Are you doing actual measurements of the dummy parts placements? or are you just doing a visual go/no go check?
If doing actual measurements, be sure to use mil inch units of measure rather than mm. Mil will give you more accurate readings. mm forces some round off that cannot be resolved within its numbering system.
Be sure to use a 20X or greater magnification when inspecting and doing a measurement. I have used the CyberOptics LSM @ 100x with good success when no other means were available. Don't rely on calipers as I've seen some engineers using.
You need to start with the machines placement accuracy spec. from the supplier. They will advertise an accuracy of +/- 5 mil.s or something to that effect.
Next, you need to characterize each machine and finds it's offset bias. It will always place a component slightly off target, a combination of +/- X and +/- Y. Determine the repeatability of bias over a large square area. It may place well when close to x=0 and y=0, but degrades when x=12 inches, y=12 inches, etc.
Using a spreadsheet (Excel or other), record daily measures of hitting the target (or how much off) when x=0 and y=0, and when x=12 inches, y=12 inches, etc.
Display the daily measures on a scatter plot graph, using the spec limits of +/- 5 mils. as upper and lower limits. Be sure your daily measures fit between the limits or initiate a recalibration.
This is just my opinion, but as a long time "quality dude", I would be satisfied with results displayed in a manner I've described above.
reply »