Frontend-Backend Integrated Circuit Device Linking Using Machine Learning Algorithms
Ingrid Kovacs, Bianca Cărbunescu-Stoenescu, Marina Ţopa, Andi Buzo, Emilian David, Georg Pelz · 2024
The current integrated circuits’qualification phase is lacking means of linking (tracking) devices without ID from backend to frontend; lot may be known, but wafer and coordinates are difficult to be linked. Therefore, machine learning techniques may enable traceability and help to support quality management actions to remove redundant tests and avoid packaging for defects. This paper proposes a methodology for frontend-backend device linking that uses the k-nearest neighbor algorithm computed on electrical parameters measurement to find a searched backend chip in a short list of targeted frontend chips. The accomplished results show that the method is able to localize a backend chip to a short list of frontend chips with an accuracy of 99.7%, when learning from some labeled chips for which frontend and backend link is known.