Rework Process for a 42mm Square, 1mm Pitch, 1657 Ball, BGA ASIC

A Networking Systems Startup OEM had two separate EMS providers develop a Rework Process for a 42mm square, 1mm pitch, 1657 ball, BGA ASIC. After six months of ASIC reworks the combined Yield for the OEM’s ASIC rework was 56%, and the EMS’s involved attributed the yield loss to the component size and rework process complexity. The ASIC then underwent a redesign (respin) and a decision was required to be made to either Rework 400 ASICs at the Current Rework Yield, build 200 Linecards from scratch at a cost of more than $2m, or carry out a Rework Process Optimization to improve Rework Yield and then rework 400 ASICs.

Deciding upon the latter option, the EMS providers involved could not improve yield, and it was decided that the process optimization would be conducted with the EMS’s Time and Materials, but the Process Optimization decisions would be driven by the Customer.

The key changes made to improve the rework process were as follows :

1. Modify mini-stencil used to a square aperture to increase solderpaste release characteristics, reduce solderpaste deposit variance and reduce average paste volume by 12%.
2. Conduct Thermal Profiling to reduce temperature differential across part.
   • A component’s temperature induced warpage displacement distance = lat
   • Where l = length between the two points in question, i.e. the center of component to the corner.
   • a = alpha, a material constant.
   • t = temperature differential between the two points in question.
   • Thus to reduce temperature induced warpage, the temperature differential must be reduced.
   • Rework Temperature Profile changes resulted in reducing temperature differential (t or delta t = t center – t corner) from 22 C to 13 C.
3. Rework site dressing was standardized and automated, to increase site flatness prior to solderpaste printing and reduce variance of flatness across site.
4. Placement force was standardized as per component manufacturer’s recommendation.

A number of example reworks were completed and were halted when the first rework failure occurred, resulting in an initial rework yield of approximately 95%.

This fallout was deduced to be a result of excessive moisture ingress, and the handling procedures were modified to reduce component exposure to atmosphere prior to rework and hence reduce moisture ingress.

Rework continued, with zero failures in subsequent processing and after 400 components were reworked, the Overall Rework Yield was increased from 56% to 99.75%.

To see the resume of the expert associated with this case study, see the link below.