A Novel Built-In Self-Repair Scheme for 3D Memory

Abstract

Three-dimensional (3D) memory products based on through silicon via (TSV) are widely developed to fulfill the ever-increasing demands of per unit area storage capacity. The yield is still one of the critical challenges for 3D memory. Redundancy technique is now widely used in industry to improve yield. How to reduce the overhead of redundancy by improving the utilization of redundancy is important to 3D memory. In this paper, we propose a row/column block-based mapping technique for 3D memory built-in self-repair scheme to improve the utilization of redundancy and low hardware overhead. Each row/column is divided into row/column block and the mapping can be performed at row/column-block level instead of the original row/column level. Therefore, more faulty cells can be clustered into the same row/column. Based on the proposed technology, a 3D-essential spare pivoting (ESP) algorithm is also proposed for the allocation of redundant rows and columns, and the area overhead of this algorithm is particularly low. The experimental results show that on an average the repair ratio of our proposed scheme is much better than the fault clustering technique by 12% and the redundancy-cost can reduce 23%.