Evaluating the Yield of Repairable SRAMs for ATE

Abstract

Abstract-An accurate yield evaluation is essential in selecting redundancy allocation and testing strategies for memories. Yield evaluation can resolve the many issues revolving around costeffective built-in self-test (BIST) and automatic test equipment (ATE)-based solutions for a higher test transparency. In this paper, two yield-calculation methodologies for SRAM arrays are proposed. General yield expressions for VLSI chips are initially presented. The regular and repetitive structure of an SRAM array is exploited, and substantial yield improvements can be achieved by the introduction of redundancy. Two repair yield-evaluation methods for one-dimensional redundant memory arrays are introduced and compared for ATE application. The first method is based on the sum of the probabilities of all repairable fault patterns; the second method is based on Markov modeling. Using industrial data, it is shown that these methods are applicable to ATE usage under different conditions of defect rate in the possible defects. Different features of the proposed methods are discussed.

Yield Evaluation of Faulty Memristive Crossbar Array-based Neural Networks with Repairability

“…The yield calculated by using two methods is compared in this section. For comparison, the Poisson distribution [8] and the Markov chain model [12] are used to calculate yield. Many characteristics of these two methods are different.…”

“…We considered a lower to higher average defect rate for comparison and these values are considered based on the data used in [12]. The evaluated yield results for comparison are shown in Tables II, III and IV.…”