SMT Final Inspection Sampling Plan for boards off an SMT line



I need to come up with a final inspection sampling plan for boards off an SMT line. We work in a new product introduction arena which means we have ever changing products (nothing sustaining). We build new P/Ns every day...sometime only build them a few times and then never again. The sizes of the builds are anywhere from 1 to ~100 boards but most in the 15-40 range. Does anyone have examples for what type of sampling they do in a similar environment? I've just inherited this group and right now they are inspecting 20% of each build.

Duke Okes

It depends on the degree of confidence you want from the sampling plan and whether you are dealing with attribute or variable data.


It is attibute data... Inspection per IPC610 guidelines for workmanship, etc...

I haven't determined the confidence levels yet and I'm very familiar with AQL/LTPD sampling plans utilizing OC curves for protection. My question is how to deal with such small lot sizes and no ability to generate history data due to constantly changing product. I just want to see what other people are doing if in a similar situation.


Sorry if this is thread necromancy, but you don't seem to have had any kind of reply yet, so I thought I'd make a few suggestions.....

1) Sampling on a not yet proven process (i.e. first runs, prototype batches etc) is a flawed concept - as your SMT line has not yet had the opportunity to be optimized and as such sampling is not appropriate - I'd suggest that for batches of less than 20, you perform 100% inspection

2) for larger batches, you need to determine what your defect rate is... there are a lot of opportunitys for defects on an SMT board - somehting like every joint x10 (if you assume each join has an opportunity for solderballing, life, misalign, insufficient toe fillet, side fillet, heel fillet, outgassing, void, flux contamination) - you said you are using the IPC standard as guidance, so you should be able to determine this yourself. - If you have such a good stable process that you have a six sigma process against these opportunities for defect, then by all means drop down to standard ship lot sampling, otherwise 100% inspect.

3) Another thing to consider of course is the application - is this aerospace, military, medical, high reliability, or is it consumer stuff that's not going to do much mroe than inconvenience the end user when his uninspected board shorts itself out, or those dry joints eventually seperate? - Your contract review process should earth out the customers expectations - if he expects 100% failure free product, then there is a cost to bear - otherwise, if he can accept a few DOAs or ELFs, then you can reduce your inspection regimen, and save him a few $$$s

I hope that this helps a bit


Top Bottom