rate PPM | 13 June, 2024

You would need to be more specific on the AOI in question here. 2D, 3D features will also have a big impact.

Are you constantly evolving or updating the models to improve the situation or going from just the base program as if low volume it can be an issue to build up that model database to cover natural variety.

reply »

MagyarT

#91065

AOI False call rate vs defect rate PPM | 13 June, 2024

Hi David,

In one of my previous jobs I worked as an AOI/AXI process engineer and we had a set target of 75% FPY. This means that 75% of the boards at IPC Class 3 standards would not have a single false call.

This also means you would need to set your processes correctly using golden (good) and silver (bad) sample boards during prototype phase. Constant program optimization in the first serial production to achieve the targets. Test coverage would start from 100%, and whatever could not be inspected as per customer requirements, would go on X-Ray (AXI) - both Viscom or combined in AOXI. Normally these were all well planned for in advance.

Of course, this was automotive industry. Low mix, high volume. But a process minded manager/engineer could do it basically anywhere. It is also dependant on the customer/internal regulations and the machines you are working with + costing.

As far as I can remember our target was to stay under 500 defects per million opportunities and under 10000 false calls per million opportunities.

The only thing I don't like in your initial statement is that for 5-10 million components you have many more defect opportunities than 20-40 million, unless you are only placing resistors and capacitors.

Regards, Tom

reply »

DavidK

#91081

AOI False call rate vs defect rate PPM | 19 June, 2024

Hello Carl, Thank you for your comment Obviously, we are evolving and debugging our test to get the best results ("no escapes no false calls") :) and using a combination of 3D and 2D tests. With the previous AOI, we had a much higher false call rate, more misjudgments, and low FPY.

reply »

DavidK

#91083

AOI False call rate vs defect rate PPM | 19 June, 2024

Hi Tom, Thank you for your comment 75% FPY is also our target, cause we gonna have the AOIs in the SMT line.

So you would expect more opportunities? I didn´t notice but there are not so many components with high opportunity rate. (the average number is around 4), But the boards optically look normal, there are some components with more leads but the number of the chips is realy high.

reply »