I admit, 16 mil QFP printing is very challenging but if all your control factors (print speed, etc...) are setup correctly (through DOE) extremely high yields can be achieved and more importantly a repeatable and robust product will be manufactured.
It is difficult to say do this and do that... It is very dependant on your solder paste, stencil type, etc... A type 3 (20-45 um) paste will work nicely. Although each suppliers paste can vary greatly even if they have similiar specs.
If using squeegee blades (and not a pump head) make sure the paste roll is consistent, this sounds silly but I did an experiment and found that when you vary the paste roll from 150 grams to 250 grams (200 grams is nominal for my product) variability is effected by ~6 db's or 50%!! This is huge!
Therefore when you do your DOE use "noise" in other words say you are doing a small full factorial that checks two control factors at two separate levels. This means you have to do 4 runs to check all the combinations...
Well, first do the 4 runs with a small paste roll and then repeat the 4 runs with a big paste roll. Obviously you will have to clean the stencil and blades after each run and throw the excess paste away.
When you have completed all 8 runs (4 at one noise factor and 4 at another noise factor) crunch your data and the results should give you the most robust process you can have.
I like doing Taguchi DOE's. I recommend a L18, they are the best. You can test 6 control factors at 3 levels and 1 control factor at 2 levels. Taguchi uses S/N ratio as his measure of variability.
DOE's take some time, but if done right they will save you headaches and $$ in the long run.
Last note throw different aperture geometries/sizes in your test matrix, for example try different sizes or shapes. That way you have data to back up any "hunches" you may have.
In the end data rules! Good luck.
reply »