BGA rework: Coplanarity of Xilinx before and after reflow.

We have the exact same issue with out 456 ball Xilinx parts. The corner balls either do not touch or are so stretched out that they are intermittant. I believe this to be a moisture ingress problem. Ive been setting up our little batch oven this afternoon to bake a tray of these parts according to J-Std-033 by tomorrow. Hopefully this will prove my theory. Ill let you know how I make out...

reply »

We believe your problem is caused by the mismatch in CTE of the material ised in your BGA. In fact, this "potato chipping" [where the corners curl-up] is present in all BGA. The issue is the matter of degree. [Mismatching CTE is how old thermostats work. Two metal strips with different CTE are bound together. As the temperature increases, the metal with the higher CTE will lengthen more than the other metal. Since the two metals are bound, the longer metal must make up the extra length by taking a longer path, or in other words, the longer metal will bend the strip.]

To prove this is correct, watch your PBGA in your rework machine. You will see it "wiggle" as the corners move from curving up to curving down to curving up etc. [Admittedly, a rework station does a more radical job of heating BGA than a reflow over, but you get the idea.] Again, this is due to the CTE mismatches before reaching Tg temperatures warping the BGA in one direction. Once either the plastic or laminate reaches its Tg, it will either exacerbate the problem or make the BGA warp the other direction. And finally, once the both the plastic and laminate reach their Tg the BGA will warp again in the direction of whichever CTE dominates.

You should find a new supplier that does a better job of selecting and controling the materials used in their product. Haaaa pretty funny, eh? Well, since that's not one of the choices, you need to minimize the problem by changing your reflow profile.

That being said, CTE is temperature dependent and not temperature change rate dependent, so you are pretty much hosed by the physics. However, we do like trying like hell to get the top and bottom of the BGA the same temperature before entering reflow. Moving into reflow slowly so the paste has time to become liquid before the BGA begins to dance. This means a long soak before reflow. Although, we admit this is probably wishful thinking since the physics of CTE doesn't care about the rate of temperature change.

reply »